child labor: cause, consequence, and cure, with remarks on

Journal of Economic Literature Vol. XXXVII (September 1999) pp. 1083–1119

Basu: Child Labor: Cause, Consequence, and Cure

Child Labor: Cause, Consequence,and Cure, with Remarks on

International Labor Standards

KAUSHIK BASU1

1. Introduction

ACCORDING TO THE Bureau of Statis-tics of the International Labor Or-

ganization, in 1995 at least 120 million ofthe world’s children between the ages offive and fourteen years did full-time,paid work (ILO 1996; Kebebew Ashagrie1998). Many of them worked under haz-ardous and unhygienic conditions and formore than ten hours a day. This is not anew problem. In different parts of theworld, at different stages of history, thelaboring child has been a part of eco-nomic life. In particular, children haveworked in large numbers in factoriesfrom the time of the industrial revolu-tion in Europe and from the mid-nine-teenth century in America. In contempo-

rary times, the incidence of child labor isvery high in Third World countries, andit has been that way for several decadesnow.

What has increased is the awarenessof and concern for children who work aslaborers. This is caused, in part, by theincreasing globalization of the world,which has brought not only more infor-mation about the condition of labor indifferent nations to academics and ac-tivists the world over, but also goodsproduced by children in faraway landsinto the hands of consumers in high-in-come countries. This has, in turn,brought two very different kinds of peo-ple onto the same platform—individualswho are genuinely concerned about theplight of children in poor countries, andthose who comprise the forces of pro-tectionism in developed countries. Thetwo have rallied together to support avariety of interventions in Third Worldlabor markets, ranging from banningimports into industrialized nations ofproducts “tainted” by child labor inputs,through setting international labor stan-dards to be monitored by internationalorganizations such as the WTO or ILO,to labeling products that involved childlabor so as to give the consumer theoption to boycott them.

1 Cornell University. For useful comments andsuggestions I am grateful to Jens Andvig, KebebewAshagrie, Alaka Basu, Clive Bell, Francois Bour-guignon, George Boyer, Dan Bromley, JeanDreze, Patrick Emerson, Gary Fields, GaranceGenicot, Noemi Giszpenc, Subbiah Kannapan,Ayal Kimhi, Elizabeth King, Luis-Felipe Lopez-Calva, Dani Rodrik, Pham Hoang Van, HenryWan, and three anonymous referees of the Jour-nal. I have also benefited from seminar presenta-tions at the Indian Statistical Institute in NewDelhi, the Stockholm School of Economics, theIndira Gandhi Institute of Development Research,the University of Wisconsin, Notre Dame Univer-sity, the World Bank, and the NEUDC Confer-ence at Yale University.

1083

Any such intervention is likely tohave not just an impact on the well-be-ing of children, but also spillover ef-fects on others. It is imperative, there-fore, that policy in this area be based oncareful analysis and research, and notjust emotion or impulse.

The literature on child labor is enor-mous, but it is scattered across the so-cial sciences and piecemeal, lacking acommon theoretical foundation. Theaim of this paper is to provide an ana-lytical survey of this field, keeping inmind that this is an area where the pri-mary reason for theorizing is ultimatelyto influence policy. The main policy de-bates and options are summarized inSection 3. Section 2 provides the factualbackground of the problem, drawing onlarge-scale data sets and on the substan-tial literature rooted in formal and in-formal micro studies. Thanks to theenormity of the problem in the last cen-tury, there is also a sizeable historicalliterature on the subject.2 I draw onthis, not comprehensively, but keepingin mind contemporary concerns and theanalytical focus of this paper, and reporton this in Section 2.2.

Sections 4 through 7 are on modelsand theories related to child labor. Thetraditional argument for government in-tervention in child labor markets isbased on the standard claim of exter-nalities. Such arguments are recapitu-lated in Section 4, which also summa-rizes the early theoretical ideas of someclassical economists. Conventional mod-els treat the household as a single deci-sion-making unit (Gary Becker 1964);but once we recognize that there maybe divergence of interests within the

household, there is scope for arguingthat children are victimized. There isnow a whole range of bargaining modelsavailable for analyzing intra-householddecision making, some of which explic-itly look into the question of child labor(for example, Carolyn Moehling 1995,and Manash Gupta forthcoming). Thesemodels are summarized in Section 5,along with some suggestions for exten-sions. If there are no externalities andno divergence of interest in the house-hold, can there still be a case for inter-vention? Section 6 sketches a model(Kaushik Basu and Pham Hoang Van1998) showing that, for a class of situ-ations, the answer is yes, because of thelabor market’s propensity to have multi-ple equilibria. The model of Section 7sketches an argument against harassmentand shows how this kind of analysis canbe used in thinking about child-laborlegislation.

Child labor regulation is an importantpart of the current debate on interna-tional labor standards. Section 8 dis-cusses some questions concerning inter-national labor standards in which theproblem of child labor arises explicitly.Section 9 consists of brief concludingremarks.

2. The Empirical Context

To place this analytical survey in con-text and also to keep our focus on real-world policy concerns, it is useful to be-gin by briefly recounting the world’sactual experience with child labor. Thisexercise is broken up into two sub-sections, one describing contemporarymacro aggregates, and another describingthe historical roots of the problem.

2.1 Contemporary

Any estimate of child labor dependson how we define “child” and “labor”and on the quality of statistics available.

2 See, for instance, Michael Anderson (1971),Claudia Goldin (1979), David Vincent (1981),Hugh Cunningham (1990), Clark Nardinelli(1990), Moehling (1995), Sara Horrell and JaneHumphries (1995, 1995a), Cunningham and PierPaolo Viazzo (1996), and Douglas Galbi (1997).

1084 Journal of Economic Literature Vol. XXXVII (September 1999)

The ILO’s Convention No. 138 speci-fies fifteen years as the age abovewhich, in normal circumstances, a per-son may participate in economic activ-ity.3 Following this, most studies treat aperson of age less than fifteen years as a“child.” As a study by the US Depart-ment of Labor notes, while ILO Con-vention 138 “has been ratified by onlyabout one quarter of the ILO member-ship, it has nevertheless been interna-tionally recognized and used as a blue-print for national policy and practicewith respect to child labor” (US Depart-ment of Labor 1993, p. 5). This obser-vation carries even greater weight be-cause the US itself is not a signatory tothe convention.

A child is classified as a “laborer” ifthe child is “economically active” (Asha-grie 1993).4 Governments and interna-tional organizations usually treat a per-son as economically active or “gainfullyemployed” if the person does work on aregular basis for which he or she is re-munerated or that results in output des-tined for the market. The Indian cen-sus, for instance, explicitly uses such aconvention. We know from micro stud-ies that if we instead include those “in-visible” workers who do unpaid workthat finds no market outlet, such aswork within the household, the esti-mates of child labor can shoot up. D.Jayaraj and S. Subramanian’s (1997) re-

cent calculations for the state of Tamil-nadu, India, show that for the 5–14 agegroup in 1983, if they use the restrictivedefinition of the kind that the ILO uses,13 percent of all children were laborers.Using the more liberal definition, thefigure jumps to 33 percent.5

Even after the definitions are sortedout (to the extent that they can be) offi-cial data on child labor tends to be defi-cient because of the likelihood of un-derreporting. In most countries thereare laws that place restrictions on childlabor, ranging from an outright ban (asin most industrialized countries) toother kinds of limitations such as anoutright ban on child labor for verysmall children, and for all children inhazardous industries (as, for instance, inBangladesh, India and Pakistan). Thusit is natural for guardians and employ-ers to hide the information of “illegal”work by children.

Keeping all these caveats in mind,and cobbling information from varioussources, Ashagrie (1993; see also Chris-tiaan Grootaert and Ravi Kanbur 1995)was the first to put together an interna-tional data set on child labor. He foundthat in 1990 there were nearly 79 mil-lion children who were economically ac-tive. Most of them (57 million) were inAsia. However, over the previous tenyears, he found that the absolute num-ber of children working was declining inAsia but rising in the Americas and Af-rica, sharply in the latter. This data issummarized in Table 1.

It is worth pointing out that while

3 While this is so under “normal circumstances,”Convention 138 does specify some special cases.Thus for “light work” the age limit is thirteenyears, and for “hazardous work” it is eighteenyears (ILO 1996). See UNICEF (1994) for a dis-cussion of Convention 138 and for alternative con-ceptions of what constitutes child labor (see alsoKarl Eric Knutsson 1997).

4 The ILO often distinguishes between “childwork” and “child labor,” the latter being used to de-scribe the more pejorative part of “child work,”whereas “child work” in itself could include doinglight household chores and can actually have somelearning value (ILO 1995, p. 1). I shall, however,here use “work” and “labor” interchangeably,while referring to what the ILO calls “child labor.”

