World Bank Reprint Series: Number Sixty-four Sebastian Pinera and Marcelo Selowsky The Opportunity Cost of Labor and the Returns to Education under Unemployment and Labor Market Segmentation Reprinted from The QuarterlyJournal of EconZom1ics (August 1978) Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized
Transcript
World Bank Reprint Series: Number Sixty-four
Sebastian Pinera and Marcelo Selowsky
The Opportunity Costof Labor and theReturns to Educationunder Unemploymentand Labor MarketSegmentation
Reprinted from The QuarterlyJournal of EconZom1ics (August 1978)
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
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
The World Bank Catalog lists all publications of the World Bank and is avail-able without charge to individuals and institutions having a serious interestin economic and social development. Address requests for the Catalog to:Publications Office, The World Bank, Washington, D.C. 20433, U.S,A., or tothe World Bank European OCfize, 66, avenue d'ilna, 76116 Paris, France.
WORLD BANK BOOKS ABOUT DEVELOPMENT
Research Publications
Housing for the Urban Poor. Economics and Policy in the Developing World byOrville F. Grimes, Jr., published by The Johns Hopkins University Press, 1976
Electricity Economics: Esa,ays and Case Studies by Ralph Turvey and DennisAnderson, published by The Johns Hopkins University Press, 1976
Village Water Supply: Econ?'nmics and Policy In the Developing World by RobertSaunders and Jerei-ny Wariord. published by The Johns Hopkins UniversityPress, 1976
Economic Analysis of Projects .1y Lyn Squire and Herman G. van der Tak,published by TW-? Joh Hopkins University Press, 1975
The Design of Rural Dev&lent: Lessons from Africa by Uma Lele, publishedby The Johns Hopkins iiiUiversitW Press, 1975
Economy-Wlde Models and Daveh.'pment Planning edited by Charles R. Blitzer,Peter B. Clark, and Lance Taylor, published by Oxford University Press, 1975
Patterns of Developmen7t, i 950-1970 by Hollis Chenery and Molses SyrquIn withHazel Elkington, published by Oxford University Press, 1975
A System of Internatiornal Comparisons of Gross Product and Purchasing Powerby Irving B. Kravis, Zoltan •Kenessey, Alan Heston, and Robert Summers,published by The Johns Hopkins University Press, 1975
Attacking Rural Poverty: Nonformal Education Can Help by Philip H.Coombs with Manzoor Ahmed, published by The Johns Hopkins UniversityPress, 1974
Country Economtilo Rtppu,N
Chad: Development Pot4ntiet and ;/ontrairfs by Richard Westebbe and others,distributed by The Johns Hopki.is Un!verslty Press, 1974
Economic Growth of Colombia: Pr oblems and Prospects by DragoslavAvramovic and others, published byThe Johnc. Hopkins University Press, 1972
The Current Economic Position end Prospects of Ecuador by Roberto Echeverrlaand others, distributed by The Johns Hopkins UnIverslty Press, 1973
Kenya: Into the Second Decade by John Burrows and others, published by theJohns Hopkins Universltv FPress, 1975
Korea: Problems a'-i lsuvie.s in a Rapidly Growing Economy by Parvez Hasan,published by The Johns r--Ipklns University Press, 1976
Lesotho: A Developmeiit 4Challenija by Willem Msane, distributed by The JohnsHopkins University Press, 1 975
(continued on inside back cover)
THE OPPORTUNITY COST OF LABOR AND THERETURNS TO EDUCATION UNDER UNEMPLOYMENT
AND LABOR MARKET SEGMENTATION*
SEBASTIAN PINRERA AND MARCELO SELOWSKY
I. Introduction, 469.-IY. Segmented urban labor markets and the existence ofvoluntary unemployment, 472.-IlI. The employment-unemployment decision-makingprocess, 473.-IV. Deriving the contribution of labor to output, 476.-V. Empiricalevaluation, 479.
I. INTRODUCTION
There is a growing concern among ecoiiomists about the appro-priateness of traditional macroeconomic models for the analysis ofopen urban unemployment, particularly in less developed countries(LDC's).1 The apparently irrational continuation of rural-urbanmigration in the face of open urban employment has stimulated thedevelopment of new kinds of models that analyze migration and urbanunemployment simultaneously. Known as job-search models, thesehave shifted the emphasis from labor demand to labor supply con-siderations in the explanation of urban unemployment. 2 They gen-erally include a rural and an urban sector, the latter characterized byan institutionally fixed wage above the market-clearing level, andconsider migration and job search unemployment as rational maxi-mizing behavior. People migrate in response to the expected rural-urban wage differential, with the urban unemployment rate actingas a regulating force.
The best known of these job-search models, that of Harris andTodaro (1970), predicted rates of unemployment that substantiallyexceeded the urban unemployment rates actually observed in mostLDC's.3 These significant gaps between predicted and observed ratesof unemployment stimulated the development of more complicatedsearch models with improved predictive performance. Most notably,these more recent models depart from the initial simplistic view thatin urban areas open unemployment is the only alternative to wageemployment in the modern sector; they do this by incorporating two
* We wish to thank Bela Balassa and Robert Dorfman for comments on an earlierversion. The views presented here are the authors' and do not represent those of theirrespective instittitions.
"The Opportunity Cost ef Labor and the Returns toEducation under Unemployment and Labor Nla rket
Segmentation," The QuarterlyJournral of Econotmtics(August 1978). Reprinted by perinission of John Wiley
& Sons, Inc.
470 Q(TAR7TRILYPif'R \'1I.1, OF E('\()MI(CS
urban sectors: A modern-protected-formal or organized sector wherewages are instil utionallv fixed above the market-clearing level char-acterized by restrictions to entry, and a traditional-unprotected-informal-unorganized or murky sector with free entry anu ;--.erewages perform a market-clearingv role.4
In introducing an iniformal-unprotectedi sector, these mnodels'basic premise is that the same forces explaining the equilibriumi al-location of workers between the rural and urban sector can also ex-plain their choices between employiment in the unprotected sectorand open unemployment while searching for a better job in the pro-tected sector.5 The notion that Unemploy ment in the presence of anunprotected free-entry sector is a resullt of a rational process of jobsearch provides a point of departure for analyzing several qUest ionsconcerning the urban economy.
