UNDERSTANDINGTHE DEVELOPINGMETROPOLISLessonsfrom the City Studyof Bogota and Cali, Colombia

Rakesh Mohan

A WoorId Bank Ilook

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Understanding the Developing Metropolis

Lessons from the City Study of Bogotiand Cali, Colombia

Understanding the Developing MetropolisLessons from the City Study of Bogota

and Cali, Colombia

Rakesh Mohan

Published for The lWorld BankOxford Universitvy Press

Oxford Universitv Press

OXFORD NEW YORK TORONTO

DELHI BOMBAY (.kLCUTTA MADRAS KARARCHI

KUALA LUMIPLIR SINClAPORE HON(; KONG TOKYO

NAIROBI DAR ES SAI-,AlAM (1APETOWN

MELBOURNE AUCKlAND

and associated companiies inBERLIN IBADAN

C 1994 by The International Bank for Reconstructionand Development / THE WORLD BANK

1818 H Street, N.W., Washington, D.C. 20433

Publishied by Oxford University Press, Inc.200 Madison Avenue, New York, N.Y 10016

Oxford is a registered trademark of Oxford University Press

All rights reserved. No part of this publicationimay be reproduced, stored in a retrieval system, or transmitted,

in any form or by any means, electronic, mechanical,photocopying, recordinig, or otherwise, wvithout the prior

permission of Oxford Universitv Press.

Manufactured in the United States of AmericaFirst printing April 1994

The findings, interpretations, and concilisions expressed in this study are entirely thoseof the author and should not be attributed in anv manner to the World Bank, to itsaffiliated organizations, or to members of its Board of Executive Directors or the

countries they represent. The bounidlaries, colors, denominiaiolns, and otherinformationi shown on any map in this volume do not imply on the part of the

World Bank Giroup any judgment on the legal statuis of anly terr-itorv or theendorsemenit or acceptance of stich boundarics.

Library of Congress Cataloging-in-Publication Dala

Mohan, Rakesh, 1948-Understanding the developing metropolis: lessons from the city

study of Bogota and Cali, Colombia / Rakesh Mohan.p. cm.

Includes bibliographical references and index.ISBN 0-19-520882-X1. Bogota (Colombia)-Economic conditions. 2. Cali (Colombia)-

Economic conditions. 3. Urban policy-Developing counltries.l. Title.HC198.1B5M633 1994333.9861'48-dc2O 93-41140

CIP

Contents

Foreword xi

Preface and Acknowledgments xiii

1. Introduction ICity Selection 4The Macroeconomic Setting: Were the 1970s Exceptional? 8Overview 1 4Notes 1 7

2. The Spatial Structure of Cities in Developing Countries 18The Emergence of Cities 18Growth and Decentralization 21Citv Characteristics: Ecidence from Different Countries 26The Consequences of Growth and Decentralization 34Notes 37

3. Growth and the Changing Structure of Bogota and Cali 38Growth and the Spatial Distribution of Population and Income 43The Evolution of Land Values and Population) and Employment

Densities 55The Changing Structure of Bogota: Some Wrinkles 63Summary 69Notes 72

4. Poverty, Distribtition of Income, and Growth 73Distributioni of Income in Bogota and Cali 1973-78 76Distribution and Characteristics of the Poor 83Summary 91Notes 95

5. Workers and Their Earnings 96The Benefits of Education and Experience 103How Segmented Is the Bogota Labor Market? 108

v

vi CONTENTS

Operation of the Urban Labor Market: What Have We Learned? 117Notes 119

6. Firms and Their Location Behavior 121Trends in the Location of Employvment 123Patterns of Employment Location in Manufacturing 128Factors Influencirng the Location of Manufacturing Firms 132Modeling the Intra-Urbain Location Behavior of Manufacturing

Firms 134Trade and Service Employment Patterns 139Implications for Location Policv 141Notes 144

7. Shelter in a Growing City 146HoIsillg in Bogota: The Instittional Setting 147Bogota's Unregulated Housing Market: The Myths and Realities

of Incremental Development 152Mobility and Tenure Choice 161Modeling Housing Demand 171Pirala Developments and Housinig Demand: Some Lessons for Housing

Programs and Policies 180Notes 183

8. Autos, Taxis, Buses, and Busetas: The Importance of Choicein Urban Trarnsport 185

The Stipplv of Transportation Services in Bogota 189The Regtilation of Supplv: Control, Suibsidies, and Incentives 191The Impact of Govern inent Regulationi, Taxes, and Subsidies 198Travel Patterns in Bogota 201Modeling Travel Demand 215Lessons for a Developing Metropolis 221Appendix: Theoretical Background for Mlodeling the Choice

of Transport Mode 225Notes 232

9. Urban Government and Finances 233Urban Government in Bogota 234The Structure of Revenue 239The Structure of Expenditures 255The Public Serxice Enterprises of Bogota 259Lessons for a Developing Metropolis: An Evaluative Summary 266Notes 271

10. Coping with City Growth 272Understaniding Behavior 275Understandiig City Structure 278Encouraging Endogenous Institutional Responses to Rapid

Growth 281

Appendix: The Data 288

CONTENTS vii

Bibliography 301

Index 317

Tables

Chapter I1-1 Economic Growth in Colombia, 1950-86 101-2 Prices and Wages in Colombia 12

Chapter 22-1 Population and Population Growth in Selected Latin American

and U.S. Metropolitan Areas 282-2 Population Density in Central and Peripheral Areas of Selected

Latin American and U.S. Cities 292-3 Population Densitv Gradients in Selected Latin American and U.S.

Cities 302-4 Population Densitv Gradients Per Kilometer in Selected Cities

Worldwide by Decade, 1880-1960 332-5 Populationi Densitv Gradients Per Kilometer in Selected Cities

Worldwide, 1950, 1960, 1965, and 1970 342-6 Central City Shares of Residential Population and Employment,

Selected Latin Americani Cities 35

Chapter 33-1 Population and Intercenisal Population Growth in Colombia 393-2 Gross and Per Capita Domestic Product, Colombia

and Bogota 393-3 Average Anntial Growth in Domestic Product, Colombia

and Bogota 4034 Area. Poptilationi, Population Growth, and Density in Bogoti

and Cali 413-5 Growth in Population and Densitv in Bogota 503-6 Spatial Distributioni of Monthlv lncome and Population

in Bogota 543-7 Change in Population) Density by Ring, BogotA and Cali 563-8 Evolutioni of Land Values by Ring, Bogota and Cali 58,-9 Employment Density in BogotA 61

3-10 Distributioni and Density of Workers' Residences and W'orkplaces,Bogota 62

3-11 I.and Value and Population Densitv Gradients in BogotA and Cali,bv Sector 66

3-12 Spatial Patterni of Housinig in Bogota, 1978 67

Chapter 44-1 Distribution of Inconme in Bogota and Cali 784-2 Spatial Distribution of Income in Bogota and Cali 804-3 Spatial Inequality in Bogota and Cali: Individuals Ranked

by tItNC(AP 82

vi i CONTENTS

4-4 Mappinig MalnTitritionl into Income Deciles in Bogota. 1978 86415 Spatial Distribution of Malnutrition) bv Age Group in Bogoti,

1978 884-6 Unemployment in Bogota by Incomile Decile and Sex 92

Chapter 55-1 Changes in Labor Use in Colornbia, 1951-78 985-2 Emplovmenit Chanige in Colombia, 1973-78 985-3 Distribution of Workers by Sex and Education Level 105-4 Mean Incomc Ratios for W orkers bv Sex and( Education

Level 10o

Chapter 66-1 Employmenit bv Firm Size and Major Industry Group, 1978 1246-2 Employment Distribution by Ring. Firm Size, and Major Industry

Group, 1978 1256-3 Employmenit Distribution in Bogota by Radial Sector. Firm Size.

and Major Industry Group, 1978 1266-4 Employment Location in Bogota, 1972 and 1978 1276-5 Distributioll of Manufactuting Employmenit by Ring: BogotA

and Cali, 1970-75 1306-6 Changes in Manufacturing Employmenit, bv Ring:

BogotS, 1970-75 131

Chapter 7

7-1 Average Data for Pirala aid .i 'orrnas ifinima5Subdivisions,Mid-1970s o58

7-2 Fiiianicinlg of Infiastructure in frafta Subdivisionis 1597-3 Alternative Average Nominial Rates of' Return

to Subdividers 1617-4 Movinig Rates by Tenure Choice. Time Spent in Previous Residenice,

andi Age of Household Head 1647-5 Probability of Moving: Coniparison of (Oh-s and 1.0(.1

Elasticities 1657-6 Mobility Rates in U.S. Metropolitani Areas by Selected Household

Characteristics 1667-7 Moves of Hlouseholds from One Ring to Another

(Recent Movers) 1677-8 Moves of Households from One Sector to Another

(Recent Movers) 1687-9 Tenure Choice by Household Income 169

7-10 Tenure Choice by Recencv of'Household Head's Migration 1697-11 Elasticities for Tenure Choice Model Comparison of ots and t.oC Ir

Specifications 1707-12 Interinational Comilpar-isonis of' Owner-ship Status Elasticities 1717-13 Character-istics of' Residences b Work Zone. Bogotul 1757-14 Analysis of Variance: Hedonic Price Equations 1767-15 Hotisinig Demand Equationis: Estimated Elasticities

for BogotA 1777-16 Raniges of'Hotising Demanid Elasticities fronm Various

Counitries 178

CONTENTS ix

Chapter 88-1 Transportatioin in Bogota 1908-2 Vehicle Stock in Colombia, 1970-80 1918-3 Vehicle Stock in Bogota, Cali, and All of Colombia,

December 1977 1928-4 Bus Subsidies in Colombian Cities, 1977 and 1978 1958-5 Adjusted C(osts, Adjusted Income, and Capital Return for Buses,

October 1980 1968-6 Formal and Actual Governmenit Interventioni in the Transpor t

Sector. (Colombia, 1980 1998-7 Vehicle Taxation Incidenice in BogotA, 1978 2008-8 Fuel Policv Incidenice in BogotA, 1980 2018-9 Tranisport Tax and Subsi(ly Policy Incidenice: BogotA,

1978--S0 2028-10 Work Trip Travel Time by Income Level: BogotA. 1972 and 1978,

and Cali. 1978 2058-11 Wor-k Trip Travel Time by Transportation Mode: Bogota, 1972

and 1978, and Cali, 1978 2058-12 Ring of Resiclentce and Ring of Emploviyment: BogotA, 1972

and 1978 2068-13 Average Comimiiuting Distance by Zone of Work, Tvpe of Worker.

and Occupationi: Bogota. 1972 iand 1978 2088-14 Average Nuliber of Trips Per Person and Average Number of Trips

Per Hotusehold: Selected Cities 2118-15 Trip Generationi Characteristics of All Travelers: BogotA.,

1972 2128-16 Trip) Generation Equatiorls: BogotA, 1972 2138-17 Work Trip by Mode of Transportation: Bogota, 1972 and 1978,

and Cali, 1978 2178-18 Change in Work T'rip Modal Shares-Direct Elasticities: BogotA,

1972 and 1978, and Cali. 1978 2198A-1 Local Registrationi Tax Incidenice: BogotA, 1980 2288A-2 Sales Tax Inci(denice: Bogota, 1978 2288A-3 (Customs Duti Imcideince: BogotA, 1978 2298A-4 Incidenice of Taxes and Suibsidies on Fuiel Consumption

of Plrivate Cars: Bogota, 1980 2308A-5 Models for Choice of Work Trip Mode: Bogota, 1972

and 1978 231

Chapter 99-I Consolidated Per Capita Revenues: BogotA, 1961-79 2409-2 Structtire of Coisolidated Reveniules: Bogota, 1961-79 2419-3 Fiianicinig of L.ocal Public Expenditiure in Selected Cities

by Type of Revenue 2429-4 Distribution of l.ocal Tax Revenue in Selected Cities

bv Source 2469-5 Structure of Consolidated Total Expenditures: Bogota.

1961-79 257

x CONTENTS

9-6 Subsidies from Public Utility Pricing: Bogota, 1974 2649A-I Per Capita Revenues of District Administration: Bogota,

1960-80 2689A-2 Consolidated Capital Expenditures: Bogota, 1961-79 270

AppendixA-1 Households Interviewed, by Barrio 296A-2 Distribution of Structure Types: Survey and Universe 296A-3 General Characteristics of Surveyed Subdivisions 297A-4 Variables in Superintendencia Bancaria Pirata Subdivision

Survey, 1977 299

Figures9-1 Total Consolidated Expenditure Per Capita of the Bogota

District, 1961-79 2369-2 Composition of the Bogota District's Real Per Capita Total

Revenues, 1961-79 2449-3 Distribution of Situado Fiscal to the Bogota District 2519-4 Distribution of Sales Tax Revenue to the Bogota District,

1975-80 2539-5 Collection and Distribution of Colombian Beer Tax

Revenue 2549-6 Composition of the Bogota District's Total Expenditures

Per Capita 2569-7 Composition of the Bogota District's Capital Expenditures 258

Maps3-1 Bogota: Ring and Sector Systems Based on 1973

Comunas 443-2 Cali: Ring and Sector Systems 463-3 Bogota: Population Density by Comuna 483-4 Bogota: Distribution of Mean Household Income Per Capita

by Comuna 525-1 Bogota: Distribution of Occupations by Sector 102

Foreword

This book summarizes extensive empirical work on Bogota and Cali. Colom-bia, carried out under one of the first comprehensive studies of its kind in adeveloping country. A few similar studies have been done on cities in indus-trial countries, the first being Anatomy of a Metropolis, by Edgar M. Hooverand Raymond Vernon (Harvard University Press, 1959). Although not amodel for this work, it was a source of inspirationi.

Our expectation when beginninig this research was that the developmentpatterns and behavior in Bogota and Cali would resemble those observed inindustrial countries in the early decades of the twentieth century. In fact,the research revealed strong behavioral similarities of households, workers,and firms to those contemporaneously observed in industrial counltries.This is true of patterns of decentralization and movement of firms, of deter-minants of household demand for housing, of travel demand and choice ofmode, and of general urban residential growth patterns. Of course, thereare differences-household incomes are lower, residential densities arehigher, transit use is greater, and service levels are lower in the two citiesstudied than in industrial countries. But it is the similarities, not the differ-ences, that have been most striking.

These similarities in behavior support the view that the basic tenets ofurban economic theory are applicable in large cities where household andfirm decisions are determined in market settings and where passengertravel and freight transport are motorized. A basic conclusion is that citiesare not chaotic collections of unpredictable activities. Their strong commonpatterns of behavior enable us to formulate policies for transport, housilng,urban labor markets, and local public finance and to improve the manage-ment of cities.

Those of us who worked on the project are particularly indebted toColombia for collaborating in this effort and for sharing with us its exten-sive data, intellectual resources, and institutional support.

Gregory K. IngramAdmrinistrator

Research Advisory Staff

xi

Preface and Acknowledgments

This monograph attempts to summarize the main findings of the WorldBank research program known as the City Study. The study, which wasunder the overall direction of Gregory K. Ingram, examined five majorurban sectors-housing, transportation, employment location, labormarkets, and public finance-in Bogota (officially, Santa Fe de Bogota)and Cali, Colombia. The goal of the study was to increase understand-ing of these sectors in order to assess the effect of policies and projectson cities in developing countries. 1

The idea of the City Study was initiated in 1975-76 by Hollis Chenery,then Vice President of the Development Policy Staff of the World Bank.It was his broad vision of development policv r esearch that enabled us tocontemplate and then implement such an intensive program ofresearch in urban economics. The research r eported in this volume wasconducted over a period of about five years, 1977 through 1981, by alarge team at the Urban and Regional Economics Division (headed byDouglas H. Keare) of what was then the Development EconomicsDepartment of the World Bank in Washington, in cooperation with theArea de Distribuci6n Espacial de la Poblaci6n (headed by Ramiro Car-dona) of the Corporaci6n Centro Regional cle Poblaci6n (ccRP),Bogota, ancl the Departmental and Munlicipal Planning Offices (Pla-neaci6n Departamental and Planeaci6n Munlicipal) of Cali. Participantsin the study directed by Gregorv Ingram were Alan Carroll, AndrewHamer, Valerie Kozel, Kyti Sik Lee, Johannes F. I.hin, Rakesh Mohan,Alvaro Pachon, Anna Sant'Anina, and Richard Westin of the WorldBank; in Bogota, Alberto lIernandez. Rodrigo Villamizar, and Amparode Ardila were led byJose Fernanido Pineda. About twentv-five consul-tants and research assistants based in both Bogota and Washingtonassisted. In addition to World Bank research funds, funding was pro-vided by Colombia's National Statistical Agency (Departamento Admi-

xiii

xiv PREFACE AND ACKNOWLEDGMENTS

nistrativo Nacional de Estadistica, DANE) for the 1978 Household Surveyand by the Bogota Chamber of Commerce for dissemination of studyresults. The project also benefited from the direction of the ColombianAdvisory Committee, consisting of Eduardo Aldana, Rodrigo Botero,Bernardo Gaitan, Pedro G6mez, Jorge M6ndez, Miguel Urrutia,Eduardo Wiesner, and the Resident Representative of the World Bank atthe time, Ian Scott. At the World Bank, the consistent interest andencouragemenit given by Ardy Stoutjesdijk and Anthony Churchill werevery helpful.

In this volume, I attempt to integrate the main findings of the differ-ent segments of the project. Although I have presented the findinlgs as Ihave interpreted them from the work of the whole team, the primarycredit for each portion of the research should go to its originators (itali-

cized in the entries).

LAIND MARKETs: Rakesh Mohan., Rodrigo Villamizar, Gregory K.Ingram, Ricardo Paredes, Guillermo Wiesner, M. Wilhelm Wagner.

HOUSING AND RESIDENTIAL LOCATION: Gregory K Ingram, Andrewflamer, Amparo de Ardila, Alan Carroll, Jose Fernando Pineda,Rafael Stevenson, Oscar Borrero, Anna Sant'Anina.

TRANSPORTATiON: Alvaro Pachon, Valerie Kozel, Richard Westin,Gregory K. Ingram, Alberto Hernandez, Jose Cifuentes, Emilio Latorre.

EMPLOYMENT LOCATION: Kvu Sik Lee, YoonJoo Lee.

LABOR MARKETS, POVERTY, INCOME DISTRIBUTION: Rakeshi Mohan,Gary Fields,Jorge Garcia, M. Wilhelm Wagner.

PUBnIC FINANCE: Johannes F Linn, Alberto Hernandez, Jeffrey Lewis,Caroline Fawcett, Juergen Wolff, David Grevtak.

OVERALL PATTERNS: Gregory K Ingram, Rakesh Mohan,Jose FernandoPineda, Alvaro Pachon.

DATA DOCUMENTATION: Yoon loo Lee. Nelson Valverde.

As may be expected from the quantitative research in this volume, alegion of research assistants has contributed to the work. Thanks aredue to YoonJoo Lee, Sung Yong Klang, Robert Marshall, Kathie Terrell,Wili Wagner, Jon Roseman, Nancy Hardinie, Mark Snydernmani, LeslieKramer, and Nelson Valverde, all of whom gave much bevond the call ofduty in managing the vast volumes of data processed in this study. InColombia we also received assistance from Sonia Rodriguez, MarcoTulio Ruiz, Maria Clara de Posada, and Alejandro Vivas.

We received extensive cooperation from maniy organizations inColombia. Work in Cali was carried out with staff from the Departmetn-tal Planning Office (Planeaci6n Departamental), headed by LacidesReves, and from the Municipal Planning Office (Planeaci6n Munici-

PRFACE AND ACKNOWLEDGMENTS xv

pal), headed by Jaime Cifuentes. We collaborated with the IntegratedPlan for the Development of Cali (Plan Integral para el Desarrollo deCali, PIDECA), directed byjulian Velasco. Among the staff associated withour work in Cali were M. Sanchez, Hugo Garcia, M. V. Fuentes, J. Oso-rio, F. Fajardo, M. Santacruz, S. Prado, A. Rivera, and G. L6pez. We werealso generously assisted by the staff of the Bogota District PlanningOffice (Departamento Administrativo de Planeaci6n Distrital), theNational Centre for Building Industry Studies (Centro Nacional deEstudios de la Construcci6n, CENAC), the Social Security Institute (Insti-tuto Colombiano de Seguaros Social, Icss), and the SuperintendenciaBancaria. Enrique Low and Clara Eugenia Lopez, Controllers (Contra-lores) of Bogota District, took special interest in the public finance workand gave generouslv of their staff time.

We were particularly fortunate to have been able to collaborate withmembers of DANE, who provided us with unfailing and careful assis-tance.They agreed to incorporate our requirements into their scheduleof quarterly household surveys, which resulted in the World Bank-DANECity Study Household Survey of 1978, on which many of our results arebased. The high quality of this survey owed much to the diligence of itsDANE supervisors, Roberto Pinilla and Maria Christina Jimnenez. Thecontributions of Gary Losee, Alfredo Aliaga,Jairo Arias, Alvaro Pachon,andJose Fernando Pineda to its design and implementation were nota-ble. We would like to thank the authorities of DANE for so r eadily provid-ing all the data sets used in the study. My own participation in theorganization and collection of the data in the 1978 Household Surveytaught me much about the practical difficulties and pitfalls involved inthe collection of primary data, on which we economists are so depen-dent but which we seldom understand.

For me, this work has spanned a period of what I presume will beabout a third of my professional life. Douglas Keare inducted me intothe World Bank in December 1976 to work in the study. Although mostof the research had been completed by late 1981, the changing interestsand reassignment of all participants made it very difficult to bring tofinal form all the work that had been done. Since 1980, 1 have returnedto the Government of India twice to work on housing, urban develop-ment, and industrial development in a policymaking capacity. GregIngram, in his various incarnations in the 1980s, was able to find theresources to enable me to extricate myself from the government ofIndia at various times, first to return to the World Bank in 1984, thenagain in 1989 and 1990 to write this volume as well as the earlier mono-graph (Mohan 1986). I appreciate his efforts in providing me theseopportunities to conclude an enterprise we started jointly in January1977. These "sabbaticals" have been invaluable in helping me keep upmy research interests and in giving me respite from the pressures of nor-mal governmental work. It was also generous of Otima Bordia and A. N.

xvi PRIFACE AND ACKNOWLEDGMENTS

Varma, Secretary, Industrial I)evelopment, Government of India, in1989 and 1990, respectively, to give me time off from the ministry dur-ing difficult times of onerous work. The constant encouragementreceived over the years from Doug Keare and johannes linn has alsobeen instrumental in enabling me to finish this work.

In Colombia, we benefited greatly from having as our main collabora-tors Ramiro Cardona and Jose Fernando Pineda. Their war m hospital-ity, generously extended to the whole team, made everv visit toColombia a pleasure. They contributed greatly to our knowledge andunderstanding of Colombia, making possible a truly collaborativeproject.

Geri Mitchell organized the final processing of this manuscript, doingmuch of it herself. She was assisted byJean Ponchamni, Maria Diniatu-lac, Maria E. Sainchez, and Helen Lee. I am grateful to them for beingable to do th1is in addition to their normal duties.

The final production of this book has taken as long as its creation.Well-considered comments from three anonymous referees led to veryuseful revisions. Its readability owes a lot to the careful editing of WillaSpeiser Jeanne Rosen, Kathryn Kline Dahl, and Deirdre Murphy.

Our quest in this study was to find general approaches to studying theworkings of large cities in developing countries and to achieve rnodes ofuinder-staniding the behavior of and interactionis among the participantsin the myriad activities in a developing metropolis. We feel that weachieved some success in this quest and that our findings have generalrelevanice to tinderstanding the growth of large cities. Our approach wasto describe and analyze the various patterns of behavior that are foundto be stable over time. For example, some of our work has been corrob-orated by the subsequent works of Steve Mayo and Steve Malpezzi onhousing ancl by Kyu Sik Lee's continuing efforts on employment loca-tion. On other issues, our results are still arnong the few available forcities in developing countries. In writing this book, delayed though it is,I hope that large cities will begin to be better understood by profession-als and policymakers alike. I also hope that more detailed work of thiskind will be forthcominlg so that we can continue to improvc our tinder-standing of the developing metropolis.

My greatest debt, intellectual and otherwise, in being able to conductthis study and to bring it to a close in this book is to Gregory Ingranm. Tohim, my deepest gratitude.

I dedicate this book to my wvife, Rasika Khanna, for allowing me todesert her at various times in the raw summer heat of Delhi in order tofinish this work, which has followed me for more than a decade. Mv chil-dren, Tarini and Rasesh, who have appeared during the course of thewriting of this book, will have to wait for my next effort to be eligible fordedication .

PRFACE AND ACKNOWLEDGMENTS xvii

Note

1. Two monographs have been published: Rakesh Mohan, Werk, Wages, andWelfare in a Developing Afetropolis: Consequences of C;rowth in Bogota, Colombia (NewYork: Oxford University Press, 1986); and Ktu Sik lee, The Location of Jobs in aDeeveloping AMetropolis: Patterns of Gro70th in Bogota and Calm, Colombia (New York:Oxford University Press, 1989).

Chapter 1

Introduction

The total urban population in developing countries has roughly quad-rLupled in the past four decades, from less than 300 million in 1950 tomore than 1.2 billion in 1990. The rapid growth of cities is among themore striking features of the developing world. In 1950 there were onlythree cities in the developing world with more than 5 million people;now there are at least twenty-four such cities in developing counltries,compared with only ten in developed counltries. The word "metropolis"used to describe Western cities such as New York, Chicago, London,Paris, and Rome. Today it is as likely to bring to mind Buenos Aires, SaioPaulo, Mexico Citv, Tokyo, Manila, Seoul, Bombay, New Delhi, or Cairo.More often than not, current discussions of urban or metropolitanproblems concern cities in developing couLntries.

Since large metropolitan centers did not appear in what are today'sdeveloped countries until their economies had reached a comparativelyadvanced stage, the emergence of such cities in poor countries has fre-quently been regarded as an unusual phenomenont. A metropolis, oncea source of national pride, now inspires awe if not fear. To many, largecities in developing countries conjure Up images of gigantic slums,floods of destitute migrants congregating in ramshackle shanty towns,cacophonous traffic choking the main arteries, and ineffective local gov-ernment strugglillg to proxide basic public services. Because of thiswidespreacl negative view and the alarm it evokes, urban policies oftenattempt to slow the growth of large cities rather than to manage theirgrowth in a healthy fashion.

Despite misdirected and even counterproductive policies, large citieshave continued to expand and have usually prospered in the process. Bythe end of the century the developing world is expected to have miorethan fifteen "megacities" with at least 10 million people each and m)or-ethan a hundred cities with at least I million people each. United

I

2 UNDERSTANDING THF DEVELOPING; METROPOLIS

Nations' projections suggest that more people in the developing worldwill be living in urban than in rural areas by the year 2020. This isalready trule in Imlost countries in Latin America, East Asia, and the Mid-dle East. South and Southeast Asia as well as Chiina are rapidly movingin the same direction, followed by Sub-Saharan Africa.

A positive approach to the problems of large cities is a must if govern-ments are to meet the challenges posed by city growth. In thle preindus-trial age the concentration of people in cities was a religious, political,or military plenomenon. Today it is an economic phenomenoll, and itmust bc analvzed as such. Evolving constructive metropolitan policiesrequires an und(ier-standing of how cities work.

The need for such understanidinig led the World Bank to launch aresearchl program in the 1970s knowvn as the City Study. This effort, forreasons outlined below, has focused primilarily on1 Bogota and secondarilyon Cali in (Colombia. The main objective of this study has been to learnmnore about the intrinsically complex and interrelated phenomena thattindenlie the rapid growth of a city. Those of us involved in the studyhave sought to understand how thie components of a city's economy-itshousillg. employment, labor force, land market, transportation systems,and public services-interact and behave in the context of rapid citygrowtlh. Our ultimiiate goal has been to bring about a more positive ori-enitationi to urban policy.

Considerable research exists on1 tile old cities in Europe and NorthAmerica. andI their strutctur-es are reasonably well understood. The work-ings of their housinig and( labor markets, the locational decisions of theirresi(denits and employers, and the patterns of their traffic and transportare all regarded as tractable. There are stvlized facts about these citiesthat appear to have wide applicability. But how relevant is this knowl-ecige for cities in developing counaries? Do we need to apply differentmodes of understanidinig to these cities? How have their populatiolnsresponided to rapicd growth? What dlo project andl city planners need toknow abotut the processes ancI patterns in these cities so they can con-tribute mor-e effectively to urbani development? Hlow much interventionin different markets is necessar-y for a city to accommodate rapidgrowth? The City Study was designedl Io illuminate these and manyrelated quuestions.

The broad theme of' the City Study is behavioral adaptation to rapidchange in the context of rapidly growinig cities in developing countries.The impetus fol this work arose from the World Bank's increasiniginvolvement in urbani projects, beginning in the early 1970s. Along withthis involvemiienit came a greater awareness of the existence of urbanpoverty and its concomitant problems. Whereas the manifold problemsof rural poverty had been studied for a long time, little information wasavailable on the dimensions and nattire of urban poverty. This led to the

INTRODUCTION 3

formation of an Urban Poverty Task Force at the World Bank in 1974.The Task Force assembled a group of academics working on these issuesand commissioned a series of studies. These efforts did not produce acoherent program for addressing the issues of urbani poverty, but theydid result in greater attentioni to urban problenms and greater apprecia-tion of the complexity and interrelatedness of urban phenomena. Italso became clear that one feature of the City Studv should be a focuson urban povert.

At the time this study was being launched, the notioln that everythinigin a city is interr-elated hacl taken hold, and it was widely accepted thatthese interrelationships were best captur ed in large-scale urban models.Researclher-s in the 1960s and early 1970s, especially in North Americaand the United Kingdom, had been very optimistic about the utilityandc feasibility of constLructing such models and usinig themil to undler-stand how cities work and hence to plan better. Two broad modelingapproaches were prevalent: behavioral or analytical models and opera-tional or planning models. A critical review of these modeling effortswas conducted as a prelude to the City Study (see Mohan 1979), ancd itled to consider-able skepticism about thle utility and practicality of large-scale models. In (lesigninig the City Study, we opted instead for detailedinvestigations of each area of urban decisionmnaking. It was more impor-tant, we clecided, to tinderstand and model the behavior of the variousactors and markets that constitute a city than to replicate its ttjnctiolIs ina large-scale moclel.

The approach we adopted was therefore 1o model cliffe rent markets,as well as the behavior of individtials. hoiseholds, firms, and the publicsector within those markets. We wanted especially to examine theresponises of these actors in the tace of extremilely rapid change. How,for example, (1o people react to shifting prices in the housing market?How much do people do for themselves, and how should the public sec-tor intervene? What kinds of supply responses can be expected whendemand for housing burgeons? How does the public sector cope withwidespread illegality in land transactionls? How do people choose theirmode of transport? W'hat kind of transpor-t services appear in respoliseto a variegated demand? How are inconmes determined in an urbani set-ting, and how are thev affected by changes in labor market conditiolns?How clo people clecide whether or not to participate in the labor mar-ket? These are the kinds of questions we sought to address through thebehavioral moclelinig of actors and markets. We felt that urball policiescould be much more effectivelv designed if there were better apprecia-tion of the underlyinig behavior that results in the phenomenia observec.

The objective of the City Study was not to propose a specific programto guide city growth, but rather to develop tools (models as well as otheranalytical niethods) that cotld be used to estimate the spatial and eco-

4 UNDERSTANDING THE DEVELOPING METROPOI-IS

nomvic effects of different kiiids of policy interventiol. The researchstrategy was threefold: (a) to svstematically describe the current spatialpatterns of' the various economic activities that constitlte a city, alongwith recenit changes in the observed patterns; (h) to model behaviorand to estimate parameters useful for understanding how the maintl-ban] constituents respond to rapidly changing conditions; and (c) toassess the policy imnpact of these findinigs. Tle svstematic descriptionWould enable us to determine howv the spatial patterns of activitiesevolve in response to changes such as the decentralization of'jobs andresidences in a growing city. These descriptions would in turni help usformulate hypotheses to he tested, andl the empirical estimations ofmodels wotld vield the paramiieter estimates. We would also be able toobserve the apparenit effects of policies that attempt to alter the spatialdistributioni of activities.

The magnitutde of this task, we knew, would depend criticallv on howsimilar the fundamenital processes of tirbani development are at differ-ent stages of economic development. If existing tools (lesignied to studycities in rich countries could be applied or adaptecl for use in poorercounJtries, thie task Would be that miuch easier. We looked for clues aboutthe transfer-ability of these tools bv comparing the spatial patterns of cit-ies in r ich and poor countries. Ouir assessment was that the similaritiesoutmeighed the dissimilarities (see Ingram and Carroll 1981). Theresearch was therefore also designed to test whether familiar toolsdevisedI to study cities in indltistrial couLntries could be transfer-red to thestucdv of growing cities in dleveloping coUntrties.

Most urban studies concentrate exclusively on issues relate(d to thesupply of'inifastlrlcttre, housing, and transport services. The City Sttudyaddresses these issues in cletail. But it gives equal attention to the work-ing of the labor market and the resulting income distribution, becausethe emphasis is on the behavior of people and firms in a situation ofr apid growth. A pr-ioI knowledge of household behavior and the labormarket activitv of' individuals and their employers is necessary beforethe housing, transport, andc infrastructuLe issues can be addressed. Map-ping the spatial distribution of income and explaining its overall trendare crticial to comprehending andl predicting the natuLre of demands forinfrastrUtcttUre, housing, and transpor-t that are likely to arise in a grow-ing city.

City Selection

Having decided to launch a program of' research designed to improveunderstaniding of how cities in developing countries work, we nextdecided to study one city in cletail rather than to attempt to draw con-

INTRODUCTION 5

clusions from cross-country patterns. This seemed the appropriate strat-egy, given our intention to focus on understanding the intracitv patternsand behavior that characterize a rapidly growing city. Resources, bothfinancial and human, did not permit a simultaneous, in-depth study ofseveral cities in different countries.

Once the decision had been made to focus on a single city, it becameapparent that Latin America was the place to look. The urbanizationexperience of Latin America in the 1950s. 1960s, and early 19 70s was aprecursor of the fast urban- growth phenomenon common to mostdeveloping countr-ies. Maniy Latin American countries had alreadyreached levels of urbanization that exceeded 50 percent, and there werea number of large cities from which to choose. Furthermore, the prox-imity of the region to Washington, D.C., where the World Bank's head-quarters are located, was a big advantage, since the study was expectedto involve intensive work in the selected city and hence frequellt travel.

Certain additional criteria led us to choose BogotA as the primary sitefor this study. First, the policy and planning climate in the citv and inthe country had to be receptive to such a study. Ae wanted the policy-makers, plannlers, and administrators to be interested in the results andcapable of absorbing them. This depended on the availability of profes-sionals working on urban issues both within and outside the govern-ment, and on1 the interest of the nationial and local governments in themanagement of city growth. The potential for the World Bank to makefuture urban-r elated investments was another considerationi.

The second criterion was a favorable research climate in the selectedcitv and country. To carry out the study in a reasonable time, it wasdesirable for the city to have both a rich data base and competent indi-viduals or institutions that could participate in the work. Moreover, thepotential usefulness of the study would be enhanced if the localresearch capability existed to utilize the results so that the knowledgecould be transferred to other researchers when the study was concluded.The availabilitv of existing data bases in time series, cross-sectional, anddisaggregated form was an important asset to be considered, as was theavailability of relevant studies.

A third criterion was the absence of an atypical development patternthat would limit the transferability of the findings. The city had to belarge enough to exhibit adequate variation in land prices, travel pat-terns, density levels, and the like so that the analyses could be applied toa wide ranige of situations. Accordingly, idiosyncratic administrative cit-ies such as Brasilia were excluded. The relative stability of thc nationlaleconomy was also important. A cultural context commnon to many COUI1-

tnes was an added asset.Given these criteria, the choice was narrowed downl to four countries:

Brazil, Colombia, Mexico, and Peru. Rio de Janeir(o was eliminated

6 L!NDERSTANDING THE DEVELOPING1 METROPOLIS

because of its highly untusuial geography, and Sao Paulo and Mexico Citybecause they are far too large, complex, and unmanageable for practi-cal studv. Government interest was found to be keenest in Colombia.From discussions with government officials, consultants, and academics,it also became apparent that a concern for intracity problems was wide-spread in that country.

Bogota exhibited characteristics representative of a cross-section ofLatin American cities (see Ingram and Carroll 1981), which greatlyenhanced the chances that the findings would be transferable to othercities. It also had a rich data base on which to build (see the appendix tothis volume for details). Moreover, Colombia had a historv of policy-related analytical work on urbani issues, influeniced in part by the life-long interest of Lauchlin Curr-ie. The importanice of urbanizationi ineconomic development and of housing as an economic activitv had longbeen emphasized by Currie. In 1950 hie led the first World Bank missionto a developing country, which happened to be to Colombia. He latermoved there and exerted a lasting influence on national planning. It isalso worth noting that Albert Hirschman derived the crucial insights forhis book, rhe Strateg- of Economic Development (1958), during a three-yearstint in Colombia's National Planning Department: that the Interna-tional Labour Organisation (tt.O) sent an infltuential employmnenit rnis-sion headed by Dudley Seers to Colombia in 1967 (see iL,o 1970); andthat a detailed country report on Colombia was among the early WVorldBank country economic reports published (Avramovic and associates1972). Of direct relevance to our study was the important urban devel-opment study of Bogota undertaken with support from the UnitedNations Development Programme (UNDP) in 1970-74 (known as thePhase 11 Study). This work had generated a substantial amount of bench-mark data, in particular a comprehensive household survey in 1972.

At the time we began the City Study in 1977, other scholars hadrecently completed or were engaged in research on Bogota. AlbertBerry and Miguel Urrutia had been working for years on issues con-cerning the labor market and income distribution (see Berry 1975a,1975b; Berry and Soligo 1980; Berry and Urrutia 1976; Urrutia 1969,1985). Gary Fields and Helena Ribe were interested in similar issues(see Fields 1975; Fields and Marulanda 1976; Fields and Schultz 1980;Ribe 1979). Richard Nelson, r. Paul Schultz, and Robert Slighton(1971) had also worked earlier on the interrelated issues of population)growth, internal migration, and labor market adjustment. Peter Amato(1968, 1970a, 1970b) had documented patterns of urban residentiallocation a decade earlier. Georges Vernez (1973) and Rodrigo l.osadaand Hernando Gomez (1976), among others, had worked on phraltadevelopments and incremental housing development. Harold Lubell

INTRODUCTION 7

and Douglas McCallum (1978) had addressed the relation of employ-ment issues and urban development. At the World Bank, several studieson Colombia had been conducted or were in progress in the mid-197Qs:Johannes Linn was carrying out a large program of research on urbanpublic finance (see Linn 1976a, 1976b, 1979, 1980a, 1980b); MarceloSelowskv was working on a study on the distribution of public services(see Selowsky 1979); William Doebele and Orville Grimes had con-ducted studies on the working of the urban land market (see Doebele1975; Doebele, Grimes, and Linn 1979); and Mariluz Cortes wvasinvolved in a project investigating the role of small-scale enterprises(Cortes, Berry, and Ishaq 1987). Thus a vibrant research ambienceexisted both in Bogota and at the World Bank that was very conducive loinitiating the City Study in Colombia.

The active involvement of local researchers was essential to the suc-cess of this endeavor. Bogota was well endowed with research institu-tions and university departments that had some experience working onti-ban issues. It was therefore easy to find researchers who could partici-pate in the study. The work was eventually based in the CorporationCentral Regional de Poblacion (('CRP)-an institution engaged in popu-lation studies, ur-ban studies, and economics research.

The interest that different levels of government have taken in thisstudy is borne out by the many contributions of public institutions.The National Planning Department (Departamento Nacional de Pla-neaci6n) was active in initiatinig and facilitating the study; the NationalStatistical Agency (Departamento Administrativo Nacional de Estadis-tica, DANE) shared raw data from past surveys and conducted the 1978City Study survev; and the Banking Regulatory Agency (Superintenden-cia Bancaria) provided data on pirate land developments and contrib-uted staff resources. The citv government of Bogota made staff availableto help conduct the public finance portion of the study.

When we were getting under way, the city government of Cali, alongwith the State Planning Office (Planeaci6n Departamental) of Valle, thestate in which Cali is situated, expressed great interest in our work. Theyinvited Us to broaden the study to include Cali, which we did. The CityPlanning Office (Planeaci6n Municipal) provided extensive staffresources to iriiplemerit the study in Cali.

Finally, the World Bank had a large program in Colombia. It wastherefore expected that the results of the study would find a ready audi-ence within the institution.

In short, Bogota satisfied all the criteria that had been laid out forselecting a site for the City Study. Cali was added because of the localinterest expressed. It proved very useful as a comparator for assessingthe generality of the findings. Whereas the primary studies were under-

8 UNDERSTANDING; THE DEVELOPING METROPOLIS

taken in Bogota, an attempt was made to duplicate the research wher-ever possible in Cali. Wherever significant differences exist, they havebeen noted. Otherwise, most general conclusions apply to both cities.

The Macroeconomic Setting: Were the 1970s Exceptional?

This study is being published many years after completion of the fieldresearch, which was carried out fiom 1977 to 1980. The question ariseswhether the results were unduly influenced by any special conditionsprevailing in Colombia from 1972 to 1980, the period covered by thestudy. A related question is whether any developments during the 1980sgive cause for reassessing the results. It is also useful to place the city-related issues addressed in this studv within the national economic coIn-text of the time.

Since the study was designed to enhance our understanding of thebehavioral underpinnings of how cities work, especiallv in the contextof rapid growth and change, the delay in publication should not makeour findings obsolete. Indeed, the economic changes that haveoccurred during the 1980s underscore the usefulness of taking a behav-ioral look at how a city functions rather than adopting a prescriptiveapproach to planning.

Viewed in retrospect, the economic environment in Colombia wasmore buoyant in the 1970s than in most other recent decades, althoughnot exceptionally so. In the early 1980s Colombia was on the verge ofeconomic crisis for reasons that were in part exogenous: a drop in cof-fee prices, high world interest rates, an international recession, and thebackwash of the debt crisis affecting the rest of Latin America. Theseexternally induced economic difficulties were exacerbated by burgeon-ing domestic problems, including high fiscal deficits and a significantappreciation of the real exchange rate. Growth in employment slowed,unemployment rose, and the external trade deficit widened. Althoughthe situation in Colombia was not nearly as strained as it was in mostother Latin American countries, many of which were highly indebted,international banks, facing difficulties elsewhere on that continent,hardened credit terms for Colombia. The optimistic tone of this study, areflection of the economic environment of the 1970s, would probablyhave been tempered if the period of observation had been the 1980s.Nonetheless, most of our findings are unaffected by the events of thoseyears.

The three decades prior to the 1980s marked an important demo-graphic and economic transition for Colombia. In the 1960s the rurallabor force ceased to grow (Urrutia 1985) and the urban labor forceexpanded rapidly. People living in urban areas began to outnumber

INTRODUICTION 9

rural dwellers: they increased from 39 percent of the poptulation in 1951to 52 percent in 1964 and 60 percent in 1973. By the late 1970s two-thirds of Colombians resided in towns and cities. Colombia, like mostcountries experiencing urban growth, had uLndergonie a striking demo-graphic transition in a short time.

In the 1980s this process slowed down; urban residents accounted for68 percent of the population in 1986. The urbanization process duringthe 1950s, 1960s, and 1970s was therefore somewhat different from thatin the 1980s. That some of our findings differ from those of earlier stucl-ies (for example, 11.0 1970; Lubell and McCallum 1978) is partdbecause we were observing the consolidation phase of' Colombia'surban history, whereas others had observed the effervescent phase,when urbatn growth seemed excessively rapid and endless. One of thelessons drawn in this study is that understanding the growth processes ina specific city requires an appreciation of the national urban growticontext.

Colombia's demographic tranisitioni is also evident fiom dramiiaticchanges in other social indicators duting the 1960s and 1970s. Thecrude birth rate (CBR) fell from 45 per thousand population in 1963 to31 per thousanid in 1974. The CBR then declined more gradually, reach-ing 27 per thousand by 1985. Similarly, infanit mortality fell from 6.5 perthousand births in 1963 to 4.0 per thousand births in 1974 and to 3.2per thousand births in 1985. Primary schooling became unliver-sal in the1970s, and secondarv school enrollment increased from 17 percent ofthe school-age population in 1963 to 39 percent in 1974 and 50 percentin 1985. These indicators confirm the extremely rapid changes thattook place in the 1960s and 1970s and the slowdown that followed in thelate 1970s and early 1980s.

The economic impact of these social and demographic shifts was far-reaching. The rapid growth in population, particularlv in urban popula-tion, placecl severe strains on the suppliers of urban services. The slow-ing of change toward the late 1970s made it easier for the housingsupply and other urban amenities to catch tip with the backlog ofurgent demand. Similarly, the impressive expansionr in the supply ofeducation starting in the 1960s, especially for females, transformed thequality of the labor force within two decades. Some of the employmentpressures durinig the 1980s may be relatedl to this expansioni, which hadthe effect of reducing the returns to secondary and higher educationand so contributed to improvinig the distribution of income (see chap-ter 4 in this Volume; Mohan and Sabot 1988).

Colombia sustained a respectable but not exceptionally high rate ofeconomic growth in the decades following WNtorld War 11 (see table 1-1).From 1950 to 1966, the gross clomestic product (GDP) grewv an averageof almost 5 percent a vear. This acceler-ated to about 6.4 percent a year

I ( UNDERSTANDING THE DEV'ELOPIN(; METROPOLIS

between 1967 and 1974 but fell back to about 5.5 percent for the rest ofthe decade. Then came the exceptional deceleration of the 1980s.Income per capita dropped during 1980-83 for the first time in threedecades. Whereas economic growth was just respectable, the increase inemplovment growth in the late 19 7 0s was exceptional (see table 1-1 andchapter 5).

The existence of relatively tight labor markets, the absence of notablesegmentation, and the measurable increase in real wages, all of whichwvere observed in our study, were clearly influenced by unusually highgrowth in employmnent during the period. In our judgmenit, however,this phenomenon was not entirely fortuitous but was in part the conse-quence of appropriate policies and sound inacroeconomic manage-ment.

Indeed, Colombia has benefited from a long record of sound macro-economic management, which has usually responded successfully tochanges in the economic environment. Like most other developingcountries, especially those in Latin America, Colombia emphasizedimport-substitution industrialization policies during the 1950s and1960s. Its periods of economic growth and stagnation have been linkedto its terms of trade, especially to the movement of world coffee prices.In the late 1970s coffee prices were booming and the economy wasdoing well: in the early 1980s the opposite was true. The post-WorldWar II import-substituting strategy led to the consolidation of industr-ialgrowth, particularly in Bogoti, which rapidly surpassed Medellin asColombia's prime manufacturing center.

The Lleras government (1966-70) introduced policy changes thathelped lay the foundationi for some of the income and employmentgains of the 1970s. Export pr-omotionl was pursued activelv bv introduc-ing various incentives; a crawling peg exchange rate policy was adoptedto maintain a realistic real exchange rate; fiscal discipline was exercised;and a mild program for the deregulation of some import controls wasput in place. This policy package was largelv continued bv the Pastranagovernment (1970-74), which, in addition, consciouslv introduced an

Table 1-1. Economic Growth in Colombia, 1950-86(percenlt)

AV.erage ainnualgrouoth Share of capital

Period (;DP Emplment Capital in G;DP

1950-57 4.9 2.5 5.3 44.11958-66 4.9 2.4 3.9 43.31967-74 6.4 2.8 4.1 42.71974-80 5.5 4.1 4.9 40.71981-86 2.7 2.8 4.2 37.8

Source: Garcia (1988).

INTR0DUcT'rI)N I I

urban development thrust by promotinig constructioni activities. Theresult was improved utilization of capacity in the industrial scctor and asignificant increase in noncoffee exports. The constructioni boom wasconcentrated primarily in Bogota and secondarily in Medellin and Cali.

In the field of housing, the most significant development was theintroduction of a new savings instrument that assured savcrs a real rateof return on their accounts through indexation. The f-unids mobilizedthrough this instrument could then be used to finanice housing.Although there is some debate about the overall impact oin housinig ofthis financial innovation, it undoubtediv helped foster middle- aiidupper-middle-income housing activity (see chapter 7). Housinig andfinancial markets did become better linked as a result.

The l.6pez administration (1974-78) deemphasized suppor-t for con-structioni and adopted an explicit urban cleconcentrationi program todirect economic activities away from the largest cities of BogotA,Medellin, Cali, and Barranquilla. Emphasis was also put on1 usirig publicresources to help the poor: spending on public health and educationwas promoted, whereas investments in infrastructure for energy andtransport were downgraded.

The net result of the policy environmiient that originated in the mid-1960s was that sectors wvith low ratios of capital to labor grew faster thanother sectors, noncoffee exports increased, and educationi expan(le(Irapidly along with other- social services (see Urrutia 1985). This led tothe remarkable employment boom in thie late 1970s.

At about the same time, however, tihe policy package began touLnravel. World coffee prices shot up to unipr-ecedenited levels, as didunrecorded foreign exchange carnings fromil the drug trade. The bal-ance of payrnents went into surplus. Silnce it was difficult to stelilizethese foreign exchange inflows, the real exchange rate appreciated sig-nificantit, and nioncoffee expor-ts began to falter. WN'hen coffee pricescrashed in the early 1980s, as showni in table 1-2, a severe balance of pav-ments crisis ensued; the currenit account deficit was 6 percent of (;)IP in1984. The policy response was to impose stringent controls on ailmost allimports. The economic difficulties were coinpouinded by rising budgetdeficits, which reached 6.8 perccnt of (.,DP in 1984. These were caused inpart by the bunichinig of some large infrastriucture investimienits, espe-cially in the power sector, and by the drop in coffee revenues. The eco-nomic downtur-n during 1980-84 caused significant unemploymcit.

Other related developments may have contributed to the slowdown inemployment growth, most notably in organized manufactur-ing. As wasdocumented in a 1987 internal WorldI Bank report, wages in the orga-nized sector began to rise in the late 1970s. and the differential betweenwages in the organized and the unorganiized sectors may hiave widened.Wage agreements. which had usually beni doncluded every other year,

Table 1-2. Prices and Wages in Colombia

lnlicatars 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988

Intlexes

Exchange rateNominal

(pesos per dollar) 20.1 22.0 23.8 27.1 31.2 35.0 36.8 39.1 42.6 47.3 54.5 64.1 78.9 100.8 142.3 196.1 243.0 299.0Real (average) n.a. n.a. n.a. n.a. 100.0 95.4 85.7 85.5 81.7 83.5 81.6 75.6 73.6 79.9 91.4 108.5 108.2 105.6

Wages' 10)5.2 108.6 110.5 111.8 111.3 10)9.2 109.7 121.7 127.6 129.5 134.2 132.6 132.9 132.4 n.a. n.a. n.a. n.a.Coffee price index 80.9 89.3 104.8 98.6 100.0 172.6 215.7 151.() 131.2 111.4 74.9 77.0 74.2 77.2 81.7 115.2 64.7 74.8

Annutalgrowth rate(pertent)

GDP deflator(1975=100) 10.8 13.0 20.2 25.4 22.8 25.5 29.1 17.1 24.1) 27.6 22.8 24.8 20.4 22.2 23.2 24.0 22.8 27.0

Exchanige rateNominal 9.4 9.7 8.2 13.8 15.1 12.1 5.1 6.3 8.8 11.2 15.2 17.6 23.1 27.8 41.2 37.8 24.0 23.0Real n.a. in.a. n.a. n.a. n.a. -4.6 -10.1 -0.3 -4.4 2.2 -2.3 -7.3 -2.7 8.5 14.4 18.7 -0.3 -2.4

Whlolesale piices n.a. n.a. n.a. nia. n.a. 22.9 26.7 17.6 27.8 37.5 23.5 24.6 18.0 21.2 23.0 24.4 25.2 283

n.a. Not available.a. 1970= 100.Source: Internal World Bank documilents (unipublishied 1987 and 1989 country economic memoranida); Thomas (1985).

INTRODUCTION 13

became an annual occurrence in the late 1970s. This put added pres-sure on real wages. One feature of the rising wage costs in the organizedsector was the increasing share represented by employment benefits.The appreciation in the real exchange rate was exacerbated by thesedevelopments, and the growth of the manufacturing sector, especially inexport-oriented industries, was adversely affected. Although this studydid not find significant evidence of segmentation in the labor market(see chapter 5), some observers in the 1980s did. It is possible that cer-tain developments noted in the 1970s could have led to increasing seg-mentationi later.

A specific feature of the rising unemploynment in the 1980s was highunemploymenit among educated women. As this study documents atlength in chapter 5 (see also Mohan 1985), the supply of women in thelabor force grew as access to education expanded in the 1960s and1970s. Some of the rise in urban unemployment rates in the early 1980sstemmed from the increasing numbers of women desiring to work out-side the home.

The travails of the early 1980s have led more recent administrationsto adopt policies encouraging the use of labor. The government hasalso kept the real exchange rate on a par with rates of the mid-1970s,reduced the fiscal deficit to manageable levels (1.4 percent of GDP by1987), and brought the current account back into balance by 1987.These initiatives have been helped by higher coffee prices. An attempthas also been made to shift public spending away from capital-intenisiveprojects. A war on1 poverty, launched by the Barco government (1987-90), emphasized extending services to the backward regions of thecountry and providing health and nutrition services, education, andbasic utilities to the poor.

Shelter is an important component of this war on poverty. A wvelcomechange in approach is the recognition that slum upgrading is a key com-ponent of housinig strategy and that communiity participation is a mustin this process. Less welcome is the return to an earlier strategy ofdirectly investing in housing through the National Housing Agency(Instituto de Credito Territorial) (see World Bank 1990). Moreover,inadequate attention is being given to rental housing as a vehicle for sat-isfying the needs of the poorest (see chapter 7). In light of C(olombia'sexperience in the 1980s, our analysis of shelter needs and patterns froma behavioral viewpoint, and our findings about effective policy initia-tives, remains valid (see chapter 10).

A final feature of Colombia's economic environment in the 1970s isrelevant to our study. Government policies after the mid-1960s did notdiscriminate against agriculture. Helped by the favorable terms of tradethat emerged, the agricultural sector prospered (Garcia 1983; Thomas1985). The coffee boom fiom about 1974 to 1980 also contributed, as

14 UNDERSTANDING THE DEVELOPING METROPOLIS

did the prospering drug trade. Furthermore, the introduction of newcoffee varieties and of new agronomic practices increased the demandfor labor, and that in turn helped tighten Colombia's rural labor marketin the late 1970fs. Since these policy and economic changes coincidedwith the demographic transition described earlier, the suipply of rurallabor tightened further and real wages rose, bringing improvements inincome distribuition (Urrutia 1985). These developments were reflectedin the urban labor market as well, putting pressure on real wages in thelate 1970s. But the deteriorationi in the general economic situation inthe early 1980s, coupled with the fall in coffee prices, erased some of thegains of the 1970s.

Overall, the late 1970s may be seen as a transitional period durinlgwhich the urbanization process in Colombia was traversing the upperbend of the sigmoid path followed bv most urbanizing countries.' Inconducting this study, we observed the rapid changes in Bogota (andCali) resulting from the rapid demographic transition that Colombiahiad gone through in the 1960s and 1970s. We also observed the fruits ofmacroeconomic policies in place since the mid-1960s that had stimu-lated labor using economic and urban growth. Conditions in the urbanilabor market benefited as well from positive developments in the agri-cultural sector. Moreover, specific government policies adopted in theearly 1970s fostered the generally high levels of urban infrastructure ser-vices noted in our study. The earlv 1980s witnessed a reversal of some ofthe economic gains posted in the 19 7 0s, but the record of the late 1980sencourages renewed optimism for the future. 2

Given the changinig natule and reqtiiremiients of matulilng urbaniza-tion (the stage when the urban population in a country exceeds 70 per-cent), the future growth of Bogota (and Cali) will undoubtedly exhibitdifferent characteristics than those clocumented in our study. But manyother countries are now undergoinig the phase of rapid urbanization.This volume will provide useful approaches to analyzing city growth inthose countries. The improved understanding of how cities work thatthis study attempts to furnish should influence the adoption of morepositive policies for managing large cities as they grow further. Develop-ing countries will continue to see the emergenice of large cities for along time to come.

Overview

The striking result of this study is that many of the behavioral relationsdocumented for Bogota and Cali are similar to those observed for othercities in both developed and developing countries. This suggests that cit-ies exhibit strong regularities in their patterns of development, which in

INTRODUCTION I 5

turn suggests that the transferability of basic behavioral findings is great.This conclusion is bolstered by the fact that Bogota and Cali have manycharacteristics typical of other large cities in the developing world.Although many of the results pertain specifically to the two cities stud-ied, the research also shows that economic forces strongly condition thecourse of urban development and that markets play an important rolein the allocation of resources to urban areas. Public use of private incen-tives can help achieve harmonious development. Efficient city growthrequires management and administration that is oriented as much toeconomic issues as it is to physical "bricks and mortar" issues of infra-structure.

The City Study was divided into seven segments covering variousaspects of urban growth and the urban economy. These include the spa-tial patterns of urban development, the operation of urban land mar-kets, employment and income distribution, the location of employmentwithin the city, housing, transportation, and public finance and adminis-tration. This book is more or less organized around these seven seg-ments or topics. 3

Chapter 2 looks at the patterns of city growth across many countriesin both the developing and the developed worlds. It offers a rationalefor the emergence of large cities and traces patterns of spatial develop-ment common to growing cities. The continuing decentralization ofboth residence and employment is a feature of most cities as they grow.

Chapter 3 introduces the cities of Bogota and Cali by summarizingthei- structure in relation to the spatial distribution of population andland values. It then describes their changing spatial patterns by tracingthe density and land value gradients as they evolved over time. The maincharacteristic of growth associated with these cities, as with otherexpanding cities, is the decentralization of residence and employmentthat has occurred.

Together, chapters 4 and 5 cover the third segment of the study, onemployment and income distribution. Chapter 4 includes an anatomyof the distribution of income in Bogota and Cali, with specific attentionto the distribution and characteristics of poverty. The central issue con-sidered is the existence and measurement of spatial inequality-that is,the concentr.ation of the poor in certain parts of the city. The variouscorrelates of urban poverty are also described, and the measurablereduction in poverty in the late 1970s is documented. Chapter 5 turns tothe work force, which is profiled not only by the usual demographic cri-teria of age, education, and income level, but also by place of residencewithin the city. Estimates of the returns to education and experience arealso presented in this chapter. An interesting featlLre of these estimatesis how the expansion of access to education has affected the returns tohigher education. The segmentation of the labor market is then exam-

16 UNDERSIANDING THE DEVELOPING METROPOLIS

ined. In addition to the usual segmenting variables such as union mem-bership and the size of firms, the influence of workers' backgrounds, asproxied by their places of origin and residence, proves relevant. Theinteraction of labor markets with city strtucture in terms of the locationofjobs and workers' residences and the extent of workers' spatial mobil-ity emerges as the theme of this chapter.

Chapter 6 reports on the changing locational patterns of employmentin Bogota and Cali, with a focus on the manufacturing sector. The fac-tors influencing decisions about where to locate are identified and theprocess of movement described. The discussion emphasizes the role ofthe central city as an incubator for entrepreneurship. The findings arethen generalized to provide some guidelines for public policy.

Chapter 7 assesses the housing situation in Bogota and the responseof housing markets to the city's rapid growth. The institutional setting isdescribed, along with the role of government intervention. The conceptof unorganized urban housing markets is afflicted with a host of mythsand prejudices; the validity of these is examined in the context of thewidespread unregulated housing developments in Bogota. To achieve abetter understanding of the working of housing markets, the functionof tenure choice and mobility as adjusting mechanisms is investigated insome detail. Similarly, the main parameters of housing demand, such asincome and price elasticities, are carefully estimated. Overall, the studyfinds that Bogota's housing market functioned relatively well and thatthe supply of housing improved despite the city's rapid populationgrowth. The sttudy points clearly to the importance of rental options insatisfying the housinlg demand of the poor.

Chapter 8 takes a detailed look at the transportation sector in Bogota.The interdependence of the city's structure and its transportation pat-terns is described. The regulatory apparatus and the supply mecha-nisms of urban transport are documented, and the distributive impactof the main taxes and subsidies affecting this sector is evaluated. Bogotahas a wide variety of transport modes, which work relatively well. Theoverall effect of regulation, taxes, and suIbsidies is progressive. The mod-eling of mode choice and other aspects of travel demand are also intro-duced. The widespread availability of transportation in Bogota ataffordable flat rates, made possible in part by government subsidies, hasdone much to alleviate the spatial disadvantage of poor workers inBogota. The importance of providing a range of choices in urban trans-portation systems is one of the important conclusions of this study.

Chapter 9 offers a window onto the structure and performance ofBogota's urban government. The seemingly unplanned and decentral-ized approach to providing urban services has resulted in relative abun-dance. The financing pattern differs from that found in most cities.

INTRODUCTION 1 7

Borrowing to finance investments is common, and user charges servicethe loans.

Finally, chapter 10 attempts to give an integrated view of the findingsfrom the City Study. The main lesson seems to be that the most practicalway of coping with city growth is to endogenize instittutional responsesto the rapidly changing and unpredictable demands made by the city'sresidents. The needs of citizens of a growing city are multifaceted.Whereas overall patterns of city growth are predictable, many individualand group needs are not. lnstitutions and methods of responding tothese changing situations must be developed in such a way that littlecentral direction is necessary. The institutions themselves should beflexible enough to respond to the changing requirements of a city.

Notes

1. As the level of per capita income starts growing in a developing country, thelevel of urbarnizationi first rises slowly, then accelerates, and finally slows down.

2. Because this study has focused on the economic sphere, the repercussionsof the deteriorating social and political situation in Colombia have beenignored.

3. Almost all the findings in this study are empirically based. An appendixdocumenits the vast data sets that have been used in the studies that underlie thisbook.

Chapter 2

The Spatial Structure of Citiesin Developing Countries

Urbanization has been the most pervasive development phenomenonin countries whose per capita income has growrn from low to high levels.In most couLntries the emergence of large cities has been part of theurbanization process. In some of these countries, one city dominates theurbani system. In others, there are a number of large cities along with acontinuum of smaller settlements, with urban activities well distributedamong them. However, a notable feature of this century, and particu-larlv the latter half of this century, has been the emergence of themetropolis as a familiar place for habitation. The greatest growth oflarge cities, both in the size of individtual cities and in the number oflarge cities, has occurred in the developing countries. As late as 1960,the number of large cities (those with a population of over half a mil-lion) in developing countries was about the same as in developed coun-tries: about 100. By 1990 there were probably more than 300 such citiesin developing counitries-about double the number in developed coun-tries. In 1960 such large cities accounted for a third of total urbani popU-lation in developing counltries; now they account for about half, aproportion similar to that in developed countries. As the nutmbers rise,it becomes incr-easingly importanit to uncderstand the reasons behind theemergenice of these cities, the dynamics of their growtlh, and theirchaniging spatial structures. Are there similarities and patterns that canbe observed and understood?

The Emergence of Cities

Urbaniization takes place when the predominant economic activity in aregion shifts from primary production (agrictulture, mining, logging, andsuch) to secondary and tertiary productioni (processing and servicing

18

THE SPATIAL STRUCTURE OF CITIES IN DEVELOPING COUNTRILS 19

activities). Primary activities are characteristically land-intensive and aretherefore spatially dispersed. Secondary and tertiary activities are capital-and labor-intensive and are therefore spatially concentrated. Becausethe elasticity of substitution between land and nonland production fac-tors in secondary and tertiary activities is greater than that in primaryactivities, a greater use of capital and labor per unit of land is possiblc insecondary and tertiary activities.

As incomes increase with economic growth, the demand for servicesand nonfood goods rises, whereas the relative demand for food falls. Atthe same time, increases in agricultural productivity make it possible toproduce more food with a smaller work force. Labor then shifts fromrural areas to those where production of nonfood goods and services isconcentrated. Those areas thtus become ut-ban.

What acttLally constitutes an urban area? The traditional characteris-tics of an urban settlement are a population above a given size, a highdensity of population, and a predominance of nonagricultural activities.Why do secondary and tertiary activities concentrate in such locations.Because such activities typically exhibit economies of scale, their exist-ence r equires a spatially concentrated work force. This work force, alongwith its dependents, requires an increasing number of activities andpeople to service its needs, as do the industrial activities themselves.Population begins to snowball as more people and businesses move intothe area to meet the needs of the existing enterprises. The growingdemand for complementary services and products produces agglomnera-tion effects-economies achieved because of the presence of a multi-tude of activities-and lead to expansion of the economy as a whole.The combination of scale economies, agglomeration effects, transporta-tion needs, and comparatively higher nonland/land substitutioni elastic-ities in industry and services produces a concentration of people andeconomic activity. Thus cities emerge.

Different industries exhibit different levels of economies of scale. Theexistence of some industries with greater degrees of scale economiesleads to the emergence of large cities. The further specialization of suchcities into high-level service, governmental, education, and financialsectors adds to their size. These growing cities offer increasingly variedemployment possibilities for workers and their households. The concen-tration of more skilled jobs also attracts more highly skilled workersfrom the smaller cities and towns. This in turn generates more jobs andbetter communication, transportation, water, and sewerage systems, allof which make it possible for cities to grow. As an economy switches intoproduction of more "moderni" goods and services, its cities continue toexpand (Henderson 1988).

Eighteenth- and nineteenth-century urbanization in European coUnI-tries was dominated by the imperatives of the Indtistrial Revoltition.

20 UNNDERSTANDINCG THE DEVELOPING METROPOLIS

Most people who came to the cities did so to work in the new mills andfactories. Current urbanization in developing countries is different inthat the tertiary sector requires more labor than does the manufactur-ing sector. The existence and expansion of cities is therefore equallydetermined by the economies of scale inherent in certain industries andby the localization economies achieved by the agglomeration of differ-ent kinds of service activities in one place.

Understanding the reasons for the emergence of large cities is essen-tial to understaniding their internal urban structure. The term "urbanstructure" refers to the kind, location, and density of activities as theyare distributed across space in urban areas. The goal of this study is tounderstand both how this structure changes as a city grows and whatdetermines these changes.

The main constituents of a city are the people who live there, thefirms that do business there, and the government. It is the residentswho generate the economic activity that supports the city. Their keydemands are for housing-which typically accounts for half of land usein cities-and for transportation to get to work, school, shopping, andleisure. The firms in a city employ people in Inanufacturing, trade,retailing, and other services. They decide where to locate based on thepattern of demand for their products, the technology they use in pro-duction, the location of their suppliers, and their distribution capabili-ties. The government's role is to supply infrastructure and publicservices and to regulate economic activity. A key area for governmentinterventioni, whether through direct supply of services or through reg-ulationi, is that of transportation.

Because people, firms, and governiments in different countries con-centrate in cities for similar reasons, and because there are commonand compreherisible patterns of human behavior, cities throughout theworld have quite similar structures. Given such broad similarity, thestructure of any city can be studied along similar lines. Differences canbe tinderstood by examining the functions of the city, the nature of eco-nomic activity found there, the characteristics of the people who livethere, and the activities of the government.

This chapter lays out the broad characteristics of cities, giving com-parative information for rich and poor countriies wherever possible. Thenext chapter describes the changes that have taken place in Bogota andCali as their structures have changed with growth and development overthe past quarter-centurv. These two cities' structural similarity to othercities in both developed and developing countries gives credence to thewider applicability of our detailed behavioral studies. The goal, then, isto look for useful ways of describing cities in a summary fashion and toidentih; the phenomena that are common to all or most of them.Decentralization, which is fundamental and pervasive with city growth,is one such phenomenon.

THE SPATIAL STRUCTURE OF CITIES IN D)EVELOPING COUNTRIES 21

Growth and Decentralization

The concurrent growth and decentralization of cities is nothing new.Bowden (1975, p. 78) reports thatJ. Stow described these processes atwork in London in "A Survey of London, Written in the Year 1598":

In 1598 London had a central business district (in the modernsense) but it had not had one for long. The London Stow describes isstill predominantly medieval in that merchants, artisans, the rich andpoor, storage and manufacturinig were found scattered in all wardsand parishes. Great parts of the city were residential. Most people(although a declining majoritv) lived where they worked..

Yet specialized warehouses were remarked in the parishes ... an(the old district of the Italian galleymen; tenements and squatter set-tlements of aliens, primarily Flemish refugees, were concentrating inthe suburbs, as well as within the city wall, many formerly clusteredactivities that we think of as dispersed were seen by Stow as dispersingfor the first time, e.g., grocers, vintners and cooks, and mantifacultr-ing-particularly that associated with the metal and leather trades-was gravitating to the north. From a population of 50,000 in the1540s, the city's population had probably trebled by the turn of thecentUry... the crucial change is the centripelal movement of the pro-genitors of the retailers of luxury shopping goods. "Men of trades andsellers of wares in this city have often times changed their places asthey have founld their best advantages." ... Most of the luxury tradersand their great merchant counlterparts had moved toward the centerof the city (often without moving their r esidences with them).

This is a good description of the processes of concentration and decon-centration of cities, and one that is quite applicable in developing coun-tries today. Typically, activities first converge in the city center and thendeconcentrate later, a process that had clearly begun in L.ondon by thesixteenthl century. Moreover, the separation of residence and workplace,while not common then, had begun. The central business district (CBD)had begun to gain specialized functions. Adna Weber, in his classic workon the structure of nineteenth-cenittury cities, said that "the most encour-aging feature of the whole situation is the tendency.., toward the devel-opment of suburban towns. The significanice of this tendency is that itdenotes, not a cessation in the movement toward concentration. but adiminutioll in the inlensitv of concenitrationi" (Weber 1963, p. 458).

Thliere has been a tendency to thlin-k of decentralizationi mainly interms of the widespread growth of American suburbs in the 1950s and1960s, which was remarkable because of the related decline in the city-center activities. I'hat was a special case of decentralizationi, perhaps typ-ical of the richest countries, but decentralizationi in general shotild be

22 UNDERSTANDINC, THE DEVELOPING METROPOLIS

regarded as a wider city growth phenomenon. Mills and Song (1979, pp.82-83) describe this process succinctly:

The most important causes of urban decentralization are thegrowth of metropolitan areas, rising real income, and improvedurban transportation. In small urban areas, the purchases of an entireurban area's population are needed to support its commercial andindustrial activities, and they are therefore located centrally. As thepopulation and real income of an urban area grow, it becomes possi-ble to support shopping and employment centers with the customersand labor force of'only part of the urban area.' Thus sub-centers ofstores and workplaces appear away from the central business district.Thus fewer people are tied to the city center for jobs and shopping,and they are attracted to suburban residences because of lower landvalties and correspondinlgly lower population densities. Real incomegrowth has an additional important effect. As income rises, a family'shousing demand rises and it is induced to move further from the cen-ter to take advantage of low land values. Improved transportation,whether- by public transit or automobile, has the same effect. Itincreases accessibility to the central business district from distantparts of the urban area, thus permitting people to take advantage ofcheap suburban land for housing.

Measurement Issues

How is decentralization to be measured? One obvious measure is theproportion of people living or working in the CBD or in the central city.As long as the boundaries of these entities are kept constant, this mea-sure should give a good sense of concentration, if' observed at differenttimes during the growth of a city. Such a measure can be supplementedby the proportion of people livirig or working in successive rings aroundthe city center. The problem with such measures, however, is thatboundaries of the CBD, central city, and iings are arbitrary (althoughoften taken to be legal boundaries), and observation of concentrationor deconcentration will depend on their inclusiveness. These bound-aries are often enlarged as a city grows. Moreover, the concept of the(CBD, although clear in principle, is not clear for measurement purposes.Where are its boundaries? As a city grows, the CBD typically expands. Forthese reasons, it is better to find methods of describing the density pat-tern of the whole city that are not so sensitive to boundary-definitionissues.

Given the observation of and theoretical justification for density pat-terns that decline from the city center, it is natural to describe them asbeing given by a functioni that shows density declining exponentially

TFIE SPATIAL STRUCTURE OF CITIES IN DEVELOPING COUNTRIES 23

from the city center: Dx = Do e-X, where D', is density at distance x fromthe CBD, Do is the estimated density at the center, and g is the estimatedrate of decline of density. Thus g may be interpreted as the percentagedecline in density per unit distance. The larger gis, the greater the rateof decline from the center and the more concentrated the city. Thedecline in g over time then measures the rate of decentralization ofthe city. And because g can be measured if there is information on thepopulation density of different parts of a city and because it is expressedin units of density decline per kilometer or mile, its value can be com-pared across cities. Vlalues are given for a whole range of cities in thenext section.

A few issues concerning measurement need to be mentioned. First isthe selection of the exponential functional form. Several articles (forexample, Griffith 1981; Andersoni 1982; Wheaton 1982a, 1982b) havequestioned the validity of the simple exponential form and suggestedmore complex forms for estimation. One criticism is on1 theoretical aswell as econometric grounds. The simple exponential form can bederived firom economic theory assuming static equilibrium in the urbanarea (Mills 1967). It is argued that, in fact, urban areas grow from a his-torical process of successive accretion, and, because of the durability ofstructures, the urban area is not in equilibrium at any given time. Thedensity functioni should therefore be derived firom a more realisticmodel of city growth. Another criticism concerns CBDs with little resi-dential population. There is a doughnut-like hole in the city center(Anderson 1982). Moving outward from the city center, the populationidensity first rises and theii begins to decline toward the suburbs. Thiis isobserved in some cities in developing countries as well as in most devel-oped countries (see, for example, Asabere and Owusu-Banahere 1983for African cities).

There is considerable merit in these arguments, and the functionalform can clearly be refined. The elegant simplicity of the negative expo-nential is appealing, however, and although inexact, it is not inconsis-tent with some of the other formulations suggested. It also has theadvantage of being the most common functional form estimated and istherefore available for comparison purposes. 2 The general problem ofthe empty city center is often solved bv excluding city-center observa-tions from the estimation sample. The recent refinements are also use-ful in interpreting the results of the simple estimation.

Chief Determinants of Decentralization

It is useful to obtain some idea of what the more powerful determinantsof decentralization are and which of these are expected to be especiallyrelevant in developing countries. The most important determinant is

24 UNDERSTANDING THE DEVELOPING METROPOLIS

increase in population. Historically, the decrease in transportation costshas also been very important; this was especially true during the majortransition from animal-drawn vehicles to motorized vehicles and laterfrom rail transportation to private automobiles (the latter particularly indeveloped countr-ies, wher e the advent of the limited-access highway fol-lowed the rise of the automobile). In developing countries many modesof transport coexist, an(d different classes of people have access to differ-ent modes at different prices (see clhapter 8). The third key determi-narnt has been rising per capita income. Again, this has been moreimportant for developed countries in the last half-centurv. The rapidrate of city growth in developing couintries has includedl high rates ofimmigration; this has meant that, paradoxically, the rate of per capitaincome growth has been low in the largest cites even when it has beenhigh for the country as a whole. In-migration has essentially been anequilibrating mechanism, reducing the disparities between large city/other and between overall urban/rural income levels. Rates of incomegrowth have therefore been high for some groups of people and corre-sponding parts of cities, and low for other groups and parts of cities.Thus, growth in population is a good explanation for residential decen-tralization in some parts of a city, whereas incoine growth may explaindecentralization in another part.

Rich people often live largely in one area of a city, and as the cityexpands and income grows, they contintte to fill spaces in the samedirectioni as city growth; the poor, meanwhile, live in another area of thecitv aind expand in another direction. The density patterns are quite dif-ferent for these two segments of the cit. The standard residential pat-tern in cities in developed countries is one of the rich living in outerrings and the poor in inner rings. Theoretically this is because thedemiianid for housinig space is highly inlcome elastic for the rich, andlower land prices at the periphery more than offset the increased trans-portation costs that result fiom living greater distances fromii the cityceenter. In cities in cleveloping countries, however, it is now quite com-mon to observe manv low-income suburban settlements, as well as a fewhigh-income residential suburbs, on the city periphery. This pattern isdocumented in detail for Bogota and Cali in chapters 3 and 4. Theselow-income settlements are often of high densities. An examination ofdensity gradients, g, in such cases yields the expected flattening out-but the flattening out is at a high density level everywhere. In cities indeveloped countries, the flatteninlg out is at the low (lensity levels thatresult from decreases close to the center and increases at the peripherv.

It is Fascinating to speculate on the reasons for these different pat-terns. One is explicit government policy. As Asabere and Owusu-Banahere (1983) have documented for African cities, in the ptursuit ofan ideal Garden City concept, it has been quite common for govern-

THE SPATIAI STRUCTURE OF CITIES IN DEVELOPING COUNTRIES 25

ments to clear congested central cities, usually populated by the poor,and move the inhabitants to distant locations, often designed as sites-and-services projects. 3 The original idea that concentration was bademanated from the dismal health conditions in European industrial cit-ies in the eighteenth and nineteenth centuries. This idea has beentransferred somewhat uncritically to developinig countries. Even whenpeople have not been relocated, it is common to design high-density res-idential complexes for poor people at the periphery, because that iswhere land is available and inexpensive.

The second factor in high-density peripheral development has to dowith the nature of housing demand and the construction technologyavailable to the poor. As chapter 7 and other studies document, thepoor spend considerable portions of their income on shelter; in fact,even for middle-income residents, housing has low income elasticity ofdemand. Among the poor, price elasticity of demand is also low. Givenlow incomiie levels, the building technology used is pretty rudinmentary,and construction is a mixture of self-help and contracting. There is evi-dence from Bogota and Cali that homes in the poorer areas have fewerfloors than homes in the richer areas (see chapter 3). Low-income hous-ing is seldom more than three stories, because going higher wouldrequire steel, more cement, and formal construction methods. Theextent of substitution of capital for land in low-income housing is there-fore limited. The high proportions of income spent on shelter and thelow income elasticity of demand also suggest some maximum thresholdfor crowding. The result of these two effects is only a small density gradi-ent in poor parts of the city: the areas near the center are only slightlymore densely inhabited than those on the periphery. It is typically largerhouseholds with greater space needs that are pushed toward the out-skirts to take advantage of lower land prices. In the poorer areas there islittle difference in per capita space betweeni the center and the periph-ery. Transportation costs affect this pattern, too. If there is a flat fare sys-tem, the poor can afford to live on the periphery, because they spendonly more time traveling-not more money.

Which Came First-Decentralization of Population or Businesses?

In cities in developing countries and industrial countries alike, employ-ment decentralization appears to lag behind residential decentraliza-tion. There has been relativelv earlv decentralization of large andpolluting manufacturing activities in cities in developing countries. Asin residential development, the Garden City philosophy has played arole in locating large-and often even small-industry in peripheralindustrial parks. Again, the desire for a clean, arboreal central city hasbeen buttressed by the availability of cheaper land at the periphery. The

26 UNDERSTANDING THE DEVELOPING METROPO.IS

deindustrialized CBD in these cities then tends to specialize in retail andwholesale trade, banking and finance, and other service functions.Given the paucitv of overall demand for services because of low incomes,these functions often become more concentrated during the initialstages of urban growth. If decentralization occurs, it usually happensafter the richer residents have left the central city and established com-muniities that can sustaini subcenters of employment. There is littledecentralization of service functions to the poorer areas, except to meetessential daily retail needs. Overall patterns of employment decentrali-zation can therefore be quite different between manufacturing and ter-tiary sectors and between rich and poor parts of the city. Given theseconsiderations, we can expect a continuiing trend toward employmentdecentr-alization in cities in developing countries when those citiesexhibit continuing population and income growth and residentialdecentralizationl.

The preceding discussion illustrates how changes in the environmentand the behavior of firms, households, and individuals are connected tochanges in urbani structure as measured by density gradients. The nextsection considers the changes in density gradients that have beenobser-ved in cities in both developed and developing countries.

City Characteristics: Evidence from Different Countries

What are some of the broad strtictural characteristics of cities, and howhave they changed over time? This section draws on research reportedin Ingram and Carroll (1981); most evidence is taken from Latin Ameri-can countries so that Bogota and Cali can be seen in their appropriatecontext.

In 1970 twenty-four Latin American cities had central-city populationsof more than 600,000 or metropolitan populations of more than 1 mil-lion. North America had thirt.-six metropolitan areas with 1 million ormore inhabitants. Of the large Latin American cities, comparable datawere available for only thirteen. Twelve large U.S. metropolitan areasfrom various parts of the country were selected on the basis of compara-tive size and rale of populationi growth dur-ing 1950-70.

Overall, urban population growth was much higher in the 1950s thanin the 1960s in both Latin America and the United States (see table 2-1).In the Latin American cities, the average annual growth rate declinedfrom 5.5 percent in the 1950s to 4.6 percent in the 1960s; in the U.S. cit-ies it declined much more, from about 4.1 to 2.5 percent. Although it isdifficult to assemble comparable information for the period since 1970,it is apparent that these trends have continued, and urban populationgrowth has declined further in both regions. In the United States most

THE SPATIAI STRUCTURE OF CITIES IN DEVELOPING COL'NTRIES 27

older large cities lost population (luring the 1970s while the newer cit-ies, mostly in the Southwest, continued to grow. Table 2-1 shows thatexcept for Mexico City, the larger cities grew somewhat more slowlythan the smaller cities, with this tendency being much more pro-nounced in the Ulnited States than in Latin America.

Preston (1979) confirms this general pattern worldwide. This sug-gests that, for the most populous cities, the advantages of lirbani agglom-eration economies and economies of scale tend to get exhausted, andsome urban diseconomies might set in. This should not be overstatecl,however, since the largest cities in developing countries continue togrow at significant rates. The main reasons for the overall decline inurbani population growth in North America and in Western Elurope arethat these regionis are approaching 100 percent urbaniization levels andtheir total population growth has become very small-near zero percentin some countries. Within the developing world, Latin Amel-ica has thehighest urbanization levels, with many countrics more than 60 percenturbaniized.

A major difference between North Americani and Latin American cit-ies is that large cities in Latin America (and in developing couLntries ingeneral) have much lower real income levels. Gross national product(GNP) per capita in 1975 was $7,1004 (in 1975 dollars) for North Arner-ica and $1,000 for Latin America (World Bank 1977). Based on growthrates reported by Kuznets (1966), North America would have had a percapita GNP of $1,000 (1975 dollars) in the 1850s. There were no cities ofmore than I million people theni. Hence, although some similarchanges are observed over time in developed and developing countries,their levels of magnitude are often quite differenit.

It is interesting to observe chaniges in the intranietropolitan popula-tion distributioni as large cities grow. Table 2-2 summarizes data for atwenty-year period for the central and peripheral areas of the ten LatinAmerican cities with available data and the twelve U.S. cities from table2-1. It is striking that across the cities on both continenits, both the cen-tral and peripheral densities vary by an order of magnitude: this is anindication of the definitional problems that are encotuntered when mak-ing such comparisons. W'hat is the central city? Where does it end andthe peripher-y begin? In most metropolitani areas worldwide, the areaconsidered to be the central city has a residential density of about 5,000to 20,000 people per square kilometer (see Mohan 1980). Central densi-ties are continuinig to increase in most Latin Amnerican cities, whereasthey are clearly stabilizing or decreasing in North America. Peripheraldensities are rising in nearly all cities in this sample, and rising fasterthan central-citv densities. Thus pervasive decentralization is takingplace in both North American and Latin Americani cities, with the trelndmore pronounlced in the former. The L.atin American central densities

Table 2-1. Population and Poputation Growth in Selected Latin American and U.S. Metropolitan Areas

Annual gowth Tate Annualg,rowUth Tate

Popfnlation (thausands) (Percent) Population (thousands) (percent)

Latin American citv 195() 196() 1970 1950-60 1960-7(0 UfS. ritY 1950 1960 1970 1950-60 1960-70

iMexicoCity 3,180 5,246 8,657 5.1 5.1 NewYork 9.556 10,695 11,572 1.1 0.8Sao Paulo 2,708 4,818 8,195 5.9 5.5 Los Angeles 4,152 6.039 7,032 3.8 1.5BuenosAires 4,723 6,739 8,189 2.8 2.0 Chicago 5,178 6,221 6,979 1.9 1.2Rio deJaneiroo 3,298 5,012 7,082 4.3 3.5 Philadelphia 3,671 4,343 4,818 1.7 1.0Limab - 1,846 3,302 - 5.4 Washington, D.C. 1,508 2,077 2,861 3.3 3.3Bogot;!' 715 1,697 2,855 6.9 5.9 Boston 2,414 2,595 2,754 0.7 0.6Santiagod 1,509 2,170 2,820 4.6 2.7 fHouston 936 1,418 1,985 4.2 3.4Caracas, 724 1,388 2,199 6.1 4.7 San Diego 557 1,033 1,358 6.4 2.8

X Recife 819 1,240 1,793 4.2 3.8 Miami 495 935 1,268 6.6 3.1Belo HorizotiLe 475 888 1,606 6.5 6.1 D)enver 612 929 1,228 4.3 2.8Guadalajara 440 851 1,455 6.8 5.5 SanJose 291 642 1,065 8.2 5.2Monterrey 376 708 1,213 6.5 5.5 Phoeniix 332 664 968 7.2 3.8Cali' 284 638 898 6.4 3.9

- Not available.a. Data are for 1947, 1960, and 1970.b. Data are for 1950, 1961. and 1972.c. Data are for 1951, 1964, and 1973.d. Data are for 1952, 1960, and 1970.e. Data are for 1950, 1961f, and 1971.Source: lngrTali and Car roll (1981).

Table 2-2. Population Density in Central and Peripheral Areas of Selected Latin American and U.S. Cities

Area (squeare Population per square kilomneter Area (square Population per square klomneterLatin Anmcan city Location, kilometers) 1950 1960 1970 US. cily Localions kWlometeri) 1950 1960 1970

Mexico City C 138 16,225 20,558 21,074 New York C 777 10,157 10,015 10,161P 2,192 432 1,101 2,675 P 4,758 350 612 773

Sao Paulo C 1 ,4 9 3 b 1,380 2,287 4,005 Los Angeles C 1,326 1,675 2,130 2,390P 6,458 79 172 343 P 9,213 210 35() 420

Buenios Aires C 200 14,952 14,872 14,897 Chicago C 578 6,275 6,140 5,825P 3,860 473 1,025 1,418 p 9,054 170 295 400

Rio dejaneiro C 1,171 2,030 2,824 3,631 Philadelphia C 334 6,202 5,995 5,835P 5,293 174 322 535 P 8,868 180 264 325

Bogota' C 304 2,352 5,582 9,391 Washington, D.C. (C 158 5,077 4,835 4,790P - - - _ P 5,936 119 221 355

Recife C 29 0 d 3,594 3,815 5,075 Boston C 119 6,735 5,860 5,387P 1,992 148 223 367 P 2,435 662 780 867

Belo Horizonte C 335 1,053 2,070 3,686 Houston C 1,028 58() 913 1.198P 3,335 37 58 111 P 15,250 22 31 49

Guadalajara C 188 2,204 3,940 6,383 San Diego C 552 606 1,038 1,2581P 1,164 52 95 220 P 10,484 21 44 63

Cali' C 85 3,341 7.506 10,565 Denver C 246 1,690 2,0(08 2,092P - - - - P 9,233 21 47 77

SanJose C 300 318 681 1,480P 3,067 64 143 202

Plhoenix C 642 166 684 906P 23,069 10 1 0 17

-Not availahle.a. C = center; P = peripher-N.b. 1,622 in 1959 and 1960.c. Data are for 1951, 1964, and 1973.d. 146 in 1950.Source: Inigram and Carroll (1981).

Table 2-3. Population Density Gradients in Selected Latin American and U.S. Cities

Latin Anmeiran fitV ph Parnae1e, 1950 1960 1970 '.S. cty ph Pareametert 1950 1960 1970

Mexico City 1.0 /O 69,00(0 62,000 44,000 New York 0.t D0 62,000 45,000 40,000g 0.37 0.27 (1.17 g 0.16 0.13 0.11

Sao Paulo 1.0 D, 8,400 12,000 18,000 Los Angeles 0.6 D, 4,800 5,300 5,800g 0.14 0.13 0.12 g 0.06 (.(6 0.05

Bueinos Aires 0.6 D, 54,000 37,000 33.000 Chicago 0.5 D, 27,000 20,000 16,000g 0.21 0.14 0.02 g 0.13 0.10 0.08

RiodeJaneiro 0.5 Do 8,700 10,000 11,000 Philadelphia 1.0 Do 20,000 16.000 14,000g 0.(9 0.08 0.07 g ().18 0.15 0.13

Bogota 0.5 D. - 37,000 26,000 Washington, D.C. 1.0 DO 15,000 11,00(( 9,000g - 0.25 0.12 g 0.25 0.18 0.14

Recife 0.6 DO 13,000 14,000 17,000 Boston 0.7 Do 14,000 11,000 9,300g 0.25 0.21 0.19 g 0.16 0.13 0.12

Belo Horizonte 1.0 Do 5.000 11,000 19,000 Hotistoni 1.0 D(, 2,100 3,500 4,200g 0.26 0.28 0.27 g 0.12 0.13 0.12

Guadalajara 1.0 D,, 14,000 28,000 39,000 San Diego 0.5 D, 2,100 3,200 3,600g 0.45 0.46 0.41 g 0.11 0.10 0.09

Moniterrev 1.0 Do 6.200 8.500 7,40(0 Miarni 0.5 Do 8,000 6,800 7,200g 0.32 0.27 (.19 g 0.23 0.15 (1.13

Cali (1.5 Do - 43,000 29,000 Denver 1.0 D)0 6,8(00 6,0()() 5,100g - 0.41 (0.21 g 0.27 0.20 0.16

San Jose 0.5 D0 620 1,300 3,500g 0.08 0.07 ().10

Phoefnix 1.0 Do 350 2.700 3,100g (0.(08 0.16 0.14

-Not available..Not: Density gradients were calCulated using the technique described in W1hite (1977).a. Parameters firom density = Del7X where xis the distanice in kilometers from the city centergis the density gradient per kilometer, and D. is the estimated

central citv density.Saurce: Ingram and Carroll (1981).

THFE SPATIAL SlTRLICTLIRE OF CITIES IN DEVELOPING COUNTRIES 3 I

are akin to older North American cities, and the newer North Americancities are structurally dispersed, with low densities all over.

Data in table 2-2 are based on arbitrary definitions of central andperipheral areas. Density functions are more robust than actual central-city densities as proxies for detecting city structures and for summariz-ing changes over time. As Mills and Tan (1980) report, urban popula-tion density ftictions have been estimated for almost every developedcountrv in the world, and an increasing number of estimates are avail-able for developing coulitries as well. As menitioned earlier, the expo-nential density function provides a good summary indicator- of citystructiure, with g measuring the percentage of decline in density perunit distance ancl D,, measuring the theoretical density at the center.

The concept behind this functional form is that of a circular mono-centric citv. There has been considerable discussioll of Lhe use of otherfunctionial forms to accounit for two problems: cities typically have morethan one large employment center, and the city center typically has verylittle residential populationi since most land is used for economic activ-ity.5 These are useftil elaborations of the simple theory and provide formore realistic estimation, but they lose the elegant simplicitv of theexponential function. Moreover, they cannot be used for comparisonpurposes because most estimates that exist are for the simple exponen-tial fo m.

Thus, g is a good summiiar y measur-e of decentralization. For a fixedradius x of a circular city, the smaller the g, the smaller the proportion ofpeople living within x distance of the city center and the more decen-tralized the city. Hence, when g declines from one decade to the next,the city can be said to have decentralized.

Table 2-3 displays estimates of density function parameters for thesame sample of Latin American and U.S. cities. The density gradient gdeclined over time in most cites, incr-easing only in Guadalajara, BeloHorizonte, Houston, San Jose, and Phoenix-and in each case in onlyone of the two decades. The decline in the populationi density gradientsin Bogota and Cali was particularly significant: both fell by almost haltbetween 1960 and 1970, implying a notable decentralization of popula-tion. Latin American cities generally have steeper gradients than U.S.cities, as do smaller cities compared with larger ones. Intercept (centralcity) densities (I),) are significantly higher in Latin American than U.S.cities; the averages are 24,000 and 10,000, respectivelv. Latin Americancity centers are also more highly congested in terms of residentialpopulation.

WNThen we combine the comparison of intercept densities with thatmade for density gradients, two major points emerge. First, there is asurprising degree of similarity betwveen the density funiction parametersof large Latin American and U.S. cities. This similarity is enhaniced

32 tTND)ERSTAND)IN(I THE D)EVELOPING MFTROPOLIS

when we compare the large Latin American cities to only the five olderNortheastern cities (New York, Chicago, Philadelphia, Washington, andBoston). The 1970 average intercept and gradient, 17,700 and 0.116, forthese five cities are similar to the 1970 averages, 24,600 and 0.12, respec-tivel1, for the large Latin American cities. It appears, too, that densityfunction parameters follow different patterns in Latin America and theUnited States: As size varies, U.S. cities have fairly constant density gradi-ents and varving intercept densities, while Latin American cities havefairly constant intercept densities and varying density gradients. Cou-pling this pattern with the similarity of parameters for large cities suggeststhat the relatively smaller Latin American cities have larger interceptdensities and steeper gradients than similar U.S. cities and therefore aremuch more centralized.