5 This point is reinforced at the all-India levelby Myron Weiner (1996, p. 3007) who argues that“most of the 90 million children not in schoolare working children.” See also Sophie Labenne(1995) for alternative definitions of child labor andhow different variables affect them in India. Someeconomists adopt the convention of distinguishingbetween “work” and “home care,” classifying chil-dren who do not attend school and are not for-mally employed as “home care laborers” (KimberlyCartwright and Harry Patrinos 1998).

Basu: Child Labor: Cause, Consequence, and Cure 1085

Ashagrie’s (1993) estimates, as summa-rized in Table 1, have the virtue of com-prehensiveness, they generally givelower values for the incidence of childlabor than other existing estimates. In-deed, more recent works by Ashagrie(1998) and the ILO (1996), which werebased on detailed experimental surveysin Ghana, India, Indonesia and Senegal,have resulted in upward revisions ofthese figures. In addition, it must bekept in mind that all figures quoted inthis section relate to full-time work. Re-cent ILO estimates suggest that if“child labor” is taken to include thosewho do part-time work, the numbersmore than double. Thus the 1995 figurefor child labor, quoted at the start ofthis essay, would rise to 250 million.

The Asian figures are very large, butit has to be kept in mind that Asia’spopulation is very large. So to get a bet-ter idea of the magnitude of the prob-lem, we may wish to look at the “partici-pation rate,” that is, the percentage ofchildren of the relevant age group whowork as laborers. In terms of participa-tion rates, for the 10–14 years category,the problem in 1990 was most serious inAfrica, with a figure of 27.87 percent.The participation rate for Asia was

15.19 percent. These statistics are avail-able from the ILO (1996a), which hasnow compiled inter-country data onchild labor from 1950 to 1995, alongwith projections up to 2010. A summaryof this information is provided in Table2 and Figure 1. The first five rows showthe distribution of child labor across themain continental regions of the world.In addition, a sample of five nations,from among those that had a partici-pation rate of over 20 percent in 1950,is represented in the table to givethe reader a glimpse of how varied theexperience of different nations hasbeen.

As this graph and table show, theproblem is enormous, but the trend,fortunately, is in the right direction.For some countries, such as China, In-dia, and especially Italy, the decline inthe participation rate of children hasbeen quite rapid. For most LatinAmerican nations, such as Brazil, thedecline is notable but less marked. Theproblem has been extremely persistentin large parts of sub-Saharan Africa, asillustrated by Ethiopia, but even herethe trend is downwards. For China, in-terestingly, the decline is most rapidbetween 1980 and 1990, which happens

TABLE 1CHILD LABOR: AGGREGATE AND DISTRIBUTION

Number of Children (below 15 years)Working (in thousands)

1980 1985 1990

World 87,867 80,611 78,516 Africa 14,950 14,536 16,763 Americas 4,122 4,536 4,723 Asia 68,324 61,210 56,784 East Asia 39,725 33,463 22,448 Southeast Asia 6,518 6,079 5,587 South Asia 20,192 19,834 27,639

Source: Ashagrie (1993).


to coincide with the period of veryrapid growth in incomes.6

The overall growth of an economy isby no means the only factor, nor forthat matter the most important factor,in the mitigation of child labor.Changes in technology, improvement inthe conditions of the adult labor mar-ket, and the availability of decentschooling can all lead to children beingvoluntarily withdrawn from the laborforce. Victor Levy (1985), in a study ofrural Egypt, found that mechanizationdid contribute to a diminution in childlabor and also in fertility. The cultiva-tion of cotton, especially weeding andpicking, has conventionally been child-labor intensive in Egypt. So changes incropping pattern, away from cotton,played a role in curbing child labor.

We shall return to the subject of howto curb child labor later. Let us, beforethat, take a look at the historical back-drop of the contemporary situation, justdescribed.

2.2 Historical

Here is an excerpt from an interviewwith a child laborer.

Q. “What were your hours of labor, doyou recollect, in that mill?”

A. “In the summer season we were veryscarce of water.”

Q. “But when you had sufficient water,how long did you work?”

A. “We began at 4 o’clock in the morn-ing and worked till 10 or 11 at night;as long as we could stand upon ourfeet.”

Q. “You hardly could keep up for thatlength of time?”

A. “No, we often fell asleep.”

This could have been a contemporaryinterview, but it is not; it took place inJune 1832. The interview was con-ducted by a British Parliamentary com-mittee, investigating the conditions ofchild labor in the United Kingdom. Inthis case the interviewee was Peter

TABLE 2PARTICIPATION RATES FOR CHILDREN, 10–14 YEARS

1950 1960 1970 1980 1990 1995 2000 2010

World 27.57 24.81 22.30 19.91 14.65 13.02 11.32 8.44 Africa 38.42 35.88 33.05 30.97 27.87 26.23 24.92 22.52 Latin America & Caribbean 19.36 16.53 14.60 12.64 11.23 9.77 8.21 5.47 Asia 36.06 32.26 28.35 23.42 15.19 12.77 10.18 5.60 Europe 6.49 3.52 1.62 0.42 0.10 0.06 0.04 0.02 Ethiopia 52.95 50.75 48.51 46.32 43.47 42.30 41.10 38.79 Brazil 23.53 22.19 20.33 19.02 17.78 16.09 14.39 10.94 China 47.85 43.17 39.03 30.48 15.24 11.55 7.86 0.00 India 35.43 30.07 25.46 21.44 16.68 14.37 12.07 7.46 Italy 29.11 10.91 4.12 1.55 0.43 0.38 0.33 0.27

Source: ILO (1996a).

6 One has to be careful not to read too muchinto these macro statistics. Most of the data areconstructed by extrapolating from a few small-scale detailed studies. Even in the US, where aconsiderable amount of money and effort goes intocollecting labor data, the amount of child labor isprone to be underreported (Douglas Kruse andDouglas Mahoney 1998). The problem must bethat much more acute for developing countries,especially ones where state control makes it diffi-cult for independent researchers to collect andverify data.


Smart, who had worked as a laborerfrom the age of five years. Peter Smartwas not an exception. The Select Com-mittee Report of 1831–32 from which Ihave taken the above interrogation (seeBritish Parliamentary Papers 1968,p. 338), is an enormous document, re-porting on a series of interviews of bothchild laborers and factory owners andmanagers. The story that emerges haslittle variation. The children workedlong hours, were frequently beaten, andwere paid a pittance.

British census data reveal that the in-cidence of child labor was very high inthe early and middle nineteenth cen-tury. According to the Census of En-gland and Wales in 1861, 36.9 percentof boys in the 10–14 age group were la-borers, and the statistic for girls was20.5 percent.7 A comparison with Table

2 shows that though the incidence ofchild labor was higher for Africa andAsia in 1950, currently no continentalregion in the world has higher partici-pation rates than Britain did in the mid-dle of the last century. However, somenations, such as Ethiopia, have a muchhigher rate (see Table 2). While theparticipation of children in the laborforce may have been particularly high inBritain, the experience of other indus-trializing nations, such as Belgium, theUS, and Japan were not very different(Rene De Herdt 1996; Donald Parsonsand Goldin 1989; and Osamu Saito1996).8

While, historically, attention wasdrawn to the child labor problem duringthe industrial revolution, there arescholars who have argued that the prob-lem was not especially acute during thistime. According to them, child labor

7 Going by the census, child labor in Englandand Wales peaked in 1861 for boys, and in 1871for girls. The participation rates dropped offrapidly after 1871 (see Cunningham 1996).

8 For a general discussion of intercountry expe-rience in the last century see Weiner (1991) andCunningham and Viazzo (1996).


was comparably widespread even in theearly eighteenth century, though thechildren did not work in factories atthat time. But most of such observa-tions are based on impressions ratherthan statistics. Moreover, the experienceof working on farms, alongside one’sparents, is arguably less grim than work-ing in factories for fourteen hours at astretch (Cunningham 1990), as suggestedby the interview quoted above.

This view finds reinforcement in acontemporary study conducted by re-searchers at the Delhi School of Eco-nomics9 and the Indian Social Institutewhich shows that child labor in rural ar-eas is often “light,” so much so thatthese children ought to be able to geteducation without seriously cutting intotheir work commitments, if they had ac-cess to proper schools (Bhatty et al.1997). This study, based on interview-ing 1,221 rural Indian parents, foundthat among the out-of-school children,about half worked less than three hourson the day preceding the survey andonly 18 percent worked more than eighthours. The story is very different forfactory workers and organized child la-bor, as, for instance, in the matchindustry of Sivakasi (Kothari 1983;Kulkarni 1983).

