Advances in Spatial Science

Editorial Board

Manfred M. FischerGeoffrey J.D. HewingsPeter NijkampFolke Snickars (Coordinating Editor)

For further volumes:

http://www.springer.com/3302

lEditors

Antonio Pez Julie Le GallolRon N. Buliung Sandy Dallerba

Methods and Applications

123

Spatial AnalysisProgress in

Editors

reproduction on microlm or in any other way, and storage in data banks. Duplication of this publication

1965, in its current version, and permission for use must always be obtained from Springer. Violations areliable to prosecution under the German Copyright Law.The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,

and regulations and therefore free for general use.

Cover design: SPi Publisher Services

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

concerned, specically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,

Professor Antonio Pez

Canada

Professor Julie Le GalloUniversit de Franche-Comt CRESE45 D, Avenue de lObservatoire25030 Besanon Cedex, Francejlegallo@univ-fcomte.fr

Professor Ron N. BuliungDepartment of Geography

3359 Mississauga Road North

Canadaron.buliung@utoronto.ca

Professor Sandy Dallerba

University of ArizonaP.O. Box 210076Tucson, AZ 85721, USAdallerba@email.arizona.edu

ISBN 978-3-642-03324-7 e-ISBN 978-3-642-03326-1DOI: 10.1007/978-3-642-03326-1Springer Heidelberg Dordrecht London New York

Library of Congress Control Number: 2009934479

Mississauga, Ontario L5L 1C6

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is

even in the absence of a specic statement, that such names are exempt from the relevant protective laws

or parts thereof is permitted only under the provisions of the German Copyright Law of September 9,

Springer-Verlag Berlin Heidelberg 2010

Advances in Spatial Science ISSN 1430-9602

Department of Geography and Regional Development

School of Geographyand Earth Sciences1280 Main Street WestMcMaster UniversityHamilton, Ontario L8S 4K1

paezha@mcmaster.ca

University of Toronto at Mississauga

For Patricia, Leonardo, and Luanna (AP)For Tara, Meera, and Emily (RB)

Foreword

Space is one of the fundamental categories by means of which we perceive andexperience the world around us. Behaviour takes place in space, and the geograph-ical context of behaviour is important in shaping that behaviour. While space byitself explains very little, spatial processes and the spatial patterning of behaviourhave long been viewed as a key to understanding, explaining, and predicting muchof human behaviour.

Whether or not spatial analysis is a separate academic field, the fact remainsthat, in the past 20 years, spatial analysis has become an important by-product ofthe interest in and the need to understand georeferenced data. The current interestin the mainstream social sciences to geography in general, and location and spatialinteraction in particular is a relatively recent phenomenon. This interest has gener-ated an increasing demand for methods, techniques, and tools that allow an explicittreatment of space in empirical applications. Thus, spatial analysis tends to playan increasingly important role in measurement, hypothesis generation, and valida-tion of theoretical constructs, activities that are crucial in the development of newknowledge. The fact that the 2008 Nobel Prize in economics was awarded to PaulKrugman indicates this increasing attention being given to spatially related phenom-ena and processes. Given the growing number of academics currently doing researchon spatially related subjects, and the large number of questions being asked aboutspatial processes, the time has come for reflecting on the progress made in spatialanalysis.

As an editor of the book series, I highly welcome the present edited volumeon Progress in Spatial Analysis with a focus on theory and methods, and thematicapplications across several academic disciplines. The effort is a worthy intellec-tual descendent of previous volumes in the series, including Anselin and FloraxsNew Direction in Spatial Econometrics in 1995, Fischer and Getis Recent Devel-opments in Spatial Analysis in 1997, and Anselin, Florax, and Reys Advances inSpatial Econometrics in 2004.

I am pleased to realize the mixture of very well-established leaders in the fieldof spatial analysis and a new generation of scholars who, one hopes, will con-tinue to carry the torch ignited more than 50 years ago at the dawn of QuantitativeGeography and Regional Science. In this spirit, it is my hour to formally profferthe welcome to this edited volume, and to the effort poured into bringing major

vii

viii Foreword

developments and applications into a single source representing recent publicationsin spatial analysis. I anticipate that this volume will become a valuable reference, asthe previous offerings in the series.

Vienna Manfred M. FischerMay, 2009

Contents

Progress in Spatial Analysis: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Antonio Paez, Julie Le Gallo, Ron N. Buliung, and Sandy DallErba

Part I Theory and Methods

Omitted Variable Biases of OLS and Spatial Lag Models . . . . . . . . . . . . . . . . . . . . . 17R. Kelley Pace and James P. LeSage

Topology, Dependency Tests and Estimation Bias in NetworkAutoregressive Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Steven Farber, Antonio Paez, and Erik Volz

Endogeneity in a Spatial Context: Properties of Estimators . . . . . . . . . . . . . . . . . . 59Bernard Fingleton and Julie Le Gallo

Dealing with Spatiotemporal Heterogeneity:The Generalized BME Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75Hwa-Lung Yu, George Christakos, and Patrick Bogaert

Local Estimation of Spatial Autocorrelation Processes . . . . . . . . . . . . . . . . . . . . . . . . 93Fernando Lopez, Jesus Mur, and Ana Angulo

Part II Spatial Analysis of Land Use and Transportation Systems

Seeing Is Believing: Exploring Opportunities for theVisualization of ActivityTravel and Land Use Processesin SpaceTime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .119Ron N. Buliung and Catherine Morency

Pattern-Based Evaluation of Peri-Urban Developmentin Delaware County, Ohio, USA: Roads, Zoningand Spatial Externalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .149Darla K. Munroe

ix

x Contents

Demand for Open Space and Urban Sprawl: The Case of KnoxCounty, Tennessee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .171Seong-Hoon Cho, Dayton M. Lambert, Roland K. Roberts, andSeung Gyu Kim

Multilevel Models of Commute Times for Men and Women . . . . . . . . . . . . . . . . . .195Edmund J. Zolnik

Walkability as a Summary Measure in a SpatiallyAutoregressive Mode Choice Model: An Instrumental VariableApproach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .217Frank Goetzke and Patrick M. Andrade

Part III Economic and Political Geography

Employment Density in Ile-de-France: Evidence from LocalRegressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .233Rachel Guillain and Julie Le Gallo

The Geographic Dimensions of Electoral Polarizationin the 2004 U.S. Presidential Vote . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .253Ian Sue Wing and Joan L. Walker

Gender Wage Differentials and the Spatial Concentrationof High-Technology Industries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .287Elsie Echeverri-Carroll and Sofa G. Ayala

Fiscal Policy and Interest Rates: The Role of Financialand Economic Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .311Peter Claeys, Rosina Moreno, and Jordi Surinach

Part IV Spatial Analysis of Population and Health Issues

Spatial Models of Health Outcomes and Health Behaviors:The Role of Health Care Accessibility and Availability . . . . . . . . . . . . . . . . . . . . . . . .339Brigitte S. Waldorf and Susan E. Chen

Immigrant Women, Preventive Health and Place in CanadianCMAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .363Kelly Woltman and K. Bruce Newbold

Is Growth in the Health Sector Correlated with Later-LifeMigration? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .381Dayton M. Lambert, Michael D. Wilcox, Christopher D. Clark,Brian Murphy, and William M. Park

Contents xi

Part V Regional Applications

Evolution of the Influence of Geography on the Locationof Production in Spain (19302005) .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .407Coro Chasco Yrigoyen and Ana M. Lopez Garca

Comparative Spatial Dynamics of Regional Systems . . . . . . . . . . . . . . . . . . . . . . . . . .441Sergio J. Rey and Xinyue Ye

Growth and Spatial Dependence in Europe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .465Wilfried Koch

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .483

Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .489

List of Figures

Topology, Dependency Tests and Estimation Bias in Network AutoregressiveModelsSteven Farber, Antonio Paez, and Erik Volz

Figure 1 LR test rejection frequency for difference levels of spatialdependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

Figure 2 The impact of sample size on rejection frequency . . . . . . . . . . . . . . . 41Figure 3 Rejection frequency curves for two different sample sizes . . . . . . . 41Figure 4 The impact of mean degree on small networks . . . . . . . . . . . . . . . . . 42Figure 5 The impact of mean degree on large networks . . . . . . . . . . . . . . . . . . 42Figure 6 The impact of clustering on rejection frequency . . . . . . . . . . . . . . . . 43Figure 7 The impact of matrix density on rejection frequency . . . . . . . . . . . . 44Figure 8 Dependence parameter estimation bias for different levels

of dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Figure 9 The impact of sample size on dependence parameter estimation

bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Figure 10 The impact of mean degree on dependence parameter estimation

bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47Figure 11 The effect of clustering on dependence parameter estimation bias 48Figure 12 The relationship between matrix density and estimation bias . . . . . 49Figure 13 Goodness of fit scatterplots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

Endogeneity in a Spatial Context: Properties of EstimatorsBernard Fingleton and Julie Le Gallo