The purpose of this paper is to use this framework to addre, s twoquestions: (a) What is the social marginal product oef types of laborthat differ in the sense that they represent different inputs in theproduction function; (b) what is the relationship between the socialmarginal product of labor and the ob)served market wage in the un-protected sector? In other words, what adjtistnietibs mujst be madeto the observed wage in the uiniprotecte(d sector-- which does resp)ondlto market forces-- in order to (lerive the true marginal montributionof labor?
The possibility of estimating the opportunity cost of lal)or fromobservable market data is operationally useful in project evaluation.Furthermore, by defining labor categories according to levels ofschooling-in this case "educated" and "Uneducated" lab)or-- -wve canevaluate the marginal contributionl of e(lucat ion and it.s relation tothe observed wage structure in the unprotecte(l sector. This becomescrucial in assessing the contribution of educational investnments fromobserved market data.
Assume that we are evaluating an educational project that"4produces" laborers of a given level of education in an urban economywhere (i) there is unemployment of that type of labor, and (ii) t.hereexists a spectrum of wages for aiiy given level of educat ion. Which ofthese wages, or combination of them, shoul( l)e used to omun)Lte thesocial marginal contribution of a worker with a given level of educa-tion? Before 1)resenting an explicit treatment of this (lue';t ionl wereview some intuitive suggestions that have been advanced( to mea-sure, under this scenario, the so cial marginal J)ro(l uct. of labor witha given education.(
(a) The marginal conitribution is zero as long as unemnployinent
THE OPPORTUNITY COST OF L.4BOR 471
of that type of labor exists. The implicit underlying hypothesis is thatunemployment is of an involuntary nature, i.e., at the prevailingwages, that kind of labor is in excess supply in every market. Underthese circurn. ,ices, an extra worker either becomes uniemployed,or by finding a vacancy, prevents the employment of another laborer.Since we are concerned with voluntary unemployment in the contextof free entry competitive markets that coexist with restricted ones,this suggestion is not relevant to our purpose.
(b) The marginal contribution should be measured as a weightedaverage of the wage earnings of that type of labor in each market, in-cluding zero for the unemployed, the weights being the fraction of thetotal labor force in each market. Implicitly this view assumes thatextra workers entering the labor force will be allocated among dif-ferent markets the same proportion as the existing labor force. As willbe demonstrated later, the acceptance of a theoretical framework ofjob-search unemnployment does not sustain this hypothesis.
(c) The ma-rginal contribution should be measured by the wagein the free entry market as long as such a market exists. Proponienitsof this view argue that if the worker is lucky and fills a highly paidvacancy in a restricted market, his net contribution is still the freeentry wage, since he has simply replaced a colleague working in thefree entry sector. If instead he enters this latter sector, then the freeentry wage again reflects his contribution. Finally, if he becomesunemployed, he induces a presently unemployed worker to accept afree entry sector job. Therefore, his net contribution is still reflectedby the free entry sector wage.
The notion that a free entry labor market coexists with a re-stricted labor market appears to reflect correctly the situation of theurban economies in most LDC's and has become an acceptableworking hypothesis in dealing with them. However, if there is a freeentry labor market, unemployment must be voluntary. In these cir-cumstances, it seems far from obvious that an extra worker in the laborsupply induces one additional worker to become employed in the freeentry sector, which would imply that the free entry sector wage cor-rectly reflects the social marginal contributioni of labor. This p)ropo-sition assumes a particular imemployment behavior that results ina constant volume of uniemployment in spite of an extra worker in thelabor force.
The purpose of this paper is to model explicitly employmenit-uniemployment behavior when there is a free entry market and toderive the true inarginal conltribution to the econiomy of additionallabor classified by education. Section IT an-alyzes the nature of labor
472 QUlARTERLY Jol RAN.. OF EC()'\O I('C8
market segmentation and the existen ce of voluntary unemploymellet.Section III spells out an emplovment-unlempl)lo.yrment strategy char-acterizing the behavior of workers outside the l)rotected sector, andSection IV integrates that behavior into a production functionframework for the assessment of the true marginal cont ribit tioni ofdifferent types of labor. Finally, Section V atteml)ts an emn)iricalevaluation of these social contributions and suLIMnarizrs the mainconclusions.
For the purpose of simplicity we assume that for each type oflabor classified by schooling ("educated" and "uneducated" wvorkers)there exist two urban labor markets with (lifferent wages prevailingin each. Wages clear the market. in the iinprotec tei free entry sector,7and the protected market is characterized by entry restrictions anda higher noncompetitive avage rate.
There is ample evidence to support this assumption. Of this widevariety of explanations for these sectoral wage differentials, the mllostcommon one is of an institutional natLre: labor unions, minimum vagrelegislation, and political pressure on governments by organized laborkeep protected .,ector wages above the market-clearing level. Otherexplanations relate to (a) the technological gap betwN-een the twosectors: firms utilizing technologies for which the on-the-job training
, are high may offer wages above the equilibriumn level as . meansof reducing labor turnover; and (b) the efficiency wage hypothesis:higher wages lead to higher productivity, aiid therefore the wage ratethat minimizes total labor costs in the protected sector may not cor-respond to the one pre-vailing in the unprotected sector.8
The literatule on wage )olicies in LDC's contains several at-tempts to separate the impact. of governmnent intervention, unionstretgtlh, and market forces on the wvage level of the p)rotected seclor.-)Regardless of the particular reasoni or comlbination of reasons for thesewage (lifferentials in an urban economy, the protectedl sector can atleast be icienit-itie(i. As Harl)erger points out:
Protected jobs e. i n ria I ilv be iidliilif l)i'(,iil'd sti flilV J)tiPOpolC %%anl them. ('ompanispav-iiig wgio.g higher than marlke levels for ;viik ilkaht skills andl working conditioinsten(1 to lave very low labor Ioirtuwirn,merid bm"" lists of .ipphtam, wlitiniug fi)r .n tivningto arise (Harberger, 1971, 1) !5GIA).
As noted above, the existence ol'free entry labor markets impliesthat observed uinemployinent must be voluntary. 'IThe und(Ierlying,
THE OPPORTUNITY COST OF LABOR 473
hypothesis is that a worker inicreases his ..,;bability of obtaining ajob in the protected sector by being unemr,ioyed and investing insearch. The increase in this probability depends on the extent to whichemployment in the unprotected sector constrains the search for aprotected job.'0 The strength of this constraint depends on institli-tional factors, particularly on the availability of job information, hirilln,practices, and the geographical distance between the two sectors.