The large incidence of child labor inthe industrialized nations in the lastcentury gave rise to a lot of debate andideas from which we can benefit to-day.10 What is quite striking and note-worthy is that child labor has not alwaysbeen thought of as an evil. There havebeen times when it was treated as un-pleasant to the child, but neverthelessdesirable, somewhat akin to our con-temporary view of education. Thus we

find an eighteenth-century writer ob-serving that “parents, whose childhoodwas spent in idleness, have contractedevery absurd prejudice against the em-ployment of children, as unnatural,cruel and unprofitable.”11 B. L. Hutchinsand A. Harrison (1903) have recountedmany instances of this attitude. Theyquote, for instance, a 1770 document,which argues that “being constantlyemployed at least twelve hours in a day. . . we hope the rising generation willbe so habituated to constant employ-ment, that it would at length proveagreeable and entertaining to them . . .[From] children thus trained up to con-stant labor we may venture to hope thelowering of its price” (Hutchins andHarrison 1903, p. 5).12

Nevertheless, as the excesses of childlabor increased through the early nine-teenth century, opposition alsomounted. And by the late nineteenthcentury, child labor was on the decline.It is true that this institution wouldsoon be “exported” out of the industri-alized nations, as the practice of childlabor got shifted to the colonies, butwithin the boundaries of industrializednations, it was undeniably on the wayout.

How did this happen? There is noconsensus on this. On the one hand,there was a series of legislative acts lim-iting child labor and ultimately declar-ing it unconditionally illegal, and there

9 I am grateful to Jean Dreze, who was partof this team, for a very valuable discussion onthis.

10 For a lucid summary account, see Cunninghamand Viazzo (1996a).

11 Quoted in Cunningham (1990, p. 120).12 By the time the child labor debate picked

up steam in the United States in the late nine-teenth century, child labor was viewed as it is to-day, as an unmitigated evil. The reasons givenagainst the institution of child labor are also ratherlike the ones given today, the important exceptionbeing the argument of “race degeneracy.” Childlabor, it was reasoned, should be ended in theSouth because that would help “the preservationof its Anglo-Saxon stock” (A. J. McKelway 1906,p. 261). Similar sentiment was expressed repeat-edly by several authors in the March 1906 issue ofThe Annals of the American Academy of Politicaland Social Science.


were also rules about compulsory edu-cation which made it difficult for chil-dren to work full time. On the otherhand, the increasing prosperity ofEurope, the USA, and Japan made iteasier for parents to pull children out ofwork without having to fear that thiswould commit the household to poverty.

All these factors must have played arole with important regional and sec-toral differences. Peter Scholliers’(1995, p. 208) study of child labor inGhent, Belgium, revealed that by themid-nineteenth century “the number ofchildren under twelve had diminishedsubstantially, and this without any legis-lative intervention.” By contrast, thelaw played an important role in the de-cline of child labor in the cotton millsof Manchester (Per Bolin-Hort 1989).Martin Brown, Jens Christiansen, andPeter Philips (1992), in their study ofthe fruit and vegetable canning industryin the US between 1880 and 1920,found that the decline in the incidenceof child labor was due to both naturaleconomic reasons and legal factors,though the economic forces were thestronger explanatory variable.

There is, however, some agreementon one matter. If one is using legislativefiat to fight child labor, it is more effec-tive to legislate for compulsory educa-tion instead of simply banning child la-bor (Weiner 1991). One good reason forthis is that a child’s presence in schoolis easier to monitor (and thus ensure)than a child’s absence from work.

There were others who, in the con-text of the historical debate, felt thatlegislation, whether it be for compul-sory education or a direct ban on childlabor, cannot be as effective as economicprogress, and that the right policy is towait for economic progress. Such a posi-tion is associated with Nardinelli (1990;and, for discussion, see Cunninghamand Viazzo 1996).

One of the more systematic investiga-tions of the historical role of law in thedecline of child labor occurs in thework of Moehling (1998). Her focus ofstudy is the United States from 1880 to1910. This is part of the period duringwhich the incidence of child labordropped off rapidly. It was also a periodof activism against child labor. In 1900twelve states had a minimum age limitof fourteen years for manufacturingemployment. By 1910 thirty-two stateshad such a restriction. The questionMoehling investigates is whether it wasthe legislation that caused the declinein child labor or it was the diminish- ing dependence of industry on childlabor that made the judicial activismpossible.

The method that she uses is the “dif-ference of difference of difference”(DDD). A more standard method (forinstance, one used by Joshua Angristand Alan Krueger 1991, and RobertMargo and T. Aldrich Finegan 1996)would have required her to compare thedifference in the difference in the occu-pation rates of children aged thirteenyears and fourteen years in states thathave a law debarring those below four-teen years old from working in manu-factures, and states that have no suchlaw, to see if the law had an impact.What Moehling does is to add anotherlayer of differencing by looking at thedifference-in-difference before and af-ter some states adopted such legisla-tion. This presumably nets out the ef-fect that there may be differencesbetween different states, stemming notfrom law but more inherent traits, suchas industrial structure. Using nationalrandom samples of households from thecensuses of 1880, 1900, and 1910 andthe method of DDD, Moehling con-cludes that minimum age restrictionshad little impact on employment in theUS at the turn of the century.


The finding is important but notconclusive. Once one begins to go theroute of adding on differences, strictlyspeaking, there is no stopping. Just asthere may be innate differences be-tween states, there may be innate dif-ferences (in particular, ones that arenot the product of law) in changes overtime across states.

Turning to the subject of state inter-vention, one must, in assessing the effi-cacy of a law, distinguish between thecase where the law is not properly im-plemented, and one in which the law isimplemented but its net effect on soci-ety is not desirable. What Moehling’sstudy points to is the inadequate imple-mentation of the US law. This does notmean that the law is not the rightmethod of intervention for eradicatingchild labor. Another country at anotherpoint in history may be able to imple-ment a law that had, allegedly, failedelsewhere in the late nineteenth cen-tury. An important policy question fac-ing us is whether a legislative interven-tion is a desirable one. Does it promotewelfare when properly implemented?To answer such questions it is essentialto have a theoretical model for analyz-ing child labor. Later, when we havestudied such models, we will see thatthere is scope for using both legislativeand nonlegislative interventions. Policytoward child labor has to be more nu-anced and context specific than whatgovernments and international organi-zations have attempted. But before weget to such a conclusion it is useful torecapitulate what the major policy op-tions are, and the role of differentinstitutions.

3. The Policy Questions

In the battle against child labor, a va-riety of laws and interventions have beentried and even more discussed. In re-

counting this debate, it is useful to dis-tinguish between three kinds of inter-ventions and institutions: intra-national,supranational, and extra-national.

Intra-national effort consists of thelaws that a country enacts and interven-tions that it plans in order to controlchild labor within the national bound-ary. This invariably comes with a certainamount of institutional paraphernalia,such as organizations set up to adminis-ter the law and otherwise dissuade childemployment. Contrary to the popularview, the question is not just betweenbanning and not banning child labor.Instead, a range of different instru-ments have been tried. This varies fromone of the earliest pieces of legislation,such as the child labor law enacted bythe state of Massachusetts, USA, in1837, which prohibited firms from em-ploying children under the age of fif-teen years who had not attended schoolfor at least three months in the previousyear, to contemporary laws, such as Ne-pal’s Labor Act of 1992 and The Chil-dren’s Act of 1992, that place restric-tions on indigenous child labor.13 Manycountries also have important nongov-ernmental, intra-national efforts to curbchild labor. The Daughters’ EducationProgramme (DEP) in northern Thai-land is one such effort. DEP tries toprevent young girls from going intoprostitution by providing education andmobilizing local opinion. Literacy pro-grams and compulsory education, to theextent that these get in the way of la-bor, may also be viewed as an intra- national intervention. I return to thesubject of education below.

Supranational interventions are those

13 The Children’s Act of 1992 prohibits childrenof age thirteen years or less from doing any workas employed laborers. The Labor Act of 1992places restrictions on the kind of labor that chil-dren can do in “enterprises” that employ ten ormore persons (ILO 1995).


attempted through international organi-zations, such as the ILO, the WTO, andUNICEF, which by establishing con-ventions, and encouraging and cajolingnations to ratify them, have tried tocurb child labor. The most powerful,and also controversial, instrument thatthe supranational institutions can use tocurb child labor is the imposition of “in-ternational labor standards,” that is, aset of minimal rules and conditions forlabor which all countries are expectedto satisfy. Since the adoption of suchstandards makes it possible to take pu-nitive action (such as imposing tradesanctions) against defaulting nations,these can be potentially quite effec-tive. Labor standards are discussed inSection 8.

Thanks to controversy and a diver-gence of opinion, the world has beenslow to adopt international labor stan-dards. This has led some developedcountries to consider legislation andother action in their own countriesthat could curb child labor in develop-ing nations. Such actions are what Ilabel as extra-national. Consider, forinstance, the Child Labor DeterrenceAct, or the so-called “Harkin’s bill,”which has been debated extensivelyin the US Congress. In a nutshell, thisis a law that seeks to disallow the im-port into the US of goods that havebeen produced with the help of childlabor. The law will work within the USbut is nevertheless expected to have astrong deterrent effect in developingcountries. There are variants of theHarkin’s bill that have been consideredand debated in the US. The “Sanders’amendment,” which seeks to amend theTariff Act of 1930 so as to deter the im-port of goods produced by unfree orbonded child labor, is another (and, tome, a more reasonable) example of ex-tra-national intervention, as are recentefforts to have imported goods that

are “child labor free” to be labeled assuch.