Figure 1 Exogenous variable spatial distribution (a) and augmented spatialDurbin parameter distribution (b, c and d) resulting fromMonte-Carlo simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Figure 2 Monte-Carlo distributions of the X parameter in (17) estimatedby fitting (18) and (11) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

xiii

xiv List of Figures

Dealing with Spatiotemporal Heterogeneity: The Generalized BME ModelHwa-Lung Yu, George Christakos, and Patrick Bogaert

Figure 1 Simulated random field realizations (top row); estimated fieldusing GBME (middle row); and estimated field using GK (bottomrow) at times t D 0 (left column), t D 1 (middle column), andt D 2 (right column) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Figure 2 Hard data (black circles), soft data in the form of uniformdistributions (white circles), across space-time . . . . . . . . . . . . . . . . . 82

Figure 3 Space-time distributions of the value of spatial order . (Left)t D 0, (Middle) t D 1, and (Right) t D 2 . . . . . . . . . . . . . . . . . . . . . . 83

Figure 4 Space-time distributions of the value of temporal order . (Left)t D 0, (Middle) t D 1, and (Right) t D 2 . . . . . . . . . . . . . . . . . . . . . . 83

Figure 5 Histograms of the estimation errors of the GBME (continuousline) and GK (dashed line) methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

Figure 6 Hard data (black circles) and uniform distributed data (whitecircles) across space-time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

Figure 7 Histograms of the estimation errors of the GBME (continuousline) and GK (dashed line) methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

Figure 8 Hard data (black circles), and Gaussian-distributed data (whitecircles) across space-time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Figure 9 Histograms of the estimation errors of the GBME (continuousline) and GK (dashed line) methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

Local Estimation of Spatial Autocorrelation ProcessesFernando Lopez, Jesus Mur, and Ana Angulo

Figure 1 Spatial regimes used in the experiment . . . . . . . . . . . . . . . . . . . . . . . . 100Figure 2 Spatial distribution of r. Lattice 7 7./ . . . . . . . . . . . . . . . . . . . . . . 106Figure 3 Spatial distribution of r. Lattice 20 20 . . . . . . . . . . . . . . . . . . . . . . 107Figure 4 Spatial distribution of r under the break. EastWest structure . . . 110Figure 5 Spatial distribution of r under the break. CenterPeriphery

structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111Figure 6 The doughnut effect and the Zoom estimation . . . . . . . . . . . . . . . . . . 112

Seeing Is Believing: Exploring Opportunities for the Visualizationof ActivityTravel and Land Use Processes in SpaceTimeRon N. Buliung and Catherine Morency

Figure 1 Critical dimensions and interactions between activity-travel andland-use systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

Figure 2 The Greater Toronto Area (GTA) and Greater Montreal Area(GMA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

Figure 3 Chronology of the spatial location of the mobile populationduring an average weekday in the GMA (1998) . . . . . . . . . . . . . . . . 127

List of Figures xv

Figure 4 Chronology of the spatial location of the mobile populationduring a typical weekday in the GTA & Hamilton (2001) . . . . . . . 128

Figure 5 GTA trip density excluding high density CBD traffic zones (2001TTS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

Figure 6 People accumulation profile in the Central Business District(GMA) segmented by region of home location (1998) . . . . . . . . . . 130

Figure 7 2003 Car accumulation profile (CAP), four districts (x: timeof day, y: number of cars) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

Figure 8 Monitoring of the number of cars parked in a specific area duringa typical weekday . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

Figure 9 Demographic structure with segmentation related to transit use(1987 & 1998 OD surveys), central Montreal . . . . . . . . . . . . . . . . . . 133

Figure 10 Geopolitical and network based conceptualizations of urban areas 135Figure 11 Network Occupancy Index (top) and Transit Network Occupancy

Index (bottom) estimated for 100 traffic analysis zones . . . . . . . . . 137Figure 12 Weighted Gaussian bivariate kernel estimation . . . . . . . . . . . . . . . . . 140Figure 13 Geovisualization of power retail capacity in the Greater Toronto

Area (19972005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142Figure 14 Centrographic estimation and geovisualization of power centre

expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

Pattern-Based Evaluation of Peri-Urban Development in DelawareCounty, Ohio, USA: Roads, Zoning and Spatial ExternalitiesDarla K. MunroeFigure 1 Study area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150Figure 2 Graphical illustration of variations in edge-to-area ratio

and the corresponding landscape shape index (LSI).(a) A square patch made up of nine individual squaresof dimension 2 2. (b) A non-square patch made upof the same nine individual squares, arranged less squarely.(c) A non-square patch made up of nine individual squares,arranged nearly linearly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

Figure 3 Landscape pattern analysis of Delaware County, 19882003. (a)Percent developed area (of total land) and Euclidean nearestneighbor distance edge-to-edge between contiguous parcels (km).(b) The number of patches (contiguous parcels sharing a commonboundary) and the landscape shape index (higher D greaterproportional edge in the landscape) . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

Demand for Open Space and Urban Sprawl: The Case of Knox County,TennesseeSeong-Hoon Cho, Dayton M. Lambert, Roland K. Roberts, and Seung Gyu Kim

Figure 1 Study area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180Figure 2 Transaction parcel with surrounding open space and 1.0-mile

buffer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

xvi List of Figures

Figure 3 Marginal implicit price of open space (10,000 square footincrease in open space) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

Figure 4 Price elasticity of open-space demand . . . . . . . . . . . . . . . . . . . . . . . . . 186Figure 5 Income elasticity of open-space demand . . . . . . . . . . . . . . . . . . . . . . . 186Figure 6 Lot size elasticity of open-space demand . . . . . . . . . . . . . . . . . . . . . . 187Figure 7 Finished-area elasticity of open-space demand . . . . . . . . . . . . . . . . . 187Figure 8 Housing-density elasticity of open-space demand . . . . . . . . . . . . . . . 188

Multilevel Models of Commute Times for Men and WomenEdmund J. ZolnikFigure 1 Population size of MSAs (n D 43) by region . . . . . . . . . . . . . . . . . . 200Figure 2 Regional differences in commute times from men-only,

women-only, and pooled menwomen multilevel models . . . . . . . 211

Walkability as a Summary Measure in a Spatially AutoregressiveMode Choice Model: An Instrumental Variable ApproachFrank Goetzke and Patrick M. Andrade

Figure 1 Map with the household locations of all the included trips . . . . . . . 222

Employment Density in Ile-de-France: Evidence from Local RegressionsRachel Guillain and Julie Le GalloFigure 1 Departments and communes in Ile-de-France. Scale: 1:9,000 . . . . 236Figure 2 CBD, new towns and highways. Scale: 1:9,000 . . . . . . . . . . . . . . . . . 237Figure 3 Geographic distribution of the density gradient for total

employment. Scale 1:9,000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244Figure 4 Geographic distribution of the density gradient for industrial

employment. Scale 1:9,000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245Figure 5 Geographic distribution of the density gradient for high-order

services employment. Scale 1:9,000 . . . . . . . . . . . . . . . . . . . . . . . . . . 245Figure 6 Geographic distribution of the density gradient for high-tech

employment. Scale 1:9,000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246Figure 7 Geographic distribution of the density gradient for standard

services employment. Scale 1:9,000 . . . . . . . . . . . . . . . . . . . . . . . . . . 246Figure 8 Geographic distribution of the density gradient for

finance-insurance employment. Scale 1:9,000 . . . . . . . . . . . . . . . . . 247Figure 9 Geographic distribution of the density gradient for consumer

services employment. Scale 1:9,000 . . . . . . . . . . . . . . . . . . . . . . . . . . 247

The Geographic Dimensions of Electoral Polarization in the 2004 U.S.Presidential VoteIan Sue Wing and Joan L. Walker

Figure 1 Electoral polarization: a conceptual framework . . . . . . . . . . . . . . . . . 255Figure 2 Box plot of descriptive statistics of the dataset . . . . . . . . . . . . . . . . . . 258

List of Figures xvii

Figure 3 Local Morans I significance maps of votes and key covariates . . 268Figure 4 Log-odds of voting republican by county clusters . . . . . . . . . . . . . . . 270Figure 5 Geographically weighted regression results . . . . . . . . . . . . . . . . . . . . 276Figure 6 Local Morans I significance maps of GWR odds elasticities . . . . . 278Figure 7 GWR odds elasticities of voting republican by county lusters . . . . 279Figure 8 GWR odds elasticities: global and local correlations . . . . . . . . . . . . 281

Fiscal Policy and Interest Rates: The Role of Financial and EconomicIntegrationPeter Claeys, Rosina Moreno, and Jordi Surinach

Figure 1 Baseline model, spatial model estimates .W D distance matrix/ . . 326

Spatial Models of Health Outcomes and Health Behaviors: The Roleof Health Care Accessibility and AvailabilityBrigitte S. Waldorf and Susan E. Chen