The mere fact of the existence of a significant level of open un-employment constitutes indirect evidence that the perceived prob-ability of finding a protected job is higher for the unemnployed. Kritzand Ramos (1976) provide more indirect evidence to support thiscontention. The three employment surveys utndertakeni byP.R.E.A.L.C. reported in their study showed very high rates of openurban unemployment: 20 percent in Managua, 15 percent in SantoDomingo, and 12 percent in Asunci6n. MWoreover, 35 percent of theunemployed in Managua, 51 percent of the unemployed in SantoDomingo, and 60 percent of the uinemployed in Asunci6n were not laidoff but left their previous jobs voluntarily to search for beetter jobs.In synthesis, unemployment reflects a rational process of job senrcli,where the costs are the present forgone earnings in the free entrysector, and the benefits are the present value of a higher probabilityof finding a protected job.
III. THE EMPLOYMENT-UNEMPLOYMENT Di,x: ISION-MAKINGP110loc.EX
Let us suppose that in each period all workers outside the pro-tected sector compare two alternative employment plans:
(a) Plan One: To accept immediate employment in the freeentry sector and remain there, unless one succeeds in the future ingetting a protected sector job. The perceived probability of getti ngsuch a job in any particular future period-when the worker has spentthe prior period working in the free entry sector- is equal to II.
(b) Plan Two: To devote the present period to inlcreasing theperceived probability of obtaining a protected sector job duriig thenext period. We denote the increased probability by i311, where ,lrepresents the probability premium associated with search activities(j > 1). As seen from today, the worker plans to en ter the free entrysector next period if he is unable to obtaini emnployment in the pro-tected sector.
In equilibrium the expected present value of the two employmentplans must be equal.1' Assuming that II is perceived as co(nstant over
474 QUAARTERLY J(OU'RNAL OF EC )NOMCS
time, we may write this equality as
(1) WT - (VT - WT)(1 - oII)(1 -I) T-1T=l (I + r)T
X V T - (VT- W7T)(10- 7T= 1 (1 + r)7 '
where Wo is the present wage in ti-et free entry sector, WT and WT arethe expected wages in the protected and free entry sectors for anyfuture period T, and r is the relevant discounLt rate."2 The left-handside of this expression represents the expected present value of thesecond plan, while the right-hand side represents the value of the firstplan. Rearranging terms, we may write the equilibrium conditionas
(2) (i)-1)MI I (-II 7( WE, -WT)(1 -TI) I'= 1+ri
If 6 = (W( - WO)/W() is define(d as the present percentage wagedifferential between the two sectors and g as the expected growth rateof real wages, the employ)menit-unemply-))nmetnt equili brilmn conditioncan be written as follows:
(3) 61( - 1)(1 + g)U1 + r) - (1-II)(1 +g)
(4) II(1 + g) a = (r - g),
where a = -1) 6 - 1. It is important to notice again that, exceptfor 5, all of the parameters entering expression (4) represenit expectedmagnitudes; they need not represent actual or effectiveparameters.
Two asstumptions have been made in arriving at expression (4).First, that the expected probabilities entering- the plan remain con-stant and equal to today's expected probal)ility or finding a p)rotectedjob next periodl. Second, workers considering the plans assutnie thatthe protected and unprotected wage will grow at the same rate. BothassuLmptions are not indepenident; rational behavior imnplies that thesecond assumption is require(d for the first one to holtl.
In cont.rast to the original Harris-Todaro specification, in whichall protecte(d jobs turn over within each period, we are making themore realistic asstumption that when an individual gets a protectedjob he holds it. Therefore, the prohability of olbtaining a protectedsector job is not the ratio of all jobs to the entire labor force, but. rather,
THE OPPORTUNITY CO8T OF LABHOR 475
the ratio of new jobs to the labor force outside the protectedsector.
If the perceived probability of entering the protected sector isequal to the real one,13 then for each type of labor, II can be def iniedas
(5)V V VS L() LL+(1-p)L
(V/L1,)(L,,/L) X.k
(-1)+ 1 ( )+1'
whereV = the expected vacancies to be open next period in the
protected sector arising from labor rotation, retire-ments, and net employment growth in the sector
S = the number of equivalent searchers who are perceivedas competitors by the worker in question
L = today's employed workers in the unprotected sectorU = the number of (volunltary) meminployed workers in-
vesting fully in searchL the total number of workers outside the protected sector
(L = L + U1)Lp = the employmenit in the protected sector
X = the rate of openings in that employment (X VILI)A = the unemploynment rate (A = U/L)
= the employment in the protected sector as a fraction ofthe labor force outside that sector (q = Lp1/L).
An equilibrium level of voluntary unemnployiment for each typeof labor requires that the expected present value of the two employ-ment plans be equal. Expressions (4) and (5) are therefore specific foreach labor category. Substituting (5) into (4), we get
Implicit in ( ) is the assumption that the free entry sector wage foreducated workers is higher than the wage for uneducated workers inthe protected sector; i.e., there is no incentive for educated wvorkersto coompet.e with noneducated workers Cor their protected jobs. Thlisassumption, although it can change the quantitative magniitudes ofour conclusions, does not chang,e the basic nature of the issues ad-dressed.14
476 QU.1ARTERLY JOURNAL OF ECONUMICS
IV. DERIVING THE CONTRIBUTION OF LABOR TO OUTPUT
Assume that the aggregate production function in the free entrysector can be written as(7) X = F[L1 , L2, K],
where now L1 and L2 represent two types of labor, uneducated andeducated workers, and K represents an index of nonlabor inputs. Thelevel of employment in this sector is determined by the condition thatthe marginal product of each type of labor equals its wage rate:(8) Fi (Li,L 2,K)= Wi fori =1,2.