These patterns are quite common the world over and in recenthistory. The regular decline in gradient over the past century (see table24) is remarkable, and the pattern clearly transcends countries and cul-tures. The pictur-e for a larger sample (see table 2-5) is a little more var-ied, but the broad patterns continuLe to hold true. The largest cities-those with population between 5 million and 10 million-appear tohave a uniformly low gradient of arounld 0.1 or less, whereas smaller cit-ies generally have much higher gradients. More densely populatedcoUltr ies, such as India,japan, and the Republic of Korea, tend to havehigher overall densities and higher gradients in their smaller cities thanLatin American coUlltinies do.

Altlhough density functionL comparisons have been based on city sizeto some extent, the data suggest that city age, transportation technology,and per capita incomie are also important determinants of the decen-tralization observed in a city (Harrisoni and Kain 1979; White 1977). Forexample, density gradients vary for North American cities of differentvintage, and they are similar for Latin American cities and older NorthAmerican cities. Older cities developed during a period when publictransit was the dominant mode, as it still is in most developing coun-tries. By the time the newer U.S. cities, like SanJose, Phoenix, Houston,and Los AnEgeles, emerged and developed, automobiles had become thedominant form of transportation. The older cities in industrial coun-tries and most cities in developing countries are therefore more central-ized and have higher densities than the newer North American cities.This also suggests that portions of developing-country cities that arenewly developed or inhabited by the rich could be similar to the newerNorth American patterns of citv development.

The level of structural similarity is quite surprising. Living conditions,per capita incomes, patterns of housing, and levels of public serviceinfrastructure vary considerably among North America, Latin America,Europe, Africa, India, and East Asia, as do the histories and culttural

THE SPATIAl. STRLICT LRE OF CFITFS IN DEVEl. PING COUNTIRIE:S 3S

Table 24. Population Density Gradients Per Kilometer in SelectedCities Worldwide by Decade, 1880-1960

Average fon- f(n/r Kingston)Year iU.S. ritiei" Chicago London Paris BRmbay (Jamalical)

1880 1.22 0.49 0.38 0.60 0.37 -1890 1.06 0.32 - - 0.33 1.0219(0 0.96 0.22 0.23 0.50 0.26 -191( 0.80 0.23 - - 0.24 0.9(1920 (.69 0.16 0.17 - 0.20 -1930 0.63 0.13 0.17 0.50 0.17 -1940 0.59 0.13 0.14 - 0.17 0.541950 0.50 0.11 0.12 0.21 (.13 -1960 0.31 - 0.09 - 0.1( 0.33

- Not available.a. Baltimore. Milwaukee, Philadelphia, and Rochester.Source: Mills anid Tall (I 980).

influiences of each of these regions. Yet decentralization of cities hasbeeni a pervasive phenomenon accompanying growth. Moreover, judg-ing from Nortil Amnerican and Western European cities, decentraliza-tion continutes evCeI after population growth stops. This is what attractedthe most attentioni in the 1960s and 1970s, because its effects-the emp-tying out of central cities and the increasing residential separation ofthe rich in suburbs and the poor in the center-are stark. The phenom-enon of decetntialization observed in developinig countr-ies is thencloser- to the earlier pattern for cities in developed counitries, whe ngrowth was accompanied by decentralization. The distinction esselitiallyis that in these periods D0 continies to increase or r emains constanit as gdeclines, whereas in later stages 1) declines as well.

The documentationi for employment dlensity gracdients is less abunl-dant because of the difficulty of obtaining spatial data on employmenit,which is not typicallv reported in population censuses. However, the pat-tern of continuiing decentralization is well documented for some couIn-tries, suchl as Brazil,Japan, the Reptiblic of Korea, and the United States(for example, see, respectively, Y. J. Lee 1985; Mfills and Ohta 1976; Millsand Song 1979; and Mills 1972). Employmenit density gradients are typi-cally muich steeper than those of populationi density, which suggests thcfamriliar overall pattern of inward coMInTLting for work. Employmentdecentralization has ustually lagged behind residential decentralization.Mantifacturing is usually the first sector to begin decenitralizing for thetechnological reasonis discussecl earlier, essentially con nected with theease of substitutinig capital for land. The portion of retail trade thatserves the dlaily needs of consuImlers typically decentralizes with popula-tion, but the other poi-tiotis remain centralized, along with much ofwholesale trade. Office activities oftetn remaini the most centialized; the

341 UNDERSTANDING THE DEVELOPING METROPOLIS

possibilities of capital/land substitution are extremely high for theseactivities. Because commercial activities are the most directly related toconstumer buying pOW7er, their rate of decentralization is closely relatedto prevailinig income levels (see table 2-6). It is therefore reasonable toexpect less centralizationi of these activities in rich cities and in richersections of poor cities. Clhapter 6 documenits these patterns as theyapplv to Bogoti and Cali.

The Consequences of Growth and Decentralization

Because we can expect continued growth of large cities along withdecentralization, it is necessary to examine some of the consequences ofthese patterns. First, we need to understandcl that although the processof ur-bani growth and decentralizationi in developing cotuntries is similar

Table 2-5. Population Density Gradients Per Kilometer in SelectedCities Worldwide, 1950, 1960, 1965, and 1970

CiIN 195() 1960 1965 197(0

Large ciliesNew York 0.16 0.13 - 0.11l'okvo 0.06 0.07 0.08 (.08L.oldo.l - 0.08 - 0.07Madras 0.24 0.24 - -Great Britain' - (.10 - (.06West. Germani'v - 0.25 - 0.21japan' - - - 0.48Seou(l - 0.35 0.33 0.22

AVedium-size citiesBirmingham - 0.12 - 0.04Manchester - (.08 - 0.06Accra - 0.52 - 0.46Poona (India) - 1.05 - 1.07I'tisani (Repuiblic of Korea) - - 0.26 0.13Kiinasi ((;hana) - - 0.61 0.55Taegu (Republic of Korea) - - 0.78 0.74Bangalore (India) (.61 0.53 - -

(Great Britainh - 0.34 - 0.25W'est GermanvO - 0.31 - 0.31japar" - - - 0.49

- N<it available ..1. Data are averages for cities with populations of rmiore than I million.bc Data are averages for cities with poptilations of 0.5 million to 1 million.Source: Ingiraimi and Carroll (1981) for New York; Mills and Ohta (1976) for Tokyo; Mills

and Tan (1980) for Madras, Poona, and Bangalore; Njills and Song (1979) for Seotl,

Pusan. and laegu; Asahbre anti Owust-Banahere (1983) fior Kmnasi:(Glickman (1979) forall others.

Table 2-6. Central City Shares of Residential Population and Employment, Selected Latin American Cities(percent)

Residentialpopailation

Ewonomically Professionalsl Empklrtnwntactive technical OJfice workers/ Senvice Blue-collar

Year population workers salespeople Zaei ns workers CGar oa'ners vlanuflaring Commerce Service

Mexicos City 195(0 73 86 81 70' _ _ _ _ _1960 59 72 66 49d - - - _

1970 39 52 47 43 28 - - -Sao Paulo 1950 85 - - - - - 84 95

1960 - - - - - - 8( 92 901970 75 - - - - 82 71 - -

BueniosAires 1950 65 - - - - - 67 - 83 h1960 46 62 55 57 - - 49c -66b,

1970 37 56 48 44 - 54 - - -Rio delaneiro 1950 76 - - - - - 77 88

1960 - - - - - - 75 86 821970 56 - - - - 79 76 - -

Recife 1950 68 - - - - - 59 86f1960 - - - - - - 63 85 861970 63 - - - - 82 63 - -

Belo Hori7onte 1950 77 - - - - - 56 911960 - - - - - - 47 92 881970 81 - - - - 90 53 - -

Guadalajara 1950 87 95 94 91t - 9 4 d - _1960 88 96 94 9() - - 92e - -1970 84 91 89 86 84 - 87 - -

Monterrey 1950 91 94 95 92' - - - -

196() 86 89 89 85' - - - -

1970 66 78 76 71 70 - - - -

-Not available.a. Service workers and blue-collar workers combined. b. (Commerce anid ser%ice combined. c. For 1964. d. For 1956. e. For 1965.Saurre: Ingram and Ca-rroll (1981).

:6 tUNDERSTANDIN(; THE DEVEIOPING; METROPOLIS

to that in developed countties, the existence of different initial condi-tions and behavioral parameters ofteni makes the restilts quite varied.The emergence of large, low-income, high-density settlements on theperipheries of cities raises the questions of where jobs and transporta-tion networks should be located and how the costs of public servicesshould be structured. For example, the establishment of a flat-fare trains-portation system would help residents overcome the disadvantage of liv-ing far from the center, but such a system would give the wrong signalsabout the associated resotirce costs of transportation. The provision andpricinig of other essential services, suclh as water, sewerage, and drain-age, presenit a similar problem. Residential decentralization stretchesoUt the supply networks of these services, which are then expensive toprovide. If the poor have little choice about where to live in the firstplace, however, they can hardly be charged high prices for essential ser-vices. Some of these issues are therefore difficult to analyze because oftheir politico-economic natutre.

Another common consequence of residential decentralization is theincreased spatial separationi of the rich in suburbs and the poor in thecentral city (see chapter 4). This separation has economic as well associal effects. Employment opportunities are likely to be more plentifulin the better-off areas, where the residents have a greater demand forgoods and services. The poor then either commute long distances towork in these areas, or they seek residential land nearby. This often givesrise to squatter settlements and slums in the middle of otherwise richneighborhoods. Because of the high land values and presstires to usethe land more "productively,' the government often relocates poor resi-dents, frequently by force.

The decentralization of employment and residences naturally causesa rapid conversion of agricultural land to urban uses. This is often per-ceived to be a major problem, and fears are expressed about the dwin-dling supply of agricultural land. This is, on the whole, a false concern.Even in the most (lensely populated countries, such asJapan and Korea,only a small fraction of total arable land is urbanized. The loss of agri-cultural land is, however, a legitimate concern for the farmers whoseproperty is tirbanized. even if they are fairly compensated. When theylose their livelihood, they often do not receive appropriate training andemployment in alternative occupations. The faster the growth anddecentralization of a city, the more serious this problem is, as the coin-petition for land at the uLrban fringe becomes intense. The challenge forpolicymnakers is to ensure that land is allocated to its most economiiicallyefficient use and that the legitimate interests of owners and users, bothold and new, are protected.

THE SPATIAL STRUCTURE OF CITIES IN DEVELOPING COUNTRIES 37

Notes

1. Compare this development with the "dispersing for the first time [ofl gro-cers, vintners and cooks" in sixteenth-century London.

2. For theoretical derivation and review of this fuinctionial form, see Mills

(1967, 1972) and Mohan (1979).3. See Berrv (1973) chapter 3 and jacobs (1961) chapter I for discussion of

the Garden C ity concept.4. All dollar amounts are U.S. dollars unless labeled otherwise.5. For example, Griffith (1981) discusses a generalizationi of the simple expo-

nenitial form to account for multicentered cities; Anderson (1982) proposes acubic spline function to account for an empty center.

Chapter 3

Growth and the Changing Structureof Bogota and Cali

Unlike many other Latin American countries-for example, Argentina,Ecuador, Peru. and Venezuela-Colombia has a well-articulated systemof cities in which Bogota accounts for only about 20 percent of the totalurban population. Colombia's primacy index (that is, the ratio of thepopulationl of the largest city to the total urbani population of the coun-try) is the lowest in Latin America except for that of Brazil (see Renaud1981). However, Bogota's share of both population and output hasincreased over the past few decades (see tables 3-1 and 3-2).

Within forty years Colombia was transformed from a predominantlyrural country to one that is predominantly urban; its urbanization levelrose from about 30 to 65 percent. During the same period Bogota grewfrom an intermediate-size city (bv today's standards) into a largemetropolis. It had a remarkably high and sustained rate of populationgrowth, of about 5 percent a year over almost a half a century (see table3-1). Tables 3-2 and 3-3 give the comparable growth in economic outputsince 1950. The growth rate of Bogota's regional product was 7 percentper year during 1950-75, while that for Colombia was 5.2 percent. Butbecause population grew faster in Bogota than in the country as awhole, the per capita growth rate for the city's gross product was only0.6 percent a year, versus 2.1 percenit a year nationwide (see table 3-3).Economic opportunities in Bogota led to substantial labor mobility, andthe per capita differential between Bogota and Colombia as a whole nar-rowed over the years until it was only about 1.5 by the mid-1970s. Itwould be even lower if price differentials were accounted for. Bogotagrew faster than other Colombian cities taken as a whole over thisperiod (and earlier-as long ago as the mid-I 800s) . and it emerged witha high concentration of people and economic activity bv the 1970s. Cali,

is

GROW[rH AND THE CHANGING STRUCTURE OF BOGOTA AND CALI 39

Table 3-1. Population and Intercensal Population Growth in Colombia

Population

7btal tlrban' Bogota Annual compound growth rale (percent)Yea, (thousands) (percent) (percent) Total Urban' Bogota

1938 8,702 31 3.8 n.a. n.a. n.a.1951 11,548 39 5.7 2.2 3.9 5.21964 17,485 52 9.5 3.2 5.4 7.31973 22.500 60 12.7 2.7 3.7 5.61983 26.965' 65b 151' 2.0 3.0 4.0(

nia. Not applicable.a. The ur ban population is the percentage living in a county seat (cabecera mvumpal).

Most cotntLy seats have 1,500 people or more, but some have fewer: moreover, not all set-tlements that large are county seats.

b. Data are estimates from Sokol and others (1984).c. Bogota's 1978 population was estimated from the World Bank-DANF Household Sur-

vey. The 4.0 pe rcent population growth rate observed from 1973 to 1978 was assumed tocontinue utitil 1983.

Source: 1938, 1951, 1964, and 1975 population censuses; Sokol and others (1984).

too, has grown consistently (see table 3-4). This record is consistent withthe idea that cities grow because they are more efficient economic pro-ducers, benefiting from agglomeration and scale economies.

The average population density of Bogota shows remarkable stabilityin the face of rapid population growth. Peak density occurred at the endof the 1950s, the period of most rapid growth. (The area data must beregarded as somewhat dubious because the juridical definition of thearea of a city is always somewhat arbitrary, as is more evident in the datafor Cali. The extension of a citv's boundaries eventually does, however,reflect the area that is regarded as belonigirig to the city at any giventime.) The pattern for Bogota suggests that the juridical area of the city

Table 3-2. Gross and Per Capita Domestic Product, Colombiaand Bogota

(olombia Bogetd Bogola;'s Bogoti s (;RP pet

(;RP as a captla as a(;lSP G' (;RP GRP percentage of percentage of

(milleions per capila (mtilon,% per capila Colombias Cooam/nas GoP

Kear Jof pesos) (pesos) ofpesos) (pesos) 6d/p per cap/la

1960 77.71 4 5,088 11,996 9,220 15.4 1.811965 97,968 5,455 17,208 9,591 17.6 1.761970 130,361 6,484 25,920 10,822 19.9 1.671975 176,478 7,352 37,671 11.779 21.3 1.60

,Vote. GDP = gross domestic product. (;Rp = gross regional product. Pesos are 1970 pesos:the exchange r ate was US$1 = Col$18.40.

So,rrc. Coloimbia (1977).

40 UNDERSTANDING THE DEV'ELOP[NG METROPOLIS

Table 3-3. Average Annual Growth in Domestic Product, Colombiaand Bogoti(percent)

Colombia Bogota

GDP GRPPeriod CDP per capita GRP per capita

1950-75 5.2 2.1 7.0 0.61960-75 5.6 2.5 7.9 1.71970-75 6.2 3.2 7.8 1.7

Note: GDP = gross domestic product. (;RP = gross regional product.Saurce: Table 3-2.

tends to follow growth in population, albeit with a lag. The constantaverage density implies that Bogota has always been decentralizing as ithas grown, in the sense that the central parts of the city have housed asuccessively smaller proportion of the total population. It would beinteresting to see whether such a pattern holds for other cities. If so, itwould go against the general notion that cities become increasingly con-gested as their populations grow. Cali's density pattern is more erraticbecause the juridical area has been changed in fits and starts, but there,too, density stabilized in the 1970s and is now comparable to Bogota's.Overall land development has clearly kept pace with population growthin these cities, and the city area appears to have become an endogenousvariable.

How does Bogota compare with other cities in the world in terms ofdensity? Bogota falls midway between the world's densest cities, such asBombay (140 people per hectare) and Calcutta (120 per hectare), andsuch North American cities as Chicago, Philadelphia, Detroit, and SanFrancisco, which have about 40 to 60 people per hectare. New York City(inclutding the boroughs of the Bronx, Manhattan, Queens, and Brook-lyvr) is about as dense as Bogota, with more than 100 people per hectare.There is a large degree of similarity in overall population densities inlarge cities in the world, and Bogota falls somewhere in the middle.

How has the physical character of Bogota changed as it has grown? Itsoriginal design was the traditional rectangular grid of streets and ave-nues, with the Plaza Principal as the focus of the city. Until the earlytwentieth century the entire built-up area of the city occupied what isnow only the central business district (CBD). The city was primarily acommercial, political, and religious center; Medellin was the industrialcapital of the country. (Medellin had more manufacturing employmentthan Bogota until about the late 1960s. It has since become betterknown for more dubious activities.) High- and middle-income residentsof Bogota began to move north in the second quarter of this century asthe southern and wester-n parts of the city began to fill up with lower-

GROWTH AND THE CHANGING; STRUCTURE OF BOGOTA AND CALl 41

Table 34. Area, Population, Population Growth, and Densityin BogotAi and Cali

Avetage annul,l DendsilPopulalan grozwtlh rate (population

(ilv and Year .4rea (hectares) /thowsands) (percent) per hectaret

Bogogali1560 20 - -

160() 56 - --

1670 129 3 - 231720 - 20, 3.9 -

1800 - 22' 0.1 -1850 294 30 0.6 1001886i 610 64 2.2 1051900 909 100 3.2 1101928b 1 .958 235 3.1 1201938h 2,514 330 3.5 1311951 - 660 5.5 -1958 8,084 1,130 8.0 14()19 6 4 b 14,615 1,730 7.4 118

19731, 30,423 2,877 5.8 951978' 30,886 3,500 4.0 113

Cali1800 50d 6' - 120

1880 l 14d 15' 1.1 1271900 _ 24r 2.5 _1918 - 45' 3.6 -1938 4001 88 3.4 2551951 1 ,2 9 0 d 2 841 9.0 1871958 1 ,8 5 0d 428' 6.0 2311964 ',1 OOl 6 3 8 h 6.3 701973 9',100' 930"' 4.2 1(31978c 9,10(0 1.10 3.4 121

- Not available.a. Data are from Lubell and McCallum (1978), table 2.1.b. Data ai-e fromil population census estimates.c. Data art from the 1978 World Bank-DANE Household Sur vey. All data for Bogota-

except for the vears marked a, b, and c-are from Wiesner (1980), table 1.d. Data are from the "Plan General de Desarrolo (le Cali suX Area Metropolitana. 1970-

1985-2000,1 an internal document of the Planeaci6n Municipal dte Cali.e. Tabares ( 1 979).f Area fbr C.ali is from the Planeaci6n Municipal de (Cali.

and lower-middle-income residents. In the mid-1950s Chapinero, origi-nally an outlyinig suburb north of the city, began to replace the city cen-ter as the primary commercial area. The city became elongated north tosouth, being bounded on the east by high mountains and on the west bvvaluable agricultural land and by low-lving land that discouraged devel-opment. Only in the 1960s and 1970s did the western parts of the citybegin to fill, giving the city its present almost semicircular shape.

4Y2 UNDERSTANDINC THE DEVELOPING METROPOLIS

Cali, which nestles into the base of the mountains to its west, has hada somewhat similar history. Until the early part of the century the built-up area was a compact, traditional rectangular grid of streets around thecentral plaza (Plaza Mayor), and the citv was mainly a commercial, reli-gious, and political center for the state of Valle del Cauca. Being small, itwas mixed in character until it began to expand in the mid-twentiethcentury, when the rich occupied the western part of the city and thepoor the eastern parts. Like Bogota, Cali developed linearly north tosouth, with greater development taking place toward the south. Theeastern part of the city consists of low-lying lands populated by the poor,who settled there in great numbers during the major land invasions ofthe 1950s and early 1960s. Since then, the city has filled out and hasacquired a semicircular shape-almost a mirror image of Bogota,although smaller.

As in other cities, it is difficult to separate the effects of institutionaldecisions on the spatial development of a city from those arising fromindividual preferences as r evealed by the market. Because much is madein this study of the particular spatial income pattern found in both cit-ies, it is interesting to note some of the main planniiing exercises thathave been carried out in Bogota. The two main influences on the shapeof the city were Karl Brunner in 1935 and Le Corbusier in 1951. Muchof Bogota's existing zoning and planning structure can be attributed toBrunner, who organized a city planning department. The northern partof the city was zoned for low-density residential development, while thesouth was zoned for high-density development. Le Corbusier's maincontribution was the creation of an industrial zone to the west of thecitv; intentionally or fortuitously, this separated the poor south-zonedfor high-density' development unider Brunner's plan-from the richnorth. These basic zoning and planning regulations have had a power-ful impact on the structuire of city growth. The low-density residentialdevelopment zoned for the northern part of the city has effectively seg-regated it for the rich. Similarly, the high-densitv residential zoning inthe southerni part of the city' made it possible for the poor to locatethere. The industrial zone in the west houses the main indutstrieslocated in Bogota.

This pattern of growth is typical of cities as they expand, except thatin Bogota land is probably more effectively segregated by use-betweenlow- and high-density residential development, and among residential,commercial, and industrial uses-than is typical for many cities in devel-oping countries. In Bogota and Cali the central cities were denselypacked to begin with and could become only a little denser as the citygrew. New growth, in both residenitial and employment activities, tookplacc largely outside the city center. Although land use planning hadsignificarnt effects on1 land tIse segregation, it is difficult to clearly iden-

GROW'TH AND THE CHANGING STRUCTURE OF BOGOTA AND CALl 43

tify the sequence of causation. Land use regulations may essentially havefollowed existing land use patterns and perpetuated their continuance.The rich suburbanized from the center in successive waves toward thenorth; the poor replaced thern, to some extent, in the center, as is typi-cal in North American cities, but they also decentralized toward thesou th.

As Bogota grew, its employmenit structur-e diversified, and now about25 percent of its labor force is engaged in manufacturing, another 25percent in commerce, and about 37 percent in services. This distribu-tion is typical of well-diversified cities-almost all large cities in thedeveloping world have more than 20 percent employment in manufac-turing activities (see Mohan 1986, chap. 3)-but some manufacturingnoncapital cities, such as Bombay, Calcutta, and Sao Paulo, have asmuch as 40 percent of their labor force in mantifacttrinig. The preseniceof the national government in a capital city clearly increases the propor-tion of people employed in tertiary activities. Cali has a higher propor-tion of manufacturinig employment and a lower share in the servicesector than does Bogota. It is important to look at the overall employ-ment structure in a growing city because different employment struc-tures have different implications for the spatial pattern of city growth,and, as shown later, the location patterns of various activities havechanged as Bogota has grown. Specifically, manitifactuirinig has led thetrend toward decentralization of employment locationi.

To summarize, both Bogota and Cali have grown at very high SUS-

tained rates for more than a century. Income has grown faster than pop-ulation over at least the past three or four decades, and real per capitaincome has therefore grown substantially. The result of growth in bothpopulation and income has been a considerable physical expansion ofthe city, and both population and economic activity have decentralized.The following sections document the magnitude of this decentralizationand note the departures from the general pattern that might be expectedfrom a growing, primarily monocenitric city.

Growth and the Spatial Distribution of Population and Income

Maps 3-1 and 3-2 show the boundary lines for rings and sectors withinBogotA and Cali. They also show the comunas-smaller divisions markedby two-digit numbers-that make up each ring or sector.' Map 3-3 illus-trates the spread and densities of population in BogotA. As might beexpected, the outskirts are relatively sparsely populated, whereas thecenter is denser; the southern part of the city is clearly more denselypopulated than the north. The center (CBI, romuna 31) is less densethan some surrounding areas. The densest areas are more than ten

44 UNI)ERSTANDING THE DEVELOPING METROPOLIS

Map 3-1. Bogoti: Ring and Sector Systems Based on 1973 Comunas

IBRD 25346

Ring System

X~~~~~~~~~~~~~~~

: -- < 5 4~ ~~~~~~~~~54

EgI Rings

93l Comunas

JANUARY 1994

SsU7rce: M<>han 1986, map 3-1.

GROWTH AND THE CHANGING STRUCTURE OF BOGOTA AND CALI 45

IBRD 25347

Sector System

25 ComunasX

JANUARY 1994

4tj UNDERSTANDING THE DEVELOPING NIETROPOLIS

Map 3-2. Cali: Ring and Sector Systems

IBRD 25343

Ring System

< b tR~~~~~~E Comnunos

JANUARY 1994

.S'nirc: Stohari 1986, map 43-2.

GROWTH AND THE CHANGINC; STRUICTLURE OF BOGOTA AND CALI 47

IBRD 25344

Sector System

< A E ~~~~~~~~~Comunas

JANUARY 1994

48 L1NDERSTANDINGr THE DEVELIOPING METROPOLIS

Map 3-3. Bogoti: Population Density by Comuna

IBRD 25348

1973

People per hectare

350 and over_ 300 to <350)

1EZ0 250to <300200 to <250150 to <200

lOO to<lS I so 771 Sto <10 I/ o as

Scrur: . <ohan 1986, map 3-3.

GROWTH AND THE CHANGING STRUCTURE OF BOGOTA AND *ALI 49

IBRD 25349

1978

People per hectore

350 and over300 to < 350250 to <300200 to <250150to <200 7}(

111100 to <150

[ <50 91

4 b j sS 54 5

241i23 Comunos 199

JANUARY 1994

50 UNDERSTANDING THE DEVEL.OPING METROPOLIS

times denser than the least dense, but some of the outlying areas in thesouth are about as densely populated as some of the inner ones. Thus,although exponentially declining population density functions fitBogota and Cali well (as documented in the next section), such a sum-mary representation hides some of the actual diversity.

The comunas are too disaggregated to provide a summary of popula-tion changes. There are two natural ways of dividing a city spatially forthe purposes of analysis: into "rings" and into "sectors," or pie slices, asshown in maps 3-1 and 3-2. (Because Bogota and Cali are semicircularin shape, so are their rings.) These gross spatial divisions are used foranalysis throughout this study. Populationi and density declined slightlyin absolute terms in the center of the city over the period 1964-78 (seetable 3-5). In 1964 the highest density was in the CBD (ring I and sector1), but ring 2 (about 1-3 kilometers from the center) was the densest in1973 even though it also lost population between the tvo census years.It is striking that ring 3 (3-6 kilometers from the center) and ring 4 hadsimilar densities in 1973 and 1978. The fastest growth clearly occurredin rings 5 and 6. Indeed, ring 5 accommodated twice as much of theincremental population during those years as the rest of the city puttogether. Land prices also increased the most in ring 5. Ring 6 was

Table 3-5. Growth in Population and Density in Bogoti

Average annual rate of grouwth in

Area, 1973 Population, 1978 population and densitY (percent)

Divjision (hectares) (thousands) 1964-73 1973-78

Ring1 398 82 -1.8 2.42 1,357 280 -0.2 -0.33 2,575 426 3.2 3.44 5,960 923 5.7 0.75 14,329 1,592 14.6 6.36 5,804 189 37.6 13.8

Sector1 398 82 -1.8 2.42 4,357 696 5.8 7.13 5,313 859 8.3 2.84 1,914 250 7.1 -2.75 3,066 213 7.4 0.26 5,673 680 10.9 6.07 5,064 325 5.7 -1.98 4,638 388 7.1 14.5

C'ity 30,424 3,492 7.2 3.9

Note: Density for all vears has been calculated based on the 1973 area.Source: 1964 and 1973 population censuses; 1978 World Bank-n.ANr Household Survey.

GROWTH AND THE CHANGING STRUCTURE OF BOGOTA AND CALI 51

almost uninhabited in 1964. The main part of population growth inBogota in the period covered by table 3-5 has clearly been in the outly-ing rings. There has been some decentralization of population, partlybecause population in the two inner rings has declined somewhat, butmainly because of the much larger additions in the outskirts. The lowerpanel of table 3-5 gives the same information for the radial sectors. Sec-tor 1, the CBD, is identical to ring 1. The southern sectors (2, 3, and 4)are almost twice as densely populated as the northern sectors (6, 7, and8). The sectors have grown somewhat uniformly, although sector 6 grewfaster than the others. Not until the early 1960s did the western part ofthe city began to fill up; before then the city had largely been a north-south strip stretching along the mountains. Sectors 4 and 5 form theindustrial corridor that Le Corbusier established in 1951. On average,the densities of rings 3 and 4 are very similar to those of sectors 3 and 4.The northern part of sector 8 is even sparser than the average.

We now turn to the spatial distribution of income (see chapter 4 for asystematic discussion of this subject). Again, comunas are too disaggre-gated to provide a useful picture. It should be noted, however, that theincome of the poorest comuna was only one-twelfth that of the richest.Each comuna was relatively homogeneous within itself. A glance at map34, which illustrates the distribution of household income per capita(HINCAP) by comuna for 1973 and 1978, shows that mean income variedacross the city in a remarkably regular fashion, increasing in income asone follows the map clockwise from south to north. There is no discern-ible pattern from the center to the periphery, however (see table 3-6).Only a mild pattern is discernible by rings. There is a tendency forincome to rise somewhat from the center toward the periphery andthen to decline; ring 5 is poorer than the rest. The differences, however,are small, and the coefficients of variation are large. For most of therings, the proportion of households in the ring is not very differentfrom the share of income, except for ring 5. Thus it is clear that theincome distribution pattern is quite different from that in U.S. cities,where income, in general, increases as one moves from the centel to theperiphery.

The data partially support the idea that the poor in cities in develop-ing countries get pushed to the periphery. The pattern is more distinctwhen viewed by sectors. The richest sector (sector 8) has a mean HINCAP

more than five times that of the poorest (sector 2). Except for sector 6,which is relatively poor, income increases as one moves from sector 2 tosector 8. The most heterogeneous zone is the CBD, with a very high coef-ficient of variation. The share of income received by the poor sectors ismuch less than their share of households. The ranking of the sectorsdid not change between 1973 and 1978. It is therefore clear that a spa-tial analysis of Bogota is of more interest when done by radial sectors

52 UNDERSTANDING THE DEVELOPING METROPOLIS

Map 34. Bogota: Distribution of Mean Household Income Per Capitaby Comuna

IBRD 25350

1973

_ 60%

43] Comunas

JANUARY 1994

Source: Mohan 1986, map 3-4.