Figure 1 Spatial linkages of a health production function (HPF) . . . . . . . . . . 344Figure 2 Cumulative distribution of physicians relative to the cumulative

population distribution across Indiana counties, 2003 . . . . . . . . . . . 347Figure 3 Spatial distribution of elderly CVD mortality (left) and elderly

cancer mortality (right) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353Figure 4 Spatial distribution of maternal smoking rates (left) and rates of

prenatal care (right) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353Figure 5 Spatial distribution of nurses per person (left) and access

to hospital care (right) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354

Is Growth in the Health Sector Correlated with Later-Life Migration?Dayton M. Lambert, Michael D. Wilcox, Christopher D. Clark, Brian Murphy,and William M. ParkFigure 1 Distribution of quantile proportions of total in-migrants

composed of individuals in the 5569 (top panel) and 70C agecohorts (bottom panel) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388

Figure 2 Semivariograms of residual error structure . . . . . . . . . . . . . . . . . . . . 395Figure 3 Top panel, unshaded counties are those with rurality indices

0:52; bottom panel, counties with rurality indices 0:49. Bothare associated with positive change in the professionalconcentration of MDs and the office-based MD sub-group . . . . . . 397

Figure 4 Marginal effects of selected demographic and socio-economicvariables on changes in location quotients measuring differentmedical professions across a ruralurban continuum . . . . . . . . . . . . 398

xviii List of Figures

Evolution of the Influence of Geography on the Location of Productionin Spain (19302005)Coro Chasco Yrigoyen and Ana M. Lopez Garca

Figure 1 Choropleth maps of relative GDP per area (1 D nationalGDP=km2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418

Figure 2 Kernel density estimates of log relative GDP per area . . . . . . . . . . . 420Figure 3 Moran scatterplot of log relative GDP per area in 2005 (left). Map

with the selection of provinces ever located in Quadrant 1, plusMadrid and Valencia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422

Figure 4 Evolution of the impact of second nature forces on GDP density . 427Figure 5 Evolution of the impact of second nature on GDP density

in two regimes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431Figure 6 Evolution of the variance decomposition of regressions

in Table 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435

Comparative Spatial Dynamics of Regional SystemsSergio J. Rey and Xinyue Ye

Figure 1 Per capita incomes in the United States, 1978 and 1998 . . . . . . . . . 446Figure 2 Per capita incomes in China, 1978 and 1998 . . . . . . . . . . . . . . . . . . . 447Figure 3 Convergence and spatial independence in the United States and

China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448Figure 4 Regionalization system in China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449Figure 5 Regionalization system in the United States . . . . . . . . . . . . . . . . . . . . 449Figure 6 Inter-regional inequality share in China and the United States . . . 449Figure 7 Local Moran Markov transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 450Figure 8 LISA time path (left: China; right: the United States) . . . . . . . . . . . 452Figure 9 Covariance networks in China and the United States (thick

segments indicate similar temporal linkages) . . . . . . . . . . . . . . . . . . . 455Figure 10 Spider graphs of Zhejiang province (China) and California (the

United States) (the links indicate similar temporal linkages andthe thicker segments highlight spatial joins) . . . . . . . . . . . . . . . . . . . . 456

Figure 11 Spatial dynamics in China (top left view: the length of LISA timepaths (1); top right view: the tortuosity of LISA time paths (2);bottom left view: the instability of LISA time paths (3); bottomright view: spacetime integration ratio of temporal dynamics) . . . 457

Figure 12 Spatial dynamics in the United States (top left view: the length ofLISA time paths (1); top right view: the tortuosity of LISA timepaths (2); bottom left view: the instability of LISA time paths(3); bottom right view: spacetime integration ratio of temporaldynamics) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458

Figure 13 Convergence classification in China and the United States . . . . . . . 459

List of Tables

Omitted Variable Biases of OLS and Spatial Lag ModelsR. Kelley Pace and James P. LeSage

Table 1 Mean Oo and E O

o

as function of spatial dependence

. D 0:75; D 0:25/ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

Topology, Dependency Tests and Estimation Bias in NetworkAutoregressive ModelsSteven Farber, Antonio Paez, and Erik Volz

Table 1 Impact of matrix density on likelihood ratio . . . . . . . . . . . . . . . . . 35Table 2 Results of rejection frequency logistic regression . . . . . . . . . . . . 50

Endogeneity in a Spatial Context: Properties of EstimatorsBernard Fingleton and Julie Le Gallo

Table 1 Spatial Durbin: 2sls-SHAC estimator bias and RMSE for b1;omitted variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Table 2 OLS-SHAC estimator bias and RMSE for b1; ignoring omittedvariable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

Table 3 OLS-SHAC and 2sls-SHAC estimator bias and RMSE for b1 . 67Table 4 OLS-SHAC and 2sls-SHAC estimator bias and RMSE for b1 . 67Table 5 OLS-SHAC and 2sls-SHAC estimator bias and RMSE for b1 . 67Table 6 OLS-SHAC estimator bias and RMSE for ; simple model;

simultaneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69Table 7 IV-SHAC estimator bias and RMSE for ; spatial Durbin

model; simultaneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70Table 8 OLS-SHAC estimator bias and RMSE for ; simple model;

measurement error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71Table 9 IV-SHAC estimator bias and RMSE for ; spatial Durbin

model; measurement error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

xix

xx List of Tables

Dealing with Spatiotemporal Heterogeneity: The Generalized BME ModelHwa-Lung Yu, George Christakos, and Patrick Bogaert

Table 1 Examples of S -KB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79Table 2 Examples of soft data with integration domain D and operator

S see, Equation (10) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80Table 3 Summary of theoretical GBME properties. . . . . . . . . . . . . . . . . . . 81

Local Estimation of Spatial Autocorrelation ProcessesFernando Lopez, Jesus Mur, and Ana Angulo

Table 1 Coefficients used in the simulation . . . . . . . . . . . . . . . . . . . . . . . . . 100Table 2 Diagnostic statistics in the static model. No spatial effects.

Lattice: 7 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101Table 3 Diagnostics statistics in the static model. No spatial effects.

Lattice: 20 20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102Table 4 Testing the SLM, under the hypothesis of stability . . . . . . . . . . . 103Table 5 Zoom estimation under the null hypothesis. Some descriptive

statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107Table 6 Zoom estimation when the DGP is unstable in . Descriptive

statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108Table 7 Percentage of cells correctly classified . . . . . . . . . . . . . . . . . . . . . . 113

Pattern-Based Evaluation of Peri-Urban Development in DelawareCounty, Ohio, USA: Roads, Zoning and Spatial ExternalitiesDarla K. MunroeTable 1 Landscape pattern analysis, 19882003 . . . . . . . . . . . . . . . . . . . . . 160Table 2 Descriptive statistics, peri-urban agricultural parcels,

and parcels developed, 19882003 . . . . . . . . . . . . . . . . . . . . . . . . . 162Table 3 Results of complementary loglog model of urban conversion,

19882003 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163Table 4 Landscape pattern analysis of actual and predicted

development patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164

Demand for Open Space and Urban Sprawl: The Case of Knox County,TennesseeSeong-Hoon Cho, Dayton M. Lambert, Roland K. Roberts, and Seung Gyu Kim

Table 1 Variable names, definitions, and descriptive statistics . . . . . . . . . 175Table 2 Comparison of performance among OLS, GWR,

and GWR-SEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183Table 3 Parameter global estimates of global (OLS) models . . . . . . . . . . 184

List of Tables xxi

Multilevel Models of Commute Times for Men and WomenEdmund J. ZolnikTable 1 Descriptive statistics for household-level dependent

and independent variables for men-only, women-only,and pooled menwomen subsamples . . . . . . . . . . . . . . . . . . . . . . . 204

Table 2 Descriptive statistics for MSA-level independent variablesfor men-only, women-only, and pooled menwomensubsamples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

Table 3 Household-level coefficients and standard errors for men-only,women-only, and pooled menwomen multilevel models . . . . . 206

Table 4 MSA-level coefficients and standard errors for men-only,women-only, and pooled menwomen multilevel models . . . . . 207

Walkability as a Summary Measure in a Spatially AutoregressiveMode Choice Model: An Instrumental Variable ApproachFrank Goetzke and Patrick M. AndradeTable 1 Descriptive statistics of all included variables . . . . . . . . . . . . . . . 223Table 2 Linear probability regression model results . . . . . . . . . . . . . . . . . . 224Table 3 Logit regression model results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226Table 4 Observed and forecasted walking mode share for the whole

dataset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

Employment Density in Ile-de-France: Evidence from Local RegressionsRachel Guillain and Julie Le GalloTable 1 Distribution of employment in Ile-de-France . . . . . . . . . . . . . . . . 239Table 2 Spatial autocorrelation LM tests for model (3),

total employment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240Table 3 ML estimation results for global employment density

functions (1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241Table 4 ML estimation results for global employment density

functions (2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241Table 5 LM tests (maximum) of spatial autocorrelation and locational

heterogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243

The Geographic Dimensions of Electoral Polarization in the 2004 U.S.Presidential VoteIan Sue Wing and Joan L. Walker