We now evaluate the total contribution to output of an additionalworker of each educational level who enters the labor force. Thiscontril)ution will differ from the observed present wages in the un-protected sector (WI, W2) and will depend on the additional em-ployment induced by the extra worker himself. This induced em-ployment effect is the resuilt of two sets of forces: First, the extraworker changes the probability that each of his colleagues finds aprotected job, affecting therefore their expected gains from beingunemployed and thus encouraging them to take employment in thefree entry sector. Second, to the extent that there are diminishingreturns and factor complementarity or substitutability, the extraworker changes the equilibrium wage of both types of labor. Thischange in the equilibrium wage has a further employment effectthrough its impact on the r elative profital)ility of unemployment. Theabove effects can be summarized as
(9) -= Wi - + w ;L for ,jdLi dLi dL i =2,J = 1,
where dLi/dLi and dlJ,/dL, represent the "own employment effect"and "cross employment effects" induced by an extra worker respec-tively. Given that the unprotected wages for both types of labor areobservable market data, the evaluation of their contribLu6tions tooutput re(uiiires knowledge of the employmenit terms. Since wages inthe protected sector are exogenously determnined, additional wvorkersin the labor force will affect neither employment nor output in th,atsector. The entire contribution of additional labor can be miieasturedby the changes in the free entry sector output..
The enmployment t.erms (lefine(i above can be obtained by dif-ferentiating expression (6):
(10 dLi - ( 1i ) ( - 1 ) (i 1A ,
THE OPPORTUNITY COST OF LABOR 477
where' 5
(iLi) ( > o,
ai > 0,
and
AFi Aj gjj
(1)dL' (d-1 (> )Li
77ii represents the own price elasticity of demand for labor i in theunprotected sector (m77i < 0). mej is the inverse of the (cross' elasticityof the marginal product of labor i with respect to the employment oflabor j. For most situations 77ij > 0, the, marginal product of labor iincreases when the employment of labor j increases. Expression (10)is positive if A > 0, this condition being fulfilled by any Concave pro-duction function.' 6 Expression (11) will be positive as long as 77ij ispositive.
The own employment effect described in equation (10) consistsof three multiplicative effects summarized by its three multiplicativeterms: a "probability effect," an "own wage effect," and a "cross wageeffect." The probability effect is positive due to the fact that an ad-ditional worker in the labor supply increases the number of searchersand therefore lowers the probability of getting a protected job. Thisinduces additional libor to accept enmployment in the unprotectedsector.17 The "own wage effect" is smaller than one if m7ii is less thaninfinite. In that case additional employment in the unprotected sectordrives down the wage rate and increases the incentive to remain un-employed, partly offsetting the positive impact of the probabilityeffect. The "cross wage effect" is positive and larger than one andreinforces the positive contributioll of the probability effect. Thispositive 'cross wage effect" occurs irrespective of both labors beingsubstitutes or complements, i.e., irrespective of the sign of 7jjj.
The employment effects described above can also be showngraphically by deriving a supply schedule for each type of labor. Frornthe behavioral (equilibrium) condition (6) we can solve for L':
(12) L- = A /(1 +Lg)V, -L - V, _ ( __
--1) (r -t g -1 (r- gi) 1,2i=1, 2,
478 QU'ARTERLY JOUt 'RNAL OF E(CON(lIC('S
0 //
W; /L / L)
N / i i1iA)
// =~i NEi (K;LI +ALI )
/ ~ ~Li,(,~
Total Effeci
FIGu TRE I
where L must be interpreted as the number of laborers of type iwilling to accept employment in the unprotected sector at a wage Wi.The demand for each labor becomes
(13) L4 = L4 (K, Lj, Wi) for i j21 ti 2,j =1.
The horizontal shift of the stuppfly curve in the face of an addi-tional worker is equal to the pure prolb,ability effect. This effect isequal to (3i/li - 1) .Xi, the change in emiployment, that would havetaken place had the wage rate remnained constant. If the demand forlabor i is not perfectly elastic, the wage rate Wi will tend to decline,having a negative effect on emp)loynment. This negative effect is the"own wage effect," represented by the second arrow. So long as thedemand for labor is not completely inelastic, the net result of thosetwo effects must be to increase employment.
Trhe cross wage effect is rep)resented by the third arrow. Noticethat Figure I has been drawn so that the sum of both wvage effects isnegative; i.e., the (negative) own wage effect is stroinger thani the(positive) cross wage effect. This doe.i not necessarily have to be so:if the cross wage effect. were stroniger than the own wage effect, thetotal wage effect Would have beeni p)ositive, reiniforcing the p)ositiveprobability effect. Therefore, we can contlucde that if the "cross wageeffect" is eqtual to or larger than the "own wage effect" (resultillgr ina zero or positive "total wage effect"), an extra vworker in the labor
THE OPPOTl'INITY COST OF LABOR 479
force will induce an increase in emplkyment, for that type of labor,of more than one job.
We want now to explore the sign and magnitude of the marginalcontributions to output of both types of labor as well as their rela-tionships to the observed wage in the unprotected :;ector. By substi-tuting (10) and (11) into (9), we can express these marginial contri-butions as follows:
(1)dX A___Lw U A~1(14) = )[(-) i) W
/f [contribution own + contribution cross1-1 A employmenteffect employment effect.
The contribution of the own employment effect is always positive. Thecontribution of the cross employment effect is positive if both typesof labor are technically complements and negativ'e if they are tech-nically substituLtes. In the former case the total contribution is un-ambiguously positive. In the latter case a sufficient condition for apositive total contribution, which is fulfilled by any kind of C.E.S.production function, is
(15) F _< FiLWi Wj
A priori we cannot speculate whether the true contribution to outputof an additional worker will be larger or smaller than its observed wagein the unprotected sector. However, we can predict that the moreelastic the demand for labor, the more complementary the technicalrelation between the two types of labor; tl.e lower the rate of jobopenings in the protected sector, the lower the rate of growth of wages;the larger the rate of discount, the smaller the relative size of theprotected sector; and the smaller the premiuim to search, the morelikely that the marginal contribution will exceed the observed wagein the unprotected sector.
V. EriPIRICAI, EVALUJATION
A. A Simplified Case
For the purpose of simplicity (and because of lack of in format ion)we shall proceed with the assumption that F12 = F21 = 0 (or equiva-lently q12 = Xq21 = ), This assumption can be consistent with two
480 QUrARTERLY JOUtRNAL OF ECONOMICS
descriptions of the unprotected sector: One where, although L 1 andL 2 enter te same production function, they are technologically inde-pendent; and another where the unprotected sector cc.nsists of twosubsectors, one employing capital and educated labor, and the othercapital and uneducated labor.