GROWTH AND HFH CHANGING STRLI(:TL RF( OF BO(;OIA AND) CAL.[ 53

IBRD 25351

1978

40

56 i ~~~~~84

VP 2~ ~~~5145.

~~~~~~~~~~14

25 WEE i

EF4 Comunas

JANUARY 1994

54 UNDERSTANDING THE DEVELOPING METROPOLIS

Table 3-6. Spatial Distribution of Monthly Income and Populationin Bogoti

1973 1978

Meaan Sharei of Meanr

Mlean house/oLd Distibution in rome in Mlean honUsehold Distributionhouse/hold inromne per of house- ring or household income per oj]house-

income caita holis sector income caJpita holds

Dimwi.mon (1973 pesos) (1973 peso5) (percenit) (perrent) (1978 peSos) (1978 pesos) (percen7t)

Ring1 2,114 884 3.5 2.6 8.343 2,49( 3.(1

(2.10) (1.03)2 2,976 1.016 10.8 11.5 16,047 5,570 9.8

(1.92) (1.04)3 3.722 1,073 12.7 16.9 19.928 5,913 13.1

(1.59) (0.93)4 3,046i 788 30.7 33.5 14,772 4,105 27.2

(1.94) (1.40)5 2.264 533 39.1 31.7 10,818 2,277 41.6

(1.86) (1.I0)6f 3,202 753 3.3 3.8 15,999 3,178 5.3

(1.93) (1.26)

.Sector1 2,114 884 3.5 2.6 8,343 2,490 3.0

(2.11) (1.03)2 1,583 414 18.1 10.2 6,833 1,585 19.9

(1.44) (0.52)3 2,002 484 24.4 17.5 9,930 2,234 24.6

(1.41) (0.86)4 2,347 646 9.8 8.2 12,843 2,800 7.2

(1.68) (0.88)5 2,453 700 7.7 6.8 13,893 3,067 6.1

(1.68) (1.04)6 2,349 5,56] 17.0 14.3 12,801 2,744 19.5

(1.66) ((.98)7 45.791 1,265 12.6 20.6 16,163 3,750 9.3

(1.61) (1.07)8 7,895 2,258 7.0 19.7 32.804 11,354 11.1

(1.50) (o90()

(Ciy 2,791 751 1(00.0 10(1 13,805 3,629 100.0(1.89) (1.22)

Nolte: Numbers in parenthieses are coeffi(-ients oF sariationi.

Source: 1 973 population cenisus sample; 1978 Vsorld Bank-DANk fIlousehold Survey.

than bv rings, although the rings do exihibit some interesting character-istics. Thle magnitude of differenices in incomile betweeni the diffelelitparts of the city is surpr-isinigly large for averages taken over large num-bers (see table 3-6).

GROWTH AND THtE CHAN(GING STRUCTURE OF BOGOTA AND CAL.I 55

Such differences are seldom encountered between different regionsof a country. In Bogoti (as in Cali) the rich clearly concentrate in cer-tain locations. Althouigh the poor are to be found in most parts of thecitv, the declining coefficient of variation between 1973 and 1978 indi-cates increasing spatial segregation of income classes. The growth of thenorthern part of the city, the resulting gradients for land values andlpopulation density, the use of transportation, an(d the type of housinigall conform milch more to patterns foundc in developed countries thanin the city's southiern section.

One other ring characteristic should be mentioned: household size,which increases from 3.83 in the citv center to 5.3 in rinig 5 and 4.9 atthe peripherv (ring 6). ILarger households require more space andhence are rnore likely to locate awvay from the center. In addition,although larger households earn higher total inconmes, the added comil-pensationi is not enougih to offset larger household size. Thus house-hold income per capita declines with increasing household size. Theperiphery of Bogota thierefore contains larger as well as somewhatpoorer households, the household income per capita measure being amore meaningful measure of welfare than total household income.

A broad pictire of the changinig structur-e of Bogota is now beginiingto emerge. Growth has entailed the densificationi of successive ringsmoving outward from the center-, but this densification is not unliformllbetweeni different sectors of the city. The rich have continued to movenorthi (sectors 7 and 8) and to settle in relatively low-densityi neighbor-hoods, whereas the poor have largely moved toward the southl andsouthwest (sectors 2 and 3) and to some extent toward the west (sector5). The striking element of this pattern is the relatively high constantdenisity that is observed as one moves from the center toward the south-ern periphery. The poor appear to live in settlements of similar dlensi-ties whether in the center or on the outskirts. There is a clear increasein household size toward the periphery, and the densiqt of dwelliigUllitS would therefore be expected to decline outward in the poor sec-tors as well. A similar pattern can be observed in Cali.

The Evolution of Land Values and Populationand Employment Densities

Table 3-7 expands on the information presented in table 3-5. The maxi-mum distance from the city center- is about 15 kilometers (to ring 6) inBogota and about 10 kilometers (to ring 5) in Cali. Growth has clearlyoccurred by accretion in the outer rings; density in the CBD hasremained at about 200 persons per hectare. Growth on the fi-inges of

Table 3-7. Change in Population Density by Ring, Bogoti and Cali

BogtadC (Ai

Den.iI) (populazliori per hectlare) Deatitv (population per hectare)

Area (hectares) 1964 1973 1978 Area (herl(iresj 1964 197 3 1978

Ring

1 398 220 180 205 140 210 150 160

2 1.357 210 210 220 1,500 140 125 135

3 2.575 1(00 140 140 3,00() 95 135 160

4 5.960 90 150 155 3,00() 25 70 100

5 14.329 25 8( 110 1,500 25 50 70

6 5,804 1 17 32 na. na. na. na.

CitV 30,424 50 95 115 9,100 70 103 121

Densitv gradient (g) 0.18 0.15 0.12 0.51 0.44 0.25

na. Not applicable.Note: All figures are r-ounded.a. City area has been kept conistant for all calCulations of city density, unlike in table 3-4. In fact, botil cities have grown duhitlg the period in question. ald

many peripheral areas inicluded here were outside the 1964 city botindaries.

Source: Data for Bogota are fiom 1964 anid 1973 population censiuses and 1978 'World Bank-D.xNE 1lousehold Sturey. Data for Cali were calculated front data

in Iabares (1979).

GROWTH AND THE CHANGING STRUCTURE OF BOGOTA AND CALI 57

the existing cities has accompanied densification of the inner rings.This process is somewhat different from the growth pattern observed inmost U.S. cities in the early twentieth century. Thus, while Bogota andCali have decentralized with growth in the sense that a smaller propor-tion of the total population lives within any area of constant radius, theyhave not yet decentralized in the manner of many cities in the UnitedStates, where central cities have actually lost populations in substantialmagnitudes.

The last line in table 3-7 gives the measured density gradients from1964 to 1978 for the two cities. The unit of observation is the barrio, orneighborhood. 2 These density gradients were estimated for Bogotafrom about 300 observations in 1964 and about 450 in 1973 and 1978and for Cali from about 130 observations in 1964 and about 200 in 1973and 1978. These data are more disaggregated than those usually avail-able for such calculations. The gradients reported for different cities inchapter 2, for example, are based on far fewer data points in each case.As noted in chapter 2, g declines as expected for both cities and ishigher for Cali, the smaller city, than for Bogoti. For purposes of com-parison, we may recall that, around 1970, gwas 0.11 for New York, 0.08for Chicago, 0.08 for Tokyo, 0.07 for London, 0.22 for Seoul, 0.17 forNMexico City, 0.12 for Sio Paulo, and 0.12 for Bogota. Among smaller cit-ies like Cali, it was 0.19 for Monterrey, 0.41 for Guadalajara (Mexico),0.27 for Belo Horizonte, 0.19 for Recife (Brazil), 0.25 for medium-sizecities in Great Britain, and 0.31 for similar cities in Germany. ThusBogota and Cali have population density gradients that are typical forcities of their income levels, size, and available transport systems. Thegradients have declined over time with increases in population andincome, much as would be expected. Both cities exhibit substantialdecentralizationi with growth but with successive densificationi of innerrings, unlike many North American cities.

Simple urban economic theory (Mills 1972) leads us to expect thatthe land value and density patterns should be broadly consistent. Thecompetition for land for its highest economic use gives land its value.Hligher density implies higher relative use, and, conversely, higher landvalue gives people an incentive to use land more intensively. In a full}developed urban model, land values and population densities wouldboth be endogenous variables that would be expected to move in a con-sistent fashion.

For most cities it is typically difficult to obtain good information onland values, especially comparable information over a long period oftime. It is also difficult to separate the value of land from that of thestructures built on it. We were fortunate in obtaining a unique data setof about 6,000 transactions in Bogota covering the period 1955-78 fromGuillernmo Wiesner of Wiesner and Cia. Ltda., a long-established Bogota

Table 3-8. Evolution of Land Values by Ring, Bogota and Cali

Bogoti Cali

Averagre A verageAverage ailcual .4ve7age annual

distancefrom growu1th mrte, distarcefrom growth rate,the centfr land values (1978 pesos per square meter) 1964-78 the center Land values (1978 peyos persquare mter) 1963-79

(kilometers) 1963--65 1972-74 1975-77 (percent) (kilomneters) 1963 1974 1979 (percent)

RitngI 0 4,250 3,900 3,100 -2.3 0 5,900 4,600 6,400 0.62 2.2 1,850 1,660 1,550 -1.3 1.8 1,100 1,100 2,400 5.63 3.8 1,350 1,350 1,32() -0.2 3.4 520 480 1,030 4.94 6.5 870 1,080 1,130 1.9 5.4 380 410 960 6.65 9.8 570 800 850 2.9 6.9 150 370 810 12.06 15.4 370 700 730 4.9 n.a. n.a. n.a. n.a. n.a.

Gradienta (h) n.a. 0.16 0.08 0.07 n.a. n.a. 0.55 0.51 0.25 n.a.

n.a. Not applicable.Vote: 1978 exchange rate: US$1 =Col$38.a. Lanid value gradienits were calculated as = - Veh', wher e V, is the price of land x from the center, V, is the theoretical value at the center, and h is the gra-

dient.Source: For Bogoti. Villamizar (1981). For Cali, Velasco and Mier (198(0).

GROWTH AND) THE CHANGING STRLUCTL'RE. OF BOGOTA ANI) CAL.I 59

real estate firm. Because all transactions observed were for land that wasnot built on, we did not have to artificially separate the values of theland and the structures. (For detailed analysis of land values in BogotA,see Villamizar 1981; Mohan and Villarnizar 1982; and WAagner 1984.)The data set for Cali was obtained from the Cali Planning Office.

Table 3-8, which shows the pattern of the evolving land price surfacefor Bogota ancl Cali by rings, is remarkable for the highly regular pat-tern that emerges and is consistent with expectations. The prices aregiven in constant 1978 Colombian pesos because the earlier prices wereinflated by the consumer price index. For both Bogota and Cali. landprices decrease from the center toward the peripherv in a regular fash-ion; central land values remain broadly constant in real terms, whereasconsiderable and consistent increases are recorded towardl the periph-ery. This is in keeping with the evolution of density patterns shown intable 3-7.

The constancy of real land prices in the central areas of Bogot.i dur-ing the 1960s and 1970s is remarkable. A word of caution is necessaryregarding the significanit decline in the C:RD (ring I ). Wagner (1984) hasdemonstrated the existence of some biases in the measulement of cen-tral land values. As may be expectcd, less and less vacant land is availablein centr-al areas over time; the ntimiber of vacant-land t-ansactions there-fore falls, and larger lots became eveni scarcer. In the CBD larger lotsfetched higher unit prices than smaller lots, but since the number ofsmaller-lot tranisactiolIs was higher, the unlweighted averaging processbiased the average CBDt land values downwarcl. Hence, the fall in realurbani land values in the Bogota CBI) duLinig the 1960s and 1970s is par-tially illusory. I.and values may be regar-ded as having bcen constant overthe period.

It is also important to note some findings that go against the conven-tional wisdom. It is widely thoughit that la.cl values have been skyrocket-ing in cities in developing countr-ies and that they will do so inclefillitely.In fact, in BogotJi between 1955 and 1977 they grew at only about 3 per-cenlt a year il real terms, a rate not far friom either the per capita GN'

girowth rate for the cotintrv or the return on any alternative asset. Theabsolute land prices were not very different for Cali. a finding consistentwith those of Mills and Song (1979). They' foutnd that, in the Republic ofKorea, land values in larger and smaller cities were not ver-v different,and the rate of incr-ease was marginally higher among cities categorizedlas second-largest than among the largest cities.

The main point is that cities grow in a regular fashion and that theevidence from BogotA and Cali is quite consistent with the theoreticalexpectations outlined in chapter 2. These regularities are, moreover,results of the r egularitics in hulimaln behavior that are investigated in therest of this study. This evolving patterin of land value gradients conifir ms

60 tUNDERSTANDING THE DEVELOPING METROPOLIS

that land values have essentially been determined by the economic com-petition for access to space according to its best use. Distance from thecity center emerges as the best explanation for variance in land valuearoulid the cit,. It is therefore the access characteristics of each plot oflandl that are the principal deter-minanits of its value. Other variables,sucIh as amenities, availability of inifrastr ucture, neighborhood charac-teristics, and zoninig, are less significant explanators of land value vari-ance in BogotA. The next section shows that although distance fromii theCBD is the prime determiiiianit of land values, other factors, such asaccess to subsidiary shopping centers and main transport arteries, arealso significanit in explaining the ridges and values encotintered in theland value sur-face of Bogota. There is little evidence that the land mar-ket has not been working competitively in Bogota.

Uhy have real urban land prices in Bogota and Cali risen at moderaterates? First, (lurinig the period tinder consideration the supply of houls-ing gener-ally matched the rapid gr(i)wth in demand (see chapter 7). Therapid land development that took place in the barrios pirata.s (extralegaldevelopimienits) essentially kept tip the stipply of developed or semidevel-oped land in response to constanitly rising demand. Second, explicitgovernmenit policv aimed at rapid investment in urban inifrastr-uctur-ehelped provide at least a minimiial level of services to the emergilig newdevelopments-both legal and extralegal. Thircl, the availability of dif-ferenit sets of instruments for investimient in Colombia discouiraged over-crowding in real estate investment. Fourthi, the competition betweendevelopers, who had relatively free entr-v into the housiig and land( mar-kets, helped keep the returins to landc dlevelopment. withinl moderate lim-its, which molre or less matchedl national income growth per capita over-this period.

Comparing the densitv and land valuie gradients in tables 3-7 and 3-8provides ftirtlher evidence:

Bogrod (:cjli

Land 7 value Den.sity Lanid vabir DentlNi

1964 -0.15 -(.18 -0.51 -0.511 973 -0.08 -0.15 -0.25 -0.441978 -0.07 -0.12 -0.23 -0.25

According to standard urbani economic theory, we wotIld expect thepatternl of land valtie gradients to he similar to density gradients. Fur-therniore. their levels are consistently lowver than the density gradients,as is often theorized (Mills and Song 1979). A comparisoni of estimated1-) (lanid value at the center of the city) with actual V( reveals that ourestimates are consistently lower than the actual values. This implies thatthe gradient of the curve shouldl, in fact, be mtich steeper at the center

GROWTH AND THE CHANGING STRUCTURE OF BOGOTA ANt) CAlA 61

of the city than we have estimated. Because of the high concentr-atioln ofeconomic activity at the center, one can expect larnd values to be deter-mined much more by employment density than by residential density.The latter falls rapidly from the center, and it is quite plausible that landvalues will exhibit a similar decline. Estimated V10 is therefore likely to belower than this central peak. The estimated densities were consistentlyhigher than the actual densities for the CBD, whereas estimated land val-ties were consistently lower than real land values. What is to be expectedis that residential densities will increase somewehat fi-om the CBD anidthenl decline, whereas land values should show a rapid decline from thecitv center at first and then a slower decrease.

L.and values are expected to be extremely high at the city centerbecause of the concentration of economic activity there. It was sug-gested in the last chapter that employment densities are expected to behigher in the center and that they typically have a steeper gradient.Table 3-9 vividly illustrates this for Bogota and Cali. Estimates of employ-ment density gradienlts are more difficult to come by, than estimates ofpopulation density because employmetnt data are not usually availablebv location. Thie gradients for Bogota in 1978 appear to be similar tothose calculated by Mills (1972) for a sample of L.S. cities in 1963. Thetable shows that manufactturing has decentralized most, as expected, fol-lowed by commerce and serxices. Finanicial activities are heavily concen-trated in the center and might have become more so dur ing the 1970s.

Table 3-9. Employment Density in Bogoti(jobs per he tare)

Mlan u-Item All facrlunng Commwrce 'inance .Srvices

jobs per lectare, 1978Ring 1 425 43 97 102 128Ring 2 158 28 36 21 54Ring 3 77 24 14 4.4 26Ring 4 42 12 8.( 1.8 15Ring 5 21 5.6 4.8 0.4 6.0Ring 6 7.0 1.1 O.fi 0.1 3.0

Employment densitygradients

Bogota. 1978 0.30 0.21 0.30 0.56 0.29Bogota, 1972 0.33 0.33 0.32 0.44 0.35

Cali, 1978 0.72 0.71 1.0( - 0.64

Average for selectedU.S. cities, 1963 0.26 0.27 0.35 - 0.33

- Not available.Sourre: Colomibian data were calcuilated from K. S. Lee (1989), 1972 Phase 11 Household

Survey, and 1978 World Bank-DANE Household Suirvey. U..S. data are from Mills (1972).

Table 3-10. Distribution and Density of Workers' Residences and Workplaces, Bogoti

Ditrnbilbion ol rwrkers1972' 1 978 b DensitY of Workers

BY residence BY voirkpluce BY residence By wvorkplare 1972 1978 1972 1978Division Area (herares) (percent) (percent) (perceit) (percenit) BY residence per hectare BY zworkplace per hectare

Ring1 398 1.7 25.7 2.4 14.5 37 73 507 4202 1,357 10.9 15.0 9.7 18.3 70 85 87 1553 2,575 13.1 16.3 13.3 17.0 44 61 50 764 5,960 34.5 20.7 26.0 21.2 50 52 27 415 14,330 37.6 20.4 43.0 25.5 23 36 11 206 5,804 2.2 1.8 5.5 3.6 3 11 2 7

Sector

1 398 1.7 25.7 2.4 14.5 37 73 507 4202 4,357 18.3 7.6 19.1 8.5 36 52 14 223 5,313 25.1 13.4 23.4 13.4 41 52 20 294 1,914 9.5 9. 3 6.5 9.2 43 40 38 555 3,065 7.1 10.9 6.3 12.6 20 24 28 476 5.673 1 6.3 9.8 17.8 12.4 25 37 14 257 5,065 13.3 9.9 10.1 10.2 23 2 4 15 238 4,638 8.8 13.4 14.3 19.M) 10 37 23 47

City 30.423 100.0 100.0 100.0 10(.(0 28 39 26 38

Note The su-rvey inc luded 877.000 workers in 1972 (incIlidi ng 786,00/) w hose worIkplaces were identified) an d 1.185 000 workers in 1978 (including1,150,00() whose workplaces were identified). Informatioun on the workplaces of sonie workers WaS Unavailable.

a. Data are fiom Pachoin (1979)) table 5. Primary data for workers bs residenice was from the 1972 Phase 11 Household Survey househiold file, and bv work-place 1rom the 1972 Phase 11 Household Survey person file.

h. Data are fiom 1978 World Bank-DANE Household Survey.

GROWTH AND THIE (CHANGIN(; STRI!(:TURE OF BOGOTA AND CM.A h:l

Employnient density gradienits are muLch steeper than population den-sitv gradients.

Cali is still very centralized. It really has only one major employmeintconcentration-the center. Because we do not have reliable (lata for ear-lier vears, it is difficult to say whetlher employment has decentralized inCali vet, except in marn ufactuLing.

The employment pattern is examined in more detail itl chapter 6.Here we merely neecl to establish that the emplovymient structule ofBogota and Cali is consistent with the expected patterni of growth anddecentralization. Bogot.i's CBD had a net loss of jobs in the 1970s, prima-IrilY in manufactUring. SoIle Of the CBD funictions shifted to rinlg 2, inthe nor-thernl sector, which is more easily accessible to the rich living insector 8.

The Changing Structure of Bogota: Some Wrinkles

So far we have treate(d the city in a relatively simple manner; The mea-surenmenit of each gradlienit assumiles that the citv is symmetric around thecity ceniter and is essenitially monocentric. This is justified in the mainbecause both BogotA and Cali are roughly semicircular, though con-strainied by moLunitainis on one sidle, with the city center roLughly at thegeograplhic center of the semicircle in each case. Thie income andi popu-

lation distributionis are not, however, svmmnletric. In BogotA the nor-thernlsectors (7 and( 8) were characterized as particularly richi anld the souith-ern sectors (2 and 3) as poor. In Cali the picture is more mixedl, butbroadly, the western sectors (2, 6, and 7) are riche r than the eastern sec-tors (3, 4, an( 5). In generalh jobs exceed the numihber of resident work-ers in the ricIh sectors, and the converse is true in the poorer sectors (seetable 3-10). As might be expected. population density is higher in thepoor than in the rich sectors. This suggests that the smooth patterllsdepicted in the precedinig discussioni hide considerable diversitv andthat it would be useful to look at the density and land( value patterns ildtifferenit sectors to observe the extenit to which earlier genieralitiesremaini true. These patterns are partictila-ly importanit for houisirng andtransport decisiois. For example, the denisity inifo-mationi presentecd intables 3-9 and 3-10 shows that emplovmenit dlensities are greater thanresideential densiiies in the inner rings but less in the ouiter r ings.

Although the commiiutinig pattern would be predominantly inward, thedecentralizationi of both employmenlt and residence suggests that adecr-easinlg proportion of trips for work would be made toward the cen-ter as tirne goes on.

A more interestinig residential and work patterni is shown in the bot-tom part of table 3-10. Most striking is the job deficit in pooI sector 2 (in

14 UNDERSTAND[NG THE DFVELOPING METROPOLIS

the south) and the job surplus in rich sector 8 (in the north), whichleads one to expect conisiderable crosstown commuting in addition toradial commutiig. This chariging pattern points to the need to increasethe number of circumfer-ential routes so that north-south crosstown traf-fic can avoid inner-city congestioni. This is possible onlv if appropriateinvestimienits and changes are made in the prevailing road structure. It istherefore importanit to observe and uniderstanidl trends in transportationipatterins in order to design workable transportatioti routes and rnakeappropriate policy decisions about public investment in infrastructure.

One anomalyv may be observed in table 310: In contrast to tihe gen-eral pattern of decentralization, there seems to have been anl increase inthe density of populationi of worker-s in Bogota's (:BD between 1972 and1978. Table 3-7 also shows some iicrease in populationi denisity in theCBD (from 180 persons per hectare to 220 persons per hectare). This ispartly explained bv the rapid change in the character of the CBD's resi-d(ent poptilationi dIurinig the 1970s. Many families moved to ouiter loca-tions and were replaced bv single-member households (single workers).The housing str-ucture also changed from apartments to roominghouses (popularly known as inquilinatos in Bogoa). Single-memberhouseholds, which are characteristically (liffictilt to cover in householdsuLrveys, were probablv undercotiuited in the 1972 sur vev. Repeat visits inthe 1978 World Bank-DANE survey ensuredl better coverage. The sub-stantial shift over time towaid single-worker- residences in central citiesis quite cojiniiion; it is well knowin in Nor tih America and EuLropean citiesand has also begun to appear in the central cities in some developingcoutnitries.

UJntil now the distance friom the city center hias been emphasized asthe kev variable affecting the access characteristics-that is, the proxim-ity of land parcels to economic oppoirtunities-of a particular locationand, therefore, the value of land. T he hvpothesis is that large or denseagglomiier-ation.s of people are instrumental in increasing these eco-nonic opportuniities, and this is the reason for the clustering of popula-tion near the city center. Thus a concenitration of economic activity inthe center produces relatively high populationi densities anid corre-spondinigly high land valLies, both of which decline with distance.

The observation that the rich live in some parts of the city and thepoor live in other areas leacis to the revision of some of these ideas. Thatmore jobs are located in rich sectors means that those sectors are eco-nomically more attractive and firms have a greater tendency to locatethere. The lower population densities of those areas are more thanenhanced by the purchiasing power of the resident population. Accesscharacteristics conifer-red by proximitv to the city center or by short dis-tance from workers' resid(ences are he ightened by the greater purchlas-ing power of richer households. Thtus the access characteristics of

GROWTHJ ANI) THE CHANGING SI UCTtURF OF BOGOTA AND (.AI I 65

locations within close proximity to richer areas of the city arie improved,and land prices are bid up. There is more competition for retail space,tor example, in these areas. The establishment of shoppinig centers andother activities then attracts greater employment, contribtitinig fur thetrto the positive access characteristics of these locations.

T'he prodtict of' populationi and mean incomile is probably not a goo(nmeasure of purchasinig power because the reqtuirements of a large inim-ber of poor people do not aggr-egate: each househlold has meagerdemands, so poor sectors can support only a limited number of cco-nomic activities-small retail activitv for subsistence daily needis but lit-tle else. Thuis, in Bogoti, much other retail and wholesale aclivitvcontinuies to be concenitrated in the center, and more people live in thepoor southerin part of the city thani there are jobs there. Conversely, ther ich neighbor-hoocds in the north have a mtich highier clemand for goodsand services than tileir population densities wotildl suggest. Moreover.office and other professionial activity has tendled to move northi in thesame direction as the rich. Thus the access char-acter-istics ot' theseneighborhoods have improved fturther andI are not adeqtiately mea-sured by populationi densities. In addition, because the infiastrulcture(for example, roads, lighting, water supply, and sewerage) is better inthe richier areas, the quality of life and the intrlinsic characteristics of' thesites are better as well. All these factors combinie to produce land valiesthat are higher in the r ich areas than in other neighborhoods no fartherfiom the citv center.

The estimates of populationi dlensity andl lanid value gradienits for eachsector in Bogota and Cali (table 3-11) are striking. The declininig expo-nential funictioni is still a good approximation of lancl valtues for eachsector. The populationi densities do not do as well; the estimated densitygradients are not significaitly dlifferent from zer-o in a tinmber of sec-tors. Indeed, in Cali sectors 4 and 5 exhibit mildly positive gradients.whereas in Bogota the density gradienits are low or insignificant for 'ec-tors 2, 3, 5, and 6. Wliat these sectors have in commoni is relatively lowmean iicome. The lancl valie graclienlts. however, are not significanitlydiffei-erit fi-om those in othier sectors.

To udiclerstanid thlese phenomenia. we need to delve furtiher into therole of land valutes and their effect on urbani structure. Wheni land val-tues are high, capital is substittited for land, and the result is the co0n-struction of' taller buildinigs. We can therefore expect to observe, onaverage, taller buildings in city centers and in zones in which land pricesare high. As land prices increase, single-family homes arc replaced bymultifamly dwellings (apartmen ts), and residlential densities rise. Althouighresidential denisities rise per unit of land area, living space per personidoes not necessarily decrease in wealthv areas. These options, however,are not open to the poor.

Table 3-11. Land Value and Population Density Gradients in Bogoti and Cali, by Sector

Bogotd ( ali

Household Householdmean inotzoe Learnd Valueh Poptulation density meant itncome Land vialued Poepulalion densiqy

indexa g R2 h Y indexa ' g R2

h R2

Sector

1 l6 in.a. n.a. n.a. n.a. 163 n.a. n.a. n.a. n.a.

2 52 -0.15 0.54 -0.11 0.06 212 -0.42 0.69 -0.13* 0.0o

3 74 -0.01 0.09 -0.02* 0.01 84 -0.45 0.73 -0.10* 0.0

1 96 1 1 -0.12 0.09 58 -0.42 0.69 +0.13 0.11

5 103 -(.08 0.71 40.05* 0.03 82 -0.55 0.77 +0.07* 0.02

6 97 -0.10 0.71 -0.05* 0.02 125 -0.21 0.41 -0.26 0.167 122 -0.08 0.72 -0.16 0.50 219 -0.86 0.59 -0.13* 0.058 236 -0.07 0.55 -0.14 0.32 n.a. n.a. n.a. n.a. n.a.

n.a. Not applicable.Note: All coefficients are significant at the 0.01 level except those marked with an asterisk.a. Percentacge of meani household incomile ior the city.b. Data are for 1975-78.c. Data are for 1978.d. Data are for 1979..Siource:Velasco and Mier (1980); Tabares (1979); Paclhoni (1979); Villainsizar (1981); City Stmdy barrio file.

GROWTH AND THE CHANGING STRUCTURE OF BOGOTA AND CAlA 67

The rates of price increase were no higher in the rich areas than inthe poor ones (Villamizar 1981). In Bogoti the rate of increase(adjusted for inflation) in the rich sector (sector 8) was about 2.5 per-cent a year between 1955 and 1977; it was about 4 to 7 percent a year inthe poorer sectors (2, 3, and 6). The price levels, however, were consis-tently lower in the poor areas. These data indicate that although eachland parcel is nonsubstitutive to some extent, a citvwide land market isfunctioning. Whereas land prices in poor areas continued to be lowerthan in richer areas, a catch-up phenomenon was observed: prices ofland parcels equidistant from the city center were not too dissimilar.The natural result of this phenomenon is that while the rich substitutefor land with capital, the poor substitute for land by crowding.

Much of the housing in Bogota still has fewer than two floors. Thenumber of floors declines rapidly from the C.BD, and the northern partof the city, the richer sector 8, has taller buildings than other areas(table 3-12). The average number of floors declines systematically byring, as does the average age of dwellings, and the proportion of single-family homes increases with distance from the center. These patterntsare very much according to expectations: capital is being substituted forland in the shape of taller buildings in the inner rings; the city hasgrown by accretion at its edges, and therefore the outer rings havenewer houses; and apartment buildings or semidetached houses arereplacing single-family houses as prices increase near the CBD. Land

Table 3-12. Spatial Pattern of Housing in Bogota, 1978

A verage age of Average diweli ngMl7ean haousehold Single-family dweledng una Auerage number space perpemxon

Division income index units (percent) (wears) of floors (squoare mtelers)

Ri nsg

1 62 39 16 7.1 142 116 57 21 3.5 233 124 74 1 6 2.8 304 112 86 12 1.7 235 82 95 9 1.8 186 122 100 8 1.5 26

Seclor1 61 39 16 7.1 142 53 96 13 1.4 123 74 91 10 1.8 194 96 91 1 1 1.9 255 103 72 18 3.4 206 97 92 10 2.1 217 122 84 1 7 1.9 308 236 59 1 0 2.9 45

City 100 85 12 2.1 2 1Soeurce: Mohan and X'illamizar (1982).

68i UNDERSTANDING THE DEVELOPING METRROPOLIS

prices are performinig their function well, and the housing marketseems to be responding as expected.

There is no clear pattern of average dwelling-unit space per personexcept that it is low in the CBD. We would expect that living space perpersoni wotild be greater in the outer rings because people would betrading space for higher transport costs. If we now look at the sectoralpattern, it is clear that the poorest sectors (2 and 3) in the south havemuch less living space per person than the rich northern sectors. Thusthe poor are substituting crowding for land, and the rich are substitut-ing capital. (Chapter 7 analyzes the hotIsing market in detail. One of theimportant findings reported there is that housing demand is inelasticamong low-income residents, and they spend a high proportion of theirincomes on hotlsinig.) Crowding in sector 2 is about the same as in theCBD; lower land prices at the periphery have not led to more space perperson. It would seem that the poor have limited choices and that thereis a limit to substitution by crowding. The percentage of single-familyunits increases consistenitly toward the peripherv, while the averagenulmber of floors per structture declines. The population density in thepoorer peripheral residential areas is not lower than in the center, butthe density of dwelling units and households is. .arger households cantake advantage of lower land prices at the periphery and live in largerdwelling unlits to have about as much per-personi living space as smallerhotiseholds nearer thie city center.

We now begin to understand why the land value gradients hold tipeven when thie cities are disaggregated into sectors, whereas the densitygradienits do not. The way the land market funictionis results in land val-ues that are not too different at similar distances from the center. Therich sectors hiave highel- land values because of better employmentopportunities, as well as better neighborhoods and inifrast-uclttre.Becauise lan(d values are relatively regular, the poor have no choice butto substituite for landc by crowding. Since it is the larger houscholds thatlocate on the periphery, they still have to live at higlh densities to coin-pensate for the land prices, which are similar to land prices in the richsubuirbs. They cannlot buyv more space by substituting capital for landbecatise housinig would be too expensive. For example, the averagenumber- of floors is 2.9 in sector 8 and only 1 .4 to 1 .8 in sectors 2 and 3;average dwelling unit space per personi is lowest in sector 2 and is alsolow in peripheral ring 5 (see table 3-12). It would have been instructiveto see how these indicators had changed since the early 1970s, btit com-parable clata were n ot readily available. In any event, we observe highpoptilationi clensities on the periphery of some parts of the citv and con-sequently there is no measurable density gradient in those sectors. Therich sectors still have a density gradient, and we can therefore observegradients for the city as a whole. Nonexistent density gradients in sornesectors of the city are consistent %4ith relatively strong land value gradi-

GROWTH AND THE CHANGING STRUCTlJRE OF BOGOTA AND CALI 69

ents. This suggests a need for caution in interpreting similaritiesbetween citywide population density and land value gradients.

One other important aspect of land value patterns merits furtherdiscussion. As a large city grows, it acquires many new competing com-mercial centers that begin to rival the old CBD. These alternative (oradditional) economic centers, in turn, are strong motivating forces forthe decentralization of residential population.

We examined this process by looking at the land value peaks alongkey urban corridors in Bogota (see Villamizar 1981 for details of the pat-tern). The detailed analysis along ridges of the land value surfacereveals small hills, in accordance with the access characteristics that gowith the higher levels of economic activity in the developing subcentersof a rapidly growing city. Because of such developments, the gradientof land prices decreased as the city grew. The relative importance ofthe (:BD declined, and secondary gradients developed around the sub-centers. In general, as a city grows, the smooth land price surface cen-tered around the CBD develops wrinkles in the form of ridges, valleys,and small hills, as observed around Bogota's new subcenters.

Mills and Song (1979) found that in Korean cities commercial landvalues were always higher than residential land values in the (:BD and inthe rest of the city at points equidistant fiom the center. The evidencewe have found is consistent with their findinigs. Indeed, at equal dis-tances from the CBD, the proportion of area covered by commercialactivity in any neighborhood is a good predictor of land values in thatarea. These results are consistent with our expectations about accesscharacteristics of neighborhood and with the observed higher land val-ues in the richer areas of the city. Commercial activity locates in the richareas of the city.

The purpose of this section has been to illustrate the complexity thatis usually hidden behind regularities in city structure. We have shown,however, that these complexities are also the results of economic behav-ior on the part of households and firms-specifically the pattern ofhousing demand by households and the location decisions of firms-incombinationi with governmental decisions regarding zoning and infra-structure. Physical characteristics such as building heights and residen-tial and employment densities are appreciated better when behavior isunderstood; conversely, behavior is affected by the prevailing structuire.Thus, it is important to study both the overall structure of a city and thebehavior of its elements.

Summary

The growth of Bogota and Cali has been characterized by continuingdecentralization of populatiotn and, to some extent, of employment.

70 UNDERSTANDING THE DEVELOPING METROPOlIS

Although cities in developed countries have experienced a similar phe-nomenon, in that the concentration of population and employment incity centers has declined, there are some significant differences. Lowersuburban population densities in North American cities are mainlyattributable to the much larger size of residential lots in the sutburbs rel-ative to those in the central city. In Bogota and Cali residential lot sizeincreases only moderately with distance from the city center. Althoughboth the rich and poor have dispersed their residences, neither grouphas developed the large-lot, low-density suburbs typical of developedcountries. The poor have moved wherever there is available space, andthev continue to live in high densities at the periphery' as well. In BogotAand( Cali some peripheral densities are as high as central densities.

It is important to unlderstand how these highi peripheral densities areconsistent with a land( value strtzctture that confor-ms to the expectationsof a regular decline from the city center. The functioninig of the landmarket is such that land values at similar distances from the city centerare quite similar, although they are slightly higher in the richer areasbecause of the higher level of economic activity there. Nevertheless, thepoor have to pay rotghliV equivalent prices, and hence they substitutefor land by crowding. Larger poor households, in particular, have tolocate in crowded conditions at the periphery. The poor cannot substi-tute more capital for land because that would mnake the housin)g tooexpelnsive.

The high level and growtlh of land prices, often attributed to undesir-able speculation and monopoly ownership, are usually regarded asunlwarranted and hienice as villains in city growth. Impressions are oftenbased on faulty observation or inadequate analysis, however, as the anal-ysis of a uniique data set spannling the past quarter century of BogotA'sdevelopment suggests. A careful analysis of this data set indicates thatthe pattern of land values is consistent with the evolving density patternof residence and employment, with land values remaining nearly con-stant in the CBD and rising at the periphery. Land prices follow a regularpattern: they are highest within the CBD and decline as distance fromthe center incre ases. They are determinied much more by standardaccessibility considerations thani by availability and quality of infrastruc-ture. Lot size also increases price. Large lots in central locations are r areand are priced hiigher per unit of area than smaller lots similarlylocated. By contrast, large lots on the city periphery are lower in priceper Unlit of area than smaller lots similarly located. I.and use on theperiphery' is not as intensive, and large parcels are commoni, so largeparcels of land are offered for sale at a discount.

There was no explosive increase in real land prices; overall land val-ties in Bogota grew in step with the city's economic activity as measuredby this real product. This implies that the aggregate renlts from land

GROWTH AND THE CHANGING STRUCTURE OF BOGOTA AND ClAtl 71

remained a roughly constanit proportioni of Bogota's product and sug-gests that land has a unitary elasticity of substitution with other factorsof urban produiction. The indications in Bogota and Cali were that land,on average, had returns quite comparable to and not higher thanreturins to other assets. The land market in these cities should thereforebe judged as operating quite efficiently and as giving appropriate signalsfor the allocation of resotirces. Increases in land prices are much morethe result of increasing opportunlity costs resuIlting from growth andagglomerationi than of the market imperfection or "rampant specula-tionI" that is ustially thotight to characterize the operation of urbaln landmarkets.

These observations have a nutmber of significant implications for pol-icy approaches to expanding cities. First, the economic forces resultingfromii household and firm behavior are powerful in pushing city struc-ture toward r elatively well-tiuiderstood and predictable changes as acity's population grows. This suggests that planninig for land-use zoning,infrastructurc, transport, and other areas of urban development woulddo well to conform to these general trends rather than fight thern, asplanining authorities ofteni do. sometimes unwittingly and sometimesintentionally. Second, fast-rising land prices at the periphery shot-ld notbe a matter of concern; they corn e part and parcel with city growth, andartificial measures that attempt to stem suchi growth in prices will sel-domn succeed. Third, given the reqtuire ments of' an expanding urbanpopulation, the utmost muist be done to let the supply of' developedurbani land expand in responise to demand; this is the sur-est way ofkeeping land prices roughly consistent with the growth of the rcst of theeconlomy.

The decentr alizationi of employment in Bogota and Cali has been ledby the decentralizationi of' mantifacturing employment, as has happenedin cities in developed counitr-ies. Yet the CBDs continue to attract serviceemployment anid to retainl large retail and wholesale trade activity. InBogota the result has been that overall C:BD employment has remainedalmost conistant in absolute terms, although its share has been falling-to about one-sixth of all citv employmenit by the late 1970s. The centerhas become more specialized in its service and retail functions. Theanalysis indicates that decentralizatioll can be expected to continuebecause the tniderlyinig causes-increases in real income, urbani popula-tion growlth and transport improvemilenits-are also expected to con-tinue. This impor-tant feature of rapicd city growth is of'ten neglected inthe planniniig of uribanl services; for example, transport investments areofien based on forecasts of risinig rather thani constant employment in acity's CBD. The existinig radially oriented system, with transport primarilyfocused toward the city center, should be supplemenlted with circumfer-enltial services.

72 UNDERSTANDING THE DEVELOPING METROPOLIS

Although decentralization has many benefits, it also has accompany-ing costs. It requires more extensive utility networks and therefore addi-tional investment in public infrastructure. The per-household costs ofmanv of these services are inversely related to density. Because the netresidential densities of most new developments in Bogoti and Cali con-tinue to be high, infrastructure costs per household are not as high asthey might otherwise be. But the location of many poor householdsnear and on the periphery raises the costs of infrastructure extension,and these costs could be difficult to recover from the poor through usercharges.

Notes

1. The size of the comunas varies significantly, the smallest being about 174hectares and the largest twenty times that size at 3,680 hectares. The populationranges from about 25,000 to 400,000. See the appendix for a description of themain data sets used in the study and more details about the spatial classificationadopted.

2. The barrio, or neighborhood, is the smallest observation unit in the censuis.Because it is much smaller than a comuina, many more observations can be usedto estimate the density gradient functions.

III

Chapter 4

Poverty, Distribution of Income, and Growth

In studVying the growth of a couintr-V it is comm>on to look at the chang-ing distributioni ot incomiie to assess the clistributioll of the gains restilt-ing from growth)1 'Ae study the distribution of incomiie in Bogot.i in asimilar conitext but with one addled dimensioni-thiat of spatial dlistribu-tion withlinl the city. This is crucial for understanding thie spatial patterniof houisilig, tranispotatioll, an d public services and for projecting thefuture demand for thiese services. One problemii in studyring the spatialdistributioll of income is tihat thiert are few esial)lishedl patterns for mak-ing comparisonis and no r obtist theory for making predictions. The onllytheory that exists (for example, Minth 1969) hias grown otit of patternsobser-ved in cities in developed coulitnles and depend ls on two crticialasstinptiois. Thle first is iliat the incomile elasticity for the demand foirlhousing (n/,) he greater than 1; the second is that the elasticity of themar-ginial cost of n-ansport (as distance increases) withi respect to incomebe small relatiVe to f11.

lln ler these assullmptions. in a inonocenitric city in eqtuilibritim, onewoul]d expect higher-income houIseholds to live farther from the citycenter than lower-incomne hiotiselolds do. There is increasilig evidencethiat the incomile elasticitv for houisinig deinanid is less than 1 (see chapter7). Hence, althotigh the thieor-v explains the existing patternis of incomedistributioni in cities in clevelopedl counitries, its methodological basiscan be called into question. Moreover, in situationis where many tranis-port modes exist and tr-aflic is slow, the second assuilmptioll mav also ben1ot -uie for cities in poor coulitries. Chapter 3 established that there islittle evidence of svstematic variationi of income bv distance from the citvcenter- but that the pattern is rnuch more distinict bv radial sector. Thereis eviclence of sonie segregation in residence by incomie, whichi alsoaffects the distributioll of employmeit. This chapter is therefor-e neces-sarily exploratory andl primiarily descriptive, concenitratilig on e stablish-

7'

74 tUNDERSTANDIN(; THE DEVELOPINC. METROPOlIS

ing the existing patterns in Bogota and Cali and exploring usefulmethods of measurenmenit and description. It is (lifficult to generalizefrom these finclings, but it is hoped that similar measurement for othercities w-ill, over timne, establish recognizable patterns that are also expli-cable t'rom a theoretical point of view. Recognition of the patterns isnecessary for understanding how Bogota functionis.

There is considerable concer-ni about the existence of "excessive" pov-ertv in rapidlv growing cities. Sometimes this is expressed as concernabout the growth and characteristics of the informal sector (see, forexample, Mazulimdlar 1976; Sethuramani 1974; Gilbert and GLugler 1982;and WXebb 1977) and sometimes as concerin about the location of thepoor in unidesirable parts of cities and the interior coniditions uniderwhich they live. Much of the literature Supports the beliet that the qual-ity ot' life for the poor has worsened in large cities in developing coun-tries. There is also a wicdespr-eacl impression that "unemployment,underemployment and misemployment add up to a massive problem inthird world cities related to dimensions of ineqtiality" (Gilbert andGugler 1982, p. 70). Much normative evaluation has characterized thework that maniv of the lpoor do (particularly in the service sector, as(lomnestic servants, street vendors, and shoeshine bo's, for example) associally wasteful o1r unproduictive. The other most commoll concern iswith thle housing conditions of' the poor. E:arlier, sluLims and squatterr esettlemenits were regarded as evidence of a cultuIre of poverty; the newview, nIow wi(dely accepted, usually sees the incremental housing typicalof manyv poor settlements as evidence ot hope, innovationi, anid opportu-nitv. It is therefore illuminating to assess the actual existence of povertyil cities like Bogota and Cali and to confront impressions and beliefswith empirical evidenice.

This chapter illustrates some of' the diftficulties of even identifyinig thepoool; clespite a plethora of'data for these two cities. It then examinles thecorrelates ofl poverty (given plausible ranges of measuremiient errors).T'he interest is as much in measurement techniques and methods ofanalysis as in the descr-iption of the situation in Bogota and Cali per se.

T'he chaniges in income distributioni ancl povertv in Bogota during the1970s are better appreciated with some uLiderstandinig of the overalleconomic conditionIs in Colombia during this period ancl in the eco-nomnic preconiditions of recent decades. The late 1940s and early 1950sin Colombia saw rapidl industrialization and creation of' new industriesunider condlitionis of substanitial protection. It has been argued thatnonagricultural income distributioni probably worsenied fronm the mid-l930s until the early 1950s while this industrialization was taking place(Berry and Uriutia 1976). Manv of the protectecl indutstries thatemerged were relatively capital-intensive and( not labor-intenisive. Giventhe skewed and highlv concenitratecd ownership of capital, such industri-

POVERTY, DISTRIBUTION OF INC(OME, AND GROWTH 75

alization could be expected to lead to a lower share of labor earnings.Because rural-urban migration was also very high during this period,providing almost "unlimited supplies of labor" at the low end, it waspossible to keep blue-collar wages down. Conversely, this burst of indus-trialization created a sudden demand for highly skilled and educatedworkers, whose real earnings rose rapidly

The whole period from the 1930s to the mid-1950s could be describedas involving major structutral changes in the Colombiani economy-inparticular, a shift from a predominantlv agricultural rural economy toone based on urban industries and services. These structural changescontintued apace until the mid-1960s, but the latter part of the period,frorm the mid-1950s to the mid-1960s, witnessed dramatic increases inblue-collar wages. Berry and UIrrutia hypothesized that during thisperiod the degree of pi-otected import substitution may have decreased,and small-scale competitive enterprises may have become more success-ful. The expansioni of primary educatioin in the 1950s also contributedto higher average levels of workers' skills. The period between the mid-1950s and mid-1960s was one of slower economic growth and was inmany ways a period of consolidation. Surprisingly, lower-skilled workers'incomes rose rapidly, even though ruial-urbanl migration continuied athigh rates.

The overall income distribution in Colombia has remained hiighlyunequal and is probably among the most inieqtual in the world. Theshare of the bottom 40 percenit has remained aroundicl I0 percenit of totalincome, whereas the share of the top 5 percent has been about one-third. 2 This is a little worse than in some comparable middle-incomecountries. In Malaysia, for exanmple, the shar-e of the bottom 40 percentwas estimated to be 12 to 18 percent and that of the top 5 percent was alittle less than 30 percent (Anand 1983). The Gini coefficient forColombia is consistenitly estimated to be above 0.5. For comparison, indeveloped couLtties the share of the bottomii 40 percent of householdsis typically estimated to be about 15 to 20 percent and that of the top20 percernt, between 35 and 45 percent of total income, with ;inicoefficients of between 0.3 andl 0.4. OUI estimates, which are only forBogota and Cali between 1973 and 1978. suggest some improvement inthe late 1970s. The poverty estimates corroborate this tendency, so thereprobably was some improvement in income distribution in Colombiaas a whole.

If such improvement did occur, it would be consistent with a numberof other developments in the Colombian economy during this period.With declines in fertility rates and the progressivelv smaller proportionof people left in the rural areas, the rates of rural-uri-bani migration andof urbanization have slowed. Moreover, there is evidence that the wagesof rural unskilled workers rose significantlv in the late 1970s (ULr--titia

76 UNDF:RSTANDING THE: DEVELOPING METROI'OLIS

1985). The Colombian agricultural economy benefitted greatly fromrelatively high coffee prices throughout the 1970s and particularly fi-om1975, wheni a booni began thiat lasted until almost the end of thedecade. Overall growth in agriculture was also consistently hIigh overthis decade, with an average anntual growthi rate of 5 percent (see Sokoland other-s 1984). Thomas (1985) has forcefully argued that thesechanges were not entirely fortuitous but rather resulted from goodmacroeconiomilic management along with progressive policies that sup-ported agricuiltural development. Ihe net result was that the demo-graphiic trailsitioln takinlg place in Colombia in the 1970s, coupled withirobust growth in the agriculture sector, proclucedI a relatively tight r urallabor market. This in turn applied pressure oni the urbani labor market,and real wages rose at the same time as uLbani employmelnt expandedsubstanitially.

Corr-espondingly, it appears tilat the noniagr-icultural sectors were alsofollowing a labor-usinig strategy that lecl to a i-ate of growth of emplov-mnent that exceedled output growth in the 1970s (Sokol andc othe rs1984). Again, apart from fortuitous circurmstances, this success can beattributed to successftil fiscal and u-ade policies that promoted labor-UsinIg manitifactur-ed exports. The extensive expansioni of uLbaln publicSerVices inlVolinig the public sector also contribtItecl significaltly to theexpansioni of demand for labor. A malor result of the expanded demanicdfoir labor was thie increased participation of unskilled womeni in thelabor force, whichi increased the houiseholdl income of poorel hlouse-hIoldCs (see Mohani 1986, UIrrtia 1985, and chapter 5 of this book for*ietails).

Distribution of Income in Bogota and Cali, 1973-78

An assessment of inlcomite distributiOnl is typically plagued by measure-ment problems. Most censuses in developing counitries do not iicilidean incomile question, and those that do merit considerable skepticismabout the veracity of responses. Household surveys are usually regardedas monre accur-ate because they are more carefulyv done, with smallersamples. Rural income surveys presenlt greater problems than urbansurveys because it is difficult to evaluate both outputs and inputs. Inuribani su-vevs, wage earniligs are easier to evaliate than income fromproperty. Thie maini sources of data that were used in this study were thie1973 populatiion census for Bogota and Cali and household surveys in1975, 1977, and 1978 conducted by DANE, the Colombiani nationial statis-tical agency. Tle 1978 survty was conclucted especially for the CityStudy, and the incomie questioni was more detailed than in otlier sur-vevs. The 1973 census seemed to have covered only about 5() percent of

POVER-FY, DISTRIBUTION OF INCOMt, AND) GROW [ft 77

total personal income, whereas the 1978 survey may have covered asmuch as 90 percernt and the other sur-veys perhaps 60 to 65 percent.

Although it is difficult to assess accurately the extent of undercover-age for each income group ancl for different years, oui overall judgmtentis that coverage is probably' better at the low end than at the top end of'the income (listribuitiol. The evidence was conflictinig, but the earningsof the least-skilled occupationial categories recorded in thc 1973 censuswere broaclv compar-able with othier data onl the labor earnings of thesecategories of workers. Female par ticipation in the labor force, however,may have been unidercounited in the 1973 census, contributing to someunderestimationi of hiousehold incomiles. Overall, although thler wassubstanltial unidercoverage of'incoine in the 1973 population census, thedislrzbulion of income may be regarded as reliable and comparable xvithother clata sets (see Mohiani 1984 and Mohan, Wagner, and Garcia 1981)lHenice, despite problems with the comparability of' differenlt dlata setsover time, it is possible to conclude from the evidenice that the (listribui-tion of income improved marginally over the late 1970s in Bogota andCali. These details illustrate typical issues that are encountered in esti-mating income distributioll trends in just two cities that are daLa-rich incomparison with most others in developing countries.

Table 4-1, which gives the decile shares of income for 1973 and 1978for Bogota and Cali ancl the valties of' two indexes of inequialilv-theGinii coefficient and the Theil index 4 -reflects the strikinig inequalitythat exists in C'olombia. The income share of the bottom 40 percenlt ofhouseholcis is less thani half the share received by the top 5 pelceIt.Even in 1978, wheni the income coverage was greater than 90 percent, itappeared that there was substanitial unider-cover-age of incomiie ftromil cap-ital. Since sulch income is likely to accrue to the richer houselholds, theactual distributioni would be eveni worse than that apparent fiom thistable. The results are similar for Bogotli and Cali andc for the dif'ferentinlconmei measures and rankinigs.?

The overall improvemenit in distributioni between 1973 and 1978 isevident. These results were also consistent for differenit incomile con-cepts, different ranking proceduires, and interimlediate years (1975 and1977) for both cities,C implving a high degree of confidlenice in theserestilts. One general restilt obtained but not repolrted here was that allthe indexes tisecl yield highel inequality of' househlolds when ranked byHINCAP. This implies that even though households of' larger size tend tohave higher total hotiselhold incorime, their per capita houisehold inicometends to be lower.

In sumimary, the levels of overall inequality foundl in Bogota and Caliare not substantially different f'rom earlier estimates for ulrban Colombia,although a slight tendenicy towaid improvement is observed in the late1970s. The distributioni of incomile is slightly worse in BogotA thani in Cali.

78 UNl)DERSIAND)IN(G THF IlEVELOPIN(; MElTROPOLtS

Table 4-1. Distribution of Income in Bogoti and Cali)peicentage shares of tota] incomyne)

B&go1a (.li

Categorv 197 3 1978 1973 1978

Moo seholds 7rnked by hmoseh(ldincotne

Bottorm 2(0 percent 3.3 4.( 3.4 4.1Bottom 4() percenit 9.8 11.9 1 0.4 1 2.4Tol) 20 pecent 62.5 55.5 60.0 54.)Top 5 percent 30.2 25.1 31.7 25.0

Gini coefficient 0.565 0.507 0. 53 0.487Theil in(lex 0.610 0.458 0.601 0.429

Individualt rainked b1, hauseholdincenne per. capila

Bottom 20 peiceit 3.3 4.0 3.4 4.2Bottom 40 perceIt 9).9 11.5 10.5 12.4Top 20 petrcernt 62.6 38.0 60.5 55.5Top -percent 30.8 29.0 31.4 27.3

Gini coefficieitit 0.568 0.522 0.554 0.500Thleil inidcx 0.649 0.508 0.632 0.470

Scmore: 1973 po)ullatio n censu s: 1978 Wcoild Bank-[)\NlN Fliousehol(i Silr ,ev.

We now addlress the issue of the distribution of hIcom11e across space inBogotai and Cali. What arc the spatial patterns? Hlow do We measurethemil svsternaticalvxS Can we trace trenids over timne? Boith Bogotui. andCali exhibit striking patterns of incoine segregationi by space. The ratiobetween meani incomes in different zotnes of the city is as miuch as I to 6.with the means taken over relatively large zones. This kind of spatial dis-parity in incomes is tuniusuial even among the regions of' a countrv, butbecause there is little intformationi of' this type for other cities, it is cliffi-cult to sa' what the nor-mii wouild be for spatial distribuitioni withinl a city.Discussioln of the extent of spatial inequality within a city and compari-son witli others are diffictilt because there are n(o naittiral units of'aialyv-sis. UInlike the world, wlichi can be divided into cotintries, or a country,whicil can be divided into states or regions. there are no natural divi-sions within a city. In the LUniteci States much cdisciission1 focuIses on thecontr asting chara(teristics of ce ntr-al cities and sibtb-hs, btit even in thiscase there is no natural definiitioni of' what constittites a centr-al city. Anydivision of a citv is arbitrary.

In this stttdv we contiitte to uise the two systems of zonificatioti intro-duced in chapter 3-that is., riigs and sectors. Muich utrban econoimictheory for a nionocentric citv wotild stiggest s'steimatic differences iy

POVERTY, DISTRIBUTION OF INCOME, ANI) (;ROW)UTH 7)

(listanice fiomi) the city center-that is, by rings-but thlere is also arespectable traclitiol), or-iginiatinig with Homer Hovt (1939, 1966), thatdivides the city into pie slices (sectors) for studyinlg the distribution ofincome. Hoyt traced the historical developmenit of a large number ofNorth Americin cities. He concluded that income grotips tended tolocate in one section of the city and that as the city developed. the richcontinlued to locate in the same dlirectioni as that in whicih the citv wasexpaniliig. Peter Amato (1968) establislhed a similar patterin for Bogota:he focttsedl on the elite and shiowecd that they have tended to locate pre-dominantly in the noirthern part of the city and have continied to m<ovein that clirection. In later articles Amnato (1970a, I9701)) notecd patterilsin QuLitc), Lima. and Santiago that were broadly similar to those inBogota, althotigih not always as distinct.

In this exploratory section we suggest the use of a more systematicmethocd of inequality measur-emiteint to assess the extent of spatial ine-qtialitv and its trenid over time. As indlicatedl in table 4-2, in 1978 mealnhotisehiold inconie per capita in the poorest sector (2) in Bogota wasabotit one-fifth of that in the richest sector (8). In (ali the poorest sec-tor (4) hiad about one-qtuarter the mean hotiselioldi income of thie rich-est (2). Some anomiialies may be note(d in the data for Cali. 'The data lortihe (CBD sector in 1978 seems to have suffered from some sample errol's:populationi is shown to have fallen from about 4.3 percent of total popu-

lation in 1973 to only 1 1 percent in 1978, whereas meani houselholdincome as a proportioni of the overall mean increased froml about 1.03to 1.78. This is not plausible. The data for Cali do suggest. however, thatthe deceiiti-alizationi process for ricih houselholds had just begun in the1970s. Many rich hoUseholds contilued to live within or near the cen-ti-al city. The movemenit of households from the nich sector- 2. locatednear the central city, to sector 6, a developing stubulbhan residenitial area.was just starting. OuL- observationis in the late 1970s wer-e made over thisperiod of rapid transformation. Cali, being a muchi smaller citv thianiBogota, still had a long way to go in terms of growth an(d decenti'aliza-tioi. althoughi the process of subtjrbaniizationi of tlhe rich had begun.Overall, however, the patterns for BogotA andl Cali wer-e similar in termsof'segregationi of'tlhe rich and poor.

Holow nicih is hiddeni behind the meanis in table 4-2?) Are there largevariances aroulnd these means, so that the means do not necessarilyimply clifterinig characteristics in the sectors? The measureilelm t of ine-quality is essentially the measur-em-Dent of varianc(e in a distribution. Whiatwe now seek to do is to clecoliupose this variance i nto two parts: tile vari-ance between the means of incomile in rings, or sectors, or any othergrouping of people and the varianice in incormes between people within

a group. The Tlheil index cani be decomposed in this way. Total ineqLial-ity can be clecom posed in to wvithi n-group (or withinl-sector, within-ring)

Table 4-2. Spatial Distribution of Income in Bogota and Cali

Bogota Cali

1973 1978 1973 1978

Share of sertor in Share of sector in Share of sector in Share of sect o in

cily population city population city population ditY pofrulation

Sertor (percent) Mean Hr\'cAP' (percent) Mean HInCAP' (percent) Mean HIN(,AP' (percen t) Alean HIM.-AP'

1 ^2.3 1.07 2.4 0.80 4.3 1.03 1.1 1.782 18.2 0.57 20.6 0.50 4.2 3.68 4.8 2.36

3 26.0 (.68 25.2 0.73 19.5 0.89 15.2 0.88

4 9.3 (.90 7.2 0.98 21.2 0.69 17.8 0.64

5 7.3 0.98 6.2 1.03 35.1 0.68 42.2 0.736 18.2 0.80 20.0 0.91 11.1 1.61 12.8 1.42

7 12.0 1.71 9.1 1.44 4.5 1.37 6.0 2.138 6.6 2.87 9.1 2.60 n.a. n.a. n.a. n.a.

Cit"v 100.0 697" 100.0 1,012) 100.0 594b 100.0 2,425b

ni.a. Not applicable.a. Mean household mo,ithlv incomile per capita taken across individuals in the sector as a multiple of overall mean HINEAP.

b. Mean household monthiv income pet capita taken across individuials in Bogoti or Cali, expressed in current pesos. Consumer price index (1970 - 100):1973 = 150; 1978 = 400.

Stnirc: 1973 population census; 1978 World Bank-DANE hlousehold Survey.

POVERTY, DISTRIBUTION OF IN(:OME, AND GROWTH 81

contribtition and between-group (or between-sector, between-ring) con-tributionis. The latter can be regarded as an index of spatial inequality inincomes.'

The Theil inequality indexes in table 4-3 show the level of inequalitywithin each sector of the two cities. In general, the Theil index for eachsector is lower tlhani for the city as a whole; that is, inequality within eachsector is less than overall inequality (see table 4-3). The lowest inequalityis found within the poor sectors (sectors 2 and 3 in Bogota and sector 4in Cali). Similarly, in 1978 the richest sectors (sector 8 in Bogota andsector 2 in Cali) also have low levels of inequalitv. This is a clear indica-tion that the rich and pooI parts of both cities became more homoge-neous over the 197 0s. The degree of spatial income segregationincreased during that period, and the rich and poor areas of the citybecame more recognizable as suchI.

The lower part of table 4-3 presents measures of the betweeni-grotipcontribution of the sectors and rings. The decompositioni of inequalitypermitted a calculationi of this between-group measure, which has beeninterpreted as a measure of the spatial inequalitv in income within a city.The spatial contributioll of groupings by sector is substantial-about 20to 25 percent of total inequality. Because each sector is distinct andinterinally homogeneous, spatial inequality between sectors is high inboth cities. A low spatial contribution (4-5 percent) by rings confirmsthat the level of ineqLialitv within rings is not very different fi-omil overallincome inequality within the whole city (see Mohan 1986). Each ring isabout as heterogeneouis as the whole citv. For rorntnas, shown in the lastrow of the table, spatial inequiality is between 32 and 40 percent, show-ing that these smaller spatial units are even more distinct in character.

These systematic measures confirm once again that there is no dis-cernible pattern in household income according to distance from thecentral city in either Bogota or Cali-as is often observed in cities indeveloped countries. Income groups are significantly segregated by sec-tor. Like households have been gravitating toward each other in differ-ent parts of the city. The metho(dology used here suggests a systematicmeasuire of spatial inequality that can be used to make similar measure-ments in other cities.

It is difficult to judge the magnitude of such spatial inequalitv. Is 20-25 percent spatial inequality low or high? I have seen no similar indexesfor other cities. Anand's (1984) work on Malaysia, however, providessome estimates for comparison. He foulid interstate and rural-urbancontributions to inequality in Malaysia to be 9.1 and 13.7, respectively,accordinig to the Theil index. By these standards, Bogota and Cali mustbe judged to be highly segregated spatially by income. Moreover, thisinequality is increasing, despite evidence of some overall decrease inincome inequiality. If spatial segregationi of income groups has long-

Table 4-3. Spatial Inequality in Bogota and Cali: Individuals Ranked by HINCAP

Bogoli CIali

Ina,me 1973 1978 Income 1973 1978rankingl ranking,

Item (197,Y) 7heil index Rankingh Theil index Rankingb (197,8) Theil index Ranking" Theil index Ranking0

Sec/or1 3 0.635 8 0.458 8 5 0.573 5 0.549 72 1 0.429 3 0.286 1 7 0.368 2 0.257 23 2 0.390 1 0.309 2 3 0.429 3 0.315 34 5 0.407 2 0.362 4 1 0.328 1 0.251 15 6 0.553 5 0.418 5 2 0.435 4 0.337 46 4 0.514 4 0.426 6 4 0.741 6 0.495 67 7 (.572 6 0.444 7 6 0.781 7 0.465 58 8 0.602 7 0.357 3 n.a. n.a. n.a. n.a. n.a.

(Citly 0.649 0.508 0.632 0.470

Intergroup contibutionto inequalily (percent)

Sectors 21.6 26.9 22.7 22.5Rings - 6.0 - 4.0(Comunas 32.6 37.8 37.3 40.6

n.a. Not applicable.- Not available.Vote: HINcLXP = houisehold income per capita.a. Ranked in ascending order of meani HiNcAP.

h. Ranked in ascending order of inequality.Soturce: Mohan (1986). tables 4-9. 4-10, and 4-11.

POVERTY. DISTRIBUTION OF INCOME. AND (;ROW(TH 8'$

term deleterious effects on the efficient and harmonious working of acity, the trend in Bogota is a cause for concern. This kind of segregationcan lead to increasing distance between the different social groups of acity and impede social mobility.

As an example, the problem of educational segregation has receivedgreat attention in the United States. One solution attempted there hasbeen to bus children to school across neighborhoods, particularly toprovide better opportunities to children coming from spatially dis-advantaged neighlborhoods. Busing caused considerable social tension.If significanit spatial differences start emerging in cities in developingcountries, social problems similar to those experienced in developedcountries may well arise. For example, spatial separation, by redticinigthe personal contacts that are so important in transmitting infornmationabout the labor market, could contribute to labor market segmentation.

If similar studies are carried out in many cities, it may also be possibleto identify the reasons behind higher and lower levels of spatial differ-entiation in cities and their effects on labor markets and social welfareas a whole. For example, are such phenomena caused by highi rates ofgrowth where similar people congr-egate near each other to begin with?Over time does fur ther sorting take place and the city become less a col-lection of differenitiated neighibor-hoods and mor-e like an organicwhole? Or are such tendenicies exacerbated as the city stabilizes.

Clearly, whatever the answers to these larger questions, the spatial dis-tribution of income in Bogota and Cali does not conform to expecta-tions from the theory for largely monocentric cities, despite regular- andexpected patterns of population densities and land valtues. It is there-fore importanit to study cities in greater spatial detail even if thiey con-form to overall recognizable patterns.

Distribution and Characteristics of the Poor

To identify the correlates and composition of poverty in Bogota andCali, we have to address the problem of how to define and measure pov-erty. Despite the high quality and large quanitity of data at hand, it is pos-sible to identify only the people who must, in all probability, be poor.And it is difficult to identify all the people who are poor, except withinrather large ranges of estimates. It is mnuch easier to identify relative pov-erty, defined merely as the bottom x percent of the population-say, 30percent. We adopt a mixture of these approaches. A lower bound is firstplaced on the proportion of people who must be poor. We can thentrace the movtmenit of this proportion over time, compute a range ofestimates of others who might be poor, and, finally, collate the charac-

RI UNDFRSTANI)ING THE DEVELOPING METROPOLIS

teristics of the bottom 30 percent, who roughly coincide with those whohave been identified as poor.

The notion of absolute poverty has been linked with that of malnutri-tioIi. BecaLuse food is the first necessity of household consumption, ifpeople are malnourished, they can scarcely be consuming other things.Estimates of malniutritioi and poverty have therefor-e been made on thisbasis for countries (for example. Ohja 1970 and Dandekar and Rath1971 tOr India) and for the world as a whole (for example, Reutlinigerand Selowskv 1976). In these studies, estimates of minimum expendi-tures on food required for adequate nutrition were compared withestimates of actual consumption by income or consumption groups.Although nutr-ition involves the intake of various elemenits such as calo-ries, Aitaminls, proteins, and minerals, it is generally agreed that if' con-sumption of calories (Ihat is, energy) is adequate, other elements arealso usually adequately consunied (see. for example, Sukhatme 1978).This, if we know the caloric content of foods, the usual food basket foidifferent income groups, the prices of foods, the expenditure on foodby different types of families. and the share of food in total consump-tion or total income, we can relate the likely level of nutrition to incomelevels.

Thle key problem in uSillg such a procedtire is that each individual's

norm is different; some individuals, for example, need only 1,800 calo-ries a day, while others need 2,500. The minimum caloric consumptionis arounld the usual requirement levels. A small percentage change inthe minimuim caloric consumption used as the basis for estimating thenumber of malnourished people can lead to substantial differences inthe estimate of malnourished people. For Colombia, the InstitutoColombiano de Bienestar Familiar (Colombian Institute for Family Wel-fare) has established 1,970 calories as the average minimum for ade-quate daily caloric intake, taking into accounlt Colombian conclitionisand the age and sex distributioni of the population.

Using the same data sour-ces as earlier (the 1973 populatioll censusand the 1978 World Bank-DANE Househol(d Survey), we attempted tocalculate the proportion of people in Bogota and Cali who can in allprobability be classified as malnourished and therefore poor. Becausewe had only incomne data from these sources, we used other informationto translate the income level of each household into likelv calories con-sumed per person, taking into account the age and sex composition ofthe household and its likelv consumption pattern. Even if it is assumedthat anyonie consuming less than 80 percent of reqtiired mean caloriesmust be malnourished, different methods gave Us a likely range of 30 to50 percenit malh-ntrition in Bogota and Cali in 1973 and about 15 to 25percent in 1978. One problematic issue was the undercoverage ofincomes in 1973 alluded to in the last section. Anothcr issue is that low-

POVERTY. DISTRIBITTION OF INCOMF. AND (GROWTH 85

income people typically eat foods that are calorie-rich; that is, the priceof each caloric is lower for many cheap food items typically consiumledby the poor. Taking thiat inito account, we conclided that otir estimate ofthe share of the populationi that must be malnouirished could be nar-rowedl downi to about 25 to 30 percent in 1973 and 12 to 15 percent in1978. These are the people whose incomes do not allow for adequatenutrition. There mav well be additional numbers of people who haveadeqtiate incomes but suffer firom manititritioni fbr other reasons.

Clearly, identifying poverty is not a simple task even with detaileddata. Table 4-4 provides a quanititative ide a of the kind of misclassifica-tion that can result if we take, say, the bottomil 30 percent of the incomnedistributioni as poor. The first row gives the breakdown, by decile, ofpeople who constime less thian 80 percent of reqtuired calor-ies (0.8R).Virtuallv all the malinotirished, so defined, fall within the bottom 3() per-cent. The next three rows give the percenitage in each decile that fallsinto each nutritional range-less thani 0.8R, between O0Rand 1.OR, andmore than 1.LOR If we defined all the bottom 30 percent as poor, wewotild misclassify as poor as many as 32 percent in the secon d decile and58 percent in the third decile who are not necessar-ily malnour-ished(with 80 percenit of requir-emenits as the norm). Conversely, it is onl, inthe top five deciles that we can be sur-e there are no maliourished orpoor. As the table shows, household size declines monotonically withincome, suggestinig that the malinouLishied are predoininiantly in largefamilies with high dependency ratios. 8

It is useful to compar-e the minimum wage ( 1,738 pesos in 1978) wsiththe average incomies presented in the last two rowss of the table. Becausethe imajo rit of the second decile and part of the third decile are mal-nourished, an intermediate lhousthold incormie level in the range of4,000 to 5,000 pesos (1978) can be utilized as necessary to attain ITini-mumit adequate nutrition. It is clear, tilen, that for adequate nutritionthose households need roughly 2.5 workers per houiseholdl at millimuLm1wages. In fact, thev characteristically have higher dependency ratios;that is, they have fewer than 2.5 worker-s per household. It would seemthat the minimiiumit wage does not provide even a small family withenough income for the barest minimnum of nutrition. The monthlywages of fully employed people in typical low-income occupations-constr-uction workers, cooks, waiters, and so on-were betA-een 2.0 arid2.5 times the miniimum wage in 1973 and 1978. Moreover, monthlywages hiad growin at an average of 8 to 9 percent a year between these*gears. It Would he dlifficult, therefore, to argue that mininuim wageswere too high durinig that period or that they contribilted to protectingthe forimial sector.

In conclusion, our r esults are mixed. First, wc must be wary of malnu-tritioni or poverty estimates reached( with the use of simple income cut-

Table 44. Mapping Malnutrition into Income Deciles in BogotA, 1978

Decile"

them 1 2 3 4 5 6 / 8 9 10 Total

Percenitagc of O.8Rin decile 46 32 20 1 1 0 0 0 0 0 100

Percentage of decile innutritionial category

0.8R 100 68 42 1.7 2.5 0 0 0 0 210.8Rto l.OR 0 28 35 40 16 3 0 0 0 0 12L.(R+ 0 4 23 58 82 97 100 100 100 1()( 67Total 100 1(0 100 100 100 1(10 100 100 100 1(0 100

Percentage of'populationiin decile 9.8 10.0 9.8 9.7 10.0 10.4 9.8 10.1 10.3 10.3 100

Average househiold size(personis) 6.1 5.5 5.4 5.3 5.2 4.8 4.6 4.3 4.1 3.5 4.8

Average monthly householdinconie per capita(1978 pesos) 440 720 925 1,200 1,470 1,818 2.30() 3,110 4,780 12,640 2,980

Average monthly householdincome (1978 pesos) 2,580 3,980 5,050 6,340 7,680 8,710 10,675 13,400 19,670 51,790 18,140

Note: O.8R riefers to people who consumiie less than 80 per-cent of the minitiiim nutrition requiremenits. I OR r efers to people who consume less than the min-imum requiremenits. [.OR+ refers to those who meet or exceed the minimum rlequiremelits.

a. The population (3.5 millioni) is ranked from poorest decile (1) to richest (10) accordinig to household income per capita.b. The minimum monthly wage in 1978 was 1,738 pesos.Source. Mhan (1986), table 5-7.

POV'FRTY, DISTRIBUTION OF INCOME. AND GROWTH 87

off levels. Second, one can get a sense of the likely ranges of malnutritionand poverty through rather detailed work. For 1978, for example, thereis almost zero probability that any malnourished people are in the topfive deciles, a small probability that they are in the fourth and fifthdeciles, and a high probability that they are in the bottom 30 percent.Taking the bottom 30 percent in this case leaves out the few who are inhigher deciles but does include some who are not necessarily malnour-ished in the bottom three. Third, there is a high likelihood that the def-initelv malnourished and poor decreased from about 25 to 30 percentin 1973 to 12 to 15 percent of the population in Bogoti and Cali in 1978.

Having found reasonable estimates of people who must be poor, wenow need to identify their demographic. spatial, and economic charac-teristics. Tlhe concenltration of the pooI in particular identifiable groupsaccording to each of these dimensions is notable. The incidence of mal-nutrition is strikingly high among children betwveen the ages of 5 and 14in large households of more than six members, a high proportioni ofwhich are located on the peripher-y of the two cities and in the particu-larlv poor sectors (in the south in Bogota and the east in Cali). (ForBogota, see table 4-5; the situation in Cali is similar.) An index ofbetween 140 and 150 for childr-en ages 5 to 14 in sector 2 means that thechildren are 1.4 to 1.5 times more likely to be malnourished than thenmean of about 28 to 35 percent of this group. Note that this mean isitself much hiigher than the overall mean of 21.1 percenit.

A few other features stand out from this table. Although, on average,the old are less likely to be malnourished and poor, those on the periph-ery (ring 5) and in the southern part of Bogota (sector 2) are excep-tions and disproportionately so. The children in these areas are notparticularly hard hit, but the adults are worse off compared with the cityaverage. Since malnutrition is accentuated by bad sanitation, poorhygiene, disease, and crowding, the purely income-based estimates ofmalnutrition are more likely to be understated for the pooI rieighbor-hoods. There, even people with somewhat higher- incomes who canafford more food could suffer from malnutrition because of chronicstomach disorders and other neighborhood effects. 9

We can now bring together a number of results that are important forunderstanding the structure of Bogota. We have pointed out in chapter3 that densities in the poorer parts of Bogota and Cali are relatively uni-form and do not decline with distance, even though land values do. Wenow find that many of the households on the distant periphery arelarge, poor households with high dependency ratios; that is, householdshave responded to lower land prices and located at the periphery; butbecause of higher household size, poptilation densito continties to behiglh in these areas. It would seem that poor households wvith moreolder people incapable of working tend to descend furthe r into poverty.

Table 4-5. Spatial Distribution of Malnutrition by Age Group in Bogota, 1978

Percentage

malnmiLlished

'Age group bidex of mia/nurition' in age grollp

Ring

1 2 3 4 5 6

0-4 160 92 54 77 117 101 19.7

5-9 28 76 77 81 120 112 28.4

10-14 126 82 76 86 112 98 34.1

15-54 77 65 45 84 133 9 2 17.7

55+ 131 97 51 68 140 88 15.4

Total 101 73 55 82 127 98 21.1

Sector

1 2 3 4 5 6 8 8

0-4 160 147 94 54 81 103 75 40 19.7

5-9 28 144 98 73 55 96 70 77 28.4

10-14 126 147 95 88 6i7 94 70 58 34.1

15-54 77 156 97 85 71 120 57 29 17.7

55+ 131 188 108 83 141 84 51 23 15.4

Fotal 101 155 98 80 72 14 61 38 21.1

a. Tlhe index of Tialnutrition is calculated as:Ai = Percentave 0.8R in age group i in sector

Percentage 0.8R in age group iwhere 0.8R refers to people who ConsuLie less tiant 80 percent of the minifmljnu nut itional requiirenilents set bs the Institiuto Colombiano de Bienestar Famiiiliar.

Source: Mohan (1986). table 5-12.

POVERIY, DISTRIBLITION OF INCOME, AN)D GROWTH 89

This tendency is exacerbated for income earners with relatively flat age-earnings profiles. As is well known (and documented for Bogota inchapter 5), people who are poorly educated and unskilled do ncot ear-nmore with experience. As these workers get older and enter the stage ofhousehold formation, tlhey have to support more people on the sameinicome. Children of the unskilled are therefore disproportionatelypoor on the whole. The situation is exacerbated when old people arepresent in the household. It is interesting to observe how this processworks itself out and how spatial effects can be identified in terms ofresidenice patterns. Although economic growth in the mid- to late 1970swas quite effective in reducing the number of people who have no poten-tial of achieving minimally adequate nutrition, the concentration ofthe poor in particular areas remains striking. The concentration does,however, allow easier targeting of antipoverty, nitrition, and educaCtion

programs.Having establishecl the demographic and spatial dimensions of pov-

erty, we now need to look in a little more detail at who the workers arewho happen to be poor. For this purpose, poor workers are identified asbeing in the bottom 30 percent of households ranked bv HINCAP.

The very fact of working does much to alleviate poverty. Onilv about20 percent of male workers and 15 percent of female workers fall intothe bottom 30 percent. Mfost women are secondary workers (that is, nothousehold heads). A household is better off if there is more thani oneincome eariner. As might be expected, almost half of ial male workerswith no education and about one-third with a primary education areclassified as poor. An analysis of poverty by age an-d educationl groupsconfirms that childreni suffer disproportioniately from poverty. Amonigboth males and females, workers in the 35-44 age group are mor-e likelvto be poor. Thus, whereas about 20 percenlt of all male workers ar e clas-sified as poor, more than one-third of the 35-44 age group is so classi-fied. Most of these are workers with little or no education, but about 15percent have a secondarv-level education. Workers with little eduicationhave relativelv flat age-income profiles (see chapter 5), and it is theseworkers who appear to fall into poverty during the period in their liveswhen their households are expanding and their children are going tosclhool. This highlights the real economic hardship unclergonie by thepoor in educating their children into their tcens rather than sendingthein out to work. Of those in the 12-14 age group who work (and thesample of this group is not large), very few are founid in poverty. Thosewho wor-k clearly contribute to houschold income and help lift thehousehold out of poverty.

Poor workers are concentrated among a few occupations and corre-sponding industries. Blue-collar occupations accoun1t for about 70 to 75percent of poor workers but only 55 to 57 percent of all workers. Produc-

4(! UNDERSTANDINO THE DEVELOPING METROPOLIS

tion, constructiorn, and transport workers are particularly likely tobe poor; along with service workers, these groups make up about 70 per-cent of all poor male workers. Among women, it is notable that maids(domestics), who account for about 20 percent of the female labor force,are not particularly poor. Almost one-third of poor female workers areservice workers (excluding maids). Among the production workers, thetraditional industries of textiles and footwear, lumber and wood prod-uicts, an(l printing and publishing account for the majority of the poorworkers. Female participation in productioni activities is mainly in thetextile and footwear industries, which do not seem to pay women well.

Do poor workers work long hotlrs or are they short of work? Poormen are strikingly overrepresented among workers who work longhoturs; about 20 percenit of poor, male workers are in this category. Onlyabout 8 percent of the poor wvork fewer than forty hours a week, andabout 55 to 60 percent work the "normal" forty- to forty-eight-hourweek. Sales, construction, and transport workers wvork the longesthours; among the poor, only professional and technical workers workrelatively short hoturs. The storv is different for female workers; theyappear to be overrepresented among those working shorter hours. (It ismainlv sales workers and maidls who work very long hours.) It seems thatmany poor, female, part-time worker-s wotuld like more work but find itdifficult to get or else cannot work longer hour s because of responsibili-ties at home.

Overall, it appears that pool male workers are not lacking for work.There is little overt unideremployment; indeed, if anythinig, the poor areoverworked but are locked into low-productivity occupatioas. implyinga need for traininig programiis to improve people's skills. The sales work-ers many actually be under-employed-that is. idle for a substanitial part oftheir workinig hour-s-but they comprise only about 25 percenit of allpoor workers. Almost half of the poor. overworked men are transport orconstruction workers. It is unlikely that they are idle during much oftheir woi-kinig day. On the other hand, a substanitial portion of thewomen are either genuliniely underemployed (willing and able to workmore hours) or uLnable to work longer hours. Poverty will therefore notbe alleviatecd by mere expansion of employment opportunities. Poorwomein need greater employment opportunities, but they could alsobenefit from day-care progr-anms-which themselves generate emplov-ment for the providers.

It is often said that recent migrants earn less than others and have totake very low-paying jobs while waiting for better-paid opportunities.The evidence is mixed. In Bogota in 1975, 22 percent of the most recentarrivals earned less than the minimum wage, compared with about 38percent of all workers. Among those who had been in Bogota elevenand twenty years, however, as many as 20 percent earned less than mini-

l'OVER-TY, DISTRIBUJTION OF INCOME. ANI) (GROwTrH 911

mum wage. It appears that low wages are spread fairly evenly amongworkers of all vintages and that experience in the city does not guaran-tee a highelr wage. On the whole, it is clear that recent migrants have nomonopoly on poverty; indeed, 60 to 70 percen)t of all workers earninigbelow minimumil wages have been in Bogota for more than ten years.

We also examined in detail who the unemployed are. Because we haveinformation about currenit income only, it is difficult to distinguishbetween people who are poor because they are unemployed and thosewho are Ul mploed because they are poor. This problem is partiallyaddressed by looking at primary and secondary workers separate1y andexamining in detail the incidenec of unemployment in each decile.Even if wc regard all unlemilployed primary workers as temporarily out ofwork, the unemployment rate in the bottom 30 percent woul( still bedisproportionately hiigh. We therefore conclude that the probability ofunemploymnenit is higher among thc chroniically poor (see table 4-6).Our coniectures are further supportecd by the particularly low participa-tion rates in the bottomil deciles. which containi a substantial proportionof people who are not working. Male participationi rates rise and thienflatten out.

We emerge withi a distinctive profile of the unlemilployed. In the bot-tom decile-that is, among the verv poorest-only about 4(0 to 50 per-cent of males and 15 percent of feImales over 12 years of age participatein the labor force compared with the overall participation rate of about70 percent for men ancd about 30 per-cent for women. Among males(over age 12) in the bottom decile who dol) participate in the labor force.almost 30 percent were otit of work. Conse(qtuenitly, onilv about 35 per-cent of all males and about 8 to 10 percent of all females over age 12 inthe bottom decile have jobs. Of thc remainidler, some are discotiragedworkers and others are incapacitated by sickness or injury. Generatingemploymerit and upgrading skills are therefore only partial solitions forpoverty removal. Many of the poor are probably unemployable, andtheir poverty must be dealt with directly tilrotigh welfare measur es.

Summary

The 1970s marked a turning poilnt in Colombia's economic history. Pru-dent macroeconomic and trade policies, along with the successful sup-port of agricultural growth durinig the late 1960s and the early to mid-1970s, had helped create a growth-oriented, labor-using environiniienit.Colombia's uniprecedenited expansion of emplovmenit during the 1970sbrought about a palpable tighteninig of the labor market. This processwas aided by the boom in intcrilational coffee prices dtiring the late1970s. In addition, this was a per iod of demographic transitioni anic the

Table 4-6. Unemployment in Bogota by Income Decile and Sex

Income deoile& Total

Sex and year 1 2 3 4 5 6 7 8 9 10 All (thonisands)

Number in the

Unentmp/vinent ?-ate (pfercent) labarforfce

Male1975 30.8 11.4 7.3 8.7 8.0 5.2 3.8 2.5 2.6 0.4 6.9 624

1977 30.2 12.2 6.2 6.5 6.6 5.5 3.9 4.4 2.7 0.5 6.2 782

1978 19.8 10.2 6.1 5.6 4.2 2.7 2.0 2.9 0.7 (.6 4.7 753h

Female1975 29.2 12.1 12.7 9.6 7.9 7.8 6.1 3.7 1.9 1.1 6.4 337

1977 16.3 11.1 9.4 9.5 7.7 6.4 5.8 4.2 1.7 2.1 5.8 469

1978 6.2 7.3 5.1 8.9 7.3 4.3 2.5 1.9 1.7 0.4 4.3 4 6 3 b

Nuzmberl)j.tburiution off the anempilved (pe..rrent) utieottiployed

Male

1975 23.8 15.5 11.2 14.2 12.4 8.1 6.6 3.7 4.0 0.6 100.( 43

1977 23.2 15.0 9.() 10.7 1(.3 10.2 7.1 8.1 5.0 1.5 10(.0 491978 30.9 16.2 10.3 11.2 9.4 6.0 4.7 7.9 1.8 1.6 100.() 35

Fernale1975 13.0 11.8 13.0 12.4 11.2 12.4 10.6 8.1 4.3 3.1 100.( 21

1977 15.1 10.0 11.(0 10.7 11.1 10.5 11.0 9.7 5.1 6.0 10(.0 28

1978 21.9 16.7 9.5 13.6 13.8 7.9 5.5 5.0 5.2 1.0 100.0 20

Note: Percentages may not add to I00 because of rounding.a. Ranked fiom poorest (1) to richest (10), based oin monthly houisehold income per capita.h. The cstimated population of BogotA was lower than expected in 1978. The estimated labor force in 1978 is also low becatise the estimates for 1975 and

1977 use tihe original r)-NE expansion factof s.Source: 1975 DANE Special Bogota Household Survey; 1977 DANr llouschold Surve%: 1978 World Bank-DANE Household Surivev.

POVERTY, DISTRIBUTION OF INCOME, AND GROWTH 93

achievement of an urbanization level of more than 60 percent. Theresult was a slowing of urbanization that was somewhat unexpected afteralmost fifty years of rapid urban growth.

The tightening of the labor market led to measurable growth inhousehold income, and poverty decreased in Bogota and Cali duringthe 1970s. According to the data at hand, the distribution of income inthe two cities seems to have improved somewhat between 1973 and1978. The quality of the data is not robust enough for this to be statedconclusively, but such a tendency would be consistent with the otherchanges in the national economv. Despite this improvement, inequalityin the two cities remained very high: the Gini coefficient remainedbetween 0.50 and 0.55.

The spatial separation of rich and poor appears to have increasedduring the period under study, despite the overall improvement inincome distribution. The spatial separation is more pronounced byradial sectors than by distance from the city center, the norm in cities indeveloped countries.

In this study, we have placed great emphasis on understanding thespatial distribution of households according to their income levels andother characteristics. This is important as a prelude to comprehendingthe changes that take place in patterns of housing demand, transporta-tioIn, and demand for infrastructure as a city grows. Whereas some ofthe specific patterns found, particularly the high degree of spatial segre-gation of the rich and poor, may be peculiar to Bogota and Cali, ourfindings illustrate the importance of understanding these patterns inorder to make policy regarding infrastructure and other urban services.

Spatial disparities in an urban area can increase segmentation in thelabor market by exacerbating differences in educational background,family background, and other determinants of earnings. The deficien-cies in household envir-onment and school quality that a poor child issubject to can be worsened by such segregation. If there is a large con-centration of poorly educated or illiterate people in specific areas of thecity, it becomes much more difficult to operate good schools in thoseareas. It is difficult to find good teachers, the student peer group con-sists of equally disadvantaged children, and the quality of education suf-fers. Similarly, the peer group of poor workers or of the chronicallyunemployed usually comprises other poor workers and those who arechronically unemployed. The diminution of aspirations and contactsdecreases the worker's chances of moving out of the low-income orunemployment trap.

The r icher areas of a city are typically better served by public utilitiesthan poorer areas. Richer residents are simply more effective in voicingtheir demand for urbani public services. If an area has a mix of both rich

94 UNDERSTANDING TH1E DEVEl OPING MEFROPOLIS

and poor residents, the poor benefit from the better sanitationi, watersupply, roads, schools, medical services, and the like that may existbecause of the influence of the rich. If. however, the poor live in large,separate tracts, as in the southerni part of Bogota and the eastern part ofCali, it becomes easier for hard-pressed authorities to neglect them. InBogota andl Cali the severity of the potential spatial disadvantage hasbeen mitigated by the operation of decentralized housing markets andprivate land development. The decentralized nature of the urban publicservice agencies has also helped make t,hese agencies more responsiveto demands, including the demands of the poor, and the service deficitsin the poor areas have not been as large as they might have been.

There is some indication that the land-uise zoning policies have con-tribtited to the prevailing spatial development pattern. Resicdential-density zoning often consciously separates lower-density areas fromhigh-denisitv areas. Although there are market forces that tend to sepa-rate the rich and the poor through the operation of the land and hous-ing markets, care must be taken in land-use zoning not to exacerbatesuch natural tendencies, but instead to promote more mixed land use.

Although we have indicated in this chapter the difficulties of measur-ing the extent of poverty in a city, we have also demonstrated the possi-bility of arriving at some robust estimates. T'he areas of Bogota and Calithat have high malnutrition anid poverty levels are characterized bylarge, low-income families with high dependency ratios becatise of thepresence of many children. Workers who have low lifetime earnings sinkinto greater poverty as their households expand. In cities in developingcountries large numbers of workers with low education levels have flatage-earnings profiles. The expansion of their families dur-ing their mid-dle age pushes them into poverty, and hence an atypically large propor-tion of children are found in the malnotirished or poverty grotip. Thisfinding provides powerful support for the idea that child-directed nutri-tion and health programs are effective weapons against poverty. If pov-ertv is concentrated spatially, it becomes easier to target such programs,bitt knowledge of city structure is then essential.

Another important finding has been that many people who are fullyemployed are poor: their low productivity, not unemployment, is thecause of poverty. We also found that unemployment is more likely tooccur among poor households. The removal of core poverty thereforeinvolves basic welfare measures: mere increases in employment,although essential, will not be sufficient. investmenits in human capi-tal-both of adtilts and of those who are potential workers-wouldenhance productivity and therefore earnings (see chapter 5).

POVERTY, DISTRIBUTION OF INC OME, AND GROWTH 95

Notes

I. See, for example, Urrutia (1985) and Aniand (1983) for excellent studies ofColombia and Malavsia, respectivelv.

9. Estimates for 1964 are available in Selowsky (1979) and Berrn and lUrrutia1976i); see Berrv anid Soligo (1980) for 1974 estimates.

3. Mor-risoni (1984) claimed that this shair-e is similar for capitalist and socialistcountr ies.

4. See Ananid (1983) and Mohani (1986i), chapter 4. for an exposition on thederivation of the two measures. Both indexes vari between 0 anid 1, with 0 beiigequivalent to complete equaility. The Theil index maxv be interpreted as measur-ing the departur-e of an incomne distribuitioni from an even distribuition.

5. A word about these differ-enlt ineasires is in order here. The most commilondistributioni used has been thiat of' househiolds ranked by houselhold income. Ithas been argued by Datta alid Meermanl (1980) and Aniind (1983) that this dis-tribution does not provide a good indication of differences in the level of livingin the poptilation. Our general concer-n is with the welfare of individuals ratherthan of households. Moreover. hotisehoilds varx bv size as well as by age and sexcompositioni Hence, for true comparisons, adjustments should be made toaccounit for these variations. The best soltitioni would be to make adjutstments bvderiving "adtilt equivalenit" measures of constiilption. Practically speaking, thisis difficult; the next-best strategy is to use HIN APt' to ranik indixiduals (or house-holds) for welfare comilparisoils. As Datta and Meermart also totind. althoughthe ranikings of par-ticillar indixiduals change substiaintially when ranked by HIN-

(,AP as opposed to household incomiie, the overall result does not. The particularrankings are clearly important wheni the aimi is to find out the character-istics of thepoor or the rich, but the distuibutioln is not too different.

6. Reported in detail in Mohani (1984), whicih also reports estimates of otherineqiualitv measures.

7. See Anand (1983) and Mohan (1984, 1986) foi a (derivation of the decom-position of the Theil index. The wvithin-group contr-ibLtioll is essentially aweighted average of the indexes of inequality wvithin each grotip, and thebetween-group contribtition is a measure of the differences in the means,appropriately weighted. In the Theil index the weight used is the group share inincomie. It can be showil that this decomposition is additive.

8. t'he dependenicy ratio is thie ratio of the dependenit populattioni (defined ias

un(ler 15 or over 64 years old) to tire working-age populationi (ages 15-64).9. This has beeen confirimied to be the case from dletailed clinical utitrition

studies in Cali assoc iated withl those repor-ted in McKav and others (1978).

Chapter 5

Workers and Their Earnings

In this chapter we describe the structure of employment and the associ-ated pattern of labor earnings. We also report on the determinalnts ofthese earnings in order to under-stand the changes in income that havealready been documetnted. And we briefly discuss segmentation in thelabor market in terms of the identification of an informal sector. Oneverv interesting feature of the labor market in the late 1970s was theincreasing participation of women in the labor force; this is also ana-lyzed. Altlhough estimations were conductecd for both Bogoti and Cali,for ease of presentation we provide the results for Bogota only. Thefindings for Cali were generally consistent with those for Bogota. Onlythe main results are presented here, becatuse details of model structureand estimation procedures ai-e available in Mohan (1986, chaps. 6, 7, 8,and 9).

The overwhelminlg message that emerges is that high urban unem-ployment, long-term employmenit problems, high poverty levels, andsegmented labor markets are not inherent in a situationi of rapid urbangr-owthi. An appropriate economic environment bolstered with a judi-cious mix of policies can lead to positive results, as occurred in Bogotaand in Cali in the late 1970s.

For backgrlound it is important to be aware of the overall changes inthe employmnent structur-e that have taken place in Colombia over thepast fifty years. We documented earlier that Colombia went through anintensive phase of ulrbanizationi over this period, going from a level ofabout 30 percent to 65 percent. As happenis in maniy countries, thechange in employment structure followed that in sectoral output, withsome lag. Hence, as is evident fi-om the data below showing each sec-tor's percentage share in total employmenit, a dramatic change in thestructure of employment OCcuIrred in the relatively short period afterthe mid-1960s.

946

WORKERS AND THEIR EARNINGS 97

Secthn 1925 1938 1951 1964 1973 1978

Primary 67.1 67.1 57.3 50.6 41.9 35.2Secondary 15.7 17.8 16.3 17.7 20.6 22.4TerLiary 17.2 15.1 20.4 31.7 37.5 42.4

The increase in the share of the secondary and tertiary sectors, whichare primarily urbani, is remarkable over the 1973-78 period. The expan-sion of education (see table 5-1) that occurred during the 1950s and1960s increased the quality of the labor force very rapidly. Emplovmentchange, however, was more marked in the late 1970s (see table 5-2). Inabsolute terms it is clear that the tertiary sector was the mnost importantcomponent of this growth in employment, but the rates of growth werealso high for manufacturing and construction-between 7 and 9 per-cent a year, higher than output growth in all these sectors. The outputgrowth in agriculture was higher than employment growth over thesame period, thereby implying significant improvements in agriculturallabor productivity. Hence the predominanitly urbani sectors have grownin a labor-using maniner. This is contrary to the usual scenario in devel-oping countries, where the urban sectors are regarded as capital-usilngand agriculture as labor-using. Stcucttl-al change in productioll usiallvtakes place faster than the change in employment struclure: outputgrowth in manufacturillg usually occurs much faster than employmentgrowth, and it is the agricultutral sector that tends to absorb the expancl-ing labor force in the absence of employment opportunities in the mani-ufacturin-g and tertiary sectors. This is the distinctive phenomenon thatoccurred in Colombia in the late 1970s, which makes this study of theBogota labor market and earninigs determiinationi particularly interest-ing as an example of rapid population growth in a city accompanied byemploynment and income growth.

There was a remarkable downturni in unenmploymenit rates in Coloin-bia in the late 1970s, implying a consi(derable tighteniing in the labormnarket: participation rates rose until the mnid-1970s because of the mas-sive expansion in education butt then began to stabilize. The last COIlIuIm

of table 5-1, labeled "Other," includes mainly housewives: theirincreased participation in market work betwveen 1973 and 1978 is clear.The expansion of edlucationi has applied equally to women, and this hasbegun to manifest itself in their greater inclinationi to participate inmarket work.

Given its large share of the total urban population and its role ascapital city, the course of Bogota in the 1970s was generally that ofurban Colombia. It had more than its share of the employmenlt boom:its unemployment rates were consistently below those of otlher cities.The constellation of economic policies after 1966 tended to promoteexports, which are usually more labor-using. The governmenit was also

98 UJNDERSTANDING THE DEVELOPING METROPOLIS

Table 5-1. Changes in Labor Use in Colombia, 1951-78

'Ibtal

population age Percentage of total

15 and over

Year (thoausands) in sehool Employed Unemployed Other,

1951 6,910 2.4 54.1 1.7 41.81964 9,528 5.3 50.2 3.4 41.11973 12,449 9.4 45.5 5.4 39.71978 14,698 10.3 52.1 3.1 34.5

Average annual rate of change (percent)

1951-64 2.4 8.5 1.9 8.1 2.41964-73 3.0 7.8 1.9 8.4 2.61973-78 3.4 5.4 6.2 -7.7 0.5

a. Nfainlsi housewives.Source: 151, 1964, 1973 population censuises; 1978 IDANE National Household Survey

(Junre): Sokol and others (1984).

more oriented toward a pattern of public expenditures that helped thepoor: for example, expenditures in public health and education wereemphasized. The improvement in the availability of public services isalso quite evident in Bogota durinlg the 1960s and 1970s (see chapter 9).All this combined to produce increased demand for labor in urbanColombia and, specifically, in Bogota.

We observed a marked increase in the participation of working-agewomen in the tightened labor market of Bogota in the late 1970s.Because the participation rate of prime-age men (ages 25-55) in thelabor market is usually close to 100 percent in any case, the increase inthe labor force has to come from people not traditionally in the laborforce-prime-age women, young men, and young women. Because ofthe expansion of educational opportunities, however, young men andyoung womeni defer their entry into the labor force and opt to stay in

Table 5-2. Employment Change in Colombia, 1973-78

Thousands Thousands Net change, Percentage Annual rate

of vorkers, oj workers, 1973-78 oftotal of change

.Sector 1973 1978 (thousands) change (percent)

Agriculture 2,3(05 2,629 324 16 2.7

Mining 68 69 1 0 0.0Manufacturing 912 1,357 445 22 8.3Construction 255 360 105 6 7.1Tertiary 2,124 3,249 1,125 56 8.9

Total 5,664 7,664 2,000 100 6.2

Source: Sokol and others (1984).

WORKERS AND THEIR EARNIN(;S 99

school longer. This tends to have a lagged effect in that people withhigher education are more likely to participate in the labor force. Theparticipation rates of women with higher education increased fromabout 43 to 53 percent between 1973 and 1978, compared with partici-pation rates of about 36 percent and 28 percent of women with primaryand secondary education, respectively, in 1978 (having changed from30 percent and 31 percent, respectively, in 1973). The increasing trendin participation was at the low and high ends of the education spectrumduring this time of labor market tightening.

Some of the improvement in poverty and income distribution indexesis attributable to the increased utilization of female unskilled labor dur-ing the boom vears of the late 1970s in Bogota. The participation rate offemales in the bottom 30 percent as classified by household income percapita (HINCAP) increased from about 17 to 34 percent in the short timebetween 1975 and 1978. A detailed analysis of the participation behaviorof women showed that the own-wage elasticity of all married wvomen wasbetween 0.6 and 0.8, whereas that of primnc-age nonmarried women wasmuch lower-about 0.15 (see Mohan 1986, chap. 7). The latter groupincluded widows and never-married women. The analysis indicates thatwomen in cities in developinig counitries can be expected to participatein market work in much greater numbers in the future. Both a secularincrease in education levels, which raises the opportunity cost of notworkinig, and the pure wage response under conditions of a tighteninglabor market will motivate the rise.

The education level of the work force changed along with its makeup(see table 5-3). By the late 1970s the Bogota work force was almost fullyliterate. Education was more advanced as well as more widespread. In1973 about 40 percent of the labor force was educated tip to at least thesecondary level in Bogota; just five years later this propoltion hadincreased to about 55 percent. The expansioin of higher educatioll wasquite dramatic, especially for the women. With this vastly increased sup-ply of workers with higher education, the earnings differentials betweenhigher and primary education have declined-which is another expla-nation of the overall improvement in income distribution (see table 5-4).

Education expanded at very high rates in Colombia right through the1960s and 1970s. In 1960 primarv school enrollment was about 75 per-cent; by the mid-1970s it was 100 percent. Enrollmenit in secondaryschools increased even more dramatically, from abotit 12 percent of therelevant age cohort in 1960 to about 40 percent bv 1975 and just under50 percent by 1980. Enrollment in higher edticationi increased from 2percent in 1960 to 9 percent in 1975 and 12 percent by 1980. The speedof expansion was particularly marked between about 1965 and 1975.The private sector has played a substanitial role in the expansion of edu-cation in Colombia. At the primary level, almost 20 percent of enroll-

1(0 UNDERSTANDING THE DEVELOPIN(; MFTROPOLIS