The trouble with such extra-nationalinterventions is that these can come tobe misused by lobbies and protection-ists representing narrow, sectarian in-terests. It is possible to take differentviews about the original motivation be-hind a bill. Krueger (1997), for in-stance, did an unusual test of the hy-pothesis that it is protectionism thatprompted the Harkin’s bill (see alsoDani Rodrik 1997). His study was basedon checking what kinds of constituencythe sponsors of the bill come from; andhe concluded that the motivation forthe bill was humanitarian concern. Onthe other hand, while the recent ver-sions of the Harkin’s bill have no men-tion of trade and protection, giving theimpression of its only concern being thehumanitarian one regarding the plightof children, an early version of the billhad openly appealed that adult workersin the United States and other devel-oped countries should not have theirjobs imperiled by imports produced bychild labor in developing countries (USDepartment of Labor 1992, p. 5; seealso Basu 1994; and Sen. Tom Harkin1994).14 Fortunately, the original moti-vation is not the relevant question.What one has to be aware of is thatsuch laws can be misused by the forcesof protection.

An intra-national intervention whichdeserves to be discussed separately iseducation and compulsory schooling.Historically this has been considered amajor instrument for eradicating child

14 Subsequently, the text of the bill (version ofApril 17, 1997) has changed further and this timein the right direction. It now tends to isolate childlabor performed “under circumstances tantamountto involuntary servitude” or “under exposure totoxic substances or working conditions other-wise posing serious health hazards” as the kindagainst which the import restriction in the US willprincipally apply.


labor. The relation between educationand child labor has been an area of ac-tive empirical investigation.15 The im-portant finding is that not only arethese not mutually exclusive activitiesbut there may be important comple-mentarities between them. George Psa-charopoulos’ (1997) study, using house-hold survey data from Bolivia andVenezuela, shows that, though workingchildren contribute substantially tohousehold incomes, the educational at-tainment of children who work is sig-nificantly lower than that of nonworkingchildren. Patrinos and Psacharopoulos’(1997) research using Peruvian data,however, reveals that child labor is notdetrimental to schooling and leaves theauthors wondering if in some cases“working actually makes it possible forthe children to go to school” (p. 398).Presumably they are referring here topart-time work which leaves children thetime to go to school. Hence, a restrictedamount of child labor and schooling canactually be complementary. This is es-pecially true in rural areas and the ur-ban informal sector where work hoursare not rigid. In his study of child laborin rural Bengal, Arup Maharatna (1997)found that male (female) children, aged10–15, worked on average 7.78 hours(4.59 hours) a day, and of that only 3.01hours (0.01 hours) were spent on formalwage employment. The importance andpossibility of work-and-school amongthe very poor is also brought out well inGrootaert’s impressive empirical studyof child labor in Cote d’Ivoire. Thisdoes not detract from the fact that com-pulsory education can play a role in lim-iting child labor (Weiner 1991). More-ver, even if the education is notcompulsory, the mere availability ofgood schools can do a lot in diverting

children away from long hours in thework place (Jean Dreze and HarisGazdar 1996; Tony Addison et al. 1997).

In debating policy questions, in par-ticular the use of compulsive measures,such as legal bans and compulsory edu-cation, it has to be kept in mind that fora child to work is not the worst thingthat can happen. So when we stop childlabor, there must be reason to believethat this will not make children worseoff, for instance, by causing starvationor bodily harm.16

Even if legal intervention in the childlabor market is found to be undesirable,this does not mean government shouldsit back and wait for natural economicgrowth to gradually remove childrenfrom the labor force. Government canintervene in the market to create a vari-ety of incentives, such as providing bet-ter and more schools, giving schoolmeals, and improving conditions in theadult labor market, which result in areduction of child labor.

The justification for a lot of these in-terventions depends on whether we be-lieve that the state is more concernedabout the well-being of children thanare the parents of the children. I be-lieve that such a presumption would bewrong when child labor occurs as amass phenomenon as distinct from casesof isolated abuse. This is not to denythat there are contexts where legalbans, total or sectoral, are desirable.But one needs careful, theoreticalanalysis to identify the contexts wherethese are likely to be beneficial.

4. Early Theoretical Ideas

Given the widespread prevalence ofchild labor in the last century, it is not

15 For instance, Peter Jensen and Helena Niel-sen (1997), Grootaert (1998), and Psacharopoulos(1997).

16 A similar point has been made by Sarah Bach-man (1995, p. 3), who observed that attempts tobar children from working in the manufacturingsector in Bangladesh have pushed some of themover to prostitution.


surprising that the origins of our con-temporary mathematical models andtheoretical constructs can be found, al-beit in a much more primitive form, inthe contributions of earlier writers suchas Karl Marx, Alfred Marshall, andArthur Pigou.

Marx, writing at a time when the inci-dence of child labor in factories was at apeak, had a lot to say on the subject.Focusing here on the theoretical ideas,it is interesting to note how Marx inCapital (Volume 1, chap. 15, sec. 3) vir-tually outlined a formal model of thecause of child labor. He first noted how,with the rise of new technology, in par-ticular machinery, there arose scope foremploying those “whose bodily develop-ment is incomplete, but whose limbsare all the more supple. The labor ofwomen and children was, therefore, thefirst thing sought by capitalists whoused machinery” (Marx 1867, p. 372).

The availability of machinery can, inan ideal world, create more time for lei-sure. But Marx noted that since the ma-chinery was owned by one agent and la-bor by another, a diminished need forlabor would tend to depress wages,17 somuch so that (1) it may be worthwhilefor the capitalist to use labor liberallyand (2) it may be necessary for workersto have their entire family work in or-der to make ends meet. Marx (1867,p. 373) writes: “[Machinery] thus de-preciates [the man’s] labor power . . .In order that the family may live fourpeople must now, not only labor, butexpend surplus-labor for the capitalist.”This argument is very close to ideas thatwe pursue later, and can give rise to thepossibility of multiple equilibria, asshown in Section 6.1.

Marx also noted the long-term debili-

tating consequences of child labor.18

But it was Marshall who pursued thisidea, to the point of sketching a dy-namic argument. Marshall (1920, p. 620),with his characteristic fastidiousness,observed that children had labored evenbefore the industrial revolution “but themoral and physical misery and diseasecaused by excessive work under badconditions reached their highest pointin the first quarter of the [nineteenth]century.” He noted (p. 469) that “themost valuable of all capital is thatinvested in human beings.”

Most interestingly, he showed aware-ness of the dynamics of these observa-tions: “The less fully [the children’s]faculties are developed, the less willthey realize the importance of the facul-ties of their children, and the less willbe their power of doing so. And con-versely any change that awards to theworkers of one generation better earn-ings, together with better opportunitiesof developing their best qualities, willincrease the material and moral advan-tages which they have the power to of-fer to their children” (Marshall 1920,p. 468). Some of the features of thisquote will be captured in a formalmodel in Section 6.4.

On policy, most early writers rangedfrom favoring a ban on child labor toplacing severe restrictions on the quan-tity and quality of child labor. Pigou(1962), who favored a ban, was awarethat a ban could cause poor families todip below their subsistence level, andso argued that a ban should be coupledwith social welfare being provided bythe state to the neediest families. Hedid not tell us what his prescriptionwould be if such social welfare was notforthcoming.

17 Empirical investigations cast doubt onwhether average wages did fall during this period.However, in certain sectors, such as the handloom,this was certainly the case (John Lyons 1989).

18 On the matter of policy, Marx was against atotal ban on child labor for the Europe of his time.He favored restrictions on working hours andcompulsory education (Marx 1875).


On what theoretical grounds didthese early writers support governmentintervention in this matter? As we shallsee later more formally, there is a hostof possible arguments, but for a longtime the most popular has been that of“externalities.” Keeping a child awayfrom education may mean missing outon benefits for society at large which donot accrue to the parent who takes thedecision. For one, the main benefitgoes to the child. As Marshall (1920,p. 470) noted: “Whoever may incur theexpense of investing capital in develop-ing the abilities of the workman, thoseabilities will be the property of theworkman himself: and thus the virtue ofthose who have aided him must remainfor the greater part its own reward.”John Stuart Mill also stressed the posi-tive externality of education, arguingthat for a parent not to educate thechild is a breach of duty not only to-wards the child but “towards the mem-bers of the community generally, whoare all liable to suffer seriously fromthe consequences of ignorance andwant of education in their fellow citi-zens” (Mill 1848, p. 319). By extension,he concluded (p. 323): “Children, andyoung persons not yet arrived at matur-ity, should be protected . . . frombeing over-worked. Laboring for toomany hours in the day, or on workbeyond their strength, should not bepermitted.”