The above assumption allows the contribution of each kind oflabor to be written as
(16) dX (1)[dLj G 1) - A l4q,oI i
The marginal contribution can be smaller or larger than thecorresponding observed unprotected wage. The probability effect(oiloi - 1), which is greater than one, is multiplied by a correctivefactor (the own wage effect), which is smaller than one.18
Before proceeding to the empirical evaluation of this last ex-pression, we must assure the consistency of the parameters to be used.In order to assure that the equilibrium condition (6) holds, theunobservable parameter 3i must be endogenously determined; i.e.,the implicit search premium must be consistent with observable dataif the world behaves as the model does. Solving for Oi from (6), weobain
This value of f3z automatically insures the consistency of themodel. An alternative possibility would be to assign values to ki ar-bitrarily so as to predict the equilibrium levels of unemployment.Since Ai is an unobservable parameter, this paper does not pursue thisline of analysis.
B. The Urban Labor Market in ASUnLiCi(), ParaguayThe standard employment survey in LDC's does not usually
collect the kind of information needed for an empirical evaluation ofthe theoretical framewvork presented above. Forttunately, an unor-thodox study undertaken by PREAI C in Asunci6n, Paraguay, canprovide us with most of the necessary information.19
By defining uneducated workers (LI) as those with six or lessyears of schooling and educated workers (L,2) as those with more thansix years of schooling, we obtain the following parameters values:2 )
THE OPPORTUNITY COST OF LABOR 481
o= 0.34 02 - 1.221= 1.08 62 = 1.00
,4= 0.08 A2 = 0.13.
The rate of growth of output in those activities most identified withthe protected sector in Paraguay ranges from 5.4 to 6.4 percent peryear for the 1962-1972 period. These figures do not represent the rateof growth of employment in these activities, which normally lags be-hind the growth rate of otitput. On the other hand, X, the rate of va-cancy openings in the protected sector, includes not only the netgrowth of employment but also the retirement and rotation rates.Given these considerations, two alternative values for X will be usedfor the empirical evaluation: X = 0.05 and X = 0.07.
The average rate of growth of real wages in Latin America wasapproximately 3 percent between 1954 and 1972. Unfortunately, welack specific information about the real rate of growth of urban wagesin Paraguay. However, we do know that the rate of growth of urbanreal per capita income in Paraguay between 1962 and 1971 was 2.3percent. This figure does not exactly represent the trend in real wages,since it is affected by an increasing rural-urban migration that resultsin an increasing volume of open unemployment. On the other hand,due to changing relative endowments of factors, wages for uneducatedlabor are expected to grow faster than those for educated labor. Weshall use here 3 percent and 2.5 percent as the growth rate in wagesfor noneducated and educated labor, respectively. (gl = 0.030, g2 =0.025). The results will prove to be extremely insensitive to changesin these values.
Finally, we need values for r, the discount rate, and 77, the ownprice elasticity of demand for each type of labor. Following thepractice in most of the empirical studies in LDC's, we shall use a 10percent rate of discount.21 In the absence of information about priceelasticities, we presenit a series of estimates based on a wide range ofreasonable demand elasticity values. Although the quantitative resultsvary with different elasticity values, the qualitative results remainunchanged.
Table I presents the implicit values of : consistent with the"observable" parameters being used. They show the proportionalincrease in the probability of finding a protected job, through beingunemployed and investing fully in search, that would be required togene-ate a rate of voluntary unemployment equal to the observed rate.The values of 1 are larger, the smaller the value of X. This is necessarily
482 QUARTERLY JOURNAL OF ECOONWICS
TABLE IVALUES OF /3 CONSISTENT WITH THE "OBSERVED" PARAMETERS
the case, since : has to induce a given rate of voluntary unemploymentdespite a lower rate of openings in the protected sector.
Table II shows the estimates implied by this theory of the mar-ginal contributions of both types of labor in terms of their respectivefree entry sector wages, as well as the relative marginal contributionof educated to noneducated labor in terms of their observed relativewages in the free entry sector.
Conclusions
The followving conclusions can be derived from Table II: First,the ratio between the true marginal product of labor and the free entrysector wage is highly sensitive to the demand elasticity for labor andrather insensitive to the value of the other parameters, g and X.22Hence, having selected a value for this elasticity, we note that thecorrective factor applied to the free entry sector wage in order to derivethe corresponding marginal contribution of labor is relatively inde-pendent of the other parameters. Second, this corrective factor is lessthan unity for demand elasticities smaller than five for educated laborand smaller than four for uneducated labor. Within these ranges, thefree entry sector wage will overestimate the true marginal product oflabor, particularly for educated labor. The correction factor rangesfrom one-half to one and from two-thirds to one for educated anduneducated labor, respectively.
The relative marginal contribution of educated and noneducatedlabor determines the profitability of investment in schooling. TableII presents these relative marginal contributions in terms of the rel-ative wages of both types of labor in the free entry sector. The cor-rection factors are again less than one for demand elasticities for laborbetween one and five. This means that, within that elasticity range,relative wages in the free entry sector overestimate the true benefitsof investment in schooling.
From the above, we conclude that, for a wide range of demandelasticity values, the wage in the free entry sector overestimates thetrue marginal contribution of labor. Furthermore, relative wages byschooling in that sector also overestimate the true contribution of
TABLE IIMARGINAL CONTRIBUTIONS (ABSOLUTE AND RELATIVE) OF EACH LABOR IN TERMS OF THE FREE ENTRY WAGE
Labor demand Protected sector Absolute contributions Relative contribution ofelasticity rate of openings Uneducated Educated educated to uneducated labor SG
-1.0 X = 0.05 0.73 Wi 0.48 W2 0.66 (W2/W 1 )X = 0.07 0.68 Wi 0.40 W2 0.59 (W2 /WI)
-2.0 X 0.05 0.90 Wi 0.71 0.72 0.79.WW/W)X = 0.07 0.87 Wi 0.63 W 2 0.72 (W2JW 1 )
-3.0 A = 0.05 0.97 Wi 0.85 W2 0.88 (W2JWI) °CcA = 0.07 0.96 W, 0.78 W2 0.81 (W2/W1 )
-X = 0.05 1.15 Wi 1.38 W2 1.20 (W2 1W1) 0A = 0.07 1.22 Wi 1.47 W2 1.20 (W 2 /W 1 )
oo
484 QUARTERLY JOURNAL OF ECONOMICS
investments in education. The more inelastic the demand for labor,the larger the discrepancies involved.