Table 5-3. Distribution of Workers by Sex and Education Level(percent)

1973 1978

E.ducation level Males Fenales All Males Females All

Nonte 5.1 9.2 6.5 2.3 5.7 3.6Primai y 55.0 53.9 54.7 42.0 42.3 42.1Secondarv 31.1 32.5 31.6 38.8 37.6 38.4Ifigher 8.8 4.3 7.3 16.8 14.4 15.9lotal 1(00.( 100.0 100.() 100.( 10(0.( 100.(

Total numilherof workers 488,000 249,000 738,000 732.000 454,000 1,186.000

lotal nmbl)erin sample 45,080 23,041 68,121 3,078 1,914 4,992

Vole: Percentages in 1973 do not total 100 becauise of some omiiitted categories.Soura': 1973 population census sample; I 978 World Bank-rXF: Houscehold Survey.

ments were in private schools; this rose to ahout 54 percent at thesecondary level and remained fairly steady at 47 percent at the higher-edtication level in the late 1960s (see Jallade 1974). Durinig the 1960sthe private sector increased its share at the primary and higher levels,whereas the public sector expanded much faster at the secondary level.Hence, the expansion of the Colombian education system was due asmuch to private initiative as to specific governmental initiatives.

Total government expenditures onl education increased from lessthan 2 percent of GDP inl 1960 to about 3 percent by 1970. As might be

Table 54. Mean Income Ratios for Workers by Sex and EducationLevel

1 97,3 1978

E.ducation level Males' Females Feimale/male Males F emales' Female/male

None 0.66 0.66 0.51 0.69 (.80 0.75Primlalry 1.00 1.00 0.51 1.00 1.00 0.65

(1,175) (595) na. (4,903) (3,176) na.Secondary 2.11 2.62 0.63 0.159 1.81 0.74Higher 6.75 6.22 0.47 4.82 3.10 0.42T'otal 1.83 1.72 0.48 1.87 1.59 0.55

Total nutmberoftworkers 488,000 249,000 n a. 732,000 454,000 n.a.

ii.a. Not applicable.a. Ratio of mearn earnings, with those of worker-s with primary education used as a base.Nole: Ntimbers in pairentheses are actual earnings (in 1973 or 1978 pesos) of workers

with primary etducadon.Sourme: 1973 popltilationi censuis sample; 1978 World Bank-A.kNF Household Survey.

WORKERS AND THEIR EARNINGS 101

expected, private-sector schools were more prominent in the larger cit-ies: government expenditures on primary and secondary educationiwere more directed to smaller towns and the rural areas. Educationfiniancing was quite complex; it included substantial subvention fromthe federal government to state (departmental) and local governmentsand to other decentralized agencies. As will be seen in chapter 9, muchof Colombia's developmental public expenditure is rotied throughdecentralized, relatively autonomotis agencies, and this *sas tuue in edui-cation as in other areas.

The period under study immediately ftollowed this very substanitialeducational expansion. The participation of the private sector indicatesthat the public perceived high returns to e ducatio>n. Moreover, employ-ment expansion in the 19 7 0s was siuch that therc was adequate demandfor this educated labor.

The proportion of professional and technical workers increased con-siderably between 1973 and 1978 among both men and women. In oth-er respects, the distribution of workers among occupationis was quitcsimilar from onc vear to the other. Among males, production workersaccounted for about one-thircl of the labor force, constructioni andtransport workers for about 16 percent, professional administrative andtechnical for another 15 percent, and otlier service for the remainingthird. Among females, domestic servants constituted the largest group,accounting for one-fifth of all women workers; professionial administr-a-tive arid technical workers were about II percent, office workers 19 per-cent, and production workers about 15 percent, with other servicesrnaking up the remaining third. The increase in the professional andtechnical workers was for both men and women. W"e found that overallmean nominal earnings increased by more than 400 percent betwceen1973 and 1978, exceeding the inflation rate of 250 to 300 percent overthat period. It is quite striking that, consistent with earlier observationsof the decline in earnings differentials between higher and less-educatedworkers, the rate of increase in the less-skilled occupationis was higherthan that for the professionals, wlho on average barely kept pace with in-flation. The larger increases occurred in lower-paid fields like procluc-tioIn, construction, and transport. Production wor-kers were among thelowest paid. Among women, even domestic servanlts were better remu-nerated than the production workers, if the servants' income in kindis also accounted for. Domestic servants typically live in with their em-ployers and get free food and some clothes in addition to their mone-tary wages.

As might be expected from the spatial distribution of incomedescribed in the chapter 4, the lower-paid, less-skilled workers live pre-dominantly in the southern part of Bogota, and the professionals andadministrators live in the north (see map 5-1). Sectors 2 in the south

102 UNDERSTANDING THE DEVElPIN(G METROPOL.IS

Map 5-1. Bogota: Distribution of Occupations by Sector

IBRD 25345

\> ) ~administra ors r

Pomucn aoundai e

, w~~~ruto o kr

Scarce: lerks rnd, ap tyists, 5-!.

| uction, co o

- Comuna boundaries >j

E l Sectors

JANUARY 1994

.Sauxrr., Mohian 1986, map 5-1.

WORKERS AND THEIR EARNING,S 103

and 6 in the west are quite similar in their worker composition,although workers in sector 2 are poorer on average. The starkest con-trast is between sector-s 2 and 8, the poorest and richest sectors. Sixtypercent of the workers who live in sector 2 have only primary educationor less; in contrast, more than 45 percent of the workers in sector 8 havehigher education. This picture has not changed over the years exceptfor a slightly increased tendency for workers with similar occupations tocluster together-a pattern consistent with the increasing spatial ine-quality based on incomes. I)espite these spatial differences in terms ofincome and occupation we find no exidence of similar differences inparticipation rates. People in the poorer areas are not less likely to work.

In summary, we observe that the recorded increases in income inBogota have resulte(d from changes in both the demand and the supplysides of the labor market. On the demand side, the continuing overallchanges in the structur-e from a predominantly agricultural economyto an industry- and service-oriented one has resulted in continuinig ex-pansion of employment in these sectors. These sectors' expansion waslabor-using in the late 1970s. with cmployment expansion being greaterthan output expansion. On the supply side, these changes have beenmatched by continuing migration from rural areas. increased participa-tion of women in the labor force, and the improvement in educationalquality of the labor force. We therefore observed overall increases inlabor- earnings. There was little perceptible change in the structure ofemployment within Bogota, with industry- or production-related employ-ment accounting for just under one-third of total employment throughthe 1970s, although there was a clear shift toward more highly skilled oc-cupations. Workers lived whele they might be expected to, given spatialinconme distribution, with less-skilled workers concentrating in the southand west of Bogota and the professional and administrative workers inthe north.

Because the earnings distribution in urban areas is often said to resultfrom the segmentationi of labor markets into the formal and informalsectors, this issue is emphasized in the rest of this chapter. We estimateearnings functions to be the predominant analytical tool used to distin-guish the detcrminants of earnings and to test whether similar peopleearn different incomes in different sectors, however defined. Overall,we find little evidence of segmentation in the tight labor market condi-tions of Bogota in the late 1970s. We also explore the carnings differ-ences that persist across space in Bogota.

The Benefits of Education and Experience

There has been considerable controversy over the inter-pretationi ofreturns to educationi as measured by differences in earnings, and much

104 UNDERSTANDING THE DEVELOPING METROPOLIS

effort has gone into estimating these returns. Psacharapoulos (1980)summarized the results for different countries. Among the more accessi-ble works on the economic benefits of education for urban workers areAnand (1983) and Mazumdar (1981) on Malaysia and Bourguignon(1983) and Fields and Schultz (1980) on Colombia. Most estimates arelower for developed countries than for developing countries. This ispartly a measurement issue, because the variance in education levels istypically lower in developed countries and it becomes more importantto measure the quality of schooling. The private returns to schoolingrange between 12 and 18 percent for developing countries and between7 and 10 percent for middle- and high-income countries. The highestrates of return seem to be found in Latin America. The typical patternin other r egions is for primary education to bring the highest marginalreturns and secondary education the lowest. Colombia conforms to thebroad Latin American pattern, but the returns for higher educationexceed those for primary and secondary education. Most of thesereturtns are calculated with the assumption of no tuition costs during thetime of schooling and are therefore somewhat overestimated. We esti-mate the returnis to schooling in Bogota in some detail in order tounderstand thie measured pattern of income distribution and to followthe changes that have been observed.

*e build on the human-capital model of labor earnings as developedby Schultz (1961), Becker (1964), and Mincer (1974).' One of the keycriticisms of the traditional human capital earnings function (equation5-3 in note 1) attempting to estimate the returns to education has beenthe omission of a measure of "ability"-that is, a measure of v. It isargued that the productixity of schooling depends on the level of initialability, and ability itself is affected by family and other background. Theearnings function should therefore, at a minimum, contain measures ofability as well as background variables.

The data sets we used for Bogota were rich but did not have anv directmeasures of ability, family background, or schooling quality. We did,however, have somewhat comparable data sets for four different yearsbetween 1973 and 1978, so some time trends could be observed. Theonly background variables that were available, however, were the loca-tion of currenit residence, location of previous residence, and place ofbirth. In view of the distinct characteristics observed for different partsof the city, we regard the location of residence as a good backgroundproxy. We also use place of birth, categorized by size of town or ruralcommunity, as another proxy. These variables may be argued to be goodproxies for family background and schooling quality because childrentypically go to neighborhood schools. The hypothesis is that quality andintensity of schooling might vary positively with city size and the overall

WORKERS AND THEIR EARNIN(;S 105

quality of neighborhood. If these are good proxies, we correct at leastpartiallv for the omitted variables bias, and the resulting estimated ratesof return to schooling may be relied on.

So far, we have neglected on-the-job training as a component ofhuman capital. This is a straightforward extension, and we add the yearsof experience as another explanatory variable. To test for segmentationwe can also add different variables measuring characteristics of employ-ment. The final equations that we estimate are then of the form: 2

y= f (schooling, experience, region of origin, currentlocation of residence, characteristics of employment).

Trhe overwhelming result of oui- estimations is that the simple humancapital model, using merely sclhooling and on-thejob-training as expla-nations of earnings, works well in Bogota and explains as much as 50percent of the variance in earnings.

WN'e paid particular attention to a couple of issues. The major expan-sion in education has been documented and commented on. It lhasbeen suggested that this may have led to a decrease in the returns to ed-ucation. The availability of data sets for four different years enabled Us

to test for this hypothesis between 1973 and 1978. We have also corn-mented on the particular expansion of higher education. It was thereforeof interest to measure the rates of return to education distinguishedby different levels: primary, secondary, and higher. Another importantissue was screening. It is alleged that part of the returns to education) areactually returns to certification, rather than to the knowledge gainedfrom an additional year of schooling: employers are willing to pay a pre-mium to someone with a high school diploma or a completed collegedegree but not to others with similar years of schooling who lack thecertification.

The schooling variable was introduced into the regressions in twoways. First was the conventional schooling variable-the years of school-ing completed by the worker. The second was a "splined" variable. 3 Theyears of schooling were broken down into different chunks for primary,higher, and postgraduate education. In addition, separate dummy vari-ables were used for those completing primary, secondary, or higher edui-cation in order to test for the certification premium. Both specificationswere used for all the years in order to trace changes.

The overall rate of return estimated for the marginal year of educa-tioII was between 12 and 14 percent for 1978. It had declined frombetween 17 and 18 percent in 1973. This decline occurred across differ-ent specifications and can therefore be stated confidently. Moreover,Bourguignon (1983) has collected results from various studies and

106 UNDERSTANDING THE DEVELOPING METROPOLIS

added his own estimates to obtain a profile of the returns to educationin Bogota from the mid-1960s to mid-1970s and also showed a consistentdecline.

The return to higher education was found to be much greater thanthat to secondary education, which in itself was greater than that of pri-mary education; and the estimated returns to each level of schoolingdeclined over time. Because about half' of total secondary and collegeenrollment was in private schools with significant tuition costs, however,the estimated rates of return to higher and secondar y schooling as cal-culated from earnings functions are then overstated, given the assump-tion of low tuition costs. Nonetheless, the magnitudes of these returnsare probably higher than those of other investments, and continue to beso despite the declines documented.

The splined specification mentioned earlier permits us to measurethe certification bonus. The additional bonus for high school gradua-tion was about 20 percent and for college graduationi about 25 percent.That is, people who graduate can be expected to earn 20-25 percentmore than their counterparts who attended as many years of' school butdid not graduate. There was n1o measurable bonus for completing pri-mary education. These estimates imply that the economic benefits ofcompleting education can be quite high, and people have high incen-tives to receive certification. The low bonus for primary school comple-tion suggests that the Bogota labor market has become quitesophisticated, with a greater demand for more highly skilled, better-educated labor. It is also possible that completing and receiving certifi-cation for a specific education level provides more information on1ability and should not be interpreted merely as certification. Certifica-tion in itself implies the ability to carry out a task to completion: a traitthat is valuable in most employment situations.

We also tested the declining rieturns for education by making separateestimates for different age cohorts: 15-24, 25-34, 35-44, 45-54, and 55-64. As might be expected, the mean years of schooling increase witheach younger cohort-except for the youngest, many of whom are stillin school. The returns to educationi were the highest for the 35-44 agegroup. This is consistent with our other results on high mean returns toeducation earlier in the 197 0s. There are two processes at work. Initially,as the proportion of people with higher and secondary educationincreases, the variance in skill levels also increases, and those withhigher education are rewarded for their scarce skills. The averagereturn to education rises as a result of "educational deepening." As thisdeepening continues, the relative shares change, the return to highereducation starts declininig, and the composition effect becomes pre-dominant. The age cohort restilts suggest that the 25-34 age group is

WORKERS AND THEIR EARNIN(;S 107

not getting as high returns to education as the preceding cohort (seeMohan and Sabot 1988).

We now have a much better understanding of the income distributionin Bogota and Cali. A portion of the high inequality observed can beexplained by the larger returns to higher education and certification,which had a greater scarcity value in previous years. Given these highreturns, public policy and private initiative have responded by expand-ing the supply of education. It is also clear that people have opted toattend school longer and enter the labor force increasing ages later. Wecan expect these trends to contintue until the returns to educationdecline further and become comparable to returns to other invest-ments.

How does experience affect earnings? It is argued in the literaturethat as people work they add to their human capital as they do withschooling but at a lower rate, because schooling is usually full time andon-the-job training (OJT) is only part time. While working. a personpartly consumes, or gets returns from, his previous schooling and par-tially invests in learning by doing. Some of the segmentation literaturesuggests the existence of an inter-nial labor market fbr firms. It is arguedthat only firm-specific experience leads to improvements in a worker'sperformance. There is also an information asymmetry: firms may not beable to evaluate a person's previous experience as well as they can evalu-ate time spent with the tirm. If this were true, labor mobility woulddecrease. It would also be the case that workers who join better firms tobegin with would be expected to earn more than people of otherwiseequivalent training and ability. We would then expect higher returns tofirm-specific experience, measured as time elapsed since leaving school.Another specific measure of experience is the number of years spent inthe same occupation. It may be expected that the returns to occupation-specific experience would be higher than those to generalized workexperience.

The data enabled us to test for these hypotheses because the 1978World Bank-DANE survey asked respondents how long they had workedfor their current firm and how long they had worked in their curr-entoccupation. It was surprising to find that job mobility has been quitehigh in Bogota. Although the mean value of the generalized experiencevariable (EXPER)

4 was about 20, the mean value of years spent with thecurrent firm (YRSFIRM) was 5.6, and in the current occupation (YRSOC,CUP), 8.5. Workers thus appear to be quite mobile between jobs as wellas occupations-a finding contrary to the general impression thaturban labor markets in developing countries are rigid.

The standard specification of the experience variable is to enter it as aquadratic in the earnings function along uith education. 5 The quad-

108 IINDERSTANDING THE DEVELOPING METROPOLIS

ratic term was expected to have a negative coefficient reflecting thedecline in the rcturns to each marginal year of experience after somepoint. The marginal returns to an additional year of experience, onaverage, were 4.6, 2.8, and 1.0 percent after ten, twenty, and thirty yearsof work experience. The earnings for those with primary educationpeak early, at the age of about forty-five, for those with secondary educa-tion at abotut fifty-two, and for those with higher education at about fifty-seven.

Each of the three different specitications of the experience variablewas cntetred in a similar way. The results were:

EXPER YRSOcCCUP . RSFIRM

'21 0.063 0.040 0.044

P22 -0.0009 -0.0006 -0.0008

where NRsocCuP is the numlilber of years spent in the same occupationand YRSFIRNI is the number of ylears spent with the same firm. Thesecoefficients indicate that, contrary to expectations, firm-specific experi-ence and occupation-specific experience are not valued more highlythan generalized work experience. The marginial contribution of anadditional year of YRSOCCUP and VRSFIRM was in each case only about 2.8percent after ten years, which is much less than the 4.6 percent esti-mated for EXPE R. There is little evidence of strong firm-specific inter-niallabor markets. In developing countries this is said to exist particularly inlarge firms in the formal or protected sector, which reportedly paymuch higher wages than) other firms. This hypothesis is not borne outfor the labor market in Bogota or for our estimation for Cali.

Having established the rclatively regular nature of the determinationof earnings in Bogota in relation to the usual human capital variables ofeducation and lengtlh of work experience, we now investigate furthersome of thc other influiences on earnings: worker backgrounds on thesupply side and firms' characteristics on the demand side.

How Segmented is the Bogota Labor Market?

Much analysis of labor markets is motivated by a desire to understandwhat the determinants of existing inequalities are. In the preceding sec-tion we found that differences in education and work experience aloneexplain about half the variation in labor earnings. We now investigatedifferenit kinds of labor market imperfections that could result in differ-ent levels of earnings for otherwise similar people.6 The task for analysisthen becomes measurement of "similarity" and of differences in earn-ings after this simiilarity has been accounted for. Somc of the confusion

WORKERS AND ITHEIR EARNINGS 19

in the discussion of labor market segmentation arises from the possiblerelationship between what might be termed the human capital marketand the physical capital market. Given an unequal physical capital distri-bution, to the extent that acquisition of human capital is related to theinitial income and wealth distribution, the resulting human capital dis-tribution would be correlated with the initial asset and income distribu-tion. Without deliberate policy intervention this situatioll couldperpetuate itself indefinitely. The main traditional intervention to breakthe link between the asset and human capital distributions has been theprovision of free education to all or the provision of tuition fee corices-sions linked to means tests. We have alreadv found that the massiveexpansion of education in Colombia has been the result of both privateinitiative and explicit government policv and that there is at least someevidence that this expansion inight have served to reduce the scarcityreturns to higher education and thereby to reduce the inequality inearnings. This inassive expansion also means that a larger proportion ofpeople now have easier access to higher levels of educationi.

The Measurement of Labor Markel .Segnentation

Considerable methodological progress has taken place in the discussionof segmentation in recent years. Early argumenits wvere based merely onthe observation of heterogeneity in the labor market. Differences inmeans in earnings between different groups wvere seen as evidence ofsegmentation. Now much of the work is devoted to identifying differ-ences after accounting for similarities between human capital variables.Thus it is argued that differences arising fiom differences in humancapital endowments are in some sense justified, or at least to beexpected. Other differences are generally attributed to inarket imper-fections, and their removal would be argued to lead to greater efficiency.An econonmetric identification problem remains even in this approach,which is followed here. If there is a protected sector that, for whateverreasons, pays higher wages, it could make its selection criteria such thatthe more highly educated people would be found in these higher-payingjobs, even if tile jobs do not actually require skills resulting from highereducation. An earnings functioni wouldl then attribute thesc higherearnings to hunanl capital variables rather thani to segmentation result-ing from protectioni. However, the existe1nce of such a situatioln is a littledifficult to argue persuasively for private profit-seeking firms. It may bea little more defensible for public sector jobs, which often use educa-tional achievements as screening criteria. Because the marginal contri-bution of public sector jobs is mor-e difficult to measure, it is possiblethat the official requir-ements could be higher than necessary. Usingsimilar arguments it is also possible to adduce sucil a situation for larger

I10 tUNDERSIAND)IN(; THEF DEVElOPING M ETR(O)POI.IS

firms, which also have much greater indirect office activity in support oftheir production activities.

In this section we investigate the issue of mar-ket imper-fections fromiiboth the supply and the demand sides. The main issue from the supplyside is that the measurement of vears of schooling masks differences inabilitv, in family background, and in schooling quality. The other keysupply-side influenice in earninlgs is hypothesized to be the existence ofUnlioIns. From the demand side, the issue is usually posed as the exist-ence of a formal or protected sector in urban labor marlkets, wheresomc workers earni more than other workers because of various kinds ofrestrictive practices. Sucih practices can result from the governmenit set-ting a minimum wage, which has the effect of limitinig employinent andkeeping wages in the "legal" or "formal" sector higher than the rest.Similar effects can be caused by governmenit-legislated social securitypaymenits, which employers mav be requir-ed toi make. It is commoni tosuggest that these characteristics are highly correlated andl exist mostlyin larger firms, which are ofteni owvned by foreign enterprises. The sizeof a firm is theni used as the indicator var-iable separating the formal andinfornal sector. All of governmenit is usually assigned to the formal sec-tor. The basic idea is to identify intervening or segmeniting variables thatserve to restrict mobility within the labor market and can then help toexplain unequal earnings between people who would otherwise beregarded as equivalenit. In this study we attempt to identify the formalsector by variables such as firm size, the existence of formal social secu-rity systems and contracts in Firms, and the type of industry or activity a

firm is cngaged in.The overall finding is that the supply-side variables have a slightly

greater influence on earnings than clo the demand-side effects. Theprotected sector is simply diffictilt to find in BogotA. It is possible, how-ever, that these results are partly due to the boom conditions that pre-vailed in Colombia during the period of this study.

The Influence of Warkers' Background.s on Farnings

There are two ways to measurie the effects of background. Wc can eitherhypothesize that people from more advantaged backgrounids have ahigher return to schooling, or that the effect of their background ismanifested as a premiumn on earnings. In the first case the sample wouldneedl to be stratified by the backgrounid variable, and we wouildl test fordifferences in the returins to schooling-that is, differences in the edu-cation coefficient. In the second case a dumimiv variable for backgrcoundwould be used in the estimation test for the existence of a prermium. WeUhave utilized both methods, even thougih the first method causes someeconometric problems. If the dependent variable-that is. carnings-ishighly corirelated with the stratifying variable, the restiltinig coefficients

WORKERS AND THEIR EARNINGS HII

would be highly biased. In this case, if people of a specific backgroundare particularly poor, the estimated returns to their education would bebiased and on the low side. This procedure must be used judiciotisly,and conclusions drawn carefully.

The region-of-origin variable used was the place of birth, distin-guished by size of settlement: rural, small town (less than 100,000 popu-lation), large town (between 100,000 and I million population), city(more than I million population) but excluding Bogoti), and Bogotd.Because our interest is in environmental effects on quality of schooling,Bogota is also regarded as the place of origin for people who migratedto Bogota before the age of ten and presumably received the bulk oftheir schooling there. In addition to the possible difference in quality ofeducation, it is reasonable to hypothesize that people who grow tip inmetropolitan areas are exposed to more varied influences and a widervariety of information, with the result that their schooling may beformed into human capital stock more efficiently. The schools in largercities are also likely to have better teachers. It is also often suggested thatmigrants are at a disadvantage in competing with natives in the urbanlabor market.

These background variables did not affect the educationi coefficients,however. There seems to be little correlation between place of originand returins to schooling. The dummy variables for region of origin, onthe other hand, were all substanitially significanit, the base comparisonbeing with workers fi-om rural backgrounds:

I'ercei tage

Plae oJ onrgin in Bogai C( efficien t

Bogot,t 46 0.10 to 0.1ILarge citv 3 0.15 to 0.18

L.arge town 6 0.19 to, 0.23Small town 26 0.08 to 0.Io

Rural (base) 19 0

Both the Bogota natives and migrants from other urbani areas appearto be 10 to 20 percent better off than rural migrants, but there is little todifferentiate Bogota natives from migrants from other urbanl areas. Ifanything. migrants from other large towns are better off than native-born Bogotanos-and this may be a r esult of self-selection. The estimateof returns to education from the stratified samples of workers producedsimilar results. There was little difference in the returns to education forBogota natives, workers in big cities, and other urban migrants, but therural migrants did get significantly lower returns. Given the consistencybetween the two methods, it is reasonable to conclude that, on average,there is no evidence of differences in the quality of education betweensmall and large towns and Bogoti, but the rural folk probably do sufferfrom poor schooling and other negative environimenital influences.

112 UNDERSTANDING THE DEVELOPING METROPOLIS

Using the location of residence within Bogota as a variable in earn-ings functions estimations is one of the mor-e controversial aspects ofthis study. Is it reasonable to hypothesize that within a city the locationof residence "causes" a different rate of return to education, or that itinfluences earnings in other ways? Is the labor market in fact segmentedin this manner? The argument is that current location is acting as aproxy for unmeastired variables. One can expect the quality of school-ing to differ by location within the city. To measure only the years ofschooling is to ignore differences in quality; however, the location of res-idence could be a proxy for schooling quality. Parents' education andincome, which largely determine where the family lives, then affect thechildren's schooling by virtue of location. Residence in a low-incomeneighborhood can also affect a person's network of contacts for obtain-ing a high-income job. Because the lower-income neighborhoods showrather flat age-earnings profiles, one can also expect that the demon-stration effect dampens residents' expectations and aspirations. 7 Thiscan have a cumulative effect on earnings, similar to the discouraged-worker hypothesis for nonparticipators. Finally, location of residence isalso a proxy for social class, which probably has significant effects onearnings. The larger question is whether this pattern is peculiar toBogota or whether it exists in other cities as well. Clearly, all cities havetheir poor and rich neighborhloods; the question is whether they areclustered in the same wav as they are in Bogota. We also need to ask howmuch spatial mobility there is across differenet parts of the cities. To whatextent does current location affect the current earnings and streams offuttire earnings?

Aill large cities in the world have rich and poor neighborhoods. Peo-ple choose their place of residence according to their income, the loca-tion of their work, and their preferences regarding the amenitiesand other characteristics of the neighborhood. Normally, income isthe main factor in the choice of residential location. People also sortthemselves out by ethnic origin and, to some extent, by occupation andclass. The issue here is a feedback mechanism that reverses the causa-tion somewhat so that a person's location affects his or her earningpotential.

We have documented that both Bogota and Cali are characterized bya high degree of spatial inequality in incomes. Recalling the pattern inBogota, the north (sector 8) is particularly rich, and the south (sector 2)is particularly poor. The other sectors are more heterogeneous, but theranking of the radial sectors in ascending order of household incomeper capita is 2, 3, 6, 4, 5, 7, and 8. Although there are virtually no richpeople in the south, there are poor people in almost every part of thecity. Historically, it seems that the rich have continuously moved northwhile the poor have located in the south and later in the west. This

WORKERS AND THEIR EARNINGS I 13

makes it even more likely that the better schools would largely be in thenor-th. When large areas of a city are known as rich and poor, as inBogota, people's addresses can become screening devices. Employers,for example, may use an applicant's address to gauge his or her likelycharacteristics-reliability, home background, and so forth-in muchthe same way that names of schools are used as screening devices.

In summary, if the residential location variables are found to hiave asignificant effect on earnings, there are at least three explanationis. First,in the human capital tradition, it may be argued that these location vari-ables act as proxies for ability, schooling quality, and the like, which areunmeasured otherwise. Second, also in the humani capital tradition, itmay be argued that they act as proxies for other productivity characteris-tics of workers (class, statLis, aspiration, attitudes, and contacts) that arecorrelated with their residence location. The third possibility is thatlocation is being tised as a partial screening device in a city character-ized by notable spatial differences in income.

There are essentially two location variables that have been used. Firstare the dummies for each residential sector, usinig sector 2, the poorest,as the comparator base. Second is the distance of residence from thecenter of the city. Dummy coefficients for sectors 3 through 7 are allbetween about 0.15 to 0.25 and are not significantly different from 0.2.Sector 8 is obviously different, with coefficients of about 0.6. These coef-ficienits measure the systematic deviation in log income from sector 2means, keeping everything else constanit, including distance from thecity center. Thus, workers in sectors 3 through 7 receive about 20 per-cent more in earnings than otherwise equivalent workers in sector 2,whereas workers in sector 8 receive about 60 percent more. The addi-tion of these variables does not appreciably increase the level of expla-nationi (R2) that is the proportion of log variance of earnings that isaccounited for. But all these coefficients are lhighly significant. We havealso discussed at some length the location of many of the poor at theperiphery. The estimated coefficient of the distance variable impliesthat workers' earnings decline by about 2.5 percent per kilometer fromlithe citv center, on average. Furthermore, the addition of the distancevariable makes the differences in earnings between radial sectors morepronounced.

As before, we also estimated the returns to education by stratifyingthe sample by residential sector. Although this procedure stiffers fromtruncationi bias, particularly for the largely poor sector 2. the estimatedreturns are fouLid to vary systematically by sector; the coefficients forschooling are 0.08, 0.09, 0.10, 0.1 1, 0.11 . 0.14. and 0.16 for sectors 2, 3,4, 5, 6, 7, and 8, r espectively. These differences are not statistically signif-icant between sectors 2 and 3 and between sectors 4, 5, and 6, but theoverall pattern is consisterit with dummy variable coefficients.

1 14 UNDERSTANDING THE l)EVELOPING MIETROPOLIS

These results were further investigated in different ways. First, do thelow-income workers who live in richer areas earn more than similarworkers in the poorer areas? The indication is that they do not. It seemsthat the differences are mainly among the better-educated workers. Aworker from the poorer areas who manages to get secondary and highereducation does earn less thani someone from the richer areas with anequivalent education. Schooling quality differences presulmably explailthis finding. For the relatively unskilled, less-educated workers, school-ing quality makes little differences. Second, how mobile are people? I)opeople who move "down" from richer to poorer locations carry theirwork characteristics with them, and vice versa? The data containedinformation about the last move of the household. We classified asupwardly mobile movers those who had moved from a poorer sector to aricher sector in the last ten years; those who had moved from a richer toa poorer sector were classified as downwardly mobile. For the 1978 data,the former comprised 40 percent of the workers and the latter about 15percent. Most of the moves were between adjacent sectors. 8 Bogotahouseholds were therefore quite mobile, but the intersectoral mobilityis between relatively similar sectors. The addition of these diummies didnot alter other results but did increase the level of explanation of earn-ings slightly. The downwardly mobile movers appear to earn about 10percent more than similar people elsewhere. This suggests that school-ing quality does matter and that people fi-om higher-income locationswho move to lower-income ones retain some of the higher-income char-acteristics. The upwardiv mobile people seem to be no different thanother similar workers. These results argue against the labeling or screen-ing hypothesis and support the hypothesis that location acts as a proxyfor schooling and other aspects of background.

The last supply-side segmenting variable is that of union membership.Only about one-quarter of the male workers in both Bogota and Calibelonged to unions. Wheni union membership was introduced asanother- dummy variable in the earnings functionis, the average earningspremium for union members was about 6 percent in Bogota and 12 per-cent in Cali. It is striking that this appears to be more important formigr-anits from riural areas and small towns than for others. Given theearlier finding that it was only these migrants who were disadvantaged,and not other migrants or natives, this finding suggests that union mem-bershiip is more important and serves as a screening device when thereare few other distinctions of labor skill or quality.

Results from both the place-of-origin variable and the location-of-residence investigations support the idea that these variables are essen-tially proxies for backgrotind, particularly schooling quality and abilityas iifliteniced by background. Although migrants are no worse off thannatives in general, rtiral migranits do appear to be somewhat worse oft:

WORKERS ANI) THEIR EARNINGS 115

The results imply that there might be income feedback effects from pro-longed residence in disadvantaged neighborhoods. One way of compen-sating for some of these effects is to belong to a union. The importanceof the background effects should not, however, be exaggerated; theyadd only about 3 to 6 percent to the level of' explanation in the log vari-ance of earnings, even though each of the coefficients estimated ishighly significant statistically.

There seems to be little evidence of the spatial disadvantage hypothe-sis usually expounded for cities in the United States. High inner-cityunemployment rates are often explained by the flight ofjobs to the sub-urbs and the resulting deterioration in the access of poor workers, Utis-

ally blacks, to these jobs. In Bogota, even though the poorer parts of thecity have far fewer jobs than richer areas do in relation to their popula-tion, there is no measurable difference between participation rates andunemployment rates. The differences are essentially those in e arnings.Access to jobs does not seem to be a problem for the urban poor inColombia. This may be partly because of Bogota's relatively cheap pub-lic transportation system and flat bus fares (see chapter 8).

The Protected Sector

On the demand side, we have attempted to identify the protected sectorby a number of different variables. First, it was hypothesized that firmswith formal social security schemes and/or formal employment con-tracts must belong to the formal sector. Second, consistent with muchother work on the subject, it is expected that large firms pay their work-ers more than other smaller firms do. Third, different types of work sta-tus (for example, employees versus self-employed workers) may' givedifferent returns. Fourth, we test to determine if different kinds ofindustry give varying returns. Last, we see if there is any measurable dis-tinction between the government and the private sector, governmentbeing an easily identifiable protected sector and often alleged to paymuch more for comparable jobs than the private sector. It is difficult tofind labor market segmentation in Bogota according to any of these cri-teria. There are some measurable differences, but protected-sector pre-miums or returns to education, measured as earnings received, areseldom higher than 10 to 15 percent.

We first introduce dummy variables for the existence of social securityschemes or employment contracts in the workers' place of'work. Abouthalf the workers, both male and female, work for employers who havethese benefits. The male workers who have contracts earn about 5 per-cent more than other equivalent workers. Similarly, workers in firmswith social security schemes earn about 7 to 9 percent more than otherequivaleint workers. The correlation coefficient between these two vari-

1 16 UNDERSTANDING T'HE DEVELOPING METROPOLIS

ables is 0.5. Next, a variable measuring the logarithm of firm size isintroduced. 9 Its coefficient implies that a doubling of firm size increasesearnings by abouLt 2.5 percent. This means that an employee of a firmwith 100 employees would earn about 10 percent more than a similaremployee of a firm with fewer than 5 employees. These results meanthat workers in large firms (with, say, more than 100 employees) thatalso have other formal sector characteristics would earn about 10 to 20percent more than similar employees in very small firms (with I to 5employees). Although these differences are significant, they are notlarge and can scarcely be said to imply the existence of a strong pro-tected sector in Bogota. The existence of relatively high labor mobility isconsistenit with this finding, as are the relatively low returns to firm spe-cific experience. Similar workers in different kinds of firms, formal andinformal, receive roughly similar earnings.

We next examine other characteristics of work that could possibly seg-ment the labor market in an identifiable manner. First, earnings func-tions were estimated for different types of workers: blue-collar andwhite-collar employees, and employers and the self-employed. Becausethe earnings and education of blue-collar workers are uniformly low,the estimationi for these workers suffers from truncation bias, andreturns to education for them are found to be low. But there is no statis-tically significant difference between the returns to education for white-collar employees and those for employers and the self-employed. Sec-ond, earning functions were again estimated, stratified by industry ofactivity classified by the standard industrial classification at the one-digitlevel. Again, no statistically significant differences were found in thereturnis to education between manufacturing, construction, trade andcommerce, public administration and other services, and financialestablishments. Workers in the tranispor-tation and communiication sec-tors, however, seem to get somewhat lower returns; in this case it doesnot appear to be a truncationi problem. Finally, different estimates werealso made for the government and private sectors.

The mean earnings for workers in government were significantlvhigher than for workers in the private sector. The rate of return to edu-cation, however, was not significantly different. The main difference isthat union members in government get about 20 percent higher earn-ings (compared to about 6 percent higher in the private sector). A nota-ble feature of the earnings function estimation for government workerswas that R2 , at 0.68, was the highest earnings function estimated in thisstudy. The human capital variables-in particular, years of education-explained more than two-thirds of the variance in earnings in govern-ment jobs. This reflects the higher degree of codification in earningsstructure in the government: earning levels are more clearly related to

WORKERS AND THEIR EARNINGS 117

formal qutalifications. The returns to work experience were lower in thegovernment than in the private sector. This probably reflects a tendencyin the government to do more formal screening at the time of entry andthen to provide slow growth in earnings with experience. Governmentemployees accounted for about 15 percent of the total in Bogota.

Given these results, it is difficult to argue that the Bogota labor mar-ket is highly segmented. Most of the variation in labor earnings isexplained by the traditional htlinan capital variables. Although differ-ences in background add little to the overall level of explanation, theymake a substantial difference in the prediction of earninigs for workerswith specific backgrounds. Although these differences have been mea-sured as the place of origin or current residence location, they areattributed to schooling quality and other background influences. Onthe demand side, a strong protected sector is simply difficult to identify.althouglh workers in formal sector firms do earn about 10 to 20 percentrnore than other equivalent workers. The only sector that seems to beparticularly protected is the government sector, witli preiniums of about15 to 20 percent being paid to workers thanks to the efforts of stronggovernment unions. Outside the government, union membershipmakes little difference except among relatively unskilled r ural migrants.

Operation of the Urban Labor Market: What Have We Learned?

The results reported in this chapter are in marked contrast to mostother studies of urban labor markets and to popular impressions of theemployment situation in cities in developing countries. The improve-ment in the urbani employment situation in Colombia in the 1970s waspartly related to the existence of boom conditions there in the latterpart of the decade. The rapid growth in GDP dturing this period resultedfrom policies that favored agriculttiral growth and expansion of export-related manufacturing activities. Governmental programs for infrastruc-ture investment also helped provide high growth in tertiary-sectoremployment in urban areas. The boom in coffee prices during the late197 0s contributed to prosperity in the agricultural sector and to unprec-edented tightening of the labor market in rural areas, which wasreflected in higher wages for tunskilled workers in urban areas.

The late 1960s and the 1970s were also periods of significant demo-graphic changes: fertility declined remarkably, resulting in slowdowns inurban population growth, overall populationi growth, and rural-urbanmigration. In Colombia this also coincided with the turning point in thepattern of urbanizationi; the effervescent urbanization between 1940and 1980, when the urban population rose from 30 to 65 percent of the

118 UNDERSTANDING THE DEVELOPING METROPOLIS

total population, gave wav to a rather slower rural-urban transforma-

tion. This is a familiar S-shaped urbanization pattern.The other reason our findings in this chapter differ from those in

other studies is that most other studies, particularly those of the infor-mal sector, suffer from the use of biased samples. The informal sector isoften identified by analyzing the characteristics of workers from slums,from specific occupations or types of firms, and so on. Appropriate coin-parisons are seldom made. Doing so means keeping other things con-stant, and this is difficult without using a careftilly drawn, citywide or all-urbani sample.

The labor market was found to operate quite efficiently. There isscant evidence of segmentation; similarly qualified people earn similarincomes in different activities; there is high labor mobility'; and return1sto humani capital are similar to those in other countries. Part of the ten-dencv toward better income distribution is attributed to the "deepen-ing" of education, with consistent and rapid improvements in averageeducationi attainment (see Mohan and Sabot 1988). There is evidenceof declines in the returns to secondary and higlher education as thestock of people in these categories has increased. Concurrently, the realwages of the unskilled have improved-a consequence of tighteninglabor market conditions in both rural and urbani areas and improvedoverall labor quality with the universalizationi of primary education.These forces have resulted in a decline in the ratio of earnings betweenthe mor-e-educated and the less-educated. These results suggest verystrongly that the continued expansion of education, along with a reduc-tion in the dispersion of completed education, will be quite effective inreducing the levels of inequality, at least in personal labor earnings overthe long run.

The evidence from Bogota and Cali indicates that the labor market isnot characterized by a strong protected sector. Individuals working in es-tablishments with formalized employment arrangements (with unions,social security schemes, and written contracts) and those in large enter-prises earn only slightly' more than others. Moreover, the stability of theestimated rates of returni to schooling and experience across the differ-ent occupational and industrial categories argue against the existenceof a highly segmented labor market in Bogota and C.ali. Much of thewriting on1 the existence of informal and protected sectors in cities indeveloping countries is connected with high rural-urban migrationi.Contrary to popular nmisconceptionis, migrants are not especially poor;thev do not concentrate in specific areas of the city-the center or theperiphery; they are not concenitrated in particular occupations or activi-ties; and thev are not less educated or less skilled than natives, on aver-age. In short, they are not found to be disadvantaged in any respect.

WORKERS AND) THEIR EARNINGS 119

Women earn less than men in every occupation and industry. Theratio of average female earnings to male earnings is between one-halfand twvo-thirds in most dfeveloping countries. In Bogota and Cali it isnearer one-half. WAhat evidence thele is suggests that women get rates ofr eturni for cducation similar to those of mcn but do not get comparableretur ns for experience. Thus, women's earnings differentials increasewith age. Womeni with higher edtucation are far more likelv to work thanothers. The fact that womens access to education is now similar to thatof men suggests that women in developing coun1tries are likelv to enterthe labor force in unprecedented numbers, mtch as they did in devel-oped countries buit perhaps even faster.

Notes

1. We can express the basic humani capital model as follows:

(5-1 ) Y = Ph- H.e"(5-2) 1I= e-e"(5-3) y = lnY = lnPh + S+ u +-1

where Y is labor earninigs, 11 is the unobserved quanititv of humall capital, Ph isthe market rental price of a unit of humani capital (which ma, vary over timeand space), and u represenlts other influenices on wages. Equation 5.2 may beinterprete(l as a production functioni for htuman capital, uSilig schooling (S) asonle input and usinig abilitv, efficiency, and so forth (denoted bv v) as otherinputs. Ph5 f then gives the flow of returns from the stock of humani capital /J,with u subsuminig other influences on earnings.

2. The equation estimated was ln(Y) = P,, + XI PI + X2 ,f2 + X< 033 + X4P4 + X5+ E, where XI = education variables, X, = experience variables, .Y3 = place-of-origin shift variables, X4 = location-of-residenct shift variables, and X5 = charac-teristics of employment. See Mohiani (1986, chap. 8) for a more detailed specifi-cation, the actual variables used, and other details.

3. Smith and iWelch (1977) also usecl a splined educationi variable to obtainestimates of the differential returins to different levels of education. In our spec-ification, the splined variable essentially decomposes the educationi Variable intostepwise dummyv (the certification premium) andl slope variables (vwar-s of cdui-cation at each level).

4. EXPER = (age - years of education - 6) years.5. That is,

IlnY= ,5,, + -1 RSEDtI + 021 EXPER + 022 (FXPFR)2 + X 03 + X4 4 + X5 05 + E

(earnings) (schooling)

where X3 = place-of-origini variables; X4 = location-of-residenice variables, X5 =employmenit chiaracter-istic variables, and E is the error term.

120 UNDERSTANDING THE DEVELOPING METROPOLIS

6. This section summarizes work found in Fields (1980) and in Mohan (1980,

1981, 1984, and 1986, chap. 9).7. See McGregor (1977) for similar arguments in relation to Glasgow.

8. See flamer (1981) for more details on residential mobility.9. The firm size itself was also tried, but the coefficient was negligible and

i nsignifican t.

Chapter 6

Firms and Their Location Behavior

The preceding chapter examined the supply side of the labor market.This chapter looks at the demand side of urban employment-morespecifically, at the location patterns of firms and the behavior underly-ing these patterns.' Household behavior is, in general, better understoodthan the behavior of firms. This is partly because the characteristics ofhouseholds are more homogenous than the characteristics of firms.Firms in retail or wholesale trade are quite different, for example, frommanufacturinig firms, which themselves can be different in each indus-try. The location imperatives for banks are quite distinct from those ofgovernment departnments. Moreover, bank branches are quite differentin behavior from their head offices. The study of employment locationtherefore has to account for a number of differences between firms andin different dimensions-in terms of their function, size, age, and soforth. Understanding workplace location is crucial to understandingresidence location and emerging urban travel patterns.

Although we established the general location patterns of the differenteconomic activities that existed in Bogota and Cali, we paid specialattention to manufacturing activities because they are usually regardedas leading other activities in a city. Their extensive backward and for-ward linkages often determine the location of other activities. In orderto focus on behavior, we intensively studied firms in a couple of selectedindustries. Before we discuss our findings, however, let us consider thecomposition of employment as a whole, the spatial distribution of differ-ent types of employment, and the changes that have recently takenplace in Bogota and Cali.

Decentralization in the two cities has been so thorough that in thelate 1970s, the central business districts of Bogota and Cali providedonly about 15 percent and 18 percent of employment, respectively, with

121

122 UNDERSTANDING THE DEVELOPING METROPOLIS

net new employment almost entirely outside the center. Employmentlocation trends are vital components of' the urban dynamic. For onethinlg, place of residence is closely related to place of work: familieslook for housinig in areas not too far from their jobs. For another, thedifferent economic activities liniked in a production process often needto be located near one another to reduce costs and facilitate tasks to beperf'ormed. If one of the segments (that is, firms) forming part of sucha technological chain moves to another location, the other segmentsare likelv to move to the same area, thus instituting a substantialchange in the spatial distributioni of employment and probably also inthe populationi, housing, physical infrastructure, trip patterns, anddemand for transportation-in other words, altering the spatial struc-ttie of the city.

In charting the changes that have taken place in Bogota and Cali inemployment location, we resorted to a large array of data sets (see theappendix). The broad patterns of employment location were estab-lished from the World Bank-DANE Household Survey of 1978 in whicheach worker was asked to provide the characteristics of the firm wherehe or she worked, including its location within the city. Trend changeswere established by comparing the 1978 pattern with that in 1972, whenanother lhousehold survey-the Phase 11 Survey-was conducted inBogotA. We also had access to the 1976 Social Security establishmentfiles for l3ogota and the 1978 files for Cali. This source naturally pro-vides poor coverage of small firms but is useftul for checking the qualityof iifor-mation in the 1978 Household Survey. For a detailed analysis ofthe manuftacttuinig sector we resorted to two additional sets of data.I)ANE publishes an annual directory of industrial firms that employ tenor more employees. The directorsv incltzdes informationi about individ-ual establishments, including locationi, production, sales, and inputsised in manufacturing. Based on these files for each year from 1970 to1975, we were able to trace the movement of each firm over this period.A special survey of a sample of nmanufacturing firms was conducted forthe City Study in 1978 to obtain further detail. It is unusual to have suchrich data in any city in developed or developing countries. We hope,therefore, that this analysis of employment location behavior has appli-cability beyond Bogota and Cali.

Among the questions this chapter raises and attempts to answer arethese: Whlat kinds of firms move? Which remain stationary? Where donew firms locate? Why do the firms that move do so? How do new firmsdecide where to locate? Are there differences in the behavior of smalland large firms? What can we expect in the future? What is the magni-tiude of capital land substitution? Ilow does this affect location in thepresence of'lan(d price gradients?

FIRMS AND THEIR l.OCATION BEHAVIOR 123

Trends in the Location of Employment

In both Bogota and Cali, employment in all sectors (manulactuijig,commerce, and services) has historically been concentrated in andaroutid the CBD, but this situationi has beeni chanlginig rapidly since the1960s as the two cities grew. The most significant anid best-documentedlocational shift may be seen in the case of manufactur-ing firms that havebeen movinig out of the CB[) to peripheral locations, thus reducinlg theabsolute number of manutactur-inig jobs in the (BD. This is consistentwith the worldwi(de trend observed in all large cities in both developedand developing counitries for which data about employment locationare available. In Bogota and Cali this decentralization has been occur-ring in conjunictioni with rapid growth in mantu,lfctUIring employment:annual average growth rates were almost 8 percent between 1973 and1978. Employmenit in the commerce and service sectors also decentral-ized but more slowly. The larger retail and wholesale trade firms contin-ued to locate in or near the CBI) while ther-e was somiec decentralizationof the smaller firmns. This was also true of the service sector, althoughthe pattern was not as clear as for trade and commercial firms. Conse-quently, the CBD became more specialized in its commerce and servicef'lnctionis, wlhile the periphery is specializinhg in manufacturinig. OverallCBI) emplovnielt has remained constant in absolute termns, but becauseof total employmnent growth, its share has been falling.

The employmenit structure of Bogota and C.ali is not very differentfrom that of other large citics in the world (see Mohan 1986, chapter 3).The share of manufacturinig employment in Bogota is 24 percent, thatof commerce 20 percent, and that of services (includinig finance) 41percent. The last is a litde higher than in many cities because of thepresence of the nationial government in Bogota, but it is similar to theshare in othe r large capital cities in the world. Cali is a little different inthat the share of manufacturing is about 30 percent and that of servicesabout 33 percent. Almost all large cities in the developing world havemore than 20 percent employmenit in manufacturing. Noncapital citiessuLch as Calcutta, Bombay, and Sao Paulo have a higher shaire of manu-facturinig employment, as might be expected. Similarly, in a typical largeNorth American city the structure is about 30 percent manufactrz-itig,30 percent commerce, and 25 percent services; utilities, co1istruction,transport, and communication account for the rest. The structule ofemployrnent in Bogota has remained relatively stable since at least 1950,and it may be regar-ded as a well-balanced metropolitan city. No particu-lar sector is overrepresented in Cali either. About half of all em)ploy-ment in both the cities was in small firms, that is, firms employing fewerthan ten employees (see table 6-1). The commerce and service sectors

124 UNDERSTANDING THE: DFVL.OPING CMFTROPOLIS

Table 6-1. Employment by Firm Size and Major Industry Group, 1978(percent)

A.1 Mlanular-

industrips' luring Commerre J,nanice Servires

BogotaSmallb 52.0 42.0 71.6 43.2 59.8Iarge' 48.0 58.0 28.4 56.8 40.2

Total 100.0 100.0 1)0.0 100.0 10(.0

Thousands ofpeople employed 1,212 286 246 98 394

Percentage share oftotal emplovmenit 100.0 23.6 20.3 8.1 35.0

Cali

Smallb 55.2 36.3 77.0 55.6 72.9Largec 44.8 63.7 23.0 44.4 27.1

Total 100.0 10(.( 100.0 100.0 100.0

Thousands ofpeople employed 368 115 80 13 107

Percentage share oftotal employment 100.( 31.2 21.5 3.7 29.0

a. InCludes other sectors in adlition to manufacturing, colmmerce, finiance, and scr-'ices.

h. Firms with fewer than ten employees.c. Firms withi ten or more emplovees.Source: K. S. Lee (1989. table 2-3), based On World Bank-i)ANF Househioll Survey 1978.

were characterized by the presence of many small firms, whereas 60 to65 percent of total manufacturinig employmiient was in large firms.

How was this employment distributed across the two cities? WAeemploy the same spatial scheme used earlier-that is, rings and sectors(see table 6-2). As might be expected, Cali, the smaller city, is more cen-tralized overall than Bogota. The large firms in the cominerce andfinanec sectors locate in the center, with rings I and 2 accounting iorhalf or more of total employment in these groups. Manufacturing is themost decentralized, and the small and large firms seem similar in theirlocation behavior. Small retail (commerce) firms are almost as decen-tralized as firms in the maniufactiuring sector.

The distribtition of employment by radial sectors (see table 63)reflects the pattern of land use specialization referred to earlier. InBogota, manufacturing employmtnt is concentrated in the "induistrialcorridor"-sectors 4 and 5. Sectors 3 (the Bosa area in the southwest)and 6 (near the airport) have recently begtun to attract manufactul-inlgactivities at their peripheries. jobs in the comrnerce, finance, and ser-vice sectors are located primarily in the r ich sectors 7 and 38 in the north,as well as in the CBD.2 T his reflects the northwar( inovement of high-

FIRMS AND THEIR LOC(ATION BEHAVIOR 125

Table 6-2. Employment Distribution by Ring, Firm Size, and MajorIndustry Group, 1978(percent)

Alan tfactunng Commerce Ftnance Servires

Small Large Small Large Small Large Small Large

Ring firmn.s firms firms firtns firmis firms firms firms

Bogota1 4.66 7.00 13.13 22.57 52.14 33.99 8.77 18.602 13.42 13.17 17.69 24.84 14.52 39.77 16.02 22.523 16.16 24.60 13.18 17.99 11.23 11.29 16.00 18.194 26.70 23.58 19.50 18.68 11.58 11.03 24.75 20.115 34.37 24.88 33.74 14.00 8.52 3.17 27.93 13.716 4.21 1.09 1.95 0.21 1.51 0.00 5.50 3.18N.ie. 0.49 5.68 0.82 1.70 0.49 0.76 1.02 3.69

Total 10(.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00

tali

1 7.02 12.61 17.02 52.76 57.20 76.04 9.53 23.512 28.74 22.81 23.29 23.42 28.51 13.51 35.38 29.873 41.14 39.06 36.64 15.03 6.25 0.00 37.76 15.264 18.24 9.31 14.42 4.77 3.12 0.00 12.89 22.615 2.58 0.63 6.06 0.00 4.91 10.45 2.59 1.53N.i.e. 2.27 15.57 2.57 4.03 0.00 0.00 1.85 7.22

Total 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.(00

N.i.e. Not included elsewhere.Note: Small firms ale those *ith fewer- than ten emplovees; large fimns. thiose with ten or

more emplovees.

,Source: K. S. Lee (1989, tables 2-5 and 2-6). based oni World Bank-n.FNE Household Sur-vey 1978.

income residents over the past several decades and the subsequentgrowth of financial and services activities in that area. This pattern ofemployment location corresponds remarkablv closely to the residencepatterns of workers shown in the last chapter. Production workers werefound to be concentrate(d in sectors 2 and 6, and professionals werefound in sectors 7 and 8. Firms have followed the rich to some extent,and the poor have adjusted their residence locations to the resultingpattern of employment location. An examination of Cali shows theexistence of an "industrial corridor" in sector 3 and the tendency of ser-vice employment to follow the rich households south. In both Bogotaandi Cali there was little manufacturing employment in the rich residen-tial sectors. The rich succeeded in keeping their environimenit relativelypollution-free.

Again, it has been useful to study the spatial distribution of employ-ment bv both concentric rings and radial sectors. An examination byrinlgs reveals the decentralized nature of employment and the concen-tration of certain activities in the C:BD. An examination of radial sectors

126 UNDERSTAN DING THE DEVELOPING METROPOLIS

Table 6-3. Employment Distribution in Bogota by Radial Sector, FirmSize, and Major Industry Group, 1978(pelrcent)

AUa ufactaring (CSmmere,' tinan ce Ser vire.

Sm(ll Large .Small Large, Small L.argr Smaill Large

J?adial secto fi rmns firms ft rrns fiZoZn firrns firms fitms firno

I 4.66 7.00 13.13 22.57 52.14 33.99 8.77 18.6()2 14.36 4.65 14.91 1.16 2.10 1.57 10.85 4.843 24.97 14.89 18. 16 8.26 6.52 0.82 12.90 7.984 8.75 20.33 12.54 7.46 2.81 2.98 7.60 4.215 8.10 27.28 7.36 17.09 4.45 13.59 7.52 12.126 18.75 7.77 11.74 4.75 5.07 2.27 11.77 16.877 11.50 5.81 8.09 13.53 8.44 11.73 13.92 9.238 8.4'3 6.60 13.26 23.48 17.98 32.29 25.65 22.44N.i.e. 0.49 5.68 0.82 1.70 0.49 0.76 1.02 3.69

Total I00.O 1(0.0( 100.(0 10(.0( 1((1.0( 1 (.0() 10(0.0( 10.0(

N.i.e. Not iocluded elsewfiere.NAle: Si nall fi1 Ins are those mwtlh fevwer than ten emiaplor>ees; large fir ins, those w%ih te,n o-

mol-e em ployees.

Soar': K. S. Lee ( 1989, table 2-8), based or)n World Btnk-n)..NFr Hotusehol]d Surrey 1978.

shows the previously noted patterin of land use specialization and therelationishiip of employment andc residential location a little morecl early.

WVe now use the 1972 Phase 1I Survey and the 1978 W'orld Bank-DANEHiouselholcd Sur vey to observe Bogota's clhatiging structure of employ-melnt locationi (see table 6-4). The two sur-vevs were not exacllv compa-rable, but thev give an idea of the broad decentralization trend inemployment. Specifically, because of the nature of the questioni asked inthat survey, employmenit in the CBD is exaggerated for 1972. The maindifference from 1972 to 1978 is the increased importance of ring 5 in itsshare of employment. Althotigh detinitiornal patterns alluded to previ-oisly do not allow strict comparisonis, it is likely that r i1g 2 alsoincreased its share in employment over this period. We miglht say that,in effect, the (.Br) has expanded into ring 2 and that we should reallythink of both rings I and 2 as the CBD now. The constrtmctioni of the newcity center, the Centro Internacionial in the northern part of ring 2 hascontributed to this expansion. In fact, a detailed analysis of the C(BD andthe CentrLo Internacionial reveals that the sumn of employment in the twolocations remained rotgilly constanit over the period. We conclutde thatoverall the center of Bogota can be regarded as stable in absolute mag-nittide of employmenit but that its share in the total has decreasedl. It isclear, however, that manutfacturing employment has decreased in abso-lute terms as well. The 1.BD is now more and more a specialized tradeand finanicial center. It is difficult to track trends in Cali because earlier

Table 6-4. Employment Location in Bogoti, 1972 and 1978(peircellt)

.All dudtstn7es Aa Manuactl 1ig Corn ,Merce Ft nan cv .Servzcf

Ririg 1972 1978 1972 1978 1972 1978 1972 1978 1972 1978

1 23.03 13.95 18.20 6.01 19.43 1 5.75 42.11 41.43 22.62 12.912 I3.61 17.74 16.07 13.47 12.18 19.77 13.69 29.38 12.74 18.683 14.62 16.40 18.94 21.54 13.35 14.83 6.89 11.13 15.88 16.874 18.80 20.60 20.27 24.89 21.83 19.37 10.00 10(.88 20.74 23.185 18.61 24.94 21.76 28.25 21.52 27.72 14.64 5.65 17.89 21.796 1.67 3.43 1.04 2.19 3.42 1.51 (1.55 0.60 1.88 4.41N.i.e. 9.67 2.96 3.72 3.66 8.27 1.05 12.12 0.63 8.25 2.15

Total 100.00 100.00 100.0 100.00 100. 1.0 1(0.00 100.00 10(.(0 100.00

Ni.e. Not in-cluided elsewhere.ai. Iicilu(les other sectors in addition to inaiinfi actirinig. conienic ce. finance, and ser-xices.Soure: KI S. Lee (1989. table 2-10). based oni World Banik-m NE I lonsehold Stit Nec 1978 and Phase 11 Survev 1972.

I28 t'NDERSTANDING THE DEVEL.OPING METROPOLIS

data are simply not available. A comparison of social security filesbetween 1976 and 1978 suggests that the (CBD there has also lost manu-facturing jobs but has attracted more trade and finance jobs.

The 1978 survey allowed a rough tracking of firms in both Bogotaand Cali over the preceding five years. This was very uisefil, becauseinformationi on movement wvas otherwsise available only for mantifacttir-ing firms. The information is, however, inexact because it was derivedfromii the workers' own knowledge about the current and previouis loca-tion of their employers. The decentralizationi pattern was clear in bothcities. Existing ftiriis were founld generally to move outward, whereasnew firms also generally located in outer rings. In Bogota there was anet movemenit of firms out of the CBD. Net losses also occurred in ring 3.The outer rings 4 and 5 clearly gained in all activities. It was interestingto notc that ring 2 gaiiecI in coinmerce, finance, and service jobsbecatise of the expansion of the Centro Interniacional adjacent to theold (:Bt). The city center had effectively expanded and moved slightlynorthward. Outer ring 5 received about one-third of all new jobs. InCali, too, there was a net loss of jobs from the city center, but all theother rings experienced net increases in employment. Because it is asmaller city, each ring outside the (CBD was experiencing net employ-melit growth. The decentralization trend was strongest in manufactur-ing jobs in both cities. It should be noted that the decentralization trendis composed of both existing firms moving out toward the peripheryaid new firms locating towarcl the periphery. Nonmanufacturinig firmsreplaced some of the maniufactuiriing employmenit that Inoved out.

Patterns of Employment Location in Manufacturing

Given the availability of the time series data in the DANE industrial direc-tory files from 1970 to 1975, the changing location of manufacttiringjobs could be charted over this period in detail. The methods employedwere similar to the empirical studies of employment location in theUnited Kingdoin and the Unitedl States, where a typical approach wasboth to investigate not only the growth aind decline of stationary firmsbut also the location choices of newly established firms and relocatingfirms and to analvze location patterns.3

It is perhaps tiseftil to compare the manufacturing firmns in Bogotaand Cali with those in large l.S. cities. The employment dvnamics inthe Colombian cities are quite different. Reflecting the vibrant growthof manufacturing employment documented earlier, the birth rate offirms in Bogota and Cali was between 7 andl 9 percent anntially from1970 to 1975. The rate for typical American cities was more like 3 to 4percent. The death rates of firmns, however, were similar: betveen 3 ancl

FIRMS ANI) THEIR LOCATION BEHAVIOR 12 9

5 percent. Thus, both Bogota and Cali have been going through a phaseof continual increase in manufactturing establishmenits, whereas thenumber of establishrnents in American cities is now essentially stable. Itis also notable that the majority of firms that die and the ones that areborn are small. In the mid-1970s the average firm size in Rogota wasabout forty, approximately half the average for U.S. cities. The averagefirm size in Cali was a little higher, about fifty, reflecting the prescnce ofa few very large firms employinig more than 500 people. Such firmlis(with at least 500 employees) typically accouniit for 35 to 50 percent of allmanutfacttur-inig employnment in U.S. cities, while in BogotA they accountfor only about 15 percenit, and in Cali 25 percent. These figujres ignorethe employmllent in the smallest firms, those employing fewer than tenpeople.

In order to trace the spatial changes in manufacturinig employmenit,the industrial directory data were again classified by location accordingto the ring and sector framework in bothi Bogotd andl Cali. As might beexpected, the decentralization trend is confirmiied (see table 6-5). Mov-ing outward, the rate of growthl increases with each ring (except f-or aslight decline in growth rate from ring 5 to ring 6 in BogotA). Note thatin Bogota the absolute increases in manufacturing employment weresimilar in rinigs 3 and 5. It would seem that the center of the city, rings Iand 2, had become saturated, anid firms were moving otit to get largelspace. NMany moved small distances to ring 3, and others moved still far-ther otUt. Cali was more centralized, with the bulk of manuLfacturingemploymenit still concentrated in r ings 1, 2, and 3.

Similar tabulation of sectors confirmedl earlier findinigs. The industrialcorridor (in sectors 4 and 5) continued to provide abotut three-quartersof all manufacturing employment in BogotA; in Cali the inclustrial corri-dor (located in sector 3) provided about 60 percenit. There was clear- evi-dence of the increasing importance of sector 3 in sotutlhwestern Bogota,the Bosa area, where manuifacturing employment expandecl from about8 percent in 1970 to II percenit in 1975. Most of this increase is in theperiphery of the sector.

Both Bogota and Cali reveal clear land use specialization, with littlemixing of residential areas with manufacturing employment. In RogotAthe residenitial areas are predominiantly in the north (rich) and in thesouth and southwest (poor), and these areas are separated by the indtus-trial corridor in the west. Much of this patter n originally resulte(d fromlanid-tuse zoninig but has been emphasized by people's residentialchoices in recent years.

We now analyze changes by the underlying components: changes inemployment in stationary firms; birth of new firms; disappearance firojmthe directory and presumned death of firmns; and relocation of firmswithin the city between 1970 and 1975 (see table 6-fi). About 25.000 jobs

1301 U'ND)ERSTANDING THF: DEVELOPING METROPOlIS

Table 6-5. Distribution of Manufacturing Employment by Ring: Bogotiiand Cali, 1970-75

1970 1975 Annual average

growtlh rate

Rilag e'rtyOns /'erceattge 1'erMon.1 Percentage (percetlt)

B(gotei

1 4,538 5.60 4.102 3.47 -2.002 11,767 14.53 14,898 12.59 4.833 34,351 42.42 47,858 40.44 6.864 18,112 22.37 25,958 21.94 7.465 11,548 14.26 24,047 20.32 15.806 391 0.48 729 0.62 13.27N.ie. 266 0.33 741 0.63 -

Total 80,973 10.00 118,333 10(.00 7.88

(Cali

1 2,600 7.65 3,064 6.95 3.342 13,836 40.74 14,381 32.60 0.783 15,192 44.73 21,704 49.21 7.404 1,761 5.18 3,361 7.62 13.805 367 1.05 10( 0.23 -22.90N.it. 219 0.64 1,499 3.40 -

Total 33,965 10(0.(0 44,109 10(.00 5.37

- Not ;vailable.N.i.e. Not included elsewhere.Vo/e: Data are fiomn establishments with ten or more emplo)ees.Swi,rm: K. S. lee (1989. table 3-3). based onl DANE industrial directors file.

were added by the birth ofl new firms and about 15,000 new jobsresulted from the growth of stationary firms during the period. The rateof growthi in jobs in each ring is similar tor stationary and new firmsexcept in ring 2, where there is high growth in new firms and lowgrowtil in stationary firms. The total employment of firms that movedwits higher in their new locations than in their old ones.

This patterni supports the incubator hypothesis, which suggests thatsmall new firms locate in centr-alized areas where thiere are already manysimilar firm1s and where it is easy to obtain specialized essential servicessuch as rea(dy production spaces and finiancial serNices (see Hoover andVernon 1959; Struyk and James 1975). NeIw firms need the benefit ofagglomeration economileis to start. A more disaggregated spatial analysisshowedl that the incubation area was concentrated in the inner area ofthe indtustrial corridor, that is, the part of the corridor that fell in ring 2.This area was characterized by a high concentration of firm births, by ahigher- increase in employment due to births rather than to the expart-sion of existing firms, and by the small average size of new firms. Anarea with similar characteristics was also found in the Chapinero districtin noi-thwest Bogot5, in ring 2, sector 7. Cali had a similar area just

FIRMS AND THEIR LOCATION BEHAVIOR 131

south of the CBD. In both cities the average size of new firms in the outerareas was large-about 100 employees.

An examination of the firms that relocated strengthened the incuba-tor hypothesis. A large number of small firms moved out of the incubatorareas, but most moved short distances-40 percent stayed within therinig, and most of the others moved from one ring to the next iing otit.Most of the firms that moved greater distances, to rings 4 and 5, werelarge.

To understand the location patterns better, we further disaggregatedthe data by industries. Decentralization occurred in most industries,except plastics, which was already predominantly found in ring 5. As inother cities, only the printing and apparel industries were concentratedin the centet; with more than 30 percent of employment in rings I and

2.4 These industries did decentralize between 1970 and 1975 but to alesser extent than other industries. An attempt was also made to mea-suL-e the cluster-ing of industries: how close to each other do firms in thesame industry locate? It was surprising to find no particular patternamong industries. Firms do not seem to gain bv locating near similarfirms in any industry.? The evidence of agglomeration economies isthen only present in the incubator area, where small new firms gain bylocating where there are all necessary facilities.

In summary, then, we found overwhelming evidence of decentraliza-tion in manufacturing employment in Bogota and Cali in even the shortperiod of 1970-75. Firms that relocate generally move outward andexpand their size, but they usually move only short distances. New, small

Table 6-6. Changes in Manufacturing Employment by Ring: Bogota,1970-75

Annual death rate

Annualgrrowth rate Annuailiirthrate (percenitage offirnis Orignri/destinatzon

ofstationarfirnuL (percentage of 7newi going omt of ratio of relocating

Ring (percent) firnmsv) huiness)l firms"

1 3.6 4.1 2.4 ?.8

2 2.9 6.4 2.7 1.53 4.5 4.5 2.1 1.04 3.9 4.6 1.7 0.55 7.3 7.9 2.8 0.16 n.a. 17.6 2.2 1.2

Total 4.7 5.4 2.2 0.8

.a. Not appli(able.Note: Data are for firms with more than ten employees.a. Based on total 1970 manufacturing employment.b. Calculated as the ratio of the number of people employed in firms that moved oUt of

the ring to the number of people emploved in firms that moved into the ring.Source: K. S. Lee (1989, table 3-6). based on DANE industrial directory file.

132 UNDERSTANDING TIHE DEVELOPING METROPOLIS

firms locate in the incubator area-in the industrial corridor but nearthe C(BD. They replace other new firms that either succeed and move outor fail and die. Firms locating at the far periphery are usually largeenough (wvith an average of 100 employees) not to need the agglomera-tion economies of the city center. The decentralization trend is com-mon to all industries.

Factors Influencing the Location of Manufacturing Firms

To explain these changinig location patterns, a special sample survey ofBogota's manufacturing establishments was conducted in 1978, usingthe industrial directory as the sample base. Data from this survey wereused to model the location behavior of manufactur-ing firms. Theunderlying theoretical framework is that a particular type of firm willchoose a site with particular attributes that are optimal to the firm interms of such criteria as profits and costs. The determining characteris-tics of the firms include type of product, production process, buildingcharacteristics, lot size, floor space, and level of skill in the work force.Important site attributes incltide proximity to product markets and sup-pliers, comnllutinIg clistanice for employees, traiisportationt modes, andthe qLuality and availability of public utilities anic municipal services.

A sample ot 126 firms was chosen firom the roughly 2,600 firm recordsin the clirectory. Thie sample was stratified by location history (newfir ms, movers, stationar-y firms), current zone of location, type of indus-trv, and establishment size as defined by the numLLber of employees. Toobtain reasonable homogenieity among firms, the sample was focusedon textile- and metal-fabricating firms, which covered about 50 percentof total manufacturing employment in Bogota. Both industries can beregarded as relatively footloose in that they require no special site char-acteristics for their location. Being footloose, they are also presumnablymore susceptible to governmenit policy regarding location. A thirdgroLlp. "other indtistries," was aclded for descriptive variet.v. In order tomaximize coverage of employees in this limiteld sample, larger firmswere oversamnpled in the survey. The geographical coverage was good inthat twentyv-seven of the thirty-eight conmiuas, located mostly in rings 3, 4,and 5 but with some in ring 2, were covered in the survey. Of the 126firms, 50 were movers (that is, firms that were relocating), 18 were new,and 58 were stationary. The homogeneity of firms from similar indus-tries allowedl the testing of behavioral hypotheses with reasonabledegrees of freedlom.

The incubator hypothesis is supported by firms in this sample. Thenew firms were smaller than others; they worked in single-shift batchprocesses: thev were housed in relatively older buildings, often with

FIRMS AND THEIR LOCATION BEHAVIOR 133

more than one floor; and they had little land for expansion. Thus small,new firms begin operations in the more centralized locations andmove to the periphery into larger sites only after they have achievedsome success.

The use of trucks has been well documented as a major explanatoryfactor in the decentralization of employment in the United States(Hoover and Vernon 1959; Moses and Williamson 1967). The same phe-nomenon was of great importance in Bogota: 95 percent of all goodsgoing to and from firms were carried in trucks in Bogota at the time ofthis survey. This is remarkable considering that the train station in thecenter of the city was an important reason for the central location offirms until just two decades ago. The change is all the more impor-taritgiven that 40 to 50 percent of the manufacturing output of these firrns isexported outside Bogota.

To gain insight into why firms located where they did, managers wereasked to evaluate their firms' locations. On the whole, the respondenitswere very satisfied with road access, proximity to clients and suppliers,and the availability of unskilled workers, but they were dissatisfied withthe quality of municipal services, zonal amenities, the availability ofskilled workers, and the cost of land for expansion. Managers of firmsthat had moved were usually satisfied with their current plant capacity,in contrast to managers of many of the stationary firms. Reflecting theirmore centralized locations, managers of the new firms were largelyhappy with the quality of mtunicipal services. 1ndeed, that quality is yetanother reason for the centralized location of new firms. Dissatisfactionwith muriicipal and public services increased with distance from thecenter. This suggests that when firms move to the periphery they leadinfrastructure investments, which follow them with some lag. This nega-tive aspect of the new locations is traded off against advantages associ-ated with greater space.

Moves are usually associated with changes in technology and expan-sion of production; the new technology often requires more space tospread out. Thus the larger firrns move longer distances to the periph-ery, whereas smaller firms move just one or two kilometers. This is pre-sumably important in order to keep the same labor force: about 80percent of employees remain with the same firnm even after the move.The average commuting distance does not change after the move,which suggests either that firms move short distances or that employeeschange tlheir places of residence to suit their new job location.

In the firms' own evaluation of factors affecting their move, theexpansion of plant capacity was the most significant. Rent paymnents,proximity to suppliers, and zone amenities were also important. Theproximity of clients, quality of public services, and cost of land carriedless weight but did enter into their (lecisionmaking. The firms'

134 lNDERSlAND)ING TtlE DEVEILOPIN( METROPOIIS

responses suggest that they pay little attentioni to public sector decisionsconcerning infrastructure availability; they assume that infrastructturewill follow with a lag.

Modeling the Intra-Urban Location Behaviorof Manufacturing Firms

How can these firms' location decisions be modeled and estimatedeconometrically? Unfortunately, theoretical and empirical literature onfirm location behavior is quite rare, which makes it difficult to draw gen-eralizations from our work.' The general idea, however, is that becauseFirms can be assumed to be profit-maximiizing, they would be expectedto respond to the existing lanid price gradlients and any other inptztprice gradients in deciding their location. The results obtained werecertainly consistent with such a hypothesis. Although firms did notexplicitly give much importance to land prices, the clesire for plantexpansion was often given as the main reason for relocation. I.ower landprices at the periphery attract such relocating firms: hence land pricesare indirectly seen as an important variable affectilg location.

Careful considerationi was given in the City Study to developing a sys-tematic modeling framework in order to fulther the unlderstanidinig ofmanufacturing firms' location behavior. Because the fill derivation ofthe model and its detailed estimates are available elsewhere, 7 here wemerely sketch the principles behind the modeling effort and provide itsmain results.

A firm's production funlction may be written as 0 =J( l.,X;Z), where 0 =

OUtplut, L= lot size, X = vector of variable inputs such as labor and capi-tal, and Z = vector of site characteristics. These site characteristics areindependenit of lot size and represenlt local public goods at a particularlocation .

Thle relevant cost components are wages, capital costs, input materi-als' costs, delivery costs of inputs and outputs. and land rent. Thesewould be the main variables of interest for calculating the optimumcombination of inputs and for optimum site selection. Following stan-dard urbani economic theor-y based on bid rents, a particular plant sitecan be seen as being occupie(d by the firm willinig to pay the highestprice. The bid price will depend on the attractiveness of the site to a par-ticular firm.

In a locational eqtiilibrium, all firms of a particular type in an urbanarea would make the same rate of profit, and no firm would have anincentive to relocate. In this case the cost tradeoff calculations by indi-vidual businesses determine the spatial distribution of firms. For exam-ple, a large manufacturinig plant may choose a site in a low-ren-t area

FIRMS AND THEIR LOCATION BEHAVIOR 135

near the urban periphery to meet its need for more space, in spite of agreater delivery distance. Small firms, in contrast, may prefer a centrallocation where the availability of various externalities more than com-pensates for high rents.

The preceding theoretical firamework leads to an empirical fi-ame-work for predicting the probability that a firm of a particular type willoccupy a site withi particular attributes (Z). Because thie firm with thehighest bid will occupy a given site, the relevant random variable fordetermininig this probability is the maximum amount a group of firmswith similar attributes would pay (that is, the maximum bid). The prob-ability (listribtifion of the random variable associated with the maximumbid leads to a multiiioinial logit specification for the firm locationmodel. 8 Such specifications are useful in modeling any behavior dealingwithi choice.

As chiapter 8 (isCLIsses, the issue in travel demandl modeling is how tomodel an individtual's choice between different modes. Similarly, theissue in residential location is how to moclel a consumer's choice of par-ticular types of hotises and locations. Here we have the employmentlocation problem of firms having to choose among locations. An appli-cation of suchi models to residential location choice by Ellickson (1977,1981) has suggested the approach adopted in this study. He formulatedhis problem: given a site of certain characteristics, what is the probabil-ity that a h)ousehcold of a certain tvpe wvill occupy it? Our problem here issimilar: given sites of certain characteristics (Z), what is the probabilitythat certain kinds of firms will occupy it? Another approach would be todefine certain zontes in the city and to find the probability associatedwith each type of firm locating in each zone. This would necessitatedividing the city into arbitrar-y zones such as rings and sectors and find-ing the type of firm most likely to locate in each zone. The approachthat has been adopted hel-e is more general; sites are specified by moregeneral characteristics. The most important advantage of this approachis that thle model allows the observation of a wide range of spatial varia-tion in site character-istics and the sample does not hiave to be broken upby zones. This type of specification also allows policy analysis. It predictsthe probability that a particular tvpe of- firm will occupy a site with par-ticular attributes. The marginal effect of changes in these attributes onthe probability of these firms' location can also then be estimated andanialvzed. Hence, coniclusions can be drawn on the efficacy of differentkinds of policy instrumnienits that affect these attributes.

In this study, the data collected in the survey of 126 manufacturingfirms were utilizedI for estimatinig the model. The firms were groupedinto four types: small and large; textile-mantifacturing and metal-fabricating. Of the 126 firms surveyed. 87 fell into these categories.Large metal-fabricating firms were usecl as the reference group. The

136 UNDERSTANDING THE DFVELOPING METROPOLIS

effect of different location attributes on the probability of location ofthe remaining three types of firms was seen in relation to the locationattributes' effect on the reference group.

The independent variables included:

* Access to local markets for output (PRODSOLD) = percentage ofproduct sold in Bogota

* Access to local markets for inputs ([NPUTBT) = percentage of inputbought in Bogota

* Proximity to residential location of production workers (WKSOUTH)

= percentage of workers living in southern part of Bogota* Proximity to residential location of administrative workers (ADM-

NORTI-I) = percentage of workers living in northern part of Bogota* Index oF quality of local public services (ELECINT) = frequency of

electricitv interruption* Presence of agglomeration economics (LOCQT) = location quotient

defined as comuna's () share of industry (i) in relation to its sharein total manufacturing employment

* Intensity of economic activity (POPDENS) = population density ofthe comuna in which the firm was located

* Measure of accessibility to the city center (DISTCBD) = airline dis-tance from the CBI) of the center of the comuna in which the firmwas located

The estimation procedure provides measures of the influence of eachindependent variable on the probability of a firm being in the groupspecified in the dependent variable. For example, the coefficient of theproximity of firms to residential areas provides a measure of the impor-tance of this variable to small textile firms as compared to the base. Theestimated coefficients are translated into elasticities such that policyconclusionis can then be drawn. The overall goodness of fit was found tobe high, providing some confidence in the results.

The multinomial logit estimation method was used to perform theseinvestigations. It became clear that small firms are quite distinct fromlarge firms in their location behavior and that textile firms displaybehavior different from that of metal-fabricating firms. These resultsindicate the interaction between generic differences found in differentindustries according to their size, the nature of their relationship withtheir suppliers, types of product markets, and their location.

Small textile firms were the most likely to locate near the city centerand in close proximity to their input suppliers and the homes of theirproduction workers. In Bogota this meant locations in rings 2 and 3 andnearer the southern and southwestern parts of the city. Large textilefirms, in contrast, choose sites farther from the city center-about three

FIRMS AND THEIR l.OCATION BEHAVIOR 137

times farthier away, in fact-and proximity to thie homes of their profes-sional and administrative workers seems to be more important. They arealso more sensitive to locating in areas with better public services (forexample, a more reliable electricity supply). They are not sensitive toproximity to local markets and are farther away fiom dense residentialareas. Small metal-fabricating firms are similar, in many wavs, to smalltextile firms: they like to be near their inputs and product markets butare not likelv to be near the citv center. All these findings are consistentwith the idea that smaller firms need to take advantage of agglomera-tion economies that are realized by locating in areas wher-e both inputand produce markets are close by and where productioni workers areeasily available. I.arge firms, on the other hand, are oriented moretoward markets outside the city and are less (lependent on the city inwhich they are located. They tend to interinalize manv of the agglomera-tion economies that smaller firms take advantage of by being near otherfirmins.

To strenigtheni confidence in these findings, the dependenit variablecould be specified in different ways (that is, firms could be grotiped indifferenit ways). Location choice could be modeled more dlirectly bygrouping the firms by ring location. In this case, firms were dixidedl intofour grotips: rings I and 2 combined; ring 3; ring 4; and rings 5 and 6combined. As noted earlier, this was an arbitrary grouping proceduire,bLut it was tised here to bolster the results found earlier. The estimationswere found to be very robtist. Firms that sold a large proportion of' theiroutput within Bogoti tend to choose sites near the center, and export-oriented firms locate nearer the periphery. Similarly, firms that rely onlocal suppliers tend to be near the center. Moreover, locally orientedfirms tend to be small. Firms that recenitlv moved are extreme;ly unlikelyto be in or near the CBD: they almost always move to outer areas wheremore space is available, thus contributing to the decentralization ofmanufacturing employment.

All of these results imply that mantifacturing firms in Bogota behavein an optimizing fashion in choosing their location to minimize overallcosts. But we still have little sense of how responsive they are to changesin factor prices. When firms move to more peripheral locations in orderto expand, how responsive are they to the land price gradient that existsin the city (see chapter 3)? Is there also a wage gradient? Additional esti-inations were made to address these questionis.

It is clear that a land price gradient exists. The 1976-78 land pricegradient was about 0.10 to 0.12; that is, land price can be expected todecline by about 10 to 12 percent with each kilometer from the city cen-ter. The sample of 126 maniufactuL-inig firms also had direct infolrmationabout the value of the land where these firms were located. Remarkably,the gradient calculated from these observationis-about 0.11-was

138 UNDERSTANDING THE D)EVELOPING METROPOLIS

exactly consistent with the other independent data. The other majorfactor of production is labor. There are few good reasons to expect theexistence of a wage gradient in a city. Firms located in the center of acity need to pay slightly higher wages to attract workers who might havea longer commute firom outer areas. But the same justification could befound for higher wages in outlying areas. Hence there are few estimatesthat suggest the existence of clear wage gradients in cities. Nevertheless,in Bogota, even after adjusting for the usual human capital variables,wages are found to have a gradient of about 0.05 as one goes out fromthe center of the city.

Ilow do firms respond to these locational changes in factor prices?The elasticities of substitution between capital and land and betweenland and labor are good measures for this response. A low elasticitv ofcapital/land substitution would suggest a greater degree of decentrali-zation in the presence of a land price gradient. The elasticity of substi-tution measures the percentage of change in capital/land ratio inresponse to a change in factor prices-that is, the ratio of land price tocapital price. As land prices vary, then, how does the capital/land ratiovary?

Again, there were few estimates of these elasticities for manufacturingfirms in other cities. The capital/land elasticities that have been esti-mated are mostly for urban housing. They range between about 0.4 and0.8 in most cases (see McDonald 1981 for a review). In this studv we uti-lized the 84 firms (out of the 126) who were owner-occupiers, and weused the market value of plant and equipment as the capital value. Theresulting elasticity of substitution was about 0.31-much less than esti-mates made for urban housing in the United States. This result supportsthe a priori assumption that the elasticity of capital land substitution formanufacturing is likely to be smaller than for housing because the possi-bilities of locating manufacturing activity in multistoried structures arelimited. Similarly, the land/labor elasticity was calculated by using thewage bill for each firm. The value of this elasticity was also about 0.30;that is, as land price increases, firms increase the intensity of labor usedper unit of land. As expected, we see greater crowding in factories incentral areas of the city.

In summary, the r esults from all these econonmetric investigations sug-gest that firm location behavior is by no means random. The estimatedmodel was derived for profit-maximizing firms, which were found tobehave much as would be predicted. The goodness of fit was satisfac-tory, and the estimated model was capable of predicting, in probabilityterms, which types of firms are likely to occupy a site with particularcharacteristics. The location patterns that the model predicted wereconsistent with those expected. Atmong simnall firms, accessibility to localinlput and output markets was the most important factor in the location

FIRMS ANI) THEIR l.OCATION BEHAVIOR 139

decision; the benefits of accessibility to the central area compensatedfor the high rents and congestion. Large establishments, which weremore export oriented and required more plant space with modern pro-duction technology, located in outer areas, where more space was avail-able at lower cost. The estimation results also indicated that the qualityof public utility services is very important for large firms and that prox-imity to the residences of administrative workers is more important tosuch firms than is proximitv to the homes of production workers.

The precise modeling framework used here was specific to Bogota.The independent Variables chosen were those available fi-om the surveyand considered important in Bogota. Similar investigations in other cit-ies could involve different specifications and different variables suitablefor those cities. Our study has, however, indicated a inodeling frame-work that is svstematically derived from urban economic theory andcontributes to an understanding of the behavior of manufacturinigfirms.

Trade and Service Employment Patterns

It is much easier to model the location of establishments that are insome sense people-serving-for example, schools, local governmentagencies, retail trade establishments, and banks-than it is to model thelocation of manufacturing companies. Access to local markets is themain determinant of the location of employment in people-servingactivities. Considerable work has been done in this area, 9 and the mnodelused is a simple analog of Newton's Law of Gravity: the level of interac-tion is directly proportional to the mass of interacting bodies (size of thegroups of sectors) and inversely proportional to the distance betweenthem. This approach is therefore referred to as the social physicsapproach. Although the model is then not couched in terms of eco-nomic behavior, it may be interpreted as reflecting firm-optimizingbehavior. A firm chooses its location in order to optimnize its access tothe customers who are likely to demand its services. Implicit in this for-mulation is that proximiiity to inputs is less important to such firms thanproximity to customers or that prices of inputs are invariant wvith respectto location. The model framework assumes that each stubarea in themetropolis attracts the various activities in relation to its relevant loca-tion attributes. These attributes will vary with the activity group consid-ered but will include both "policy" attributes such as available land forthe activity aind "given" attributes such as current populationl or employ-ment levels and land value (Lakshmanan 1964). Only the broad outlineof the model and the main results are provided here; details are avail-able in K. S. Lee (1989).

140 UND)ERSTANDIN(; 'tHE DEVELOPING METROPOLIS

Because we did not conduct a special firm survey of firms in theseactivities at the establishment level. information was not available forthis exercise. Instead, the activity location information in the 1978Househiold Survev had to be used. From this survey we knew the loca-tion of employment of all the workers in the sample. 'I'he behavior offirms was therefore deduced from the location of workers in those tradeand service sector activities. The specific activities considered in theseexercises were wholesale trades; retail trades; restaurants and hotels;financial establishments; real estate and business services; governmentservices; sanitary services; social and commtinitv services; recreationand entertainment; and personal and household services.

The standard model repr-esents the hypothesis that the trade anld ser-vice firms in the ith zone serve people in all zones in the city but thelikelihood of service declines with distance. 10 This implies that a firm ofa particular type will locate in a zone with particular attributes.

Three meastur-es of market potential wer-e used in the estimates. Thefirst two were the numnber of hotLisehlolds and population in each zone.As might be expected, thev gave similar results. The third was puL-chlas-ing power of a zone. The dependent variable used was employment den-sitv of each clifferenit activity in each of the thirty-eight comunas inBogoti. The model was quite successftul in explaining the location ofthese tracle and service activities. The more directly people-serving anactivitv was, the better the model performned. Thus, the best fits wereobLained for reuail trade, social and commnunity services (which includededucationi), and personal and household services (which included domes-tic workers). The least satisfactory explanation was found for the loca-tion of government workers and sanitar-y service workers. This is asexpected, since these latter groups serve the public less directly.

The purchasinig-power model performed better than the populationandl household measures of market potential for all activities exceptretail trade, where it was about the same. This is consistent with the over-all employment location pattern mentioned earlier-that is, the loca-tion of maniy more employment activities in the richer parts of the city.Specificallv. the purchasing-power measure improves the degree ofexplanation significantly for- the location of employment in restaurantsand hotels, finance, recreationi and entertainment, and personal andhousehold service. Clearly, all of these activities are highly dependelnton customers with high incomes. Banks, restaurants, and domestic ser-vants are all Inore likely to be found in the richer areas of the city.

One other experiment was performed to test the degree of associa-tion between the market potential proxv and employment density ofeach zone. Sensitivity tests were performed by varving the value of g-that is. changing the decay factor of market potential by distance. If ahighel g resullts in better goodness of fit, the more localized the activity

FIRMS AND TIIEIR LOCATION BEHAVIOR 141

is. Recreation and entertainment, social and community services, per-sonal and household services, and restaurants and hotels turn out tohave significantly improved goodness of fit with an increase in g: these

activities are more dependenii on local clientele.In summary, it was again found that the location of employment in

different activities was as woulld be expected from rational profit-maximizinig kinds of behaviolr Employment in retail trades, social andcommunity services, and personal and household services follow thepatterns of residential location. Other activities, such as recreation andentertainmentt, restaurants, hotels, and finanicial services, nee( to be nearhigh-income districts of the city. Government services and wholesalersare relatively independent of the residenitial location of hiouseholds.These investigations illuminiate the interactioni between residential andemployment location and highlight the imipor-tance of lhigh-income res-idential locations.

Implications for Location Policy

This chapter emphasizes the pheniomenoni of decentralization thataccompanies the growth of cities. It focuses oni the role of the manufac-turing sector as the leading factor in the decentralization of growing cit-ies. In tryinig to uLnderstand the economic forces at work in this process,we have also begunl to better understand that certain activities arc likelyto continue to locate near the center of the city as it grows and thattherefore the (CBD is likely to remain an important concentration ofemploymnent in large cities. What do we learn from all these findingsthat could help shape policy in regard to the structure of cities?

TIhe growth and diversification of industr-y in a large city would almostinevitably lead to decentralization of the more successful segments ofindustry. The rnost important reason for firms to relocate toward theperiphery of Bogota was their desire for expansion, usually accompa-nied by techniiological change in the productioni process. Mass produc-tion requires assembly lines or continuotus flow process technology,which implies the need for single-floor plants and therefore more space.This is another way of saying that for such processes the elasticity of cap-ital/land substitution is low, and( firms move down the rent gradient tothe periphery to acquir e the larger plots they need for productioni, gain-ing space for future expansion in the bargain. Moreover, large plots ofland that are not btiili up are usually available only at the periphery.

In contrast to larger firms, smaller manufacturing firms, especiallynewer ones, still prefer locations close to the city center. They locate inpremises vacated by other firms and near to the many input services-financial institutions, raw materials suppliers, equipment repair shops,

142 UNDERSTANDING THE DEVELOPINC. MET ROPO LIS

and residences of prospective workers-that they typically require fortheir dav-to-day funictioniing. A clue to this behavior is found in the highbirth and death rates of firmns in Bogota during the period under study.New, small entr-ants into the manufacturing business are inherenitlyunstable and therefore naturally likelv to economize on overhead andfixed costs of all kinds. Many services that a larger firm would inter-nal-ize are r ented temporar ily by such a firm until it reaches greater stabil-itv. Many new entrepreneulls, for example, are technicians with littlecapital. starting out on their owin after breaking away from their originalemployers (see Cortes, Berry, and Ishaq 1987). They can afford to takefew risks. A thriving, concentrated area of manufacturing firnms thatgives agglomeration economies to budding entrants is therefore likelyto impart health to a city seeking industrial expansion.

Many urban policies are clesigned to thwart the location of new manu-facttur-inig firms in large cities. In couLntries as dlifferent from oneanotlher as Indlia, the Republic of Korea, and Venezuela, similar policiesprohibiting manufacturing firms from locating in the large metropoli-tan cities have been followed. These policies have ignored the impor-tant role of large cities as incubators of entreprenieuirslhip. The' haveresulte(d from a lack of understaniding of the locational dynamics ofindustrial growth. Prohibiting new manufacturing activity in large citiesalso has the effect of reducing effective death of sick firms. The death offirms is easier for a city's economy to sustaini if the displaced labor has agood opportuLnity to find work in new firms. If births are prohibited,resistance to death increases and the city acquires an environment ofstagnationi. Apart from enforcing pollution regulationis and( preventinghazardotus industries fromn locating in densely populated areas, it shouldbe understood that the locational dynamics of maniufactturinig firms aresuch that large firms or expanding and successful small firms will them-selves decentralize away from central cities.