Table 1 Spatial Durbin model results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272

xxii List of Tables

Gender Wage Differentials and the Spatial Concentrationof High-Technology IndustriesElsie Echeverri-Carroll and Sofia G. Ayala

Table 1 Determinants of (log of) individual hourly wages for maleworkers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299

Table 2 Determinants of (log of) individual hourly wages for femaleworkers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301

Table 3 Decomposition of the gender wage gap . . . . . . . . . . . . . . . . . . . . . 305

Fiscal Policy and Interest Rates: The Role of Financial and EconomicIntegrationPeter Claeys, Rosina Moreno, and Jordi Surinach

Table 1 Data sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319Table 2 Baseline model, pooled and panel estimates; and spatial panel

lag model (W-matrix D distance) . . . . . . . . . . . . . . . . . . . . . . . . . . 320Table 3 Baseline model, spatial panel error model

(W-matrix D distance) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322Table 4 Augmented model, spatial panel lag model, spatial fixed

effects, specifications (W-matrix D distance). See (4) . . . . . . . . 323Table 5 Baseline model, spatial panel lag, country groups

(W-matrix D distance) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328Table 6 Baseline model, spatial panel lag model, various weight

matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330Table 7 Augmented model, spatial panel lag model, spatial fixed

effects, specifications (W-matrix D distance). See (4) . . . . . . . . 333

Spatial Models of Health Outcomes and Health Behaviors: The Roleof Health Care Accessibility and AvailabilityBrigitte S. Waldorf and Susan E. Chen

Table 1 Physicians and nurses per 100,000 residents in 2004 . . . . . . . . . 346Table 2 Variable definitions and descriptive statistics . . . . . . . . . . . . . . . . 348Table 3 Spatial autocorrelation (Morans I ) of variables across Indiana

counties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352Table 4 Outcomes as a function of primary care availability (NURSE) 355Table 5 Behaviors as a function of primary care availability (NURSE) 356Table 6 Outcome as a function of accessibility of hospital care

(HOSPITAL) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357Table 7 Behavior as a function of accessibility of hospital care

(HOSPITAL) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358

List of Tables xxiii

Immigrant Women, Preventive Health and Place in Canadian CMAsKelly Woltman and K. Bruce Newbold

Table 1 Definition and coding of covariates . . . . . . . . . . . . . . . . . . . . . . . . 368Table 2 Multilevel logistic regression models: lifetime Pap uptake . . . . 370Table 3 Summary of variance (standard error) components, multilevel

logistic regression, lifetime Pap uptake . . . . . . . . . . . . . . . . . . . . . 372Table 4 Multilevel logistic regression models: regular Pap testing . . . . . 374Table 5 Summary of variance (standard error) components, multilevel

logistic regression, regular Pap use . . . . . . . . . . . . . . . . . . . . . . . . . 376

Is Growth in the Health Sector Correlatedwith Later-Life Migration?Dayton M. Lambert, Michael D. Wilcox, Christopher D. Clark, Brian Murphy,and William M. ParkTable 1 Summary statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 386Table 2 Model specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394Table 3 Regression results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396

Evolution of the Influence of Geography on the Location of Productionin Spain (19302005)Coro Chasco Yrigoyen and Ana M. Lopez Garca

Table 1 Variable list for the Spanish provinces . . . . . . . . . . . . . . . . . . . . . . 414Table 2 Descriptive Statistics of Relative GDP per area . . . . . . . . . . . . . . 419Table 3 Normality and spatial autocorrelation tests of log relative GDP

per area. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420Table 4 Second nature on first nature OLS regression results . . . . . . . . . 424Table 5 Instruments and endogeneity tests in second nature effect

regressions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427Table 6 OLS regression results of GDP per area on second nature

variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428Table 7 OLS regression results of GDP/area on second nature in two

spatial regimes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430Table 8 First and second nature joint effect on GDP density . . . . . . . . . . 433

Comparative Spatial Dynamics of Regional SystemsSergio J. Rey and Xinyue Ye

Table 1 Local Moran transition matrix in China (ND/D) . . . . . . . . . . . . . 451Table 2 Local Moran transition matrix in the United States (ND/D) . . . 451Table 3 Spatial dynamics in China . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453Table 4 Spatial dynamics in the United States . . . . . . . . . . . . . . . . . . . . . . . 454Table 5 Relative mobility of classic and local Moran Markov in China

and the United States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 459

xxiv List of Tables

Table 6 Local Moran transition probability matrix in China . . . . . . . . . . 459Table 7 Local Moran transition probability matrix in the United States 459

Growth and Spatial Dependence in EuropeWilfried KochTable 1 OLS and spatial error model (level model) . . . . . . . . . . . . . . . . . . 473Table 2 OLS and spatial error model (level model) . . . . . . . . . . . . . . . . . . 474Table 3 Spatial Durbin model (level model) . . . . . . . . . . . . . . . . . . . . . . . . . 475Table 4 OLS and spatial error model (convergence model) . . . . . . . . . . . 478Table 5 OLS and spatial error model (convergence model) . . . . . . . . . . . 479Table 6 Spatial Durbin model (convergence model) . . . . . . . . . . . . . . . . . . 480

Contributors

Patrick M. Andrade 150 North Martingale Road, Schaumburg, IL 60173, USA,Patrick.Andrade@nielsen.comand150 North Martingale Road, Schaumburg, IL 60173, USA

Ana Angulo Department of Economic Analysis, University of Zaragoza,Gran Via 2-4, Zaragoza 50005, Spain, aangulo@unizar.es

Sofia G. Ayala IC2 Institute, University of Texas at Austin, 2815 San Gabriel,Austin, TX 78705, USA, sofia ayala@mail.utexas.edu

Patrick Bogaert Department of Environmental Sciences & Land Use Planning,Universite Catholique de Louvain, ENGE Croix du Sud, 2, bte. 16 a` 1348,Louvain-la-Neuve, Belgium, patrick.bogaert@uclouvain.be

Ron N. Buliung Department of Geography, University of Torontoat Mississauga, 3359 Mississauga Road North, Mississauga, ON L5L 1C6,Canada, ron.buliung@utoronto.ca

Susan E. Chen Department of Agricultural Economics, Purdue University,403 W. State Street, West Lafayette, IN 47907-2056, USA, sechen@purdue.edu

Seong-Hoon Cho Department of Agricultural Economics, Universityof Tennessee, 321 Morgan Hall, Knoxville, TN 37996-4511, USA, scho9@utk.edu

George Christakos Department of Geography, San Diego State University, 5500Campanile Dr., San Diego, CA 92182-4493, USA, christak@geography.sdsu.edu

Peter Claeys AQR Research Group-IREA, University of Barcelona, AvingudaDiagonal 690, 08034 Barcelona, Spain, peter.claeys@ub.edu

Christopher D. Clark Department of Agricultural Economics, University ofTennessee, 321 Morgan Hall, Knoxville, TN 37996-4511, USA, cdclark@utk.edu

Sandy Dallerba Department of Geography and Regional Development,University of Arizona, P.O. Box 210076, Tucson, AZ 85721, USA,dallerba@email.arizona.edu

xxv

xxvi Contributors

Elsie Echeverri-Carroll IC2 Institute, University of Texas at Austin,2815 San Gabriel, Austin, TX 78705, USA, e.carroll@mail.utexas.eduSteven Farber Centre for Spatial Analysis/School of Geography and EarthSciences, McMaster University, 1280 Main Street West, Hamilton, ON L8S 3Z9,Canada, farbers@mcmaster.ca

Bernard Fingleton Department of Economics, Strathclyde University,130 Rottenrow, Glasgow, Scotland G4 0GE, UK, bernard.fingleton@strath.ac.uk

Julie Le Gallo Centre de Recherche sur les Strategies Economiques, Universitede Franche-Comte, 45D, Universite de Franche-Comte, 25030 Besancon Cedex,France, jlegallo@univ-fcomte.frAna M. Lopez Garca Dpto. Economa Aplicada, Facultad de CienciasEconomicas y Empresariales, Universidad Autonoma de Madrid, Carretera deColmenar Viejo Km. 15.500, Madrid 28049, Spain, ana.lopez@uam.esFrank Goetzke Department of Urban and Public Affairs, School of Urbanand Public Affairs, University of Louisville, 426 W. Bloom Street, Louisville,KY 40208, USA, f0goet01@louisville.edu

Rachel Guillain LEG-UMR 5118, Universite de Bourgogne, Pole dEconomieet de Gestion, BP 21611, 21066 Dijon Cedex, France, guillain@u-burgogne.frSeung Gyu Kim Department of Agricultural Economics, Universityof Tennessee, 321 Morgan Hall, Knoxville, TN 37996-4511, USA, sgkim@utk.eduWilfried Koch Laboratoire dEconomie et de Gestion, LEG-UMR 5118,Universite de Bourgogne, Pole dEconomie et de Gestion, BP 21611, 21066 DijonCedex, France, wilfried.koch@u-bourgogne.fr