Externalities are such a well-knownargument for intervention that it isoften used too cavalierly by economistsand noneconomists. It is frequentlyused as a facade when the real reasonsare more self-serving. In the context ofchild labor, the externality argumentneeds to be made carefully for it tocommand attention. An excellent state-ment of this occurs in the recent workof Grootaert and Kanbur (1995). Groo-taert and Kanbur consider the possibil-

ity that the social returns to educationmay exceed private returns. So govern-ment intervention to direct childrenaway from work and to the classroommay be desirable. The ideal policy forachieving this, according to them, is tobolster the returns to education. Theyconsider a ban on child labor to be asecond-best intervention.

An extreme case of externality arisesin the model of Gupta (1998) in whichthere is a total bifurcation betweenagency and welfare, since parents andemployers take the child employmentdecision entirely in their self-interest.The child is simply an instrument oftheir bargain. The model in section 5.2may therefore be viewed as a model ofextreme externality.

5. Bargaining Models

The formal analysis of child labor isclosely related to the modeling ofhousehold behavior. The early venturesin this direction (for instance, MarkRosenzweig and Robert Evenson 1977;Goldin 1979) were models of householddecision-making, which tried to simulta-neously explain decisions of consump-tion and child labor and, at times, alsochild schooling and fertility. The speci-fications were kept simple enough to al-low for this greater generality and alsoto allow for empirical testing. Sub-sequent work moved away from this toallow for the possibility that a house-hold’s behavior is not determined byone benevolent dictator, but instead isthe outcome of internal bargains andpower struggle. Models involving bar-gaining can and have been used to ex-plain child labor and the level of well-being of children, though by and largesuch efforts treat child labor not as thefocus of analysis but a fallout of gen-eral household modeling. In this sec-tion, I present the main outline of these


models, focusing on and drawing outtheir implications for child labor.

Bargaining models of child labor maybe classified into two distinct kinds, de-pending on who the agents involved inthe bargain are. According to one view,the bargain occurs within the family,between the parent and the child. Theother approach treats the employer andthe parents of the child as the agentsinvolved in bargaining.19 These twomodels are the subject matter of thenext two subsections, respectively.

5.1 Intra-Household Bargaining

The traditional model of the house-hold, known as the unitary model, char-acterizes the household as a single unitof decision making (Becker 1964). Thisis a valid model if one person in thehousehold happens to be a dictator or allpersons have the same utility function.

There is, however, increasing evi-dence that a household’s consumptionpattern tends to change as the composi-tion of who earns how much changes,even when the total earnings of thehousehold are unchanged (see, for ex-ample, Duncan Thomas 1990; John Straussand Thomas 1995; and Moehling 1995).This is usually taken to indicate that thehousehold is not a single conflict-freeunit of decision making but, instead, anarea of bargaining, where a person’sbargaining power depends on the re-sources one brings to the household andone’s fallback options.20

A general representation of these ap-

proaches occurs in what is known as thecollective model (Francois Bourguignonand Pierre-Andre Chiappori 1994).Moehling (1995) has adapted this modelfurther by taking explicit account of thechild. For a simple version of thismodel, we could think of the householdas being characterized by one parent(agent 1) and one child (agent 2). SinceI am here not interested in the demandfunctions for different products, I shallassume that there is only one good inthe economy and xi is the amount con-sumed by agent i. Let the unit of thegood be chosen such that its price hap-pens to be 1. As in Moehling’s model,suppose each agent in the household isconcerned about the consumption of allmembers of the household.21 Let ui beperson i’s utility function.

The household’s utility function is aweighted average of u1 and u2 wherethe weight attached to the parent’s util-ity, α, depends on the incomes of theparent and the child, denoted respec-tively by y1 and y2. In other words, in ahousehold’s utility function, who getshow much weight depends on whobrings how much money.22 Hence, inthe collective model, the household’sdecision-problem is as follows.

Max{x1,x2}

α(y1,y2)u1(x1,x2)

+ [1 − α(y1,y2)]u2(x1,x2).

subject to x1 + x2 ≤ y1 + y2.It is assumed that

∂α∂y1

≥ 0,∂α∂y2

≤ 0,∂u1

∂x1 > 0,

∂u1

∂x2 ≥ 0,

19 These studies, therefore, exclude the case of“street children” who live on their own and belongto no household as such. The present paper alsodoes not address the problem of homeless chil-dren. This is, however, not a negligible category(see William Myers 1988).

20 See, for instance, Marilyn Manser and MurrayBrown 1980; Marjorie McElroy and Mary Horney1981; Nancy Folbre 1986; Amartya Sen 1990; Mar-tin Browning et al. 1994; Grootaert and Kanbur1995; Christopher Udry 1996; and Bina Agarwal1997.

21 It is worth stressing that the important specialcase of this model is where each person is con-cerned about his own welfare, and the final house-hold consumption and labor supply are entirelyconsequences of the power of different individuals.

22 In this model the household allocation is al-ways Pareto efficient. The empirical validity of thisis not uncontested (see Udry 1996). Many take y to be i’s unearned income. I prefernot to do this for reasons explained below.


∂u2

∂x1 ≥ 0,

∂u2

∂x2 > 0 and 0 ≤ α ≤ 1.

Moehling estimates a closely-relatedmodel using household data from earlytwentieth-century urban America, andfinds that working children receive alarger share of household resourcesthan nonworking children.

This model can be taken further toexplain a child’s participation in thework force. To see this, note that, strictlyspeaking, members of a household con-sume not just goods and services butalso leisure. For simplicity, assume thatthe adult always works. Let e be thework done by the child, where e ε[0, 1].The child’s consumption of leisure is,therefore, 1 − e.

Given that each person’s utility de-pends on x1, x2 and e, a natural exten-sion of the above model is to think ofthe household as facing the followingdecision problem, which I shall call the“collective maximization problem”:

Maxx1,x2,e

α(y1,y2)u1(x1,x2,e)

+ [1 − α(y1,y2)]u2(x1,x2,e).

subject to x1 + x2 ≤ y1 + y2.

This is, however, a more complicatedproblem than appears at first sight,since the child’s income, y2, depends onthe choice of e. Thus it is not clear thaty2 can be thought of as an exogenousvariable.

There are two alternative routes thatone can take from here. One way out ofthis is to treat α as a function of theprice vector, as in Bourguignon andChiappori (1994). Then, since the wagerate is a price, α will turn out to be afunction of the wage rates of adults andchildren (w1 and w2) rather than y1 andy2. And of course the budget constraintwill now be:

x1 + x2 ≤ w1 + ew2,

keeping in mind that the adult alwaysworks full-time.

This gives us a straightforward opti-mization problem, which avoids thecomplication of simultaneity. The em-pirical plausibility of this approach, how-ever, seems open to question. It impliesthat a person’s bargaining power in thehousehold depends not on the person’sactual share of the household incomebut on the wage that one could earn, ifone worked full-time. Sociological stud-ies, on the other hand, suggest that awoman’s bargaining power in the house-hold is diminished if a woman does notdo outside work (Nancy Riley 1997).

A second way out of this difficulty isone that rectifies this weakness and as-sumes that a person’s power depends onhow much income the person actuallybrings to the household budget. This is,however, technically more complicated.Note that in the collective maximizationproblem the decision maker is, effec-tively, an amalgam of the members ofthe household with the preference ofeach member receiving some weight.Once the child’s effort (or, for that mat-ter, anybody’s effort) is thrown in as oneof the variables over which a decisionhas to be reached, the weights that indi-viduals receive (here captured by α and1–α) depend on what decision getstaken. In other words, who the decisionmaker is depends in part on what thedecision is. Hence, the collective deci-sion problem cannot be thought of as anormal-form game, since in such agame the players or the decision makersare primitives. One direction that onecan pursue is to take y2 as exogenousand have the household decide on x1, x2and e; and then check whether we re-cover the same y2. If we do, then wehave a “household equilibrium.”

In order to do this, let us take achild’s wage rate to be w; that is, a fullunit of work by a child yields a wage of


w. If a child works for e units, his totalincome, y2, is ew. It is assumed that, asfar as the household is concerned, w isgiven exogenously. With y1 and w givenexogenously, we describe (x1

∗, x2∗, e∗) to

be a household equilibrium if (x1∗, x2

∗, e∗)is the solution to the above collectivemaximization problem, with y2 = e∗w.

This equilibrium is not necessarilyunique, nor is it obvious that it alwaysexists. The existence problem is bestunderstood by converting the house-hold’s problem to a search for a fixedpoint of a correspondence.

Let Φ = [0,1] → [0,1] be a correspon-dence defined as follows. Given anye′ ε[0,1] , we can take y2 = e′w (recall w isgiven exogenously). Now solve the col-lective maximization problem taking y2to be e′w (recall y1 is given exog-enously). All values of e that are a partof such a solution are the set Φ(e′). Thisis what defines the correspondence Φ.It is now plain that a household equilib-rium exists if and only if Φ has a fixedpoint.