It is also commonplace to find in(lustrial estates located on theperiphery of cities; these are ofteni designed to attract small firms. Theyhave seldom been successful. Our results provide persuasive explana-tions for suchi failure. Industrial estates are unllikely to become incuba-tors of entreprenLeurship until a large nulnber of successftil firms arelocated there already. Moreover, new firms cann1ol. tolerate interrup-tions in basic services like electricity, water, andl transportation. Newdevelopments in cities typically suffer from teething u'otubles in earlystages; if they do not, it is at very high investment costs. Enliglhteniedindustrial location policv in cities would therefore not aim for premna-ture decentralizationi of industrial activities; it would occur anyway withsuccessf'tl growth. The role of policy is to facilitate suCh) growthi and planfor the provision of public services in response to suchi growth. It is thelarger firms that are more likely to move to periplheral locations andf

FIRMS AND THEIR LOCATION BEHAVIOR 143

they are also more likely to be able to pay for the various necessary pub-lic services.

Attempts are also often made, at very high initial infrastructure costs,to persuade small and medium-size manufacturing firms to locate innew "satellite" cities, which are typically 30 kilometers to 100 kilometersfrom the large city. Examples of such attempts include Banweol, nearSeoul in Korea, and Kalvan, near Calcutta in India. Attempts at suchdevelopment are usually miserable failures. The actual cost of locatingin such places tisually far exceeds the benefits from any tax or otherincentives offered by the government.

The pressures for decentralization are relentless with growth, how-ever, so public policy must also be careful not to thwart it. Public trans-portation routes, for example, are typically radially oriented toward thecenter of the city, in response to the traditional location of jobs. Anexpanding city with a growing and decentralizing manufacturinig sectoralso suggests the need to instittite circumferenitial routes so that poorworkers can reachi their destination without wasteful trips through thecenter (see chapter 8).

Many of the relocating large firms in Bogota planned their moves wellbefore the availability of public services in the new location. Thereseemed to be a general belief that once they moved, public services likeroads, water supply, and power would follow. This may have been pecu-liar to Bogota, where the decentralized agencies (see chapter 9) are par-ticularly responsive to demand and have the autonomy to act. It was alsothe case that much industrial location in and around Bogota conformedto the industrial zones provided accordinig to land use regulationis. Thispoints to the value of realistic land use zoning that is itself responsive tofinms' preference and able to give public authorities the flexibility toprovide services in response to demand.

W'hat abouLt other activities? ManLifacturing, after all, seldom employsmore than one-third of the workers in a city. For most commerce andservice employmilent, we have found a pattern opposite to that of manul-factturinig. Larger firms locate much more in the center, whereas smaller-firms are more decentralized. Overall, as we also saw in chapter 3,employmenit in manufacturing is more decentralized than in servicesand commerce. Indeed, it can be said that the CRD is specializing moreand more in nonmanufacturing employment, and the periphery, inmanufactur-inig. The reason, although this has not been determiieclempirically, seems to be that office activities, trade activities, and othersimilar activities are likely to have a higher capital/land substitutionelasticity. It is easv to pile up such activities in multistory buildinigs. Atthe same time, many activities that are directly people-serving, sucIh assmall retail stores, other personal services, local goverriment services,and schools, hiave to be near their- customiiers and therefore decenitralize

1 44 UNDERSTAND)ING THF. DEVELOPING METROPOLIS

as their customers do. Residential location and manufactturing locationtherefore lead these other activities in their decentralization pattern asa city grows. But most of these activities are sensitive more to purchasingpower rather than to mere mass of people, so they are much more likelyto be found in the richer areas than in the poorer ones. In fact, themany retail activities found in the C:BD are located there because of thelack of effective demand in the many poor areas of cities in developingcountries. A poor neighborhood cannot support many activities. Thisexplains why some activities are likely to continiue to be more central-ized in cities in developing countnes until incomes increase in thosecountr-ies and cities.

It has also been observed that competing peripheral "city centers"emerge as a city expands. The indication is that the relative poverty ofcities in developing countries makes this process slow, however. Cityplanners often prematturely plan competing centers for retail, whole-sale, office, and other activities, whereas they attempt to restrict theaccretion of new activitv. in the traditional city center. Our study suggeststhat this should be done onlv after careful consideration of the popula-tion's demand for such activities. Otherwise, much inefficiency canresult from the loss of agglonmeration economies inherent in the citycenter.

These observations illustrate the strong linkage between the patternof residence location, the spatial distribution of income, and the loca-tioin of employvment, and show how their mutual interaction producesthe city structure. The spatial inequality of incomes influences the distri-bution of' emplovyment location and explains the lack of employment inthe poor areas of Bogota. Once again, the effect on the design of publictransportation is clear. We can better predict the course of a city'sgrowth once we understand these linkages.

Notes

1. This chapter summarizes the work reported in K. S. Lee (1989).2. The traditional commercial district (Chapinero area) in sector 7 used to be

the "Fifth Avenue" of Bogota; in recent years, however, retail stores and restau-rants have sprung tip along 15th Avenue northward, and a large shopping cen-ter (called Unicentro) was developed in the northern part of sector 8. Thedeterminanits of commercial and service employment location were also testedusing a gravity model. For details, see K. S. Lee (1989. chapter 6).

3. As reviewed in Kemper (1973). See also Struvk and James (1975),Hanushek and Quigley (1978), Leone (1971), Schmenner (1973, 1982), andCameron (1973).

4. See Henderson (1988) for a broader discussioni of the kinds of firms thatare found to continue in large cities.

FIRMS ANI) THP IR I OCAFI ON R.HAVIOR 145

5. Henderson (1988) r eached a similar conclusion.

6. Works on this subject as it applies in the United States include Hlanushekand Song (1978), Erikson and Waysleniko (1980), Schneniner (1973, 1982), andCarlton (1979, 1983).

7. The derivation was first provided in K. S. L.ee (1982) and is also available inK. S. Lee (1989, chapter 5).

8. The application of the multinomial logit method to urbani economicrcsearch became popular withi McFadden's work on travel demand studies(1973, 1974. 1976) and wtith work by Friedman (1975), Lerman (1977). andQuigley (1976) on residential location studies.

9. There was muci initerest in the location of'service-oriented establishmentsin the United States in the earlv 1960s (for example, Lowrv 1964; Lakshmanan1965; W. (;. Hansenl 1959; W. B. Hansen 1961: Pendletoni 1963; Huff 1961).

10. Details of the model may be found in K. S. Lee (1989. chapter 6). Briefly,the model f'ramework is as follows:

kE f(G,, )

where Ek = number ot'jobs in activity k in zone i; G1 = a gravity measul-e of themarket potential of zone i; and ET = the total nuImber ofjobs in zonie i.

The rneasure of m'iarket potential can be expressed as

z(; = , i, j = .I

dg

where Z j = proxy fotr market poten tial of zonie i with respect to zone; dg = dis-tance between zonie i and zonC]: and g= distance decav weight.

If, for example, we measure market proxy by the population in each zoniC,and( if there are, say, three zones in the city and g= 2, thenl:

p, PP2P I P.2I= _ + +

2 2

dil d12 " 143

where d, I is a measure of the size of zonie I-for example, the radius of zonie 1.GI¾ is then a measure of interaction betweeni zonie I and all other zolnes in thecity: the influence of each zone on zone I de(reases by the square of the dis-tance between the two zonies. A retail store, for example, is likely to colisider thepeople in nearby neighbor-hoods much more important than those in far-off'neighborhoods.

Chapter 7

Shelter in a Growing City

Colombia has been exceptional in the level of importance it hasassigned to housing throughout the past several decades. Although thelack of adequate shelter is seen as a serious social problem in mostdeveloping countries, higher priority has generally been given to otherpressing economic issues, such as agricultural development, povertyalleviation, nutrition, and industrial development. In Colombia the gov-ernment consciously attempted to use housing as a leading sector topromote economic development in the late 1960s and early 1970s.Moreover, it is probable that Colombia was one of the first developingcountrlies to create an organiized system of housillg finance. It is there-fore of great interest to assess the housing situation in Bogota and toexamine the efficacy of different housing policies.

A country undergoing rapid urbanizationl faces special difficulties inexpandinig housing supply quickly enough to keep pace witlh thie butr-geoning demanid. In countries that are already highly urbanized, theannual increase in housing demand is essentially a combination ofgrowth in the number of households and the need to replace depreci-ated housing. In countries with extremely low levels and growth rates ofurbanization, similarly low growth rates of housing demand are encoun-tered. It is in the large group of middle-income and rapidly urbanizingcountries that the problem of adequate supply of urbani housing is mostserious. Because of the movement of hotiseholds from rural habitats tourban areas, and the unusually high rates of growth in income duringsuch periods, the increase in demand for urban housing far exceeds therate of growth in the total number of households in the country. Notonly do materials have to be provided in adequate quantity, but land hasto be developed and infrastructure services provided at an acceleratedpace. At the same time, there is a natural tendency for the price of eachof these commodities anid services to rise; failing this, the public sector

14fi

SHELTER IN A GROWING; CITY 147

tends to ration these services. The expansion of infrastructure and pub-lic services is typically a responsibility of the government or public utili-ties, and they are often hampered by a lack of adequate resources. Evenif the pricing of public resources is appropriate, such activities still needexternal resources for investment purposes. Hence, there is typically alag between the expression of housing demand and the supply of thefull range of housing and infrastructure services needed to satisfy suchdemand. The result is pressure on prices and/or rationing.

Colombia experienced this type of expansion from the early 1940s tothe late 1970s. Overall, Bogota may have been exceptional in copingwith the problems usually associated with this phase of development.This has happened somewhat fortuitously as private hiouseholds, privatedevelopers, public utilities, local governments, and the national governi-ment have separately responded with relatively innovative and flexiblesolutions in the face of adversity. Between 1964 and 1978, for example,the stupply of housing in Bogota grew faster than the population did,leading to improved dwelling conditions. The housing stock in Bogotaremained crowded, however: the average number of houselholds perdwelling unit was 1.4 in 1978, a small increase from 1.3 in 1964. Despitethis increase, there were actually fewer persons per dwelling unit (adecline from about 8 to 7) because of a precipitous drop in averagehotusehold size. This fell from 6.2 members in 1964 to 5.2 in 1973 and4.9 in 1978.

The traditional approach to housing has been to estimate a housingdeficit or shortage. "Quantitative deficit refers to the material lack ofhousing, the result of subtracting the existing 'adequate' houses fromthe total number of families in the city. Qualitative deficit refers to thosehouses which fail to meet habitability requirements and are then classi-fied as subnormal (dirt floor, lack of water supply, lack of sewerage sys-tem and the like)" (Stevenson 1984, p. 2). This kind of approachtypically results in a massive estimate of a housing shortage that has littlehope of being met in the immediate or foreseeable future. Such an esti-mate paralyzes policvmakers into complete inaction.

Housing in Bogota: The Institutional Setting

The main public institution responsible for housing in Colombia hasbeen the Land Credit Institute (Instituto de Credito Territorial, or IC1).Originally founded in 1939 to help finance sanitary housing for field-workers in rural areas, it became a primarily urbani housing institutionin 1956 when the rural housing functions were transferred to the Agrar-ian Credit Bank. Its approach has been essentially a physical, construction-oriented one, initiating major projects in the largel and medium-size

148 UNDERSTANDING THE DEVELOI'ING METROPOIIS

cities. It has always attempted to promote organized physical urbandevelopment planning, but it has met with only limited success.Although its professed aim has been to provide mainly low-incomehousing, its own periodic assessments indicated that it felt unable to doso. For example, in a document issued by the ICT in 1955, it was esti-mated that 83 percent of all Colombian families could not pay for ade-quate housing (ICT 1955). Furthermore, the "housing deficit" wasprojected to increase continuLously as the city grew. Although in itsreport the ICT recognized the problems of delivering fully finished unitsto low-income households, its actual approach remained one of build-ing units for sale or rental.

The ICT achieved a degree of success during the early 1960s, whenlow-cost fuLnds were r eadily available from the United States through theAlliance for Progress started by the Kennredy administration. Largehousing programs such as Ciudad Kennedy, Timiza, La Esmeralda, andGarces Navas were implemented in Bogota at that time, but the pro-grams were halted when the funds dried up. These programs onlyserved the elite among low- and lower-middle-income groups, and eventhen the ICT suffered from low collection rates and erosion in itsresource base because of inflation anid large increases in building costs.As a result, the ICT suffered from constant problems of decapitalization.

Thr-oughout most of its operations, until the mid-1970s, the ICT builtabout 25 to 35 percent of all its units in Bogota, accounting for 25 to 50percent of its investments. The greatest number of units built in Bogotawas about 10,000 in 1962; it varied betveen 3,000 and 5,000 throughoutthe late 1960s up to the mid-1970s, a period when the number of house-holds was increasing by about 40,000 to 50,000 annuially in Bogota. Itwas calculated in 1971 that, according to the ICT's repayment require-menits, about 60 percent of families in Bogota simply would not qualifyfor loans from the ICT (Stevenson 1984). Other studies reached similarconclusions during the 1970s: more than half of Bogota's householdswere ineligible for the cheapest public housing (Valenzuela 1970; Ver-nez 1976; A-rias 1977).

All accounts suggest that the ICT's ptiblic housing programs are likelyto have reached groups in the sixtieth to ninetieth percentiles ofincome distribution. Thus, despite its explicit mandate to serve thehousing needs of low-income, urban residents, the ICr was not able toreachi the poor. Its financial base was never well structured; forced chan-neling of funds at low cost formed the main portion of its financialresources, and this was never a secure and self-sustaining base. Oneobserver concluded that, in an attempt to reach the elusive low-incomemarket, the ICT had adopted a strategy of building verv small shoddilyconstructed units on land far from the center (1Laun 1977).

SHELTER IN A GROWING CITY 149

The other main housing finance institution that has existed for a longtime is the Banco Central Hipotecario (BC.H). which was founded in1932, replacing an earlier entity, the Central Mortgage Bank of Bogota,founded in 1905. The BCH has administered the mortgage bond marketand raised funds from other instruments such as social security bondsand indexed bonds. It was the main financial institution operating as asavings and loan institution for housing loans to builders as well as con-sumers. Until 1972, when the indexation savings and loan system forhousing came into being, the BCH essentially catered to the top 10 to 15percent of the income distribution, accounting for about 4,000 to 9,000housing loans annually in Bogota in the late 1960s and early 1970s. Thisconstituted about 35 percent of its loans nationwide and 40 to 50 per-cent of the total amount loaned. Through the 1960s and early 1970s itsaverage loan amounted to between Col$250,000 and Col$300,000 (1975pesos). Until 1972 the BCH typically accouLnted for 70 to 80 percent of allformal housing credit in Colombia, with the [CT accountifrg for most ofthe rest. Since the institution of indexed mortgages in 1972, the BCH hasshifted its lending to low- and middle-irncome households.

The big change in housing policies came in 1972 with the announce-ment of the Cuatro Estrategias, or the "Four Strategies" (DNP 1972).Urban construction was to be given the highest priority, and housingwould be treated as a leading agent or promoter of development. Thegovernment believed that housing construction would not be import-intensive and that its expansion would help promote demand fordomestic industrial and agricultural products as well as for transporta-tion and services. The main innovation was the introduction of theUtnidad de Poder Adquisitivo Constante (UPA(.) system of indexeddeposits and mortgages. It financed mortgages at positive ex ante realinterest rates and paid depositors a similarly indexed returni. Thesefinancial flows were to be routed through new Caja de Ahorro y Vivi-enda (Savings and Loan Associations) regulated by the Fondo deAhorro y Vivienda createcl by the Banco de la Republica, the central bank.

This was a major policy initiative and had wider ramifications for thefinancial sector in Colombia. Because Colombia has typically had regu-larly high rates of inflation (20 to 30 percent per vear), the availability ofan indexed finanicial instrument also provided an avenue of safe invest-ment for funds that wotuld otherwise have gone into other inflation-proof investments, such as land. Although investrnents in housingthrough formal channels did increase substantially as a result of thisreformn, it is not clear how much difference it made in overall housinginvestment. Some of the shift may have been from the informal housingmarkets to the formal markets. As will be seen, the poor are not gener-ally able to avail themselves of these formal channels. This policy initia-

150 UNDERSTANDING THE DEVELOPING METROPOLIS

tive did serve to indicate forcefully that the government was concernedabout raising housing investmenit in the country. One result of this pol-icy initiative was that the funds controlled by the BCH fell dramatically,and most of the formal housing credit (70 to 80 percent) began to bechanneled through these savings and loan associations. Total invest-ment in housing through formal channels increased from aboutCol$3.2 billion in 1972 to Col$10.3 billion in 1976. It has been arguedby some that this system was even more effective in denying low-incomefamilies access to formal housing credit (see, for example, Stevenson1984). There was clearly a construction boom in Bogota between 1973and 1975, but there is little evidence of any longer-lastinig effect onincreased housing investment rates.

The new development plan for 1975-78, Para Cerrar la Brecha, didnot mounat anyi new specific initiatives on housing, but it shifted theemphasis toward providing public services and encoturaging communityparticipation. On the private supply side, there appeared to be a com-petitive, legal, housing constructioni business supplying housing tomiddle- and higher-inconme groups. The ten largest firms supplied lessthan one-quarter of the units offered for sale on the legal market. In1969 the constructioni sector in Bogota included 741 companies, ofwhich 40 percent were individually owrned, 41 percent were limited lia-bility companies, and 3 percent were corporations. Of these companies,35 percent were very small, employing fewer than 10 workers, while thetop 5 percent had more than 200 workers each. About fifty firmsaccounlted for 20 percent of the total construction market in Bogota. Ofthe total construction, about 80 percent was for housinig. Thus, therewas a relatively vibrant housillg construction businiess sector that did riotseem to be concentrated and hence could be expected to operate rela-tivelv competitively.

Formal institutional initiatives had very little impact on the housingof low-income farnilies in Bogota. The vast inajority of households in thebottom 60 percent of income distribution had little or no access to theformal housinlg market, which is characterized by formal tenure, formalcredit systems, and the concept of a house as a fully finished, deliveredunit. Yet the quality and qtLantity of housing in Bogota as it existed bythe end of the 1970s was relatively good for the level of per capitainconiie that Colombia had achieved. More interestingly, there was evena hiigh level of public service in most housing settlements. Most infra-structure investinents typically followed the formation of settlements,but people seem to have felt reasonably sure that such investmentswotild appear at some point.

How did this happen? Most of the rest of this chapter is devoted todescribing and analyzing the housing responses in the unorganizedmarlket, which resulted in a relatively high level of housing services in

SHELTER IN A (;ROWING CITY 151

Bogota. This has been made possible by unbundlinig the usual bundleassociated with housing. Most households could not qualify for evensubsidized low-income housing because the unit price of the completehome was beyond their reach. The cost of a house consists of five maincomponents: the cost of land, the cost of building materials, the cost ofthe labor that goes into construction, the cost of obtaining access toinfrastructure and other public utilities, and the cost of obtaining accessto housing financing.

The informal housing market helped households reduce the accessprice of each of these components. At the time of occupation, land wasoften only semideveloped and hence much cheaper than a fullv devel-oped plot. Furthermore, many families moved into semifinished units atfirst. At the same time, building materials used (including recycledmaterials) were also often cheaper. Hence, at the time of entry, the unitcost of the finished unit was also reduced. Expenditure on wages wasreduced through full or partial use of unpaid family labor. A plot wasusually occupied before public services like paved lanes, water supply,and sanitation services had reached the subdivision (alternatively, utili-ties were tapped into illegally). Again, the effect was to reduce the priceof public service access at the beginninig. The developers themselvesusuallv provided credit for land purchase, thereby openinig access tofinanicinig for low-incomiie households. The process of increiienltal devel-opment then upgraded the dwelling unit, with the household buildingup equity in the house in the process. The incremental constructioni wasoften financed by income gained from the rental of rooms.

The informal nature of the supply' side helped in this process. Thedevelopers obtained largely undeveloped land at low prices and thenpassed on semideveloped lots to the consumers while financinig a signif-icant portion of the lot price; their intermediation and transaction costswere probably lower than a formal credit institution's would be. Theywere able to do the financing because they were reaping the uniearnedgainis in land values inherenit in any land development business; butthey clearly shared a portion of this with the lot buyers. In this process,the effective cost of financinig was reduced for the buvers.

The public sector also participated in this process by providing accessto public services, albeit with some delay and by tolerating a certainlevel of illegal activity. The public sector itself unbundled its investmentactivities in infrastructure through agency decentralization. As will beseen, there was explicit recognitioni of the need to lower infrastrLucturestandards in order to make housing affordable to the poor: this wasdemonstrated through the launchinig of the Normas Minirnas program(discussed below). Even this proved too expensive for the poor, how-ever. In general, public infrastructure services were extended inresponse to demand and as the need arose. Demand was often

152 l'NDERSTANDING THE DEVELOPING METROPOLIS

expressed through political means at the time of local elections orthrough particularly influential or vociferous local representatives. Thisinstituitionial process of decentralized and relatively tinplanned develop-ment (see chapter 9) can be seen as having promoted more rapid hoLIs-ing (levelopment in Bogota than in many other cities, where morecomprehensive attempts are made to deliver fullv completed housingunits with a full complement of infrastructure services.

This process effectively enabled people in the low- and lower-middle-income categories (perhaps the thirtieth to seventieth percentiles ofincome distribution) to enter the housinig market. The poorest were stillunable to participate as homeowners, but they were able to live as rent-ers in the dwellings of those slightly better off.

Maniy of the first-time entrants in the housing market among the lesswell-off consist of inquilinatos: usually single people with uncertainemployment who live as roomers in other low-income households. Astheir employment situation improves and as they are able to form fami-lies or to brinlg their families from rutral areas or smaller towns, theymove to other rental dwellings. Much of the older housing stock in thecity center has also been subdivided to provide for such low-incomeroomers. If their income levels improve substantially, they are then ableto become homeowners, typically in an urbanizacion pirala, or pirate sub-divisiom.

ele examine this process by (a) analyzing in detail the extralegal hous-ing settlements in Bogota known as barrios piratas, (b) looking at themobility and tenure patterns of hotisehiolds, and (c) modeling housingdemand behavior in BogotA. The 1978 World Bank-DAINE HouseholdSurvey information was supplemented by information from special datacollected from pirata developers (urbanizadores piralas) thernselves, froma sample of low-income households, and firom a special mobility survey.

Bogota's Unregulated Housing Market: The Mythsand Realities of Incremental Development

The gap between the conventionally estimated housing supply anddemand has been successfully filled in Bogota through the widespreadsystem of pirata subdivisions. This system is different from the widelyobserved squatter settlements in other Latin American cities and inother developing countries. The difference is that these pirata subdivi-sion settlements did not r esult from land invasions: the land has actuallychanged hands through legal purchases. It is the subdivision itself that isusually illegal. BLIt these settlements are better described as extralegalrather than illegal. Low-, lower-middle-, and middle-income families,havinig been shut out of the formal housing market, buy lots from entre-

SHELTER IN A GROWING CITY IS 3

preneurs who acquire tracts of undeveloped land and subdivide themwithout conforming to zoning laws, subdivision regulations, or serviceprovision standards. The lots sold usually provide only a bare minimumof services, often nothing more than some streets and water standposts.Typically, this rudimentary infrastructure is incrementally upgradedafter initial settlement has taken place. It was found that buyers typicallymade down payments of about one-quarter to one-third of the lot priceand then paid off the balance in monthly installments over one to fouryears. The pirate developer himself usually financed these installments.

There have been no accurate land records in Bogota to indicate theextent of pirata-supplied housing. Most estimates suggest that 30 to 40percent of the total residential land in Bogota has been developed inthis fashion; this implies that about 50 to 60 percent of the city's dwell-ing units have been constructed incrementally in these developments-residential density in these areas is higher than in other areas of the citv(DAPD 1980; Vernez 1973; Borrero and Sanchez 1973). Our- own house-hold survey, in 1978 indicated that betveen 1971 and 1976, abotut 60percent of all construction was unlicensed. According to Valenzuela andVernez (1974), in the early 1970s, 45 percent of households in BogotAlived in units developed as pirata developments, about 43 percent inlegal private developments, 11 percent in public developments, andonly I percent in squatter developments. Various studies corroborate thenotion that pirata developments cater mainly to low- and lower-middle-income households. For exarnple, Vernez (1973) calculated that thesedwellings provided shelter for only 28 percent of the poorest 10 percentof households in Bogota, about 75 percent of the next quarter of house-holds, about 65 percent for the next quar-ter, and about 10 percent ofthe top 40 percent.

Many of the originally illegal pirata developments were eventuallylegalized. By 1980 both the area covered by illegal development and thepercentage of the population living in unauthorized settlements haddeclined substantially. \Various estimates, including official estimatesmade by the District of BogotA, suggest that only 10 to 15 percent ofhouseholds lived in pirata developments in 1980. There is also evidencethat new residential land development in the late 1970s had increasingproportions of legal development. This may have resulted from both aslowed citv growth rate and an accelerated process of legalization.

Government Response

The significance of pirata development was recognized increasingly bythe government starting in the early 1970s. One governrnental responsewas the iVormas Minimas approach to development (see Consultecnicos1971). Normas Minimas are planned and approved "sites-and-services"

154 UNDERSTANDING; THE DEVELOPING METROPOLIS

projects. The idea behind such projects is to provide a developed site,often with a core sanitary structure, along with full public services, andto let the household do its own unit construction. Regtilations wereamended in 1972 to lower minimum design standards for residentialdevelopmenits and to allow private sector participation in these deve-lopments. The idea was to replicate the incremental development pat-tern of pirata developments in a legal and more organized manner,with a higher level of public services. The implementation of this newconcept face d many obstacles. First, the approval process itself was verycumbersome, consisting of two stages of review and approval. Between1973 and 1977, for example, only 28 Normas hlinimas subdivisionswere giveni final approval-out of more than 250 applications. Second,there was a lack of land zoned as high-density residential within theurban perimeter, and such land was the only kind eligible for NormasMinimnasdevelopment (Paredes 1984). Third, there was social opposi-tion to the zoning of such land because it was seen as legitimizing slums.Fourth, the approval process took one to two vears-long enough to dis-courage many would-be developers. The simplification of the approvalprocess from two stages to one improved matters somewhat and in 1979allowed developers to begin selling lots before making infrastructureinvestments.

Under laws passed in 1963 and 1972, the district government has pur-sued a policy of systematically legalizing unauthorized subdivisions. Thelegalization process includes upgrading public services with contribu-tionis coming in different proportions from the government, develop-ers, and the community itself. Since 1972 the legalization process hasalso inchlded the installation of provisional sources of wvater supply andother public services such as road lighting. That the district governmenthad a relatively positive attitude to the legalizationi of these develop-ments is signified by the fact that the government was required to payfor these emergency services.

The legalization process involved reaching an agreement between theoriginal developer or the neighborhood association and the district.This agreement specified a subdivision plan and the provision of a num-ber of collective services. Although the developer was legally required toprovide some minimum services, in practice there were loopholes thatpermitted the developer to plead ignorance of these provisions and yetbe legalized. For subdivisions created after 1972, there was an effort inthe late 1970s to enforce stricter standards and greater financial contri-butions from developers for the infrastr-uctur-e provision. In practice,this did not prove too effective. Although legalization was supposed to takeplace only if certain minimum standards in public services had beenmet, legalization was actually much more liberal: in effect, a subdivisionbecame legal if the district government simply pronounced it as such.

SHELTER IN A GROWING CITY 155

Once the subdivision was legalized, it became feasible for lot ownersto get legal lot titles, but this process was made very difficult by the legalrequirements. Once the lot purchaser had paid for the lot in full, hecould have the title transferred fi-om the developer. But he was requiredto have the following documents: a draft contract of sale; evidence ofpossession of title by the seller; income tax payment certificates of boththe buyer and the seller; a property tax payment certificate; a mu-nicipaltax payment certificate; citizenship papers; registry of certificates; and,for males over eighteen, a certificate of military service. As might beexpected, it was extremely difficult for low-income househol(is to assem-ble all these documenits; moreover, it was expensive. As a result, few peo-ple had legal titles to their lots, even if they were entitled to them. Thecontiact documrients between the buyer and seller (promesa de venta,escritura de comtpraventa) became recognized as adequate for extralegallot sales. D)espite the legal obstacles, considerable tenure stability wasachieved in these informal imnarkets, and lot sales were surprisiligiv brisk,despite the extralegality of the system. The main disadvantage, ofcourse, was that the lack of lcgal title still impeded access to the orga-nized housing finanlce system, which requlired legal mortgages. Becausehouseholds usually possess at least one document giving proof of owner-ship, though with varying degrees of legal recogniitioni, the fully legaltitle has little effective valie, except for those owners who want access tofinancinlg from the official sector.

These legal and economic obstaclcs to authorized development havegiven rise to widespread pirata development in Bogota. The need forshelter has inspired innovative responses essential to cope with therequirements of a rapiIly expanding city. Unlike many other cities,where land invasions are encouraged bv extensive government owner-ship of land, in Bogota there has been a large supply of privately ownedland on the periphery of the city. The original owners of this ustiallyrural/agricultural land were induced to sell to urbanizadores pirata for anumber of reasons: the developers were often offering the best price;legal transfer procedures took longer than sales to pirata developers; orthe land might not havc been zonied for developmcnt. Piratas have alsobeen known to coerce landowners with the threat or fact of squatting.

The developer sells the lots on an installment plan. Lot buyers areustially first-time entrants in the housing market, with limited resources;they are often renters who have been in Bogota for some time. Thetransfer of the lot typically takes place without an adequate supply ofutilities. Most buyers do not havc access to the organized credit marketand therefore have to resort to financing from the developers them-selves. The terms of such financing are rarely explicit, and it was seldompossible to find the cash price anid instal[lmienit price of lots separately, soestimates of the implied ratc of interest could only be calculated indi-

156 UNDERSTANDING THE DEVLLOPING METROPOLIS

rectlv. Carroll (1980) estimated these implied rates of interest fromhedonic price equations at the aggregate subdivision level and from dis-aggregated data at the lot level. The aggregate estimate was 46 percent,whereas the estimate derived from lot level data was 13 percent. Thelower estimate was considered more reliable. Given average inflationrates of 20 to 25 percent, it seemed that the lot financing was done atrelatively low or even negative rates of interest by the developers.

Once the lot is bought, the structure is built increinentally accordingto available resources and the housing needs of a growing household.The resources are commonly augmented by renting a room or two torenters, and family members and friends provide significant unpaidlabor. Some labor, mostly for specialized tasks, is hired; the extentdepends on the family's circumstances.

Amparo de Ardila conducted a special survev of 212 representativelow-income hiouseholds for the City Study in 1978 (see the appendix fordetails of this data set). Of these households, about tvo-thirds were liv-ing in substantial structuLes of three to six rooms. About 90 percent ofthese households had been tenanits prior to their current residence; andtwo-thirds had lived in one-room quarters. This clearly indicates theimprovements achieved by means of pirata development. The economicnecessity of entering the housing market by way of pirata developmentcan also be illustrated by citing actual costs of acquiring land and build-ing in 1978: a 120-square-meter lot would have cost about Col$60,000; amodest 40-square-meter structure would have cost about Col$80,000;and thie cost of full utility services would have added aniotherCol$20,000, for a total of about Col$160,000. The median householdmonthly income was about Col$8,000 in Bogota, and the medianmonthly income among the households in the low-income homeownerssurvey was Col$5,000. Such a household would have to pay about eightmonths' salary as a 25 percent down payment and then about two-thirdsof total monthly income in installmenits if these were paid over threeyears-and without interest. Given that such a household would tVpi-cally spend almost 60 percent of its income on food alone, this would beclearly impossible. Hence, the only way to enter the market is throughincremental development. The household typically buys an unservicedlot and resorts to clandestine utility hooktups to save on utility connec-tion as well as running costs. Occupancy of the lot would then typicallybe delayed to build up resources for building the structure, which wouldbe a modest one initially. It is also common to delay installment pay-ments; the developers, on the other side, have to factor these eventuali-ties into their lot prices.

According to the survey, two-thirds of recent lot occupants reportedno official connection to the wvater netwsrk, yet two-thirds of the samegroup reported access to piped water; sinmilarly, 90 percent had nio of fi-

SHELTER IN A GROWING CITY 157

cial power connection, yet 90 percent had actual connections; and 80percent had no official sewer connections, yet almost half had toiletsconnected to a septic tank or sewer system. The discrepancies seem tobe excessive. Part of this may have been the result of imprecise question-ing. For example, those repor-tinig "access" to piped water supply do notnecessarily have an individual water connection in their home but couldhave effective access to public standposts. ln any case, these data indi-cate the ways in which low-income households achieve effective accessto public services. Even if many connections are illegal to begin with,they tend to be regularized later with appropriate payments to the utilitycompany. This may have been why the utility companies tolerated illegalconnections to begin with; effective policing would, in any case, be toodifficult. Overall, it was found that water and electricity were widelyavailable in Bogota, and some kind of road existed in most areas. Onlythe sewerage system was slow to catch up.

About one-quarter of households were found to have delayed occu-pancy of their lot by a year or more. Missed or delayed installment pay-ments were common among one-third to half of all households, as wasthe practice of building only small core structures initially. Resales oflots were common-betweeni one-third and half of current homeownershad bought their lot from prior lot owners rather tlhani fronm the devel-opers themselves.

It is interesting that the mean size of vacant lots acquired in these sub-divisions was about 160 square meters, quite a substantial lot size for low-income households. The initial structure built was tvpically about 25square meters or less, leaving considerable scope for expansion. Abouttwo-thirds of the households in the 1978 survey reported some form ofincremental construction. Almost half of the reporting households didthis incremental constructioln using only unpaid labor; one-fifth usedpaid labor only and another fifth used both paid and unpaid labor. Asmight be expected, the tendency to use paid labor was higher amonghouseholds that had higher incomes or were headed by older people orfemales. The major part of financinig for construction comes fromsavings or sale of other assets. But almost half of the householdsreported reliance on unspecified loans and cesantras, as compared withonly a third who used these sources to finance their purchase of thevacant lot. 1

Because pirata development has formed such a large part of Bogota'shousing supplv, it is important to understand the workings of this mar-ket. We were fortunate in gaining access to a special survev conductedby the Superintendencia Bancaria in 1977.2 About 120 pirata developerswere questioned about approximately 150 subdivisions (see the appen-dix and Carroll 1980). Data included acquisition price of the land tract,when the land was bought, the time profile andl prices of lot sales, ancd

158 UNDERSTANDING THE DEVElOPING METROPOLIS

public services and infrastructure installation. fnformation was alsoobtained about the Normas Minimas developments. These data wereused to asses the rate of return to pirata developers, to compare thepirata and Normas Minimas developments, and to understand the deter-minanits of lot prices.

As noted earlier, the Normas Mlinimas subdivisions were intended tosupplemenit and preferably replace pirata developments. That this didnot work in practice is illustrated by the comparative data obtained fromthis SUrvey (see table 7-1). On average, the hormas Minimas lot wvas signif-icantly smaller than the pirata lot but substantially more expensive persquare meter. Whereas the average down payment (1976 prices) for aNormas Mfinimas lot was Col$14,700 and the monthly installment wasCol$700, for a pirata lot they were Col$9,500 and Col$615, respectively.Thus, Normas Minimas developments catered to somewhat higher-income groups; for the lower-income households, access to the housingmarket still had to be through the pirata market. Because the NormasMinimas developments had to follow formal (although considerablydiluted from the normal) planning standards, thev typically devoted

Table 7-1. Average Data for Pirata and Normas Mini mas Subdivisions,Mid-1970s

Pirata iriside oIVrmas Pirata oLttsideCharacteristir (average) uirban perimneter Minmmas' urban perimeter

Tract price per square tneterb(19X6 pesos) 50 64 18

Tract sizeb (square meters) 35,528 98.700 84,758Number of lots 156 585 150Lot size (square meters) 125 92 315Lot price per square meter

(1976 pesos) 253 456 109Total lot price' (1976 pesos) 31,625 41,952 23,008Subdivider expenditure

(1976 pesos) on infrastructtireper salable square meter(usable area) 22 64 9

Subdivider expendituire on infra-structure per lot ( 1976 pesos) 3,256 5,955 3,695

Percentage of subdivision in openspace 10 22 10

Open space per lot (square meters) 26 39 38Total number of subdivisions 109 14 24Total number of lots subdivided 16,994 8,191 3,604Total number of lots with sale data 11,540 5,916 2,099

a. All iVarmas Minimas subdirisions are inside the urban perimeter.b. "Tr-act" refers to the whole parcel of land that is pur-chased for subdivision.c. Total lot price is the sum of undiscounted inistallments plis the down payment, per

standaid practice in the pirata subdivision business.Soiir: CCarroll (1980).

SHEI.TER IN A GROWING CITY 159

about 20 percent of their area to communal and green areas-aboutdouble the proportion for pirata developments. Accounting for another20 percent devoted to streets, that left about 55 to 60 percent for salablearea in the ANormasAMinimasand 70 to 75 percent in pirata developments.This also contributed to higher unit land prices in the ANormas Minimas.

Infrastructure investmenits uisiually took place well after initial subdivi-sion: at least a year elapsed between the sale of lots and the installationof iifiastruicttire in three-quarters of the subdivisions studied. Even insome NCrnmas Minimas subdivisions, investment did not begin untilalmost all the lots had been sold. The developers clearly use lot sale pro-ceeds for infrastructure expenditures: they simply do not have resourcesprior to the sales. The absolutely essenitial components-streets andwater standposts-are the first elements of infrastructure to be installed.The community itself ended up financing infrastruictiur-e investment to aconsiderable extent-more than 30 percent of costs in most cases (seetable 7-2). In formal sector land development in most cities, the govern-rnent usually bears the initial cost of infrastructuLre investment andthen attempts to recover varyinig degrees of costs throuigh user chargesor taxes. A novel feature of the pirata development scenario is the exante sharing of costs among the developers, hotiseholds, and the gov-ernment. This sitiuation has resulted from sheer force of circumstancebut has resulted in faster availability of public services in Bogota thanelsewlhere.

Lot buyer-s end up participatinig in this cost sharing in two ways. First,if the pirata developer does not install the infiastructure he is supposedto, and cannot be persuaded, the lot buyers have little choice but tomake the investment themselves. This is sometimes done by explicitly

Table 7-2. Financing of Infrastructure in Pirata Subdivisions

Percenteage of i frast-rIwture (oss paid 6)'

Both

sztbdn&iid.r Number oJ

and subdivisions.

Item Subdivider Commutnzty commMUit C;ov'ern ment I ith service

Sewers 43.5 32.2 16.1 8.1 62Water 62.8 23.2 5.8 8.0 86Electricity 48.8 31.7 6.1 13.4 82ITelephone hookups 4.8 61.9 4.8 28.6 21Streets 81.4 12.8 2.3 3.5 86Sidewalks 21.0 47.4 5.3 26.3 19Curbs 53.3 30.0 3.3 13.3 30

Naote: In all fotiu-teen Noarmas Minima.s subdivisions, 100 percent of all services werefinanced by the subdividers. The Narma. Miimmas subdivisions are excluided from thistable.

Source: Carroll (1980).

1l60 UNDERSTANDING THE DEVELOPING METROPOLIS

sharing costs with the developer. Second, the lot buvers often have topay when obtaining services from the government or the utilitv compa-nies. In an,i case, this process was cost-effective: for example, averageinfrastructure costs per square meter in these developments were only25 to 33 percent of costs budgeted for similar investments by the WorldBank in sites-and-services projects in El Salvador in the mid-1970s.

One of the main reasons for antipathy toward pirata land developersin Bogota was a general perception that they made high unearned prof-its. It is always difficult to judge what is a "high" undeserved rate ofreturni and what is reasonable. One yardstick is the existing rate ofreturn in alternative investments. In Colombia in 1979 it was easy toobtain relatively safe nominial returns of more than 30 percent frominvestments in officially available financial instruments. In some instru-ments, such as the Titulos de Ahorro Cafetero (coffee savings bonds), therate of return was just over 50 percent. Hence, a pirata developer wouldhave to get nominial annual returns of substantially more than 30 per-cent to compensate him for his risk in an economic environment wherethe average rate of inflation between 1972 and 1977 was about 25 per-cent.-3 The most realistic estimates are with payment default assump-tions, where the median rate of returni is only 25 percent for pirataswithin the ul-ban perimeter, 33 percent outside the urban perimeter,and 39 percent for Normas AMinima.5 (see table 7-3). These are clearlymodest rates of return, given the risk taking and effort involved in piratadevelopments. Of great interest is the wide variation among developers:a minority of developers clearly do make extremely high profits, andpopular impressions are obviously keyed to these high flyers.

These results help explain the reluctanice of developers to make ade-quate infr-astruLcture investmenits: most simply wotuld not have theresources to do so without losing money. This exercise also focusesattentioni on the much higher returns made by Narmas Minimas develop-ers: they were clearly making rents from the district's restrictive approvalprocess. The high return-ts also suggest that people are willing to paysome premium foi- legality.

An attempt was also made to unlderstand the determinants of lotprices. As might be expected, lot prices were higher nearer the city cen-ter and for preferred locations in the north and west of the city (the pre-dominlanitly poor areas being in the south). Amonig elements ofinfrastructure, there scerned to be specific willinigness to pay for waterand sewerage. This may partly result from an inadequate degree of vari-ationi in other infi-astr-ucture such as streets, which are available almostuniiversally. Proximity to water sources is clearly valued, as is hygienicwaste (lisposal.

In sumnmiiary, thie pirata development process in Bogota operated rela-tively well in responding to the heterogeneotis needs of a fast-growing

SHELTER IN A GROWING (CFF\ 161

Table 7-3. Alternative Average Nominal Rates of Return to Subdividers(annual percentages unadjusted for inflation)

Piraoa Piteita

su/diu si(ons vibdvisionste isie iirbao NVorans out ode urban

Allernaoivesfor rate ofreturi eshimates penme/e? Mitmmai pertmfeler

Payment defaull assumptionsJ

Median 25 39 33Mean 38 1 1 55Standaid deviation 42 94 75

(N) (101) (14) (21)

Mediuim at,ssmptions'Median 33 78 41Mean 53 125 64Standard devtation 53 92 78

(N) (102) (14) (21)

Favorable a.ssumnp/ions'Median 46 146 6()Mean 72 246 9(Standard deviation 8( 315 106

(N) (95) (13) (20)

Vole: Rates of return are internial rates (IRR) averaged for individual subdivisions. Calcu-lations are based oti co<mplete revenue and expetiditure figures, with dates, loi- eac h subdi-vision. Basic asstlImptions in all iRR calculations: All lots in the subdivision are sld if thifetotal number of lots is greater than the number of lots foir which sale data are given. the"excess" lots are coulited as sold at the average price (in 1977 pesos) and oni the last d(ate

of sale aniong the lots.a. The data assume that a cer-tain proportion of defaulting lot bluyers is ginen per sutidni-

vision: this incidentie of default is asstined to occ rl i andomlv over the period oF lot sales;and defaultng lot buvers are assumed to pay onilv the doowni payilmetit.

b. Assumptions are that no defatlt occurs: all lots are ftil paid for On tim e.c. Assumptions are that overrheac7 costs are limited to Olnl stanidard deviation above thie

nican of figures for Vormioas Minimas developmenlts.Source:(Carroll (1980).

city. It would have been difficult for the governmenit to improve on thisprocess in terms of response speed and flexibility or cost-effectiveness. Itis difficult for the forrnal sector to do this kind of incremental develop-tnent of land, infrastructure, and house constructioll. The unibundlingof these componenits of shelter is essential in order to make affor-dableshelter accessible to low-income hotiseholds.

Mobility and Tenure Choice

Household mobility rates are another measure of a city's hotising mar-ket. The characteristics of households change over time in various waysthat alter their requirements for hotising. An increase in household

162 UNDERSTANDING THE DEVELOPING METROPOLIS

income enables consumption of greater housing space. Expansion of ahousehold throuLgh marriage and child rearing increases the demandfor space. The relocation of workplaces through job change or the work-place itself changes the proximity of a dwelling unit. Changes in citystructure and in availability of transportation services also alter the char-acteristics of a dwelling unit. We have observed that in Bogota, particu-larly among low-inicome households, much adjustment takes placethrouLgh the process of incremental construction. The other means ofad justmenit is to imove to a new dwelling unit. High rates of mobility maybe regar(led as one sigll of a housing market that is functioning well.Mobility, however, involves costs. Tllese costs are expected to be muchless for renter-s than for owners. Tranisactioni costs in buying and sellingreal estate can be quite substantial. Hence, one expects significant dif-ferenices in the mobility behavior of renters and owners. It may beexpected that a household is far more likely to own a house wheni it hasreached stability or does not expect to needl to move rapidly.

The need for mobility is likely to be greater during a phase of rapidcity growth. In many cities this need may remaini unftulfilled because ofrigiclities in the housing market. As we saw in the last chapter, employ-meint cleceintr-alized fairly rapidly in Bogota during the 1970s. This waspartly because of the relocation of existing workplaces and partlybecause of a changinig location patter n of new employment. Both kindsof chanige imply continually evolving city structure in terms of proximityor desirability character-istics of different neighborhoods. Chapter 3 alsodiscussedl the changing nature of the land price profile over time inBogota. Neighborhoods change through dcensification and sometimestlhrough physical deterioration. All these factors point to the need forhotiselholds to move in response to varying circumstances. Rigiditiesrelated to housinig supply conditions, legal provisions concerning rent-als and tenure, land sales regulation, and other factors can contribute toreduced mobility and to mismiiatches between household requir-ementsand dwelling unit characteristics. A high mobility rate implies theabsence of such rigidities and probably contributes to a higher level ofhousehold welfare. As household characteristics change, people have tomove to findl a dwelling to match their new requirements. If they cannlotdo so, their quality of life may decline.

One of the more striking featuL-es of household behavior emergingfrom the World Bank-DANE 1978 Household Survey was the rather highrate of hotuseholdl mobility. In both Bogota and Cali, 23 percent ofhouselholds had resided in their present dwelling for less than one year.Just over one-thlird of houlseholds had lived at their currelnt address forless than two years and just over half for less than four years. These rateswere similar to movring rates observed in the 1972 Phase II HouseholdSurvey. What is most remarkable is that these rates are almost identical

SHELTER IN A GROWING CITY 163

to mobility rates observed in U.S. metropolitan areas. Further corrobo-ration of mobility rates in Bogota and Cali was obtained from a moredetailed, longitudinal record of moves in a special household surveyconducted by Roberto Corno in 1975. This SuLrvey focused on a sampleof 1,730 adults between the ages of twenty and forty-five and recordedall moves throughout their lives. We concluded that the mobility ratesobserved were quite reliable and stable across time and cities in Colom-bia. It would be very interesting if similar data were available for cities inother countries to help us determnine whether Bogota is typical or atypi-cal in this respect.

Wliat were the characteristics of the movers? How does householdincome affect the tendency to move? Higher income implies both agreater ability to move and less need to move. Higher-income house-holds woul(d be more stable and more likely to be owners rather thanrenters. Accordingly, higher-income households exhibited lower mobil-ity rates than lower-income households did: only 17 to 20 percent ofhouseholds with incomes over the median level were moved in the pre-viotis years, compared with 27 to 30 percent amonig the lower half. Asmight be expected, multiple-worker households move less often, but thedifferences were not large. The age of hotLsehold head is probably thebest predictor- of household mobility: the older the household head, themore stable household characteristics would be and the less likely thehousehold to move (see table 7-4). As might be expected, workplacelocation correlated closely with residence location. Between 40 and 50percent of households whose head changed jobs within the previousyear moved. As discussed, tenure status and mobility are highly corie-lated. In 1978 about 40 percent of all renters had moved in the previotusyear, but less than 10 percent of owner-occupiers had done so. It alsoappears thiat a large proportion of those who move do so frequenitly.Furthermore, recent migrants who have been in the city for less thanfive years are the most frequlent movers.

These insights were gained by merely examininig the various charac-teristics of movers. Given the expected correlations between variablessucIh as income, age of h1ousehold head, and length of time in currentjob, it was instl uctive to imodel the imobility beliavior. The basic hypoth-esis was that the variouis household characteristics can be used to preclictthe probability of moving. Estimationi was done by using both the OLS

anid LOGIT techniques, and the results were essentiallv as expected (seeHamer 1981 for details). When household income is controlled for, theprobability of inoving declines as household size and length of resi-dence in currenit location increase. The probability of moving is clearlythe highest for the most recent imigrants. A couple of other results arewortlh mentioning. It may be expected that once household income hasbeen controlled for, the location of residence should have no effect on

164 INDERSTANDING THE D)EVELOPIN( MFTROI'OLIS

Table 7-4. Moving Rates by Tenure Choice, Time Spent in PreviousResidence, and Age of Household Head(percent)

Time in pyfz.4au.i

residence

I,eS5 MOr Agef olhousehcdd head

7enurp *how than one tha i() and Ovper All

Cit( /leriod (hOn Rent yevar one Vear iinder 31l-4) 4() movers

Bof)goa, 1972Moved in 1971-72 15 54 - - - 37 20 32

Bogoli, 1 978

Moved within past year 8 37 45 20 41 21 14 23

(Cali. 1978

Moved within past vear 5 43 49 20 42 25 13 23

-Not ivai lable..Soe:ru Hairier (1981).

mobility. A dummy variable for the poorer sectors was used to test theirindependenit effect and found to be significant. Poorer neighborhoodsseem to have a more unstable pattern of residence. Finally, it was foundthat female-headed households exhibit no significantly different mobil-ity behavior than male-headed households: both were equally likely tomove. Overall, a priori conjectures were confirmed by the systematicestimations. Moreover, the lO0GIT estimates were quite similar to the OLS

procedure (see table 7-5, which gives comparable elasticities). For pur-poses suclh as this, where the reason for estimation is essentially toobtain an idea of independenit influences on mobilitrv as distinguishedfrom specific interest in the magnitiude of variables, it would probablybe adequate to conduct Ot.S estimations rather than the more complex1.1(;1Ir procedure.

NWhat is most remarkable, again, about the results obtained is thatvery similar patterns are generally observed in the United States (seetable 7-6). As in Bogoti, income, age of household bead, family size,and tenure status are importanit determinants of mobility, whereas thesex of household headl is not.

What is the pattern of houisehold moves? In 1972 there was clear evi-dence of decentralizinlg moves (see table 7-7); that is, movement ofhouseholds away from thie center. This tendenicy was muclh moie imoder-ated in 1978, possibly because of a slowdown in the growth rate of thecity. More interestinigly, most moves were within the ring of previous res-idence. In general, houselholds move to relatively nearby locations. Inmost sectors a majority of relocating households moved to new locationswithin their sector of previous residence. This tendency was strongest in

Table 7-5. Probability of Moving: Comparison of OLS and LOGIT Elasticities

Bogntd, 1978 Cali, 1978

LOGOI sample LOGI1 sampleVaniable OLS LOt.IT model I LOCIT model H mean OLS LOGrin model I LOCIT model II mean

lrncomiie -0.0981 -0.0940 -0.2580 12.51 -0.10 -0.04 -0.14 11.58cr Ageofhouseholdhead -0.8694 -1.1677 -1.2603 41.24 -1.11 -1.09 1.11 42.80

Family size -0.5146 -0.6250 -0.6736 4.95 -0.48 -0.51 -0.59 4.92Years in current residence -0.1086 -0.0470 -0.0382 5.08 -0.11 -0.17 -0.17 4.55Years in current job -0.0487 -0.1068 -0.0840 6.06 -0.06 -(.09 -0.08 6.16Numiiber of workers in household 0.1377 0.0468 0.0970 1.74 0.11 0.10 0.13 1.72

Note: The probabilitv of moving was 0.346 in Bogota anid 0.356 in Cali.Source: llamiier (1981).

166 UNDFRSTANDING THE DEVEL,OPING METROPOLIS

Table 7-6. Mobility Rates in U.S. Metropolitan Areas by SelectedHousehold Characteristics

Pen(-enf7gr' of h0UW.sehods

Characlerrsti moving

Incoine'$6,999 or Iess 25$7,000-$14,999 25$15,000 or mole 18

Age oJ household headUinder 30 4630-45 20Over 45 8

Familh sizeULnder 4 244 or more 1 7

Tenure rhoiceRemt 38Own 1I

.Sex of household headMale 22Female 22

a. 1977 dollars.Source: Derived from U.S. Departmenlt of C(ommerce (1977), table A-1 data for house-

holds inside meLropolitari areas.

the sectors of greatest poverty (2, 3, and 6 in Bogota; see table 7-8). TheCorno survey data, which had much longer histories of mobility, foundfurther evidence of this tendency. As expected, migrants into the citvshowed much greater mobilitv than natives. Natives were more likely tobe found in their sector of birth thani migranits in their first sector of res-idence in the city. Migrants live wherever thev can whern they first arriveand then adjust their residence after achieving some stability. The poorsectors (2, 3, and 6) showed higher retention rates than other sectorsbut much more so for natives than migrants. Once thev achieve stabilityin poverty, it would seem, natives are not able to move out of their poorneighborhoods. Migrants, by contrast, appear to be more mobile, resid-ing in the poorer neighborhood to begin with and then moving up.

Much of the discussion of housing supply earlier in this chapter wasrelated to ownership housing. How important is ownership housinlg inBogota? How easily is it accessible? What are the determinlants of tenurechoice? We have already seen that renters are likely to be more mohilethan owners. Hence, one would expect that all the intflences thatincrease mobility rates would decrease the probability of homeowner-

SHEITFR IN A GROWING CITY 167

Table 7-7. Moves of Households from One Ring to Another(Recent Movers)(percent)

Movped within Molved lo ringfarther Molved lo rng closer

Rinig oj presentI residence sam noTg from center to center

Bogot(i, 19721 35 65 n.2 36 62 23 28 56 164 59 29 125 73 1 266 18 n .a. 82

Total 53 31 16

Bogotei. 19781 17 83 n .a.2 58 30 123 35 43 224 55 23 225 74 2 246 46 n.a. 54

Total 59 18 23

Ceali, 1978

1 41 59 n.a.2 67 33 53 73 14 134 61 3 365 39 n.a. 61

Total 64 19 17

n.a. Not applicable.Source: Hanier (1981).

ship: the age of household head, the length of residence in the city, sizeof the houselhold, and income would all be expected to be positivelyassociated with the probability of owning a house. This leads to thequestion of whether, all other thinigs being equal, people in better-oftareas of the city are more likely to own a house.

In both Bogota and Cali, almost half of all households own the housethey live in (see table 7-9), and higher-income househol(ds are morelikely to be owniers than renters. Even among the poorer households,however, one-third or more of the households are homeowners, testify-ing to the efficiency of the noniformal modes of housing supply. In thelate 1970s the overall level of homeownership matched that attained inthe United States as late as the early 1950s. A large proportionl of thehouseholds own houses despite the lack of access to home financing foranyonie but the better-off grotips.

As expected, homeownership increases rapidly with the age of thehousehold head. Few household heads tinder the age of thirty own their

168 UNDERSTANDING THE DEVELOPING METROPOLIS

Table 7-8. Moves of Households from One Sector to Another(Recent Movers)(percent)

.Sector ofpresent Moved within Moved toresidence sector oforngin other sector

Bogota, 19721 35 652 57 433 64 364 46 545 45 556 68 327 45 558 46 54

Total 56 44

Bogota, 19781 17 832 58 423 69 314 55 455 46 556 78 227 34 668 51 49

Total 58 42

Cali, 19781 41 592 57 433 59 414 65 355 76 246 76 317 0 100

Total 66 34Source: Hamer (1981).

own homes; those beyond forty are much more likely to own. Similarly,job stability contributes to homeownership: among household headswith less than one year in the current job, as many as 75 percent wererenters. Recent migrants, who are also very mobile, are likely to be rent-ers, and older residents are much more likely to be owners (see table7-10). Female-headed households in Bogota exhibit tenure choice char-acteristics similar to those of male-headed households.

How do these patterns compare with other countries? Informationabout developing countries is still scanty, but it does seem that home-ownership rates vary with income levels of countries as a whole. Forexample, whereas only 33 percent of households in Bangalore, India,were homeowners in the early to mid-1970s, in Korean cities and in

SHI-ELTER IN .A CROWING CITY li69

Table 7-9. Tenure Choice by Household Income(percent)

Household income

Up to half lalfof median

Tenure choice of media ni to median Over me(lian All households

Bogold, 1972Own 33 39 57 46Rent 67 6] 43 54

Bogolid, 1978Own 33 39 57 47Rent 67 61 43 53

Coa/i, 1978OwnI 40 43 59 51Rent 60 57 41 49

Source: Hamer (1981).

Dakar, Senegal, almost half were homeowners, and in the United Statesthe figure was as high as 61 percent. Ownership increases moresmoothly with income in the United States, but, as in Bogota and Cali.the jump in ownership comes just above the median income level. Oneimportant difference in the United States is that female heads of house-holds are much less likely to owni their own homes. This is probably dueto the race-related issues in U.S. cities. Although the trends are, in gen-eral, similar benveen the U.S. and Colombian cities, the differences inownership levels are much greater than differences in mobility levels.Income levels are clearly much more important for homeownership.

The probabilitv of homeownership was also estimated UsinIg dataavailable in the World Bank-DANE 1978 Household Survey, and( all vari-ables were found to have the predicted signs. 4 The elasticity of eachcharacteristic is nearly identical in magnitude but opposite in sign to

Table 7-10. Tenure Choice by Recency of Household Head's Migration(percen t)

Arrived within post

Jrenu re, clcoice jive veanr Long-lerm residew

Bogota, 1 972OwII 26 56Rent 74 44

Bogoki. 1978Own 23 5(Reint 77 50

Cacli, 1978Ouwnl 1 6 5f6Rent 84 44

Source: Hamer (198 ).

170 UrNDERSTANDlIN(G THE DEVE1OPING METROPO.IS

those estimated for mobility decisions (see table 7-1 1). Hence, decisionsabout homeownership are clearly very strongly influenced by the house-hold's need to be mobile. As a household matures and becomes morestable, either because of its members' characteristics or because ofincreased job stability, the household head decides to own a house andexpects not to move too frequeintly. After becoming a homeowner, he ismost likely to continue in that status even if he does move.

International comparisons of such elasticities are difficult to obtain,but some sketchy infornmation is provided in table 7-12. These compari-sons need to be interpreted with some care, because homeownership isstrongly influenced by institutional structures related to housingFinance and legal systems. Nonetheless, it is interesting to observe thatincome elasticities of ownership are surprisingly uniformly low acrossthe three countries comnpared. The age of the household head is atypi-callv impor-tant in Colombia, presumably because of the lack of access ofmost households to housing finance: they have to save for some timebefore gaining access to hoineownership.

Household character-istics are the main determinants of householdmobility and homeownership. The household life cycle is the mostimportant of these. In a fast-growing city such as Bogota in the 1960sand 1970s, the progress of a migrant household through these life-cyclestages may lag behind those of natives, but the essential behavior is nodifferent. Bogota and Cali had relatively high mobility rates-quite simi-lar to those in the United States. However, these rates were relatively lowamong homeowners in Colombia; the relatively higher proportion ofrenters had higher mobility rates, bringinig up the average. The morewidespread availability of housing finance and smoothly functioning sys-tems of property transfer help ease mobility among homeowners in the

Table 7-11. Elasticities for Tenure Choice Model Comparison of OLSand LOGIT Specifications

Bogoki Cali

l.X;;, J.(XL rr

sa mpi sa tnpe

Independent variable 0ol WfI (;f ean ?I 01S 1n(;Fr mean

Incomi-e 0.11 0.24 12.51 0.12 0.16 11.58Age of household

head 0.96 1.39 41.24 0.65 1.1( 42.78Fanmily size 0.38 0.73 4.95 0.46 0.60 4.92Years in previotts

dwelling unit 0.11 0.06 5.08 0.09 0.13 4.55Years in presentjob 0.05 0.06 6.06 0.03 0.03 6.18Number of workers

in houselhold -0.12 -0.25 1.74 -0.12 -0.16 1.72Nole: The probability of ownership is 0.49 for Bogota and 0.54 for Cali.Snmve: Hanier (1981).

SHELTER IN A GROWING CITY 171

Table 7-12. International Comparisons of Ownership StatusElasticities

Colombia Eleven Korean catn's United Slales

Phita- lThirtv-7ariable Bogola Cali Low Mean fHigh detphia4?ise cnties

Income 0.11 0.12 0.03 0.06 0.12 0.30 0.18Age of household

head 0.96 0.65 0.06 0.23 0.50 0.19 -Family size 0.38 0.46 0.20 0.43 0.54 0.25 -Length of

occupanicy 0.11 0.09 0.12 0.15 0.18 - -Number of

Workers inhousehold -0.12 -0.12 - -0.08 - - -

Female house-hold head(coefficient) -(.01 0.07 - 0.10 - - -

- Not available.a. Estimates are fbr married couples.Source: Bogota and Cali data from OLS estimates in table 2-29 of Ha1ner (1981); Korean

data from Lim, Follain, and Renauid (1980): Philadelphia data fiom Fredland (1974);other Lt.S. city dar. f<ron Struyk and Marshall (1975).

United States. Expectations about mobilin, seem to determine whether-a household ownls or rents, and these are based largely on factors exoge-nous to the household.

Modeling Housing Demand

So far we have attempted to understand the behavior of the householdas a demander of housing and the behavior of both public and priiatesuppliers of housing in the particular institutionial setting of Bogota. Inthis section we focus on the more traditional dimensions of housingdemand: how income-elastic and price-elastic is housing demand, andhow can this be measured properly? Although there is considerable lit-erature concerning housing demanid in th)e ULnited States, very few stud-ies exist for developing countries. 5 This situation has improved in thelast few vears, 6 but information available is still limited to a relativelysmall group of couLntries, including Egypt. El Salvadot; lndia, the Philip-pines, and the Republic of Korea.

Modeling housing demand is intrinsically difficult because it is hardto find "correct" measures of both quantitv and price. Moreover, hous-ing markets are very heterogeneous, and housinig prices t}pically varyover space within a city. It is difficult to define a "standard" house suchthat different quanitities consumed can be measured. Similarly, in theabsence of a natural unit of quantity measurement there can be no asso-

172 UNDERSTANDING THE DEVELOPING METROPOLIS

ciated Ultit price. What is more readily observed is expenditure on hous-ing and variotis attributes such as extent of development in the area,size of the lot, number of rooms, location, neighborhood quality, andlevel of public service. Typically, a means of simplification is then foundin order to make the modeling problem tractable. The easiest solutionis to assume away price variation in a housing market and regard expen-ditures as proxies for quanitity (see Muth 1969). Another solution is tosegment the hlousinig market in a city into neighborhoods within whichhomogenieity, and thus price uniformity, can be assumed (see King1975). In more complex approaches, indirect methods are used toderive price indexes by means of hedonic price-index estimation or theestimationi of a housing prodtictioni function with varying input prices(see Polinsky and Elwood 1979; Witte, Sumka, and Erekson 1979).

Eachi of these approaches causes specification problems that involvethe inappropriate limitation of' choice. Households typically searchwithin and across neighborhoods: the price variation encountered isthen not just withini but across neighborhoods. Similarly, estimatinghedonic prices unbtindles the dwelling into its constituent parts,whereas thie household may really regard the house as a compositegood. As has been argued, this is an empirical issue, because low-income households have succeeded in unbunidling their shelter into dif-terent components that are sought in a house sequentially.

The problenm, therefore, is to find a tractable method of defininghotusitng quantity and price that conforms to our theoretical under-standing of urbani housing markets. It would then be possible to esti-mate hotusinig deinand elasticities with respect to income and pricevariation. Another issue is to understancl the biases that result fromincorrect data aggregation. It is not always easy to obtain disaggregateddata at the household level. Elasticities are then estimated from differ-ent kinds of aggregated data that grotip households in different ways.Are aggr-egated results consistent withi disaggregate demanid estimates?

A relatively simple application of residential location theorv suggestsan alternative way of incorporating price variation into a demand equa-tion for housilng as a composite good. We have observed that residentiallocation is related to the workplace location of the household head. Thehousehold heacd can theni be seen as optimizing the location of resi-dence in relation to his workplace: housing prices, R, would typicallyvary with distance, d, fi-om his workplace, whereas travel costs, 1, wouldincrease with distance. For any given amount of desired housing, H, thehouselhold's problemil is to minimize total expenditure on housing, Z7,includinig the implied transport expenditure. Thus:

(7-1) Z = R.(d) -H+ t(d)

where the subscr-iptj r efers to workplace j.

SHELTER IN A (GROWINC (.ITY 173

For each qualLtity of H desired, there would be an optimal distancefrom the workplace-this is essentially the monocentric city modelbeing generalized to a city that has workplaces in all areas. For eachindividual household, however, the city looks like a monocenitric citywith the workplace at the center. A different optimal location profilewill be obtained for each workplace, because the price variation andtransport costs variation would be different for each workplace. Theconsumptioni of the same quantity of housing, H,,, will imply a differenttotal cost for each workplace. Thus, for example, the total hiousinlgexpense for H, for a person working in the CBD is likcly to be higherthan for one working on the periphery; this variation in cost acrossworkplaces for consumption of the same housing quanitity may beregarded as constituting the price difference wvithin a housing market.This also implies the existence of a wage gradient: a worker at theperiphery can be paid less than one at the center and yet have the samehousing consumption and the same utility level. This was empiricallyfound to exist in Bogota (see K. S. Lee 1989).

The procedure used for estimating housing demand was as follows:the city was divided into j work zones, each with Nj households associ-ated with the workplace. The total number of sample households was M.Each hotIsehold i at workplace j spends Rig on housing that has a set of kcharacteristics, ijk. A hedonic equation relating housing expenditure R,with housing characteristics can then be estimatecl:

K

(7-2) R,I = jk , hx,,

h- I

Thus Pjk can be estimated and

(7-3) h

calculated where R, is the expenditur e on housing associated wvithworkplace j on a house with standard characteristics XA. This set of stan-dard characteristics Xk could be defined as the average across the city.This procedure serves to define the dif'ferent expenditures on the samestandard hlouse as related to different workplaces. The housing priceindex for workplace j is it, = R, /Rp, where RI can be arbitrarily used asthe numneraire.

Each household's housing expenditure, R1J, can then be normalizedby iT . R /ic is then the quantity of' standard house consumed bv house-hold i as'sociated with workplac"e j. The demand equation

174 UNDERSTANDIN(; THE DFVEINGPl\( METROPOLIS

(7-4) 0). = R,/ =i J(nj, Y, HC ,, dgjI i I i f 1r

can then be estimated relating quantitv of housinig (Qj) with plice (n.),income (Yr), household characteristics (HC-1), and distance of resi-dence (dOj) fiom workplace (j). This equation can be estimated acrossall Al households.

Such an equation was estimated for BogotJi for 1972 and 1978 and forCali for 1978. Bogota was divided into thirteen work zones for bothyears. Separate equations were estimated for owners and renters, usirigvalue of house and rent paid, respectively, as the dependent variable.For 1972 only renters' equations were estimated because house valueinformation was not available for owners. It may he seen from table 7-13that there was considerable variation in average commuting distance.For owiners the longest commuting distances were for the central zonies,1, 2. 3. and 4, andl for zones 8 and 12, which broadly correspond to ourstandard sectors 7 and 8 in the north. It had been observed earlier thatthere was consi(lerable in-conimtitinig into the relatively better-off sec-tors 7 and 8. The commutifng pattern for renters was similar except forwork zone 2, which had a very low average commutinig distance in both1972 and 1978. Trhis corresponds to one of the "incubator" areas identi-fied in chapter 6. The central areas of the citv had many renlters, andtihey appeared to be working in central zones such as zonle 2.

The hedonic equations (7-2) were estimated first usinig several inde-pendent variables: area of dwelling Unlit: lot area; dummyilv variables foravailability of phone, separate kitchen and bathioom, aid muniicipalgarbage collection service. The last variable was seeni as a proxy forneighiborhood quality or amenities. People woIrking in work zones 9 and

10 in the southwest and the south of Bogota, who were likely to be low-incoine blue-collar workers, showed the lowest indexes for these facili-ties. The most striking difference observed betws'een reniters and ownerswas in the size of the dwellinig unrit area. The average for owners illBogota was about 170 square meters, whereas for r enters it was about 70square meters in 1978. The comiiparable figures for Cali were 125 squaremeters and 65 square meters. Interestingly, the differences in lot areawere not as large: 150 square meters for owners and 125 square metersfor renaters in Bogota (130 square meters and 125 square meters, respec-tively, in Cali). There was intich more variationi in the dependelit vari-able (rent and value) across work zones than in the independentvariables. The variables used for the 1972 estimationi were differcntbecause the same information was not available in that data set.

The hedonic price equations were estimatedl for each work zone andfor renters and owniers in both Bogota and Cali, resultinig in forty-twoequationis. The most important variable to emerge was dwelling unitarea; lot area did not perfor m as well. People apparently value dwellinig

SHELTER IN A GROWINC; CITY 175

Table 7-13. Characteristics of Residences by Work Zone, Bogoti

A verage distance of residence from work -one Derived pnrce index

Work zone Renters, 1972 Renters, 1978 Oumers, 1978 Renters. 1978 Ovners. 1978

1 5.7 6.5 8.0 100 1002 3.8 3.5 6.2 91 633 5.2 6.0 6.9 83 904 5.6 6.7 8.1 105 1075 3.2 2.6 3.2 74 766 5.8 5.4 6.8 75 857 3.1 4.5 5.4 81 808 4.9 5.8 7.3 98 1019 3.3 2.6 3.3 84 61

1 0 6.0 5.2 4.2 77 641 1 5.8 4.5 6.0 91 791 2 5.9 7.8 7.3 99 1361 3 8.6 5.1 6.1 84 87

Note: Zones 1, 2, 3. and 4 are central ciry zones; 9, 10, 11. and 13 are outer zones; and 5,6, 7, 8, and 12 are intermediate zones.

Source: Ingram (1984), exhibits 1, 3, arid 4.

unit space more than lot area. The dummy variables for phone access,exclusive bath and kitchen facilities, and garbage collection also per-formed relatively well. These variables do not perform well for workzones 2, 9, and 10, where many of the less well-off workers work. Thiscorrelates with dwelling unit area: the workers in these work zones whodo have access to these facilities probably have large dwelling unit areasas well. A measure of the explanatory power of these equations is pro-vided in table 7-14: the equations succeed in explaining about 45 to 70percent of the variation in housing expenditures. The workplace stratifi-cation is more important for owners. This is as expected, because theprice gradient of owner-occupied tinits is usually steeper than that ofrented units in most urban areas. Finally, the rents and values for theaverage "standardized" home were obtained by usilng the average char-acteristics of each zone along with the estimated price coefficients. Thenormalized price index for the work zones is displayed in table 7-13,using zone 1 as the normalizing work zone. There is considerable varia-tion, up to 40 percent for owners and 25 percent for renters. The high-est price indexes are found for zones 1, 4, 8, and 12-the corridor thatextends from the CBD to the north and where much of employmentexists, as well as much in-commuting and the longest commuting dis-tances. The lowest indexes for owners are for zones 2, 9, and 10, whichalso have shorter commutes on average. The higher price indexes are,in some sense, for the more preferred or competitive workplaces wherepeople are willing to work even if it involves a longer commute in orderto live in the same quality of hotise.

17I UtNDERSTANDING. THE: DEVELOPIN; METROPOLIS

Table 7-14. Analysis of Variance: Hedonic Price Equations(percenlt)

Varnaliono explained btr

Woerk zone

Househ'ole(l stMalfication Equations Total variation

1972 Bogota renters 4.7 49.3 54.01978 Bogota renters 2.5 47.6 50.11978 BogotAi owners 8.7 36.4 45.11978 Cali renters 1.9 64.3 66.21978 Cali owners 8.0 60.9 68.9

Soum-. Ingr-am (1984). exhibit 13.

flaving obtained the price index, ir , the demand equation (7-4)could then be estimated usinig househoid income, price (t I), ctistancefrom workplace, and other household characteristics as the indepen-dent variables. 8 Two functional forms-log and linear specifications-were estimated. Income was clearly the most important explanatory vari-able. T'he housing price index was significant only in two of the five sam-ples (Bogota owner 1978 and Bogota renter 1972). Distance was alsofound to be generally significant and negative in sign, as expected. Ageof'household head and family size were usually significant, whereas sexof'household head was not, although it was negative in sign. Althoughthe price coefficients were not in general found to be significant, itwas important to conduct the elaborate exercise constructing the priceindexes because the unbiased estimation of other variable coefficientsrequires proper price normalization. In addition, the omission of pricevariables could have led to biases in other coefficients as a result of theomitted variable problem.

The characteristics are calculated at the mean value of each variablein the linear equations except for income. The estimated linear elastici-ties for Bogota are shown for approximately the first, second, and thirdquartiles of each sample's income distribution (see table 7-15). In eachcase, the sample mean and seventy-fifth percentile of the income distri-bution are essentially identical. All income elasticities are clearly posi-tive, significant, and less than one, including those for Cali (notreported here). At the sample mean they lie in a narrow range between0.6 and 0.8. The elasticities seem to increase with increases in income.At low income levels, income elasticities are relatively inelastic. Theprice elasticities vary more in magnitude but are always negative and lessthan I in absoltIte magnittude; they become qtuite small at high incomelevels. The distance elasticities are usually small and negative. Familysize elasticities (nlot reported here) showed an interesting pattern,being negative for owners and usually positive for renters. Rentersapparentlv vary their house size according to family size. Owners had

SHELTER IN A (.ROWIN; (lIY 177

Imuch1 larger holtises oni average and do not seem to vary their hoLIse sizewith increatses int faimilv size. As rnight be expected, the age of hotuse-hold head had a consistently positive demand elasticity. When) the term(age of hotisehol(l head) 2 was used in the linear eqttatioris, hottsingdemaniid seemed at its peak wheni the hotisehold hea(d was betweell theages of fifty and tifty-seven.

How do these resUlts compare wvith informationi from other coUnl-tfies? Most income elasticities are beLweeni 0.2 andl 0.8, with most clus-tered between 0.5 and( 0.8 anid some near unity (see table 7-16). Y Ingeneral, renter elasticities appear to be lower thani those for owners.The medliani renter elasticity is about 0.45, whereas for owners it is abouit0.6i5. Price elasticities are generally founid to be quite low anid nega-tive-ustially lower in absoltite valiue than income elasticities. However.comparisonis are difficuilt to make across sti(lies becatise of (lifferentspecification and estimationi methodlologies. Malpezzi aln(d Mavo (1987)attempted to solve this problem b% estimatirig a consistenit equationi fora large number of rifferent data sets for sixteen cities in eigilt countrlies.These restilts confirmed the general finidlings rlescribed here. In addi-tion. Malpezzi and Mayo observedl that renit-to-inicome ratios decline sys-tematically within cities (as wotild be expected ti-om incomiie elasticitiesless thanl u1nity) Iiut inreasme nv',h average income .arrav.o citiies. They also coni-

Table 7-15. Housing Demand Equations: Estimated Elasticitiesfor Bogota

Renters, 1972 Rlenern, 1978 Owners, 1978

P1rcent(i1es h11 ear 1-ag-log linear l./ag-tag Linear log-aog

In Cromf

All income percenitiles 0.77 0.72 0.7825th percentile 0.32 0.55 0.3350th percentile 0.45 0.71 0.4775th percelitile' 0.59 0.80 O.i0

l'rice

All incomite percentiles -0.70 -0.28 -0.4425th percentile -0.91 -0.17 -0.3150(th percenitile -0.75 -(.11 -0.2475th percentile -40.55 -0.08 -0I 91)j.stan cI 'e

All incomiie percenitiles -0.(6 -0.06 -0.0225th percentile -0.05 -0.23 -0.0250(th percentile -0.04 -0.15 -(.0175th percentile,' -0.04 -(.10 0.01

.a. (ot respo,nds to sam ple ineani.'SoIurce: I ngrmin (1984), exhib its 16i arni I17.

Table 7-16. Ranges of Housing Demand Elasticities from Various Countries(based on observations ot households)

Elasticztv with respect to

Sex of howsehold head

Cou ittrv Current income Pyice ramil.size Age of household head (I = MaIle

Ren7erJ

Colombia 0.2 to 0.8 -0.1 to -0.7 -0.1 to 0.4 0.1 to 0.6 -0.01 to -0.2United States 0.1 to 0.4 -0.2 to -0.7 Uncertain Uncertain Consistenitly

negativeRepublic of Korea 0.12 -0.06 to 0.03 0. 15 to 0.25 - -

Egypt 0.25 to 0.5 - -Ghania 0.33El Salvador 0.25 to 0.45Iidia 0.6Philippines 0.6 to 0.9 - - - -

Owners

Colombia 0.6 to 0.8 -0. 15 to -0. 40 -0.2 to -0.35 0.1 to 0.4 -00. 2 to -0.1ULlited States 0.2 to 0.5 -0.5 to -0.6 UTncertain Uncertain NegativeEgvypt 0.2 to04 - -0.Republic of Korea 0.21 -0.05 to 0.07 -0()2 to 0. 15El Salvador 0.45 to 1 0hidia 0.4Philippines 0.6 to 1.0

- Not available.Soutrce: Colombia fronm Ingramii (1984), exhibit 18; United States from Mayo (1981); Korea from Follain, Lim, and Reniauid (1980): all others fromii Malpezzi

and Mayo (1987).

SHELTER IN A G;ROWING CITY 179

firmed that rent-to-income ratios are consistently higher for ownersthan for renters.

The curious finding across cities was reflected in another finding: per-manent income elasticities of demand for housing are greater than cur-rent income elasticities. In the short term, housing consumption doesnot adjust as fast as increases in income levels, but it catches up in thelong term. It is also possible that housing and land prices increase alongwith economic development, thereby leading to higher housing expen-ditures as proportions of income. Thus, cross-country data tend to showhigher proportions of housing expenditure with increases in per capitaincome. It is also interesting to note that the U.S. results show relativelyinelastic estimates that are generally lower than for developing coun-tries. This suggests that the tendency of housing expenditures to be agreater proportion of income might taper off at the highest income lev-els and be the highest for the middle-income countries. This would beconsistent with the fast-increasing housing and property prices observedat middle-income levels during the period of rapid urbanizationi.

It is too time-consuming to estimate housing demand parametersfrom disaggregated data, especially when these estimates are needed forpolicy purposes. The short-cut is usually to take available informationfrom aggregate statistics based on rent and income data for groups ofhouseholds. Many such estimations result in income elasticities of hous-ing demand of greater than 1, a finding that is now generally agreed tobe incorrect. This result is usually the conisequence of incorrect aggre-gation when the stratification procedure used is based on stratificationof the dependent vfariable, that is, on categories of housing expendituresitself. The correct aggregation procedure is to stratify by the indepen-dent variable-that is, on income categories (see Ingram 1984 fordetails and simulations).

In summary, we have found that housing demand estimations have tobe made with some care in order to separate price and demand effects.Although housing quantity and prices are intrinsically difficult to mea-sure, a careful theoretical appreciation of the nature of the housingmarket in its spatial setting suggests a computationally tractable proce-dure for analvzing housing demand. The key theoretical construct is thetheory of residence location based on the relation of residence to loca-tion of workplace. Our principal findings have been that income elastic-itv of demand for housing is typically less than 1; that the elasticity islikely to vary with income level, being lower for low-income household;that there are systematic differences in the behavior of owners and rent-ers, with the requirements and characteristics of renters being different;that owners typically spend more oni housing; and that other householddemographic characteristics also matter. The important policy conclu-sion is that the heterogeneity in housing demand requires a heteroge-

I80 UNDERSTANDING THE DEVELOPING NMETROPOLIS

neous response. The affordability that may be expected by observingr enter behavior is likely to seriously understate the ability of householdsto pay for housing. It is also a mistake to assume that all households candevote a similar proportion of their income to housing. Finally, housingpolicy should also ensure provisions for adequate supply of rental hous-ing: ownership housing is not always a substitute for rental housing.Most housing programs in developing countries are designed to pro-mote ownership housing, thus neglecting the needs of renters.

Pirata Developments and Housing Demand:Some Lessons for Housing Programs and Policies

The most remarkable finding of this study was that the supply of hous-ing in Bogota and Cali grew faster than the population did during thehigh growth period of the 1960s and early 1970s. The average numberof persons per dwelling unit fell from 8 in 1964 to 7 in 1978. Because ofa significant drop in household size from 6.2 members in 1964 to 4.9 in1978, however, the number of households per dwelling unit remainedalmost constant at betwveen 1.3 and 1.4. At the same time, the access tobasic infrastructure improved during the 1960s and 1970s. The popula-tions of Bogota and Cali clearly had better shelter in 1978 than in theearlv' 1960s or before.

Although the government in Colombia has exhibited deep concernfor housinig throughout the period under study, the government had lit-tle to do with the successful expansion of housing supply. The formalhouisinig credit and construction agencies encountered great difficultyin effectively channieling their efforts toward the poor. If the poor wererelatively well housed in urbani Colombia by the beginnling of the 1980s,it was mostly due to their own efforts and ingenuity. The evidence, how-ever, is that different governmenit authorities acquiesced in providingincremental infrastructure. In this sense the infrastructure developmentin Bogota was clearly demand-determined. Infrastructure generally fol-lowed development rather than the other way around. The regulariza-tion of pirata developments was actively pursued by the authorities, thusalso contributing to the development of reasonable housing stock andurbani public services.

The estimated income elasticity of housing demand was less thanunity, varving largely between 0.6 and 0.8 at the sample mean. In fact,the income elasticity increased with rising income, and low-incomehouseholds effectively exhibited low-income elasticities. The result isthat low-income households have to devote a large proportion of theirincome to housing; as expected, housing behaved as a basic necessity.The average proportion of income spent on housing was 16 percent in

SHELTER IN A GROWING (-llY 181

Cali and 22 percent in Bogota, with some of the poorer householdsspending as much as 50 percent of their income on housinig. Theseresults are quite consistent with those found in other developing coun-tries. One interesting finding from cross-country information is thatpermanent income elasticities of housing demand are higher than cur-rent income elasticities. Although almost all income elasticities are lessthan 1, rent-to-income ratios increase with income across cities as aver-age inicome increases. This is one reason that earlier estimates of hous-ing demand elasticities, which are often based on averaged data acrosscities or countries, werc usually found to be greater than 1.

A number of important policy implications arise from these findings.Housing programs that are based on the ability to pay have to bedesigned very carefully, given the variation of income elasticity ofdernand with income within a city or country and the variation of aver-age income across cities. The proportion of income devoted to housingwould seem to depend on the place of a household in the prevailinginternal income distribution, as well as on the absolute average level ofincome. The latter may also be a result of prices for housing materialand land that increase along with average incomes. It is common prac-tice to make affordability calculations based on a fixed rent-to-inicomeratio. The result is that housing programs ostensibly targeted at specificlow-income groups prove unsuccessfuil, not becatise implementation ispoor (as is usually supposed) btut because of design errors. Heterogene-ity in the demand for housing must be respected in designing housingprograms.

Most housing programs in developing counltries are aimed at owner-ship housing. One of the important findings of our study is that even ina middle-income country like Colombia, the poorest third of urbanhouseholds have difficulty in achieving effective access to the ownershipmarket. Indeed, the formal legal organized housing market essentiallycatered to the top third of households, the extralegal market enabledthe middle third to buy homes, and the bottom third were basically inthe rental market. Apart from problems caused by the low income level,it was household instability-in terms of incomc itself, job location, andhousehold formation-that made poorer households find shelter in therental market. As some of these households graduated up, they enteredthe ownership market. There, low-income r entei-s gener-ally riented fronmthe middle- or lower-middle-income owner-occupants, thereby alsohelping them finance their housing investments in an interesting symbi-otic relationship. It is important for housing policymakers to recognizethe predominant preference of the poor to r ent (rather than own), so thatthis section of the housing market can be developed more effectively.

The chief lesson obtained from the extralegal developments ofBogota is that uiibundliing the housinig package can substantially ex-

182 UNDERSTANDING THE DEVELOPING METROPOLIS

pand lower-income groups' access to higher levels of shelter satisfaction.This unbundlinig is done both by separating different elements of ahousing package at a given time and by separating them over a periodof time. Unbundling is known more popularly as incremental housingdevelopment. Given that shelter is usually seen as a package of infra-structtlre services, house strulcture, and land, the entry costs of acquir-ing this package often become prohibitive, effectively excluding all butthe top few income deciles in a developing countrv. In a situation of' rap-id urban growth, like that experienced in Bogota, households foundshelter by means of such innovative approaches. They were willing to de-velop the housing package incrementally: to buy undeveloped land withminimnal infrastructure, build a core house to begin with, and thenachieve access to improved infrastructure while expanding the houseover time. This was made possible from the supply side by the intervenl-tion of small, independent developers who rapidly jumped in to fill thislarge demand niche. The government seemed to recognize the effec-tiveness of this process and responded by incrementallv imiproving pub-lic services and eventually legalizing the developments.

Unfortunately, few governments recognize the real potential of' thisprocess. The standard model for delivery of housing to the poor consistsof public sector construction of fully finished unlits, or sites and services,along with a full complement of infrastructuire. Households are thenassisted by the provisioni of subsidized loans. The cost of formally builthousing units is seldom low enough for most households to afford, withthe restlt. that governments invest large resoturces in directing subsidiesto the better-off, while the real target group remains unaffected. Despitethe relatively good, although seemingly chaotic, results achieved inBogota in the 1960s and 1970s, the government of Colombia has revertedto the previously abandoned IcT-led housing construction model, withapparently predictable results (see WVorld Bank 1990). As this study doc-uments, much better results would be achieved if institUtions affectinghousing supply remain decentralized and demand-deterinined.

Durinlg rapid urbani growth, governments seldom have the resourcesto lead with timely infrastructture development in tirban areas. In anycase, housing development patternls must be guided by people's prefer-ences. What may work in Bogota may not work elsewhere. But we dolearn that people's requirements and preferences in seeking shelter aresufficienitly heterogeneous so that the government must facilitate a flex-ible suipply response to match their demands. Public policy must be ableto harness households' own energy in providing shelter for themselves,yet public services must be provided to the households. The finding thatpirata developers gained returnis that were approximately comparable tothose fi-om alternative investments in the economy should promote the

SHELTER IN A GROWING CITXY 183

idea that land and housing development is like any other economicactivity. Hience, like other actihities, it should he promoted, particularlybecause substanitial infiastructure investment is then made in a rela-tively cost-effective way.

Notes

This chapter draws on work reported in Carroll (1980), Hamer (1981, 1985),Ingram (1984), Paredes (1984), Stevenson (1984). Pineda (1981), and Ardilaand flamer (1979). Anna Maria SanrttAnna participated in the iiitial stages ofthe project. The first section of the chapter, which discusses the iistitutionial set-ting of housing in Bogota, incorporates research reported in Stevensoni (1984)and Paredes (1984). The second section, on Bogota's uniregulated housing mar-ket, is based on Carroll (1980), Paredes (1984), Hamer (1985), and Ardila andHamer- (1979).

1. Ctsanfia5 are severance pay futids that mav be given as advances for houseconstructioll purposes. They are not techniicallv allowed to be given for piraladevelopments, btIt this conditioni does not seem to have been enforce(l.

2. The Stiperintendenicia Bancai-ia, which is otherwise concerned with regu-lating bankinig operations, was put in charge of regulatinig land sales controls in1968.

3. Thiese returns may be compared with effective aniual rates of returil avail-able in Colombia toward the end of the 1970s:

In5trinwnl Percent

UlA(: (index savings and loan) ordinary deposits 19.30

u PA(. savi ngs accoun ts 24.95LPAC savings certificates

Six months 2fi 14One year 27.33

Term deposit certificatesThree months 26.82Six monthls 28.07

Savings accounts 19.(0Agroi idustrial Tilulos (bonds)

Three months 27.50Six months 28.00

Bonos Cafeleros (coffee bonds) 22.00ITitulos de Ahorro Cafetero (coffee savings bonds)

at 65 percent for a two-year term 51.29Certificates of exchange 35.60

These figures are from the June 1979 issue (no. 23) of Astralegia Er1onomica yFinanciera, page 10. cited in Carroll (1980).

184 UNDERSTAND[N(G THE DEVELOPING METROPOLIS

4. See Hamer (1981) for details of model specifications. Here we report onlythe elasticities estimated.

5. For summaries of U.S. hlousinig demand literature, see Quigley (1979) andMavo (1981).

6. See Malpezzi and M.ayo (1987) and Malpezzi, Mavo, and Gross (1985) for

recent reviews.7. See, for example, Follainl and Jimenez (1985); Follain, Lim, and Renaud

(1980); ,Uayo (1982);Jimenez and Keare (1984).8. Othel- household characteristics include sex of household head; famih' size,

(family size) 2 , age of household head, and (age of household head) 2 .9. See Malpezzi and Mavo (1987) andc Malpezzi. Mayo. and Gross (1985), for

example, for reviews of other studies.

Chapter 8

Autos, Taxis, Buses, and Busetas:The Importance of Choice in Urban Transport

The spatial distribution of population, employment, and income deter-mines transportation patterns in a city. Personal travel almost invariablybegins and ends at home. The distribution of employment within a citytells us where people work, shop, and generally transact business. Anurban area's growth and the development of its transport network areinterdependent. Urban population growth accompanied by spatialgrowth stretches the transport network required to service the city. Atthe same time, the availability of an expanding transport network makescity expansion feasible. Thus, the growth of a city is both constrained bythe transport network and facilitated by it.

Transportation technologychanges over time. Radical changes in trans-portation technology can also fundamentally change the physical struc-ture of cities. The introduction of streetcars in the latter part of thenineteenth century made it possible for cities to suburbanize. Later,changes in highway construction technology and the availability of truck-ing made it possible for industrial locations to decentralize. Before theadvent of trucking, intracity goods transportation costs were muchhigher than intercity rail transportation costs. Hence, even heavy indus-try was typically located adjacent to railheads in large central cities, suchas in Sheffield, England.

Technology use changes along with the technology itself. Thesechanges are often linked to income changes. Higher incomes make pos-sible widespread car ownership, for example. This phenomenon leadsto the existence of spread-out suburbs; in the United States, for exam-ple, the introduction of almost universal automobile ownership is gen-erally held responsible for the proliferation of post-W'orld WAar 11residential suburbs. Cities in developing countries are fortunate in that

185

186 UTNDERSTANDING THE DEVELOPING METROPOLIS

a wide variety of transportation technology is already available; thischoice depends crucially on the level and distribution of income in thecity, however. As incomes rise, the possibilities increase, and the introduc-tion of new technologies can have a significant impact on city structure.

Although there are many similarities between the decentralization pat-terns observed in cities in rich countries and those in developing coun-tries, there are also important differences, and these differences aredeterminants of people's transportation behavior. In cities in rich coun-tries, generally, the rich live on the periphery and the poor nearer in orwithin the central city. The resulting transportation pattern includesheavy automobile use. This decentralization also gives rise to vocaldemand for rapid modes of public transport such as subways or efficientcommuter railways. In developing countries, because of the absence ofsuch rapid transpor-t modes, the lack of highway infrastructure, and thedurability of old building structures, the rich continue to live relativelynear the central city, and the poor are often found at the periphery.This raises the costs of providing transportation, and the public trans-port system has to be widespread because the poor do not have access toprivate transport modes.

Aniother difference observed in cities in developing countries is thatem plovment is relativelymore concenitrated in the central city. This makesit feasible to operate high-occupancy vehicles, such as large buses, onradial rotites exhibiting high-passenger density. In cities in developedCOUlltr-ies, employmiienit often has become so decentralized and auto own-ership so high that it is no longer possible to operate high-occupancytranspor-t modes in an economically efficient manner. As a result, publictransport systems in developed countries are typically government-owned and run on continual subsidies. Because the structure of cities indevelopihg coUntlties makes it feasible to operate public transport cost-effectively, it is quite common for these cities to have privately run,unsubsidized public transport networks.

It is therefore important to understand the causes, consequences, andpatternsofurban decentralization. In general, if residencesdispersefasterthan employment, an increase may be expected in average trip distancesand trip times. A continued high concenitration of employment encour-ages a public transport system that is based on high-occupancy public vehi-cles. An alternative that has emerged in cities in developing countries inthe past two to three decades is the use of vans or small buses, known asmicrobuses or minibuses. Because they are smaller vehicles, thev canmaintain high occupancy with smaller loads. Because of the low wagespaid in developing countries, it is more economical to operate thesevehicles there than it is in developed nations. In the United States, forexample, the bus drivers' wage costs account for over half the total oper-ating costs whereas, in cities in developing countries, buses are usually

AUTOS, TAXIS, BLUSES, AND BUSETAS 187

operated by two people-the driver and the conductor. On smallerbuses, like the busetas in Bogota, the driver can collect the fare. Thismakes it feasible to runi the smaller buses at low fares-not much higherthan the fare on a standard full-size bus. These smaller buses bringgreater flexibility to the transport system. Indeed, even with the decen-tralization of employment and residences, mini- and microbuses can be

used to run effective public transport systems. This has been the publictransport response in Bogota.

The decentralization of employment in manv cities in developed coun-tries has meant the actual loss ofjobs from central cities; radial transportnetworks then become inefficient and the low density ofjobs in peripheralareas makes it difficult to operate public transport networks. In cities indeveloping countries, there are similar declines in the employmlent den-sity gradients (see chapter 6), but that is because of an increase in jobs inall areas of the city rather than a decline in the central city. Hence,employment density in central Bogotai has remained high, keeping theradial transpor-t system relatively efficient. To the extent that employmentremains more densely concentrated in the center than do residences,some of the more distant residence locations can be served efficiently byminibuses.

Much of this chapter focuses on understanding the importanice ofincome in determining transportation patterns and choices. The wealthyhave more leisure time and hence indulge in more discretionary travel, atendency that is reinforced by auto ownership. City structure itself isrelated to income distribution within the city. As discussed in chapter 6,higher-income residential areas in the city also have more jobs. Higherincomes generate greater demand and hence more jobs. There is then asurplus ofjobs in these areas and a large degree of in-commutting. Poorresidential areas have significantjob deficits. Overall, the poor make longcommutes to their jobs. This feature of city structure adds to the disad-vantage suffered by poor workers because of adverse residence loca-tions. In Bogota this is exacerbated by a regular pattern of segregationof residenice by income (see chapters 4 and 5). The poor live in thesotith of the city and the r ich predomii.antly in the north; the poor alsotend to live in the outer rings. However, a key finding of this study is thatthe relatively high service levels provided by Bogota's public transportnetwork and the prevalence of flat bus fares substantially mitigate thespatial disadvantage suffered by the poor in that city.