In this model the measure for the total contribution of labor tooutput (dX/dL) also represents the social price of labor, i.e., thewelfare cost of removing one worker from the labor force.23 The socialprice of labor is positive only to the extent that removal induces-viathe employment effect (dL/dL)-a decline in the unprotected sectoremployment and output. Declines in unemployment (dU/dL) inducedby the removal of that worker do not generate a welfare loss andtherefore have no bearing on the social price of labor. If the own em-ployment effect is 0.73 (value for uneducated workers in the first rowof Table II), it means that the removal of 10 workers from the laborforce will come at the expense of 2.7 unemployed workers and 7.3workers previously employed in the unprotected sector.24 The socialprice of those 10 workers is 7.3 times the unprotected sector wages,since the unemployed workers do not have any bearing on the socialcost.
This result is different from the one derived by Harberger (1971)where unemployment is a result of labor-leisure choices and whereunemployed workers have a supply price above the unprotected sectorwage. In Harberger's framework the cost of labor is equal to the un-protected sector wage if the worker is removed from that sector andis equal to the supply price or value of leisure if it comes from the poolof unemployed. The unprotected sector wage becomes therefore thelower bound for the social price of labor, the discrepancy becomingsmaller, the smaller the rate of unemployment.
The basic difference in the two approaches are the factors ex-plaining the preferences for being unemployed. In Harberger's ap-proach leisure is the alternative to an unprotected sector job, i.e., thechoice of unemployment is Pareto optimal. In the approach outlinedhere, unemployment is a result not of labor-leisure choices but ofsearch activities perceived as affecting the individual probability ofobtaining a protected sector job. Given that the number of vacanciesin the protected sector are given, unemployment is simply a mecha-nisin for redistributing a constant aggregate probability acrossworkers; unemployment therefore represents a dead weight or anon-Pareto-optimal situation.
A second differenice is that the nmacroeconomic nature of themodel presented here captures, through the employment effect, theeffect of removing one worker from the labor forces on the employ-ment of other workers. Other workers' choices are affected to theextent that removal affects the previous equilibrium between the
THE OPPO'0RTUTNITY COST OF LABOR 485
employment-unemployment plans. If the removal of the worker di-minishes unemployment in the rest of the labor forces, the worker hasinduced a positive externality; other workers increase output withouta social cost given the absence of leisure beniefits. If the effect is toinduce additional unemployment in the rest of the labor force, theworker has generated a negative externality; the decline in that em-ployment generates a loss in output without gains in leisure.
From that above, it is clear that the factors explaining a worker'srefusal to accept an unprotected sector job-either leisure or searchfor a protected sector job-are crucial in deriving the social price oflabor. If these factors are search activities, the model presented herewill be an appropriate framework to derive these types of esti-mates.
U.N. ECONOMIC COMMISSION FOR LATIN AMERICAWORLI) BANK
NoTES
1. See, for example, Hall (1975), and Harris and Sabot (1976).2. See Todaro (1969), Harris and Todaro (1970), Phelps (1970), rodaro (1971),
and Johnson (1971).3. See Turnham (1974).4. See, for example, Harherger (1971), Zarembkla (1972), Mazumdar (1975), Fields
(1975), Stiglitz (1976), and Harris and Sabot (1976). Other extensions of the initialHarris-Todaro model are the recognition of the heterogeneity of migrants (Mazumdar,Stilitz); a more general approach to job search (Stiglitz, Harris and Sabot); the laborturnover approach (Stiglitz, Fields); recognition of employers' preferential treatmentof the better educatied workers (StigliLz, Harris and Sahot); consideration of the fi-nancing of the investment in job search and the direct costs of migration (Harris andSabot); the prohahility of a rural resident obtaining an urban joh (Fields); and finally.consideration of the appropriateness of the use of present versus current values in themigrant decision-making process (Fields).
5. Fields (1975) analyzes these two decisions in a simultaneous framework.6. It is important to notice that this issue is independent and different from that
coincerning the "value added" of education. It has been argued that even if one acceptsthe notion that wage d ifferent ialk by schoo.ling retlect productivity differences, t hesedifferentials do not necessarily indicate a positive value added of education from thepoint of view oft he produict ion functio,n. The screening hypothesis is perhaps the bestknown of these argoments: "It suLggests that inter-educational earnings differentials,even when standardized for differeLces due to non-educational factors, reflect no dlirectpr(dltictivity enhancing effects of educationi but only it.s effect. as a device for signalingpre existing abiliLy diflerences" (Layard and Psacharopoulos, 1974).
'The "valuLe added" arguments, those emhodying the hypothesis that wage dif-ferentials overstate the true c,ntribution of education can, in principle, be tested. Theyare empirical rather than theoretical consideration,s.They l)asicallv represent a "nmissingvariable hias" argument: if one had data on all backgroaund variables correlaled witheducation an(l having an independent effect oIn wages (including preschool levels ofabilihv, to take care of the screening hypothesis), one could estimate the true valueadded of schooling.
7. This market includes self-employed, as well as hired labor as long as the freeentry asstumption holds.
8. See, lor example, Stiglitz (1974, 1976), Herrick (1974), Mazumdar (19751, Harris
486 Qi 'AI 7'EI?LY Jf 'RUNAL, OF ECONV(H)fICS
and Sabot (1976), Webb (1977), Leibenstein (1957), Turnham (1974), and Harberger(1971).
9. See, for example, Berg (1970), Turner and Jackson (1970), Frank (1968), Ramos(1970), Isbister (1971), Kilby (1967), Knight (1967), and Sabot (1975).
10. I'his is a basic assumption of search models developed for LDtJsb. See Maz-umdar (1975), Fields (1975), Sabot (1975), and Harris and Sabot (1976). It is also acrucial assumption of a number of more general job search models based on St igler'sclassic paper on information on the labor market (Stigler 1962). Most of these mnoregeneral search models view the unemployed as sampling from a inrolability distributionof prospective wages. Under certain conditions, the optimal search rule is to establisha reservation wage. However, these job opportunities would not exist had the workernot been employed and investing in search. See, for example, Mortensen (1970), Gronau(1971), Ilolt (1970), McCall (1970), Phelps (1970), Kohn and Shavell (1974), Lucas andPrescott (1974), and Eaton and Naher (1975).