Dayton M. Lambert Department of Agricultural Economics, Universityof Tennessee, 321 Morgan Hall, Knoxville, TN 37996-4511, USA,dmlambert@utk.edu

James P. Lesage McCoy College of Business Administration, Departmentof Finance and Economics, Texas State University-San Marcos, McCoy Hall 504,San Marcos, TX 78666, USA, jlesage@spatial-econometrics.comFernando Lopez Department of Quantitative Methods and Computing,Technical University of Cartagena, Paseo Alfonso XIII, 50 Cartagena 30203,Spain, fernando.lopez@upct.es

Catherine Morency Department of Civil, Geological and Mining engineering,Ecole Polytechnique de Montreal, P.O. Box 6079, Station Centre-Ville, Montreal,QC H3C 3A7, Canada, cmorency@polymtl.caRosina Moreno AQR Research Group-IREA, University of Barcelona, AvingudaDiagonal 690, 08034 Barcelona, Spain, rmoreno@ub.edu

Darla K. Munroe Department of Geography, The Ohio State University,154 N. Oval Mall, Columbus, OH 43210, USA, munroe.9@osu.edu

Contributors xxvii

Jesus Mur Department of Economic Analysis, University of Zaragoza,Gran Via 2-4, Zaragoza 50005, Spain, jmur@unizar.esBrian Murphy Department of Agricultural Economics, Universityof Tennessee, 321 Morgan Hall, Knoxville, TN 37996-4511, USA, bmurphy7@utk.edu

K. Bruce Newbold School of Geography and Earth Sciences,McMaster University, 1280 Main Street West, Hamilton, ON L8S 3Z9, Canada,woltmak@mcmaster.ca

R. Kelley Pace College of Business Administration, Department of Finance,Louisiana State University, Baton Rouge, LA 70803-6308, USA, kelley@pace.am

E.J. Ourso College of Business Administration, Department of Finance, LouisianaState University, Baton Rouge, LA 70803-6308, USA, kelley@pace.am

Antonio Paez Centre for Spatial Analysis/School of Geography and EarthSciences, McMaster University, 1280 Main Street West, Hamilton, ON L8S 3Z9,Canada, paezha@mcmaster.ca

William M. Park Department of Agricultural Economics, Universityof Tennessee, 321 Morgan Hall, Knoxville, TN 37996-4511, USA, wpark@utk.edu

Sergio J. Rey Department of Geography, San Diego State University and Schoolof Geographical Sciences, Arizona State University, P.O. Box 875302, Tempe,AZ 85287, USA, Sergio.Rey@asu.edu

Roland K. Roberts Department of Agricultural Economics, Universityof Tennessee, 2621 Morgan Circle, Knoxville, TN 37996-4511, USA, rrobert3@utk.edu

Jordi Surinach AQR Research Group-IREA, University of Barcelona, AvingudaDiagonal 690, 08034 Barcelona, Spain, jsurinach@ub.eduErik Volz Department of Epidemiology, School of Public Health, Universityof Michigan, 109 Observatory Street, Ann Arbor, MI 48109, USA,erik@erikvolz.info

Brigitte S. Waldorf Department of Agricultural Economics, Purdue University,403 W. State Street, West Lafayette, IN 47907-2056, USA, bwaldorf@purdue.edu

Joan L. Walker Department of Civil and Environmental Engineering, Universityof California, Berkeley, 760 Davis Hall, Berkeley, CA 94720-1710, USA,walker@ce.berkeley.edu

Michael D. Wilcox Department of Agricultural Economics, Universityof Tennessee, 321 Morgan Hall, Knoxville, TN 37996-4511, USA, mwilcox2@utk.edu

Ian Sue Wing Department of Geography & Environment, Boston University,675 Commonwealth Avenue, Boston, MA 02215, USA, isw@bu.edu

xxviii Contributors

Kelly Woltman School of Geography and Earth Sciences, McMaster University,1280 Main Street West, Hamilton, ON L8S 3Z9, Canada, woltmak@mcmaster.ca

Xinyue Ye Department of Geography, Joint Doctoral Program of Geography,San Diego State University and University of California-Santa Barbara, 5500Campanile Drive, San Diego, CA 92182-4493, USA, xinyue.ye@gmail.com

Coro Chasco Yrigoyen Dpto. Economa Aplicada, Facultad de CienciasEconomicas y Empresariales, Universidad Autonoma de Madrid, Carretera deColmenar Viejo Km. 15.500, Madrid 28049, Spain, coro.chasco@uam.esHwa-Lung Yu Department of Bioenvironmental Systems Engineering, NationalTaiwan University, 1, Section 4, Roosevelt Road, Taipei 10617, Taiwan, R.O.C.,hlyu@ntu.edu.tw

Edmund J. Zolnik Department of Geography and Geoinformation Science,George Mason University, Fairfax, VA 22030, USA, ezolnik@gmu.edu

Progress in Spatial Analysis: Introduction

Antonio Paez, Julie Le Gallo, Ron N. Buliung, and Sandy DallErba

1 Background

With its roots in geography and regional science spatial analysis has experiencedremarkable growth in recent years in terms of theory, methods, and applications. Theseries of books, that in the past decade have collected research in spatial analysis andeconometrics, provide both documented evidence and a powerful platform to furtherthis upwards trend. Among the collections that have done so stand those compiledby Anselin and Florax (New Directions in Spatial Econometrics, 1994), Fischer andGetis (Recent Developments in Spatial Analysis, 1997), and Anselin, Florax and Rey(Advances in Spatial Econometrics, 2004). In the spirit of this series of volumes, thepresent book aims at promoting the development and use of methods for the analysisof spatial data and processes.

Traditionally, the core audience for the spatial analysis literature has been foundin the Quantitative Geography and Regional Science communities, but also increas-ingly within the allied disciplines of Spatial and Regional Economics, Urban andRegional Planning and Development, Civil Engineering, Real Estate Studies, andEpidemiology, among others. Previous edited volumes, in particular the two spatialeconometrics collections cited above, tended to emphasize, in addition to theoreticaland methodological developments, economics and regional economics applications.In this book, we have made an attempt to capture a broader cross-section of themes,to include fields where spatial analysis has represented in recent years a boonfor applications, which have in turn encouraged further technical developments.Besides the disciplines represented in previous collections of papers, up-and-comingareas that are seen to be making more extensive use of spatial analytical tools includetransportation and land use analysis, political and economic geography, and theanalysis of population and health issues. In order to provide a faithful picture of the

A. Paez (B)Centre for Spatial Analysis/School of Geography and Earth Sciences, McMaster University,1280 Main Street West, Hamilton, ON L8S 3Z9, Canada,e-mail: paezha@mcmaster.ca

A. Paez et al. (eds.), Progress in Spatial Analysis, Advances in Spatial Science,DOI 10.1007/978-3-642-03326-1 1, c Springer-Verlag Berlin Heidelberg 2010

1

2 A. Paez et al.

current state of spatial analysis it is also our wish to present recent theoretical andmethodological developments. Together, this collection of theoretical and method-ological papers, and thematic applications, will project, we hope, the image of athriving and dynamic field, with wide-ranging intellectually stimulating challenges,and rich opportunities for applied research that promises to promote and advancedata analysis in a variety of fields.

In terms of the contributions collected for this volume, the papers represent aselection of research presented at the 54th North American Meetings of the RegionalScience Association International celebrated in Savannah, Georgia, in November of2007, as well as a small number of invited papers. All contributions were subjectedto a strict process of peer review; the outcome is a set of papers that have been orga-nized, in addition to a section on Theory and Methods, into four thematic sections:Transportation and Land Use Analysis, Population and Health Issues, Political andEconomic Geography, and Regional Applications. Some of these areas have tra-ditionally been associated with the use of spatial analytical tools (e.g., regionalapplications). Others represent nascent opportunities for the development and useof spatial analysis (e.g., transportation and land use, population and health). It isour hope that this edited volume will simultaneously help to consolidate the reputa-tion and value of spatial analysis established by previous titles in the series, and toincrease awareness about the utility of spatial analysis in other application domains.

2 Theory and Methods

Five chapters comprise the section on theory and methods. Pace and LeSage, inchapter Omitted Variable Biases of OLS and Spatial Lag Models, address a ques-tion that has received relatively little attention in the spatial econometrics literature,namely, the effect of omitted variables in regression analysis. This research is moti-vated by the conjecture that omitted variable bias is less severe in spatial models thanin ordinary regression approaches. One of the bases for this conjecture is that theadditional components in a spatial model are perhaps sufficiently capturing missingrelationships to offset the effect of bias. The problem of omitted variables in spatialanalysis, on the other hand, is complicated by the fact that spatial variables oftendisplay non-negligible amounts of spatial autocorrelation. Most likely, this will bethe case for both the included and the omitted variables. In order to sort out whatthe impacts of this are, the authors develop a very general framework that allowsthem to derive results for a wide range of situations likely found in applied research.The analytical derivations presented in the chapter are backed by extensive simula-tion experiments that help to give a feeling for the magnitude of bias under differentcases. The results indicate that, contrary to the original conjecture, omitted variablebias is magnified by the presence of spatial dependence. Several implications leadto useful guidelines for applied research.