Cities in developing countries tend to have larger households than citiesin developed countries, and many of these are multiple-worker house-holds. Moreover, as primary and secondary education become universal,the household's decisioni about where to live becomes more complex. Theexistence of the informal sector also complicates matters. With many peo-ple dependenit on irregular- employment with a variety of potential job

188 UNDERSTANI)ING THE D)EVELOPING METROPOLIS

locations, there is a greater tendency to locate in densely populated areasthat are likely to have more employment prospects.

The participationi of the private sector in the provision of an urbanitransport systeimi has become quite common in many cities in developingcountries. There is often a vacuum in the range of transportatioll servicesbetween the privately owne(d automobile or two-wheeler (mopeds, scoot-ers, motorcycles) and the publicly provided full-size bus in cities wherethe minibus has not yet colime into widespread use. Cities with a widerange of public transportationi services include Istanibul, where the vari-ous transportation modes include the dolmus (sharecd taxi), the minibus(8 to 10 seats), the midibus ( 13 to 15 seats), and the standard largc bus;Hong Kong, which has large buses. double-decker buses, minibuses (14seats), a large taxi fleet, and the metro; Manila, which also has a verylarge number of "jeepneys" (14 seats) supplementing the full-size busesand a limited light rail system; and Buenos Aires, which, althoughserved by an extensive unider-groulid metr-o system, also has a largc fleetof Colectivo.s (21-seat minibuises) that are estimated to carry about halfthe metropolitan-area passenger-s (Walters 1979; Feibel and Walters1980).

The intioductioni of inter mediate-size buses makes it possible to servethe varying needs of a large passenger populationi efficiently. Large-capacity vehicles can operate efficiently only in densely traveled trunkcorridors. As cities grow anid get more spread out, these vehicles are lessable to satisfy the transportaLtioni needs of people wvhose origins and des-tinations fall between the corridors. MNSoreover, large-capacity vehiclesare better suited to serNing radial routes in cities with dominiiatinig citycenters. As a city expands and both residence and employment loca-tions decentralize, there is an icireasing demand for circumferentialtrips. These routes are unilikely to be densely travele(d routes: large-capacity vehicles can then proxide onlv infrequernt service. The result isa greater tendenicy for travelers to opt for private niodes in order toavoid the inconvenience and lengthy trip times implied by iifrequjenitservices. Although the fares on intermediate-size vehicles are usuallyhigher than those on full-size buses, they are onow enough to cater tomuch of the population. These intermediate-size vehicles provide a low-cost, effective alternative to private modes of transportation: becausethev are smaller and cheaper, manly mnore people can invest in operatingthem. A 15-seat minibus can operate three times as often as a 45-seat busin a corridor of similar transportationi densitv. In Bogota, for example,the break-evcn point in terms of load factor was much lower- for the buse-tas thiani for the full-size buses. More frequent service on less denseroutes was thus possible, waiting times were reduced, and convenie nce

began to approach that of privatc vehicles.

AUTOS. TAXIS, BUSES, AND BUSETAS 189

All these issues of interdependence between city structure, level andspatial distribution of income, labor market and household characteris-tics, and transport demand and supply have been founid to be importantin Bogota.

The Supply of Transportation Services in Bogoti

The residents of Bogoti and Cali have had a rich supply of transporta-tion services from which to choose. These services include full-sizebuses (68 passengers), trolley buses (104 passengers), busetas (32 passen-gers), microbuses (15 passengers), and regular and collective taxis(rolertivos). The full-size buses and busetas operate on1 a flat-fare basis ondesignated fixed routes but are free to stop to pick up passengers any-where along the routes: there are no fixed bus stops. The taxis cruisealong streets to pick up passengers. The colctivos operate on fixedroutes, usually during the peak hours, also on1 a flat-fare basis. Walkingand driving privately owned automobiles are two additional options.Motorcycles, scooters, and bicycles are very rarely used in Colombiancities.

UJnlike most cities in developed countries, where public tranisporta-tion is typically government-owned and -subsidized, almost all of thepublic transportation in Colombian cities is privately owned and pro-vided, although it, too, is subsidized by the government to an extent.Similar systems exist in other Latin American countries, includinigArgentina, Brazil, and Chile. The District of Bogota does run a publiclyowned bus company (Empresa Distrital (le Transporte Urbano [EDTU]),

but it accounts for only about 1 percent of the city's transport system.The governmenit provides subsidies for runninig the full-size buses,which keeps the fares low, but the busetas are not subsidized.

Substantial growth took place in Bogota's transportation services be-tween 1972 and 1978 (see table 8-1). In 1980 the buses and busetas wereoperated by thirty-nine urban transport companies, of which twelvewere cooperatives, twenty-five were limited liability companies (affiliatedcompanies), and one was a publicly owned transportation company.Rights to route operation are awarded by the state to the companies.which then lease the rights to affiliated bus owners. Some companiesoperate their own buses. The managers of these affiliated cornpanieswield considerable power in the transportation sector, but the system ofcooperatives and affiliated companies allows the participation of manysmall owners in the public transport system.

Bogota has had a high level of public transportation services for sometime. In 1980 the buses and busetas supplied 123 seats per 1,000 resi-

Table 8-1. Transportation in BogotAi

Appr/xmimate n umberofvehirles Fare strucitrre,

I pe of service ehiicles 1972 1978 Passenger capacity 1978 1980 0wn7Oship

Taxi European and 7.600 13,530 4 and 5 seated $12 tlag T he drivers own approximately 70Amer-ican cars $1 ( per km percent

$25 more from8 P.M. tO 5 A.M.and Sundays

Colectivo I uropean anld 370 606 4 andc 5 seated $25 8 AM. to Same as aboveAmerican cars 5 P.M.

Microbus Small buses w/light 996 252 15 seated $3.00 $6 39 private companiestruck chassis $3.50 $6.50 nights

and Sunidays

Buseta Small buses w/inter- 615 3,289 24 seated $4.00 $7.50 Same as abovemedliate truck chassis 8 standing $4.50 $8.00 Sundavs

Bus Buses w/school- 2.782 6,289 38 seated $1 .50 $3.0() Same as abovetype body & large 30 standing $3.50 nightstruck chassis $4.00 Sundays

Bus (diesel) 194 35 38 seated $1.50 $3.00 Urban transport district company30 standing $3.50 nights

$4.00 Sundays

Trolley bus Russian and some 118 37 48 seated $1.50 Same as above Same as aboveAmerican buses 56 standing

Source: Pachon (1981b)

AUTOS, TAXIS. Bt'SES, AND BtSETAS 191

dents, and taxis and colectivos added another 12 seats per 1,000 resi-dents. This compares very favorably Wiith Hong Kong, a large citygenetallv regarded as being well proxided by public transportation ser-vices, which had 108 seats per 1,000 supplied by buses and public lightbuses (minibuses). about 37 by rail services, and an additional 14 bytaxis (Hong Kong Gover-1nment Transport Department 1983). In coIn-trast, the net transportation supply in Lagos was only about 39 per 1,000residents-15 supplied bv buses, 12 bv miniibutses, 6 by taxis, andanother- 6 by other informal modes (Transpoconstilt 1976; WilburSmith and Associates 1979). The supply in Cali was also almost 100 seatsper 1,000 residents in 1980.

The level of automobile ownership in Colombia has been relativelylow. Automobiles show the highest aniniual growth rate, at about 8 per-cent, followed by trticks at 7 percent, and buses at 5.5 percent (see table8-2). Bogota' accounts for a large proportioni of vehicles in the coun-try-about 25 percenit (see table 8-3). In 1978, however, there were onlyabout 45 auitomobiles per 1,000 residents in BogotA and 32 per 1,000in Cali-as compared with 78 per 1,000 in Mexico City, 74 per 1,000 inBuenos Aires, and 91 per 1,000 irn C,aracas (UIrrutia 1981 ).

The Regulation of Supply: Control, Subsidies, and Incentives

This largely private provision of public transportation in Bogota is gov-erned by a relatively complex system of government regulation and sub-sidies. Governm11enit agencies set routes and fares for buses and buselas aswell as for taxis and the colectivos; subsidies are provided to bus compa-nies for their bus operations; subsidizecl credit is made available for thepurchase of both buses and buselas; and the number of licenses grantedfor bus and busela operation is kept tinder control. Thie government

Table 8-2. Vehicle Stock in Colombia, 1970-80

Year Cars Buses 71'1'5Ibtol

1970 284,252 28,875 47.318 360,4451971 303,712 30,065 49,368 382,1451972 324,712 31,315 51,548 407,5751973 347.378 32,625 53,843 433.8461974 374,778 33.782 57.168 465,7281975 397,641 35,273 60,447 493,3611976 423,655 37,422 63,894 524,9711977 463.817 39.810 69.437 573.0641978 511,822 42.288 79,985 634,0951979 553,955 46,693 88,316 688,9641980 601,463 50,167 94,888 746,518

Sourae: Pachoni (1981a).

Table 8-3. Vehicle Stock in Bogoti, Cali, and all of Colombia, December 1977

Bogota (ali Remainder of Colombia 7btal

V'ehicle N.vmber Percen,t Number Percent No ,nber Percen I Numler Percent

CarsPrivate 95,451 24.69 30,553 7.90 260,543 67.40 386,547 1((.00Public 13,147 15.70 3,580 4.27 67,010 80.02 83,737 10().00Totala 114,057 23.58 35,231 7.28 334.383 69.13 483,671 100.00BusesPrivate 1,282 22.65 283 5.00 4,094 72.35 5,659 100.00Public 10,538 25.83 2,8(0(0 6.86 27,458 67.30 40,796i 100.00

;D TotalP 12,402 25.94 3.122 6.53 32,284 67.53 47,8(08 100.007Tr cktPrivate 15,40f; 31.43 2,837 5.79 30,770 62.78 49,013 100.((Public 5,653 10.03 2,310 4.1(0 48,374 85.87 56,337 100.00Totala 22,297 20.26 5,609 5.10 82,16(0 74.65 110,066 100.00

otIalPrivate 112,139 25.42 33,673 7.63 295.4(07 66.95 441,219 100.00Public 29,338 16.22 8,690 4.80 142.842 78.97 180,87(0 100.00Total, 148,756 23.19 43,962 6.85 448,827 69.96 641,545 100.00

a. Iiclides vehicles registered k,r official use as well as puhiliclv and privately owned vehicles.Source:Pachon (1981a).

Al.TOS, TAXIS, Bt SLtS, ANI) Bt SE FAS 14:

maintains a mild policy of restrictive entry-a license k-nown as "notaopci6n" is needed before a vehicle enters public service. Although theadministration of tilese regulations and subsidies Couldl be made moreefficient, the system has provided a high level of transportatioll servicesat relatively low social and private cost. Bus subsidies made possible faresthat are below the average costs of tihe suppliers, thus providing thepoor with better access to urbani transport. Our analysis shows that thewhole regulatorv and subsidy system performs in favor of the poor.

The responsibility for the overall regulationi andc policy for an urbaintransport system rests with the Instituto Nacional (Ic Trarisporte (INi RA).

INFRA typically delegates route setting and enforcemiient of transporta-tion legislation to the respective mayor's offices. In Bogota the relevalltlocal authoritv under- the mayor's direction is the Departamento Admin-istrativo de Transporte y Trinsito (DAtY). DA-TT is supposed to reviewtransportation demanidl constantly and assigo new r outes or extendexisting ones. It does this by inviting tenders foI bidding the numberof vehicles ancl frequenicy of operation. l)emancl is ofteni identifiedthrougih pressuire from commtinity actionl boards (fnntas de Acci6nCoomunal) This meanis that the identificationi of tranlsportatioll needsand the subsequent openinig of r-outes are subject to political pressure.and the process is often strongly criticized on this account. Actual r outeassignment is done on the recominmenicdationis of DATI by a committeeconsisting of the mayor, representatives of INIRA, the Ministr y of PublicWorks and Transportation (Minister-io de Obras Publicas y Transporte),DArT, and the Departamnento Acdi nistrativo dc Planeaci6n l)istrital(Da,PD) (Cifuenites 1984). The route is generally granted if the biddingcomnpanly canl conxince the authorities that therc is enough demanidand that they would not be cr-ow(litig out existing operators. Almost allthe bus companies are affiliated with one of the two national urbanitransport organizations-National Corporation of UTrban Buses (Corpo-raci6n Nacional cle Buses Urbainos, (c;R0PBusEs) and th(e ColombianFederation of U rban Transportatioln (Federaci6n Colombiano de Trans-por te l rbano. FECOTRANS)-which are r ecognizedl by the governmentas representinig urbani transpor-t interests. Once a companv receives ther ights to a r oute, that route is assigned to the individuial affiliated opera-tors on a leasing fee basis.

Despite complaints about the politicization of the routinig process, thesystem operates well in matchirig tranisportation suppIV with perceivedand expressed demand from consulmers. It mav be judged superiol tomost systems, which are centrally run ancd where; routes are assignedl in amore technical, information-intensive manner: "It has been estab-lished that the existing route assigiinmenit svstemIl in Colombia produces anetwork with wide coverage that has adapted to changes in trip dlemadiclpatterns" (Paclioni 1987, p. 146). Tler-e are, however-, som1e negative

194 trNDERSTANDING THE DEVELOPING METROPOLIS

results, such as congestion on some desirable routes, particularly in thecenter of the city.

Drivers' remuneration is based on the number of passengers trans-ported. This leads to some chaos on the streets as drivers of competingbuses ruslh to get passengers. Drivers get only about 10 to 15 percent offare box revenues (Avendano and d'Anico 1981)-about 12 percent in1980-compared with the 83 percent that bus drivers in American citiesaverage (.4PTA 1980). This emphasizes how different the economics ofbus operations are in low-wage countries and why it becomes efficient tooperate smaller buses in such cities.

The cost of operating this system is quite low (Urrutia 1981; Intercon-sulta I.tda 1970). The private operator's cost was about Col$352 per pas-senger in 1980. and per-kilometer operating cost was about Col$21. Thestate-owned companiy, Empresa Distrital de Transporte Urbano (EDTU)

had muchi higher costs. mainly because of higher wage costs per workeras well as the larger number of' workers employed per operating bus.Drivers for private companies are relatively poorly paid and work longhours-the average driver worked about twelve hours a day, about halfwor-ked seven days a week, and most of the rest worked six days a week.Their wages were just over minimum-wage levels. In 1980 bus driversearncd an average of Col$6,400 per month, whereas buseta driversearned about Col$8,300 per montlh. They generally have no sick pay,health insurance, or unemploynmenit benefits. Their earnings are roughlywhat might be expected for their level of education and skills in theBogota labor market. When bus services are run directly by government-owned agencies, wages of drivers and other workers generallv rise as aresult1 of' union and other pressures, leading to higher- transportationcosts. The relatively low cost of the transport system in Bogota is in largepart a benefit of its privately owned, government-regulated, decentral-ized status.

The rationiale for the urban tranisport policies maintainied by the gov-ernment is therefore clear. The government has wanted to maintain lowpublic transportation prices so that all city dwellers have reasonable accessto employment and education opportunities.

In order to mainitaini low fare levels, all buses are given flat-rate, per-vehicle monthly subsidies that roughly cover the fixed costs and returnto capital. New vehicles receive much larger subsidies than old vehiclesin order to encourage companies to purchase newer vehicles with loweroperating costs (see table 84). The Financial Transportation Corpora-tion (Corporaci6n Financiera (le Transpor-te, CFI') provides subsidizedcredit fto r the pturchase of new buses; this amounts to nearvly one-fifth ofthe vehicle's value. Overall, the subsidy structure encouraged vehiclereplacement and reduced the average age of the fleet. Analysis alsoshowed that it benefited the poor more than others.

AUTOS. TAXIS. BUSES, AND BLUSETAS 195

Table 84. Bus Subsidies in Colombian Cities, 1977 and 1978(Colombian pesos in current prices)

Model Near of vehicle

1959 and 1974 andYear and quarter earlier 196%0-64 1965-69 1970-73 later

1977

January-March 9,383 11,133 12,633 13,633 21,133April-june 9,620 11,370 12,870 13,870 21,370july-September 10,640 12,390 13,890 14,890 22.390October-December 10,610 12,360 13,860 14,860 22,360

1978janutarv-March 11,630 13,380 14,880 15,880 23,380April-June 11,600 13,350 14,850 15,850 23,350july-September 12,630 14,370 15,870 16,870 24,370October-December 12.590 14,340 15,840 16,840 24,340

Source: Cifuentes (1984).

The fare structure in 1978-80 has already been outlined (see table 8-1).To put the data in perspective, it should be noted that bus travel costsonly about 3.8 cents per trip (1978 U.S. dollars), or an average of 0.6cenits per passenger kilometer (average bus trips were 6.5 kilometers pertraveler). Buselas cost about 10.3 cents per trip. At these rates, even alow-income household in one of the poorer peripheral parts of Bogotawould spend less than 2 or 3 percent of its monthly earnings on worktrips. ln 1980 revenues from bus fares covered the variable costs of busoperation (about two-thirds of total costs), and subsidies covered thefixed costs and return on capital. Through this system, passengers getthe benefit of low tranispor-tationi costs while bus owners get healthyreturns oni their capital. In the case of busetas, the fare (Col$7.50 pertrip) more than covered total operational costs along with a healthyreturn on capital, so no subsidies were needed. It is not surprising thatthe buseta fleet grew much faster than the bus fleet in the 1970s. Onereason for the existence of these profits is the governmenit's mild policyof restrictive entry.

The subsidy levels are continually varied to account for increasingcosts. The authorities maintain an index of input costs to reflect chang-ing transportation costs. The changes in this index help to inform theauthorities and bus owners when the fares and subsidies have to beadjusted.

Is there an adequate rationale for giving differential subsidies by age?It is difficult to analyze this issue because many variables enter the actualeffective costs of bus operation. Some sttidies have found no age-i-elatedstatistical differences in operating costs. These studies did not adjustadequately for the differences in operation between buses of different

Table 8-5. Adjusted Costs, Adjusted Income, and Capital Return for Buses, October 1980(Colombian pesos per month)

E'stimated AdjustedModel Age (Nears) vanable cost' Jixed cot5 h Iistalcosts Fstimated income' Subsidies Gapital return

1959 2Y1 55.8 3.7 50.5 47.1 24.9 12.41960 20 56.8 3.9 60.7 48.2 25.8 13.31961 19 57.9 4.0 61.8 49.2 25.8 13.21962f 18 58.9 4.1 63.0 50.3 25.8 13.11963 17 60.0 4.2 64.2 51.4 25.8 13.01964 16 61.1 4.3 65.4 52.5 25.8 13.01965 15 62.2 4.4 66.0 53.7 26.7 13.81966 14 63.3 4.6 67.9 54.9 26.7 13.71967 13 64.5 4.7 69.1 56.0 26.7 13.61968 12 65.6 4.8 70.5 57.3 26.7 13.51969 11 66.8 4.9 71.8 58.5 26.7 13.41970 10 68.1 5.1 73.1 59.8 30.4 17.11971 9 69.3 5.2 74.5 61.1 30.4 17.01972 8 70.6 5.4 75.9 62.5 30.4 17.01973 7 71.9 5.5 77.4 63.8 30.4 16.91974 6 73.2 5.7 78.8 65.2 42.8 29.21975 5 74.5 5.8 80.3 66.7 42.8 29.11976 4 75.9 6.0 81.9 68.1 42.8 29.11977 3 77.2 6.2 83.4 69.6 42.8 29.(1978 2 78.6 6.3 85.0 71.1 42.8 29.01979 1 80.1 6.5 86.6 72.7 42.8 28.91980 0 81.5 6.7 88.2 74.3 42.8 28.6

Note: In notes a, h, and c above, numbers in parentheses are t-statistics.a. In (variable costs) = 11.31 - 0.018 (age)b. In (fixed costs) = 8.81 + 0.028 (age)c. In (income) 11.22 - 0.022 (age)

R2= 0.959 (8.4) R2=0.985 (14.1) R 0.937 (6.7)Source: Pachon (1987).

AUTOS, TAXIS, BLUSES, AND BUlSETAS 197

vintages, however. For example, newer buses, being more trouble-freethan older buses, are generally used more intensively than older ones.Bus operators may press the older buses into service only in the rushhours, for example. It was estimated that the per-kilometer cost of oper-ation fell by only about I percent for every four years of age, but thisresult itself could be different if all buses were operated for the same dis-tance covered: the operating cost of older buses would be likely toincrease with distance traveled.

It may be more useful to examine the effect of age on costs, revenues,and return on capital (including subsidies) on a monthly basis. Thiswould substume the adjustments in operation made by the bus operatorsin response to the subsidies offered. Because the subsidy is given month-ly, they are free to alter the intensity of bus use in order to maximize in-come (see table 8-5). Data on bus operating costs and revenuesaccording to model year of buses were obtained from INTRA and FECOL,

TRAN, and relationships were found between variable costs, fixed costs,revenues, and age of bus. It may be observed from table 8-5 and fromthe estimated equations cited that the revenue falls faster with age thando the variable costs. The gap between revenues and costs increasedwith age. Much of this is clearly related to differences in intensity ofuse-newer vehicles were used with greater intensity than older vehi-cles, and higher ratios were obtained with newer vehicles.

Profitability rates must be calibrated with respect to capital prices,however. In order to investigate the comparative profitability of busesand lusetas in relation to their age, price equations were estimated posit-ing a simple exponential form of depreciation (Pachon 1987).' The

equation for buses is:

In PIa= 14.4 - 0.069a R2 = 0.77(141.1) (8.3)

and the equation for busetas is:

In P9 a = 14.1 - 0.069a R2 = 0.85(141.1) (8.6)

where PI,a is the price of a bus at age a, and P2a is the price of a buseta atage a. (The numbers in parentheses are t-statistics.)

It is remarkable that both equations reveal consistent results of about 7percent annual depreciation. This is lower than in the cost studies madeby the bus companies and used for setting fares and subsidies. The higherdepreciation rates actually used help produce higher real profitabilityrates for older buses.

198 UNDERSTANDING THF DFVEI.O'IN(; METROPOIIS

The results from these equations show that, because of the level of subsi-

dies given, older buses had higher profitability ratios, whereas the profit-ability profile of busetas was constant until the busetas were about tenyears old, and then it declined somewhat. Because busetas did notreceive subsidies, their results may be seen as the no-subsidy counterfac-tual case. Whereas some allowance may be made for errors in price esti-mation, the result indicates that bus subsidies could be more smoothlyand sharply tapered with age. Adopting a more realistic rate of deprecia-tion could help correct the subsidy profile. It is also possible that actualoperatinig costs of older buses may have been underestinmated. It is, forexample, difficult to account appropriately for the costs arising from theamount of time owners spend on repeated repairs of olcder buses.

These results suggest ways in which transportationi regulations (grant-ing of routes, provisioni of subsidies) can be analyzed. The system ofurban transport that exists in Colombia has beeri operating remarkablyefficiently, benefiting suppliers and demanders alike. There are, how-ever, several areas that could be refined witlh the help of this type ofanalysis. Administration of subsidies could be improved. Congestion insome routes could be removed by more careftil route assignments or byraising the bid price for routes. Traffic regulation and establishment ofbus stops would remove sotme unnecessary chaos on the streets.

The Impact of Government Regulation, Taxes, and Subsidies

An extensive evaluation of the redistributive impact of government taxesand subsidies in the transportationi sector was conducted by Pachon(1981b). In addition to thebussubsicliesdescribedearlier, thecColombiangovernmenit had, for a long timie, kept fuel prices low and taxed privateautomobiles relatively heavily. The task of evaluation is to estimate the netredistributive impact of these subsidies and taxes on different incomegroups. By conventioni, a policy can he described as progressive if thefraction of income paid as taxes increases with income level or if thefraction of income received as subsidies decreases with income level.See table 8-6 for a summary of the various tax/subsidy interventions.

Private cars were subject to iour taxes in Colombia. In 1980 the cus-toms duty on imported cars was 150 percent, and onl componenits forcar assembly it was 10 to 15 percent; sales tax on private cars was 35 per-cent; a registration fee was collected at both national and state levels, ata rate of about 0.05 percent of capital valuLe; and the value of cars wasalso assessed as part of the wealth tax. The information about thenational registration fee and wealtlh tax was inadequate to be accountedfor in this evaluation.

AUTOS, TAXIS, BISES. ANI) BUSETAS 199

Table 8-6. Formal and Actual Government Intervention in the TransportSector, Colombia, 1980

Formal inoerlenlitn Actual itter ention

Regislration jees for vehiclksCol$0.40 and $0.50 per each $1,000of'vehicle's valie The same

Sales tax35% of notionali7ed value 22% on capital cost

Customns duty150% cIF V alue of impo>rtedcars; 10-2W0 (CIF value imported 37% on capital cost; 5.64%compouentsa oni capital cost

Fuel tax110% on refinery price; 12% sales tax The same for private vehicles

BusfaresSubsidv for owners according to age Transferred to consumersof vehicle for public transportation

Fuel subsiidyDoes not exist Col$27.00 per gallon

Note: Capital cost is calculated as an annual implicit rental pr7tce of capital after account-ing for depreciation; returin is calculated as an original outlay ift real terms.

a. Coist, insuranice, and fieigit.Sounre: Pachon (198 1b).

A 12 percent sales tax was levied on the wholesale price of fuel, and110 percenit tax for thle Highway Trust Funid (HTF), was levied on refin-ery prices. In 1980 the sales tax amouLited to Col$2.48 per gallon andthe HTF tax to Col$9.10 per gallon. These taxes were actually based onsubsidized fuiel prices. The government petroleum agency, ECOPETROI.,

sold gasoline at a price below opportunitv cost: in May 1980 a gallon offuel was bottght at US$1 or Col$47 per gallon but sold by ECOPETROL atCol$20.12, implying a subsidy of about Col$27 per gallon. Taxesreduced this subsidlv to Col$ 15.40 per gallon.

In analyzinig the impact of these interventions on income distribu-tion, a number of assumptions hiad to be made. The travel informationgleaned from the 1978 World Bank-DANE Hlousehold Survey was used toassign the incidence of taxes and subsidies according to the trip infor-mation for each inconme group. Private cars were assumed to get 40 kilome-ters per gallon ol ftuel. Fuel consumption per trip by public transportationwas assigned on the basis of average trip distance. Given the iniformationon vehicle makes and age from the survey, it was possible to calculatethe average price of cars by make in Bogota in 1978 and then to find aw^eighted average for all makes. The incidence of local registration fees

200 UNDERSTANDING THE DEVELOPING METROPOLIS

could then be assigned. Similarly, the incidence of sales taxes was calcu-lated after accounting for the age distribution of cars and the increasein the vehicle's average price attributable to new vehicle taxes. A salestax on new vehicles is also passed on to the value of old vehicles: becauseof high sales taxes, the value of old cars gets inflated and the incidenceof sales tax on new vehicles is also passed on to used-car buyers. It waspossible to estimate the impact of sales tax on car value by age and thento assign it according to ownership of vehicles by income groups. Differ-ential effects had to be calculated for imported and domestic carsbecause of the difference in customs duty rates (see tables 8A-1, 8A-2,and 8A-3 in the appendix to this chapter).

The incidence of fuel taxes and subsidies was calculated by assumingan average consumption of fuel at the rate of I gallon per 40 kilometersand a distance driven of about 1,250 kilometers per month. The tax/subsidy incidence was estimated according to cars owned and averageincome by quintile group (see table 8A4 in the appendix to this chap-ter). Similarly, the incidence of bus subsidies and of fuel taxes and subsi-dies on users of public transportation was also assigned by incomegroup according to the pattern of modal choice and trip data collectedin the Household Survey.

Vehicle taxes appear to have a V-shaped pattern of incidenice, beingregressive at low income levels and progressive at high income levels(see table 8-7). It is somewhat puzzling to observe high tax incidence atlow-income levels. Two explanations are possible. First, the data Lised inthis exercise were based on current monthly household incomes: thefew automobile owners in low-income categories could have been soclassified because of transitory low incomes during the period of thesurvey. As may be seen in tables 8A-1 through 8A4 in the appendix tothis chapter, there were only five to six cars per 100 households in thefirst two quintiles. With average incomes being very low, the averageincidence still comes out high. Second, the calculations have been made

Table 8-7. Vehicle Taxation Incidence in Bogota, 1978(estimaLed tax paid as a percentage of family income)

Incomnquintiles Sales tax Customs duts Regisration fee 'lotal

1 1.22 1.09 0.14 2.452 0.61 0.54 0.07 1.223 0.53 0.47 0.06 1.064 0.71 0.63 0.08 1.425 1.01 0.89 0.12 2.02

City average 0.87 0.77 0.10 1.74

NVole: Quintiles arle in ascending order of household income.Source: Tables 8A-I throtigh 8A-4 in the appendix to this chapter.

AUTOS. TAXIS. BUSES, ANt) BtISEIAS 2(11

Table 8-8. Fuel Policy Incidence in Bogot., 1980(net subsidy as a percentage (If tamily income)

Incomequinbloes Car Bus Busela 7;1(al

I 0.79 1.17 0.25 2.212 (0.39 0.87 0.23 1.49.9 0.34 0.65 0.23 1.224 0.46 0.35 (.18 (.995 0.65 0.06 0.07 0.78

City average 0.57 0.29 (.13 (.99

No,n: Quiinmiles are in ascending order of household income.Source: Tables XA-I through 8A-4 in the appendix to this (hapter.

assuming that car ownership and usage are homogeneous over all car-owning households. It is quite likely that low-income car-owning house-holds own, on average, older and cheaper cars and tise them less. If thiisis the case, the methodology used would overstate the incidenice of velhi-

cle taxes on low-income groups.The results fiom the incidence of net subsidies give an unequivocal

progressive incidence (see table 8-8). This is a conseqtuence of the netsubsicly (througih lower fuel prices) to buses and busetas. This somewhatcounterintuitive result is a suiprise because it has tisually been argtuedthat fuiel subsidies are essentiallv regressive in nature. The result is aconsequen-ce of the heavy use of public triansportation, which over-whelms the effect of fuel subsidies resulting from autonmobile use. Thenet result of the subsidies and taxes on BogotA's transport systemii isclearly progressive, with the bus transportation subsidy dominiating theoverall effects of the system (see table 8-9).

These results have been presenited in some detail in or(ler to illustratethe possibilities of analyzing quantitatively the effects of governmentpolicies. These policies are often made in an ad hoc manner. It is oftenassumed that it is not possible to conduct such policy atialyses. Our taskwas made feasible bv the availability of a rich set of household level datacollected especially for this purpose and supplemented by other data onstocks of vehicles available, operational details of public transport, andl

the like.

Travel Patterns in Bogota

The design of an tirban transport system requires an understandingof travel patterns as they exist in a city. There are various ways to look attravel pattern1s in a citv. One is to look at the outcoime of the variegatedtravel needs of city residents: this may be done bv looking at travel flow

Table 8-9. Transport Tax and Subsidy Policy Incidence: Bogota, 1978-80(tax or net subsidy as a percelltage of family income)

Income Public Privaltquintile Transport subsids Bus Buseta Total Vehicle taxation Fuel Net tax Fotal

1 3.63 1.17 0.25 +5.05 -2.45 +0.79 -1.66 +3.392 2.71 0.87 0.28 +3.86 -1.22 +0.39 -0.83 +3.033 2.95 0.65 0.23 +3.83 -1.06 +0.34 -0.72 +3.114 1.09 0.35 0.18 +1.62 -1.42 +0.46 -0.96 +0.665 0.19 0.06 0.07 +0.32 -2.02 +0.65 -1.37 -1.05

City,average (.99 0.29 0.13 +1.41 -1.74 +0.57 -1.17 +0.24Note: Quinitiles are in asceniding order of household income. Net subsidies are shown as positive and taxes as negative.Source: Pachon (1981 b).

AtlTOS, TAXIS, BLUSES, AND BUSETAS 203

between various areas, average speeds in different transportation corri-dors, travel time and distances traversed, and the varied patterns thatemerge from travel for different purposes. Another way is to look at therequirements for travel at the household or individual level. Why c(opeople travel within the city? What determines their frequency of traveland the speed and distance traveled? How important are variables likeincome, age, sex, worker status, and residence and work locations? Forsome purposes it may be adequate merely to observe the outcomlles; forothers, particularly long-term planniing for urban utansport, it is equallyimportant to understand the whys and wherefor-es.

A standard way of describing travel patterns is to construct origin-destination matrixes based on some spatial disaggregationi of the city. Inthe case of Bogota, we continue to use our standar-dized zonal disaggre-gation by ring and sector. The trips undertakeni can also be broken upaccording to their purpose-commuting between residence and work-place, commuting between residence and school, and trips not relatedto work or school. Among the latter, if the data are rich enough, it mayalso be possible to distinguish between necessary trips (such as dailyfood shopping) and discretionary trips. Once the residential and work-place location pattern is well understood, it is relatively easy to predictthe nondiscretionary travel pattern in a city.

It is also necessary to describe the travel pattern according to mode oftransportation used. An intensively used public transport system re-quires less road space than a tuansport network primarily based onprivate transportation. Household income levels emerge as prime deter-minants of transportation mode as well as of travel frequency. A particu-lar concern in evaluiatinig the travel pattern in Bogota has been therelative accessibility of low-income workers to workplaces and other des-tinations. The highly income-segr-egated pattern of residlence makes thisan especially key issue.

Our main sources of data were the 1972 Phase 11 Survey and the 1978World Bank-DANE Household Survey. Wher-eas the 1972 survey hadexhaustive information on all trips, the 1978 survey had information onwork trips only.2 According to the 1972 survey, travel in Bogota was pre-dominantly nondiscretionary-40 percent of all tnips were work trips, 36percent were school trips, and only 24 percent were other. Of the lastcategory, 18 percent were home-based and 6 percent non-home-based.Hlence, nondiscretionary travel probably accounted for more than 80percent of utips. Information fiom cities in other dieveloping countriessuggests that nondiscretionary travel may be expected to account for 60to 80 percent of trips in most of these cities. In the tJnited States, bycontrast, only about 25 percent of trips in medium-size cities are charac-terized as work trips, as many as 55 percent may be home-based non-work trips, and almost 20 percent are characterized as non-home-based

201 U'NDtERSrAN DING; THE ID)F:VEl0t'NG(; METROPOi.S

trips. Ther-e is cleariv mucii mor-e discretionary travel in rich countr-ies,because of higher incomes, the flexibilitv provided by widespread autoownership, and the spread-out natture of services and employment thatmakes multiple-destination trips necessary. An interesting findinig forBogota is that, adjusted for income, automobile-owning householdsdo not make more cdiscretionary' trips than similar households Witiloutaiitos.

Dib/sa e and T ravel Time

T'he average distance traveled byadult w(orkerswas aboutt8.3 kilometers-8.5 kilometers for workers without automobiles and 7.6 kilometers forautomobile owners. Studenits' school trips were mnich shorter-5.3 kilo-meters on average, 5.1 kilometers for those from households withoutautos ancd 5.8 kilometers for- thiose frtom auto-owning households. Foradult nonworkers, the average trip was 7.6 kiloineters-8.0 kilometersfor those from households without autos and 6.5 kilometers for thosetromil auto-owning houselholds. An inter-esting fact is that auto-owningworkers tralveled shorter- distances in contrast to patterns toulid in citiesin rich coulinties. Richer households live closer to the central city andpoore r housellolds are clisproportionately located at the city peripheryThc average time spent in claily travel bv workers and studenits alike wasabout I 10 minutes Jor non-auto users and 83 minutes for people fromatito-owninig househiolds. On this basis the speeds work out to 12 kilo-meties to 15 kilometers an hour. The substantiallv shorter distances forstu(lents are mostly because of' the proximity of primary schools to localneigihborhoods: many children walk to school, and, although the dis-tanlce is m11uchi less than for workers, the time spent traveling is similar.Autom11obile travel was abouit 3 kilotmeters to 4 kilometers per hour fasterthani the public transpor-tationi.

One issue worth commentinig on1 is Zahavi's theor-v of travel behavior:

he posited that people have relatively fixed "travel-time budgets" inde-pei(lenit of income. They choose their residential location partly on thebasis of the transportation available (Zahavi 1979). Our data are clearlyinconlsisteint withi this conjecture: average time spent traveling per dayfalls as incoime rises (sec table 8-10). There is much more to a house-hold's bchavior regar-dinlg its choice of location for residence and itstravel pattern thanl the concept of a fixed "travel-time budget."

Because informationi was available for work trips made in both 1972anl 1978, a number of utseful comparisons can be made Average traveltimle increased ftor all income groUps; the larger increases were amongmiddle-income grouips (see table 8-l(). T1'ravel time increased for bothipublic transportation ancd automobiles At the higher income levels, twocontlictinig trenidis are at work. Automobile ownership increased sub-

AULTOS, TAXIS. BUSES. AND BUSETAS 2(5

Table 8-10. Work Trip Travel Time by Income Level: Bogota, 1972and 1978, and Cali, 1978(niinltes per trip)

Atfon,lhl hkusehod untcomne(1972 (ol$) Bogohi, 1972; Bogoli, 1 9 78 b Cah, / 9 7 8 )

0-500 43.7 41.2 27.3500-1,00() 40.2 43.6 34.11.000-1,500 39.3 43.8 30.21,500-2,000 40.6 46.6 39.62,000-3.000 36.8 48.6 37.43,000-5,000 33.3 44.8 36.95.000-15.000 28.5 38.4 32.415,000-30,000 22.6 30.9 20.9More thani 30,000 24.4 27.6 -

Overall average 35.4 42.3 34.6

- Not available.a. Calibrated travel times.b. Repor-ted travel times.Source: 1972 Phase If Survey; 1978 Worl d Ban k-D,xNF Household Sursev.

stantially between 1972 and 1978-this tactor would tend to result inreduced travel times. Concurrently, there was an outward shift in resi-dence locationi patterns. The net result seems to have been an increase

in average time traveled, but the increase was muich less than the in-crease in travel time for bus and buseta trips (see table 8- l ). Buseta tripstook longer mainly because of the expansion of buseta ser%ice: in 1972buse-a service was rare in peripheral poorer areas. The increased bus triptimes are partly the resuilt of increasing congestion in the city center andpartly the resuilt of the poor's decentralizinig pattern of residence loca-tion. The distance data were unfortunately not comparabie across thetwo surveys.

Table 8-11. Work Trip Travel Time by Transportation Mode: BogotSi,1972 and 1978, and Cali, 1978

Aiznue.s per trip Kilometers

ButAetal Bus etalC(ar Bus ,n,icohu.s (Car Bus mirrolnms

Bogotai

1972 23.5' 40.6 ' 33.5' 6 .7 h 8.0" 6.3'1978 27.6' 54.8' 47.8' 50"d 6.5"1 6.0"

Cali1978 24.5' 44.7' 47.8' 3.3d 3.3"1 4.4d

a. Calibrated travel time.b. Reported tr-ip distances, all trip purposes.c. Reported travel timie.d. Airline (listances, centroid to centroid.Source: 1972 Pliase 11 Stirvey: 1978 World Bank-i)u,xF Household Suirvev.

Table 8-12. Ring of Residence and Ring of Employment: Bogota, 1972 and 1978(percetnt)

PReiden,e kmrploed in ring I Emploved in ring 2 Employed in ring 3 Fm.iployred in nrig 4 Emploved in ring 5 Employed in ring 6 Totalnng 1972 1978 1972 1978 1972 1978 1972 1978 1972 1978 1972 1978 1972 1978

1 66.0 71.9 8.5 16.1 6.6 5.5 10.8 4.6 8.1 1.9 (.0 0.( 1(0.0 100.02 27.1 18.2 50.9 59.6 9.0 8.9 t6.7 7.4 5.9 5.2 0.4 0.7 100.0 100.03 30.3 12.8 9.5 15.2 43.3 52.1 9.5 10.0 6.1 8.7 1.4 1.2 100.0 100.04 25.6 14.7 9.9 15.3 13.6 12.8 39.2 45.1 10.3 11.3 1.3 0.8 100.0 10(.05 21.1 11.1 11.7 12.4 12.1 11.7 13.0 14.9 40.7 48.1 1.3 1.8 100.( 100.06 '30.8 10.0 8.8 9.3 9. 5 13.2 8.0 16.8 9.4 9.4 33.3 41.3 100.( 1(0.0

Total 25.5 14.5 15.1 18.7 16.2 1 5.9 20.8 21.4 20.6 26.1 1.9 3.4 100.( 1WO.0

Sofre : 197 2 Phase 11 Smmrse : 1978 W'orld Ba nk-D.A N Ho usehold SU rvev.

AU FOS. l AXIS, BUSES, AND BLASE FA.-S 207

How did the spatial pattern of work travel change between 1972 and1978? The most striking change was the large decline in the proportioniof workers employed in the (CBD (ring 1), although the decline was slightin absolute terms. Between 1972 and 1978, employmenlt in the CEBD fellfrom almost 25 percent to about 15 percent of total city employment(see table 8-12). Tlle CBD lost only about 36,000 jobs in this period, butthe rest of the city gained almost 350,000 jobs. Anothier interesting fea-ture of the change was the increase in intra-ring commuting-a sign ofemploymenit decentralization that will probably lead to a greater cle-mand for circumfer-ential trips. (A similar table for students' travel pat-ternis in 1972 reveals that the majority of stuidenits traveled to schoolwvithin close proximity of their residences-within the same ring as thelocation of their residence.) Another way of looking at the pattern ofwork travel is to observe the "relative employment absorption" by riigof employment (see table 8-12). The index measures the number of jobsin the ring filled by workers who commtited from other rings. The r atio,as might be expected. declines as one moves outward from the (:1D. Thetable clearly points to increasinig emplovment and residential decenitial-izationi. The outer rings exhibited a high growth rate in external trips.

Tllis conhtinuinig decentralization of employment and residential pat-terns in Bogotai may have a serious and adverse effect on the quality ofBogota's public transportation. Mass transit serves radial, concenitratedtrips more efficiently than it serves less concentrated, circumferentialtravel. The evident increase in llse of intermediate-capacity vehiclesmav be a direct response to decenitralizationi. As mentioned before,iltermediate-capacity vehicles can operate profitablv with lower loadfactors thani full-size buses. Break-even load factors for- busetas were 45 to50 percent at existing fare levels, compared with 75 to 80 percent forbuses. As comniLting patterns become less concentrated along specificradial corridors, intermediate-capacity vehicles will begin to supplanltlarger vehicles.

7The Effect of Income and Other Household Characteristics on 7ravel 7ime

We found(i convinlcing evidence that workers from poor householdsspenit more tinme traveling and traveled loniger distances than workelsfrom better-off households. In both Cali and Bogota, the r icher r esiden-tial areas have a higher ratio ofjobs per resident than do the poorer res-idential areas. Hence, it seemed that the pOor had relatively worseaccess to work opportunlities. In 1978, workers living in the richest seC-tor of Bogota traveled an average of only 3.7 kilometers per trip-mtichless than the 6.2 kilometers averaged by the population as a wvhole. Aver-afge speeds of commuters from the poorest sector were more thai 25percent slower thiani the speeds of' commtiters from the richest sector.

Table 8-13. Average Commnuting Distance by Zone of Work, Type of Worker, and Occupation: Bogoti, 1972 and 1978(kilometers)

All zones ( entral zones Other zonesPercentage Percentage Perrentage

1972 1978 increase 1972 1978 increase 1972 1978 increase

Houtsehold headsProfessioiials 3.5 4.1 15.7 4.2 4.9 17.6 3.2 3.7 16.8Emplovees 4.2 4.4 5.3 4.4 5.1 16.0 4.0 4.1 2.5Retail employees 2.0 3.4 15.5 4.2 5.1 23.4 2.4 3.1 25.4Others 3.4 4.0 16.4 4.2 4.6 8.5 3.2 3.9 23.4

Secondary workersProfessionals 3.5 4.0 12.9 3.9 5.7 47.8 3.3 3.7 12.1Ernplovees 3.9 4.4 12.9 4.2 4.7 13.9 3.7 4.2 14.0Retail employees 2.9 3.1 8.4 4.3 4.6 5.5 2.4 2.9 19.0Others 3.3 4.1 23.6 4.5 4.9 10.7 3.0 4.0 32.5

Source: Pineda (1981).

AUTOS, TAX[S, BUSES. AND BLUSETAS 209

However, if the choice of transportation mode is controlled for, the dif-ference in speeds is small but the distance and time differences remain.

The evidence is that the level of transit service is roughly equalthroughout the city. If the poor are transportation-disadvarntaged, it isbecause of the location of activities and riot because of the level of trans-portation services available. Indeed, the relatively high level of serviceavailability throughout ihe city and the prevalence of the flat-fare systempartially mitigate the locationial disadvanitage of poor workers. Funda-mentally, the transport system per se does not contribute to any loca-tional disadvantage suffered by poorer workers in Bogota and Cali.Instead, locational problems are caused bv the imbalance between jobsand workers in rich and poor neighborhoocls. However, the longer timethat the poot spend traveling does impose higher effective costs onthem .

Relocatingiemploymnent and residencewould dloagreat deal to alleviatethe locational disadvantage oftpoorer workers. but this is typically difficultto accomplish. It has also been founid that the net effect of the taxes andsubsidies affecting tranispor-tation in Bogota has helped reduce the disad-vantage of the poor. More attention could be given to infrastructureinvestment by fturther imiproving the road system in a wa' thiat wouldreduce commuting time for the poor, who reside in disadvantaged loca-tions. Furtlher expansion of public transportationi services and somerealignment of routes, particularlv circumferenitial routes from the poorsoutherni areas of the city to the wealthy north, wotild also help.

So far we have focused on one determininlg characteristic of travel pat-ternls-incoiiie. For a richel appreciation of'travel patterns, it is necessar-yto observe the effect of other variables stich as hiousehold size and com-position, occupationis oF individcual workers, employment clensities, andhousinig and ownership staits. For example, households with mor-e thanone worker- have to assess the competing travel neecds ofditferent house-hold memiiber-s in choosing the best residence location. In 1978 as mallyas half of all houselholdls had more than onie worker-this hacd increasedfi-onm 40 percent in 1972.

This issue was examined in great detail by Pineda (1981) as part of theCity Study. His findings suggest that there is a great degree of interdepen-dence in the commutinig decisions of multiwor-ker households. House-holds do tend to optimize the joint satisfaction of different hotiselboldmembers' travel needs. Primary workers have longer work joulr-neys thansecondary workers. This suggests that r esidence location is chosen to max-imize the secondary workers' access to likely employment prospects; ingeneral, the location of the primar-y worker's job is likely to be known inadvance ancI would be more fixed. Primarv workers were also morelikely to be professionals and white-collar employees whose jobs weremore likely to be farther away in the central city (see table 8-13). Homiie-

210 UNDERSTANDING THE DEVELOPING METROPOLIS

owners travel I kilometer more on average than renters; this is consis-tent with patterns in cities in developed countries. It also conforms toexpectations, because renters are likely to be more mobile and there-fore to have greater flexibility in their choice of residential location.

In attempting to understand the travel pattern of workers according totype of work done, Pineda found a clear distinction between what hecchar-acterized as type I and type 11 secondary workers. Type I secondary work-ers had more than a primary educationl, were neither family workers norself-employed, and were at least twenty-one years old. Their characteristicsresembled those of household heads or primary workers. Type I workers'commutes were longer than those of type 1I workers and not very differentfrom those of primary workers. Type I workers had increased in propor-tion from about one-third of all secondary workers in 1972 to about halfin 1978. There seems to be a process of "formnalization" of the kind ofjobs engaged in by secondary workers.

After controlling for other variables such as housing status, mode oftravel, and quality of dwelling unit, Pineda also estimated a number ofmultivariate regressions in order to estimate better the effect of multi-worker households on distance traveled by primary workers. The keyresult was that the skill mix of workers in a household was important indetermining residence location. The more homogeneous the characteris-tics of differentworkers in the housellold, the more likely they were to liveclose to areas of the citv where all could find work. In households withtype 11 secondary workers, the household hiead commuted farther.

The conclusion of this investigation is that a city with increasing laborforce participation has an incr-easinig niimber of secondary workers whosetravel requirements are more variegated than those of primary workers.Hence the citv then needs a more flexible transport systemn.

Travel Schedules

Another useftul way to look at urban travel patterns is to observe the fre-quency of travel and the time profile of travel over the day. Frequency isusually conmputed on a daily basis, on the basis of trips per person or perhousehold. Analyzing the time profile of travel helps in assessing the ade-quacy of existing utansportation services to meet peak hour demand;travel rates vary widely over the day. Typically, about 75 percent of all tripsare made duringjust 25 to 30 percent of the day. It is also useful to under-standwhy trips take place: household size and composition and individuallifestyle differences all contribuite to differences in type of transportationiused, as well as time, origin, and destinationi of trips. Income and access toa private vehicle are other key factors.

Characteristics suchi as income, household size, and automobile owner-ship affect trip rates in cities in both developing and developed couii-

AUlTOS, TAXIS, BTSES, ANI) BLISETAS 211

Table 8-14. Average Number of Trips per Person and Average Numberof Trips per Household: Selected Cities

.4 verage

A verage Tnps per 7ripsper nuimber ofrarsil/v Year househol jize person homseholb per hou.sehold

Lagos 1975 5.4 0.79 4.3 0.18Bangkok 1972 6.5 1.11 7.2 0.28Bogota 1972 5.7 1.15 6.6 i 1.17Hong Kong 1973 4.7 1.27 5.9 0.14Singapore 1968 5.2 1.40 7.2 0.21Buenos Aires 1972 3.8 1.42 5.4 0.30Sao Pauilo 1977 4.8 1.50 7.2Kuala lumpur 1973 5.8 1.74 10.2 0.42Mexico City 1972 3.5 1.77 6.4 0.45L.os Angeles 1967 2.9 2.28 6.7 1.28Chicago 1971 2.9 2.45 7.2 1.04Denver 1970 3.1 2.83 8.8 1.40

- Not available.Note: Motorized trips only.Source: BogotA-1972 Pliase 11 Survey tabulations; I.agoS-TR5NSP0(:ONst[LL (1976); Bue-

nos Aijes-Ministerio de Obras y Servicios Publicos (1972); Sao PaUI-FEMPLASA (1978);Hong Konig-Wilbur Smith and Associates (1976); Bangkok-cited in Zahavi (1979);Koiala 1Lmpur-Wilbur SmiTh and Associates (1974); Singapore-Zahavi (1979); MexicoCitv-Barton-Aschnian Associates (1983); Los Angeles, Denver, and Chicago-Lecvinoson(1978).

tries. (See table 8-14 for a sample of data for motorized trips fiom the1970s. Although these data are now somewhat dated, they are still quiteinstructive. The data exclude walking trips becattse comparable informa-tion for different cities is difficult to obtain.) Trip rates are progressivelyhigher in higher-income cotintries. There is considerable variation inhousehold size, the average household size being small in richer coUII-

tries. Automobile ownership seems to lead to higher trip r ates.Trip patterns in Bogota were examined by level of household income,

household size, auto ownership, and indixidual characteristics-workers,students, noniworkers. Within BogotA, both trip rates per traveler andnumber of travelers per household increase as household income rises.A striking fact is that, once household income is controlled for, autoownership has a minimal impact on trips per traveler but a significantimpact on travel participation rates-that is, nuniber of travelers perhousehold. The net effect is that household trip rates are higher forautomobile owners than for nonowners at similar incomne levels. Theimpact of automobile ownership is greatest on workers: presumablythose who travel by private automobile to work are also likely to makemany more side trips. Controlling for income, automobile ownershiphas little impact on travel frequenicy of other household memnbers suchas students and noniworkers. The fact that members of higher-income

Table 8-15. Trip Generation Characteristics of All Travelers: Bogota, 1972

AvJerage t ips per traveler per day Av erage travelers per houtsehold per day A verage ntips per household per dayMlonthly hoau.mhold income(("ol$) .Vo mar Car 7btal NIo car Car 7Ttal No cad Car 7otal

0-50() 2.2 - 2.3 1.7 - 1.7 3.9 - 3.9500-1,000 2.4 - 2.4 2.0 - 2.0 4.8 - 4.91,000-1,500 2.4 - 2.4 2.3 - 2.3 5.8 - 5.81,500-2,000 2.6 2.7 2.6 2.6 2.6 2.4 6.7 6.4 6.72,000-3,000 2.8 2.9 2.8 2.8 3.6 2.9 7.9 10.2 8.23,000-5,000 2.8 2.8 2.8 3.0 3.1 3.0 8.1 8.8 8.35,000-15,000 2.8 2.8 2.8 3.( 3.4 3.2 8.5 9.7 9.2More than 15,000 - 2.9 2.9 - 3.5 3.5 - 10.4 10.4

Weighted average 2.6 2.8 2.7 2.5 3.2 2.7 6.7 9.4 7.2

- Not applicable.Source: Westin (1980); data from 1972 Phase 11 Stirvey.

AUTOS, TAXIS, BUSES, AND Bt!SETAS 213

households stay in school longer is brought out very sharply in students'trip rates per household as household income increases.

It is interesting to compare U.S. urban travel characteristics withthose in Bogota (see tables 8-14 and 8-15). At the 1972 exchange rate ofCol$25 to US$1, a monthly household income of Col$5,000 in Bogotawas equivalent to annual household income of US$2,400. This corre-sponded to the lower end of the income scale in the United States.Automobile ownership in Bogota was only 0.17 per household, com-pared with 1.31 in the United States; the average household size was 4.6in Bogota, compared with 3.06 for the United States. Travel frequencoincreased with income in both places; at straight exchange rate equiva-lents, household trip rates were higher in Bogota. The data indicate thatautomobile ownership has a quantitatively larger effect on trip rates inthe United States than in Bogota for a given level of income. This illus-trates the benefits gained from the operationi of a relatively good andeasily accessible transport system such as Bogota's. In the United States,non-automobile-owning households are put at a disadvantage by theexistence of a highly dispersed pattern of residence and employment.Moreover, high levels of automobile ownership ptit those without cars ata further disadvantage, because it is increasingly difficult to operate a

Table 8-16. Trip Generation Equations: Bogota, 1972

Wlorketrs Non workers

1`anablh Total tnrps Work tnps Nonwork ntrps

Mean trip rate 2.74 2.51 2.54

Sex (female = I) -0.12* -0.03 -0.11Age (years) 0.04** 0.04* 0.08*

HI ours worked - 0.004 -

Number of workers in household - - -0.11 *

Transportation accessibilityWork tnps

Automobile owniers 0.57 0.19*Non-owners -0.29* 0.28* -

NVonwork trips

ALutomobile owners -0.22 - 0.21Non-owners 0.07 - 0.40

lntercept 2.83 2.35 2.64

W2 0.057 0.055 0.051

-Variable not included in the equation.* Denotes significance at the 99 percent confidence level.** Denotes significance at the 95 percent confidence level.Note: Household income niot included as a variableSource: Kozel (1981).

214 UNDERSTANDING THE DEVELOPING MEIROPOLIS

public transport system economically as fewer people come to dependon it.

Urban transport planners in developing countries must understandthis issue well. The continued provision of good public transportationcould help slow increases in automobile ownership as incomes rise. Thevalidity of this conclusion was investigated by estimating trip generationmodels for workers and nonworkers. (See table 8-16. For details, seeKozel 1981.) The most important variable was a measure of transporta-tion accessibility. This index was calculated as a measure of transport sys-tem attractiveness as defined by times and costs to desired locations andthe availability of transportation modes. 3 The model follows the originalspecification by Lerman and Ben-Akiva. The estimates shown excludehousehold income as a determining variable because it was foundthat income was highly correlated sith the index of transportationaccessibility.

The equations provide evidence of a strong causal link between fre-quencv of travel and quality of transport system. The measure of accessto non-work-related destinations was highlv significant for nonworkingadulLs. Households without automobiles were more affected by improve-ments in the transport system. Although our overall conclusion in thisstudy has been that the transport system serves the poor relatively well,the high correlation between the measure of transportation accessibilityand household income suggests that the poorer areas of the city are lesswell served by the transport system.

In cities in developed countries, peaking takes place during rush hourtraffic in the journey from home to work in the morning and the returnhome in the evening. In many developing countries, particularly inLatin American cities, there is considerable travel from work to homefor lunch, creating another peak travel period. The proportion of work-ers going hiome for lunch in Bogota seems to have increased from about25 percent in 1972 to 31 percent in 1978. In Cali this proportion wasabout 50 percent. As may be expected, automobile owners are muchmore likely to lunch at home-about 70 percent of automobile ownersdo so in Cali and about 40 percent in Bogota. Among workers withoutcars, just under 50 percent in Cali and 28 percent in Bogota lunched athome. This additional trip home contributes to traffic congestion in themiddle of the day, hampering the movement of both business-relatedgoods and passengers. An interestinig fact is that the number of peoplelunching at home seems to decline as city size increases-presumablybecause of longer average commutes. Thus, the increase in lunch travelbetween 1972 and 1978 in Bogoti is hard to understand.

A close examination of the peaking pattern by transportation modeshowed that bus travel peaked earlier in the morniing than buseta and

AUTOS, TAXIS, BUSES, AND BtlSETAS 215

automobile travel. The morning peak was the highest because trips toschool and work coincide. The evening peak was more spread out,because school days end earlier than most jobs. The travel distributionsfor busetas over the course of the day closely resembled automobiletravel distributions-again suggesting that buseta transportation is aclose substitute for private transportation.

Modeling Travel Demand

The design of urban transport systems and infrastructure requires a sys-tematic understanding of how and why people choose among variouspublic and private transportation modes. Understanding their choicesand being able to encourage them to use mass transit are particularlyimportant because collective public transportation modes usually makemore efficient use of road space and fuel and other resources. It is espe-ciallv desirable to promote such collective modes in resource-poor devel-oping countries. Private modes, however, provide faster, more comfortabletravel, so there is a natural tendency for people to switch to these modeswhenever possible. Can behavioral modeling help in understanding thedeterminants of travel behaviors? What characteristics of urban trans-port supply do travelers value most? What really irks people in theirdaily urban travel patterns? What, if anything, can be done to improveurban public transportation services and thus slow the rapid shift to pri-vate motorized modes of transportation as incomes increase? The mod-eling of urban transport demand would be useful if answers to suchquestions become available.

Every traveler is faced with limited transportation choices, which are,interestingly enough, often greater in cities in developing countriesthan in developed countries because of the emergence of a number ofintermediate modes. Our descriptive work has already shown that modelchoices are strongly influenced by income. Automobile ownership isparticularly influienced by income level and mode choice is thenaffected by whether or not the household owns an automobile. Trip dis-tance also affects mode choice: very short trip distances, for example,are likely to be traversed by walking. City structure also influenices modechoice: if a city has relatively concentrated central city employment,those working in the city center are more likely to use public transporta-tion than those employed in the periphery. Modeling travelers' behav-ior thus involves the modeling of discrete choice. In chapter 6 wereferred to the modeling of location, also a discrete choice problem forthe firm. In that formulation, the employer's problem was to choose thelocation that maximized profit subject to various constraints. This was

216 UNDERSTANDING THE DEVELOPING METROPOLIS

transformed into an empirical framework for predicting the probabilitythat a firm with certain characteristics would occupy a site with certainattributes. Similarly, in modeling urban transport demand, the individu-al's problem is to chose the most useful travel mode that maximizes util-ity subject to various constraints specific to the individual and othersspecific to the chosen mode. Using this conceptual framework, one canestiimate the probability that an individual with certain characteristicswill choose a specific travel mode. Among individuals regarded ashomogeneous in tastes, except for measured characteristics, the varia-tion in travel choice results from variation in unmeasured characteris-tics and tastes: the utility function is therefore probabilistic in form. Thetheoretical derivation and various applications of urban travel demandmodeling are found in a number of sources and will therefore not berepeated here. 4 A skeletal introduction to the essential theoretical basisof such modeling is given in the appendix to this chapter, however. Theestimation method is the multinomial logit technique that was used foremployment location.

This approach, focusing on the behavioral modeling of the individ-ual, may be contrasted with the traditional social physics approach ofurban transport modeling. The latter attempts to replicate the observedtravel patterns in a statistical fashion by modeling the volumes of trafficflow, for example, based on the available information of residential andworkplace densities, the available travel routes, and the like. The goal insuch modeling is essentially to observe statistical regularities and to rep-licate such regularities through modeling. These models do not provideinsights into behavior and are very data-intensive. Our approach is moreeconomical in the use of data, although the estimation techniques andmodeling theory are quite involved.

In this approach, it is necessary to carefully identify the choices facingeach individual. Not all individuals face the same choices. For example,individuals in households without cars do not have the option of travel-ing by private car.

In Bogota the available choices were defined as follows (see table 8-17):

Transport mode A vailable to:Private car All members of automobile-owning

householdsTaxi and colectivo All persons living on colectivo routes

and in the inner three rings of BogotaBus All travelersBuseta In 1972, all travelers living within

I kilometer of a buseta route; in 1978,all travelers

Walking All travelers making a trip of less than10 kilometers

AUTOS, TAXIS, BL'SES, AND BUSETAS 217

Table 8-17. Work Trips by Mode of Transportation: Bogoti, 1972 and1978, and Cali, 1978(percentage shares)

Mode Bogotd, 1972 Bogoid. 1978 Cali, 1978

Automobile driver 9.0 9.3 6.2Automobile passenger 2.9 2.7 2.0Taxi 1.3 1.2 2.7Ciolertivo 0.8 0.3 0.1Bus 69.0 51.8 62.1Buseiaa 8.7 17.7 2.5Walk 7.3 12.9 16.3Otherb - 4.0 8.1

- Not available.a. Inciudes inicrobuses.b. Includes bicycle, motorcycle, and vehicle owned by emplover.Source: Kozel (1981); 1972 Bogota tabulations from Phase ll Survey; 1978 BogoUi andl

Cali figures from 1978 World Bank-DANE Household Surves.

The key determinants of modal choice are household income, compara-tive costs in terms of both money and time, motorization rates, andurban structure. Other socioeconomic variables such as household size,composition, and the traveler's position in the household (for example,household head) can also affect the mode of travel chosen.

In modeling the mode choice of travelers in such a discrete choiceframework, where the probability of choosing a mode is determined bysuch explanatory variables, a number of useful by-produicts emerge forpolicy analysis. The valuation of time by travelers can be estimated. Thevaluation of time can vary for different modes, and for different levels ofincome. This is usually done as a proportion of existing wage rates. Istravel time valued by workers at the same rate as wages or at a lower orhigher rate? Such travel time valuation is needed for conducting projectevaluations of different transportation projects. The estimations alsoyield direct estimates of elasticities of mode choice (derived in theappendix to this chapter). The evaluation of the elasticities can be usedto predict the effect and mode choice resulting from time reductionsthat may emerge from traffic management improvements or transporta-tion infrastructure investments. Similarly, the effects of increased faresor rising incomes can be evaluated.

In this study, a large number of different specifications were used tomodel mode choice behavior in both Bogota and Cali (see Kozel 1981and 1986 for details). In the case of Bogota, data were available for bothwork trips and nonwork trips for 1972. For 1978, data were available forwork trips only but on a comparable basis for both Bogota and Cali. Thespecifications had to be different in the different years because the 1978survey had only travel-to-work information and the data available in thetwo data sets were not on identical variables.

218 UNDERSTANDING THE DEVELOPING METROPOLIS

Our interest in modeling is to understand the influence of these vari-ables oil the probability of choice. If the model is well specified and theestimation is found to capture the revealed choices well, it can be usedto predict future travel demand in response to expected changes. Thereare several ways to judge the performance of such models. First, as inordinary least-squares regression, the significance of the coefficient ofeach variable is an indicator of the importance of each variable in themode choice decision. Second, statistical measures like log likelihoodratio and likelihood ratio index may be used to measure goodness of fit.Third, the percentage correctly predicted gives a good idea of the mod-el's ability to capture the revealed preferences of the population. Oncea model is calibrated o01 a base year, it can be used to predict the out-comes of projected changes in a future year.

Because of the lack of variation in travel cost caused by the existenceof flat bus and buseta fares, the cost coefficients were not very robust.

Otherwise, the estimated coefficients were significant and of appropri-ate size and magnitude. I'he models estimated performed about as wellas might be expected from other similar exercises: in general, the mod-els were able to achieve 70 percent predictive capability. Had there beenmore variation in travel cost, there would have been even better predic-tive performance. Moreover, wherl there are more than two choices, asin Bogota, the modeliing problem acquires greater complexity.

It hacl been hiypothesized that time spent in different transportationmodes could be valued clifferently. The time spent in bus travel, for ex-ample, would be seen as more undesirable than time spent in a privatecar. It was found that the generic time coefficient is not significantly dif-ferent from the alternative specific coefficients, and with the bus beingthe most dominant mode, the bus-specific time coefficient was the near-est to the generic coefficient. Household characteristics were also foundto be important: travelers from better-off households placed a highervalue on their travel time; travelers from larger families appeared to beless willing to pay to save time, reflecting lower hiousehold inconle percapita. It was also interesting to note that the time spent walking wasfounld to be the most onerous. It was also founld that the travel accessi-bility variable-distance to the nearest bus or buseta route-was signifi-cant for automobile owners. Proximity to a public transportation routereduced the probability of automobile use for automobile-owninghouseholds. This reduction was particularly pronounced regarding buse-la accessibility, a fact suggesting that a buseta was seen as a possible alter-native by autoinobile-owninig households.

Althiough it is not very useftl to provide the detailed estimates ofequa-tions here, it is qtiite instructive to give an indication of the elasticitiesobserved. It is interesting to compute the actual change in modal shareresulting from a I percent change in the explanatory variable: this is the

AUTOS. TAXIS, BUSES, AND BUSETAS 219

percentage change in the share (or the elasticity) multiplied by theshare itself. Because the elasticity estimates are nonlinear, they have tobe evaluated at a given point; in this case the effects are evaluated at theobserved modal shares. The results indicate that, whereas automobileuse was quite sensitive to cost, walking, bus, and buseta use were rela-

tively more sensitive to time variations. In Bogota, for example, in 1978,a I percent increase in time spent walking would have reduced the walk-ing share from 12.9 percent to less than 12.5 percent; and a similarincrease in buseta time would have reduced its share from 17.7 percentto 17.4 percenit. In contrast, a I percent increase in automobile timewould have had almost no effect on the 9.3 percent share of automobileuse (see tables 8-17 and 8-18). Cross-elasticity effects can be calculatedin a similar fashion. As may be expected, increases in bus times increasethe use of busetas, but increases in automobile journeys have negligibleeffects on other modes.

How do the models perform intertemporally?5 Because we had datasets for 1972 and 1978, it was possible to test the predictive accuracy ofthe 1972-based parameters using 1978 data by predicting modal sharesin 1978. Parameters were also estimated for 1978 and compared with

Table 8-18. Change in Work Trip Modal Shares Direct Elasticities:Bogoti, 1972 and 1978 and Cali, 1978

Mode 7imw Cost Household incone

AutomobileBogota, 1972 - -.0144Bogota, 1978 -.0026 -.0502 -Cali, 1978 -.0144 -.0807 .0070

Taxi, colectivoBogota, 1972 - -. 0030 -Bogota, 1978 - -.0192 .0176Cali, 1978 -.1108 -.0775 -

BJ3setaBogota, 1972 - -.0029 -

Bogota, 1978 -.2936 -.0286 .1704Cali, 1978 -.0851 -.0175 .0070

BusBogota, 1972 - -.0032 -

Bogota. 1978 -.3309 -.0211Cali, 1978 -.1440 -.0196

WalIBogota, 1972 - -

BogotA, 1978 -.4350Cali, 1978 -. 3194

- Not available.Scnsrre: Kozel (1982).

220 tUNDERSTANDING THE DEVELOPING; METROPO015

the 1972 parameters. It is important to note the main exogenoiuschanges that occturred betweeni 1972 and 1978. Automobile ownershipincreased from 167 to 217 per 1,000 households; household incomeincreased bv about 33 percent in r eal terms; proportion of total emplov-ment in the central city fell from about 40 percent to 33 percent; and alltravel tines increased with considerable city expansioni (particularly bustrips, which increased fi-om 40 miniutes to 55 minutes on average, and of'buseta trips, which increased fi-om 34 minutes to 48 minutes); finally,supply of huselas increased tremendously-in 1978 almost all areas wereserved by both buses and busetas.

A number of' tests were performed to analyze the transferability ofthe estimated models. The 1972 BogotA estimates were sirnulated toobserve how well the 1978 modal split would be predicted. Similarly, the1978 estimates were used to simulate the 1972 split, and the Bogota andCali estimates for 1978 were interchianiged. Overall, the model estima-tion was similar for work trips over the Iwo years (see table 8A-5 in theappendix to this chapter). There were some difficulties, however, insuch comparisons. As mentionied earlier, because of differences in dataavailability, the specifications of models estimated for the two years weredifferent. Hence, the compar-isoni had to be done on the basis of a com-moni set of variables. Another general difficulty encoutntered was in tilecost variable. Because of the prevalence of flat-fart svstems in Bogota,there is no variation in costs between joulnlevs of clif'ferent clistances.Thus, the coefticients of'cost variables were generallyt unsatisfactory, anda significanit diff'erence was fotiud betweeni the parameter estimates forthe two years. The stability of coefficients across the two models wastested statistically thioughi likelihood ratio tests.

The most interestinig test of a mio(del for planning purposes is in itspredlictive ability. The 1978 predicted modal shares from the 1972 mod-cl give good results-the maini error being in the share of' buseta rider-ship. This is not surprisinig, because the supply of buselas had beenlimited in 1972. The 1972 coefficients therefore modeled mode choiceutnder such supply constrainits. Different simulations were conducted tosuccessively substitute 1978 parameters step-by-step for the 1972 param-eters. The predictive success of the model improved with each succes-sive step. In particular, b,useta ridership is approximated well after thesesubstitutiois. Automobile use falls and walking increases, however. An-other experimeint was done to predict mode choice bv income class.The predictions for- the lower income grotips perform mutch better tihanfor the highier income groups. Once again, this suggests the greater dif-ficultv in mnodeling situations where choice is greater than two modes.

One problem found thioughout this modeling effort and one that isdifficult to explain is that the valuationl of time spent in traveling isfounid to be a multiple (of'3 or 4) of hourly wages. This is quite counter-intuitive and very- differenit frotml modelinig efforts in dleveloped coun-

AUTOS, TAXIS, BLUSES, AND BUSETAS 221

tries. Travel time is usually valued at half or three-quarters the rate ofhourly wages. There could be two explanations. First is the lack of costdata for bus and buseta trips. Second is that the time cost becomesimportant mainly for the richer, auto-owning travelers. It is possible thatthe average reduction of time being obtained really reflects the valua-tion assigned by high-income travelers.

Similar modeling methodology was used to estimate the probabilityof a household owning an automobile. Overall, the models performedwell in predicting automobile ownership levels in 1972 and 1978. Asmight be expected, income was the key determinant of automobile own-ership. Homeownership, used as a proxy for wealth, was also significant.In fact, an alternative specification approximating permanent incomewas found to be a better explanation. Estimated demand elasticitiesranged between 1.0 and 1.3-consistent with cross-section elasticitiesestimated elsewhere. It was also found that, at similar levels of income,households living farther away from the center were more likely to ownan automobile. What was most interesting was that a variable attemptingto measure accessibility to transportation performed well: better trans-portation availability reduced the chances of automobile ownership.This is a most important result: the general inexorable trend towardgreater private transportation ownership and use that accompaniesincreased incomes can at least be slowed writh better provision of publictransportation. The projection for the year 2000 suggests that automo-bile ownership levels at that time will be less than one-third of currentU.S. levels.

What have we learned from this relatively sophisticated modelingeffort? The disaggregated travel demand modeling methodology con-trasts with the macro or social physics approach used earlier in model-ing urban transportation patterns. This approach is clearly much moreeconomical in data requirements and even estimation; even relativelysmall data sets can be used. However, it must be admitted that this meth-odology demands data of high quality, if not heavy volume. Moreover,because it is not easy to gain an intuitive understanding of the theoryand estimation technique, it can be difficult to interpret the estimateswell. Nonetheless, the models performed quite well and even had rea-sonably good predictive capability. The estimation of such models alsofosters an understanding of the behavioral rationality (or irrationality)of travelers. Overall, the verdict is mixed: better understanding has beenachieved, but with considerable effort.

Lessons for a Developing Metropolis

The structure of a city is deeply affected by, and at the same time itselfdetermines, the transport system at any given time. The spatial distribu-

222 UNDERSTANDING THE DEVELOPIN(; METROPOLIS

tion of activities in a city depends on the qualitv of the transport system,its spread and coverage, and the costs of transportation. In turn, the pat-tern of ur-ban transportation is determined by the existing structure ofactivities in the city: an efficient svstem responds to effective demandsfor travel as they become evident. In a city in a developing country, as inBogota', incoome growth causes changes in city structure throughchanges in housinig patterns, residence location, income distribution,and shifts in occupations and activities; these changes then have animpact on travel demand. In addition, income changes affect traveldemand more directly thirough the impact on automobile ownershipand the valuation of time.

We founld that Bogota had a rich supply of transport services. People ofalmost all income levels had easy access to motorized transportation. Busservice was available to most residents within about 500 meters of theirhome. About 125 seats were available for every 1,000 residents. This com-pares well with richer cities, such as Hong Kong, which are generallyregarded as being well served by public transportationi. The cost of travelwas low because of the tax and subsidy systenm adopted by the government.

Unlike most cities in developed countries, and many large cities indeveloping countries, where the government typically operates publictransportation services and usually subsidizes them, in Bogota the govern-mnent had a negligible presence in actual sulpply. Almost all public trans-portation was privately owned and operated. The government (lid,howevel-, regulate the system physically through route allocation andcontrol, and regulate the fares through tariff setting and subsidies.Overall, the system succeeded in being progressive: the bottom 40 per-cent of the population received between 4 and 6 percent of theirmonthly incomes as transportation subsidies. Although the tax and sub-sidv system was by no means ideal and could be improved, the combina-tion of fuel subsidy, vehicle taxation, and bus subsidies helped makeurban public transportationi accessible and affordable to almost all resi-dents of Bogota. The situation was similar in Cali.

One result of this high level of transportation availability was thatBogota households exhibit a high level of mobility. The number of tripsper household was seven per day, similar to the trip rates recorded forsome U.S. cities. The trip rate per person, however-1.4 in Bogota com-pared with two in U.S. cities-was much lower because Bogota house-holds are larger. One interestinig feature of trip generation in Bogotiwas that autolniobile owner-shlip was not founld to increase the mIlliber ofjourLIneys per household. This was probably because of the high level oftransportation availability: the absence of a private mode did not ham-per mobility. This suggests one of the more important findinigs of thissttid(. In developing countries, it may be possible to slow down theincreasing use of private transportation modes, particularly autornio-

AlUTOS, TAXIS, Bl'SES, AND BUSETAS 223

biles, if credible alternative modes are provided. In Bogota, busetasappeared to be substitutes for automobiles.

What are the characteristics of private automobile travel? Auto-mobiles provide great convenience: door-to-door service, no waitingtime, twenty-four-hour availability, and the possibility of multipurposetrips. Thie drawback, of coulse, is thie high cost. Automobile use wasfouLnd to be quite sensitive to costs, so it is clear that automobile travelcan be reduced, or at least its growth curbed, if public transport modescan be made sufficiently conveniienit and economical. The success of thepublic tranisport system in BogotA and in a number of'other cities in var-iotis middle-income countries has demonstratecl that convenience andeconomy are possible to achieve in developing country systems.

In developed countries, urbani public transportation is dif'ficult to runwithout subsiclies. The driver is the most expensive element in the struc-ture of'costs of bus operation: typically, more than 80 percent of'fare boxrevenues in U.S. cities go to the bus driver. Econonmical operationi there-fore implies large buses serviing heavily traveled corridors. Other loca-tions are then dif'ficult to serve economnically by means of'large vehicles,thus promiioting private transportationi, which itself reduces public tran-sit ridership.

In developing counitries, on the other hand, where wages are low, thedriver does not cost much at all: in Bogota, the typical driver receivedonly 10 to 15 percent of fare box reveniues. It is therefbre possible tooperate smaller buses, with jtust 10- to 25-passenger capacity. These canlapproximate the convenience of private transportation. They can servethe less heavily traveled routes at reasonable frequencies and( at lowercosts than automobiles.

As a city expands and incomes grow, the city decentralizes. In cities indeveloping countries, although decentralization of jobs commonlyoccurs, there is seldom a decrease in the absolute number ofjobs. Hencethe demand for radial travel does not fall in absolute terms. The decen-tralization of jobs and residence implies a greater demand for circum-ferential travel and provides new opportunities for distributing loadsmore evenly in the road system bvy means of low-capacity vehicles likehusetas or even smaller vans. A r acdial and circuLmlfer-ential transit systemcan coexist: this seemed to be the case in Bogoti as the city expanded insize and the transport system responded to the new demands.

How was the systetn able to respond so well? The private provision ofpublic transpor-tation through the participationi of small bus ownersoperating as cooperatives lent great flexibility to the system. As the citygrew, new neighbor-lhoods sprang Up, and employmenit location patternschanged; it was easy for small indepenclent bus enterprises to respond tonew demands. The governmenit authorities, on their part, r esponded byissuing permits for new route allocations. Sttch flexible r esponses in the

224 UNDERSTANDING THE DEVELOPtNG; METROPOIAS

high-growth situation of Bogota in the 1950s, 1960s, and 1970s, wouldprobably not have been possible if there had been a monolithic large pub-lic transport company.

The government in Bogota succeeded in inducing private participa-tion by making good use of private initiatives. Overall costs per buskilometer were low; subsidies amounted to about half of costs; and as aresult fares were low, which improved access of the poor to transporta-tion. The presence of many private operators introduced competitivebehavior, which helped keep costs low. At the same time, the govern-ment was saved from making heavy capital investments in bus purchases.The introduction of busetas, which were not subsidized, meant that thebus subsidies helped the poor because the higher-cost, more convenientbusetas attracted the better-off riders away fiom the subsidized buses.Lesson: the existence of a full-cost, upmarket service can help concen-trate subsidies where they are most needed. Although the results mavhave been achieved serendipitously, the flexibility inherent in Bogota'stransport system provides good pointers for effective transport policy incities in developing countries. The introduction of even greater hetero-geneity through smaller vans with superior levels of service would haveadded to this flexibility.

One implication of the patterns observed in Bogota is that when auto-mobile ownership rates are low, as they were in Bogota, it is a good time toestablish restraints on automobile use. If upmarket residents have a rea-sonable alternlative to driving, controls on automobile use (such as exclu-sive bus lanes, high fuel prices, parking charges, and restrictions onprivate cars in congested areas) could do much to slow down the growth ofautomobile use. None of these measurles has been introduced in Bogota.Such measures are desirable because private transportation uses roadspace and energy inefficiently, increasing congestion. Reducing automo-bile ownership would redLIce the need for premature heavy investmentsin road and other infrastructure.

As the city grows and both residences and jobs decentralize, thedemand for radial trips decreases while that for circumferential tripsincreases. In Bogota, for example, there is a good deal ofcross-commutingfrom poor sector 2 in the south to rich sector 8 in the north. The availabil-itv of circumferential routes would reduce both transportation time andcentral-city congestion.Another lesson from Bogota was that circumferen-tial roads can eliminate unnecessary radial trips through the center, reduc-ing both congestion on the radial routes and the time required to travelthe ci rcumferential routes.

What did we learn from our behavioral modeling efforts? The behav-ior of Bogota residenits was broadly similar to that observed elsewhere.Choice of transportation modes is more sensitive to income and servicelevels than it is to relative prices of competing modes. At low income

AllTOS, TAXIS, Bl SES. ANI) BUSETAS 225

levels, there is effectively no choice: people either walk o(r use publictranisit. Betweeni the fiftieth and eightieth percentiles of' income distri-bution, people xwill switch to better service levels if they are available. Atthe hightest inconie levels people switch to cars but can be restrained ifviable alternatives exist. Wher-eas autoniobile ownershiip rates wouldcontintue to intcreise, it may be possible to restrict aUt(mobile USe 6V

simtiltanieously iniproving puiblii tranisportationi service levels anidincreasinig costs of aLltom(nbile use.

The meaini lesson for ur-bani transport provision in developing coulitriesis that, if the systemn is madie flexible enough to respond to chaniges in

clemannd, it is possible to cope with rapid growti.

Appendix: Theoretical Background for Modelingthe Choice of Transport Mode

For an indivifual, each alternative j = 1.I has a vector of' observedattributes Xi. Each individual I is also sLibject to socioecononlic charac-teristics S, which are invarianit with the alternative chosen. The LltilitV

function) of an individual may then be writte n as

l = U! ( XI ,. e)

forj= 1I- wher-e e is an unobserved vector contairning all the attributesof the alternative and individual char-acteristics that are not amenable tomeasurement. lHence the valuie of' U at any X, S, is a randomTI variabledependinig on the random variable £.

The individual will choose alternative i if

1'(X' S) >V( S

for all j* iand j= 1.1 Hence the probability of choice i can be written as

Pi = Prob I (Z( X ' S j > lU'(X 7 S) I

forj• i, j= 1.1 The stochastic utility fuLiction may be writtein as

I( X, A) = V ( X. -S ) r1 ( . .V)

where V is nolistochastic ancd 11 is stochastic. The probability of, choice imay then be rewritteni as

P = Prob [11 (X S) -n (XX S) < V'(X' S) - lV(X! S)I

22 6 UNDERSTANDIN(G THiE DEVELOPING METROPOLIS

for j i, j= 1..1 Once the Tj (X S) are given a functional form (D, theprobability may be written as

Ij= J¢(tX, S)dt

where <P is the cumulative joint distribution function of the randomcomponents rij of the stochastic utility function.

The only computationallv tractable distribution function that leads toa plausible stochastic specification is the Weibull distribution, whichma'y be wuritten as

(D ( 11) = r- iT' + !' ex p [ e- rT ] c

where a is a parameter.The Weibull distribution is similar to the normal distribution but is

skewed withi a thininer left tail and a thicker right tail. Two properties ofthe Weibull distribution are relevant for the travel choice problem: themaximum of a Weibull distribution is itself a Weibull distribution, andthe difference between two Weibull distributions has a binary logisticdistributioni. Thus.

exp (V I+ a.)Prol) [i V + 11 V. + j for-j (l*I, i•)I = ----

exp (V,- a.)I~

whicih may be rewritten (suppressing a() as

exp I V(Xj , S)]P. =

,, exp [ V(Xj .S)]

jE I

This provides a formulation for multinomial logit estimation. If Vis lin-ear in parameters. Imay be written as

'(Xj. , .S) = 'ZI

AUTOS. TAXIS. Bl'SES, ANI) BtLSETAS 227

and

exp (PZ)

,exp (P'Z)

tE I

This expression may be transformed to a logit foirm:

plogit (PI) = In

where the logit function is estimated by maximuLn likelihood tech-niques. It is important to understand that the coefficients D must be esti-mated to be interpreted correctly. The easiest way to understanid them isto derive an expression for the slope of the estimated logit function. It isfounld that the slope with respect to a specific variable may be written as

(8 1i) / ( zid PP,(1I - ,)

where is the estimated coefficient. Elasticity of probabilitv Pi withrespect to variable Zjk iS

P, = , (1 -P) Zlk

Here, both the slope and elasticity in such an estimationi are nonliniearand need to be evaluated at a specific point.

The Accessibilitv Index

The accessibilitv measure used in the trip gener-ation equations was com-pu ted from the parameters esti mated in the mode choice equations. It wasa composite measur-e of accessibility as perceived by the individual trav-eler. In principle, this accessibility measure is an index of the maximumutility derived by the traveler by choosing thie most appropriate alterna-tive-that is, Max tU, fOr all i e I, where 1 is the choice set of traveler 1.and U', is the utility of traveler I associated with alternative i. The accessi-bility measure is the expected value E [Max l.',j for all i E I, This mnea-sure can be computed fi-om the parameter estimates derived in themode choice equations, using the same stochastic specifications of theutility function as previously derived.

228 UNDERSTANDING THIE DtVELOPING METROPOLIS

Table 8A-1. Local Registration Tax Incidence: Bogota, 1980(I 98f) Colombian pesos)

Local InidewnceIneome Ca ys Pr 1l0) Income registralion (percentage ofquinfil,e bou.eholds per month fee month/v income)

1 5.6 3,388 4.9 0.142 5.8 7,102 5.1 0.073 8.0 11,288 7.0 0.064 17.9 18,788 15.7 0.085 72.0 53,523 63.0 0.12

WAeightedaverage 21.7 18,659 19.0 0.10

Nole: The tax is Coi$0.45 per 1 ,000 pesos of vehicle valie; the 1980 vehicle value was esti-mnated from the aver-age valie in 1978. Quinitiles are by ascending order of householdincome.

Souroe: Pachion (1981 b); numliber of cars fiom 1978 World Bank-DxNF HouseholdSurvey.

Table 8A-2. Sales Tax Incidence: Bogoti, 1978(1978 (olombian pesos)

Inridence.Income (Caf per 100 Lfffert on Effect I Monthl/v (percentage ofqoitdtiep heuseholds rarprite on in(om,e income monthlT income)

( 1 ) (2) (3) (4) (5)

] 5.6 1,663 29 2,378 1.222 5.8 1,712 30 5.019 0.613 7.1 2,587 12 7,921 0.534 17.9 5,371 94 13,149 0.715 72.0 21,602 378 37,569 1.01

W'eightedaverage 21.7 6,518 I1 15,094 0.87

Note: Quii tiles are by ascendling order of household income.SSourre: Pachon (1981b).

Table 8A-3. Customs Duty Incidence: Bogota, 1978(1978 Colombian pesos)

Effect on car price Fffect on cost of (apital Inuienlce(percentage of

Income quintie Imported Domestic Weighted average Imported Domestic Weighted average monthly income)

1 2,640 443 1,476 46 8 26 1.092 2,764 469 1,045 48 8 27 0.54

3 3,789 642 2,118 66 1 1 37 0.474 8,525 1,445 4,765 149 25 83 0.635 34,283 5,813 19,164 600 102 335 0.89

Weighted average 10,340' 1 ,7 5 3 b 5,779 181 31 101 0.78

Vote: Quintiles are by ascending order of houLsehold income.a. Effect on price for imported cars is 37 percent of the average price of the car.b. Effect on price for locally assembled cars is 5.64 percent of the average price.Source: Pachon (1981 b).

Table 8A4. Incidence of Taxes and Subsidies on Fuel Consumption of Private Cars: Bogoti, 1980

Net subsidyIncome hiel consilfnption Subsiukd (pesos 7hx (pesos Net so bsuiy Subsidy (percentage lax (percentage (percentagequintile (gallons per monih) per month) per montih) (pesos per month) of incole) of incorne) of income)

1 1.74 +47 -20 +27 +1.39 -0.60 +0.792 1.82 +49 -21 +28 +0.69 -0.30 +0.393 2.50 +67 -29 +38 +0.60 -0.26 40.344 5.62 +151 -65 +86 +0.81 -0.35 +0.465 22.61 +611 -262 +349 +1.14 -0.49 +0.65

Weightedaverage 6.82 +184 -79 +105 +0.99 -0.42 +0.57

Note: Quin tiles ate by ascendinIg order of household income.Source: Pachon (1981h).

AULTOS. TAXIS, BUSES, AND BUSETAS 231

Table 8A-5. Models for Choice of Work Trip Mode: Bogoti, 1972 and 1978

E.stimnated coefficient

Independent variables Bogvitd, 1972 Bogotd, 1978

Travel time (minutes), all motorized modes -0.0257 (1.3) -0.0675 (2.5)

Travel time (miitites), walking -0.0688 (6.0) -0.1083 (9.4)

Cost of travel (Colombian pesos/'wage rate per minute)h -0.3013 (1.0) -1.083 (4.0)

Gross familv income'Automobile driver/passenger 0.00013 (3.2) 0.00017 (6.6)Taxi/coltjchvo 0.00018 (4.2) 0.00018 (5.8)Bus 0 0Buseia/microbus 0.00010 (2.7) 0.00011 (6.1)Walk (<30-minute trip) -0.00003 (0.4) -0.00002 (0.6)

Head of household(1 = head, 0 otherwise)

Automobile driver/passeniger 1.61 (5.2) 2.192 (7.8)Taxi/colectivo 0 0Bus 0 0Buseta 0 0Walk 0.263 (0.6) 0.672 (2.5)

Car competition variable(number of vehicles/number of workers)

Automobile driver/,passenger 3.450 (4.0) 1.107 (2.2)Taxi ,'colectivo 0 0Bus 0 0Buseta 2.333 (3.4) -0.117 (0.3)Walk 2.702 (2.4) -1.707 (2.1)

Buseta/microbus accessibilitiv dummvvariables

Poorly served (1) -0.465 (1.4) 0Well served (0) 0 0

Allernalive-specific conslantsAutomobile driver/'passenger -1.696 (3.0) -2.410 (4.5)Taxi/colectiv'o -2.167(5.70) -3.749 (6.9)BUs 0 0Buseta -1.302 (6.2) -0.761 (6.2)Walk 0.754 (0.9) 2.287 (4.9)

Modelstalislics Bogvoi, 1972 Bogrmi, 1978

Log likelihoodAt zero -894 -1,538At convergence -524 -1,014

Likelihood ratio index 0.414 0.340

Number of cases 732 1,244

(7abl. continues on the following pagr.)

232 UNDERSTANDING THE DEVELOPING METROPOLIS

Table 8A-5 (continued)

Sample constitution: 1 97 2 d Number of travelers in sample who choose

Automobile driver/passenger 223 (305 have alternative in choice set)Taxi/colectivo 38 (338 have alternative in choice set)Bus 321 (733 have alternative in choice set)Busela/microbus 115 (520 have alternative in choice set)Walk 35 (590 have alternative in choice set)

Sample constitution: 1978 Number oJ travelers in sample who choose

Automobile driver/passenger 323 (417 have alternative in choice set)Taxi/colectivo 21 (614 have alternative in choice set)Bus 467 (1,239 have alternative in choice set)Busela/microbus 378 (1.239 have alternative in choice set)Walk 146 (909 have alternative in choice set)

Note. Available alternatives include drive/shared ride, taxi/colectivo, bus, buseta, and walk.a. Asymptotic t-statistics are in parentheses.b. Cost of travel measured in Colombian pesos. Wage rate measured in Colombian pesos

per minute.c. Measured in 1972 Colombian pesos per month.d. All individuals in households owning automobiles have driver/passenger in their

choice set. Households living in inner areas of the city (rings 1-3) have taxi/colectivo. Alltravelers are assumed to have access to standard buses. Households living in regions servedby busetas (according to route maps) have the buseta/microbus alternative-all householdsin 1978. Travelers living within 12 kilometers of their work location are assumed to be ableto travel there on foot.

Source: Kozel (1986).

Notes

This chapter is based on work originally reported in Cifuentes (1984), Kozel(1981, 1982, 1986), Pachon (1979, 1981a, 1981b, 1981c), and Westin (1980).Alberto Hernandez and Emilio Latorre also contributed to the research, andGregory Ingram directed the research design. The section on "Modeling TravelDemand" reports the work in Kozel (1981, 1982, 1986).

1. Data about the value of vehicles by age were available from informationassembled by the Centro de Estudios para Desarrollo Economico of the Univer-sidad de Los Andes, Bogota.

2. A trip, incidentally, is taken as a one-way trip: travel from home to work andreturn is counted as two trips.

3. See the appendix to this chapter for the derivation of the accessibilityindex.

4. See Domencich and McFadden (1975), Maddala (1983), Hensher andJohnson (1981), Charles River Associates (1976), Ben-Akiva (1973), Lerman(1975), Maniski and Lerman (1977), Hensher and Stopher (1979).

5. See Kozel (1986) for a detailed evaluiation of transferability of such models.

Chapter 9

Urban Government and Finances

There is a wide variation in the range of administrative, service, and fi-nancial responsibilities borne by local governments. Most cities exhibita complex web of relationships, usuially hisiorically determined, be-tween the local muniicipal governmenit and state and national govern-ments. These relationships become more complicated in capital citiesbecause the national governments often display greater interest in therunning of these cities. Municipal bodies in most cities are responsiblefor proxiding local public services, such as water supply, sanitation andsewerage, solid waste disposal, roads, and street lighting. They also regu-late such activities as land use, land developrment, and constructioni, andthev provide public safety services in the form of a police force. In addi-tion, in many cities, local governments are responsible for providinigpublic transportation, power, telephones, public housing, education.and health care. In some cases there is a distinction betveen investmentand mainteniance responsibilities, because trban infrastructure invest-ment is characteristically expensive and local bodies seldom have ade-quate resources for making such investments.

In order to provide services, municipal governments have certain tax-raising powers, most commonly related to taxing real property but alsoin many cases related to taxing industrial and commercial activities.Because muniicipal governments are typically responsible for controland regulation of land use, zoning, building by-laws, and developmentcontrols, it is also common for them to have overall responsibilities formedium-term investmernt and physical plals. As may be expected, allthese functions have to be performed by some entity: what variesbetween cities and countries is the degree of private involvement andhow responsibilities and powers are shared between the different levelsof government, as well as between government and empowered para-statal public agencies.

233

234 UNDERSTANDING THE DEVELOPING METROPOIIS

Urban Government in Bogota

In the case of Bogota, the district government is responsible for a largernitmber of functions than usual for local government in developingcountries. The city is an autonomous administrative entity, the DistritoEspecial (Special D)istrict), and does not form a part of any state. This isa commoin pattern for many capital cities, including New Delhi, MexicoCity, and Washington, D.C. The Distrito Especial was carved out of thestate of Cundinamarca in 1954 by consolidating the Municipality ofBogota with a numniber of adjacent municipalities. The mayor of Bogota(the alcaldle) is directly nominated (and removed) by the president ofColonibia. He has no fixed term and is not elected. Over the last seventyyears, the average tenuLre of the mavor has been just over one year!

The district government of Bogota is responsible for providing almostall public services: health, education, roads, water and sewerage, power,telephones, and waste disposal. It has minimal presence in the supply ofpublic transport and public housing, hiowever. The administrative struc-ture of Bogota that delivers all these services is highly decentralized butessentially organized along functional lines. The mayor is assisted by acabinet of five secretaries: Interior (Gobierno), Finance (Hacienda),Education (Educaci6n), Health (Salud), and Public WVorks (ObrasPfiblicas). In addition, there are four administrative departments (De-partamentos Acmi nistiativos): the planning department (DepartamentoAdministrativo de Planeaci6n Distrital, DAPD), the transport department(Departamento Administrativo de T'ransporte v Transito, DATT), socialwelfare (Departamento Administrativo de Bienestar Social), and com-muniity development (Departamento Adininistrativo de Acci6n Com-munal). Each departlnlelnt is headed by a director who reports directlyto the mayor. There are three other officers: the legal representative(personero), the treasurer (tesorero), and the controller (contralor). Eachof' these is appointed by the city cotlncil, an elected body of twenty mem-bers and tweinty alternates.

Services are delivered primarily by several decentralized agencies.The main agencies were the power company (Empresa de Energia E1ec-trica de Bogoti, EEEB), the water and sewerage company (Empresa deAlcantarillado v Acueducto de Bogota, EAAB), the telephone company(Empresa de Telefonos de Bogota, ETB), and the general public servicecompany (Empresa Distrital de Servicios Pt6blicos, EDIS). The decentralizedagencies account for a major portion of the district budget, in particularthe capital buLdget, and are largely self-financinig. Another implement-ing agency, mainlv concerned with capital infrastructure works, is theInstitute of' Urban Development (Instituto de Desarrollo Urbano, IDU).

It was funided mainly from valorization charges and was very active inthe early 1970s but has declined in importance since. The (listrict also

UtRBAN (;OVERNMENT AND FINANCES 235

has its own social security agency, the Caja de Previsi6n Social; a smallpublic housing agency, the Caja de Vivienda Popular; and a small publictransport company, Empresa Distrital de Transporte Urbano (EDTU). Asthis list suggests, most of the local government functions are carried outthrough relatively autonomous decentralized agencies.

The management of many large cities in the world suffers from exces-sive geographical fragmentation. It is not unusual for a large metropoli-tan area to fall into the jurisdictions of a number of local governments,as do the Washington, D.C., Calcutta, and Mexico City metropolitanareas. Such fragmentation creates a number of problems, includingunieven tax rates, management of services across boundaries. and vary-ing building and zoning regulations. The problem of fragmentationdoes not exist in BogotA because the (listrict covers almost the entiremetropolitan area, thanks to annexation of selected adjacent municipal-ities. After official extension of the urban perimeter, the major publicutilities had even begun to service some areas outside the district towhich development had spread. As the city expands, it may becomemore difficult to continue annexing municipalities, however, and frag-mentation ofjjurisdiction may become more of a problem in Bogota.

An interesting feature of the extensive functional decentralization inBogota is that some agencies have to conform to national policies andcontrol despite being essentially autonomous. Service charges for water,electricity, and telephones must conform to nationial tariff policies. Edu-cation and health services are regulated by national tax subventions andcontrols. The borrowing ability of the agencies is limited bv the need forpermission and oversight by the national government. However, there islittle interference by the national governmenit in the day-to-day adminlis-trative activities of the district government and its decentralized agen-cies. The district government is also quite independent in its authorityover urban planning functions and the preparation and implemenita-tion of its annual budgets.

There is no centralized consolidated district budget in Bogota. In-deed, in conducting this study it was not easy to obtain the full pictureof district finances over a reasonable historical period.' We launched aspecial effort to compile these data for 1961-79 in order to gain an ap-preciation of the trends in the district's revenues and expenidituires.There was a rapid 23 percent annlual growth in total district governmenltexpenditure in nominal terms. Between 1961 and 1979 consolidateddistrict expenditure grew by a factor of 37 in current prices. In constantprices there was little growth in per capita expenditures-about I per-cent per year. This trend is not easy to discern, however, because growthin constant prices fluctuates widely along with inflation rates (see figure9-1). It is obviously difficult to calibrate taxes and user charges so thatthey keep up with inflation in a regular sm)ooth fashion and so that ex-

2:36 UjND)ERSTANDIN(G THE1: I)EVELOP[NG METROPOIAS

penditures can be kept up in real terms. It is interesting to note that ex-penditur es in Cali were at a similar absolute level and exhibited a similartrenid over the same period. Although it is creditable for a local govern-ment to be able to keep up real per capita revenues and expendituresover a long period of extremely rapid population growth along with in-flation, the growth of per capita real income in Bogota was still higher,at about 1.7 percent per year from 1960 to 1975 (see chapter 3, table 3-2). Local government's share of total city income therefore declinedover this period. The revenue yields closelv followed expendittires. Thedistrict government itself accoulited for only about 20 percent of expen-ditures over the whole period. The various decentralized agencies ac-coUnltedi for the rest: the three big utility companies-EEEB, EAAB, andETm-had a combinied shiare of more than 60 percent, increasing to al-most 75 percent in Ilhe late 1970s.

Ain Overview of Financial Struc/ure

In Bogota durinig the period studied, user charges financed more than50 percenit of all expenditures, particularly for the utilities. This mighthave beeni expected from the decentralizecd nature of the city's utility

Figure 9-1. Total Consolidated Expenditure Per Capita of the BogotaDistrict, 1961-79

Pesos

4,00(

1961 1 963 1965 1967 1969 1971 1973 1975 1977 1979

Yea-R

.Smrnre: Linn (1984).

URBAN GOVERNMENT AND FINANCES 237

companies. Similarly, borrowing, at 20 to 30 percent of total revenues,financed much of the capital expenditure. Per capita revenues fromtaxes fell in real terms over the period, implying a failure of the tax sys-tem to fully compensate for inflation. Transfers and revenue sharingshowed a slight increase in real per capita terms and went up to about15 percent of all revenues in 1976. By the late 1970s, the revenue struc-ture was quite changed from that in the early 1960s. The reduction inreliance on tax revenues and transfers was matched by an increasingshare of user char-ges.

As might be expected from the highly decentralized nature of publicservice agencies in Bogota, district budgeting and investment planniirgare highly decentralized. Although there is an elaborate coordinationsystem designed to link the activities of various agencies with those ofthe district government, there is very little effective coordination. This ispartly because of a lack of genuine medium-term urban planning andsystematic evaluation of urban investments. It is also a restult of the self-financing nature of the main decentralized agencies. Their budgets areconsiderably larger than that of the city government per se, and theyneither receive nor pay any funds to the district budget. Moreover, theirstaffs have greater continuity than the leaders of the district govern-ment. In reality, then, the decentralized agencies are left pretty mulchalone to plan and implement their projects and programs.

Within the district government, the budget for current expendituresis prepared in the Finance Secretariat (Hacienda) and the investmentbudget is prepared by the District Planning Department (DAPD). For-mally', the budgets have to be approved by the District Council. Addi-tions and amendments to the budget are frequent over the spendirngyear. Similarly, the budgets of the decentralized agencies, after approvalby their respective boards of directors, are presented by the secretary offinanice to the District Council for approval. When changes in the bud-gets are approved, implementation can take place, but the payment

systems and auditing system are quite bureaucratic. The control experi-enced is essentially "numerical-legal" rather than administrative (Wolff1984). The agencies supposed to supervise the executioll of the budget,the Finance Secretariat and the DAPD, are too weak professionally toexercise effective control. However, the auditing proctedures are quitestrict and cumbersome: there is 100 percent auditing rather than ran-dom sampling. The controller's office has to verify all expenditurestwice-once before and once after the expenditure. For the maindecentralized agencies, the auditing agents are revisoresfiscales who arenominated by the mayor and the D)istrict Council.

Credits have historically financed the major part of investment financ-ing. Between 1971 and 1976, credits financed about 60 to 90 percent of

218 LUNDERSTAND)ING THE I)DEVEKL.OPING MF1 ROP'OlIS