11. This equilibrium conditioni assumes that workers are risk neutral. Never-theless, the introduction of n)nneutrality- - which implies an analysis in terms of utilityrathler than of income-does not change the essential aspects highlighted in thepaper.
12. Equation (1) assumes that workers live forever; this simplifying assumptionallows the use of formulas of perpetuity at a later stage. This approximation is notunreasonable given the high discount rate and the rather young labor force charac-terizing LDC cotntries.
13. Imperfe~tions of information mean that individuals might misperceive theactual probability of obtaining a job in the protected sector. For the optimistic ones,the subjective probability will overestimate the true one, the opposite being true forthe pessimistic ones. The notion that subjective probabilities are equal to the effectiveones implicitly assurhes some kind of BaYesian learning process that tends to guaranteethis equalitv.
14. This assumpti;'n was made for the potirpose ofrsimnlicity and because it cor-responds to the situation of the urban economy of Asuncoic, Paraguay, which will beused for the empirical evaluation of the model. See "SitIoa-ion y Pc r..npectivas delEmpleo en Paraguay," P'REA LC, I.L.O. 7 I5). 1.
15. F-nom the equiilibrium condition (4) we can see that a, will be )l1iite1 as longas (' - jj ) iS. po )sit ive. This c ,nd it ioin characterizes most empirical situations; at the sameiime it is also a sufficient condition of convergence for the series being used to transform(2) into (3).
16. A sufficient condition for A > 0 is the following relationship among the secondderivativec of the production function:
F <F,, F,,.
This conidition is fulfilled by any concave produiction function.17. This probability effect is also larger than one, i.e., one additional worker in
the labor force induces an increase in employment in that type of skill by an amountlarger than one.
18. From expression (14) we (b,serve that if the two types of labor are technolog-ically substitutes (Fij < 0), then our results wvill overestimate the true contribution oflabor. The opposite will haplwen if, as in most of uhe cases, the factors are technologicallycomplements (Fj > 0). The not ion of capital and an aggregate labor input index en-tering the pr(D(liction function,
'= F[K, LtL,, I1)],
tends to supl)port the first hl)othesis. On the other hand, the slecitication of a pro-duction function considering capital and educated labor as one in)ut entering theproduction function and uneducated labor as the other,
Y= t[(Q(K, L,), L,],
tends to support the second hlvothesis.19. See "Situllcilon y Perspectivas de Empleo en Paraguay," PREALC (ni?m-
ployment Program for Latin Amierica, l.I,.O.), 1975. The detailed sources of data arerel)ortcd in rviera and Selowsky (1976).
20. The rate of voluntariy uiineniplovnment, defined as a fraction of the labor forceoutside the protected sector can be wri ten as
THE Q)JP/(I7'( ?7VNI7' (COST OF L-8I01? 4-87
Hi=(I + 00i (g.-*)
where A' is the standard definition of the rate of unemployment (defined as a fractionof the total labor force) andg* is the rate that ca ble htiributed to seasonal and fric-tional unemployment. The unemployment rate q was 12 percent ror both types oflaiborin Asunci6n. We have ass.umed that half of that rate could be expainel be hyseasnialand frictional unemployment. Therefore,
1.34 (0.12 - 0.06) 0.0= .
= 2.22 ((0.12 - 0.06) = (1.13.
21. We refer here to a 10 percent real rate. All this analysis is carried out in realterms, and we assume that no agents suffer from money illusion.
22. Experiments with alternative *alues fo(rg are reported in Pinera and Selowskv(1976).
23. This obviously assumes no difference(s in nonpecuniary benefits betweenworking in the tunprotecte(i sector ain(d searching for a protectCe(l djl. To the extent thatthey do exist, the social price of lahor should include, in add itioni, a term capturing thechange in these .benefit.-_
24. Given the definition L L + UI, we get
dL dL(1') 1 -dL +dL,
Berg, E., 'Wages Policy and Employment in L.ess D)eveloped Countries," presentedto the Confere.-ice on Prospects for Empltymennt Oppomiunit itis in the Nineteen-seventies, U1niversity of ('ambridge, England. 197t).
Eaton, C., and P. Neher, ' !nemplovroent, Un deremployment, and Optimal 'JobSearch," Journal of Plolitical Economy, LXXXIII (April 1975), 355 -75.
Fields, G., "IRural-LUrban Mligration, I lrrban Unemployment and IInderemniployment,and Joub-Search Activity in LD(C'= Journol of L)Deelo'lpmlent Economics, II (1975),165-87.
Frank, C., "Urban Unemployment and Economic Growth in Africa," Paper No. 120,Yale Economic Growth Center, (1968), pp. 262-65.
Gronau, R.. "Information and Frictional Unemployment," American EconoMih RJvieut'tLXI (June 1971), 290-0301.
Hall, R., "The Rigidity of Wages and the Persistence of UInemployment." Dtepart ni(entof Economics, M.I.T., prepared for the Brookings 1'nnel of Economic Aclivity,1975.
Harberger, A., "On Measuring tie Social Opportunity Cost of Labor," InternationalLabor Revieu', C XX X, No. 6 (1971), 559. 79.
Harris, J., and R. Sabot, "'Ulrban Unemploynent in I.DCs: Toward a More GeneralSearch Model," paper presented t ob the Workshop on Rural-Urban Labor MarketInteractions, IIRD, February 1976.and M. Todaro, "Migration, I Tnenig [lovment and Development: A Two -Sect or
Analysis," American Economic rh'm'ir'w', L.X u 1970), 1264:3.Herrick. B., "Urban Self Employment andc Changing Es.xpecmalions as InfluenCeS (tll
tUrban 1Nligration,'' Ilen rtnt men ofEcomics. UIniversity of ('Clitornin, Ios An-geles, 1974.
Holt, ('., "Job) Search, Phillips Wage Relation and Union I1nfluence: Theory and Evi[dence," in E. Phelps et al., Microeconomic Foundations of Employment anldInflation Theorsy (New York: Nofton, 1970).
ILO, PREAI.(C, "SituLaciony Perspectivas del Empleo en Paraguav," 1 975.Isbister, J., "Urban Wages and Emiploymnent in a D)eveloping Economy: 'I'le ('ase of
Mexico," Economic Development and Cultural Change, XIX (Oct. 1971), 2446.