Given that the maximand in the col-lective maximization problem need notbe concave, Φ need not be convex-val-ued and we are not able to use Kaku-tani’s fixed-point theorem. Indeed it ispossible to construct reasonable exam-ples where Φ has no fixed points. Onthe other hand, it is also possible tothink of restrictions on the parameterswhich ensure that an equilibrium exists.One simple example of this is the text-book case of selfish individuals: Let usassume

u1(x1, x2, e) = f1(x1)u2(x1, x2, e) = f2(x2) − c(e)

That is, the parent is interested in hisown consumption, the child in thechild’s own consumption. It is worthemphasizing that this is still a collectivemodel of the household, since thehousehold’s maximand continues to be a

weighted average of the utilities of themembers of the household. Hence, theallocation of resources continues to bePareto efficient. Let us also, as is usual,assume, f1′, f2′, c′ > 0, f1′′, f2′′ < 0, and c′′ > 0.

These assumptions ensure that forevery y1 and y2, there is a unique (x1, x2,e) which solves the collective maximiza-tion problem, and that the solution var-ies continuously with y2. Thus, ignoringother variables, we can write Ψ(y1, y2)to be the value of e which solves themaximization problem, given y1 and y2.Since for every e ε [0, 1], y2 = ew, wecan define Φ(e) = Ψ(y1, ew). Since Φ isa continuous function from [0, 1] to [0,1], by Brouwer’s fixed point theorem,there exists e∗ such that

e∗ = Φ(e∗) = Ψ(y1, e∗w) (1)e∗ is the amount of labor that the childwill supply in equilibrium.

Note that e∗ can be written as a func-tion of y1 and w, by using (1). Thus

e∗ = e∗(y1, w).In other words, the child’s labor supplydepends on the adult wage (y1) and childwage (w) that prevails on the market. Itis possible to develop a more elaboratemodel with more exogenous variables. Itwill be interesting to draw out the em-pirical implications of such a model andput them to test. Another direction thatcan be pursued is to model the interac-tion between household members as asequential game in which a player’spower in period t depends on his incomeshare in period t-1. Instead of pursuingthis here, I shall instead move on now todiscuss models that were developed ex-plicitly to explain child labor. One suchmodel is that of Gupta (1998).

5.2 Extra-Household Bargaining

Citing a survey work currently afootin some villages of West Bengal, India,Gupta (1998) argues that a child hasnegligible bargaining power in the


household and is, effectively, an instru-ment for the parent’s maximization ef-fort. In addition, he assumes that par-ents are entirely selfish in the sense ofbeing uninterested in the well-being ofthe children per se. While a parentowns his child’s labor, he is unable tomake the child work productively forwant of complementary resources, suchas land or cattle. He found that in WestBengal villages the bulk of child labor isdirected to the maintenance of cattle.So for work the child has to go to anemployer, who has the resources.

The bargaining that occurs in thismodel is between the parent and theemployer over (i) the wage, w, that is tobe paid for the child’s work and (ii) thefraction, λ, of the wage that is to be paidin the form of food to the child. Gupta(1998) assumes that the parents spendthe cash component of the wage entirelyon themselves. I believe that this is em-pirically questionable, but let us goalong with it as a simplifying assumption.

Using an efficiency wage argument,assume that the output, x, produced bya child laborer is a function of hisconsumption. So

x = x(λw),where x(λw) = 0, for λw less than somepositive number, b, and concave thereaf-ter. That is, for all λw > b, x′(λw) > 0 andx′′(λw) < 0.

Let us assume that if the bargainfails, the parent and the employer earnYP and YE respectively. Hence, thethreat point of the bargaining problemis given by (YP, YE). If the parent andthe employer agree on (w, λ), their in-comes are, respectively, (1 − λ)w andx(λw) − w.

Therefore, the Nash bargaining prob-lem consists of solving the followingproblem:

Max{w,λ}

[(1 − λ)w − YP][x(λw) − w − YE]

This gives us the first-order conditions:

[(1 − λ)w − YP]x′(λw)w − w[x(λw) − w − YE] = 0

[(1 − λ)w − YP][x′(λw)λ − 1] + (1−λ)[x(λw) − w − YE] = 0

By rearranging these we get:

x′(λw) = 1 (2)and

x(λw) + λw = 2w + (YE − YP) (3)

Since λw is, effectively, the child’swage, (2) says that the child’s marginalproduct is set equal to 1. This is thestandard result of efficiency wage. Tosee this, use v to denote a child’s wage.Then the net income generated by achild is given by x(v) − v. If v is chosento maximize this, we get x′(v) = 1, whichis exactly what (2) is. In other words, inthis model a child is paid a wage, λw,that maximizes net returns. The valueof this is independent of the threatpoint. Once we know the value of λw,we can use (3) to solve for the valueof w. This depends on YP and YE. As(YE - YP) increases, w falls and λincreases, as is evident from (3).

It is now possible to derive compara-tive statics results by postulating howYE and YP get set and considering vari-ations of this. Gupta (1998), for in-stance, considers the case where YE de-pends on the adult wage that prevailson the market. In particular, he as-sumes that YE falls as adult wage rises.This implies that as adult wage rises,the child’s wage, w, will rise.

One can, however, hypothesize otherexplanations for YE and YP and deriveother theories of how w will depend onthe parameters of the model. What isinteresting about this model is howsharply it contrasts with the model thatwe are about to encounter, one in whichthe parent is altruistically concerned


about the child’s welfare. In the modeljust described, the child’s welfare is no-body’s concern. The child is valued inthe same way as the goose that lays thegolden eggs.

6. Multiple Equilibria and GovernmentIntervention

What the early models seemed tooverlook is that a labor market, wherechildren are potential workers, will beprone to having more than one equilib-rium, and if it did, then this would raisea variety of interesting policy questions.This is demonstrated in the model ofBasu and Van (1998). It is worth askingwhy the model of Section 5.1 over-looked the possibility of multiple equi-libria. Note that the collective max-imization problem reduces to theunitary model of the household if (i)u1(x1,x2,e) = u2(x1,x2,e), or (ii) α(y1,y2) = 1,for all y1,y2, or (iii) α(y1,y2) = 0, for ally1,y2. But in allowing for this generality,what the model glosses over are somenatural restrictions on the utility func-tion, which is the starting point of themodel of Basu and Van (1998). The lat-ter is based on the unitary model. Thatis, it assumes that (i), (ii), or (iii) isvalid, but this is not its distinguishingmark. I would conjecture that themodel’s central results would remainvalid even if we allowed for bargainingwithin the household. A simplifiedmodel of Basu and Van is presented insubsection 6.1. Subsections 6.2–6.4 dis-cuss the relation between child laborand, respectively, social norms, adultunemployment, and dynamics.

6.1 Model with Altruism

I shall here demonstrate, followingBasu and Van (1998) but using somevery special assumptions, the possibilityof multiple equilibria. The special as-sumptions are for reasons of conve-

nience. The only two essential assump-tions are the following:

Luxury axiom: A household wouldnot send its children out to work if itsincome from non-child labor sourceswere sufficiently high.

Substitution axiom: Adult labor is asubstitute for child labor, or more gen-erally, adults can do what children do.

To give the simplest sketch of thismodel, let me strengthen these two as-sumptions to assert that (1) for everyhousehold i, there exists a critical wage,Wi, such that the household will sendits children out to work if and only ifthe adult wage prevailing in the marketis less than Wi; and (2) adult labor andchild labor are perfect substitutes sub-ject to an adult equivalence correction.Both these assumptions can be relaxedenormously without hurting the conclu-sions of the model. Let us defineW— ≡ Max

i Wi and W__ ≡ min

i Wi.

Suppose that a child’s labor is equiva-lent to γ units of an adult’s labor, where0 < γ < 1. In other words, I am assumingthat adult and child labor are perfectsubstitutes subject to an adult-equiva-lent scale correction of γ.23 A more con-venient way of thinking of this is as fol-lows. Production depends on the totalamount of labor used; and each adult,working all day, produces 1 unit of la-bor, whereas each child, working allday, produces γ units of labor.

23 Admittedly this is a simplifying assumption.What is really needed is the assumption thatadults can do what children do. This is contrary tothe pervasive “nimble fingers” belief—that forsome activities, such as carpet-weaving, childrenare essential. There is however very little empiricalsupport for this belief (ILO 1996). One of themost careful examinations of this occurs in thestudy of India’s carpet industry by Deborah Levi-son et al. (forthcoming). By collecting informationon actual productivity—for instance, square inchesknotted per hour—they reach the conclusion thatfor no activity are children essential. Adults canalways replace them. This is not to deny that suchsubstitution may cause costs to rise, since adultwages are typically higher.