investments in different years. To raise any credit over Col$1 millioni(about US$22,000 in 1980) required formal approval by the nationalgovernniel1t. This approval consists, at different stages, of the concur-

rerice of the National Planning Departmenit, the Ministry of Finance,and in the case of foreigni credits, the Central Bank (Banco de laReptiblica). The Monetary C(ouncil (Junita Monetaria), attached to theFinance Ministry, also has a say in the approval of credits. The NationalPlanning Department evaluates the economic viability of projects beforegranting approval. For the decentralized agencies, foreigin credits havebeen the most important source of fun(ding. The existence of these vari-ous controls, in addition to the technical supervision of the interna-tional lending agencies, has conitribtited to upgrading the quality ofprofessionalism in the decentialized agencies. In the absence of a coor-dinated overall investmlenit plan. however, each agency essentially makesinvestments in response to its own perception of increasing demandsancl proceeds incrementally. This has often resulted in poor timing orsequenicinig of investments by the different service agenicies.

Overall, the adminiistiative fragimienitationi of Bogota is reflected infragmentation in the budgetary pIrocess. Although much attention hasbeen paid to organizing formal mechanisms of control anid coordina-tion among the different agencies, in practice there was little effectivecontrol. The maini decentralized public service agencies are more pow-er-fili thani the district government and hence operate quite autolno-mously. The lack of continuity in the district governimient is anotherfactor- in this relationiship. As alreadv mentioned, the mayor seldomholds office for much longer than a year. The city council is electedeverv two years ancl meets four times a vear for a month each time.There is little continiuity among council members, unlike the U.S. Con-gress, for example, where many members are reelected many times.Representatives of the couLIcil who serve on agency boards exercisesome influence but not a significant amouLnt. The general conclusioni isthat "the executive ullits exercise more political power than the legisla-tive District C.ounlcil" (WNtolff 1984). The continual chaniges in officials,council members, and mayors contr-ibute to the difficulty of makinigcoherenit long-ter-nm strategies and approaclhes.

The paradox is that, in spite of all the problems, public service deliv-ery has been relatively efficient in Bogota in terms of people's access toessential services such as water supply. urban transport, sanitation andsewerage. roads, power, andl even telephones. A fragimienited structu]remade up of power-ftil public service agenicies is probably well stiited tothe needs of a rapidly expandinig citv. The agencies' relative autoniomycontr-ibuites to their ability to be morc aware of changinig needs andthen responidinig to themi inicremcnitaliv. Their fiscal autonioiiy has also

ULRBAN GOVERNMEN1 ANI) FINANCES 2 39

helped make them more reliant on user charges, whichi in turni proba-bly makes constimers more demanding. This itself contributes to theagencies' sensitixity to consumer demand for their services. It is quitepossible that a tight, more centralized governmental system might haveperformedl less well in the context of a rapidly growing citv.

The Structure of Revenue

The compilation of the district government's total consolidated reve-nues, incluidinig those of the decentralized agencies, revealed that taxrevenues have become less important. In 1961, taxes contributed about27 percent of total revenues; they had declined to about 16 percent in1979, an improvemiienit from just 12 percenit in 1974. In constant 1976prices, per capita tax revenue fell from Col$482 in 1961 to Col$381 in1979. More noteworthy is the significant increase in the contribtitioll of"self-financed" reveniues. These inclucled service charges and develop-ment contributionis. In relative terms these contributions grew from 45percenit of total revenues in 1961 to about 60 percent in 1979. Revetnuesfrom borrowing fluctuated significantly over the years, depending esseni-tiallv on the investment program requirements of the main decentral-ized agencies. The reventues from national grants and shared taxesvaried between 8 and 15 percenit over the whole period. Grants hadmainly been for education services. Education was "nationialized" in1976 and these funids transferred to a special educationi fttnd. Hlence,grants fell to zero by 1979 (see tables 9-1 and 9-2, and figure 9-2).

It is interesting to compare the revenue structur-e of Bogotai withother cities in developing countries. Comparable data on local fiialicesare not easy to come by, however, and coverage of services by local gov-er-nimiients is dlifferenit in different cities. Broadly comparable data per-taining to the early to mid-1970s were compiled y I.vinn (1980) and arein table 9-3. Bogoti and Cali are unusual in their degree of self-financ-ing charges as souL-ces of revenue. This pattern is a commoni one for theColombian cities listed in the table-Bogota, Cali, and Cartagena. Theother cities exhibiting a similar pattern are Ahmedabad and Bombay inIndia and Lusaka in Zambia. The only other city with a high degree ofborrowing was Karachi, but it had a very low degree of self-financing. Itappears that borrowing costs were being financed througlh high mobili-zation of local taxes. Bogoti, Cali, and Cartagena are outliers in theirlow reliance on local taxes as revenue sources. The Colombiani citiesalso have little dependence on grants anid other devolution froml higherlevels of governments. This would be even more pronouniced if the edu-cation grants were taken out of these data. Overall, Colombian cities

Table 9-1. Consolidated Per Capita Revenues: Bogota, 1961-79(1976 Colombian pesos)

Re-venue 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979

lax revenue 482 460 373 338 354 288 325 324 358 248 364 357 383 325 357 350 326 345 380(,eneralnoni-taYxrevenue 25 26 21 21 25 8 10 12 9 8 12 8 5 10 8 6 7 1l 21Self-financed revenue 792 847 803 740 815 778 895 1,061 1,285 1,348 1,474 1,373 1,317 1,300 1,241 1,395 1,451 1,560 1.432Shared taxes 142 131 185 201 179 150 159 141 171 235 284 255 224 233 259 205 191 199 210G Grants 16 48 73 61 50 57 46 79 43 102 107 86 113 108 133 115 22 31 0Borrowing 327 530 575 440 352 289 572 665 509 488 562 575 557 658 787 516 419 413 322

Total revenue 1,784 2,042 2,030 1,8011 1,775 1,570 2,007 2,282 2,375 2,429 2,803 2,654 2,599 2,634 2,785 2,587 2,416 2,558 2,365

Source: Linn (1984b).

Table 9-2. Structure of Consolidated Revenues: Bogota, 1961-79(percentage distuibiitioni)

Revenue 1961 1962 1963 1964 1965 1966 1967 1968 1969 1 970 1971 1972 1973 1974 1975 1976 1977 1978 1979

Tax revenue 27.0 22.5 18.4 18.8 199 18.3 16.2 14.2 15.1 13.8 13.0 14.5 14.7 12.4 12.8 13.5 13.5 13.5 16.0

General non-tax re\enic 1.4 1.3 1.0 1.2 1.4 0.5 0.5 0.5 0.4 0.3 0.4 0.3 0.2 0.4 0.3 0.3 0.3 0.4 0.9Sell-tinanced revenule 44.4 41.4 30.5 41.0 45.9 49.3 44.3 46.5 54.1 53.3 52.6 51.7 507 4t.3 44.5 53.9 6(1 60.9 f0.3

Shareo taxes 8.0 6.4 9.1 11.2 1(l.1 9.6 7.9 6.2 7.2 9.3 10.1 9.6 8.16 8.8 9.3 7.9 7.9 7.8 8.8(; a I-IL m 0.9 2.4 3.6 3.4 2.8 3.6 2.3 3.5 1.8 1.0 3.8 3.3 4.1 4.1 4.8 4.5 0.9 1.2 0.0

Borirowinig 18.3 26.0 28.3 24.4 19.8 18.I 28.5 29.1 21.4 19.3 20.0 21.7 21.4 25.0 28.2 19.9 17.3 16.1 14.0

Total revenutie I 0.() 100.0 100.0 100.() 100.0 1()(.0 100.() 100.0 10)).() 100.() I ().O 00.)) 100.0() 1()(. 10 100.0 10().() 100.0 1(0().0

.Sozire: Iable 9-1.

Table 9-3. Financing of Local Public Expenditure in Selected Cities by Type of Revenue(percent)

A B C D X V Z T

Self-financing Otherlocal Total local Grants Nei Total externalCity 1'ear Local taxes services revenue revenue and taxes banroziwng' revlenue Total

Francistowni, Botswaana 1972 46.8 56.1 n.a. 102.9 1.9 -4.8 -2.9 100.0Mexico City, Mexico 1968 70.9 5.2 25.8 101.9 8.9 -10.8 -1.9 100.0La Paz, Bolivia 1975 61.9 3.6 31.5 97.0 9.( -6.0 3.0 100.0Tunis, Tunisia 1972 36.8 7.1 50.0 93.9 0.7 5.4 6.1 100.0Kitwe, Zambia 1975 35.0 53.1 4.6 92.7 2.2 5.1 7.3 100.0Valencia, Venezuela 1968 44.8 13.4 32.6 90.8 9.2 n.a. 9.2 100.0Lubumbashi, Zaire 1972 72.8 n.a. 17.7 90.5 9.5 n.a. 9.5 100.(Rio deJaneiro, Brazilh 1967 74.5 7.2 6.7 88.4 1.7 9.9 11.6 100.0Ahmedabad, Ildia 1970-71 38.6 41.8 5.9 86.3 4.2 9.5 13.7 100.0Bombay, India 1970-71 37.9 38.7 8.0 84.6 1.0 14.4 15.4 100.0Karachi, Pakistan 1974-75 67.6 2.2 14.3 84.1 2.8 13.1 15.9 100.0Seoul, Korea 1971 30.3 36.3 13.4 80.0 15.8 4.1 19.9 100.0Jakarta, indonesia' 1972-73 40.6 15.2 23.0 78.8 21.1 n.a. 21.1 100.0

(43.7) (15.2) (23.0) (81.9) (18.1) (n.a.) (18.1)

Lusaka, Zambia 1972 39.3 36.9 2.0 78.2 6.0 15.8 21.8 100.(Cali, Colombia 1972 15.6 57.5 1.3 74.4 2.8 22.9 25.7 100.0Calcutta Corp., India 1974 64.4 n.a. 9.4 73.8 19.4 6.8 26.2 100.0Cartagena, Coloimibia 1972 23.3 43.3 3.8 70.4 12.8 16.8 29.6 100.0Mbuji-Mayi, Zaire 1971 66.5 n.a. 2.7 70.2 29.8 n.a. 29.8 100.0

Manila, Philippines 1970 10.0 5.0 30.0 55.0 n.a. n.a. 30.0 100.0Bukavu, Zaire 1971 67.4 n.a. 2.5 69.9 30.1 n.a. 30.1 100.0Madras, India 1975 54.5 3.7 11.0 69.2 25.1 5.7 30.8 100.0Eogota, Colombia 1972 13.7 48.5 0.3 62.5 14.0 23.5 37.5 100.0Tebrani, Iran 1974 42.8 n.a. 4.1 46.9 45.2 7.9 5'3.1 100.0Kingston,Jamaica 1971-72 23.9 2.7 3.4 30.1 67.2 2.7 69.9 100.0Kinshasa, Zaire 1971 25.4 n.a. 1.5 26.9 73.1 n.a. 73.1 100.0

Median (average)' 42.8 7.2 6.7 78.8 9.5 5.1 21.1 10).0_________________________________ (46.0) (19.3) (11.2) (76.6) (17.7) (5.7) (23.4)

n.a. Not applicable.Note: D = A + B + Cl; Z X + ; T = D + Za. Net boriowing conisists of loan finanicinig minus net changes in financial assets or reser-ves.h. As a r-esult of the exclusion of autonomous agencies, the contu ibritions of self-financing service revenue and of all locally raised revenne are probably

understated, and the remaining enutries overstated.c. Figures not in parentheses include shared taxes under grants; figures in parentheses inclu(le shared taxes under local taxes.d. lotal revenues are used instead of total expenditures..Su rce: Linn (1980).

244 UNDERSTANDING THE DEVELOPING METROPOLIS

Figure 9-2. Composition of the Bogota District's Real Per Capita TotalRevenues, 1961-79

Percent of total

80

60- 60 ~~~~~~~~Self-financed revenues /

40Tax and renewal

Borrowing non-tax revenies

20

Shared txes

1961 1963 1965 1967 1969 1971 1973 1975 1977 1979Year

.So1rct. Fawcett (1984).

have a different financing pattern from most other cities in the develop-ing world: they exhibit much greater reliance on user charges as ameans of financing urban services.

The property tax is thie most cominonly levied local tax in almost allcounitries in the world (see table 9-4). On average about half of all taxescome from the property tax. Another 25 to 40 percent of' tax reveniuelypically comes f'rom some tax related to economic activity. Among mostHispanic counitries, this is a tax levied on industry and commerce; inInidia the cor r espondinig tax is octroi (although it is generally agreed thatoctroi should he replaced by a more general tax related to economicactivity); in other countries soimle kind of incomiie tax is levied. In theColombian cities. the property tax fetches almost 60 percent of' tax reve-nue; the industrv and commerce tax are the other major taxes.

Between 1960 and 1979, receipts from the property tax fell in con-stant per capita ternms. (See table 9A-1 for revenues of' the Bogota Dis-trict Administration from 1960 to 1980). The fall in relative importanceof' tax revenules via-a-vis the total district revenues is largely the result ofthe declining performanice of the property tax. Wk'hy did this happen?The basis for property taxation in Bogota is capital value of property, as

URBAN (;OVERNMt:N r ANUI FINAN(:tCS 245

distinguished fiom the "annual value" taxation fotind in most formerBritish colonies. Whereas it is difficult to prescribe a normative level ofproperty tax, the rate of property taxation seemed low in the late 1970s.The effective taxation rate-the ratio of property tax collection to esti-mate(d market value of properties-was less than 0.5 percent. In theUnited States it is generally greater than I percent. One caveat needs tobe expressed here, however: a number of the service user charges inBogota are hasecd on property values, so at least a portion of the usercharge collection is related to property values, thus raising the effectiveproperty tax rate in Bogota.

The data system related to property taxation is relatively well svstem-atized in Bogota and has been computerized for some time. The city hasalso been quite ef'ficient in adding newv properties to the system: forexample, the numnber of properties in its files doubled between 1975and 1979. The District Cadastr-al Of'fice also m1ainitains a land valie map,which is continiuially updated. It was interestinig to find that the assessedvalues in this data set kept up well withi market values: they were evenoverestimated in maniy cases. Yet assessment ratios for property tax pur--poses hadl declined during the late 1970s. The statutory tax ratio was0.84, but it was estimated that the effective assessment ratio had proba-bly declined to below (04 in the late 1970s. It seemed that valies in thieland value map were seldom used for revaluilig properties for tax pur-poses. Thiere was also a lag in following up chanlges in land uses-forexample. from residenitial to commnercial uses.

As a result of concern over the performance of the property tax andthe overall fiscal systemi. particular ly the shar-ing of fLund(1s andl respolisi-bilities aniong national, (lepartineital (state), and muniicipal govern-ments, the government appointed a Coommission on InteigovernimentalFinance in 1980. Withi changinig circtmmstances at both the local andnationial levels, it is usefil to appoint such special commissions at regu-lar intervals-say, every five or ten years. The actual arrangements madedepend on the particular circumstances of the country and its tax svs-tem, but the principles should be to ma;ke revenue sources commenisu-rate wvith the responsibilities of' providing services and to encour-agedecentralization of service provision as well as revenue raising.

In this case the commnissioni's basic concltision was that local govern-inent functionis should r ely on local resources for ftmunding. This r esultedin new legislation aimed at strengthening local government institutionsand tax sourices. Specific attention was given to the admi-nistrationl of theproperty tax. All cadastral values were updated to a base of' 1983 valuesby inflating all values bv an annual incr-ease of 10 percenit fi-om the yearof' their last valuation. Subsequientlv, there was to be an anniual increaseby a price index. This adjustment has not kept up with full inflation, butit is, nonetheless, a step in the right directioni in injecting greater tax

Table 94. Distribution of Local Tax Revenue in Selected Cities by Source(percemt)

A B C D E F ( fl / J K 1. T

Local taxes Inditstr5as perrent Prop,ert Enter- and Ala/or

of lvwal Propert iransfer In-orne General Gasoline lainment commerne vehicle Camb- All otherCita Year expenditure tax tax letx sales tax Octroi Beer tax ltax tax tax tax ling tax taxes Total

Managua, Nicaragua 1974 84.3 - - - G9.4 - - - 2.3 21.1 3.2 - 4.1 100.0Mcxico City, Mexico 1968 70.9 33.5 2.8 - - - 1.1 2.6 44.2 - - 15.8 100.0Valencia, Venezuela 1968 44.8 21.4 - - - - - - - 66.7 11.8 - - 100.0Bogota, Colombia 1972 13.7 58.4 - - - - - 1.8 7.0 18.2 5.1 - 9.5 100.0Cali, Colombia 1974 23.2 54.0 -a _ 6.1 27.8 4.3 - 7.8 1O0.0Cartagena, Colomibia 1972 23.3 61.2 - - - - - - 4.4 12.2 2.1 5.8 14.2 100.0La Paz, Bolivia 1975 61.9 5.2 - - - - 7.1 - 1.5 73.8 - - 12.4 100.0Manila, Philippines 1970 55.0 61.9 - - - - - 2.2 - 32.1 - - 3.8 100.0

(Group average 37.0

Klarachi, Pakistan 1974-75 67.6 46.0 - - - 49.9 - - - - 3.0 - 1.( 100.0

Ahmedabad, India 1971-72 38.6 43.0 - - - 52.0 - - - - 2.0 - 3.0 100.0Bombay, India 1971-72 37.9 55.6 - - - 37.7 - - 0.3 - 3.7 - 2.7 100.0

Calcutta Corp. India 1974-75 64.4 64.8 - - - 27.1b - - - - - - 8.2 100.0Group average 41.7

Francistown, 12.9Botswana 1972 48.8 - 61.1 - - - - - - - - 26.0 100.0

Lusaka, Zambia 1972 39.3 74.6 - 25.4 - - - - - - - - - 100.0

Ndola, Zambia 1972 n.a. 75.6 - 24.4 - - - - - - - - 1 00.0Kitwe. Zambia 1972 n.a. 80.0 - 20.0 - - - - - - - - - 100.0

Kinslhasa, Zaire 1971 25.4 - - 14.4 - - 62.5 - - - - 23.1 100.0Bukavu, Zaire 1971 67.4 - - 3.7 - - 87.0 - - - - - 9.3 100.0Mbuji-Mayi, Zaire 1971 66.5 - - 62.7 - - - - 37.3 100.0Daegu, Korea 1976 n.a. 49.5 21.2 9.1 - - - - 10.4 - 5.4 - 3.5 100.()Gwangju, Korea 1976 n.a. 50.3 23.1 13.2 - - - - 6.6 - 4.1 - 2.9 100.0D)aejeen, Korea 197ti n.a. 51.0 20.1 9.7 - - - - 10.7 - 5.5 - 3.0 1().0Jeonju, Korea 1976 n.a. 52.0 24.4 8.9 - - - - 7.5 - 4.9 - 2.1 10o.0

Group average 23.0

Seoul, Korea 1971 30.3 20.6 34.8 - - - - - 16.4 - 22.2 - 6.0 100.0Madras, India 1975-76 54.5 68.9 5.1 - - - - - 16.0 - - - 10.0 1(0.0Tehran, Iran 1974 42.8 55.3 - - - - - - 9.1 - 10.1 - 25.6 100.0Tunris, Tunisia 1973 36.8 82.6 12.8 - - - - - 4.6 - - - - 100.0Jakarta, Indonesia 1972-73 43.7 - - - - - - - 16.9 - 50.2 26.9 6.0 100.0l.agos, Nigeria 1962-63 50.9 100.0 - - - - - - - - - - - 100.0Kingston,Jam1aica 1972 23.9 10(.0 - - - - - - - - - - - 100.0Rio dejaneiro, Brazil 1967 84.4 3.9 1.0 - 89.2 - - - - - - - 5.9 100.0

Overall median 51.0

Overall average 44.6

n.a. Not available.-Negligible.Vote: Cities grouped roughlv according to similaritv of financial structtlre. T= the sunm of colhmns A-L.a. Share tax r-eceipts not shown under local tax reventle.b. Shar e of octroi receipts collected bs the Calcuitta Metropolitan Development Authoriw.Somrce: Linn (1984b).

248 UNDERSTANDING THE DEVELOPING MFETROPOLIS

buoyancy into the system. Further flexibility has been given to munici-pal governnments to fix the tax ratio at any point between 0.4 and 1.2percent, compared with a fixedl rate earlier. These changes are quiterealistic and pragmatic. Most cities in developing countries suffer fromthe problemii of inefficiently updating property tax valuatiolls. Some areconstricted bv rent control legislation, which naturally provides a cap onpossible property valuations. Now that property records can easily becomputerized, it is much easier to keep propertv records up to date.The application of gradual anniual incr-eases along with inflationi alsoprovides a convenient method of tipdating values: this is also unllikely tobe resisted strenuously by taxpayers, because it is easier to sustain small,regular changes thani any large change.

An inte r esting feature of property' tax administration in Bogota is theapparent progressiveness in the incidence of the property tax. It is oftenassumedl that the propertv tax has regressive incidence, but in Bogotaand Cali it was found that the exemption of properties of low value doesmuch to promote progressiveness while not losing any significant reve-nue. There was also a tendency to underassess lanid values in the poorerareas of the city and to overassess those in the better-off areas, whichhelped lend a degree of progressiveness to tax incidence.

The other main tax in Bogota and Cali is the induistry and commercetax. Earlier, the tax was levied on different activities at different rates-mostly specific rates. As a result it had a declining trend until 1974,when it was transformecl into a uniform ad valorem rate based oni grosssales. This made it into the most buoyant tax in the district. This reformwas extended to other Colombian cities by the Commission on Intergov-ernmental Finance in 1983.

Other taxes include the motor vehicle tax and the urban constrtictiontax. The former is another commonly levied tax in most cities and onethat can be a very buoyanit source of revenue. Automobiles are majorusers of infrastructure that is not easily amenable to incremental usercharges. Compared with public transport, they are also inefficient usersof road space. Ownership, natuLrally, is also with a small (although in-creasing) affluenit minority in developing countries. Typically, however,taxation of motor vehicles is inadequate because of the natural politicaland bureaucratic influence of car owners. In Bogota revenues from au-tomobiles declined in real per capita terms until 1976, but posted a verysubstantial increase between 1976 and 1980. As with property, it is essen-tial to continually update the tax rate with inflation. Given the indefi-nitely' ising trends in automobile owner-ship in cities in developingcountries, serious thought must be given to making this tax a significantsource of revenue for the funds reqtiired for urban infrastructure main-tenance and investment. It would be rational to link the level of the mo-tor vehicle tax to an index of road maintenance and investment costs.

URBAN GOVERNMENT AND FINANCES 249

The Commission on Intergovernmental Finance also recommendedincreasing the devolution of national tax collections from the sales taxto municipal governments. The sharing was to be increased from 30percent in 1986 to about 50 percent by 1992, and the sharing formulahas been skewed toward smaller municipalities and those who succeedin raising their owIn resources.

Valoiization

The other source of revenue in Bogota that has been an importantmeans of financing intrastructure investment has been the system of val-orization charges, which has frequently been used to finance infrastruc-ture improvements in built-up areas. The idea is to recover the cost ofinfrastructuire investments from properties that are in the vicinity of theimprovements and/or likely to benefit from capitalization resultingfrom these improvements. This method was used extensively in Bogotaduring the late 1960s and earlv 1970s. The implementing agency wasthe IDU. At the peak of its operations, valorization revenues amountedto almost two-thirds of the property tax revenues in 1969 ancl financed16 percent of all consolidated expenditures. This was the phase ofheightened urban constructioni activity, which was then regarded as thelead activity promoting economic development and employment. Valo-rization-financed expendittires fell to 8.6 percent of the total in 1976and 3 percent in 1979. The procedure of lev,ing valorization tax hasbeen quite complex: properties are levied a differing rate of tax depend-ing on their proximity to the improvement financed. In the case of roadimprovement, for example, the rate of valorization varies with distancefrom the r oad.

One reason that the importance of valorization-related activities hasdeclined has been that the system has encountered problems in fullyrecovering its investments. This has been caused both by ar r ears in col-lection-particularly from public agencies-and bv a feeling that peo-ple in low-income neighborhoods could not be expected to pay fullvalorization charges. The resulting shortfalls have necessitated transfersfrom the district government, leading to a reduced enthusiasm for valo-rization projects. Another problem that has cropped up is that legalnorms restricted the concuL rrent coverage of one property by' more thanone valorization charge. Hence, new projects are difficult to carry out inpreviously covered areas. Moreover, cost overruns are difficult to passon through valorization charges: if they result from inefficienicy there isno reason the beneficiaries should suffer. Finally, the lack of adequateinfrastructure investment planning is also responsible for a slowdown invalorization-financed investments. It shotild be noted that one elementof good planning is the involvement of beneficiaries: when beneficiaries

250 UNDERSTANDINC; THE DEVELOPING MEIROPOLIS

have been involved in planning projects, it has been easier to collect thecharges.

Despite the problems encountered, Colombian cities have effectivelyused the valorization process to raise revenues for financing urban infra-structure investments. The method is well stiited for financing invest-ments in built-up areas and can be used to great benefit in othergrowing cities, although its potential as a method of financinig urban in-frastructur-e is limited in cities where government agencies owIn a largeproportion of land and propertv. Some of the problems encountered inBogota also suggest that it may not be feasible to finance improvementsin infrastructure from this source repeatedly. Taxpayer resistance is like-ly to develop if the same properties face valolization charges a second orthird time. Some resistance could be avoided by better participation ofaffected communities. Even if valorization charges cannot be levied tocover the full costs of construction, they could be uised to recover a sig-nificant portion of the costs.

Revenue Sharing

Shared taxes from the national government have generally ranged from7 to 10 percent of total consolidated per capita revenues for the BogotaDistrict. Between 1961 and 1978, real per capita shared taxes grew by alittle more than 2 percent a year, a performance that was much betterthan the growth of local taxes. The shared taxes consisted of the beerconstumption tax, the shared sales tax, the gasoline tax, and the tobaccotax. The most important of these for Bogota has been the beer tax,whose revenues often exceeded those from the property tax. The salestax r evenues were primarily earmarked for education, specifically teach-ers' salaries. Since 1976 these funds have been directed to the RegionalEducation Fund (Fondo Educativa Regional, FER) and no longer enterthe district's budget. The FER also receives direct grants from thenational government under a revenue-sharing program called the Situ-ado Fiscal, which has been in operation since 1969 and formalized since1971. Fifteen percent of total national revenues, excluding taxes alreadyshared with subnational governments, was assigned for revenue sharingunder this program. Funds received by the district government underthis program were earmarked solely for education and health: 74 per-cent for primary education and 26 percent for local health care (see fig-ure 9-3).

The Situado Fiscal was introduced to help reduce disparities, particu-larlv in health and education services, between regions in Colombia.The formula for distribution favored the most rural regions: 30 percentof the fund was distributed equally to all departments (states) and theother 70 percent was based on 1964 census population shares. This obvi-

URBAN GOVERNMENT AND FINANCES 251

Figure 9-3. Distribution of Situado Fiscal to the Bogota District

Total national current revenues(net other tax-sharing transfers)

85% reveniues 15% set asidefor general btidget for Situado Level I

expenditure Fiscal

70% 30%

I l Level 2

7.75% 3.03%( Bogota) (Bogota)

74% 26%Ministry Ministry Level 3

of Education of Hlealth

.1 Fond(o Senricios

Educativo Seccionales Level 4Regional de Saltid

IPrimary Localeducation health

cost services

Source: Ruiz Betancourt (1978).

252 UNDFRS'1ANDIN(; THE DEVELOPING MFTROPOLIS

ouLSy acted against the interests of Bogota. Originally 6.5 percent ofnational finds was destined for the Situado Fiscal; the amotint in-creasedl to 15 percent in 1975.

The national sales tax was started in 1965 as an emergency tax but hasstayed on the books. It is a general levy on the sale of all goods and ser-vices except for food, druigs, and a few other basic items. It contribtitedabout 15 percent of total nationial reveniues in the late 1970s. Since1971, 30 percenlt of total revenues from this tax have been shared withstibnational governmenits. The sharing formtila is the same as for theSituado Fiscal. (See figtire 9-4 for furthier- sharing agreements.) Ten per-cent of Bogota's shares is channeled throtigh the Ministry of Health for-local health progr-amiis, and 13.1 percent goes through the Ministry ofEducationi for local education progr-ams. Of the remaininig 77 percent,half is assigned to the FER, with the district government receiving only38.5 percent of the subvention for its general revenues.

A detailed analysis of' the acttial revenute sharinig unider both the Situ-ado Fiscal and sales-tax-sharing programs found that the actual distribut-tionis were less than the statecd objectives of these programs (see Fawcett1984). Supplemental grants wvere given on a discretionary basis over theyear in addition to the formula-based appropriationl. Thle result wasthat, althotigh the program was stIccessftilly redistributive il the begin-ni ng, the supplemiletital appropriations dilutedc the original intenitionlover the years. Bogotai was clearly a beneficiary of this trend, althotigh itstill received less per capita tlhan other r egions of the counitry in the late1970s. This was as intenided, btit the process illustrates the difficulty ofdevisinig fiscal programs to benefit poorer areas and to take aid awavfromii the better-off large cities. The political and bureaucratic pow er oflarge cities tends to dilute such programs. Thle district governimielit feltthat the revenue-sharing arrangemenits biased against Bogoti wereunftair.

The beer tax has been an importanit souirce of' tax revenue for the dis-trict. Begutn as a local district tax in the 1920s, it was levied at a specificrate, which had to be iicr-easecl successively. In 1966 it was fixed at 60percenit, thieni lowered to 48 percent in 1971 (8 percent as a national taxand 40 percenit as a departmental tax). The tax applies to the producerPlrice and, like most excisc taxes, is collected directly fi-om the ptocduc-ers-in this case, the breweries. The "national" portion, 8 percelit, waspassed on directiv to the local social security administiationi in eachdepartmetit for health-related expenditures, without entering thenationial budget. The remaininig 40 percent departmental portioni wasdistributed according to consumptionl; the distributionl formula basedon both consumption and distributioni clata was reported to the natiolialtax administration by the departiment governments (see figure 9-5).BogoI.i hadl the highest per capita bcer consuimptioni in the country,

URBAN GOVERNMENT AND FINANCES 25.3

Figure 94. Distribution of Sales Tax Revenue to the Bogota District,1975-80

Total nationial sales tax revenues(100%)

Level 1 70% revenue 30% set asidefor general budget for shared

expenditLire sales tax

Level 2 70% based on 30% divided1964 population as equal shares

7.74% 4.35%(Bogota) (Bogotai)

1<~~~~~~~~~~~113.1% 76.9% 10% C.P.S.

Level 3 Ministry District/departmenital (through Ministryof Education allocation of Health)

38.45%Earmarked for

Level 4 nationalizationi(50% of 76.9%)

38.45% DistrictDistrict receipt C.P.S.(50% of 76.9%)

DistrictFondo District

Educativo treasurvRegional

Source: Caceres Bolanos (1978).

254 UNDERSTAND)ING THE DEVELOPING MFTROPOIIS

and so it received almost one-quarter of the proceeds from the beer tax.Because this tax is not earmnarked (althiough it is supposed to be devotedto euitcation-related expenditures), it is seen as a most useful tax by thegovernmenit. It has also been a very buoyanlt one. Over the vears, thecontribution of beer tax revenues to district government revenues, notincludinlg decentralized agencies, increased from about 15 percent to40 percenit between the early 1960s and late 1970s. The real annualgrowth rate was more than 9 percent over this period. Similar taxes lev-ied on tobacco and liquor have not been important sources of r-evenue.

The tax on gasoline was discussed in the last chapter. Because the rateof tax was fixed at Col$0.06 per galloni over the whole period, reveniuesfrom gasoline taxes declined in importance. The revenues from this taxwere distributed according to consumptioni. which meant that Bogotareceived mnore than one-quarter of the national proceeds from it.

In summary, althougli revenue sharing from the nationial governrilenithas not been as importalit for Bogoti as it is for most cities in develop-ing counttries, the arrangements for tax sharing are quite complex andhave a long history in the Colombian fiscal system. The beer tax ilitIs-

Figure 9-5. Collection and Distribution of Colombian Beer TaxRevenue

Beer soldBeer proildicer to constlinelespavs nationial imicludes taxa(icl local tax in price per bottle

District /departmenital Natiorial bt er taxtax of 40%7 distribultedl of 8C% directlvbased Oil C0I1SUMiptiOll reiilitte(i tO local Servicios

Seccioilales (le Salucl\ ~~~~~~(local health agencN)

L/ Otlhet 23 Dist rict treasulrv

clepartniieiits rc-ccives 20-25f'vrecehc reinauinlg of total70-75'/7c of total

.Source.-Based oni intrlIvsiewsivith (listric t adiiiiinistraDoZ officials.

LIRBAN GOVERNMENT ANt) FINANCES 255

trates how important even a single-source tax can be: in this context itwas clear that mitch more could be collected from motor-vehicle-relatedtaxes than has been the case in BogotA. Not only are both beer and vehicle-related taxes easy to collect, but the tax base is highly income- and pop-ulation-elastic. These taxes are ther efore vervy useful sotirces of revenuefor fast-expanding cities in developilng countries.

The Structure of Expenditures

Expenditures on general administrationi have generally varied betveen6 and 10 percent of total consolidated expenditure in the district: thereseems to have been a slight increasing tendency. The average share ofgeneral administration) expenditture increased from 6.35 percent in the1960s to about 6.95 percent in the 1970s. There was, however, large vari-ation betweenl different vears, largely related to cliffering inflation andreflecting lags in wage adjustments. The share of the maini decentralizedagencies on expenclitures relatecd to water and sanitation, electricity,and telephontes varied between 50 and 60 percent of total consolidateddistrict expendittures throughotut the 1960s and 1970s (see figure 9-6),with considerable variation between the agencies.

tUntil fthc micl-1960s, expendituires on power predominated, amotulnt-ing to about one-thir-d of district expendittires. The share of power hadFallen to 10 to 15 percent in the early 197(s but rose to about 25 percentin 1979. Obviously, these movemetnts are related to the investmentcycles of specific projects, which are tylpically bunched with constructionof large power projects. Water and sewerage expendittires were around10 percent in the early 1960s; they increased to about 15 to 18 percentin thle late 1970s (see table 9-5). These increases in the share of waterand sewerage svstems and telephones have clearly been reflected in therelatively good availability of these services in BogotA.

Debt service fluctuated between 14 and 20 percent over the twodecadles, exhibiting n1o trend. Each indixidual agency is prohibited bystatute from exceeding a debt service ratio of 20 percent. Some of theagencies have reached this limit from time to time, but the districtadministration has always beenl substanitially below it.

The expenditures on education show a clear increase in the 1970scompared with the previous decade. 2) (The higher education levels dis-cussed in chapter 5 reflect these expenditures. Thie increase in school-ing in BogotA was palpable in the 1970s and was clearly a considerableachievemenit foi- the district.) Althoughi the financinig of these expendi-tures is now directly controlled by the centr-al government, the districtretained important responsibilities in the running of schools and hastherefore contintued to take considerable interest in the devoluition of

2 56 UNDERSTANDING THE DEVELOPING METROPOLIS

Figure 9-6. Composition of the Bogoti District's Total ExpendituresPer Capita

Percent of total

80

60

40

Other decenitralized agen(cies

20

1961 1963 1965 1967 1969 1971 1973 1975 1977 1979

Year

&Smrce: Linn (1 98 4a).

education funds. One issue of interest in this context is that the sub-ventions to the FER can, by law, be applied only to teachers' salaries. Thefinancing of school equipment, teaching materials, textbooks, and soon relies on extremely limited resources, and provision of these itemssuffered.

Another noteworthy feature in the structure of expenditures inBogota was the wide fluctuation in spending on roads and bridges. Asalready discussed, a significant portion of infrastructure expenditureson roads and bridges came from valorization, and this is reflected inatypically high expenditure between 1966 and 1970 (see table 9-5 andfigure 9-7). This was also the period when a conscious effort was madeto further the development process of the country through urban hous-ing and infrastructure construction programs. The city's road systemwas significantly upgraded in both quality and coverage. Major new traf-fic management schemes, including the building of overpasses andtinderpasses in the central areas, were undertaken. Apart from the par-ticular circumstances related to the history of valorization in Bogoti,expenditures on roads and bridges are generally related to the rate ofexpansion of a city. The effervescent growth of Bogota took place in the1960s into the earlv 1970s. One may therefore expect a rising pattern of

Table 9-5. Structure of Consolidated Total Expenditures: BogotA, 1961-79(percernt)

(Caleg~r) 1961 l 9%2 1963 1964 1 965 1906 1%'t 1968 19%9 1970 1971 1972 1973 197J4 1975 1976 1977 1978 1979

General admirristiationi 6.4 5.1 6.0) 5.5 5.8 5.8 4.7 6.3 9.4 8.5 7.3 8.6 6.5 10.9 6./) 6.7 9.2 7.9 6.4Police and prisonis 0.9 0.8 (.9 0).9 1.0 1.0 0.8 0.7 0.6 0.5 0.5 0.5 0.5 0.5 0.5 0.4 0.5 0.4 0.5Fire protection 0.3 0.2 0.2 0.2 0.2 0.2 0.2 0.1 0.2 0.2 0.2 0.2 0.2 0.1 0.2 0.3 0.2 0.2 0.2Roads and bridges 4.4 3.4 2.7 4.7 4.9 6.4 14.6 17.0 9.9 7.1 6.1 4.5 5.2 8.1 5.1 7.0 6.7 4.5 4.9Parks and recreationi 0.2 0.2 0.2 0.2 0.3 0.3 I.0 1.3 0. 9 0.5 0.3 0.5 ().8 0.7 0.8 0.8 0.5 0.6 0.4Educatiorn 6.4 5.9 6.0 6.5 7.6 8.8 7.5 5.7 6.3 10.6 11.2 11.5 1().7 9.8 10.7 10.5 5.1 1.8 1.2Hlealth 2.1 1.7 1.5 1.8 1.9 2.1 1.5 1.4 1.3 1.4 1.(i 1.5 1.5 1.4 1.2 1.2 1.3 1.1 1.3Tr ansportation 4.3 3.2 3.0 3.) 3.1 2.8 2.2 5.2 2.0 3.7 3. 4 2.0 1.7 2.2 5.0 1.9 1.6 1.8 0.6Welfarc 1.6 1.8 2.1 1.3 0.8 0. 9 0.7 0.7 1.0 1.3 2.6 2.7 2.6 2.4 1.1 1.8 2.0 1.7 1.7Water anll( sewerage 9).1 9.1 9.7 1(0.7 11.1 13.2 13.9 16.8 23.f 622.5 16.0 19.7 16.6 13.9 19.0 18.0 17.5 15.7 17.0Other saniitationi 2.1 2.2 4.1 6.2 3.2 3.1 2.6 2.7 2.5 3.3 3.8 3.6 3.4 3i.9 2.8 3.0 3.4 2.9 3..3Electricitv 34.8 :4.2 37.5 32.9 27.7 18.3 13.2 15.1 15.8 12.4 14.1 12.7 16.2 1(0.7 15.7 15.3 16.4 23.6 24.3Teleplonmes 5.9 8.2 6.4 5.3 6.9 10.8 15.9 7A. 8.7 10.0 1(0.4 12.2 9.9 14.1 11.5 12.4 12.4 15.6 19.3Housinig 2.0) 1.5 (.9) 2.1 2.1 2.6 ().8 1.4 1.0 0.9 1.) ().6 (.). 1.2 1.6 2.1 2.1 2.1 0.()[)istrict social sCcuritv 3.8 4.9 4.1 4.1 5.0 1.() 4.6 3.1 2.9 3.2 :3 7 .6 3.9 3.8 3.8 3. 7 3.9 4.3 I.6[)ebt service 15.5 17.4 14.8 13.9 18.6 19.7 15.7 15.2 14.0 13.9 17.7 15.8 19.8 1f6.4 14.8 11.8 17.0 16.t) 17.3Transfer to outside

distr-ict (1(1 0)() 0.0 ((.0 0.0 (0) 0.0 ( ( ()) 0.0 0. 0 ) 0 1.0 (o.) ) 0. (( 0.() 0.0 0.0

Tota1 100(.0 I(X).0 l(X).() 1 (1.0 1(X).() 10).0 I )t).( 10(.).) 1000 1(X)) 100.))10)1.0 11()).() 1100.1.0 1(X).() 1)1O.0 1(0).0 100.0) 1(X0.0

oer: ( Caregories arrangedl is shown in Bogota b1)cdget.Source: Linn (1984b).

258 LUNDERSTANDING THE DEVELOPING METROPOLIS

Figure 9-7. Composition of the Bogoti District's Capital Expenditures

Perceln of total

100

80

60

40Other deceint ralizedl

ageic-ies

20

Districad(lini istrato

01961 1963 1965 1967 1969 1971 1973 1975 1977 1979

Yea r

Source: Linri (1984a).

expenditures on such infrastructure to be related to city expansion. Val-orization-related expenditures have proxided Bogota with a good roadsystem that provides relatively good access to almost all parts of the city.Unlike many other cities in developing countries, even poor neighbor-hoods are generally well served by roads. This was also reflected in thealmost uniiversal accessibility of bus transport.

One interesting feattire of the infrastructure system in Bogota,although small in terms of expenditure, is the system of communityaction (acci6n communal) grants. Funds for this program come partlyfrom the national Ministry of the Interior. The district's CommunityAction Administration (Departamento Admiinistrativo de Acci6n Com-munal) channlels these funlds to six hundred community action groups.The distiict governmlienit is required by law to carry out small publicworks projects if the community collects a certain percentage of thecosts in advance. A number of small projects specifically demanded bythe commutity get implemented under this program. Another provi-sion through which such projects can be executed is that of auxilios par-lamentarios (parliamentary grants), which are made on the initiative ofindividual councillors who determine their tise. These auxilios can beused, for example, to fund minor public works, the neighborhood com-mittees, and scholarships. Typically the District Council adds thesefunds to the district budget: in 1980 they amounted to about 1.5 percentof the budget (excluding decentralized agencies). This system of direct

URBAN GOVERNMENT AND FINANCES 259

response to the demands of the community does much to help forge alink between the people and the district government. The prihciple ofmatching grants in response to the community's own raisinig of ftunds isin itself a good mechanism for indtucing community action.

The structure of capital expenditures illustrated in figure 9-7 showsthe importance of the main decentralized agencies in infrastructureinvestment in Bogota. The dip in the late 1960s and in 1974 is almostfully accounted for by the valorization-related expenditures made by IDIU

(induced in other decentralized agencies in figure 9-7). Overall, BogotAhas been served relatively well by its local governments through the(lecentralized agencies, which are discussed in the following section.

The Public Service Enterprises of Bogoti

The three main public utilities of Rogota-the water and sewerage, elec-tricity, and telephone companies-were all originally founded as privateentities in the latter part of the nineteenith centurv. Each has a historyof gradual takeover by the government. The water company wasfounded in 1888, acquired by the city in 1914, and given its current sta-tus as a decentralized autonomous agency, the EAAB, by the city councilin 1954. The power company began operating in 1892: it later mergeclwith the Conipafiia Nacional de Electricidad, foundted in 1923, and thecity acquired a 51 percenit interest. The merged company was fully takenover by the city in 1951. when it became known as the EEEB. The tele-phorne company, founded in 1884, was taken over in 1940 and has oper-ated as the ETB ever since. By contrast, BogotA's public service company,the EDIS, was government-run from the start. Founded by the district in1958 as a refuse-collection and street-cleaning company, it expanded in1960 to cover functions like administerinig the municipal slaughter-house, running markets, and maintaining cemeteries.

Each of these decentr-alized agencies has a similar governing boardconsisting of six or seven members and presided over by the mayor. Twomembers of the board of each agency are nominated by the DistrictCouncil. The other four members of the E.AB board are nomniated bythe main bond holders of the company-the Banco Central Hipote-cario and the Colombian central bank, the Banco de la Repilblica. TheEEEB has one member nominated by the president of the republic andone each nominated bv different associations-the manufacturers, themerchants, and the banikers. The ETB has the personero (district legalcounsel) on its board, along with two additional members nominated bythe mayor and approved by the District Council. The EDIS also has thepersonero on its board, along with two ex-councillors chosen by the Bancode la Repi[blica and two members recommended by the city banking

2fiO UNDERSTANDING THE t)EVEI.OPING NIETROPOLIS

associations. In each case the board appoints the general manager, whois in charge of all the company s operations. The board, however, has toapprove all major decisions involving, for example, the budget, invest-ment plans, changes in tariffs, and changes in administrative structuire.

Each of these public utility agencies is a relatively large entity andsaw major expansions during the 1960s and 1970s. Bv 1972 the EAAB

had about 2,300 employees, the EEEB about 1,600, the ETB over 2,000,and the EDIS over 3,000. Each had more than doubled its staff size dur-ing the 1960s, but the output per employee generally increased at thesame time.

The service levels achieved were generally satisfactory by the late1970s. Over 90 percent of the population received water, supplied bythe EAAB. Per capita consumption of electricity grew by about 6 percentannually between the early 1960s and late 1970s; almost all householdshad a power connectioni by 1978. Similarly, the EDIS provided more than95 percent of the city with street cleaning and refuse collection, a levelachieved through continual expansion and upgrading of equipment.As manv as 60 percent of houiseholds reported access to a telephonein 1978.

The autoniomy of these public utilities has been useful in insulatingthem from day-to-dav political pressures on their operations. (;ivcn thepresence of the mayor and District Council members on each board,the agenicies are obviously not immune to political pressures. However,the administiationi of these enterprises is left largely to professionalmanagers and technical personnel more removed from political inter-ference than are the staff of the district administration itself. The enter-prises have succeeded in maintaining a high degree of professiolialcompetence and technical expertisc and in keeping a stable and experi-enced staff. They have had broad financial autonomy and remained sol-vent through most of the period under study. Thiey have essentially beenself-financinig, borrowing for capital expenditures and financinig theborrowing by levying overcharges.

The SYstem of User Charges

The tariff rates of all public services in Colombia are r egulated by theNational Tariff Board. Each enterprise has to follow the guidelines setby that board, which must approve tariff changes. In principle, all agen-cies apply a long-run average-cost pricing criterion: operating costs anddebt charges that finance past borrowing for capital expenditures aresupposed to be covered from operating revenues. Capital expendittiresshould be covered by new borrowing anid by connection charges. Inaddition, equity considerations have been seen as very important in tar-iff setting in Colombia. Most public service tariffs in Colombia have an

URBAN GOVERNMENT AND FINANCES 261

equity element built into them: tariff charges paid by more affluent cus-tomers are higher, being related to either property values or volume ofconsumption. Changes in the tariff pattern, guided by these broad prin-ciples, were frequent during the 1960s and 1970s.

Most user charges are divided into two segments: monthly charges forservices consumed and connection charges payable on initial connec-tion of the service. The monthly charges go into the current revenues ofthe enterprise, whereas the connection charges are capital contribu-tions financing capital expenditures. Monthly charges for water supplywere themselves divided into two parts. One was a lump-sum chargerelated to a minimum allowance of water supply per connection and theother was a unit charge for consumption beyond the minimum allow-ance. This lump-sum charge was graduated according to seven classes ofproperty values. The minimum allowance was 15 cubic meters permonth and uniform across all property classes. Prior to 1970, the chargewas graduated according to twenty-six property classes and the minii-mum allowance itself ranged from 20 to 60 cubic meters per month.Consumption beyond the minimum allowance was levied a unit charge,which itself increased according to property value as well as progres-sively with higher consumption. Residential rates at the lower ends werelower than the charges for commercial and industrial use. For higherconsumption levels at high property values, however, residential chargeswere higher than charges for commercial and industrial use. For indus-trial consumers, the minimum allowance was abolished in 1970. All tar-iff rates increased in step jumps in 1961, 1965, 1968, 1970, 1973, and1974. The result was that the average tariff rate (total revenue divided bytotal sales) in real terms looked like a ratchet pattern, with stepincreases at the time of tariff revision followed by a decline until thenext revision. In the late 1970s small monthly marginal adjustmentsreplaced this pattern. The progressiveness of rates increased over time.The ratio between the highest and lowest increased from just about 6 inthe 1961-65 period to over 40 in the late 197 0s, if the rates for the mini-mum allowance are accounted for. For total water consumed, this ratioin the late 1970s was about 10. Almost all connections consumed morethan 15 cubic meters per month, or the minimum allowance; this factsuggests the minimum allowance was less than actual minimal house-hold requirements. In other Colombian cities, the minimum allowancewas 25 cubic meters per month.

In addition to the monthly charges, the EAAB levied rather stiff con-nection charges. The connection charge itself had four elements: a con-nection fee set at 3 percent of the property's assessed cadastral value; a"supply fee" set at a flat rate per square meter of developed land on theproperty (with a lower rate for certain specified low-income areas); anetwork fee levied as a capital contribution for the construction of the

262 UINDERSTANDING THE DEVELOPING METROPOLIS

secondary distribution network designed to connect a new develop-ment; and a fee to cover actual installation charges (this also variedaccording to property value).

Sewerage charges were essentially linked to the water charges becauseit is difficult to charge according to use. The monthly charge variedfrom 5 to 20 percent of the water charge depending on property value.The capital charges for sewerage had two components: the networkcharge analogous to the water network charge, and valorization chargeslevied by the IDU to finance large-scale sewerage projects.

The EAAB made a great effort to collect user charges. Given the differ-ent objectives of levying user charges-the financinig of both operatingand capital costs on the one hand and promoting equity in order toimprove access to basic public services on the other-the resulting tariffstructure was quite complex. The tariff structure looked like a largematrix rather than a rate. It turned out that the capital contributionswere higher thian most poor families could afford, despite the progres-sive structure. Even if they paid in installments spread out over three orfour years, a low-income family woul(d have had to pay 9 to 10 percent ofits income for the water supply connection alone. This led to many ille-gal connections. Over the years the share of capital revenues increased,whereas the share of capital expenditures fell. As a result, during themid- to late 1970s the capital reventues were helping to fund operationalexpenditures.

The structuire of electricitv pricing was guided by similar principles.The ain was to cover the long-run average cost of producing energywhile incorporating redistributive criteria in the tariff' structure. Resi-dential areas were charged a unit rate in two slabs; the ratc increasedwhen tisers went beyond a certain level of montlhly consumption. In ear-lier years there were a larger number of' such slabs and a monthly metercharge. Industrial and commercial users were levied higher rates, withsrnaller firms receiving preferential rates. Some elements of efficiencypricing were iitrodtuced through time-of-day variations with higherpeak-period charges. Installation fcees were charged according to capac-itrv installe(d and varied between residential, commercial, andl industrialuses.

Electricity connlection charges were nowhere near as onerous as thosefor water supplv. An analysis of actual collections showed that commer-cial consunmer-s paid thie highest charges, followed by industiial consuim-ers, and then residential consumers. Residential consumers paid justabout the same rates as those paid for public lighting. An element ofcross-subsidy was therefore built in on the principle of' apparent abilityto pay. The actual incidence was difficult to calculate because the indus-trial and commercial consumers wotuld presumablyI pass on their higher

t RBAN GOVERNMENJ AND FINANCES 263

charges. Clearlv, the r ate differences were not based on the relative costsof supply to different users.

Unlike the EAAB, the EEEB's revenue structure moved from larger cap-ital reveinues in the early 1960s to greater current revenues in the 1970s.The structure of expenditures moved in a broadiv consistent direction,

and the surplus from curr-ent revenues helped finanice capital expendi-tures througlhout the whole period.

The tariff strticture for the telephone company, the ETB, was alsosupposed to cover long-run average costs while providing for internalcross- subsidies for different income groups. It is probably unusual fortelephonie tariff rates to vary accordinig to property values. The maincomponenit of the tariff structure was the basic monthly rate, which waslcvied according to six residenitial property valuc categories for residen-tial areas and four registere(d asset value categories for indtistrial andcommercial users. The difference between the lowest and highest rateswidened durinig the 1960s and 1970s, but all had fallen behindc the rateof inflation. The capital contributioni was collected in four segments: alump sum, a returnable security deposit, and an installation charge. allof which were cqual for all users, and a connection fee that rose steeplywith property value. These rates were revised fi-cqtuently to keep up withinflation. Thlic ETs financial position was generally healthy, with cur-rent surpluses contributing to capital expendittires throughout the1960s and 1970s.

Service charges for garbage collection by the EDIS were also leviedaccording to property value assessed at a constant ad valorem rate basedon the capital value of properties. Low-value pr<operties were exemptedfrom any charge. The EDIS also charged for its other operations.

This summar-v of user charges shows the importance attachled tocross-subsidizationi in providing urbani services in Colombia-a politicalsystem not otherwise known for- its concern for the poor. The govern-me(nt s considerable effort to make basic urbani services accessible to thepoor was partly the result of the self-financinig and fragmented naturie ofurbani service enterprises in Bogota. Because the enterprises had to relyon their own finances and ther e was no possibility of cross-subsidizationbetweell utilities, there had to be stronger cross-subsidization acrossincomiie groups. In Cali, where all public services arc uniifiecd in one pub-lic utility, E:mpresas Municipales dc Cali (EFMCA.I), there was significantcross-subsidlizationi between the different services.

For almost all the titilities, average Unlit costs increased in real termsover time. This implied that Unlit tiser charges shoulci have beenadjusted more than inflation throughout the period and( that marginialcost exceeded average cost: utilitv pricing based on marginal costswtould have more than finariced the expanding network of services.

Table 96. Subsidies from Public Utility Pricing: Bogoti, 1974

Water and sewerage Electricity Telephotnes Garbage collection Total

.Subsidy Subsidy Subsidy Subsidy Subsidy Subsidy Subsiuy Subsidy Subsidy Subsidy

(Col$ per (percent of (Col$ per (percent of (Col$ per (percent of (Col$ per (percent of (Coas per (percent of

Population decile month) income) month) income) month) income) month) incomen) month) income)

0-10( - - 1.13 0.18 - - - - 1.13 0.1810-20 - - 4.66 0.40 - - - - 4.66 0.40

20-30 - - 5.07 0.34 - - - - 5.07 0.34

30-40 16.72 0.91 5.37 0.29 - - 13.75 0.75 25.84 1.9540-50 16.48 0.72 6.72 0.29 - - 11.12 0.49 34.22 1.4750-60 14.10 0.49 6.63 0.23 2.48 0.09 10.72 0.38 33.93 1.1860-70 14.10 0.38 5.78 0.15 2.96 0.08 3.21 0.09 26.06 0.7070-80 9.02 0.16 4.28 0.07 2.14 0.04 (1.74) (0.03) 13.70 0.2480-90 (15.23) (0.18) (8.92) (0.10) (2.05) (0.02) (9.86) (0.12) (36.07) (0.42)90-1(0 (147.65) (0.82) (20.33) (0.11) (9.57) (0.05) (44.94) (0.25) (222.50) (1.24)

Gini coefficients

(without charges.0.5103) 0.5098 0.5074 0.5102 0.5097 0.5070

- Not applicahle.Note: Figures in parentheses are negative subsidies.Source: Linn (1 976).

URBAN GOVERNMEN r AND FINANCES 26Y5

How far did the attempts at cross-subsidization succeed? Linn (1976)reported a 1974 study conducted for the National Tariff Board that

showed that the top two income deciles of the population essentiallysubsidized the rest of the population. The main beneficiaries were thethird to sixth deciles, who were estimated to gain I to 2 percent of theirmonthly income through these cross-subsidies. The lowest incomegroups also gained, but not as significanitly (see table 9-6). This may beexpected; it was observed in chapter 6 that the lowest three deciles alsodid not benefit as much from pirata housing developnment. It is simplytoo difficult to pass on public subsidies to the bottom three deciles.Where a significant proportion of households in the bottom threedeciles are not homeowners and are rnot connected to the public servicesystem, a graduated tariff structurje cannot hope to succeed in redistrib-uLting effective income.

Overall, the existing system of user charges enabled the three mainpublic service enterprises in Bogota to operate throughout the 1960sand 1970s on a financially self-sufficienit basis without receiving or con-tributing funds to the district government. There was a mild t-end ofincreasing real unit costs for each of the agencies, and henice thev raninto difficulties whenever inflation ran ahead of their tariff adjustments.The EAAB was most affected and had to resort to occasional borrowingin order to finance operating delicits as they evolved. In the late 1970ssmall monthly charges in tariff rates began to be made for water andsewerage as well as electricity in order to alleviate the problemii caused byinflationi.

How efficient was this system of user charges? It was clearly niot based

on marginal cost considerations. Hence, for different services, the unitrates differed quite markedly from marginal costs. For electricity, themarginal cost was higher than the average tariff, whereas for water it waslower. There were other inefficiencies, too. Electricity charges for indus-trial and commercial users were higher than for resiclential consumne rs,when the relative marginal costs of supplv would be in the oppositedirection. It is difficult to evaluate, however, how impor-tant these iieffi-ciencies were in comparison with the perceived gains from redistribu-tive tariffs in as unequal a society as Colombia. The driving force behindthe tariff structure in Bogotav was the need for financial self-sufficiency.However, the same ain could have been achieved with greater consider-ation to marginal cost pricing and a different division betweeni marginaltinit consumption charges and fixed monthly charges on the one handand connection charges on the other. Progressive connectioni chargescan be supported strongly: low-inconie households should not be dis-cour-aged from gaining access to these basic public services.

266 tlNDERSTANDING THE DEVELOPING Mt:TROPOLIS

Lessons for a Developing Metropolis: An Evaluative Summary

Bogota has long had a rathier elaborate system of urban government,one that has been supported by public service institutions that also havea irelativelv long history. Yet Bogota's government has been more of afollower than a leader in guiding the growth of the city. This was at leastpartly because of the effervescent growth that the city experienced fromthe early 1950s to the early 19 70s. Few governments in the world couldhave led such growth. The spatial developmernt of the city has beendominated more by the rapid development of pirate settlements than bylegallv zoned formal housing developments. It is creditable, however,that public infrastructure was largely able to keep up with this growthand even marginally improve on per capita service levels while certainlyimproving the coverage of the expanding urban population.

This was accomplished with negligible infusions of f'unds from out-side sources-except for international borrowving that was sustained bymanageable and prudent levels of debt servicing. Bogota and other cit-ies in Colombia have been unusual in the extent to which urbani publicservices have been financed by user charges. Some degree of cross-subsidy between income groups reduced the burden on the poor, butthey still had to pay substantial charges for the installation and use of'public services, often forcing the poorest among them to resort to ille-gal tapping of these services. To the extent that each agency was finan-cially solvent, this illegal tapping effectively amounted to a large subsidyfrom the better-off'users to the poor. In many other cities in developingcountries, urban public services are designed to subsidize the poor, but

the main beneficiaries are the better off. In these cases powerful vestedinterests then succeed in preventing future increases in user charges.The resuilt is that the expansionl of public services is limited by paucity of'resources, and the most severely affected are the poor. It is thereforevery important to develop sound user-charge-based systems of financinigurban public services in order to fun1d the ever increasing needs of'growing cities.

It has been observed that there is no systematic attempt at urban in-vestment planning in the medium term. Zoning and building regula-tions exist but are frequently breached without fear of effective sanctions.We have also found that accessibility and variety of ur-ban transportpr-ovision impr-oved over the period studied. How did all this occur? Par-adoxically, the very decentralized nature of urban public service provi-sion in Bogota made it possible for the government to cope with veryrapid city growth. By design or accident, Bogota's government and pub-lic service agencies, as well as the private sector providers of service,have all responded well to expressed demands. This has enabled the un-buindlinig of services that are often a prohibitivelv expensive package inother cities in developing countries. The key to successful unburndlirig

URBAN GOVERNMENT AND FINANCES 267

has been a very effective sharing of costs among all concerned. The ur-ban government system was able to adopt the incremental constructionparadigm of informal housing because of the autonomy of the publicservice enterprises, which in turn enabled greater flexibility in serviceprovision than is generally possible. The valorization and tiser chargesystems institutionalized the idea of cost sharing between public serviceagencies and the beneficiary public. The private sector contributed bytaking part in supplying transportation. The government further con-tributed through regulation and tariff-setting policies that made theunit costs of services progressive, making it easier for the poor to obtainaccess to basic services.

It is argued that, paradoxically, the fragmented nature of decisionmaking in Bogota and its lack of centralized and effective urbani invest-ment planning have contributed to the development of a relatively effi-cient public service delivery system. It is probably the case that theneeds of a rapidly growing large city are intrinsically difficult to serve ifan attempt is made to meet these needs in a plannied and coordinatedfashion. It is probably easier for autonomous agencies, which are forcedto be financiallv independent, to perceive the growing demand for ser-vices and to respond incrementally. Such a decentralized system lendsitself to financial self-sufficiency, achieved through the levy and collec-tion of user charges. Excessive reliance oni usually scarce budgetaryresources only leads to inadequate growth in the supply of urban publicservices.

The key lesson to be learned here is that much can be gained byobserving people's preferences and devising systems of government thatare able to respond to expressed demands. In Bogoti this was done byinstituting a fragmented system of city government. In other cities therecould be a stronger system of community developmnent, which wotuldalso be effective in making perceived demands felt but perhaps in amore organized manner than in BogotA. There is, however, consider-able room for improvement in Bogota. Although the pricing system per-formed well in financing the public service enterprises, it could beimproved in structure to better reflect efficienlcy considerations withoutlosing its effectiveness as an adequate resource-raising mechanism. Theenterprises could improve their performance further bv a little moreforward investment btidgeting and planning-but not in such a wav thatwould reduce flexibility and become dysfunctional. The valorization sys-tem had fallen on hard times: it could be revived to finance furthelimprovement of the infrastructure. Much more needs to be dtone as wellto improve the qualitv of education and health service in the poorerneighborhoods. Because BogotA is one of the more spatially differeniti-ated cities by income group, a more directed effort will have to be madeto provide services and upgrade neighborhoods in order to integratethe city into an organic whole.

Table 9A-1. Per Capitat Revenues of District Administration: Bogobt, 1960-80( 197(i Cxiionihtian pescis)

Rewe/ri tt, 19eil I .96 1 1962 / 963 1964 196.5 1966 1967, 946 1909 1970 1 971 l 972 J'973 1974 1 975 1976 1977 1978' 1979 1980

Loca i]/ X"venw'l 3-17.9 456.2 43'4.1 '349.8 3X15.3 328.4 2)57.5 290.3 28l\.2 303.1 314.f; 298.3 27fi.4 276.7 240n0 2l.1 '2fi9.7 2*6/.7 279.fi 317.7 36'2.6

Propev-t .& complemtent 19L'.7 258.0 259./ 2()9.5 19'>.7 I'J8.9 lfi3.8 1 89.4 1 72 1) 187.1 1 88.8 16;9.4 166(. 8 172.fi 1-14.() 158.5 141.5 108.7 122.5 121 tj 146.0

Indi(isti -s -comiillerte 411.6 11 1.0 94.() 79). fi5.2 7().1 44.6 51).7 48.fi 51.'9 59.3 56.4 Sl.9 58.3 5 1.9 83.7 88.2 95).9 1()1.2 13().4 13().3

Puiblic enter tainmefit ().0 1.8 0.4 0. () O.l) 0.0 0.1 O.Q 0.0 1 0.9 10.fi 11.1 8.4 G.!) 5.9 6.2 5.7 6.2 6.1 6.3 5.5

Advei tisemietit O.l 01.0 0.5 ().5 ().4 ().()4 /) 5 1.5 4.0 41.4 4.7 2.3 0.8 1.0 0 zl 0.9 0.5 ().6 ().4 ().4 ().3

*t hicle 3/.3 HU.7 3 8.1) 25.9 21.2 25.4 20.1 18.7 21.3 1fi.5 18.8 2l).() 14.7 1().1 11.7 9.l) 8.8 16.3 19.8 29.3 '25.0

Ui-ban cosu-ticticxn I I.(i 10.0 1 1.9 1().7- 14.2 1 2.3: 9.7 tl.'J 15.1 I-1.L' 15.2 12.5 10.3 1(1.6t 9.9 8.6 9.2 13.8 12.4 13.3 31.2

Puiblic i-oad IlSc' 0).6 ()) (1.8 0.8 ().7 o.G ().5 2.3 1.9 1.4 1.5 1.() 1.1 ().8 1.1 0.7 ().7 3.4 3.9 4.() 9.3

x Ra111les ().8 0.8 0.) ().7 ().9 ().9 l).'J 0.9 0.4 1.6 H. 1.0 1.1 0.5 ().fi 1.5 1.1 2.1 2.0 1.5 1.3S

getting 18.5 11.1 10.0 9.2 8.6C 9.5 8. 4 8.(; 8.!1 7.5 7.(i 90. 10.7 9.1 10.2 8.5 1().5 8.1 6.4 4.3 3.1I

Gasoline ($().()4'gaw1.) 14.8 17.9 1 7.t) 1'1.1 11.4 1().7 8.9 S.3 8.() 7.6 7.() 7.1 5.8 5.3 4.2 3.5 3t) 2.5 2.1 1.fi 1.3

Foi-eigii cigarettes ().() ().() 0.() ().l) ().0 ().0 ().0 ().l) ().() ().() ().() 8.5 4.S 1.5 ().1 0 t0 0.5 '3.1 2. 8 5.() 9. 3

Othcers 0.0 O.l O.l ().0 ().() ().() O.( O.() O.() ().l) l).O 0.() 0,.0 O.( (1.() ().(: ().(1 ().(1 ().( l)() ().(

Locat nosn-taxYevevule A 22.4 25 '3 2fi.4 21.() 21.5 25.5 8.6 1().3 12.2 'J.8 8.0 12.7 8.3 5.7 ll).8 8.8 6 ).9 7.4 10.4 21.3 43.9

Saniitatioii licenises 1.7 1.6 1.1 ().8 ().8 ().8 ().7 ().6 0.5 ().4 0.5 O.fi) 0.1 0.0 ().0 (1.0 ().() (.( ().0 0.( 0.()

NCowvnclahra urz6ana ().1 ().1 .(1 (1.() O.() O.() 0.0 0.0 0.1 0.1I (1.0 ().0 ().0 ().0 o.n o.o O.( O.() O.t) o.o o.o

Laboratoi v services 0.9 0.7 0.5 ().4 ().3 ().3 0).2 0.2 (1.3 0.2 0.1 0.1 0.0 0.0 (1.0 ().0 ().0 ().() ().0 0.( 0.()

Inteiest receipts 3.1 1().(1 1:3.fi 11.1 12.5 11.7 0.0 0.0 0.0 (1.0 ().0 ().2 ().(1 ().(1 0.1 1.8 1.3 6.2 4.3 10.8 .39.8

Fihies 7.9 6.4 fi. I 6f.1 5.9 5.9 5.9 7.2 9.3 7.2 1.6 1.7 4.2 i.9 .5.2 5.4 3.4 ().(1 4.5 9.4 0.()

6;onfiscated land sales ().1 ().1 ().1 0.1 0.() 0.0 0.0 0.() 0.2 0.1 ().1 ().2 ().1 ().1 ().( 0.() 0.(1 0.1 0.0 0.0 0.0

Reniial receipts 2.6 1.9 1.1 ().5 0.3: ().4 0.1 0.1 0.1 0.0 0.0 0.1 0.1 0.1 ().1I ().1 ().1 ().( 0.( 0.() 0.()

Real estate sales ().0 ().0 ().1 0.2 0.2 5.1 0.6f. 0.4 0.0 (0.0 ().0 ().2 ().0 ().(1 0.( 0.() 0.1 0.1 0.1 0.0 2.4

OLtiers 6.0 4.5 3.8 1.8 1.5 1.3 1.1 1.8 1.7 1.8 5.7 9.6 3.8 1.6 .5.4 1.5 2.0 I.() 1.5 1.1 1.7

Shared taxes 105.6 142.4 131.4 185.7 202.0 179.3 150.6 159.6 141.4 171.1 235.2 284.8 255.5 224.1 233.1 259.7 205.5 191.7 199.1 210.1 243.1Sales tax 0.0 0.0 0.0 0.0 0.0 0.0 11.1 0.7 0.0 6.6 13.9 31.1 36.4 21.5 37.7 62.6 31.0 17.6 28.8 16.1 23.3Beertax 84.7 94.1 82.2 145.4 139.6 123.5 116.1 122.8 113.2 133.3 186.0 204.6 177.2 150.2 148.8 157.9 138.1 138.6 138.1 161.0 185.7Gasoline ($0.06/gal.) 19.3 26.9 25.4 19.6 17.1 16.1 12.6 12.4 11.9 11.2 10.0 10.1 7.9 6.4 6.1 5.7 4.5 3.6 3.1 2.1 2.1Other 1.6 21.4 23.8 20.7 45.3 39.7 10.8 14.1 13.4 20.0 22.3 22.1 19.1 20.7 23.7 16.1 17.0 16.6 18.1 19.6 21.0Tobacco tax 0.0 0.0 0.0 0.0 0.0 0.0 ()() 9.6 2.9 0.0 3.0 16.9 14.9 25.3 16.8 17.4 14.9 15.3 11.0 11.3 11.0

Grants 25.1 10.9 31.9 45.7 42.8 38.1 43.0 31.8 40.7 34.5 73.3 80.0 69.1 86.1 82.0 117.8 97.4 0.0 0.0 0.0 1.8For teachers' salaries 4.6 5.9 29.8 38.2 42.0 37.2 42.3 31.6 26.5 28.5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ().0 0.0 0.0 0.0For regional educationi

fund 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 71.9 78.5 69.0 86.1 81.5 113.8 97.4 0.0 0.0 0.0 0.0For school cafeterias 0.3 0.4 0.3 0.2 0.2 0.3 0.3 0.2 0.2 0.2 0.0 ().0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 (1.0 0.0For health services 18.9 3.4 0.8 6.5 ().0 0.0 0.0 0.0 0.0 0.9 1.0 1.5 0.1 0.0 0.0 1.0 0.( 0.0 0.0 0.0 0.0For social welfare 0.0 0.0 0.0 0.0 0.0 0.0 0.0 (.0 0.0 ().0 0.1 (.0 0.0 ().() 0.0 0.0 0.0 (.( 0.0 0.0 0.0For economiic develop-

ment works 0.0 0.0 0.0 0.0 (1.0 0.0 0.0 0.0 14.0 4.9 0.3 0.0 0.0 0.0 0.() 0.0 0.0 0.0 0.0 0.0 0.0From decentraliLed

agencies 1.3 1.2 1.0 (0.8 0.6 0.6 0.4 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.5 3.0 ().0 0.0 0(. 0.0 ().0

Capital reuourres (debt) 75.0 5.1 4.5 0.0 (1.0 0.0 0.0 168.9 186.5 4.2 15.9 22.4 3.1 13.4 39.1 91.0 39.0 7.6 7.6 8.8 286.0Loanis 75.0 5.1 4.5 0.( 0.0 0.(1 0.0 58.0 11.3 4.2 15.9 17.7 3.1 13.4 39.1 91.0 39.() 7.6 7.6 8.7 216.7Bond issues 0.0 0.0 ().0 0.0 ().0 0.0 0.0 110.9 175.2 0.0 0.0 4.7 0.( 0.0 0.0 0.0 0.0 0.0 0.0 0.1 69.4

Total revenues 576.0 639.9 628.3 602.2 581.6 571.3 459.7 660.9 661.0 522.7 647.0 698.2 612.4 606.0 605.0 758.4 618.5 467.4 496.7 557.9 937.5

Source: Linn (1984b).

Table 9A-2. Consolidated Capital Expenditures: Bogotai, 1961-79(percentage distribution)

Revenue 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979

Genleraladministration 0.0 0.0 3.0 1.6 1.4 0.0 1.4 5.3 40.2 24.1 12.3 9.9 12.2 20.7 10.8 12.4 2.1 3.3 9.0

Police and prisons 0.0 0.( 1.2 0.0 (.0 (.( 0.0 0.0 (.0 (.( 0.( 0.0 0.0 1.5 0.( 0.0 0.0 (.0 1.8Fire protectioni 0.0 0.0 0.0 (.0 0.() 0.0 0.0 (.0 0.0 0.0 0.0 (.0 (.( 0.2 2.5 2.4 (.0 0.1 0.0Roads and bridges 48.6 38.5 24.2 56.6 50.1 73.3 235.3 323.6 2()1.4 117.0 110.1 76.0 92.6 173.2 98.8 132.1 97.7 69.7 60.1Parksand rea-ceation 0.0 0.6 0.1 0.3 0.3 0.4 17.9 27.2 21.3 11.0 6.5 4.5 4.1 4.0 6.9 8.8 3.3 2.3 4.3Educatioii 4.1 1.1 16.2 3.0 7.1 11.3 21.3 4.2 35.2 37.0 37.6 11.6 4.2 9.8 4.5 5.0 0.5 1.9 6.4hlealth 0.() 0.1 0.2 1.2 1.5 4.5 1.9 0.9 1.8 0.8 1.2 ().0 0.0 0.0 0.0 0.0 (1.0 0.( 0.0Transportation 0.0 2.1 0.5 0.0 (.0 0.0 0.0 68.9 0.0 32.3 33.1 0.5 (.0 19.3 99.( 6.2 0.0 (.1 1.5Welfare 0.0 0.6 6.3 0.9 1.7 0.0 0.0 (.9 6.2 15.9 48.3 51.2 4(0.6 40.0 8.7 18.6 16.1 10.8 7.4Water aid sewerage 96.4 109.1 126.4 110.6 93.6 11(0.4 153.6 240.9 463.8 402.7 277.3 370.1 225.2 180.9 335.1 293.5 223.3 222.1 151.7Other sanitation /).() 11.0 42.8 71.0 8.6 3.3 1.2 6.1 4.9 12.3 27.2 7.5 4.2 21.2 2.0 (.1 0(.( 0.( 1.5Electricitv 572.3 693.1 725.9 518.9 354.5 166.9 105.4 202.7 261.4 157.6 201.9 136.1 2336.1 99.9 196.6 131.7 88.1 120.7 57.5Telephones 53.5 111.3 75.7 46.0 57.2 109.3 228.8 89.5 126.2 141.2 171.6 217.9 135.5 258.4 199.1 191.5 136.2 182.7 191.6Housing 1.5.1 13.1 3.2 23.0 17.8 20(.6 4.0 18.2 16.2 12.9 15.1 6.5 4.2 27.7 39.1 45.9 36.1 41.8 ().()Distr-ict social

securitv 1.4 14.0 5.7 4.3 4.9 12.8 6.4 4.6 11.5 14.2 6.7 1.2 2(0.1 10.0 0.8 (1.2 (1.4 0.3 0.2Debt ser%ice 0.0 ().() 0.0 0.0 0.0 °0.0 0.0 () ° ° ()() o(1) 0 0(I (.0 ) 0. (.0 0.0 (°.0 0.°1Transfer to outside

district 0.( ().8 0.0 0.0 0.0 ().0 ()() 0.0 0.0 (1.0 0.0 0.() 0.0 0.( ().( 0.01 0.0 0.0 ().0

Total expenditures 791.4 984.4 1,031.4 837.4 598.7 512.8 777.2 993.0 1,1.0.1 979.() 948.9 893.6 779.0 866.8 1,003.9 848.4 6()3.8 655.8 493.11

Sou/rce: Linn (1984b)

URBAN GOVERNMENFNT ANI) FINANCES 271

Notes

This chapter is drawn from works b Linn (see bibliography) and from Con-traloria de Bogota D.E. (1982), Fawcett (1984), Greytak (1984), and Wolff(1984). The section on "The Structure of Revenue" summarizes researchreported in l.inn (1984a) and Fawcett (1984). The detailed work on consolidat-ing the finances of the district was conducted over a period of three years byAlberto Hernanidez, Marco Tulio Ruiz,Jeffrey Lewis, and Jose Fernando Pineda,directed by Johannes l.inn, and published in Contraloria de Bogota, D.E.(1982). The section on "The Public Service Enterprises of Bogota" is drawn firomresearch reported in L.inn (1976b) and Greytak (1984).

1. OCne bv-product of the City Studv was the compilation of the full series ofdistrict and agency revenues and expenditures for 1961-81. This informationwas subsequently published in the statistical bulletini of the conitroller of Bogota(Contraloria de BogotA. Bolelin EstadiAtico, vol. 1. no. 5, December 1982).

2. The fall in education expenditures after 1975, shown in table 9-5, reflectsthe channelinig of all education expendittires throtigh the FFR.

Chapter 10

Coping with City Growth

The aim of this study has been to improve our understandinig of the waycities in developing countries work. What have we learned from the var-ious detailed investigations that have been reported in earlier chapters?What are the messages that can be flashed to policymakers in develop-ing countries as they confront the daunting problems posed by therapid urbaniization that most developing countries are experiencing?

One of the broad themes of this book has been that an understandingof cities requires an appreciation of the behavior of economic agentswho r espond to economic signals in a relatively rational and predictableway. There are no easy or straightforward "policy recommendations"that can be generically prescribed for managing growth in the cities ofthe developing world. Differenit cities and countries exist in specific set-tings that need to be understood and analyzed case bv case. Differentcountr-ies also have quite differenit adminiistrative systems. Policymakersneed to observe and understand the rationiale behind the various phe-nomena observed in cities and to react accordingly. The integratingtheme of our study is that the var ious patterns observed in cities are a re-sult of relatively rational behavior onl the part of their various economicconstituents-*vorkers and their families, employers, government, hous-ing suppliers, and so on. The problems arising from rapid city growthare best dealt with by observing and understanding these patterns anddesigning flexible and self-correcting institutional frameworks that arecapable of respondinig to the variegated needs of the citv's compolnents.

Bogota and Cali emerge from this study in an optirnistic light. Wehave depicted them as having had fair success in withstanding the chal-lenges of sustained and extremely rapid growth over the past fifty or so

years. Much of this growth was concenitrated in the 1940s, 1950s, and1960s. W4e also found evidence during the 1970s of a significanit increase

272

COPING W'ITH (.ITY GROWTH 27 3

in household incomes and some decline in the high inequality that hadcharacterized Colombia for a long time. WAe have documented that thetwo cities are relatively well served with the essential elements of infra-structure, such as water supply, roads, transport, and electricity.

Consequently, we have an optimistic view of the prospects of manag-ing city growth in developing countries. This is in marked conitrast tomuch writing concerned with urbanization in developing countries. Wedo not believe that cities in general are growing out of control in devel-oping countries. WAe do not believe that the vastlv expanding urban pop-ulation in these countries is necessarily condemnled to a shelterlessexistence. We do not believe that it is impossible to provide a modicumof necessary urban services that are affordable and manageable. We donot believe that cities are being swamped by a flood of destitutemigrants who have no productive employment prospects. We do believe,however, that some clhanges in attitude are necessary if this optimisticvision of cities in developing couLntries is to become a generalizable real-itv. The study of Bogota and Cali suggests that maniy solutions are bestfound by the urbani constittuents themselves and that the job of the pub-lic authlorities is to develop institutionis and systems that are sensitive tothe emerging needs and preferences of households and firms and arecapable of reacting accordingly.

Our observations of Bogota and Cali are colored by the relativelyhappy state of the Colombian economy in the 1970s. Tlhis was fortuitousin many ways but was also the result of the prudenit national economicpolicies followed durinig the late 1960s and throughl the 1970s. The stateof the urban labor market and urban income generation is much mioreconditioned by national economic condlitions and policymaking than byany specific local policies.

In retrospect, it turns out that the period during which this researchwas conducte(d (1977-81) and the period under study (1972-80) wereremarkable in recent Colombian economic history. The average growthrate of real gross nationial prodtuct was 6 percent from 1960 tight tIp to1980, althouglh it showed signs of slowing toward the end of this period.During the 1973-78 period an unprecedented 6 percent anntual expan-sion in employment took place. It was in this overall nationial economicclimate that we observed the tightening of the urbani labor market inBogota and Cali that was accompanied by a rise in real wages. The early1970s witnessed a major turninig point for Colombia: real wages in agri-culture began to tise after stagnating for at least forty years, and the agri-cultural labor force began to stabilize or contract in absolute terms. Inaddition, this was also a period of hiigh world coff ee prices, which madepossible a more rapid rise in real agricultural wages in Colombia in the1970s. By the end of the 1970s urbariizationi had reached a level of

2 74 UNDERSTANDIN(; THE DEVELOPING METROPOLIS

about 70 percent. This is beyond the point at which the rate of urbangrowth typically slows. The process of urbanization in general and thegrowth of particular cities follow a sigmoid path, with an accelerationin growth followed by deceleration. Colombia had reached the turninigpoint of the curve by the late 1970s. The effervescent growth of the pre-vious fifty years was therefore drawing to a close during our period ofobservation. This is an important factor since the period encouraged amore optimistic view than that held by earlier observers, whose expecta-tions of unending rapid and bewildering change had been conditionedby the experiences of those fifty years.

One lesson of our study is that understanding city growth requires aknowledge of the overall process of urbanization. The transition from arural to an urbani economy is veryi rapid in historical terms for mostcountries. The trajectory for a country moving from a level of about 30percent urban to about 70 percent urban is steep, and it has usuallybeen traversed over a period of fifty to seventy years. During this periodmost cities in an urbanizinig economy grow at unrprecedernted rates. It isunderstandable if observers and administrators are driven to despairduring this period of seemingly unending rapid change. The task ofmeeting all the demands for jobs, shelter, water, roads, transport, andother urbani infrastructure is daunting.

It is therefore all the more important to tLnderstand that rapidchange durillg such a period is a norm of urban development, not anaberration. Once this is accepted, it follows that policies for urbaniza-tion and provision of urban infrastructure and employment must bepositive, not negative. Urban growth must be accommodated and insti-tutional rnechanisms derived to cope with suclh growth. Most cities andcounTtries are faced with acute fiscal pressure during such a period. Thisin itself points to the need for fiscal conservatism in the provision ofurbani services and for innovative financing mechanisms, like thosefound in Bogota and Cali, to provide for the necessary urball infra-structure.

This study differs from others in a number of ways. First, it is probablythe most comprehensive study of any city to be uLndertakeni sinceHoover and Vernon's (1959) study of New York in the 1950s. Second,whereas some other studies have attempted to understand city structuLreby developing comprehlensive models of cities, we eschewed a large-model approach, preferring to conduct detailed studies of behavior.Partial modeling was done to test hypotheses regardiing labor markets,employment location, r esidence location, housing choice, transportchoice and trip patterns, and the effects of some government policies.Thiird, we have attempted to link the changes in city structure witlhunderlving behavioral patterns, a step seldom taken in other studies.Fourth, and as a direct consequeniCe of the overall approach adopted,

(COPING WITH CITY (;ROWTH 275

this study may be regarded as being an exercise in positive (as distin-guished from normative) analysis.

Understanding Behavior

How do households in cities in developing countries behave? Can weunderstanid the workings of labor markets in these cities? Is the labormarket highly segmented, thereby constraining hiousehold choices?What determninies decisions about housing r-egarding owniership status,location, quantity, and so on? What determines the movement of firmsas they grow? Where does a firm locate to begin with and why? How aretravel patterns determined? What is the range of effective choice intransport modes for different kinds of households? These are the kindsof questions we addressed in attempting to model behavior.

The broad result is that many of the behavioral relations documentedin this studv are similar to those obtained from studies in other cities-mainly in developed countries. We ma,v infer that the transfer of behav-ioral findings is high and that the worlkings of cities in developing coun-tries are therefore quite comprehensible.

Income level is a kev determinianit of household behavior. In unider-standinig the determination of income itself, wse have fotunid that thehumani capital model of earnings works well and explains a substantialportion of the variation in earninlgs. The explanatory power of such amodel was higher in Bogota and Cali than is ustually the case in devel-oped countries. Mluch of the dispersion in earnings is explained by thedispersion in educational achievement. There was little evidence of sig-nificant segmentation in the labor markets of Bogota and Cali; workerswith similar backgrounds were found to have similar earnings in differ-ent activities. Hence the informal sector was hard to identify. If therewas segmentation in the labor market in these cities, it was probably inthe government sector, which paid workers with similar backgroundsslightly higlher wages than private-sector employers did. There was alsoevidence of segmentationi by sex: by and large, other things being equal,women earned substantially less than their male counterparts.

The other major explanation of household behavior is its stage in thedemographic cycle. Participation of household members in the laborforce, their demand for housing, and their travel patterns are all condi-tioned by characteristics that are related to the household's size arid theage composition of its members. For a large proportion of workers, whohave relativelv flat age-earnings profiles, the advance into the middlestages of their life cycle of household formation and expansion alsomeans a descent into relative poverty as family size increases but earn-ings remain flat. This phenomenon is more noticeable in cities or coun-

2'76 UINDERSTANDIN(; THE DlEVELOPING. METROPOLIS

tties in which average education levels are relatively low, as was still thecase in Bogota and Cali. As may be expected, women's participation inthe labor force depends greatly on whether there are small children inthe household. Because the value of womeni's time in child care andother hotiselhold activities is traded off against the expected benefits ofearninlgs, woineil with higher education are more likely to work outsidethe home. The expansion of higher education in the 1970s pro'ided asupply-side reason for increased participationi of women in the laborforce. Overall, women were found to respond well to earnings opportu-nities in the lahor market: as real wages rose in the late 1970s the laborforce participation of women increased measurably. This has been afamiliar phenomenon in developed countries in the post-World War IIperiod, wvhich has seen a secular- increase in the participationl of womenin the labor force.

The techiniqLues of analysis and the basic principles of behavior usedin investigatinig labor market behavior and labor force participationwere the conventional ones used in developed countries. The resultsobtained are quite robust and conform to expectations. The under-standing of income determinationi, labor market dynamics, and incomegiowthi helped significanitlv in tracing the changes in city structurebrought about by inconie growthi and( changes in demographic andhousehold characteristics.

Transport choice is also determined largely by income levels; morejourneys are ma(le as income rises. As maw' be expected, income levelalso largely determines what mode of transport is chosen. In Bogota andCali the level of service provided by busetas makes them a credible alter-native to private automobiles and pro%'ides an important clue to devisingstrategies aimed at containing the growth of private automobile use.Choice is an essential feature of such a strategy. Transpor-t choice andtravel patterns are more depenidenit on available service levels and indi-ces of accessibility than on price levels. The availability of adequatechoice also mitigates the tendency of households to use private trans-port as their income increases. Understanding the heterogeneity ofhousehold characteristics and demands leads to the conclusion thattransport supply must also be heter-ogeneous to reflect such denlands.

Similarly, income levels and household characteristics are the maininfluerices onl housing demanid. As has been documented in developedcountries, the income elasticity of housing demand is less than unity,impling that households with low incomes spend a substantial propor-tion of their income on housinlg, leaving little discretionary income forother expenditures. The nature of the demand for housing changeswith shifts in income levels and movement along the life cycle. Peoplewith low and uncertain incomes find it difficult to enter the owner-occupied housitig market and prefer to rent, which offers greater

COPING WITH CITY GROWTII 277

choice and mobility. Indeed, the mobility rates for households inBogota were found to be similar to those in American cities. The infer-ence is that, as needs change, households change their hotisirig con-sumption bundle accordingly. The flexibility of supply in the housingmarkets in Bogota and Cali has permitted households to exercise awider degree of choice. Choice is, however, clearly limited for the poor,who-particularly if they have large families-tend to live on theperiphery of the city, where they can get more housing for less moneybecause land is cheaper. Average family size was found to increase withdistance fiom the city center in both BogotA and Cali.

The behavior of the land market also broadly conforms to expecta-tion. The trade-off between land price and transport costs, particlltarlytime costs of travel. are evidenit in Bogota and Cali, and regular landvalue gradients are observed, as in most other cities. Densities are alsoquite high, however, in the periphery, where the poor live. Whereas therich tend to substitute capital for lancd when land prices increase, thepoor stibstitute "crowding" for land. For those with larger families, aslide down the land value gradient is essential to preserve a modicum ofhousing consuLMption, but density rtemains high as more people crowdonito somewhat larger plots. In this case, predictable and unlder-stanid-able behavior leads to a different result: in developing countries thepoor live in crowded conditions on the periphery of cities, whereas indeveloped countries the rich live in large-lot suburbani developmnents.

Central cities have oftten been observed to be "incubators" for en trepre-neurship. The employment location of firms is well explained by locationmodels based on standard profit-maximization or cost-minimizationprinciples. Firms minimize localization costs by locating near their sup-pliers and their markets. Locations nearer the center of the city are alsousuallv better served with essential infrastructure. New firms respond tothe same conditions and tend to prefer these locations. As firms growanid their technology changes, requirinig greater space, they move out toareas where more space is available. Failures are intritn'sic to new firmdevelopment. Firm death rates in BogotA and Cali were found to be sim-ilar to those in North American citier. but firm birth rates in the twoColombiani cities were vastly higher, indicating the conditions character-istic of a developing counltryv

The adaptability of firms and households to rapid change is an all-pervasive characteristic. A city's structuri-e changes inl respoi'se to de-manlds made bv those who live or woIrk there. These constituents in tUlrnadapt to the chaniging city structure. As land prices rise in the city cen-ter, firms and households decentralize. This act itself moderates the rateof increase of land values. New center-s of activity develop elsewhere inthe city, and households and firms again respond by clustering aroundthese newer centers.

278 UNDERSrANDING THE DEVELOPING NIFTROPOIJS

Understanding behavior also means understanding the heterogeneityinherent in the characteristics and consequent demands of both house-holds and firms. This is particularly important in a growing city, wherethere is much more ferment than in a city that has reached a position ofstability. A key lesson from our studies of behavior is the resilience ofhouseholds and firms in the face of rapid change and their willingnessto adapt to chaniginig coniditionis. City structuLres and urbanl policies musttake these characteristics into accounit in order to provide for the varie-gated dernands of a heterogeneous set of city constituents.

Understanding City Structure

A pervasive phenomenon characterizing a growing city is the decentrali-zation of residential and economic activity. Changes in citv structurealso result fiom changes in prevailing technologies. Because we havefounid the transfer of behavioral findings to be high, we also find manybroad similarities in uirban structure in developed and developingcoUilltrieCs.

A conventional method of observing the structure of a city has been tolook at the prevailing patterns of population density and land prices,which reflect much of the change that takes place. Both land prices andresidential and employment densities are usually highest at the city centerand decline exponentially toward the periphery. As a city grows, the gradi-enit of these exponential curves declines and becomes flatter. Within thisbroad generalization, many subpatterins exist in different cities accordingto their peculiar circumstances and topographical characteristics.

Bogota and Cali conform to the generalized pattern. Land value andpopulation density configurations are consistent with expectations, asare the changes in these patterns ovetr timile. Land values declinled expo-nentially in all directions from the citv center. This suggests that theland market works as might be expected and does not suffer from muchdistortion. Land values are relatively similar at equivalent distances fromthe city center, a fact indicating the absence of significant segrnentationin the land market. Density gradients, however, differ substantiallybetween the different segments of the city. The de nsitv gradient is steepin the rich northern sector, but it is quite flat toward the poor souther-npart of the city.

How is the density gradient consistent with the regularly declininigpattern of land values? The rich sectors have higher land values becauseof better employment opportunities as well as better neighborhoodquality and infrastructure. Because land values are relatively regulam, thepoor have no choice but to substitute land by crowding. But it is thelarger househiolds that locate there. Poor, large households settle at the

COPI N WI FIH CITY CRO WTH 279

periphery to take advantage of the lower land prices-although theseprices are in fact similar to land prices in the rich areas at equivalent dlis-tances. Taking advantage of lower land prices, they (the poor) buy orrent somewhat larger lots. Because the poor are uniable to substittutecapital for land (that is, to build more floors), their onlv choice is crowd-ing. The rich can build more floors on the same lot size and redutce thenumllber of persons per room. Hence, even the periphery exhibits highipOpulatioin denisities in the poor areas. A flat density gradienit is thusconsistent with the existence of a relatively strong land value gradient.This example shows how overall regularities of structure can mask sig-nificant variations at a more disaggregated level, variations that clearlyrclate to uniderlying behavioral patterns.

A significanit difference between the decentralization observed in cit-ies in developed countries and the cases of Bogota and Cali is the persis-tence of high employment densities at the city center. Whereas indeveloped countries the decentralization of employment has usually

meant an absolute decline in the level of economic activity in cenitral] cit-ies, in Bogota and Cali it did not. This phenomenon is related to the lowincomes that contillUe to prevail for a large proportioni of the people.An important aspect of city structure emphasized in this study is the sep-aration of rich and poor in different parts of the city. In addition, therich parts of the city had more jobs than they had workers. The clemandfor goods and services being higher in the richer parts of the city, thedemand for labor is also higher in these parts. Poor parts of the city,conversely, have a deficit of jobs. Poor hou-seholds have meager de-mands because poor sectors can support only a limited number of activ-ities, mainly small retail businesses that satisfy daily subsistence needs.Thus, much of the retail (and wholesale) activity continues to be con-centrated in the center in Bogota and Cali, and there is an excess ofworkers residing in the southern part of Bogota. The relatively low in-come levels in the city therefore explain to a considerable extent thecontinuation of high levels of economic activity at the center. In richercities, retail and other service activities decen-tralize much more in re-sponse to demand.

Another consequenice of this pattern of residence and employmentlocation is the increasing incidence of cross-commtuting as a city growsand employment decentralizes. IMulch commuting is radially directedtoward the central city. As long as employment density gradients aresteeper than residential density gradients, the predominant patterni ofcommuting is inward, toward the city center. With the decentralizatiollof employment, particularly toward the richer section of the cit,v. thepoor need to cross-commute across the center to the richer pat-ts. Thisneed leads to a demand for circumferenitial trips, the availability ofwhich would eliminate the inconvenience of traversing the congested

280 UNDERSTANDING THE DEVELOPING METROPOLIS

city center. It would be beneficial if road patterns and public transportroutes were designed in response to this changed commuting pattern.Tlhe example above illustrates the manner in which household behavior(with regard to hotlsinig consumiiptioni and residenice location) and firmbehavior (with regard to location) interact with the operation of theland market to produce a predictable and comprehensible city struc-ture. Knowledge of this interactiofi could lead to the provision of appro-priate transport serxices.

Changes in technology also influence city structure greatly. The move-ment in Iilanufactturing technology from batch processing to continuousprocesses that require much greater space has contributed to the decen-tralizationi of manutifacttLirinig employment. Conceptually, such a techno-logical change r educes the elasticitv of substitution between capital andland: it is difficult to stack a contioLous manufacturinig process on dif-ferent floors. Hence, there is greater demnand for land, and a firm is bet-ter off locating at the periphery of a city, where land values are lower.

Older cities-for example, those in Europe-were heavily dependeniton the railwavs for intercity transport of goods. Transport costs there-fore indluced maniy manufacturing enterprises to locate close to the rail-head, usually in thie city center. The advent of trucking and new highwaytechnology has made it profitable for firms to locate at the peripheryand so take advantage of the flexibility and convenienice of trucking.The decentralization of manufacturing activity in Bogota and Cali hasclearly been related to the conveniience of trucking and the increasingavailability of roads. There is a natural tendency for manufacturingfirms to relocate toward the periphery of a city as they expand. Prema-ture forced location at the periphery, however, woul(d be harmful to thehealthy' generationi of economic activity in the city.

We have documented the importance of income as a determinant ofhousehold behavior with r egard to demand for housinig and transport.How does this affect city structure? Rising incomes lead to a demand forlarger houses. A general increase in incomes in a city can be expected tolead to decentralizationi of residence, with each household demandingmore and better housing as incolnles rise. Thus, each household wouldmove out to where land valties are lower, and the city would decentral-ize. In fact, in developing countries, income growth is accompanied by acontinutial increase in population. What is observed therefore is a gen-eral increase in densities in most areas of the city and an expansion ofthe citv itself. As each area becomes denser, land values increase andnew accretions take place at the edge.

The spatial expansioni of cities into the surrotundinig agricultural areasis sometimes viewed with alarm, yet the densification of cities is alsoseen as leading to excessive congestion. Our studv of behavior and itsinterrelationi with city structure suggests that both patterns are inherent

COPING WIT'H (ITY (;ROWTH 281

in the dynamics of city growth. The real issues relate to the sequenicingof such city growth and the provision of matching uLrban services.

Encouraging Endogenous InstitutionalResponses to Rapid Growth

The theme of this study has been the extent to which behavioral adapta-tion to change is facilitated in the context of the rapid city growth char-acteristic of developing countries. Observation of the processes ofchange in Bogoti and Cali has led tis to believe that rapid city growthcan be managed in such a way that most residents of cities experiencingsuch growthi can be provided with urban services that are commensu-rate both with their needs and with the delivery capabilities of the gov-ernment and other ptiblic and private agencies. The key lesson fromthis study is that city growth is easier to manage successfuilly if the tasksof managemiienit are decentralized in such a way as to encouLrage endoge-nous responises to the changing situation.

The rate and natture of city growth are seldom precisely plredictable,but the broad contours of the process are well known, and even the gen-eral patterns of structural change that will be experienced by a growingcitv can be anticipated. Within these patterns, however, there are con-siderable tlicertainty and unpredictability, requirinig the establishmentof institutional mechanisms that can respond actively and innovativelyto rapidlv unfolding, new situations. We have emphasized that thisresponse is more likely to be possible if we unider-stand better the behav-ior of the various constituients of the city-households, firms, and gov-ernrment agencies-and the impact of their behavior patterns on theevolving city structure.

A standard governmental response to rapid city growtlh is to prepal-ecomprehensive metropolitan plans. Such plans attempt to shape thegrowth of a citv for perhaps five to twenty years. delineating land use ina detailed manner and freezing citv structure for- the planned period.It is also typical to do physical planning for all the anticipated infrastruc-ture needs. Often, however, the costs of stich infrastr-uicture are not calcu-lated realistically, and, when these costs turn out to be excessive, the planbecomes nonioper-ationial or implemenitationi falls far short of what wasintended. Large portions of the growing city are then deprived of essen-tial needs. Success in drawing up physical plans depends heavily on effec-tive forecasting of population growth: as we have seen in Bogota andCali, such forecastinig is fraught with difficulty and is seldom successful.

The preparation of detailed plans is necessarily time-consuming.Moreover, in most counltries the adoption of an urban plan also involvesdifferent levels of political consultationi-another time-consuming pro-

282 UNDERSTANDING THE DEVELOPING METRoPOIIS

cess. By the time a plan becomes operational, it may alreadv havebecome obsolete.

Largely as a result of these difficulties, the experience with urbanplanning has been less than happy. But the desire to under-take similarexercises remains widespread. One lesson of our study is that citieswould be better advised to decentralize the instrumiienlts of iiftastrtic-ture provision that are able to finatice themselves and to respond flexi-bly to the changinig demands of a growing city. The mechanism of self-financing is important becatise it serves as a self-correcting procedurewhereby higher-priority projects are implemented first and realisticplanning becomes a necessity. Self-financing by an agency does not nec-essarily implv commner-cial financing. It can incilide governmenit subven-tions, commercial loans, governmilent loans, and soft loans, as well aspublic contributionIs from users. It cloes, howeveir meani greater agencyautoniomy than in a system in which infrastr-ucture programs are ftillyfunded andc handed dowsn from above. WN`hat is important is that the de-centralized autonomous agencies develop the ability to respond to theemer ging (leniatids of the growing city.

Given the limited ability of governmnent agencies to deliver the fullinfr-astr-ucture and other requirements of a rapidly growing citv, it is alsoprudent to allow private initiative to flourish whenever possible. This isnot so much a matter of ideology as of' necessitv. The relatively happyexperience of' Bogota and Cali with hiousinig provision is a case in point.Public agencies were clearly not in a position to meet the f'ull require-mnents of' landl development or housinig durilng the 1950s, 1960s, and1970s; it was the relatively tlihindet-ed operations of the ur/anizadorespirtlas that helpecl house the burgeoninig population of these cities dur-ing that period. Similarly, the transport systemn has expanded adequatelyto cope with the cities' needs, through extensive use of' private incen-tives, althIougil this time within a public r egulatory framework. The scar-city of' pIblic resources that usually constricts the expansion of publictranspor-t was not an issue. Private investment in public transport wasallowed to flourish in such a way that high serxice levels were achievedat low cost. The system was also allowed to respond to the heteroge-neous demands arising from increases in household incomes, and aniicreasijig variety of- transport modes was encouraged.

It is difficult to assess how mtici conscious policy-level thinkinig wentinto devising the kIind of institutional arranigetimetits that were in placein Bogota and (Cali in the 1960s and 1970s. It is likely that institutiotnsand policy responses emerged somewhat endogenously in response torapicl url)anization. It is true, however, that the highest policymakingautilor-ities had a positive attitude toward urban developmett, invest-ment in urbanl infrastructure, and housing during that period. Innova-tive schemiles stich as the indexed UPACl hoUsing finanlce instruments

(COPING WITH CITYGROWTH 281

were the result of such high-level policy concern. Some of the innovativeand flexible policy responses that we observed in Colombia may there-fore have emerged as a result of a generally positive attitude, rather thansimply fortuitously. It is useful to delineate some of the ir tin illstitu-

tional developments that were founid to assist the rapid growth in thesecities.

One problenm that a rapidly growing city typically faces is how to man-age a fast-expanding supply of dleveloped land for housinig and otherpurposes. Rapid in-migrationi makes the demand for lhotisin,g growmuch faster than normal population growth, and entirelv new areashave to be developed. Land development includes investment in infra-str-ucttire sucih as water supply, sewerage, roads, and power supply. Allthis requires substanitial front-end investment, which public authoritiescan usually ill afford. On the demand side, households face similarproblems: most households do not possess the wherewithal to enter theowner housing market by purchasing a fuilly finished housing bundle.There is usually a time lag between the expression of housing demandand supply by hotisinig services.

Different cities have attempted to responid to this situation in differ-ent ways. In Delhi, for example, the government effectively nationalizedall land that could possibly be urbanized with the idea that the ptiblicauthorities would develop it and make sites-and-services projects avail-able to the poor at affordable prices. In addition, the public authoritieswould build houses. Thus, uliIdesirable speculation would be avoidedand the poor would have better access to shelter.

fIn fact, the public authorities have not been able to keep pace withthe bLtgeoniing demriand, and the legal housinig they provide is only asmall proportioni of the total. Because of this supply-demand imbalance.land prices have risen tremendotisly, limiting access to hotising anddefeating the originial purpose of public land acquisition. In Colombia,too, the national hotisinig agency, the IcT. initially attempted to stipplyfinished houses for the poor-. As many as 60 percelnt of all urban house-holds could not afford the lowest-cost ICT houses, however. Thlis is notatypical-in fact, it is a common situation: houses fully constructed bypublic agencies are seldom accessible to the poor. Specific housinigprojects can rarely address the problem of' providing shelter tor thepoor. More general approaches devoted to improving access and reduc-ing effective prices are more likely to be successful.

What, then, is the solution? The conceptual breakthrough is to r ealizeand accept that under almost no circumstances can the poor in develop-ing countries afford a complete housing bundle. In principle, a well-functioning hotisinig finanice market should solve this problem. In prac-tice, the earning pattern of the poor is such that financial institutionisfinid it difficult to finance them; moreover-, low-income householdls are

284 UINDERSTANDING THE DEVELOPING ME rROPOLIS

loath to get into long-term debt relationships in the face of other press-ing consumption demands and uncertainties about their earnings. Thepredominanit solution in Bogota and Cali, and in many other cities, is to"unbundle" the housing and infrastrticture package and to invest incash and kind in an incremental fashion. The public agencies alsosolved their problemii by supplying iifi-astr-ucture services piecemeal,unibunidlinig the infrastructiu-e package (a solutioni not permitted inholistic planning fi-amewor-ks).

The housinig bunidle consists of developed land, the actual housestructure, 'and a package of infrastr-uctural serxices. Households solvetheir problem by buying relatively undeveloped land to begin with,building a core house, and then obtaining access to imiproved infra-structure over time, while augmeniting the house for househlold expan-sioin as needed and as incr-eases in incomiie permit. From the supply sidethis was made possible by the interventioni of small, indepenident devel-opers who jumped in to fill this large demand niche. Althouigh govern-merit regulations and other rigidities often do not sanctiori this kind ofincremenital solution, it is a rational response to household needs in arapidly growinig city. Buying a semideveloped site reduces the effectiveland price for the household, which then benefits from the large in-creases in land values that occur after full development. By reducingthis price, the household can afford better and larger housing over timethan. if it attempted to start with a fully finished unit. The cost of thestructure is also reduced, to much less than the cost of an equivalenthouse built in the conventional way, by the use of recycled materials andfamily labor. Similar-ly, at the beginniing, the household makes do with alower quality of ur-ban services. In such an incremental development,various agencies bring in different services at different times. In a fullybuilt-up developmenlt all services have to be provided simultanieously,which can be so costly as to be prohibitive; hence the development ofteidoes not take place.

The poorest 30 percent were not even able to enter the incremiientalowner-ship housing market. They were mostly renters, partly becausetheir incomes were so low they cotld not afford even core housing sites,but also because of uncertain employment, which necessitated frequentresidential moves. The poor therefore rent rooms from people who areslightly better off, thereby becoming housing finance agents for theirlandlords in the bargaini. The slightly better off enterinig the housingmarket often rent roorns to the less well off as a source of funds withwhich to finance their next stage of development. Governments havetypically done little to encourage a rental market that facilitates shelterfor the urban poor. Most government shelter programs, even sites-and-serxices projects and other programs targeted at the poor, are aimed atincreasing home ownership. The behavior and preferenices of poor

COPING WITH CITY GROWTH 285