Johnson, G., "The Structure of Rural UTrban Migration Models," Eastern AfricaEconontic Review. III (Junie 1971, 21 28.
Kilby, P., "InduLstrial Relations and Wage I)etermination: Faiilure ol'the Anglo-Saxon
488 QUIARTEl?l,Y JOURNAL OF E(VONO)AfI(CS
Model," Journal of Developing Areas, I (1967), 489-520.Knight, J., "The Determination of Wages and Salaries in tUganda," Bulletirn of Oxford
Institute of Economics and Statistics, XIX (1967), 250.Kohn, M., and S. Shavell, "The Theory of Search," Journal of Economic Theory, IX
(1974), 93-123.Kritz, E., and J. Ramos, "The Measurement of Ulrban llndereinployrnent: A Report
on 3 Experimental Surveys," International Labor Review, CXIII (1976), 115f-27.
Layard, R., and G. Psacharopoulos, "The Screcning Hypothesis and the Returns toEducation," Journal of Political Economy, LXXXlXI (1741, 985 88.
Leibenstein, H., Economic Backwardness and Economic (Growvth (New York; ,JohnWiley & Sons, 1957).
Lucas, R., and E. Prescott, "Equilibrium Search and Unemployment," Journal ofEconomic Theory, VII (Feb. 1974), 188-209.
Mazumdar, D., "The Theory of Urban Underemployment in Less I)evtelopedCountries," World Bank Staff Working Paper 198, 1975.
McCall, J. "Economics of Information and Job Search," this Journal, LXXXIV (1970),113-26.
Mortensen, D., "Job Search, The Duration of Unemployment, and the Phillips Curve,"American Economic Review, LX (1970), 847-62,
Phelps, E., et al., Microeconomic Foundations of Employment and Inflation T'/wory(New York, W. W. Norton and Co, 1970).
Pihera, S., and M. Selowsky, "Unemployment, Labor Market Seg,mentation, TheOpportunity Cost of Labor and the Social Returns to Education," World BankStaff Working Paper No. 233, 1976.
Ramos, J., Labor and Developm ent in Latin America (New York, Columbia UlniversitvPress, 1970).
Sabot, R., "The Meaning and Measurement of UTrbin Sturplulls ,Labor," Minmeo, Eco-nomics Department, I.B.R.D,, 1975.
Stigler, G., "The Economics of Information," Journal of Plolitical Economy, LXIX(1962), 312-25.
Stiglitz, J., "Alternative Theories of Wage Deterniinat ion and Uneinploymenl in I.I)C's:The Labor Turnover Model," thisJournal, LXXXVIII (May 1974). 94 227."The Structtire of Labor Markets and Shadow l'rices in LDC.," Employmnit
and Rural Development Division, I.B.R.D., 1976,Todaro, M., "A Model of Labor Migration and Urban Unemployment in Less Devel-
oped Countries," American Economic Review, LIX (1969), 138-48.-- , "Education and Rural Urban Migration: Theoretical Constructs and Empirical
Evidence from Kenya," paper prepared for a conference on urban unemploymentin Africa, Institute of Development Studies, UTniversity of Sussex, England,1971.
Turner, H,, and D. Jackson, "On the Determination of the General Wage I.ex'el: A WorldAnalysis; or Unlimited Labor Forever," Economic Journal, LXXX (Dec. 1970),827-49.
Turnham, D., The Employment Problem in Less Developed Countries A Revieu ofEuidence (Paris: O.E.C.D., 1974).
Webb, R., "Wage Policy and Income Distribution in Developing Countries," in C. Frankand R. Webb, eds., Income Distribution and Growth in Less-Developed Countries(Washington, D.C., Brookings Institutimni, 1977), pp. 215-47.
Zarembka, P., Toward a Theory of Economic Development (San Francisco: Holden-Day, 1972).
Nigeria: Options for Long-Term Development by Wouter Tims and others,published by The Johns Hopkins University Press, 1974
The Current Economic Position and Prospects of Perui by Jose Guerra andothers, distributed by The Johns Hopkins University Press, 1973
Senegal: Tradition, Diversification, and Economic Developmont by HeinzBachmann and others, distributed by rhe Johns Hopkins University Press,
1974Turkey: Prospects and Problems of an Expanding Economy by Edmond Asfour
and others, distribuled by The Johns Hopkins University Press, 1975Yugoslavia: Development with Decentralization by Vinod Dubey and others,
published by The Johns Hopkins University Press, 1975
World Bank Staff Occasional Papers
Economic Evaluation of Vocational Tralning Programs by Manuel Zymelman,published by The Johns Hopkins University Press, 1976
A Development Model for the Agricultural Sector of Portugal by Alvin C. Egbertand Hyung M. Kim, published by The Johns Hopkins University Press, i 775
The Future forHardFibers and Competition from Synthetics by Enzo R. Grilli, dis-tributed by The Johns Hopkins University Press, 1975
Public Expenditures on Education and Income Distribution In Colombia by Jean-Pierre Jallade, distributed by The Johns Hopkins University Press, 1974
Tropical Hardwood Trade In the Asia-Pacific Region by Kenji Takeuchi, dis-tributed by The Johns Hopkins University Press, 1974
Methods of Project Analysis: A Review by Deepak Lal, distributed by The JohnsHopkins University Press, 1974
Road User Charges in Central America by Anthony Churchill, distributed by TheJohns Hopkins University Press, 1972
Cost-Bene ItAnalysis in Education: A Case Study of Kenya by Hans H. Thias andMartin Carnoy, distributed by The Johns Hopkins Press, 1972
Other PublIcations
Size Distribution of Income: A Compilation of Data by Shall Jaln, distributed byThe Johns Hopkins University Press, 1975
India: The Energy Sector by P.D. Henderson, published by Oxford UniversityPress, 1975
The Assault on World Poverty: Problems of Rural Development, Educatlon, andHealth, published by The Johns Hopkins University Press, 1975
Redistribution with Growth by Hollis Cheniery, Montek S. Ahluwalla, C. L. G. Bell,John H. Duloy, and Richard Jolly, published by Oxford University Press, 1974
Population Policies and Economic Developrnernt: A World Bank Staff Report byTimothy King and others, pujblish,ed by The Johns Hopkins University Press,
1974Prospects for Partnership: Industrialization and Trade Pollcies In the 1970s
edited by Helen Hughes, published by The Johns Hopkins University Press,1973