Progress in Spatial Analysis: Introduction 3

Chapter Topology, Dependency Tests and Estimation Bias in Network Autore-gressive Models, by Farber, Paez and Volz, also deals with a specification issue inspatial modeling, namely the definition of spatial weights matrices, the instrumentused to specify how spatial cross-sectional observations are connected. While thismatrix is usually defined based on geographic criteria (e.g., contiguity, distance-based matrices, nearest-neighbors matrices etc.), there has recently been increas-ing interest in using a network-based connectivity specification. The subject of thischapter is the structure of the weights matrix and the effect of network topologyon the estimation of network autocorrelation models and statistical tests of depen-dence. The authors investigate, both analytically and through extensive Monte-Carlosimulations, the power of the likelihood ratio (LR) tests for network dependencein SAR and SEM models. They first show that for all the model specifications,the level of network dependence is the most significant factor in predicting thepower of the LR test, albeit with a non linear effect and differently for SAR andSEM models. Second, the effects of network density and clustering on the powerof the LR test are analyzed. Finally, the relationship between bias and the varioustopological properties of networks are graphically illustrated. In sum, the vari-ous results unambiguously show that the topology of the weights matrix used inautocorrelation models has a strong impact on statistical tests and the accuracy ofmaximum-likelihood estimates.

Fingleton and Le Gallo, in chapter Endogeneity in a Spatial Context: Propertiesof Estimators, are concerned with the important issue of identifying appropriateestimators when dealing with endogeneity in a spatial econometric context. Whilethe appropriate treatment and estimation of the endogenous spatial lag has receiveda good deal of attention, the analysis of effects related to other endogenous vari-ables has been less popular. Based on their previous work, the authors focus onthe case where endogeneity is induced by the omission of a (spatially autoregres-sive) variable. They show the inconsistency of the usual OLS estimators induced byomitting a significant variable that should be in the regression model but which isunmeasured and hence is present in the residual. A simulation experiment is imple-mented that demonstrates how an augmented spatial Durbin model with a complexerror process is a reasonably appropriate estimator. This is estimated using 2SLS(2 Stage-Least-Square) and SHAC (Spatial Heteroskedasticity and AutocorrelationConsistent) estimator for the variance-covariance matrix. This estimator performsbetter in terms of bias and Root Mean Square Error than the OLS-SHAC estimator.They reach the same conclusion when they modify the properties of the omittedvariable used in their Monte-Carlo simulations. The discussion moves on to thecase where endogeneity is a consequence of simultaneity and errors in variables.The authors conclude again that the 2SLS-SHAC estimation of the spatial Durbinmodel is better than an OLS-SHAC estimation of a single equation model where theendogeneity problem remains untreated.

Chapter Dealing with Spatiotemporal Heterogeneity: The Generalized BMEModel by Yu, Christakos and Bogaert, discusses a stochastic approach for studyingphysical and social systems and their attributes, when these systems are charac-terized by heterogeneous space-time variations under conditions of multi-sourced

4 A. Paez et al.

uncertainty. The proposed Generalized Bayesian Maximum Entropy approachemerges from the fusing together of generalized spatiotemporal random field the-ory and a Bayesian Maximum Entropy mode of thinking. The result is a versatileapproach to conduct spatiotemporal analysis and mapping that exhibits a numberof attractive features, including the following: the approach makes no restric-tive assumptions concerning estimator linearity and probabilistic normality (i.e.,non-linear estimators and non-Gaussian distributions are naturally incorporated);it can be used to study natural systems with heterogeneous space-time dependencepatterns; it can also account for various kinds of physical knowledge (core and case-specific) concerning the system under study; and it provides a general frameworkfrom which mainstream methods can be derived as special cases. The proposedspace-time approach is applicable in a variety of knowledge domains (e.g., phys-ical, health, social and cultural). Numerical experiments provide key insights intothe computational implementation and comparative performance of the approach.

Along the lines of spatial heterogeneity, a long standing question refers to insta-bility or nonstationarity in spatial models. Although this issue can be traced backto the development of Casettis expansion method in the early 1970s, it has claimedrenewed attention in light of newer methods for exploring local variations in spa-tial autocorrelation patterns and multivariate relationships (e.g., LISA, Getis-Ordstatistics, geographically weighted regression or GWR). The problem of spatialinstability is important as it refers to the well-known problem of the complex rela-tions between spatial heterogeneity and spatial autocorrelation. The last chapterin this section by Lopez, Mur, and Angulo, approaches this issue and investigatesmodels where the intensity of spatial autocorrelation depends on the geographicallocation of each observation. In this respect, the chapter first presents a simple LMtest of parameter instability for the spatial autocorrelation coefficient in a spatial lagmodel. Second, an extensive Monte-Carlo exercise is undertaken to study the distor-tions affecting the usual cross-sectional diagnostic measures (spatial autocorrelationLM tests, Jarque-Bera, Breusch-Pagan, White and RESET tests), when the assump-tion of constant spatial autocorrelation does not hold. Third, a local estimationalgorithm labeled zoom estimation, which can be considered an extension of theSALE model (Pace and LeSage 2004) is suggested and its performance with regardto the zoom size is evaluated with Monte-Carlo experiments. Finally, a strategy toidentify spatial regimes in the spatial autocorrelation coefficient is proposed andcompared to four other strategies based respectively on the k-means algorithm,Gaussian mixture models for multipolarity, Getis-Ord statistics, and trimmed meanclassification rule. By providing novel information regarding the effect of spatialinstability on usual diagnostic measures and specification search strategies, as wellas giving suggestions to identify the presence and form of spatial regimes, thischapter represents a valuable step toward increasing our understanding of spatialinstability in spatial econometric models.

Progress in Spatial Analysis: Introduction 5

3 Thematic Applications

3.1 Spatial Analysis of Land Use and Transportation Systems

The impressive visual qualities of transport and land use systems and processes inthe real world have arguably not been matched by an equally impressive and con-structive exercise in abstract data visualization. The availability of both proprietaryand open environments for data analysis and visualization, coupled with the imple-mentation of innovative approaches for data visualization presents an opportunityto advance the state-of-the-art with regards to the visual communication of spa-tial, temporal, and social qualities of transport and land use processes. Moreover,progress in automatic data collection through onboard GPS, cellular phone traces,or smart cards increases requirements for useful approaches and tools for summa-rizing and communicating the complexity and relevance of emerging modalitiesfor communication and spatial interaction. Chapter Seeing Is Believing: Explor-ing Opportunities for the Visualization of ActivityTravel and Land Use Processesin SpaceTime, by Buliung and Morency, has as its objective to introduce recentinnovations with regards to both platforms and approaches for the visualization oftransportation and land use processes. To draw a parallel with the arts, visualiza-tion can be compared to an anamorphosis interpreter wherein the act of visualizationmakes use of specialized devices (e.g., computer programs, statistical tools, GIS,interactive spreadsheets), or compels the viewer to occupy a specific perspective(e.g., spatial, temporal or social feature), with a view to reconstituting the originalfor the purpose of developing a clearer understanding of process. Using examplesdrawn primarily from Montreal and the Greater Toronto Area, Canada, this chapterdemonstrates how visualization techniques and tools can be used, often in a comple-mentary way, to clarify transport- and land use-related spatial, temporal and socialprocesses.

In addition to the exploration of transportation and land use processes, throughvisualization techniques, there has been considerable recent work on the confirma-tory analysis, via multivariate techniques, of transport and land use phenomena.Four papers in this section apply spatial analytical techniques to the investigation ofdifferent aspects of land development and travel behavior. The first contribution inthis group, by Munroe, is concerned with the expansion and rapid growth of urbanareas, a process that can occur unevenly across space and through time. The avail-ability of detailed spatial and temporal data describing land use, combined withthe application of spatial and temporal modeling approaches (e.g., spatial logis-tic regression, hazard models), facilitates, in Chapter Pattern-Based Evaluationof Peri-Urban Development in Delaware County, Ohio, USA: Roads, Zoning andSpatial Externalities, the detection and description of the global and local spatialproperties of urban growth i.e., dispersion, decentralization, fragmentation. Themore abstract conceptualization of urban sprawl, as a somewhat even and regularexpansion of urban areas into rural or peri-urban places, can be replaced by a moredetailed, empirically informed view of key development processes and outcomes.

6 A. Paez et al.

Investigation of peri-urban development in Delaware County, Ohio, is based on adiscrete time-to-event model for Delaware County, one of the fastest growing coun-ties in the US, located north of the state capital of Columbus, Ohio. Overall, theresults suggest that the process of urban expansion/dispersion has simultaneouslyincluded an increase in the local clustering of development. A simulation experimentexamines the sensitivity of predicted patterns of residential growth to policy and/ormarket-based drivers of growth processes including: density (intensification), accessto roads, and development externalities. Controlling for the timing of development,avoidance of development, maximum density zoning policies, and distance to majorroads emerge as factors contributing to the fragmentation of residential developmentin the county. From a policy perspective, the findings suggest that cooperative landuse management at the township level, and open space preservation, are potentiallyuseful approaches to control the growth processes described within the chapter.