In Figure 2, let the vertical axis rep-resent adult wage (i.e. the wage paid toan adult for a full day’s work). Considera competitive model in which all agentsare price takers. Let AA′ be the supplycurve of the aggregate adult labor in theeconomy. For simplicity we show it asperfectly inelastic. Next consider the to-tal amount of “effective labor” that allthe children can supply. If there are Xchildren in the economy, this will beequal to γX. Add to the aggregate adultlabor supply the effective labor that thechildren can potentially supply in theeconomy and draw another line repre-senting this. Let T′T be this new line.Thus A′T is equal to γX, the totalamount of labor available from thechildren in the economy. In otherwords, if the country had a law that eve-rybody would always have to supply la-bor, then the aggregate supply curveof labor in the economy would indeedbe T′T.

Now it is easy to figure out the ac-tual, aggregate supply curve of labor. Ifthe market (adult) wage is below W__,then all children are sent to work; sototal labor supply is OT. If the marketwage exceeds W—, no child is sent to

work; so total labor supply is OA′. Aswage rises from W__ to W— one householdafter another withdraws its childrenfrom the labor force; so the total supplyof labor keeps decreasing, as shown bythe curve CB. Hence the total supply ofall kinds of labor (that is, adult pluschild) plotted against alternative adultwages gives us the curve ABCT. Let mecall this the “hybrid supply curve,”which should serve as a reminder that itis not quite the standard supply curve.

Not only is this supply curve back-ward-bending but its composition keepschanging as we move along its contour.Along AB it consists of pure adult labor;as we move from B to C it includesmore and more child labor; and from Cto T it consists of all available labor inthe economy.

The possibility of multiple equilibriais now transparent. I shall assume, with-out loss of generality, that wheneveradult wage is W, child wage happens tobe γW. Keeping this in mind, supposethe aggregate demand curve for labor inthe economy is given by DD′. That is,DD′ shows the total effective labor de-manded by firms for every possibleadult wage W. Then there are threeequilibria.24 Let us ignore the unstableone and focus on the ones depicted byE and F in Figure 2. If an economy iscaught at point F, wages will be low(WL for adults and γWL for children)and children will be working. The sameeconomy, however, can be in equilib-rium at point E, where wages are highand children do not work.

If the economy is at equilibriumpoint F, there is scope for an interestingpolicy intervention, which in Basu and

24 An important exception is a sufficiently open,small economy, where wage would be determinedby world prices, and so the demand curve wouldbe horizontal, thereby destroying the multipleequilibria result (Avinash Dixit 1998). Even in thiscase, however, there may be multiple equilibria ina dynamic sense as shown in section 6.4.


Van’s paper is called a “benign interven-tion.” Suppose child labor is banned.Then effectively, the supply curve of la-bor becomes AA′. So if demand condi-tions are unchanged, the economy willnow settle at the only equilibrium, at E.What is interesting is that once theequilibrium settles at E, the law ban-ning child labor is no longer needed(since E was anyway an equilibrium ofthe original economy). It is in this sensethat the intervention is described asbenign. After its initial effect, it goesdormant and can actually be removedwithout loss.25

If the demand curve intersects thesupply curve only once and on the seg-ment CT, then a ban on child labor maywell cause a decline in welfare of theworkers, including the child laborers.26

If this model were fitted into a Wal-rasian description of the entire econ-omy, each equilibrium would be Paretooptimal. So between the equilibria de-picted by E and F neither Pareto domi-nates the other. However, working-classhouseholds are necessarily better off atE. To see this, consider a point (call itE′), where wages are the same as at Ebut all children work. Clearly workerhouseholds are better off at E′ than atF, since in both cases everybody worksbut in the former the wages are higher.

Next, since given the wages at E house-holds prefer not to send the childrento work, by revealed preference weknow that E is superior to E′. Hence, bytransitivity E is preferred to F.

The contestable assumption of thismodel is the luxury axiom, which takesfor granted parental altruism towardsthe child. In the early nineteenth-cen-tury, when child labor took some of itsworst forms in industrializing Europe, astandard critique of the British elitewas that child labor was an outcome ofparental callousness (see Nardinelli1990, p. 94). There is counter-evidence,however, including from those whothemselves worked as child laborers,that the parents sent their childrenout to work typically when they werecompelled to do so by acute poverty(Anderson 1971; Vincent 1981).

Turning to more contemporary evi-dence, Neera Burra (1995) has reportedevidence of parental callousness; andGupta (1998) has cited sources whichsupport this viewpoint (see also Parsonsand Goldin 1989). The much discussedissue of discrimination against the femalechild in several developing regions, no-tably in northern India, suggests that theanswer may also differ depending onwhether we are talking of the male childor the female child.27 In addition, a studyby Alaka Basu (1993) of some slums out-side New Delhi brings to light anotherperverse causation between adult wagesand child labor. When the wages for fe-male laborers rise, starting from a suffi-ciently low level, this often prompts themother to take up work outside thehome, which in turn means that she

25 The model also is suggestive of the link be-tween technology and child labor. If technologychanges so that children become relatively lessproductive than adults (perhaps because of therise of the use of the computer), then we couldthink of γ as falling. From Figure 2 it is clear thatthis would result in T moving left and the inferiorequilibrium, where children work, vanishing.

26 Arguing from a different perspective, HananJacoby and Emmanuel Skoufias (1997, p. 331)reach a similar conclusion: “ . . . our results sug-gest that efforts to expand educational opportuni-ties for the poor [ . . . ] without an understandingof the economic risks and constraints they facemay be met with only limited success. Moreover,compulsory schooling laws or laws against child la-bor, to the extent that they can be enforced inrural areas, could substantially lower householdwelfare.”

27 The source of this discrimination need not bethe household. If job opportunities for females arelimited, it may be a rational response of the house-hold to first educate the boys before turning to thegirls. In the context of adults, the more puzzlingpossibility of gender discrimination in the labormarket originating from within the household ismodeled elegantly by Patrick Francois (1998).


takes the daughter out of school in or-der to have her do the housework. It isarguable that if the mother’s wage risessufficiently, the daughter would get putback into school with household help nowbeing hired from outside. This suggestsan inverted-U relation between adultfemale wage and child labor, especiallythe labor of the female child.

A recent empirical test of the luxuryaxiom, by Ranjan Ray (forthcoming),suggests that the verdict is mixed. Heuses data from Peru’s Living StandardsMeasurements Survey 1994, and Paki-stan’s Integrated Household Survey1991, to check if the luxury axiom isvalid. For this test he constructs adummy variable which takes a value of1 for households with income below thepoverty line and 0 otherwise, and thenuses probit and logit models to test theimportance of this variable on thehousehold’s decision to send its chil-dren to work. The estimated coeffi-cients lend support for the luxury axiomin Peru but not in Pakistan. While thesetests may be interesting in their ownright, they do not really test the luxuryaxiom, because Ray identifies Wi en-tirely with the poverty line, whereas no-where does the axiom suggest that thisshould be so. What does come outrather interestingly from Ray’s paper is,as in Basu (1993), the differential re-sponse of child labor to changes in themother’s and the father’s wages.

One of the best crafted empiricalstudies of child labor is Grootaert’s(1998) work on Cote d’Ivoire. It isbased on annual data collected for 1600households over the years 1985 to 1988.Sorting out carefully various definitions,he goes on to use a sequential probitmodel to find the determinants of childlabor. He corroborates that the charac-teristics of the parents matter and alsothat who has these characteristics, thefather or the mother, matters. It is evi-

dent from his study that the very poorhouseholds often critically rely on thechildren’s income, and he concludesthat, initially, interventions should aimto make possible combining light workwith schooling, instead of a suddenstoppage of child labor.

This is not the forum to settle thiscontroversy but to pursue the conse-quences of alternative viewpoints. Theextreme case in which parents have noaltruism or commitment towards thechildren but treat them purely as labor-producing machines was modeled inSection 5.2. My own preference is forthe altruistic model of this section,while admitting that it is best viewed asa polar characterization of reality. For aclearer conclusion we will have to waitfor further econometric work.

6.2 Norms

With the idea of multiple equilibriain the core, there are several directionsthat one can pursue. Albert Hirsch-man28 has rightly argued that the deci-sion to send a child to work is partly amatter of social norm.29 This may bemade more precise by assuming thatsending a child to work makes the par-ents incur a social stigma cost, c; and ifmany children work, then c is smaller.The latter captures the idea that whilemaking a child work is socially frownedupon, the ferocity of the frown isgreater if you live in a society where vir-tually no one but you sends a child towork. This can yield the result that if allparents send their children to work, it isworthwhile for each parent to send hischild to work, and if no one sends their

28 Personal communication to the author dated15 February, 1995.

29 “Social norm” has many interpretations (Basu1998; Partha Dasgupta 1993). I use it here as asocial influence that changes our preference. Therelation between child labor and cultural norms hasbeen discussed in the book by Gerry Rodgers andGuy Standing (1981), especially in chapter 1.


child to work, each parent may find itnot worthwhile to send his child to work(because the social stigma would be toohigh). The argument here is analogousto the one used by Assar Lindbeck, StanNyberg, and Jorgen Weibull (1999)30 toshow how the number of people whoare unemployed and live off social wel-fare can settle at very different valuesdepending on which of the possiblemultiple equilibria is realized.