households, however, clearly indicate that it is not generally feasible topush them into the ownership housing market, nor is it usually in theirinterest. Closer observation of the operation of housing markets in cit-ies in developing countries and of household behavior would havemade this apparent earlier. Yet most government programs continuiie tofocus on ownership housing. The reality that the poor are seldom candi-dates for homeownership should be reflected in the policies of housingfinance agencies that serve this group. Rental of rooms to the poorest bythe slightly better off should be made easier; the poorest will then behoused and a general expansion of the housing stock achieved.

The overall lesson of these observations is that governmernt shouldrecognize the effectiveness of this process and facilitate institutionalarrangements that enable people to express their preferences andinvest their own ingenuity and resources in providing for their own shel-ter. On the supply side, governments should be more realistic in assess-ing their owIn capabilities for supplying the requisite urban services anddeveloped land. It should be possible for small, independent developersto stipply the needs that governments are clharacteristically incapable ofsupplying. This broad approach may be characterized as promoting thepublic use of private incentives.

Transport

The managemtienit of transport in Bogota and Cali provides a similarillustration of a public service system's responding well to householdneeds and with built-in flexibility ancl adaptability. A rapidly expandingcity requires a rapidly expanding transport system. One approach toproviding this kind of service is to operate a publicly owned transportcompany and to develop a metropolitan plan for expansion of services.This requires heavy capital investments in equipment. heavy expendi-tures on operating costs, and, in a large city, the ability to cope with themanagement problems inherent in the operation of a large system. It isdifficult for such a system to responid to rapidly changing consumerdemands and to provide for different levels of service quality demandedby households of different characteristics. It is also difficult to makesuch systems pay for themselves because of constlner resistance to real-istic fare setting in a government-owned public scrxice. Regulatory sys-tems should allow for the entry of different types of transport supplierswho provide a variety of transport services in response to the typicallyvariegated demand found in a growing city in a developing country.

The institutional ftamework in Bogota and Cali has allowed for adecentralized system that can respond quickly to changing situationis.The route allocation system allows for new routes to be opened inresponse to consumer demand. The introductiLon of busetas introduced

2.Yfi UNDERSTANDING THE DEVELOPINGl METROPOLIS

both greater flexibility and a higher level of sersice. Other cities haveallovwed even greater heterogeneity in service levels. At the same time,goverinimienlt regulation of fares and routes prevents a free-for-all andinjects some order into what might otherwise be too chaotic a system.The system of fare subsidies performs relatively well as a subsidy to theless well off because only the lowest level of transport service-full-sizebuses-receives such subsidies. Our evaluation of the whole transportsubsidy and regulatory system suggests that the system. in general, workswell. This uniderliines the fact that a government subsidy system to pri-vate. cecentralized operationis is probably cheaper to operate than afully owned public transport system. Another significant feature of thetranspor-t system in Bogota and Cali is that flat fares are effective in miti-gating the spatial disadvantage suffered by the poor', who are pushed todistant peripheral residential locations.

We have emphasized that the decentralization of employment sug-gests that circumferential routes also have to be introduced as a cityexpands, although the continiuied concentrationi of employment in citycenters in developing countries makes possible the continued use oflarge high-occupancy buses on trunk radial routes. (In developed coun-tries the loss of employment in the city center, coupled with highincomes, makes it increasingly difficult to run high-occupancy vehiclesofteni, even on trunik routes, without subsidies.) Low wages in develop-ing couLitries also make it econoiiicallv feasible to operate smaller busesand vans, whereas, in richer cities, drivers' wages make it uneconomicalto runi anything but large buses with a minimum number of passengers.

These arrangements illustrate a point: that in cities in developingcounitries it is feasible to devise public transportation systems that mini-mize the use of public resources and respond to the needs of the vastmajority of city residenits. The specific needs of different cities may vary,but the behavioral principles do not. Household and firin behavior havea bearing on city structure, and such knowledge can help in designingappropriate systems for public tranisport provisions in a responsive man-ner. Thle tr-ick is to devise a system that economizes on information col-lection and processing and responds positively to emerging needs.Nlajor ex ante public decisionmakinig is not needed in such a system.

Infrastructure

The governments of Bogota and Cali have acted as followers rather thanleaders in guiding city growth. There were no systeinatic attempts atoverall urban investmenit planning for the rnedium term. Yet theachievement of adequate public service levels has been better than inmost cities in developing countr-ies. How did this happen? Our view isthat, paradoxically, it is the decentr-alized Ilature of urban governnient

COPING WITI H CITY GROWTTH 287

in these cities that has led to such a creditable performance. As men-tioned earlier, decentralization facilitates the effective sharing of thecosts of urbanization among all concerned. A more centralized systemtends to shift all present costs to the government, which usually finds itdifficult to bear the burden. The result is inadequate investmenit inurbani infrastructure and the emergence of significant public servicedeficits. The autonomy extended to public service agencies has helpedfoster a culture of user-financed public services. Capital costs have usu-ally been funded by loans, wvhich have then been effectively serviced byuser charges. The autonomy of public service entities also helps themconcentrate on providing service and respondinig to demand.

During the period of particularly high growth, the system of valoriza-tion helped provide for road infrastructure, which is usually seen as aclassic public good, financed solely fiom the governmelnt budget.Although the system fell into disuse later because of practical difficul-ties, it served its purpose at a crucial time. Again, the principle was thatpeople benefiting from the development of infrastructure should beaware of such development and should share in paying for its costs.

This system stands in sharp contrast to more centralized systems thatare based on a priori impressionis of requirements and are niot geared tomeeting the preferenlces of urbani residenits. Our observationi of tirbanbehavior, of the tinfoldinig citv structure, and of the government agen-cies stipplying services in Bogota and Cali leads us to the conclusion thatcity growth is manageable only if institutions are designed to be respon-sive to the behavior and needs of the various constituents that make tip

a city. The institutionial framework must have an endogenous ability toser-ve grcatlv heterogeneous demands, and an ability to change as thesedemands change in a rapidly growing city.

Appendix: The Data

The City Study assembled a data bank of existing sourcts of informationabout Bogota and Cali in the form of copies of the original computertapes prepared by the respective originators of the data. All have beendocumented in detail by NValverde (1978) and Y J. Lee (1979, 1981).A number of primary surveys were also carried out specifically for theproject. The main comprehensive primary data collection effort wasthe 1978 World Bank-DANE Household Survey conducted in Bogotaand Cali, a joint project of the Bank and the Departamento Adminstra-tivo Nacional de Estadisticas, Colombia's national statistical agency.The main data sources used in the studv are brieflv described in thisappendix.

The 1973 Population Census

Before 1985, the most recent censuses taken in Colombia were in 1964and 1973. Unfortunately, the 1964 census did not report incomes.There was widespread skepticism about the coverage of the 1973 cen-sIus,1 but Potter and Ordonez (1976) concluded, after careful demo-graphic analysis of an advance sample, that the information appeared tobe of good quality, at least in relation to previous censuses. They esti-mated that the overall undercounting for Colombia as a whole was prob-ably 7 percent. The tape proxided by DANE for public use was a 4 percentsanmple of households. For Bogota, however, the tape covered all house-holds living in the buildings that happened to house the 4 percent sam-ple. We used the entire sample for the tabulations in this book andexpanded it to reflect the size of the city's population and estimatedundercounting. Because this study was particularly concerned with spa-tial distributions within the city, the expanded sample was used so that it

288

APPENDIX: THE DATA 289

would be representative of the various subdivisions within the city'sboundaries.

The census containis information about dwelling and household char-acteristics, demographic data for all individuals, labor force informationfor workers, and fertility information for females. Although we cannotcomment otn its coverage, we agree with Potter and Ordonez (1976)that the overall quality of the information in the sample appears to begood. Nonresponses seem to be distributed randomly; the only obviousbias is that single-person households predominate in the "no informa-tion" categories for income and the labor force. One of the most usefuldistinguishing features of this data set is that the location of respon-dents is coded down to the block (manzana) of residence. Income datawere obtained from only one questioni: "What was your income from allsources last month?" Only about 12 percent of the sample did notreport income information-a proportion that compares well with noni-responses in the U.S. census.

The extent of income coverage of the 1973 census as well as the 1977Household Survey and the 1978 World Bank-DANE Household Surveywas estimated (see Mohan 1986, appendix C, for details). The aggrega-tion of all incomes reported in the census appeared to be no more than50 percent of the estimated total personal income for Bogota. Variousfactors are responsible for this unlderreportinig:

* Twelve percent of the people gave no information.* When onIlv one question is asked, much of the nonlabor income is

probably not reported.E Income in kind-for example, as received bv domestic servants-is

probably not reported.* Many earners receive end-of-the-year bonuses; these are character-

istically not covered in one-shot cross-sectional surveys such as thecensus, unlless the question is asked specifically.

In view of these factors, it is not surprising that the income coverageof the censtus was only 50 percent. 2

Household Surveys

DANE has conduLcted a regular program of houselhold surveys since1970; the main objective has been to collect information about thelabor force. Since 1975, these surveys have been conducted quarterly,alternating betweenl Colombia's four largest and seven largest cities,.along with an occasional national survey. We obtained the computertapes for 1972 (Encuesta de Hogares 6, Fuerza de Trabajo: EH 6-FY),1975 (EH8E), and 1977 (EHI5). 4 The 1972 survev was a national one

29o UNDERSTANDING; THE DEVELOPINC; METROPOLIS

and covered 6,371 households, of which 1,348 were in Bogota. It con-tained informnation about housing as well as demographic and laborforce characteristics. The survey did not provide the intracity location ofthe respondenits. The 1975 survey was a special one for the city ofBogota; it sampled 3,953 households and contained information onlabor force and demographic characteristics only. The 1977 survey wasconducted in the four largest cities and sampled 6,082 households,3,161 of whiclh were in Bogota. Starting with this survev, DANE began touse "rotational sampling": 67 percenit of dwelling units sampled remainin the next survey and 33 percent are new. Both the 1975 and 1977 sur-vevs contain the location of respondents' residences. The 1972 and 1975surveys had information about the numiiber of workers in each respon-dent's workplace. This question was not included in the 1977 sample.

In these samples, DANE classified neighborhoods into six socioeco-nomic strata: (1) low-low; (2) low; (3) medium-low; (4) medium; (5)meditum-hiigh; and (6) high. At the conclusion of a survey, weights wereassigned to each of these strata and applied to the meinbers of the stratafor all expansions of the sample. These "expansion factors" are sup-posed to take into account the over- and undersamplinig that may occurover the course of the siirvey. The expanded sample should then be cor-rectly representative of the city as a whole.

The 1977 survey is uinusual in that as many as 20 percent of the work-ing respondenits did not report their incomnes, a proportion that comn-pares unlfavorably with the census. Consequently, a method was devisedto impute incomes to them. (See Mohan 1986, appendix B, whichdescribes the calculation of incomiies Ior all respondents and the imputa-tion method for nonirespondents.)

The coverage of income in the 1977 survey was only slightly betterthan that in the 1973 census, despite the more detailed questionis asked.Labor and nonilabor income data were taken separately, and income inkind was estimated as well. Even when the imptlted incomes wereincluded, the survey covered only about 61 percent of the estimatedtotal of personal incomes in Bogota. It also appears that the highestincomes were either unlderreported or unldersampled. If the incomesreported in the 1977 sample are converted to 1973 pesos, there is littlereal income growth on average, and those in the highest categories actu-ally decline.

The 1978 World Bank-DANE Household Survey

In 1978 the WA'orld Bank's City Study team and DANE jointlv conducted asurvey of about 3,000 households in Bogota and 1,000 households inCali. This was larger and more carefully conducted than previous stir-

APPENDIX. THE DATA 291

vevs. It had five main parts: household and dwelling characteristics;(demographic characteristics of'all individuals; worker characteristics, in-cluding information on workplace; information about the unemploled;and information about the vehicle ownership and journey-to-work char-acteristics for the employed.

A partial recount of dwelling units was done to take into account theexpansion of Bogota since the 1973 census (earlier surveys were allbased on a 1972 sample frame). Information about income was elicitedmore carefully. In earlier surveys, data on the earnings of all workers ina household were usually obtained from any available adult respondent.In this survev, all worker information was obtained from each workerdirectly, even if this requir ed more than one visit to the hoLischold. Fur-thermore, income questions were asked of all members of the hotise-hold even if they did not work. Two questions were asked to obtain laborincome: the amount and periodicity of wage payments and total earn-ings in the previous month. Income in kind was imputed. Various nonla-bor sources of income were specifically mentioned to obtain nonlaborincome. As a result, the income coverage of this survey was about 90 pcr-cent, which was a great improvemenit over previous surveys.

Only about 3.6 percent of responses gave no infol-rimation aboutincome, and these responses have been imputed by the same metho(dused in the 1977 Household Survey. All regressionis were conductedlafter weightinig eachi observation witil expansioni factors, as in the proce-diure for 1977. I'hese expansion factors take into account over- and(undersampling.

Note on Household Survey Samples and the SpatialDisaggregation of Bogoti and Cali

In map 3-1 Bogota was divided into romuinas, rings, and sectors, but thebasic socioeconomic spatial unit in the city is a barrio, or neighborhood,of which there were about 500 in 1973 and about 700 in 1978. DANE geo-coded each of these units with a four-digit nutimiber. The first two digitsidentify a romuna-a collection of barrios. The last two digits identifybarrios writhin the comuna. The comunas were theni furthier aggregateclfor City Study purposes into rings and sectors. The bounidaries of theC0o1U711as shown in map 3-1 are the principal streets in Bogota. The city isbotLmded on the east by mountains and therefore has an approximatelysemicircular shape, although the north-south axis is longer. Notc thatthe first digit goes from I to 9 and roughly rotates (increasing) fromsouth to north by sectors (or pie slices). The seconid digit ranges from Ito 6 and corresponds roughly to rings centered in comunas 31 and 81,increasing from south to north. Similar principles were followed in

292 UNDERSTANDING THE DEVELOPING METROPOLIS

geocoding Cali (map 3-2) so that rings and sectors could be identifiedthere.

DANE, along with the Ministry of Health, compiled an inventory ofmanzanas and of dwelling units within the city before the 1973 census,and that inventory has continued to form the sample frame of all subse-quent surveys in Bogoti. Thus, none of these surveys had sampled thenew neighborhoods that had developed since 1972. The sample framewas therefore updated for the 1978 World Bank-DANE Survey. The sam-pling was designed to make equiprobable the possibility that any dwell-ing unit in the city according to the 1972 inventory might be selected.The basic sampling unit (unidad prinmaria de muestra) is a block, withinwhich all households in all dwelling units are interviewed. Provision ismade for differenit sizes of blocks. Since all the sampling was based onthe 1972 sample frame, it was difficult to trace time trends for changeswithinl the city. Moreover, caution must be exercised in drawing any con-cltisionis about the changing character of neighborhoods. If differentregions of the city differ systematically from one another and if oneregioni chaniges clharacter over time, the later samples will no longer- berepresentative. Sampling is based on the classification of neighborhoodsinto the six socioeconomic strata. If neighborhoods change character-that is, filter up or down in the socioeconomic scale-the resulting sam-ple will no longer be representative. Hence, drawing conclusions aboutfinie changes in the income distribution from two household surveys attwo points in time is hazardouis in the absence of detailed knowledge ofthe samplinlg procedures used. If, however, rates of change are not high,such difficulties are mini1mal-but then one would have less interest intracing time trends anyway! These remarks can be extended to nationalsurveys, where the heterogeneity of regions is perhaps typically morepronounced than within a city.

These details have been offered here because they are seldom givenby users of household survey data. They become particularly importantwhen comparisons are made between surveys undertaken in differentyears; information about each data source is necessary in order to beaware of biases that may arise from differences in survey design andcoverage.

The 1972 Phase II Household Survey

Phase 11 of the Bogota Urban Development Study occurred in 1971-72and was ftinded by the United Nations Development Programme(UNDP) under the auspices of the World Bank. The project included asurvey of 4,675 households in 1972 (for a detailed description of thedata set, see Valverde 1978). Althouglh the survey focused on demo-

APPENDIX: THE DATA 293

graphic and housing characteristics, it also include(d questions about therespondent's workplace, including its location and type of industry.Extensive information about work trips was also collected. This data setwas used extensively in the City Study to obtain a comparative picture ofBogota in 1972, particularly in the employrnent location, housing, andtransportation segments of the study.

The Social Security Files on Establishments

These data (see K. S. Lee 1989) provide a complete listing of all estab-lishments affiliated with the social security system. The Colombian So-cial Security Agency (Instituto de Seguros Sociales) released to the CityStudy a copy of the complete files of Bogota for 1978 and 1979 and ofCali for 1976. For each firm, the tapes containi the address, the numberof workers, and the S1C (Standard Industrial Classification) code; eachestablishment's location was then geocoded at the barrio level at theCorporaci6n Centro Regional de Poblaci6n in Bogota.

The main limitation of this data set was its poor coverage of smallfirms. Even though the law requires that all employees in private andpublic enterprises and self-employed individuals register with the socialsecurity system, those with small family shops and independent andcasual workers tend to avoid membership. Government employees andmilitary personnel have separate social security programs.

For the country as a whole, the social security data compare favor-ablywith other information, such as the industrial census and labor forcesurveys, particularly for the manufacturing and finance sectors (DANE

1975; further descriptions and analyses of the data appear in DAINE 1974and 1976). DANE concluded that the coverage is almost complete forlarge manufacturing establishments (with ten or more workers) andfirms of all sizes in finance but is poor in the commerce andl service sec-tors (which include many small firms).

DANE's Annual Industrial Directory File

D,ANE compiles an annual directory of all industrial establishments. Thisdata set is very similar to the Dun & Bradstreet Marlket Identifier (DMI)

data in the United States, which have been used in a numbei of employ-ment location studies. The DANE directory includes information aboutall establishments with ten or more employees. This covers about 70 per-cent of total maritifacturinig employment in BogotA. The data in thisdirectory include information about location, production, sales, andinput uses of individual establishments. The annual directories for the

294 tUNDERSrANDING THE DEVELOP'IN(. METROPOLIS

years 1970 to 1975 were computerized for Bogota and Cali in order tostudv the trends in employment, births and deaths of firms, and loca-tion of jobs.

The World Bank City Study Surveyof Manufacturing Establishments

In 1978 K. S. Lee organized and conducted a survey of manuifacttiuriigestablishments in Bogota to cxplain the changirig location patternsfounid fiom analyzinig the data from the DANE inidustrial directories.5The establishmnents in the industrial directory were used as a samplebase. The sample of' 126 establishments interviewed in the survey wasdrawIn from DANE's records of 2,629 firms in the industr-ial directory filefrom 1970 to 1975. The sample of firms was stratified by location history(that is, whetlher they were new or mature firms or movers); zone oflocation, defined by thi-tv'-eiglht comttnas; type of industrly, defined bythree-digit sic codes; and size, defined by the number of employees.

To miiliniize the cost of samplinig while obtaininig a sufficient numberof observations for econometric estinationi, the survey focused on tex-tiles and fabricated metals. These industlies have no particular loca-tional requirements, such as proximity to a river or mnies, and shouldbe more amenable to spatial policies thani others suchi as cement orsteel. Moreover, together these two iiniistries accounted for 50 percelntof total manulfacttirinig employment in Bogota The homiiogeneity offirms within each iiduIstL-y also makes it possible to test behavioralhvpotheses with suf'ficient clegrees of' freedom. A third group of "oiherindistries" was added, hiowever, to allow descriptive studies of otherkinds of firms.

The secondz major- consideration in the sampling process was to over-sample large firms in order to maximize the nulmber of workers includ-ecl. Moreover, there was an attempt to cover a wide geographic area thatwould allow estimationi of rent and wagc gradients for the entire city.Thle goal was to obtain a sample of at least 120 firms, with equal repre-sentationi of the three types of firnis. There were 58 mature firms, 50mover-s (including 2 firms that moved to Bogota from outside the city),and 18 new firms. The sample coverage across zones was satisfactory;twenty-sevenl comnunas were represelited and allowed a fairly even spreadover the three rinlgs with high densities of manufacturing employment.Only a small ntimber of establislhmenits, however, were selectedl fromrinig I (thie cential businiess district [1:DB]) and ring 6 (three residentialcomnunas in the north).

In some cases the four-way stratification severely limited the possibilityof'drawinig sample establishmcnits fiom a specific poptulation category.For examlple. not enouigh textile firmn-s were located in certain comu77as.

APPENI)IX T11E DATA 2Y5

Additional establishmenits were therefore selected from the apparelindustry to stipplement the textile industry sample and from the non-electric machinery industry to supplemenit thie fabricatedc metal indus-try sample. The final sample hiad fairlv even shares among the threeindtistry groups: about 35 percent each for the two main industr-ygroups and 30 percent for the "other" categorv.

The average size of mature firms in the sample was about four timeslarger than that of new firms, and( more than twice that of' movers. Thisr esults from the oversamplinig of large firms; the average size of firms ilthe sample (135 employees) is aboutL twice as lari-ge as the averageamong Bogota's mantifacturinig establishments.

The World Bank Housing Survey of Low-Income Households

Twenty-six bar-rios were drawn f'romi- two separate unliverses of' low andlvery low socioeconomic status subdivisionis created either between 1963and 1971 or after 1971 (see 1lamer 1985 for details) i6 These lists werefurther reduced to exclude barrios for which there was no informationon legal standinig at the District Planninig Office. In addition, areas ofthe city with little new low- or very low status subdivision formation wereeliminated from considerationi. Civen these two universes, one for eachtime perioc, a stratified random sample of 1(0 percenit of the barrioswas extracted from each, reflecting the geographic location, legal stand-ing, and socioeconomic status inherent in the sample frames. From theset of' twen-ty-six barrios, thiirteen were eliniinated from thc househ1oldsurvey sainple frame because thtey wvere government-sponsored, hadvery low lot occupanicy (below 50 percent) , or were very small (fewerthan fifty dwellinlg units). Another- barrio was eliminated because it hadbeeni extensively survevedl in a previouis study. Other corrections weresubsequently made, and the restilt wits the set of twelve bairios listed intable A-1 7

TI'he number of' structures in the surveyed barrios totaled 2,433; ofthese, 9 percent were surveyed. Following the recommendationis of localconsultanits, a decision was made not to take a compiletely random sam-ple. lInstead, giveni the paramount interest in the charactelistics of theclifferent prevailing dwelling types ancl in the way thev were modifiedover time, an effort was made to sample roughily the same number ofhousehiolds in each of several clwelling types in the barrio. In each bar-rio roughly twenty houIsehlold(s were interviewed. The dwellinig typessturveyed in any given barrio did not necessarily correspond to the clistri-butioni in place. In addition, on enterinig any given barrio, the surveyorscarried instructions about which tvpes of uLnits to sample randomly onthat particuila day, in order to maintain the reqUiredl structural dliversityat the aggregate twelve-barrio level. That could result, for exam pIe, in

296 UNDERSTANDING THE DEVELOPING METRO)POLIS

Table A-1. Households Interviewed, by Barrio

DA NE: code Nu mber of

Bartio riarne nuumber hkousehaods

Quindio Viejo 1318 13Quindio Nuevo 1399 8Managua 1406 14Molinos del Sur 2506 20Pachon de la Torre 2528 21Pastranita 4525 21De (Gaulle 4527 21San Ignacio 5607 21Prado Pinzon 9105 12Los Naranjos 9204 20Rincon 9205 21Tibabuves 9209 20

.Sitnre: Haomier (1985).

the inclusion of only one or two dwelling types in a given barrio, simplybecause earlier survey work had not managed to uncover enough unitswith those characteristics in other barrios. Thus the survey could beused to characterize the conditions in any one barrio. Finally, and bydefinitiorn, the survey was not designed to derive means and medianseven for the barrios as a whole, except for illustrative and very roughapproximations of actual behavior. It is only a felicitous coincidencethat the distribution of sampled structure types corresponds in a roughway with the actual distribution of structures in the barrios, as noted intable A-2. Such a coincidence allows the cited means and/or medians tobe taken more seriously than would otherwise be possible, because itcan be shown that the different structure types are associated with afairly predictable and distinct set of characteristics. 8 Table A-1 presents abreakdown of actual interviews by neighborhood, and table A-3 pro-vides a review of selected characteristics of the barrios surveyed.

Table A-2. Distribution of Structure Types: Survey and Universe(percent)

Categury SurveV t!.niverse

Tugurioa 14 16Casalote" 26 24Si ngle-story 32 39Multistorv 29 21

Nate: Thcse categories are aggregations of yet more carefullv diffelrentiated structurerypes.

a. A shack, usually of ver y small dimension and intended as temporary shelter.b. Onie or two multiple-use rooms attached to the wall defining the perimeter of the lot.Souroe. Hamer (1985).

Table A-3. General Characteristics of Surveyed Subdivisions

Quindio Poahon .oliinons Los Ptodo .San Qoitndio(hararleristr V'wjo di' ln Torre d'l Su Rimon uNaranjos Pmnzon D), (; aulli Mlanoagun Palrvoi/n Ig7nat70 libabues .uo

Age' I I I I 1 1 1 2 2 2 2 2Stratumi I 1 2 2 2 3' 1 2 1 2 1 1Distanced 2 3 2 3 3 3 3 2 3 2 3 2Legal 78' 3 4 2f 4 3 3 4 3 3 3 4 5Legal72k 2 2 if 2 1 1 2 2 2 1 0 3Waterh 78 77 62 74 75 64 0 73 75 70 75 0Sewerh 0 0 62 78 0 72 0 0 0 74 0 0Powerh 62 73 62 64 65 64 10 73 70 68 75 0LIn its 77i 370 218 541 195 119 497 88 221 195 209 54 223Gross areaJ 20 7 13 6 4 24 1 8 5 8 2 3Net areak 7 5 9 5 3 18 1 5 4 6 1 2P'Op oDE-N 4130 320 400 308 230 222 508 3)00 303 300 260 659I'OP DE\-(," .1300 228 266 243 225 166 430 170 236i 225 175 467One fooloor 99 99 56 82 87 76 91 9() 85 85 96 99Two floor-s" I 1 42 17 13 22 9 10 15 15 4 1Three tloors" 0 0 2 1 0 2 0 0 () 0 0 0(.auz 77" 15 27 10 25 20 20 39 31 25 30 31 17jgugiios 77' 32 19 1 6 2 7 9 11 1 1() 15 54

hlb/i (ontin ues oi Mhejollou'ingpage.)

Table A-3 (continued)

a. Age: I barrio formiied berween 1963 anid 1971 2 = barrio formed al'ter 1971.b. Barrios classified( in ascenDdilng order of socioeconomic raniik. "I" being the lowest: the ranking in the whole critm tins from I to 5.

c. Aithotigh the Pra(lo Piuizon barrio as a whole was classified as having a midile-lowt stamis, the area interviewed w,as then a recent nutcleus that had an atvpi-cal, low statis.

d. Distance: 2 b barrios locate(d at iiteimediate distance fi-om the central htisiniess district (trill), between 1.750 andci 8.750 meters; 3 - barrios located farthierthan 8,750 meters.

e. 1978 slatms of barric, provided by Bogot.i District Planninig Department (Departanienlto Adniinistrrativo dc Planeaci6n. Distrital [DAPrt]): I = niiniiilumnor m development; 2 - legal, stanidard barrio; 3 = barrio in process of being regularized; -1 = Unatlthoriued barr io 5 = squatter- settlement.

£ Althouigh the Molilros bar-rio as a shole swas classified as lcgal. the area interviewed was a then-recent nuclens that had been cleveloped illegallyg. 1972 statis of barrio provided by DAiDt I legal; 2 = in process of regulaiization; 3 illegal; 0 inot applicable.It. Year of barrio's coniniectioni to serice; 0 no connection.i. Barrio dlwelling units in place in 1977.j. 'otal arca in bectares.k. Total area in bectares excluding noniresidenitial use in 1977.1. Net residential derisirv in persons per hectare.m. Gross residenitial denisity in persons per hectare.Tm. Percentage distributions of dwellirig uinits in bar rio accordinig to uniriiber of floors.o. Perceiitage of cdwelling inits classified as suich in 1977.Soure: H1lamer (1985).

APPENDIX: THE DATA 299

Survey of Pirata Developers Conductedby Superintendencia Bancaria

In 1977 the Superintendencia Bancar-ia in Bogota collected data on 135pirata subdivisions as well as 14 Normas Minimas developments (Carroll1980). The Housinig Division sent out questionniiaires to some 200 subdi-viders whose names were in its files and in records of the District Plan-ning Department. The respondents were informied that they werelegallv bound to supply the info rmation and that their subdivisioniscould be "regularized" if thev cooperated. I)ata requested for each sub-division included the size of the original tract; its price and date ofacquisition: the nuniber of lots; the total street and open space areas;the amounits and costs of services installed (sewer, water, telephone,electricity, and paving); and, for eachi lot sold, its size, price, terms, anddate of sale (see table A-4 for a complete list of variables included in thesurvey). The 149 pirata and Normas Miniinas subdivisions for whicil com-pleted questioninaires were returtned represented the efforts of 121 indi-vidual subdividers. No data on dwvelling unlits or on) the characteristics offamilies buying lots were collected.

Table A-4. Variables in Superintendencia Bancaria Pirata SubdivisionSurvey, 1977(from questionnaires)

Seial

nu mber Va2nable description

1. Name of subdivision2. Namlle Ofl)ANt (census) barrio ofsubdivision3. ID ntiniber of DA NE barrio (1973 codes)4. Date of initiationi of physical works5 .Nuniber of lots6. Date of DxAP approval, if anv7. Date of Superinitenidenicia Bancaria appro\al, it anv8. Size of tract (square meters)4. Date of acquisitioni of tract

10. Date of tract acquisition dociiumenit1t. Price of' tract12. Value of installmenits (up to tive) for tract13. Dates of installmi-enits14. Interest rate on tract instillmzienits15. N umnber of buis rouLtes withiin ten min utes walkt6. Existence of neighborinig subdivision with water network17. Subdivisioni inside or outside urbani perimeter?18. Subdivisioni above or below water companiv elevation limit?19. If'above, source of water supply20. For sewer, water pipes, stanidpipes, electricitv, telephones, streets, side-

walks, and cturbs: quanitity installed, cost of installationi, source of finalic-ing, date of installationt

300 UNI)FRSTANI)ING THE DEVELOPIN(G MFTROPOLIS

Notes

1. See, for example, Lubell and McCallum (1978, p. 126), who regard theBogota 1973 results as "simply not usable" and therefoire do not rely on the cen-sus. Their calculations and projections are based on the 1972 Urban Develop-merit Study Household Survey, which covered 4,675 households.

2. This may be comparedl with an estimate of' undercoverage in the Brazil1960 and 1970 censuses bv Pfeffermiiani anid Webb (1979, p. 16), who fitid thatthe censuises cover about 57 to 58 percent of incomes.

3. In descending order by sie, these are Bogota, Cali, Medellin, Barranquilla,Bucamaranlga, Manizales. and Pasto.

4. EH = Enctiesta de Hogares; E = Especial (special sample of Bogota).5. For more details about firm.s in the sample, see chapter 4 of this book; for a

copY of the questioinnaire kised, see K. S. l.ee (1989, appendix B).6. The classiticationi schemile was developed for the Special District of Bogota

byjairo Arias and reporteI in "Estu(lio de EsLratificaci6n de Desarrollo Social,"Departaineito Administrativo de Planeaci6n Distrital, n.d. The schleme wasbased on a poinLs index based oni type of dwelling, type of structutre, type of coll-structioll, level of crowding, availability ot public services, and tamily incomie.

7. One barrio on the original list was r eplacecl by another because of commu-nitv hostility. Anothlel of the barrios chosen for the Sur vey terminated iLs involve-mienit in mlid-streamil, and a substitute barrio was cholsenl to complete therequisite number of inter-views. See table A-i for the final list of barrios andtable A-3 foi a listinig of general barrio characteristics.

8. This can be demonstrated from the suirvev diata. even thouigh these sufferfrom a peculiarit, namiaely that the bulk of the casalole dwellers are drawin frombarrios witi a somewvhat higher socioeconomic ranlkinig than those fronm whichsingle-story occupanLts were selected. Thus, for example. the mean householdincomiie of Iugmrio dwellers is Col$4,400: tor casalote and single- story occup)ants itis Col$6,500; for houselholds living in two-story str-uctures it is Col$9,300; and for

families Wtuinid in thl-ee-stor-v structiures it is Col$13,900. (See table A-2 ftor rlefini-

tiois of' ctaae and Iugunrio.)

Bibliography

7'he uoord "processed" describes informally reproduced works that ma?y not be commonlyavailable through libraries.

Abt Associates; Damses & Moore; General Organiiation for Housing, Building,and Planning Research; and USAII). 1982. InJormal Housing in Egypt. Cam-bridge. Mass.: Abt Associates.

Amato, Peter. 1968. "An Analysis of the Chaniging Patterns of Elite Resiclentiall.ocations in Bogota, Colombia." Latin American Studies lProgram, Disserta-tion Series, no. 7. Ithaca, N.Y.: Cornell lUniversitN.

- 1970a. "A C:omparison of Population Densities, Land Value and Socio-Economilic Class in Four Latin Americani Cities." Land Economics 46 (4): 447-55.

- 1970b. -Elitism and Settlement Patterns in the l.atin American Citv."Journal of Amnerican Institute of Planners (Marrch).

Ananid, Sudhir. 1983. Inequality and Poverty in Malaysza: MVeasurement and Decompo-sitionz. New York: Oxford University Press.

Anderson, Johni E. 1982. "Cubic-Spline Urban Deensitn Finctions." Journal ofLbi,an Economnics 2 (2): 155-67.

APTA (American Public Tranispor-t Association). 1980. Transat Fact Book, 78-79.Washingtoni, D.C.

Ardila, Amparo de, and Andrew M. Hamer. 1979. "Selected Characteristics ofthc Barr-ios of Bogota." City Study Intermediate Paper, no. 21. World Bank,Development Eco0loTnics Departmenit, Washinigton, D.C. Processed.

Arias, jairo. 1977. "Modelo de Desarrollo Urbano." In (oicultitra: l;ida Urbana yUrbanismo. BogotS, D.E.: Colcultura.

Asabere, Paul K., and K. Owustu-Batiahere. 1983. "Population Density Functionisfor Ghanaian (African) Cities: An Empirical Note." Journal of Urban Economics14 (3): 370-79.

Avendano, R., and A. d'.Anico. 1981. "Bogota Bus Drivers: Asesinos o Santos."Bogota. Processed.

Avramovic, Dragoslav, and Associates. 1972. Economic G;rowth oJ C.olombia: Problemsanzd Prospects. Raltimore:Johns Hopkins University Press.

301

302 UNDERSTANDING; THE DEVELOPIN(. METROPOLIS

Barton-Aschnman Associates. 1983. "Transportation Plannririg Technical Assis-tance for Metropolitan Mexico City: Narrative Summary and Technical Sum-mary," technical appendices. Mexico City.Jine. Processed.

Becker, Garv S. 1964. Human Capital: A Theoretical and Empirical Analysis NewYork: Columbia University Press.

Ben-Akiva. M. E. 1973. "Structure of Passenger Travel Demandl Models." Ph.D.diss., Massachusetts Institute of Technologv, Cambridge, Mass.

Ben-Akiva. M. E., and Steven Lerman. 1986. 'Trazvel Behav7ior: Theeory. Models, and

Prediction Metlods. Cambridge, Mass.: MIT Press.Berry, Brian J. 1.. 1973. T'he Human Conseqwunces of tUrbanizalion. New York: St.

Martin's Pr ess.Berry, R. Albert. 1975a. "Open Unemployment as a Social Problem in Urban

Colombia: Myth and Reality." Economic Development and Cultural Change 23

(Janiuary): 276-91.1975h. "Wage Emplo-ment, Dualism and Labor Utilization in (Coloill-

bia: 'rf ends over rime."Journal ofDeveloping Areas 9 Uuly): 563-76.Berry, R. Albert, and Ronald Soligo. 1980. Economic Policy and Income Distribution

in (Colombia. Bouilder, Colo.: Westview.Berry, R. Albert, and Miguel Urrutia. 1976. Income Disiribution in Colombia. New

Haven, Conn.: Yale University Press.Borrero, Oscar. 1979. "Actividad Edificadora Segun Firmas Constructoras." City

Studv Intermediate Paper, no. 8. World Bank, Development EconomicsDepar-timienit. Wasliingtoin. D.C. Processed.

Borrero, Oscar, and Sonia Sanchez. 1973. "Mercados de Tierras en Barrios Clan-(lestinios de Bogota." Bogota, D.E.: Departamento Administrativo de Pla-neaci6n Distrital.

Bourguiginon, Francois. 1983. "The Role of Education in the Urban l,abour

Market Durinig the Process of Development: The Case of Colombia." In VictorL. Urquidi and Satil Trejo Reves, eds., Hutnarl Resoturces, Employment, and Devel-

opment. Vol. 4. London: Macmillan.Bowden, Martyn J. 1975. "Growth of Central Districts in Large Cities.' In Leo F.

Schnore, ed., 'I'he Noew Urban History Prinicetoni, N. .: Princeton lUiversityPlress.

Caceres Bolanos, Robert. 1978. "La Situaci6n Finaniciera de Cunidinamarca."Revista C'imara tie Comercio de Bogoti30-31 (June-September 1978): 123-33.

Cameron, Gordon C. 1973. "Intra-Urban l,ocation and the New Plant." Papers of

the Regional Science Association 31: 125-44.Cariton, Deninis W. 1979. "Models of New Business Locations." In William C.

U'hieatoni, ed., Interregional Movements and Regional (Growvth. Washinigton, D.C.:

Urban I nstituLte.1983. "The Location and Employment Choices of'New Firms: An Eco-

inomilic Model w ith Discrete and Continuous Endogenous Variables." Review ofEconomics and Statistics 65 (3): 44(-49.

Carroll. Alan. 1980. Pirate Subdivisions and the Market for Residential Lots in Bogot.

Staff Working Paper 435. Washington, D.C.: World Banlk. (In Spanish as AlanCarroll, "Las Surbdiyisiones Piratas y el Mercado para Lotes Residenciales enBogota," Revisla C(imara de Comerrio de Bogolti 41-42 [December 1980-March1981 1.)

BIBLIOGRAP'HY .1)3

Charles River Associates, Iic. 1976. Disaggregate Travel Demand Models. Vols. I and2. Boston: National Cooperative Highway Research Program.

Cifuientes, J. 1. 1984. "Urban Transport in Bogota." In "Housing and UrbanDevelopment in Bogota, D.E.: The Institutional Backdrop." Discussioll Paper51. World Bank, Water Supply and Urban Development Department, Wash-ingtoni, D.C. (First published in Spanish as "El Transporte en Bogota," RevistaCdmara de Comerrio de Bogotd 43 (Junte 19811: 151-84.)

Consultecnicos Ltda. 1971. Estudios de Normas Nlinimas de Urbanizzaci6n. Bogoti.Contraloria de Bogota, D.E. 1982. "Finanzas Ptiblicas Consolidadas de) Distrito

Especial de Bogota 1961-198] ." Boletin Fstadistico 2(5).Cortes, Mariliz, R. Albert Berry, and Ashfaq lshaq. 1987. Success in .Small and

Medium-Scale Enterptises: 7The Evidencefrom Colombia. New York: Oxford lUniver-

sitv Press.Dandeker, V. M., and Nilakantha Rath. 1971. Poverty in India. Bombay: Sanieck-

sha Trust.DANE (Departamenito Administrativo Nacional de Estadisticas). 1974. Statistical

Bulletin 279 (October).. 1975. Statistical Bulletin 293 (December).. 1976. Statistical Bulletin 301 (August).

DAPI) (Departamento Administrativo de Planeaci6n Distrital. 1978. Applicaix6n delas Normas Mlinimas de tUrbaniza•-i6n y de Servicios. Bogota, D.E.: Uniidad dleMejor-amienito y Coordinaci6n de Barrios.

. 1980. "Crecimiento de Bogota, D.E. Colombia. 1890-1980." Bogota.Datta. Gautam. and Jacob Meerman. 1980. Ifousehold Income or lhousehold Income

per Capita in We?fi7re Comparison. Staff Working Paper 378. Washingtoni, D.C.:World Bank.

D)NP (D)epartamento Nacional de Planeaci6n). 1972. (Cuatro Estrategias: PlanA'arional. Bogota, D.E.

- 1977. (Cuentas Regionales de Colombia, 1960-75. Bogota, D.E.Doebele, W'illiamil. 1975. "The Pirate Subdivision of Bogota." Disctissioll Paper

D75-1 1. Hiarvard University, Department of City and Regional Plannlilig, Cam-bridge, Mass.

Doebele, Williarn, Orxille F. Crimes. andJohanines F. Linn. 1979. "Participationof Beneficiaries in Finanicinig Urban Services: Valorization Charges in Bogota,Colombia." Land Economics 55 (1): 73-92.

Domencich, Thomas A., and Daniel McFadden. 1975. U!rban Travel Demand: ABehaviouralAnalvis. Amsterdam: North-Holland.

Ellickson, Bryan. 1977. Economic Analysis of Urban Hlousing Markets: A ANeuApproach. Santa Monica, Calif.: Rand Corporation.

1981. "An Alternative Test of the Hedonic Theory of Housing Markets'"lournal of Urban Economics 9 (1): 56-79.

EM,I'ISA (Empresa Metropolitana de Planejamenito de Grande Sao Paulo).1978. "Basic Results of the 1977 Origin Destinationi Survev." Sao Paulo. Pro-cessed.

Erikson, R. A., and M. Wayslenko. 1980. "Firm Relocation and Site Selection inSuburban Municipalities." journal of Ur-ban Economics 8 (1): 69-85.

Fawcett, Caroline S. 1984. "The Revenue Sharinig System in Colombia: A CaseStudy of Bogo1tA." In Johannes F. L.inn, ed., "Urban Finanices in Bogota,

304 UNDERSTANDIN(G THE DEVELOPING. METROPOLIS

Colombia." Discussioni Paper 39. World Bank, Water Supply and Urban Devel-opment Department. Washington, D.C. Processed.

Feibe]. Charles, and Alan A. Walters. 1980. Ownmhip and EffiriencY in Ulrban

Bluses. Staff Working Paper 371. Washington, D.C.: World Bank.Fields, Garv' S. 1975. "Rural Urban Migration, Urban Employment andi Under-

Employment and Job Search Activity in T.D(:s.`Journal of Devel)pment I Eonomics

2 (2): 165-87.-_____ 1980. How Segmented Is the Bogotai Labor Mlarket? Staff Wor-kinHg Paper 434.

Washington, D.C.: World Bank.(ln Spanish as "Que Tan Segmentado es elMercado Laboral in Bogota" Revista Cnmara de Comerrio de Bogotfi 41-42

[December 1980-March 19811: 293-353.)Fields, (;ar' S., and Nohra de Mar-ulanida. 1976. "IIntersector-al Wage Structtire

in Colombia." Economic Growth Center Discussion Paper 251. Yale Univer-

sity, New Haven, Conn.Fields, (Gary S., and T. Patul Schultz. 1980. "Regional Inequality and Other

Sources of Income Variationi in Colombia." Economicr Development and Cultural

Change 28 (3): 447-67.Follain, james. and Emmanatel Jimenez. 1985. "The Demand for HoLtsiligChar-

acteristics in Developing C(ountries." Urrban Studies 22 (5): 421-32.Follain, james. Gill-Chin L.im, and Bertrand Renaud. 1980. "The Demand for

Housinig in Developing Countries: The Case of Korea." Joutrnal of LUrban Ecou-

iOtniics 7 (3): 315-36.Fredland, D. R. 1974. Residential Mobility arid Home Purchase: A Longitudinal Per-

splctive on the Family Life Cycle and the Housing Market. Lexington. Mass: L.exing-

ton Books.Friedman,Joseph. 1975. Housing Location and the SupplY o] Local Public Services.

Santa Monica, Calif.: Rand Corporation.Fueintes. Alfredo, and Rodrigo Losada. 1978. "Implicaciones Socioecon6micas

cle la llegalidad en la Tenenicia de la Tierr-a Urbana de Colombia." C.ovuntura

Econ6inica 8 (1).Garcia, Jorge G. 1983. "Aspects of Agriculttiral Developmiienit in Colombia."

World Bank. Washington, D.C. Processed.(;arcia, Jorge G., and Sisir-a jayasuriva. Forthcoming. StabilitY, Adjustment, and

Groith. 7'he Colombian Economy since the 1960.s. Washington, D.C.: W'orld Bank.Gilbert, Alan, and joseph Gugler. 1982. Cities, Poverty and leveilopmnit: lrbaniza-

tiotl in the T'Ihir d World. Oxford, U.K.: Oxford University Press.(;lickman, Norman J. 1979. T1'he Growth and Management of the Japanese UJrban

Managemenl .System. New York: Academic.Greytak, David. 1984. "Revenule and ExpendituIre Forecastinlg in Bogota." In

Johannies F. L.inn, ed., "Urbani Finances in Bogota. Colombia." DiscuLssioInPaper 39. World Bank, Water Supply and Urban Development Department.

Washington, D.C. Processed.Griffith, Daniel A. 1981. "Modelling LUrban Population Density in a Multicen-

tered City.'Journal of Urban Economics 9 (3): 289-310.Hamer: Andrew M. 1981. "Houselholds and Housinig: Residential Mobility. Ten-

nLre Choice, and Space C01onsumTption in the Developinlg Metropolis." UIrbaniand( Regional Report 81-20. World Banlk, I)evelopment Economics Depart-

menet, Washington, D.C. Processed.

BIBLIOGRAPHY 30T5

1985. Bogotea' Unregulated Subdivisions: 7he AMiths and Realities of Incremmen-tal HJouising Construction. Staff Working Paper 734. Washingtoni, D.C.: WorldBank. (In Spanish as "Los Mitos v Realidades de la ConstrucciOn de ViviendaSuplementaria." RevistCO (amara de C(.nercio de Bogota 45-46 [December 1981 1.)

Hansen. Walter G. 1959. "How Accessibility Shapes Land Use." fournal of Ameri-catz Institute o'Planners 25 (1): 73-76.

Hansen, WVillard B. 1961. "An Approach to the Anahsis of Residential Exten-sion."JournalofRegionalScence 3 (1): 37-55.

1lanuishiek, Eric A., and John Quigley. 1978. "AII Explicit Model oi Intra-Metio-politan Mobility." Land Economics 54 (November): 411-29.

1979. "The D.,nainics of the Housing Market: A Stock AdjustmentModel of Housinig Consumption. " Journal of Urban Economics 6 (I): 90-1 1.

Hanushek, Eric A., and Byung Nak Song. 1978. "The Dynamics of PostwarIndustrial I.ocationT. Rezview of Econosnici and Statist?cs 6i0 (4): 515-22.

Hlarrison. David, Jr., and John F. Kain. 1974. "CLImulative lUrball Growth andUrbani Density Funiction. "Journal oJ Urban Econolmics 1 (1): 61-96.

Hienderson, J. Vernoni. 1988. Urban Developmenet 'Iheorv, Fact, il/nson. New York:Oxford University Press.

Helisher, D. A.. and [.. W. Johnsoni. 1981. Applied Discrete Choice Modelling. Lon-don: Croom Helm.

Hensher, D. A., and P. R. Stopher. 1979. Behavioral Trazel .11odelling. London:Croom Helm.

Hirschnian, Albert. 1958. The Strategy oj' Economnic Dezelopment. New lHaven,Conni.: Yale LIUniversity Press.

Hong Kiong Government Transport Departmenit. 1983. 7'rafirc and TransportDigest. Hong Kiong.

Hoover, Edgar M.. and Raymivond Vernon. 1959. Anantomy of a Metropolis. Cam-bridge, Mass.: Harvard LUniversity Press.

Hioyt, Homer. 1939. The Structure and GrowIth of Residentzal Neighborhoods in Ameri-can Cities. Washingtoni. D.C.: U.S. Government Printinig Office.

. 1966. W7here the Rich and Poor- Live: The L,ocation of Residential Areas 0(cCn-pied 1by the Hlighest and l.owest Income Fainiiies in American Cities. Washinigtoni,D.C.: lUrbani Institute.

Hluff, P. L. 1961. "Determination of' Intra-Ui-bani Retail Trade Areas." Real EstateResearch Program, University of California. Los Angeles, 1961.

ICT (Inistituto de Credito Territorial). 1955. U'na Politica de Viviienda Para Colom-biha. Bogota.

Ingramil, Gregory. 1982. "Land in Perspective: Its Role in the Structure of Cities."

In Matthew Cullen and Sharon Woolerv, eds., Wor1rld Congress on l.and Policy,1980, Proceedin7gs. [.exington. Mass.: D. C. Ifeath.

- 1984. Housing Dernand in the Developing Metropolis. Staff Working Paper663. Washingtoni, D.C: W'orld Bank. (In Spanish ais "Demanda de Vivienda enlas Metr6polis en Via de Desarrollo." in Revista Ce:nara de Conmercio de Rogotci45-46 [December 1981]: 82-121.)

Ingram, Gregory, and Alan Carroll. 1981. 'The Spatial Structure of L,atin Ameri-can Cities." Journal of Urban Economics 9 (2): 257-73. (In Spanish as "Estruc-tura Espacial de las Ciudades de America l.atina," Rem sta CI'imara (le Comerciode Bogota 43 [June]: 9-28.)

306 UNDERSTANDIN(; THE [)FVEI.OPING METROPOLIS

Ingram, Gregory K.,jose Fernando Pineda, and Alvaro Pachon. 1986. "El Estu-dio Urbano: Resumeni de Resultados e Implicaciones de Politica." In AlvaroPachon, ed., Lecturas Sobre Economnia Urbana. Bogota, D.E.: Fundaci6n Sim6nBolivar, Instituto de Estidios Politicos.

Inter-coiisuilta ltda. 1970. "Estudio de Organizaci6n Administrativa del Trans-porte Colecfivo de Bogoti." Bogota, D.E.: INTRA (Instituto Nacional de Tranis-porte).

International Labour Of(fice. 1970. 7a7nard Full Employment: A Program f2/r Go/am-6ia. (Geneva.

Jacobs. Jane. 1961. TI'le l)Peah and l.Vfe (If Greal American Cities. New York: VinitageBooks.

Jallade, Jean Piere. 1974. Public Expenditures on Education and Income l)Ditriulionin (olamhiba. Baltimore: Johns Hopkins University Press.

Jimenez, Emmanuel, and Douglas H. Keare. 1984. "Hoisinig Constimptioln andPermanent Inconme in Developing Counres: Estimates frorn Panel Data in ElSalvador."]cnlrnal of U'rhan Economics 15 (Marchi): 172-94.

Kemper, Peter. 1973. "The Location Decisions of ManiiLfctiur-inig Firms withinthe New Yor-k Metropolitan Area." Ph.D. diss.. Yale l.'niversity, New Haven,Conn., Departmenit of Economics.

King, Thomas. 1975. "The Demandc for Housinig: Integrating the Roles of Jour-ney to Work, Neighbourhood Quality and Prices." il N. Terleckvj, ed.. Hlowse-hold Produrtion and Consumplion. New York: Colutimbia Universitv Press.

Kiozel, Valerie. 1981. "'Travel Dlemand Models for Developing Countries: TheCase of Bogot., Colombia." Urban and Regional Report 81-26. World Bank,Development Economics Departmen t, Washington, D.C. Processedl.

1982. "Coomparisons ot Disaggregate D)emand Model Parameter StabilitVOver Time in a D)eveloping Counitry: A Colombian Example." ULran andRegional Report 82-2. U'oirld Bank, Development Economics Department,Washington, D.C(. Processed.

1986. "Tempor-al Stability of Transport Demanid Models in a ColombianCitv." In Behavicmral Research for T'rainsporl PoliNy. Proceedings of' the 1985 Inter-nationial Conference on Travel Behaviour. Amsterdam: North-Hollaid.

Kunziets, Simon. 1966. Modern Economir Growth. New Flaven, Conn.: Yale Univer-sitv Press.

Lakshmariani, T. R. 1964. "A Model for the Distributioni of Urban Activities: For-mulation, Evaluation, and Reformnuilation ." Paper presented at the seminaro)n Models of' Land Use [)evelopmen t, U niversity of Pennsylvaniia, Inistituteofl' Uibai Studies, October 22-26. Alan M. Voorhees and Associates, Washing-toni, D.C.

1965. "A Theoretical and Empirical Analysis of Intra-Urbani Retail L.oca-tion." Ph.D. dliss., Ohio State ULniversity.

Latorre, Emnilio. 1981. El fodelo de f)efmanda de 7irnspotle para Cali: 7assimt- Cali,D)escrpridn 1v Utldizarcin. Working Paper 10. Bogoti, D.E.: Corporaci6n CentroRegional (de Poblaci6n.

Laun.J. 1977. "'Analisis (le Urbanizaciones del I.C.T." In Colcultura, Vida Ulrbanay Urbanismo. Bogota, D.E.: Colcultura.

L.ee, Kvu Sik. 1978. "Distribution of Manutfactutring Establishments and Employ-menl in Bogoti at(. Cali." City Study \W'orkshiop Paper. World Bank, Develop-menit Economilics Department, Washington, D.C;. Processed.

BIBLIO(,RAPHY 30 7

. 1981. "Intra-Urban Location of Manufacturingjobs in Colombia."Jour-naloflUrbanEconomics9 (2): 222-41. (In Spanish as "Localizaci6n Intra-urbanadel Empleo Manufacttirero en Colombia," Revista Camara de Comercio de Bogota

43 [June]: 49-72.)- 1982. "A Mkfodel of litra-Urban Emplovment Location: An Application

to Bogota, Colornbia."Journal of Urban Ecoornewis 12 (3): 263-79. (In Spanishas "Modelo de Localizaci6n del Emp]eo Manuifacturero en Bogota," Revisla

Cdimara de Comercio de Bogotai 43 [Junie 1981]: 29-48.)- 1989. The Location of Jobs in a Developing Metropolis: Patterns oJ Growth in

Bogota and Cali, Colombia. New York: Oxford Univer-sity Press.Lee, Yoon Joo. 1979. "The Citv Study: The Available Data." Vol. 2. Urban and

Regionial Repor-t 79-"13. World Bank, Development Economics Department,Washington, D).C. Processed.

. 1981. "The City Studly: The Available Data." Vol. 3. Ul-ball and RegionalReport 81-15. World Bank, I)evelopment Econiomics Department. Washinig-toni, D.C. Processed.

1985. Thhe Spatial Str?cture ofthe fetropolitan Regions of BraziL Staff Work-

ing Paper 722. Washington, D.C.: World Bank.Leone, Robert. 1971. "Location of Manutacturing Activity in the New York Met-

ropolitan Area." Ph.D. diss. Yale Universit., New Hlaveni, Conn., Department

of' Economics.Lermain, S. R. 1975. "A Disaggregate Behavioral Model of Urban Mobility Deci-

sions." Ph.D. diss.. Massachutsetts Institute of Technology. Cambridge, Mass.- 1977. "Location, Housing, Automobile Owwnership and Mode to Work: A

Joint Choice Model." Transportation Research Board Record (Washingtoll, D.C.)610: 6-11.

Lermani, S. R., and M. Ben-Akiva. 1975. "A Behavioral Analysis of Automobile

Ownership and Modes of Travel." Washinigtoni, D.C.: U.S. Departmelit ofTransportation. Office of' the Secretaryv and the Federal Highwav Adniinistra-tion. Pr(ocesse(l.

Levinison, Herber-t S. 1978. Characteristics oJ Urban 7iransport Demand: A Handbookfor 'lransportation Planners. Washingtoni, D).C.: U .S. Department of Transportatioln.

L_im, (Gill-Chin, Jarnes Follain, and Bertrand Renaud. 1980. "Determinants of

Hlome-O(wnership in a Developing Economy-Case of Kiorea." Urban Sudies 17

(1): 13-23.Linn, lohannes F. 1976a. The Di.stibnutive Effecti of Local (;ove-nment Finances in

Colombia: A Review oj Evidence. Staff Working Paper 235. Washington, D.C.:

World Bank.1976b. "Public Utilities in Metropolitani Bogota: Organieation, Service

Levels and Financing." Ulrban and Regional Report 76-9. World Bank, Devel-

opment Econiomiics Departmenit, Washington, D.C. Processed.1979. "Thle Incidenice of Urban) Property Taxation in Colombia." In Roy

W. Bahl. ed., 7htI 7axalton of LUrban Properlt in L.es Devealoped Countries. MadisonI.Wis.: University of Wisconisini Press.

1980a. "Property Taxation in Bogota, Colombia: An Analysis of PoorRevenue Perforimianice." Public Finance Quarter&l 8 (October): 457-76.

1980b. "UIrban Finances ifi Developing Countries." In Roy WV. Bahl ed.,

tLTrban Government Finance: Emerging 7rendi. Beverly Hills, Calif.: Sage Puiblica-

tioIls.

o)8 L NI)DERSIANI)INC TDI-IT lE DV [.()PI . MPFNrR(O)POL S

- 1981a. "Administraci6n v Firaiacianien1to de Servicios Pfiblicos en elDistrito Especial de Bogota." Working Paper 11. Bogota, D.E.: Corporaci6nCentro Regional de Poblaci6n.

. 1981b. "Impuesto Predial en BogotA, Colombia: lin Analisis de BajoRendirrienito de los Ingresos por este Concepto." Revzista CAmara de Comercio deBogolad43 (lu-e): 73-90.

- 1 984a. "Administrationi and Financing of Public Services in Bogota,D.E.: Sumi imiary of Findinigs.' In Johanines F. Linn., ed., "l!rban Finances inBogotai, Colomnbia." Disciussioni Paper 39. World Bank, Water Supply andiUrban Developmenit Department. Washiurgton, D.C. Processed.

- - (ed.). 1984b. "LUlrban Finances in Bogota, Colombia." Disclissioll Paper39. World Rank, Water Supply and Urban Development Departmenit. Wash-ingtonl, D.C. Processed.

L.osada, Rodrigo, an(d Hernanido G6mez. 1976. L/a Tierra en el Mercado Piralt deBogoti. Bogoti, D.E.: Fedesarr-ollo.

L.osada, Rodrigo, and( L.. Pinilla. 1980. I.os Barrios legales de Bogota. Bogota, DE.:Ceuitro de I rvestigaciem y Desarrollo de Pedro G6mez v Cia.

L.owrv, 1. S. 1964.A Alodelb?fAfe!rapolis. Santa Monica, Calif.: Rand Corporation.L.ubell. Harold, and Douglas McCaliim. 1978. Bogota: Ulrban D)evelopment and

EnplaVirmenl. (Genevas: Inteinationial L.abour Office.Maddala, C. S. 1983. Limited Dependent ond Qualitaliv!e 'Vanables in Fconometrics.

Cambridge: Cambridge University Press.Malpezzi, Stephen, and Stephen K. Mavo. 1987. "The Demanid for Housilig in

Developinig Countries: Empirical Estimates from Hotusehold Data." EconomicDevelopment and Cultural Change 34 (4): 687-721.

Malpezzi, Stephen, Stepheni K. Mavo, and David J. Gross. 1985. llousingDemandin Dev!eloping (;ountries. Staff Working Paper 733. Washlinigton. I).C.: WorldRank.

Manski. Charles F., and S. R. L.erman. 1977. "The Estimation of Choice Probabil-ities ftoiom Choice Based Samples." Econuonetnra 45 (80): 1977-88.

Mayo, Stephen K. 1981. "Theory and Estimation in the Economics of HousingDemand." /ournal of Urban Fconomics 10 (1): 95-1 ]6.

Mazunidar; Dipak. 1976. "The Urban Informal Sector." UWorld Development 4 (8):655-79

- 1981. Urban labor Market and Income!istnbution: A Study (ojMalaysia. NewYork: Oxford Universitv Press.

McDonald, john F. 1981. "Capital-l.and Substitutioni in Urbani Housing: A Suir-vey ot Empir-ical Estimates." Journal of Urban E(onomicsi 9 (2): 190-211.

McFadden, Daniel. 1973. "Conditional l.ogit Analysis of Qualitative ChoiceBehavior." In Patil Zarembka, ed., Fmnlie-s in Econometrics. New York: Aca-demic.

- 1974. "The Measuremenit of LUrbani Travel Demand."-Journal of PublicEconomics 3 (4): 303-28.

1976. "The Theory and Practice of Disaggregate Demand Forecastingfor V'arious Modes of Urban Transportationi." Urban Travel Demand Forecast-inig Project. Working Paper 76-23. Berkelev: Institute of Transportationi Stud-ies, University of California.

BIBLIOGRAPHY 309

McGregor, Alan. 1977. "Inter-Urban Variations in lJnemployment Duration: ACase Study." Urban Studies 14: 303-13.

McKay, Harrison, and others. 1978. "Improving Cognitive Ability in ChronicallyDeprived Children." Science 200 (4339): 270-79.

Mills, Edwin S. 1967. "Aggregative Model of Resource Allocation in a Metropoli-tan Area." American Fconomic Review 57 (2): 197-210.

1972. Studies in the Structure of the Urban Economy Baltimore: Johns Hop-kins University Press.

Mills, Edwin S., and Katsutoshi Ohta. 1976. "Urbanization and Urban Prob-lems." In Hugh T. Patrick and Henry Rosovskv, eds., Asia's New Giant: How0 theJapaneseEconomy Works. Washington, D.C.: Brookings Institution.

Mills, Edwin S., and Byung Nak Song. 1979. Korea's LUrbanization and Urban Prob-lems, 1945-1975. Cambridge, Mass.: Harvar-d University Press.

Mills, Edwin S., and Jee-Peng Tan. 1980. "A Comparison of Urbani PopulationDensity Function in Developed and Developing Countries." UIrban Studies 17

(December): 313-21.Mincer,Jacob. 1974. Schooling, Experience and Earnings. New York: Columbia Uni-

versity Press.Mohan, Rakesh. 1979. Urban Economic and Planning Models. Baltimore: Johns

Hopkins University Press.1980. The People of Bogoli: W'ho They Are, Whal 7hey E'arn, Where iThe Live.

Staff Working Paper 390. Washington, D.C.: World Bank. (In Spanish as"Poblaci6n, Ingresos y Empleo en Una Metr6poli en Desarr-ollo," RevistaCdmara de Comercio de BogotM, 41-42 [December 1980-March 19811: 9-116.)

- 1981. T'he Determinants of Labor Earnings in Developing Metropolises: Esti-mates from Bogotei and C.ali, Colombia. Staff Working Paper 498. Washington,D.C.: Wlorld Bank. (In Spanish as "Determinantes de los lIgresos l.aboralesen Una Metr6poli en Desarrollo," in Alvaro Pachon, ed., 1986, Lecturas SobreEconomia Urbana, Bogota, D.E., Fundaci6n Sim6ri Bolivar. Institutto EstudiosPoliticos.)

1984. An Anatomy of the Distribution oJ Urban Income: A 'hlae of 7wo Cities inColombia. Staff Working Paper 650. Washington, D.C.: World Bank.

- 1986. IWo,*, Wages, and Kielfare in a Developing AMetropolis: Consequenres ofGrowth in Bogote, Colombia. New York: Oxford University Press.

Mohan, Rakesh, and Nancy Hartline. 1984. 7'he Poor of Bogot: W'ho They Are, WhatThey Do, and Where T'hey Live. Staff Working Paper 635. Washington, D.C.:World Bank.

Mohan, Rakesh, and Richard Sabot. 1988. "Educational Expansion and the Ine-qualitv of Pay: Colombia 1973-1978." Oxford Bulletin of Economics and Statistics50 (May): 175-82.

Mohan, Rakesh, and Rodrigo Villamizar. 1982. 'The Evolution of Land Values inthe Context of Rapid Urban Growth." In Matthew Cullen and Sharon Wool-er,v eds., 14hrld Congress on Land Policy, Proceedings. Lexington, Mass.: D.C.Heath. (In Spanish as "La Evoluci6n de Los Precios de la Tierra eni el Con-texto de un Rapido Crecimiento Urbano," in Alvaro Pachon, ed., Lecturas

Sobre Ecoonomia Urbana, Bogota, D.E.: Fundaci6n Sim6n Bolivar, Instituto de

Estudios Politicos.)

310 U'NDERSTAND)ING TH t)E VEL.OPING IMN TROPOLIS

Mohan, Rakesh, M. Wilhelm Wagner, and jorge Garcia. 1981. Measuring UrbanMalnutrition at1d Poverty: A Case Stud'y of Bogotet and Cali, Colombia. Staff Work-ing Paper 447. Washington, D.C.: World Bank.

Mor rissoii, Christian. 1984. "Income Distribution in East Europeani and WesternCoun tries." fotrnalof'Comparati,e Economnics 8 (2): 121-38.

Moses, Leon, and H. F. Williamson), Jr. 1967. "I'he Location of Economic Acth-ity. " Amric7ian EconomnirReview 57 (2): 211-22.

Muth, Richard F. 1969. Cities atnd Housing. Chiicago: Univelrsity of Chicago Press.Nelson, Richaid R., T. Paul Schltitz, and Robert L.. Slighton. 1971. Structural

Chanige in a D)eveloping Economy: Colombia s Problems and Prospects. Princeton,N.J.: Princeton University Press.

Ojha. P. D. 1970. "Configuration of Indiani Poverty: Inequality and l.evels oflearning." Reserve Bank of India Butletin 24 (Janiary): 16-27.

Pachon, Alvaro. 1979. "Urban Structiure, Modal (Choice, and Auto Ownership inBogota. 1972.' Citv Studv Intermediate Paper ino. 31. World Bank, Develop-meiit Econiomilics Departnmenit, Washington, D.C. Processed.

, 1981a. "El Autom6vil en Dos Metr6polis del Tercer Mundo." Revi.staCdrmara de Cormercio de Bogotdi 45-46 (December) : 188-222.

1981 b. "El Impacto Redistributivo de la Intervenci6n del Gobierno en elTranisporte Urbano." Revista Ctimam de (Comerio de Bogota 43 (June): 185-216.

1981c. "Situaci6n del Transporte en Cali y Bogota Decada de los Set-eta." DocuLimenito del Triabajo 18, Corporaci6n Centro Regional dePoblaci6n, BogotLa.

1987. "Urban Transpor-t Policy: Colombia." In George Tolley and VimiodThomas, ecis.. 7'he l'conoinics of Urbanization and CUrban Policies in DevelopingCountries. Washington, D).C.: WVorld Banik.

Paredes, L.. Ricardo. 1984. "Colombia's Urban l.egal Framework." In "Housingand Ulr,bain Developmelnt in Bogota: The Instituitionial Backdrop." DiscussionlPaper 51. World Bank, Water Supplv and Urban Development Department,Washinigtoni, D.C. Processed.

Pendletoni. W. C. 1963. "The Value ol Highway Accessibility." Ph.D. diss., Univer-sitv of Ch icago.

Pfeffecminiatii ,uy 1.R, and Richard C. lVebb. 1979. The Distrniution of Inciome inBraziL Staff Working Paper 356. Washington), D.C.: World Bank.

Pineda, Jose Fernanido. 1981a. "Residential Location Decisions of Multiple*Worker Households in Bogota, Colombia." Ph.D. diss., Univelsity of Califor-n1ia, Berkeley, Department of City and Regional P'lannling.

1981 b. "Residential l.ocation Decisions of Multiple Worker Householdsin Bogota, Colombia." (In Spanishi in "Decisiones de l.ocalizaci6n Residenicialet' Hogares cont Multiples Trajabadores en BogotA, Colombia," Revisla Ctimarade Comercio de Bogolci 45-46 [December]:163-87.)

Pinteda,Jose Fernanido, and Sonia Rodriguez. 1982. La Pobkl.ricin de Bogvf a. Work-ing Paper 26. BogotA: Corporaci6n Centro Regional de Poblaci6n.

Polinsky, A. Mitchell, and Daid M. Elwood. 1979. "An Empirical Reconciliatiollof Micro and Grotiped Estinmates of the Demanid for Housing." Review of Erco-noritrs and Statistics 61 (2): 199-205 .

Potter; joseph, and Myriam Ordoniez. 1976. 'The Completeness of Enumerationiin the 1973 Census of thie Populationr of Colombiat." Population Index 42 (Jtily):377-403.

BIBLIOGRAPHY 311

Preston, Samuel H. 1979. "Urban Growth in l)eveloping Countries: A Demo-graphic Reappraisal." Population and Development Review 5 (2): 195-215.

Psacharapoulos, George. 1980. "Returins to Education: An Updated Internia-tional Compatison." In Timothy King, ed., Education and Income. Staff Work-ing Paper 402. Washington, D.C.: World Bank.

Quigley,john M. 1976. "Housing Demand in the Short Run: An Analysis of Poly-tomotis Choice." Explorations in Economic Research 3 (1): 76-102.

- 1979. "What Have We Learned About Urban Housing Markets?" InPeter Mieszkowski and Mahlon Straszheim, eds., Current Issues in Urban Eco-

nomirs. Baltimore: Johns Hopkins University Press.Renatid, Bertrand. 1981. Vational tJrbanization Polio' in Developing (Countries. New

York: Oxford University Press.Reutlinger, Shlomo. and Marcelo Selowskv. 1976. Malnutrition and PovertY: Mag-

n.tude and Polico Options. Baltinmore:Johns Hopkins UIIniversitv Press.

Ribe, Helena. 1979. "IInconme of Migrants Relative to Non-Migrants in Colombia:An Econiomic Analysis of the 1973 Census Sample." Ph.D. diss., Yale Uiniver-

sity, New Haven, Conns., Departmient of Economics.Ruiz Betancourt, Nlario Christina. 1978. "El Situado Discal." Investigaci6rn de

Grado para optar al Titulo de Economista. Universidad de Gran Colombia.Proc essed.

Schmeniner, Roger- W. 1973. "Citv Taxes and Industry l.ocation." Ph.D. diss., YaleUnhiversity', New Haven. Conn., Departmenit of Economics.

1982. Making Bzusiness Location Decisions. Englewood Cliffs, NJ.: Prentice

Hall.Schultz, T. Paul. 1969. "Population Growth and Internal Migration in Colom-

bia." Ranid Cor porati on Memorandum 5765, Santa Monica, Calif.Schultz, T. W. 1961. "'Investmenit in Humnani Capital." American Economic Reviezw 51

(1): 1-17.Selowsky, Marcelo. 1979. W1ho BenefitsJfrom Government Expenditure? A Case Study of

Colombia. New Xork: Oxford Ulniversity Press.Sethul-amari, S. V'. 1976. "Towards Definitioni of the Informal Sector." Research

Working Paper. International I.abour Office, WVorld Employment Program,Genieva. Processed.

Slighton, Robert. 1987. Relative W1ages. SkillShortages, and Changes in IncomeDistri-

buntion in Developing Countries. Washingtoni, D.C.: World Bank.Smith, James P., and Finis Welch. 1977. "Black-White Male Wage Ratios 1960-

1970." American Economic Rezvieun67 (3): 32.3-38.Stevenson, Rafael. 1984. "Housing Progr-ams and Policies in Bogota: A Historical

and Descriptive Analysis." In "Housing and Urban Development in Bogota:The Institutional Backdrop." Discussioni Paper 5]. World Bank, Water Supply

and Urban Development Department, Washington, D.C.Struvk, Raymond)J., and Franklin J.James. 1975. Intrametropolitan IndustrialLoca-

tion. L.exington, Mass.: Lexington Books.Struvk, Ravmoiid J., and S. A. Marshall. 1975. "Income and Urban Home Own-

ershiip." Rervieuw of Economics and .Statistics 57 (1): 19-25.Sukhatme, P. V. 1978. "Assessment of Adequacy of Diets at Different Income

levels." Economic and Politiral Weeklh 13 (31, 32, 33): 1373-85.Tabares, Hlenr. 1979. "Poblaci6n de Cali: Series Hist6ricas y CaracterisLicas."

PIDF(:A, Document 3. Planeaci6n Mttnicipal, Cali. Processed.

312 UNDERSTANDING THE DEVELOPING METROPOLIS

Thomas, Vinod. 1985. Linking Macroeconomic and Agricultural Policies for Adjust-ment urith Growth: The Colombian Experience. Baltimore: Johns Hopkins Univer-sity Press.

Tolley, George S., and Vinod Thomas, eds. 1987. The Economics of Urbanizationand Urban Policies in the Developing Counties. Washington, D.C.: World Bank.

Transpoconisult. 1976. Lagas: Metropolitan Area Transport Study. Lagos, Nigeria.Urrutia, Miguel. 1969. The Development of the Colombian Labor Movement. New

Haven, Conn.: Yale University Press.- 1981. 'Evaluaci6n del Sistema de Transporte Publico en Bogota." In

Fedesarrollo, Evaluaci6n del Transporte Urbano de Bogota. Bogota, D.E.: Fedesar-rollo.

- 1985. WVinners and Losers in Colombia's Recent Crowth Experience. Baltimore:Johns Hopkins University Press.

U.S. Department of Commerce. 1979. Annual Housing Survey: 1977. Part 0.Housing Characteristics of Recent Movers. Washington, D.C.: GovernmentPrintinig Office.

Valenzuela, Jaime. 1970. "El Papel de la Vivienda Popular: El Caso de Colom-bia." Centro de lnvestigaci6nes en Desarrollo (CID). Bogota, D.E. Processed.

Valenzuiela, Jaime, and George Vernez. 1974. 'Construcci6n Popular y Estruc-ttia del Mercado de Vivienda: El Caso de Bogota." Revista SIP, no. 31 (Sep-tember).

Valverde, Nelson. 1978. "The City Study: The Available Data." Vol. 1. Urban andRegional Report, no. 70-6. World Bank, Development Economics Depart-ment, Washington, D.C. lProcessed.

Velasco, Juliania, and Gilberto R. Mier. 1980. "Valores v Caracteristicas de laTierra en Cali." Planeaci6n Municipal, Cali. Processed.

Vernez, Georges. 1973a. "Bogota's Pirate Settlements: An Opportunity for Met-r opolitan Development." Ph.D. diss., University of California, Berkeley.

1973b. "Pirate Settlements, Housing Development by IncrementalHousing Development and Low Income Housinig Policies in Bogota, Colom-bia." New York: Rand Institute. Processed.

1976. "Traslados Residenciales de Migrantes de Bajos Ingresos." Bogota,D.E.: Corporaci6n Centro Regionales de Poblaci6n (CCRP). Processed.

Villamizar, Rodrigo. 1981. "Land Prices in Bogota between 1955 and 1978." In J.Vernon Henderson, ed.. Research in Urban Economics. Greenwich, Conn.: JAI

Press. (In Spanish as "Los Precios de la Tierra en Bogota [1955-1978],"Revista (amara de Comercio de Bogold 41-42 [December 1980-March 1981]:117-70.

Wagner, M. Wilhelm. 1984. 'Fhe Market Prices and Assessed Values in the IJrban LandMarket in Bogotd, Colombia. Staff Working Paper 651. Washington, D.C.: WorldBank.

Walters, Alan A. 1979. Cost and Scale oJ Bus Services. Staff Working Paper 325.W\'ashinigton, D.C.: World Bank.

Webb, Richard C. 1977. Government Policy and the Distribution of Income in Peru1963-1973. Cambridge, Mass.: Harvard University Press.

Weber. Adna F. 1963. The Growth of Cities in the Nineteenth Century Ithaca, N.Y.:Cornell University Press.

B[BLIOGRAP11V 11 3

WestLi. Richard. 1980. '"Travel Patterns in Bogota." City Studv IntermediatePaper 11o. 45. World Bank, Development Economilics Departnienit, Washing-ton, D.C. Processed.

Wheaton. William. 1982a. "Urban Residential Growth unider Perfect Foresight."joural of Urban ElConomicv 1 2 (1): 1-2 1.

- 1982b. "Urban Spatial Development with Durable and Replaceable Cap-ital."Journal ol Urban Economi.cs 12 (1): 5.3f-67

White, Michelle J. 1977. -On CUmulative Urban (Grow'th and Urbani DensitvFunctions."]Jornal of lUrban Economics 4 (1): 104-12.

5 Aiesnier, (Guille:rmo. 1980. "The History of Land Prices in Bogota between 1878and 1978." (In Spanish as "Cieni Afios de Desarrollo Hist6rico de los Preciosde Ia Tierra en Bogot`i," Revis/a Cemanzra de Comeniro de Bogoal 41-42 [Decem-ber]: 171-208.)

Wilbur Smith and Associates. 1974. "Urban Transport Policy and Plannililg Studyfoir Metr-opolitan) Kiuala l.umpur." KLuala Lumpur, Malavsia. Processed.

1976. "Hong Koong Comprehensive Transport Study." Hong Kiong. Pro-cessed.

1979. "Masterplan for Metropolitan l.agos.' Lagos, Nigeria. Processe(l.Witte, Anin D.. Howvard Sumlika, and Homer Ereksoni. 1979. "An Estimnate of a

Strticttiral Hedonic Price Model of the Housinig Market: An Application ofRosen's Theoryv of Implicit Markets." I L'conooinca 47 (5): 1151-73.

Wolff, Juergen 11. 1984. "BBudgetinig and Investmenit Planniniig in Bogota. InJohanzies F. L.i nn, ed., "Urban Finanices in Bogouta Colombia." DiscussionlPaper 39. World Bank, Water SuppIl and Urban Development Department.Waslhington. D.C. P'rocessed.

World Bank. 1977. W;orld Bank Atlas. Washington, D.(C.1990. Colotnbia: .Soriat Programsfor the Atlrlevalion of f'ove-rte Washinlgton,

D.C.WVorldl Bank andJose Sokol. 1984. Colomnbia: Economic Development and Polircy under

C.'hanging Conditions. World Bank Country Study. Washington, D.C.: W;orldB1ank.

Zahiavi, Yacor. 1979. "Urban Travel Patterns." Washingtoni, D.C. World Bank,Economic Development Institute. Processed.

Index

African cities, 24-25 location in, 121-22. 124-25, 1226, 128,Agencies. See Poubfic service agencies 129; firim size in, 129, 1 3 1: (t)p in, 38:Agglomeration economies: city expansioni growth characteristics of, 15-1 i: holie-

and, 19, 20: for new firms, I 3(0, 141-42 ow-nershiip in, 167-68; houiseholdAgrarian Credit Banik, 147 mobilityj in, 114. 162-63, 170: hotusinigAgicultuTr-al sector: decenitralization and, patter-ns in, 67-68: inicome distribution

36: oUtputI growtil in, 97; primary activ- in, 77-78, 79, 81, 83, 101, 103, 112;ities in, 19: prosperity in. 13-14, I 17: inifrastrticttnie expenditures in. 256.utansition fromii, 75-76 258; laboir force in, 89-91 ; laboir mar-

Ahmedabad (Indlia), 239 ket segmenLtationi in, 1 17, 1 18, 275:Alliatice fir Progress, 148 land use in, 42 -43: la id v-aloes iii. 58,Aiiato, Peter, 6. 79 6(0. 70-71 : malnutrition in. 84-85;Ananid, Snidhir, 81 , 95 n5, 10)4 inanilfacttirinig sector in, I0. 4 3, 123,Aiidersrsn, john E., 37 o5 128-29. E3()32, 14.3; /97Kheml)losioentA.rdila, Aminp.r de. 1.56 in, 97-98; oetuipatiional distribution in,Asabere, Patil K., 24-25 11t. 103: peripheral developmient in,Autitoiobile owner shipi.S. &eCar ownership 25; phiysical char acteristics of, 40-4 1.S1ax,ho pasJurlarnuni'si (parliamjentarv 42; pzuwau sobdivisions in, 152-53; pup-

grants), 258 olanon densities in, 39(t4, 43, 5(0-51,55-56f, 63; prior research on, 6-7: poh-

Banco Central Hipotecario (BU H). 149, lic hotising pro)grams in., 148-49; p)b-1511, 259 lie serNices in, 94, 234-35., 259-60, 266i-

Baico rle la Rep&ihlica (Central Bank), 67; residential lot sue in, 70: returns to149J 238. 259 eedl cariiii in, 1 06.-107. I J lII>2; re%enicn

Bangalore (India). 168 s1ilmces in.2329244-45,2 4 8-50,254-55;Bankiig Regtilatorv Ageuc.eiice Superin- transpoi oljon services in, 187. 189. 190),

tenideticia Barncaria 222. 22'3-24, 285-86; travel eJchiracteris-Banweol (Korea), 143 tics ot, 20'3, 20)4, 21:3, 214; tirmio ri meni-Barco governimient (1987-90), 13 bershii) in, 114 See alsoii:n (cenitialBarrios: and densit y gradienits, 56, 72 n2; bhsiniess district): Govei mlg1zenit, (listri(ct

in stirsey dat. 2"91. 2959-6, 299, :30)) Bonmba, 4(), 43, 239n7, 3()(1 n8 Bourguignoin. Frarvmois, 1 )5

Barnos pira(as. See flra/n subdivisions Bowden, Nlai tvn f.. 21Rt:H (Banco Central H-ipotecario), 149, Bridges. 256

150, 259 Britumier, Kard, 42Becker (Gary S., 1(04 BUdgets, district, 235-36. 237. 258-59Belo Horrionite, 30.,5fi Buenos Air es, l88Beni-Akiva a.I. E., 214 Buses: fare strtictiire of, 195, 223; gover-n-Berrv, R. .Mbert, 6, 75 ment subsidy cit, 189, 191, 193. 1 94BogOt:i: beer cornstimptioni in, 252, 254; 195-96, 198; operating costs tf. 194;

car ownership in, 191, 214; Citv Studs's (oi iiershlip) of. 189; pl ofitabili R, of. 196.selection of, 5-6; employnient 198; seaLs pe- 1,000 )residenLs on. 189,denisities in, 6.3-64, 279; employment 191. 222: sizes of, 189; tiavel peaks oni,diversification in. 43. 123: emproslo ent 211 -1 5; travel timles on, 2(5, 22(

3 5

34 I 6 NO:EKS rAN'I)f NO. tt E I)IEVE t Pi NC; %V FR FCPo I.S

JulwtIIs: (I riveis' earn ings oni, I 94, 22'3: r1' (C( orpi - i6'a n Filaiaciera cl e [ranis-tiae stir iicitre of, 187, 1 95; noinsuhsidv porte). 194of, 224; ownership of, 189: priofitabiliri ( hapineo co inimer-cial distri( I, 41, 130,ol, 196. 198, 207; seats perI 1,000 resi- 144 r2dents on. 189, 191: servi( e teqnenc'v Chicago. 4(, 56of. 188: travel peaks on. 214-15: unavel Children, pool. 87, 94iiimrsiiii, 20)5. 220 (Cities: characteristics of, 19-2(: deceiitral-

izatiol of, 21-22; in deseloping coon-(Cia tIe Allot Iro v \'ivieiida (Savings andll tries, 18: development regolatrities il,

loa-nn .sso(ciatioins), 149 14-15: emptx(ei ttenhi, 23,. 3(0,'37 n5:(Cap t(le Preisi6iii Snila!, 235 N;NP TTilMrinp isi insi 27'; lIrge-s(ale(;aja (Ie Viienida Popti h., 235 modtels ofi, '3: ini Latin America, 5; miLl-CalIclitta, 44,4.3 ti(nccited,c30, 37 'L; pilicvy etom-(:al: cal owilershilp ill, 191,214; Cuit nieiendations loin 71, 2 273,281,282;

Studv's selection of, 7-8; emplovinlellt PoPIlItaion densities in, 27, 30; poptila-decentializalioni in, 63, 121-22, 123; tinio growth in 1- 2,8-9, 26-27; povertytiil ploimelnt densities ini 279: employ- iit, 2-3' spatil divisiolls if, 50, 78-79.

iclit lixiersiiication in. 43: fi rii size in. See also Bogota:; Call; i:iBD129, 131: 11 ii i, :38;growth ch a-actenis- (itr Plaitiiiiiig Office (Plaieacii3n Munini-tics oif, 15: hoiliiowniership in. 167-(i8: pal, (Cali), 7liotistfiiolf miibilit' in, f 62-63, 1704: City r-ings: einploiymeiit distribution by,iwico iiti distribUotiOin ii, 77-78, 79, 81, 124. 126: hiouseholId size by, 55; income112: labor iiiar ket swgiclianltion in, cfisu illuion byv 51, .52; land valcies by.118, 275; laitic valties il, 58, 70-7 1; niial- 58: population dellsides by, 54-51; inlii itiolii i, 84-85; mtarni factlUrillg Sct- sUr v data, 291: 'Theil indexes ior, 81

tor ii, 413 128: physical cliat acteristics City sectois: emplcylilelnt density by, 63-of, 42; pioipilaion denlsities ii, 56; (34: rmpiymentn distributionoit by 24-

pl)lllit sersices in, 94, 2h3; residential 26; ilicoiie distribtion by. 51, 54, 63,lot size ill, 70: IreXeInue soLu-ces ill, 239: I 112: laud( Naltie gradieuits bv, ;5; ccct-traits5liO ri ii services in, 189, 191 pationilal distribhition by, 1011 103: P)p-222, 285-863: travel patter iis il. 21 4: tIlation dciisi ties by, 5)-5 (i 65; I ettlI is

Ullinin Ult eiihelsiip Hi, 1 14 lii ecdfti-aon . I 1 :i; ii sill vevx (ata,Caltii COIisMillpti\Oi anld lintlitioll, 84 291: l'hrtieil iidxes for, 81Car owiiesluiip: per liottsefiold. 21.3.224: (2 tv su lt- ttLre: defilued, 20: and einplo-

per resi deii t, 191: p oblic transportatiml ni eii t location. 1 22: an d tralnspotltation

atnd,f,214, 221, 222-2:3, 224, 273i; taxes svsiliii, 16. 1 85, 2844 aniI n avel miiode,in, 198. 248; aiid trIivel timies, 2()4-2145: 2155-164 221-22

aii( utip Laws, 2 1 1, 213,222 ( ii Sti(tv ceisits (1 973) data for, 288-Carroil. Alan. 24i, 156 89; (celayed publication of, 8; hotise-

(:artagenia, 2:39 hol(l stirvev data for, 2894-3, 295-96,B Bt) (cell business (lisu-ict): lxoiditlaries 299, '3( n6ti, 34444 118: maniiafactitrinig

of, 22: rnillIloment densities in, 22- cfita for, 294-95; iof(lelinig approacf

23, 63-;4, 279: emploielltn liocationi by, 3. 274: objective of, 2, 3-4; aiid

in.6ti3, 71. 121-22, 123, 2ti. 128, 247, rither stuidies. 117, 118; Ie(oillmmelclda-

2204: land valile pat terns ii, 58. 61, 64, tiois by, 71, 181 -83, 272,27'3,281,282:65, 647 , 69. 74, 278: stubcetitel-s' 'clomple- research p ior to, ti-7; adi(i researchi

titiotl itti, 41. 69, 1:31, 144. 144 n2 transferahility', 5, 14-15,275; selection(:BR (( twid bhi th rate), 9 r i lenia ot, 5-6; social sectirity Ia ta oir,c:r:1R t(Corporacin Cenritral Regiomtil (le 2913: toipics oi, 1 5-1(, 17 n3

Pobla(cid) , 7. 29:4 (Citdad kernmedy hicusing program, 148(:enital Banik (Banric de Ia Rep(iblica) . (:offee prices. 8,1 f),11, I 3-14, 76, 117,

149, 238, 259 27:3CCeIutnal bLsiIes distict See CBD Ctlecl)ivos (collective taxis), 188, 189, 191.(:enln'r Ilitel'liacioiiala 123, 128 .Sn alhn Btises; ffitselat

Certification, 14)5. 1446. See alseo dccaiorll. (:Colilbia: tar ossnelship in, 19 1, 198;retilnrIs tt demogr-aphic Wransitioll in, 8-9, 14,

(C'esaniaas (severance paW intds!), 157, 1 1 7: educationi expalision in, 99-1(01,18:3 Ti 1 144; el lploymienit structutre in, 96-97;

INDEX 317

*DP growth rate in, 38; homeowvner- DAPD (Departamento Administrativo deship in, 181; hotisehold mobility in, Planeaci6n Distrital), 193, 234, 237163, 170; housinig programiis in, 146, DATT (Departamento Administ-ativo de147-48; income distribution in, 75; Transporte v Tridnsito), 193, 234industrializationi of, 74-75; investment Danta. Gautam, 95 n5rates of return in, 160, 183 n3; under Decentralization: consequences of, 36:Ileras government, 10: under 1l.per described, 21-22; measurement of,government, 11; 1980s economn in, 8. 22-23. 30. See also Employment decen-14, 17 il;l1970s economy in, 13-14, 76, t-alization; Population decentralization

97-98, 117, 273-74: ntutritional Delhi, 283requirenment in, 84; under Pastrana Density gradients: exponiential functionalgovernment, 10-11; primacy index of, form for, 22-23. 30, 37 n5; and inter-38; prior research on, 6-7; reveoine cept densities, 30, 32sources in. 198-99, 239, 244, 260-64: Departarnento Administrativo de Pla-transportationi services in, 189. See also neaci6n Distrital (DAPD), 193, 234, 237Bogota; Cali Departamnento Administrativo de Trans-

Colombian Federation of Urban Trans- porte y Transito (DnAT), 193, 234portation (Federaci6n Colombiano de Departamento Adininistrativo NacionalTr-ansporte Urbano. FcoLTKAN's), 193 de Estadistica (DANE), 7, 76-77, 288,

Colombian Institute for Family Wtelfare 289-90, 293-94, 30)0 n3. See also WNorld(Instituto Colombiano de Bienestar Bank-DANE ( 1978) Householdl SurveyFamiliar), 84 Dependencv ratio: defined, 95 ns8; of mal-

Columbian Social Secuinty Agency (Insti- notit-ished. 85; in peripher al densities.tuto de Seguiros Sociales), 293 87

Coomiier-ce sector, 19, 123-24, 143-44, Det-oit, 40144 n2 Developed countries: decentralization in,

Commission on Inter-governmental 33; income distribution in, 75, 95 n3;Finance, 245, 248. 249 populationi densities in, 24: returns to

(:omnunitv action (ac-i6nt commutnat) edutcation in, 104; transportationi pat-grants, 258 terns in, 186, 214, 223: urbanization in,

Comnisinity Action Administration 18, 19-20; utrban research on, 2, 4, 275.(Departamiiento Adminisu-ativo de See aLso North American cities; U.S. cit-Acci6mn Communoal), 234, 258 ies

Commuting patterns. SeeTravel patterns Developed estates, failure of, 142Compalsia Nacional de Electricidad, 259 Developers, residential, 151, 152-53, 154(.'omtunas: income distribtttion by, 51, 81; Developing countries: employment

poptulation density of, 43, 50; size of, decentralizationi in, 25-26, 33-34;43, 72 nl; in sul-vey data, 291 female labor in, 99; homeowsiersiip in,

Consolidation phase of urbanization, 9 168f-69, 181; household size in, 187-88;Controller (contralor), 234, 237 labor-uising sectors in. 97; land valuesCorbusier, Ie, 42, 51 in, 58; manufacturing employment in,Cotno, Roberto, 163, 166 123; popitlation density in, 24-25: pov-t:oRPOBtuSES (Corporaci6n Nacional de erty in, 74; real iticome in, 27; i-ettirns

Butses Urbanos), 193 to education in, 104; tranisportationCorporaci6n Central Regional de modes in, 24; travel patterns in, 186,

Poblaci6n (CCRP), 7, 293 187, 214; tt-baniizationi in, 1-2, 14, 17(:orporaci6n Financiera deTransporte nl, 18, 273. See alo Colornbia

(( F-r), 194 District Cadastral Office. 245Cortes, Mariluz. 7 District Council, 237, 238, 258, 259-60Crude birth rate (t:BR), 9 Distrito Especial (Special District,Cuiatro Estrategias ("Fotir St-ategies"), Bogota), 234, 235. See also Bogota; Coy-

149 erintent, districtCu:ndinanaarca, 234 Doebele, William, 7CLurTrie, L.atchlin, 6 Dolmus (shared taxi), 188

Dwelling units. See Housiig unitsDakar (Senegal), 169DANE. See Departamento Administrativo EAAB. SeeEmpresa de .Ucantarillado y

Nacionial de Estadistica Acued(ciito de Bogoti

318 UNDERSTANDING THE DEVELOPING MET'ROPOLIS

Earnings. -Se Iabor force earnings 123, 124, 143-44, 144 n2: deterini-East Asia, 2 nanits of, 132, 141, 280; household sur-Economies of scale, city expansion and, vevs oni, 122, 126, 128; and housing

19-20 demanid, 172-76; modeling frameworkEconomy: under Lleras government. 10; foi, 134-41, 145 n7, 145 n8; and public

Uinder I.Opez government, 11; 1980. policy 142-43; reasons for changing,cisis in, 8, 14, 17 nl; 1970sh boom in, 133-34: by rings and sectors, 124-25,13-14, 76, 97-98,117, 273-74; tinder 129, 130, 131; and transpor-tation, 64,Pastrana governmienit, 10-1 1; rural- 186, 207urban migrationi in, 8-9, 75, 274 Empresa de Alcantarillado v ActedLuctCt

Et:OPETROL, 199 de BogotA (ExAsA), 234, 255, 259, 260,EDIS (Empresa Distrital de Servicios P6ibli- 261-62

cos), 234, 259-60 Empresa de Energia Eletrica cle BogotaFODTU (Emnpresa Disutital de Transporte (FEEB), 234, 255, 259. 260. 262-63

t'rbano), 189, 194, 235 Empresa deelefonos cle BogotA (ErR),Education: government expansioti of, 234, 259, 263

100-101, 255-56, 268 n2; higher, 99. Empresa Dist-ital de Ser vicios Ptblicos100, 103, 104, 106; povert, and, 89; pri- (LDIS), 234, 259-60mary and secoridary, 9, 75, 99-100, Empresa Distrital de Transporte I rbarno103, 1(04, I 05-106: private expendi- (EoD'). 189, 194, 235tures for; 99-101; revenue sources for, Empresas Mtrnicipales de Cali (EMCAIA1,

239, 250, 252; U.S. segregation in, 83 263Education, ietuirns to: by age cohorts, Esmeralda (.a) hotisinig program, 148

1(06-107: and certification bonus, 105, FTB (Emnpresa de Teletonos de BogotA),106; declines in, 105-106. 118; and 234, 259.263earnings, 103-107, 114, 116-17; by Europe, 19-20educational level, 104, 105-106, 118;1by Exchalige rate, 8, 1(), 13industy and occupation, 116, 118; by Expendcituires, capital: by decentralizedresidence location, 112, 113; using agenicies, 236, 255; bv district govern-splined variable, 105, 106, 119 n3; for ment, 235-36, 255. 268 nl; financiigwomen wo)rkers, 13. 97, 99, 119, 276; ol, 236-38, 249workers' backgrounidc) and, 11)0-11 Exponential functional form to meaiseire

Edtication (EdCIcaci6n) secretary, 234 densit: 22-23, 30EFEB. See Empiesa de Eneigia Electrica de Export promotion, I0

Bogo tAEftervescenit phase of urbanization, 9 Fares, uransportation., 25, 187, 194, 195,Ellickson, Bryan, 135 223, 224, 285-86El Salvador- l6i( FLr ot:tRA\s (Federaci)n Colorrobiano deFM(IAI.I (Empresas iiunicipales de Cali), Transporte Lr bano) . 193

263 FER (Fond(lo Educativa Regionial), 250,Employment: by firm size, 123-24; 1980s 252, 256

slowdown in, 11, 13; 1970.sboom in, 19, Fieldcs, Cary S., 6, 10)411, 97-98, 117: by sectoral share, 43, Finanice (Hacienda) secretary. 234, 23796-97, 123. See also Firms; labor torce; Finanicial Tranispor-tation CorporationManutacturinig firms; Services sector (Corporaci6n Financiera de Tranis-

Employmenit decentralization: in BogotA, porte, (:Fr), 19463, 71, 128, 129; in Cali, 63; by cori- Firms: characteristics of, 121; construc-merce aid services, 33-34,61. 123; tiont, 150; as earninigsvariable, 115,conisequjences of, 36: by nanatactur- I 1 6; location trends of, 122. 277; struc-ing, 25-26, 33, 61, 71, 123, 131; and tural role of, 20; years spent with, 107-transpyortation. 143, 207, 221). 223. 224. 108. See alto Manufacturiring firms; Ser-279. 286: trtUckirrg indtistry and, 133 .ices sector

Emnplovinent dlenisities: and land valties, Fondo de Ahorro y Vivienda. 14961; and populationl densities, 33, 61, Fondo Ediucativa Regional (IFR), 250,63: by rings aird sectors, 61, 63-64: and 252, 256t-avel patterns, 187 Fornial sector See Protected sector

Employment location: in CBD, 71, 121-22, Friedman, joseph. 145 n8123: of coriirerce ansd services, 33-34, Fuiel taxes. 199. 254, 255

INDEX 319

Garce Navas hioUsin1g program. 148 on periphery, 55, 87; ptiblic servicesGarden City concept. 24-25 uise by; 260: single- member. 64; size of,German cities, 56 55Gini coefficient, 75, 77, 95 n3, 95 n4 Household sur%evs: k)ANE, 76-77, 289-90,Gomez, Hernancdo, 6 300 n3; Phase 11(1972), 122, 126, 162,Government, district: budgetar-y expendi- 292-93. See also UWorld Bank-DANF

tures by. 235-36, 255, 268 rof; and City (1978) fHousehold SurveySttidv, 7; District Council of, 237. 238, Housing: availability of, 152, 15fi; as258. 259-60; educationi irole of, 255-56, Colombian prioritv, 146; conistrLction268 n2; Naormas Mitnimas bv, 153-54: of, 10-11, 25, 150; government pro-public works projects of, 258-59; in rev- grams for. II, 147-5(, 182, 283; incre-conel1-sharing program. 250, 252; seo- mental development of, 181-82, 283-vice responsibilities of, 233, 234: 84; locationi of, 112, 113, 114-15. 208.sutlctLiral role of. 20; subdivision legal- 214); Normas Miniras, 153-55, 158-59,ization by, 154, 18(\, 182: transporta- 16ff tinder I'astrana governmet. It10-tion role of, 189. 191. 193, 194, 222. See 11: policv recommendationis on, 181-also Public services; Public service 83; for poor, 13, 25, 74; quantity andagencies price measUtIemlents of, 171-72; tradi-

Governmient sec tor: earnings in, I1. 13, tional approach to, 147; dur ing urban-I 1 0, 1 16-17; edticationi expendituties siation.1 16. 146-47. Sere aLa;, Piralaby, 100-(I01 . 255-56 subdivision,,

Grants, governamentei, 239, 258 Houising demand: income anid price elas-Great Britain, 56 uiciues of, 176-78. 180-81) infiuenicesGriffith, Daniiel A., 37115 uon. 276-77; land prices anid, 60; mod-Grimiies, Orville, 7 cling, 171-78; real income growth and,Gross domestic pr odIuct ((nt)r 9-10. 38- 22; residenitial locatioIn theor y and.

39 172-75; stidies oni, 171; workplaceGross nationial pr-oduct ((.NP). 27, 37 n4 location anid, 172-76Guadatlajara (Mexico). 30, 56 Housinig developers, 151, 152-53. 154

Housing uniits: householcds per. 147; ntim-Health care pr ogramii, 250, 252 her of floors in, 67, 68; siue of; 174-75:Health (Salsid) secretary. 234 space per person in, 68; World BankHighway T'r-uist Fuiiid (triF). 1)99 research on, 295-96, 299, 300) n8

JItNC:.xt' See nlcomie per- capita, hotisehol(l Houston, 30

Hirschmani, Albert, 6 Hovt. liiomei, 79Homeownership: and dwelling tinit size. Hunan-capital model of earnings, 104-

174-75; goverinmelntal focus on, 284- 1)5, 1(09, 119 n I , 19 n2, 27585; househol(d characteristics of, 166-68, 170)-71, 181; hotiseholcl inconie iti (Iristittito de Crclito Territorial),and, 152, 153, 15, 168-69; and hous- 147-48. 182, 283inig mobilits; 162, 163. 1469-70, 276: ani Int (Institito de Desarrollo Ltbana),travel patterns, 174, 208. 21(0. See also 234, 249, 259Lot bcuyers (pa(onesa de venaa) II.O (Initerniational LIabour Or-ganisation),

H ong Kong, 188, 19I, 222Y 6Hoover, Edgar I, 274 Immigration. See MigrantsHouselhold incomyie. See lncomie per linport-suhsfilttting strategy, 1(

houtiehold Income disuibtution: in Bogota, 77-78,Hotisehold income per capita. Seelncome 79, 81, 83, 101, 103, 112; fiN(tAP mea-

per capita, houisehold suremenit of, 95 n5; and housingHousehold mobility: eariniigs andc. 114; demand, 73, 177-78. 180-81, 276;

high rate of, 162-63, 1770; household improvement in, 93; during indtistrial-character istics anid, 161-64, 170-71; i7ation, 74-75; ineqtialities in, 75, 77-

household i)comne annd, 1 63, 277; ten- 78, 79, 81, 83, 92; measurement of, 73-tire status anid, 162, 163, 166-67, 169- 74, 78-79; in North Amer ican cities.7(), 276 79; survey information on, 76-77;

Households: behavioral cleter-ninants of, travel patterns and, 185-86, 187, 276;161462, 275-76; car owniersihip per, vehicle taxationi and, 200-20 1. See a!lso2(0. 213, 2241; per dwelling unit, 147; Labor force ear nings

320 LUNDERSTANDING; THE DEVELOPING METROPOLIS

Income elasticities of housing demand, and. 107-108, 199 n4, 199 n5; and firm73, 176,177-78, 180-81,276 size, 116,120 n9; in government sector,

Income per capita, 10, 24 11, 13, 110, 116-17; household mobil-Income per capita, hotisehold (HIN(CAP): ity and. 114: human-capital model of,

by comuna, 51; household size and, 55; 104-105, 109, 119 nl, lIIn 2, 275; ofmalnutrition and, 85, 87: as measure- migrants, 90-91; in 1970s, 273; andment tool, 95 n5; by rings and sectors, nutrition, 85; occupational differen-51, 54, 79 tials in. 101; on-the-joh training and.

Income per household: census data on, 107-108; in protected sector, 11, 13,289, 300 n2; iHIN(AP and, 95 n5; and 110, 115-16; residence location and.homeownership, 152,153.156,168-69; 112, 113, 114-15, 208, 21(0; ofruraland household mobility: 163, 277; workers, 75; from survey data, 291;household survey (1977) on, 290; union membership and. 114; workers'lhousing loans and, 149; travel time by. backgrounds and, 1 10-11. 114-15204-205, 207-208; trip patterns by. Lagos, 191211. 213. See also Labor force earnings Land: incremental development of, 152-

Incubator hypothesis, 130, 131 53, 284; nationalization of, 283; forIndia, 244 .Normrs Minimas, 154Industrial countries. See Developed coun- Land Credit Institute (Instituto de

tries Crdito Ter ritorial, i(.r), 147-48, 182,Industrial Revolution, 19 283Infant mortaliry, 9 Land values: access characteristics and,Infrastructure services. See Puiblic services; 58, 6(, 64-65, 69, 70, 278; in Bogota

Public service agencies :BD, 58, 60; and density gradients, 56,Ingram, Gregor y, 26 58, 60, 61, 277, 278-79; in poor sectors,Inquilinalos, 152 67, 68, 70, 248; and public service tar-Institute of Urban Developmenit (Iristi- iffs, 261, 263; in rich sectors, 65, 67, 68,

tuto de Desarrollo lUrhano. tu). 234, 248; tax assessment of, 248249, 259 Latin Americani cities: debt crisis in, 8;

Instituto Nacional de Transporte (IN rRA), education benefits in, 104; GNP' per cap-

193 ita in, 27. 37 n4; population densitiesIntercept densities. 30, 32 in, 30, 32; population growth in, 5, 26,Interest rates, 8, 155-56 27; primacy index in, 38Intergovernmerntal Finance, Commission Lee, Kvu Sik, 139, 145 n

7, 145 lnO, 294

on, 245, 248. 249 Lee, Yoonjoo, 288Interior (Gobierno) secretary, 234 Legal representative (personero), 234International Labour Organisation (wLo), 6 Lerman, S. R., 145 n8, 214INTRA (Instttito Nacional de Transporte), LhinJohannes, 7. 239, 264

193 Lleras government (1966-70), 10Istanbul, 188 LoGn estimates: of moing probability, 164

London, 21, 56Kalvan (India), 143 l1opez government (1974-78), 11Karachi, 239 Losada, Rodrigo, 6Kennedy administration, 148 Lot buyers (promesa de v/eda), 155-56.Korean cities, 56, 58, 69, 168-69 159-60

Lot sellers (esfriturn deromprarventa), 155Labor force: construction, 150, 157; ediu- Lo)t size., 70

cation and, 9, 13, 75, 89. 97, 99. 103; Lubell, Harold. 6-7occupational distribution of, 101; poor Lusaka (Zambia), 239workers in, 89-91, 208; rural, 8. 14, 75-76; segmentation in, 13. 15-16, 93, McCallum, Douglas, 6-7118, 275; spatial distribution of, 101, McFadden, Daniel, 145 n8103; travel patterns of, 204, 207. 208, Malaysia, 75, 81210; urbanization of. 18-19; women Malnutrition, 84-85, 87, 94, 95 n9workers in, 77, 90, 91, 97, 98, 99, 101 Malpezzi, Stephen. 177-78

Labor force ear-nings: of bus drivers, 194, Manila, 188223; bv city sectors, 113; education vari- Manufacturing firms: average size of, 129;able and, 103-107, 114; experience DANE directory of, 293-94; death rates

INDFEX 321

of, 128-29, 277; incubation area for, Nota opcion (vehicle license), 193130-31 : labor force share bv, 43, 97, Nutrition. See M.tanutt-ition123; location behavior of, 124-25, 128,129, 131-33, 141-42, 280: modeling Occupations: of poor workers, 89-90; spa-locauon of, 134-39, 145 n7, 145 n8; pri- tial distributio bv, 101, 10)3; tr-avel patt-mary aictivities by, 19; reasons for mTov- tem ns bv. 21(0; vears spent in same. 1(07-ing, 1 33-34, 280; in satellite cities, 143; 108surveys on, 122, 1 26, 128; WVorld Bank o i (on-the-ljob trainini g). 1(15, 1017-1018survey of, 294-95. See also Labor force ot.s estimates of moving probabilirv, 164

Mayo. Stepheni K., 177-78 Ordone7, Mvriam. 288. 289Mayor: role of, 234 Owuisi-Banahlier(. K., 24-275Maz7umdar. Dipak, 104Medellin, 10, 40 Pachori, Alviaro, 198Meernsanjacob, 95 n5 Para Cerrar Ia Brecha. 15()Mexico City, 6. 56 Pasu-ana government ( 1970-74), 1(-1 IMicidle East. 2 Phase 11l lotiSehold Stirvev 1 1972), 122.Migrants. 24, 90-91, 166, 1 70 1216. 162, 20';. 292-93Mills. Edwin S., 22. 3(0. 58, 61. i9 Philadelphia, 40Mincer,Jacob, 1(04 Phoenix, 30Minibuses. Seel BlIseiej Pinedaj osL Fernando. 21 (Ministr-y of Finance, 238 Pi-atla developers (urbanizadoors pirants).Ministry of Public W'orks and Transpor ta- 152. 155-56. 160, 183 n3, 282, 299-300)

tioni (Ministerii, de Obr as Pfiblicas v Pialsa subdivisions: res rbetd. 152-53;Transporte). 193 government respoisst to, 153-54, 18));

Modeling: of car owntership probability, inc remncrtal development oif, 64), 153,221; bv City Studs. 3, 274; of housing 15(-57, 159: lIot puce in, 158; lot trans-demilanld, 1 7 1-78; (if iiantiulactiiri nig fer in. 1 55-56: p-(fiso oii . I 60, I 83 ris:1locations behIvior. 134-39, 145 is7. 145 Stipern iteii (eicia Banicarai survev of,n8: ofservices locatson behavior, 139- 41, 299-30))145i n9. 145 nis); of travel rmiode, 21i5-21 Pla7a MIavor (:atii), 42

Mohali, Rakesh., 96 Pla7a Pr incipal (Bogot)i , 4))Monetarv(Coinc il (utinta Monetaria). 238 Policvy rcomissed(lations bv City Stu(ly.Moiltelrre, 56 71, 181-83,272.273.281,282MNlintIiioiial Iogil ni ethod, 135, 1 45 n8 Poor, [Ise: Iw age arild cr ticrationm 89: dlen-M1iticipal goverismisnti. SeeGovernment, sitV patterns of, 24. 25, 278-79: ins

distr-ict developing (oinitries, 74; ensplone nentlocation and, 63. 64, 65, 187; bome-

Nationsal ( orpo ration ol Urbhan Buises ownership hv, I 67; hosseliold msobhility((:orpo raci6;ti Nacionial de Buses oi, 164, 1 66 277; hott intg ho r, 1'3, 25,Urbaiso s, ((muixsitts). 1 93 148, 181-82, 283-84; identifi(iatialon

Natioisal lhouising Agencv (Inistittito) de oi, 85-84; in labor foice. 89-91; laudCErdito Tt riitorial). 13 alties oif 67. 68, 7(1, 248: living spa( e

Nationial Plaisiiisg Departmenit (Departa- of, 68: malnourishissisut of, 84-85, 87,issenuo Nacioisial de Planeaci6it). 7, 238 94, 95 n9: public services fom, 94, 263-

Natiosial Statistical Office (Depar tameiito 64, 266; residutsitial locittson of, 24,Administrativo Nacional de Estadis- 36, 63. 112-13: spatial vegregations of,tica., I).NI ). 7, 76 15. 36, 52, 55, 63.i 93i: Thli iJ ndi xc

Naionial Tariff Board, 26i). 264 for, 79. 81; t-avel patterns of, I 86,Nelvois, Richard, 6 20)7-2(08, 279: WN'old Batik cicarchNew Vitrk City, 4(0, 56 on. 2-;). 295-96, 2919, 31)1 rifi. 301) s7.Norman Minaims stubdivisions, 153-55. 3)11) ts8

158-59, 161, 299-300 Poputliations ( cnisi (1973). 7(-77. 288-89,No rtls Ass e ricris citics: dicce rstrali7rationli 300 i 1, 300 is2

itl, 33: incitse distribntio sn in, 79: issani- Poptilation dccritrali/iatios: conse-Llfacttiring erisp1woxisetst in, 123: popUt- qustces :o, 3f; costs ot, 72: deteriti-lation deissities its, 30. 32, 4tl, 56: iistmvs ol, 23-24, 32. 280; as growthlpopulation growth it, 26-27: residei- phenoirsenont 21-22,33: ttseasiirtlilliuttiial lotts in, 701 See W/t, U .S. citic s of, 22-23: sitbheister a iand, 69

322 UND)FRSTANDIN(; THElU )Et.DE 1EOPINC METROPOlIS

Population densities: in Bogotia, 39-40. housinig of, 13, 174; travel patterns of,43, 50-51. 56; central and peripheral. 174, 21027, 30. 36, 70, 87, 277; in developing Research: transferabilitv of, 5, 14-15, 275.coulitries, 24-25; and land values, 56, See also City Studv58, 60, 61, 277, 278-79; by irings and Residence location: and housing demand,sectors, 50, 55, 63-64, 65; worldwide 172-75; as labor foice variable, 1 12,patterit in, 30. 32; and zoninig policies. 113, 114-15.208.21094. See also Density gradients Residential decentralization. See Popula-

Populationi growthi: and average densit. Lioll decentralization'39-40; decentralization and. 22, 23- Residenitial densities. See Population den-24; pi tintacv index and. 38; by rings sitiesanid sectors, 50-51: itl urbani centers, Residenitial (leelopers, 151, 152-53. 1541-2, 8-9; 9o -ldwicde, 26-27 Residential location theory, 172-75

P sotter, IOeph, 288, 289 Revenues: fromil public sel vice tariffs,IPover-ty. See Poou, the 236-37, 26044; sour-ces of, 236-38,Plowet companiy See Emiipr-esa de Enei gia 239; from valorization char-ges, 234,

Hera i(a de Bogota (PEER) 249-50, 256. 258, 287. See also TaxesPreston, Sa isiuel H ., 27 Ribe, I lelena, 6Priiiyacw iiilex of Colombia, 38 Rich, the: employment location and, 63,Prinitat y sectotr. Se Agriculttiral sector 64, 65, 124-25, 187; income of, 79, 112;Privat sector: eatrnings in, I I 6; education laud values of, 65, 67, 68, 248; living

fiinancing by, 99-101; hotising inter-ven- space of, 68; public services for, 93-tiori bx 282; t unaspor tation participa- 94; residential locationis of', 24. 36, 63,tion bv 188, 189, 222, 223, 224. 282 112-13; Theil in(lexes foi; 79. 81; travel

Ploperty taxes, 244-45 patterns of, 186, 207Propet ty valtues. See l,anid values Ritigs. See City ritigsPr-otecte(d sectotl 11. 13, 110, 115-16, Rio dejaneiro, 5-6

120 n9. See alts' Govern nient sector Roads, 256, 258I'saclarapoilos. (Geoige, I 14 Ruir-al at-eas. 2, 8, 250, 252. See Also Agr-ictil-Puiblic service agencies: buidgetinig by, itiral sector

237, 238; expendittites by, 2'36. 255;imiain, 234-35; responisiveness of, 94. Samplintg unit. 292238-39, 266-4i7, 282, 286-87: strutctir-e San F aticisco. 40and( operationi of, 259-60; tat-iftf' SanJose, '30chlarges bv, 236-37, 260-64, 26t6 Sa:o Patilo, 6, 43, 56

Public se-Nices: availability of, 98, 15()-51; Sa.tellite cities, 143Cit' Study recommendationss for, 282; Savings arid loan Associations (Caja dedecentralizationi and, 36, 72, 142-43; Ahor-ro v Viviendai), 149iticrentienital development of, 153, 156- Schools. See Eduicationi57, 180, 182, 284; lowered standards Schiltz. '1: Pattl, 6, 104fbr, 151-52, 153; in .Vorrmas, AIiimas, Secondar'sector. SeekManufacturing154, 159: in poiralos, 153, 156-57, 159, Secirr180; shareed finan(inig of, 159-60; spa- Secto rs. See C.it sectorstial disparities il, 93-94; tax-flinanced, Seers, Duidley, 6245, 248-49; diriitsg urbanization. 146- Selowsky, Marcelo, 747, 281: valorization-finaniced, 249-50, Seoitl. 5fi256, 258, 287 Services sector: labor requiretenits of, 20,

Public W*or-ks (Obras Pfiblicas) secretary, 103; labor shar-e by. 97; locational pat-234 terins of. 123-24, 143-44, 144 n2;

locatiots ntodel foir, 139-41; poor'sQuigleslJohim M.. 145 n8 eritplovnienit in, 74; prirsary activities

itn, 19; uirbani conscentration of, 19, 26,Real income, deceittiralizatiots and, 22 71Recife (Brazil), 56 Situado Fiscal. 250, 252Regional Education Fmnd (Fonsdo Educa- Slighlton, Robert, 6

tiva Regional. FFR), 250, 252. 256 Social welfare department (DepartmentoRentters: and dwelling Utlit size, 174-75; Adissinistrativo de Rienestar Social),

and hbomeowitersltip accessibility, 284- 23485; hsousehold itiobility of. I 62. i 63; Song. Ruiyng Nak, 22. 58, 69

INDEX 323

Soutih Asia, 2 poor, 186, 207-208, 279; of students,Southeast Asia. 2 204, 207: by transportation mode. 203;Splined education variable, 105, 106, 119 of workers, 204, 207, 208: by work

n3 zones, 174State Planninig Office (Planeaci6n Depar- Travel time: for btis trips, 205, 220; car

Lamental, Valle del C.auca), 7 ow-nership and, 204-205; by house-Stow,J., 21 hold income, 204-205. 207-2)8: andi'he Strategy ofEronomi Develeopment travel mode, 218-19, 220-21

(Hirscnmai). 6 T'ravel trips: defined. 232 n2Sub-Sahar-an Africa, 2 Treasurer- ( tesorero), 234Superintendencia Bancaria, 7, 157-58. Trucking iicnustry, 133

183 n2, 299-300"A Survey of l.ondon. Writteni in thie Year LINDP (United Natioins Development Pro-

1598" (Stow), 21 graorne), 6, 292Uneneployment, 8, 13, 91, 94, 97, 115

Tan, jee-lPeng, 30 Lnidad de Poder Adquisitvo ConstanteTariffs, public service, 236-37, 260-64 (tep ), 149Taxes: beer, 250, 252, 254-55: indtustrv Unllion memilbership. 114

and cotmmerce, 244, 248; proper-ty, United Nations, 1-2244-4A, 248: sales, 250. 252; vehicle, Unlite(d Nations Developireiit Programme198, 199-200, 248. 254, 255 (tNDP), 6, 292

ITaxis, 189, 191 URiAC (Uliri(la(i de Poder AdqUisitivo Con1-Telephore comilpaniy (Eniplesa de Tele- stanite). 149

fonos (le Bogota E1-B), 234, 259, 263 Ulrlbaiizationi: causes of, 18-19: ini Colorn-Tertiar-y sector. See Set vices sector bia, 8-9, 38, 273-74; in deselopingTheil index, 77, 79, 81, 95 n4, 95 n7 countries. 14. 17 n 1; in EuriopeaniT'Ihomilas, Vin(od, 76 countries. 19-20; housinig response to,l'iniciia housillg program. 148 16, 146-47: mattirinig versus rapid, 14;'I'itlols di Ahorra Cafe/era (coffee savings phases of. 9, 273-74; policv responses

bonids), 16(0 to, 274, 282-83Tokyo, 56 Urbanizinig (ounitr-ies. See DevelopingTriade deficit. 8 couinitriesTransportation department (Depar-ta- Urban Poverty Task Force. 3

mento Administrattvo de Transpor te y ['rru-tatia, Miguel, 6, 75Triansito t)ATT), 193, 234 LI.S. cities: decentralizationi in, 21, 56:

Transportation systerm: accessibility of, edrrcatioial segregation in), 83; GŽNPt per208, 214, 222: as alternatise to cars, capiut in, 27, 37 n4; homeownership214, 221, 222-23, 224, 276; and city rates in, 169; houisehold mobility ratesstructurre, 16, 185. 280; decentr-aliza- in. 1i62-63: inaritlacmtring firmis in,don and, 24, 32, 143, 223, 224, 280. 128, 129; population density in, 30;286; eniploniltelit cortcentratiour arrd, popurlationi growth tir, 26-27: tiavel in,186, 207: far-e stnutictur e of, 25, 187, 194, 203-204, 213, 222. 223; unemployment195, 223, 224, 285-86: governmelt sub- in, 115sidy of, 191, 193, 194, 195-96, 198, 20. User charges. 236-37, 260-64201. 222, 285-86; operating costs of,194: private sector participation in. Valen-zuela,Jairoe. 153188, 189, 222, 223, 224. 282; rotute pro- Valle del Cauca. 42cess of, 193-94; velsicle types in. 188, Valori7ation charges. 234. 249-50, 25fi,189, 191. See aLso Buses: Bose/as 258,287

Travel nsode modelinig: discrete choice Valverde, Nelsori, 288framewsork of, 215-17; predictability Verrsez, Georges, 6. 153of, 220, 221: theoretical backgrouLid Verrnorn, Raymond, 274flo, 225-27 anid travel timre aluatiLo),217, 218-19, 220-21 Wages. See Labor force eaarrinigs

Travel patterris: car owrsership and, 213- Wagner, M. Wilhelmn, 5814, 222: emploviisetit decenltralizatiors Water and sewer-age cornpany. Seearid, 64, 207, 279; freqrenicy profile of, Enmpresa de Alcantarillado y Acue-210: iicome andI, 187, 211, 213,276: by drrcto de Bogota (FAAB)occupatiorn. 210; peakirsg in, 214; of Weher, Acina, 21

12-1 t'NDERSTANDING TIIE DEVELOPIN(. METROPOLIS

Western European cities. 27.33 WVorld Bank-DANE (1978) Household Sur-Wiesner, (;uillermio 56 vev: construction data from, 153;Wiesiier- anid Cia. Ltda., 56. 58 described, 290-91; employment dataWoNomeni: education supply and, 9, 13, 97, from, 107, 122, 126, 128; homeowner-

119, 276: househol(d mobility of, 164; ship data from, 169-70; householdin labor force, 77, 90, 91, 97, 98, 99, data from, 64, 162-63; sampling frame10(1 or, 292; sources for, 288-96, 299-300;

Wor-kers. See labor force travel data from, 199, 200, 203Workplace locationi anid houtsing demand, World War 11, 9

172-76World Bank, 2-3, 6, 7, 11. 13, 294-95, Zahavi, Yacor, 20)4

300 o1i Zonificatioti svstems. See Citv rings; Citysectors

he _World Bank -r ; .3

nhis book is among the most comprehensive studies of a city since Edgar M.;;:,.Hoover and Raymond Vernon;':analyzed New;,York City th'iivWe, years ago ',t ' ll>-",

, ; '.*itPi'''i'0 cl assic,`Anatomj.'f aMei;opa&-The -minles6"tohiuried inthe u-, <+i.ren' bo'ok, istliat'ithe most practical way6of coi 'th' &' "etablishnstitutional mechanismsthat can respond toiildlychanging;dand unpredictabledemands of a citys'residentst Unlikc much writing on cit-A'7- ies and urbanizatGion'in the develo'pi'n"'g"w'orldsthi'stidy hificluaes reasons for,

*~ V ern o n analrzed N ;,at< iI . .',: optimism that the'ex'panisio'nof fast-growing cities in the developing world

pdeth the 'structuredof two Colombian cities,Bogoth and Cali, by'modeling ,different markets'and the behavior of indiiduals, houeholds, firms, and

* governments within these markets. He underlines theimportance of under- k* standing the behavior of the various actors in a city,'and his use of simple

economic reasoning'contributes much to comprehension not only of urbanbehavior but also of the structure'of the city itself. Approaches developedhere are also broadly applicable for analyzing cities in developed countries.

This study is unusual in that it brings to the lay reader, in accessible form,* the rationale of and results obtained from the sophisticated techniques

* used in the analysis of urban housing and transportation patterns, laborforce behavior, and industrial location patterns within a city. Whereas manyurban studies concentrate exclusively on issues related to infrastructure

| requirements and delivery problems within a city, this study links infrastruc-ture requirements and supply to the behavior of households, firms, and gov-ernment and to the existing income distribution in the city. This betterappreciation of the underlying behavior-which determines what citieslook like-could result in much more effectively designed urban policies.

Other books that have come out of this study are Works, Wages, and Welfarein a Developing Mctropobis: Consequenacs of Growth in Bogotd, Colombia, byRakesh Mohan, and The Location ofJobs in a Developing Mctropolis: Patterns ofGrowth in Bogtd and Cali, Colombia, by Kyu Sik Lee, both published by

'Oxford Uniest rs.*. ~.

Wi RaehMhni conomicdvsrotegvrmnofniamheMity ; ! ,kof Industry.4A f&9r World Bank staff member,' hie the author of two pre-;' -

vious books'on urban economic modeling and urban labor marketsa'At pub-< lication-,he h ahgdju'stifiished a"term' as Distinguished'Visiting Fellow' at the

United Nations University Instituite for New Technologies in' the Netherlands. 9

UNDERSTANDING DEV METROP

WaUon Am4uoooo 1219s 3 9 5