Also related to the topic of sprawling development, Chapter Demand for OpenSpace and Urban Sprawl: The Case of Knox County, Tennessee by Cho, Lam-bert, Roberts and Kim, is concerned with the demand for open space. While thereis limited consensus in the literature regarding the conceptualization and measure-ment of urban sprawl, scholars, practitioners, and governments, consider the studyand implementation of growth management to be an important intellectual andpractical exercise. The research reported in this chapter makes use of a two-stepspatial modeling approach to examine the efficacy of open space conservation asa policy tool for managing urban sprawl. The conceptualization of sprawl chosenby the authors includes processes of expansion or encroachment into rural areas,and the leapfrogging of development. The case study of Knox County presentsan interesting situation because the county has experienced rapid growth overall,with some local heterogeneity (spatially and temporally) in the pace of residen-tial development. Analysis is supported by a very detailed spatial database of theregion obtained from secondary sources, and the use of GIS techniques and remotesensing data to quantify household access to open space. The spatial modeling taskcombines hedonic price modeling with geographically weighted regression. Com-parative analysis of model results indicates that the GWR (spatial error) modelprovides an important complement to the global (OLS) alternative. With regard topolicy, the results appear to be open to several interpretations; acting freely in themarket, affluent households may be willing to pay (i.e., buy into policy) to preserveopen space, on the one hand, or demand open space at the edge, potentially givingrise to additional and perhaps undesirable patterns of growth particularly in theabsence of an appropriate regulatory framework.

The next two chapters are concerned with issues in travel behavior. One themethat has interested geographers and planners for some time is the existence ofdifferential patterns of mobility by gender. The tenth chapter is entitled MultilevelModels of Commute Times for Men and Women. In his contribution, Zolnik exam-ines, from a spatial perspective, the well-documented issue of the commuting-timegender gap. Research has often presented evidence suggesting that women typi-cally have shorter commutes than men. Sociological and economic explanations

Progress in Spatial Analysis: Introduction 7

have been advanced, with some recent evidence suggesting some convergence in thegeneralized cost of commuting, particularly at the margins of male/female incomedistributions, and within certain occupational or ethnic groups. The research pre-sented in this chapter draws independent (male, female) and pooled (male andfemale) samples from the 2001 US National Household Travel Survey, which areused to estimate multi-level models of self-reported journey to work commutetime. The samples included individuals who worked and commuted (within a sin-gle Metropolitan Statistical Area) by private vehicle the week prior to the survey.Income and occupation effects appeared to be stronger for women, while access toprivate vehicles appeared to have a stronger positive influence on commute times formale workers. Interestingly, Zolnik concludes that his findings lend little support tothe household responsibility hypothesis. Apart from strong congestion effects dif-ferentiable by sex, his findings suggest only marginal commute time savings withchanges in development intensity and the mixing of land uses.

The final chapter in this section, by Goetzke, is concerned with two topics of cur-rent interest from the spatial analysis and travel behavior perspectives: the role ofspatial effects in choice models, and the possibility that information spillovers maylead to interdependent choice processes. The research reported in chapter Walka-bility as a Summary Measure in a Spatially Autoregressive Mode Choice Model:An Instrumental Variable Approach is motivated by the difficulties posed by thenon-linear functional form of spatially autoregressive binary choice models (logit orprobit models), especially if the analyst does not wish to assume a conditional spa-tial structure, which has the disadvantage that it imposes a strong restriction on themodel. On the other hand, a linear probability model (LPM) can easily be extendedto a spatially autoregressive model with few additional difficulties. However, aLPM exhibits by definition always heteroskedasticity, which makes the estimationinefficient. Empirically, the model proposed is demonstrated using the 1997/1998New York Metropolitan Transportation Council comprehensive regional householdtravel diary survey, in analysis that aims at determining whether social spill-overeffects exists for walking commutes in Manhattan (i.e., a large enough sample sizeto adequately capture pedestrian behavior). The spatial process is modeled usingthe instrumental variable 2SLS method. In a third step, the LPM is additionallycorrected for heteroskedasticity using a weighted least square approach with theassumption of a binomial distribution in the error term. The estimation method pro-posed is extended to a probit/logit model where the spatial process is also modeledusing the instrumental variable approach (spatially autocorrelated IV probit/logitmodel). The results of both models are compared with the results of a condi-tional spatially autoregressive probit/logit model. This application shows that theinstrumental variable method for estimating spatially autoregressive probabilitymodels is able to overcome the shortcomings of a conditional spatially autoregres-sive binary choice model, besides being relatively straightforward to implement.

8 A. Paez et al.

3.2 Economic and Political Geography

The second thematic section is comprised of four chapters dealing with varioustopics in economic and political geography. In Chapter Employment Density inIle-de-France: Evidence from Local Regressions, Guillain and Le Gallo addressthe issue of suburbanization in the Ile-de-France region in France. With the devel-opment of peripheral employment centers, the spatial organization of the regionsactivities does not necessarily correspond to the traditional monocentric model. Theaims of this chapter are first to understand whether the Central Business District(CBD) does still influence the employment distribution in Ile-de-France, and sec-ondly, if so, whether this effect differs by sector. In order to answer these questions,the authors identify first the location of employment centers by measuring the spatialagglomeration of economic activities, with global and local spatial autocorrelationstatistics. Second, they conduct an in-depth analysis of the centers by identify-ing their sectoral specialization and their attractiveness for strategic activities. Theauthors use various spatial econometric specifications of the density functions andperform local regressions, using geographically weighted regression, where the rateat which density falls with distance from the CBD is estimated for each observation.The local regressions facilitate the detection of changes in density by distance (het-erogeneous distribution) and direction (anisotropic distribution) from the CBD. Themain results of this study indicate that the CBD still influences total employment inIle-de-France but that its influence varies by sector, distance, and direction from theCBD. From a political viewpoint, their conclusions provide new insights about thelocation strategies of households and economic activities in Ile-de-France.

Chapter The Geographic Dimensions of Electoral Polarization in the 2004 U.S.Presidential Vote, by Sue Wing and Walker, is motivated by the apparent divi-siveness of the 2004 US presidential election. This observation gave rise to theexploration of the hypothesis that the U.S. electorate is geographically polarized.Using spatial econometric analyses, these authors investigate the effects of the char-acteristics of populations and places on voter turnout in favor of George W. Bush.Specifically, the authors identify key factors affecting Bushs odds of success at thenational level, and demonstrate how these aggregate effects vary over finer spatialscales. The results provide an intriguing first look at overall spatial patterns in thecorrelates of voting behavior, and argue for a new way of thinking of polarization asa phenomenon which occurs within individual sub-groups across space, with geog-raphy playing a crucial role at both local and regional scales, but in ways which arenot easily categorized or explained.

The topic analyzed in chapter Gender Wage Differentials and the Spatial Con-centration of High-Technology Industries, by Echeverri-Carroll and Ayala, dealswith gender wage differentials in cities and is relevant for the study of the issuesof agglomeration, the productivity of cities and the existence of localization andurbanization economics. From a methodological perspective, the work deals withspecific econometric problems linked to the analysis of spatial microeconomicdata: heteroskedasticity and endogeneity. Previous studies have found that maleworkers attain higher wages in cities (high-tech cities in particular) with a large

Progress in Spatial Analysis: Introduction 9

endowment of human capital than in those with a low endowment. New Eco-nomic Geography models maintain that the higher wages of males are linked toproductivity-enhancing effects from the (formal and informal) exchange of knowl-edge that characterizes high-tech cities. The authors question whether women enjoysimilar productivity-enhancing effects. A large sample is drawn from the 5% PUMSof the 2000 Census of Population, and is used to estimate regressions separatelyfor a sample of male and female workers, accounting for arbitrary clustering, het-eroskedasticity in the error terms, and endogeneity. The estimates show that aftercontrolling for individual- and city-level variables that affect wages, male workersthat live in a high-tech city and work in a high-tech industry, holding other factorsfixed, indeed earn more than comparable female workers. The results support theview that women might benefit less from knowledge networks that are predominantamong high-tech industry workers in high-tech cities and from the demand for talentexercised by these industries.

Among the wide range of spatial econometric applications, fiscal and mone-tary economic applications remain quite scarce. Chapter Fiscal Policy and InterestRates: The Role of Financial and Economic Integration, by Claeys, Moreno andSurinach, fills this gap by analyzing the role of spatial spillovers in the crowding-out effects of fiscal expansion on interest rates. The chapter parts from the commonbelief that fiscal expansion raises interest rates. However, the crowding-out effectsof deficits have been found to be small or non-existent. One explanation is thatfinancial integration offsets interest rate differentials on globalized bond markets.As a result, the authors measure the degree of integration of government bond mar-kets, using spatial modeling techniques, with a view to taking this spillover effecton financial markets into account. Using a panel of 101 countries and annual dataon interest rates and fiscal policy covering the period 19902005, the main findingis that the crowding out effect on domestic interest rates is significant, but that itis reduced by spillover across borders. The detected spillover effect is importantin major crises or in periods of coordinated policy actions. The result is generallyrobust to various measures of cross-country linkages, and indicates strong spillovereffects among EU countries.

3.3 Spatial Analysis of Population and Health Issues

The next three chapters in the volume are concerned with the spatial analysis ofvarious aspects of population and health. In chapter Spatial Models of HealthOutcomes and Health Behaviors: The Role of Health Care Accessibility and Avail-ability, Waldorf and Chen address the question of whether poor spatial accessibilityto health care providers leads to poor health outcomes. Their work focuses on the 80counties of Indiana, a state that, as many other across the US, experiences a spatialmismatch between the location of supply and demand for medical care as well asspatial variations in the quality of medical facilities. This problem presents a chal-lenge for policymakers who need to determine how to equitably allocate medical

10 A. Paez et al.

resources to improve public health in general and help medically underserved ruralareas in particular. The study is grounded on the measurement of accessibility(as opposed to availability) of health care providers in order to better capture thedistance-cost faced by patients wishing to receive treatment. It is worthwhile to notethat while accessibility to health care has been extensively studied, the approachpresented in this chapter is innovative for two reasons. First, accessibility to healthcare is linked through a modeling framework to health outcomes. Second, the mod-els are estimated after the inclusion of spatial dependence effects. In the case ofhealth outcomes, spatial dependence may be a statistical artifact, but it can also begrounded in behavioral processes such as imitation behavior and the spatial diffu-sion of cultural norms influencing health care utilization. These effects could alsobe a result of underlying factors such as poor labor market conditions which affectpeoples access to health insurance and thus ultimately peoples health. The mod-els reported in the chapter are estimated for six health outcome variables relatingto the health status of infants and the elderly, and four health behavior variables.The results indicate that the impact of health behaviors, health care accessibility,and spatial dependence varies across the various health outcomes investigated. Theauthors conclude that, from a policy perspective, it is important to recognize thatefforts to improve health behaviors in one county could impact health behaviors inneighboring counties, eventually trickling down through an entire state.

While the research of Waldorf and Chen is concerned with the effect of acces-sibility to health care on health outcomes, the work of Woltman and Newbold inchapter Immigrant Women, Preventive Health and Place in Canadian CMAs isrelated to the utilization of health services, with a particular focus on immigrantsin Canada. While the health status of immigrants has been studied extensively, thehealth service challenges facing immigrants are perhaps less understood. This chap-ter advances current thinking on the use of health care services by immigrant womenin Canadian Census Metropolitan Areas (CMAs), and more specifically, the utiliza-tion by immigrant women of cervical cancer screening. Analysis is conducted byexamining the multilevel factors associated with Pap (smear) testing in native-bornand immigrant women. Cross-sectional multi-level logistic regression analysis isthen used to detect individual and neighborhood level correlates of lifetime uptake(i.e., ever had a Pap test), and regular use (i.e., a test within the last three years) ofPap testing. Individual data are drawn from the Canadian Community Health Survey(Cycle. 2.1, 2003) for the population of interest, namely immigrant and native-bornwomen between the ages of 18 and 69, living in the Montreal, Toronto, and theVancouver Census Metropolitan Areas. Contextual factors are constructed by link-ing individual level data with census tract profile data from the 2001 Census ofCanada. The results indicate the presence of between-neighborhood variation inuptake. Immigrant status and cultural origin appear to be significantly associatedwith lifetime uptake, although uptake appears to be less common amongst recentimmigrant women and women of Chinese, South Asian and other Asian back-grounds. The results also suggest that neighbourhood disadvantage (i.e., a compositeindex) and immigrant concentration are positively associated with regular Pap test-ing. Findings concerning the role of culture and immigration status, coupled with

Progress in Spatial Analysis: Introduction 11

the reported neighbourhood effects, lend support to the development of neighbour-hood level interventions focused on increasing the awareness of recent immigrantwomen of the availability of cervical cancer screening services.

The last chapter in this section, by Lambert, Wilcox, Clark, Murphy and Park,combines in the most explicit way the two themes of population and health.The question posed for this chapter is the extent to which new-generation retire-ment communities are responsible for agglomeration within the health care sector.Demographers estimate that over the next 18 years at least 400,000 retiring babyboomers will migrate beyond their state borders each year, carrying with them anaverage of $320,000 to spend on a new home. It is no surprise then that migrat-ing retirees can stimulate economic growth and development in their host ruralcommunities. Factors of import to migrating seniors with respect to residential siteselection include health care service availability, recreational amenities, affordablehousing, low taxes, and proximity to friends and family. The geographical focusof this paper is the Southeastern US, an area that has experienced an extraordinaryinflux of retiring seniors since 1990. As more retiring seniors choose a particularresidential location, demand for health services will presumably increase, creat-ing new employment opportunities. On the other hand, migrating seniors may beattracted to communities with a wider array of health care services. This prob-lem is reminiscent of the jobs-to-people/people-to-jobs conundrum. In order totease out these relationships, the authors draw from recent developments in the spa-tial econometric literature to develop a regional adjustment econometric model thataccounts for endogeneity and heteroskedasticity. The results of the analysis suggestthat rural communities, able to support diversified health services are at a compara-tive advantage with respect to attracting retirees, whereas provision of such servicesin counties near metropolitan centers appears to be of reduced importance. In addi-tion, there is evidence that retiree in-migration is correlated with overall growth inthe health sector.

3.4 Regional Applications

The last set of papers in the book includes applications of spatial analysis toquestions focused on regional systems. Chapter Evolution of the Influence ofGeography on the Location of Production in Spain (19302005), by Chasco andLopez, is concerned with the relative importance of geographic features on the loca-tion of production in Spain. Based on a panel of 47 Spanish provinces and the19302005 period, they quantify how much of the spatial pattern of GDP can beattributed to only exogenous first nature elements (physical and political geography)and how much can be derived from endogenous second nature factors (man-madeagglomeration economies). The authors employ an analysis of variance (ANOVA)to infer the unobservable importance of first nature indirectly in a stepwise proce-dure. In order to disentangle the two net effects empirically, as well as their mixedeffect, they control for second nature because every locational endowment will be

12 A. Paez et al.

reinforced and overlaid by second nature advantages. In a dynamic context, theyalso estimate how much agglomeration can be explained by both gross and net sec-ond nature with the aim of isolating the importance of first nature alone. The authorsstress the fact their results could be biased if some potential econometric questions(multicollinearity, relevant missing variables, endogeneity and more particularlyspatial effects) were not properly taken into account. They conclude that produc-tion is not randomly distributed across Spanish regions: 88% of the GDPs spatialvariation can be explained by the direct and indirect effects of geography. Aftercontrolling for agglomeration economies and interaction effects of the first/secondnature, the net influence of natural geography goes from 20% in 1950 to 67%nowadays. On the other hand, while second nature agglomeration forces (e.g., trans-port and communications) were dominant in the 1930s, they were overcome by firstnature geography by the end of the period. These results also differ across the twospatial regimes that characterize the country: the coast plus the Madrid metropolitanarea, and the hinterland. Overall, the research presented in this chapter representsan innovative way to measure the extent to which regional policies are able to favoragglomeration in areas without clear geographic advantages.

The spatial dynamics of regional systems is the topic of chapter ComparativeSpatial Dynamics of Regional Systems, by Rey and Ye, and in particular, thedynamics of income convergence. These authors note that the focus of research, hav-ing shifted from the national to the regional perspective in the early 1990s, continuedto be dealt with using the same theoretical and technical frameworks underpin-ning national growth research. By the end of the 1990s, however, the geographicaldimension of convergence issues had already attracted substantial attention. Thischapter contributes to the literature on income convergence by considering two ofthe worlds largest and deeply entangled economies, the US and China, at differ-ent developmental stages, and by bringing to bear some of the most recent tools inexploratory spatial data. In addition to their use for convergence analysis, the newset of statistical measurements introduced in this chapter open up new opportunitiesfor scientific visualization and the generation of hypothesis in other fields that dealwith dynamic space-time processes.

The closing paper, contributed by Koch, examines regional growth and conver-gence. The literature focusing on issues of growth and convergence from the specificperspective of spatial econometrics techniques is today extensive. The studies inthis area focus on the interdependence between nations and regions, highlightinghow the economy of one country or region is not independent of the economiesof neighbouring countries or regions (and perhaps non-neighbours as well). How-ever, a common feature of these papers is that the spatial econometric specificationsare introduced in an ad hoc way, i.e., spatial lag or spatial error models are esti-mated, and the choice of the specification tends to be based on statistical criteria.Recently, theoretical foundations of spatial dependence have been suggested. Chap-ter Growth and Spatia