This is a simple point analytically butis probably of considerable practical im-portance in the context of child labor.Also this is an idea that can, potentially,be developed much more and used toaddress questions of policy.

6.3 Distribution and Unemployment

In the model of section 6.1, workersare assumed to be ex ante identical, andin the equilibrium that occurs they turnout to be ex post identical. A conse-quence of this is that the model is com-pletely silent on questions of incomedistribution and unemployment. The lit-erature has relatively little to offer onthese topics. The distribution questionis raised in a short theoretical paper byKenneth Swinnerton and Carol AnnRogers (forthcoming); there is little thatis available on this empirically. On theother hand, the relation between adultunemployment and child labor has beenthe basis of some empirical investiga-tion (see Goldin 1979; Patrick Horanand Peggy Hargis 1991; Michel Bonnet1993; C. P. Chandrasekhar 1997); thehiatus here is in the theory. Modelingthe relation theoretically is a large andinteresting topic, but some initial stepstowards it are easy to take using themodel of section 6.1 above.31 Before do-

ing so, I want to briefly recount Swin-nerton and Rogers’ argument.

Their paper begins with the observa-tion that if a parent’s decision to sendthe children to work depends on(among other things) the parent’s in-come, then we should look at not justthe parent’s wage but also any profitsthat may accrue to the parents. So theymove away from the polar assumptionof Basu and Van’s model by supposingthat the firms are owned by a fraction λof the workers. So for workers in thiscategory, income equals wages plus ashare of profits, while the remainingworkers earn only wages. They thenshow that, if in an economy there existsa good equilibrium, then there does notexist an equilibrium in which the divi-dend-earning workers send their chil-dren to work. This directs the policydiscussion to matters of distribution.Giving workers a share in profits turnsout to be a method of curbing child la-bor. This is a valid argument, thoughthe feasibility of such policy changesmay be restricted for reasons of politics.

Turning to the subject of adult unem-ployment, let us suppose that there areN worker households and each house-hold has one adult and m children. Asbefore, child non-work is a luxury goodand, in particular, for each household ithere is a critical adult wage Wi suchthat if the adult wage rises above Wi,the household withdraws its childrenfrom working. Adults, on the otherhand, always prefer to work, no matterwhat the wage. This immediately givesus the aggregate labor supply curve asin Figure 2.

As before, let us assume that onechild manages to produce the equivalentof γ units of adult labor. Hence if theprevailing market child wage, Wc, hap-pens to be equal to γW, where W is theprevailing market adult wage, then firmswill be indifferent between employing

30 For the use of similar assumptions, see MarkGranovetter (1978), and Timothy Besley andStephen Coate (1992).

31 A different model linking adult unemploy-ment with child labor occurs in Gupta (1997).


adults and children. Let us focus on thecase where γW = Wc. I should clarifythat I am not assuming Wc to be ex-ogenously fixed. Rather I am trying toisolate an equilibrium in which Wc hap-pens to be equal to γW.

Within the confines of this case, wecan think of the total effective demandfor labor to be a function of W. Letus denote such an aggregate demandfunction by d(W). Let us assume thatthis is downward sloping and take DD′in Figure 2 as representing such a de-mand curve.

To study the impact of adult unem-ployment on child labor, we first need amodel in which there can exist adult un-employment. The best way to developsuch a model is to use a structure thatcan endogenously explain wage rigidi-ties. I shall, however, for reasons ofsimplicity, stay away from such compli-cations32 and simply assume that adultwage is exogenously fixed at W∗, whereW∗ is such that d(W∗) < N. In otherwords, we are focusing on cases abovepoint E in Figure 2. The questions are:How many adults will find employment?Will children be working and if so whatwill be the incidence of child labor?

In answering these questions I shallconsider the case where γm < 1 andd(W∗) − γmN > 0. This is merely a suffi-cient condition for the existence of anequilibrium in which γW∗ = Wc. Afuller model will have to go beyond thisand consider other cases as well, butthe logic of the relation between childlabor and adult unemployment comesout clearly even in this limited model.

Let us use E to denote the number ofadults that will find employment. Sincethe total demand for labor is d(W∗), itfollows that d(W∗) − E will be theamount of labor demanded from chil-

dren. In other words, the number ofchildren demanded by the firms will be[d(W∗) − E]/γ. Since child wages are notrigid, demand for child labor must beequal to supply of child labor. So whatremains to be done is to determine thesupply of child labor. Note that if onlyE adults find employment, N − E house-holds will be earning zero income fromthe adults. So these will be the house-holds that will send the children out towork. Hence, if E adults find work, thesupply of child labor will be given bym(N − E). In other words, E is equilibriumadult employment if:

m(N − E) = [d(W∗) − E] ⁄ γ.To remind ourselves that the value of Edepends on W∗, let us write the solutionof this equation as E(W∗). Evidently,

E(W∗) = [d(W∗) − γmN] ⁄ (1 − γm).Since d(.) is a downward-sloping func-tion, it follows that as W∗ rises, adultemployment, E(W∗), will fall.

If we use C(W∗) to denote the numberof children who find employment whenadult wage is fixed at W∗, then we have:

C(W∗) = m[N − E(W∗)] = m[N − d(W∗)] ⁄ (1 − γm).

Given that we are considering the casewhere γm < 1, it follows that C(W∗) > 0.Further, as W∗ rises, C(W∗) rises. Sincea rise in W∗ increases adult unemploy-ment, a rise in adult unemployment isassociated with an increased incidence ofchild labor.

This confirms intuition and empiricalfindings but also cautions us about theright policy intervention.33 If marketwages rise, we have good reason to be-lieve that child labor will decline, but

32 For a model of child labor and efficiencywage, see Garance Genicot (1998).

33 The model also allows us to derive sometheoretically puzzling results, such as how theabolition of a legal minimum wage can causewages to rise. To see this, start from a minimumwage, W∗, below point E in Figure 2 and derivethe equilibrium as done above.


that does not mean that this impact onchild labor can always be replicated byraising adult wages by fiat. If the adultlabor market has oligopsonistic ele-ments, a rise in the legal minimumwage can boost adult employment andcause a decline in child labor, but incompetitive labor markets the conse-quence of raising the minimum wagecan be ambiguous.

6.4 Dynamics

One big caveat in the large literatureon child labor is the treatment of dy-namics. Yet the dynamic consequencesof child labor are likely to be large,since an increase in child labor fre-quently causes a decline in the acquisi-tion of human capital. If a child is em-ployed all through the day, it is likelythat the child will remain uneducatedand have low productivity as an adult.

While the analysis of the long-runconsequences of child labor is almosttotally ignored in the literature (Jean-Marie Baland and James Robinson 1998being an exception), one can draw onsome papers on human capital acquisi-tion and its long-run consequences, a laOded Galor and Joseph Zeira (1993),Abhijit Banerjee and Andrew Newman(1993) and Lars Ljungqvist (1993) toderive some hints for modeling the dy-namics of child labor.34 This is what Iattempt here and in the process developthe idea of a “child labor trap.” I usesome strong assumptions to develop myargument but it should be possible torelax these.

Since my aim is to explain multipleequilibria in the long run, it cannot becriticized if I make an assumption thattilts the description of the single-periodmodel away from multiple equilibria.Note that in Figure 2, if the demand

curve is sufficiently elastic, there willexist only one equilibrium in the single-period model. To ensure that this is thecase, I shall assume that the demand forlabor is perfectly elastic.

Consider an overlapping generationsmodel in which each person lives fortwo periods, first as a child and then asan adult; and at the start of the secondperiod gives birth to a child. As a childa person can either work or go to school(that is, acquire human capital). So ifwe denote a full day by 1 and a childworks for a fraction e of the day, thenthe amount of human capital acquiredby the child is 1 − e ≡ h.

The productivity of an adult dependson the amount of human capital, h, heacquired as a child. If we think of L asthe number of labor units produced byone adult, then what we are assuming isthat

Lt = L(ht−1), L′ > 0, L′′ < 0.35 (4)Let us make the normalization assump-tion that L(0) = 1. In other words, theamount of labor produced by an un-skilled adult is defined to be 1. Thatthere is an inverse relation betweenchild labor and the child’s productivity in

34 For an overlapping generations model of pa-rental schooling decisions, with implications for childlabor, see Gerhard Glomm (1997).

35 So we are assuming that if a child worksmore, his productivity as an adult falls. We justifythis by assuming that a child’s nonworking time isspent on studying. Even without this assumptionone can argue that child labor diminishes adultproductivity. As Pigou (1920, p. 751) had noted,“Many forms of unskilled labor at present open toboys not merely fail to train, but positively un-train, their victims.” Madhura Swaminathan’s(1997) study of child labor in Gujarat, India,weakly confirms this. She found that, for boys, en-tering the labor force as a child lowered produc-tivity when they became adults, though for girlsthere was no significant effect one way or theother. It is however noteworthy that in some spe-cial cases this relation tends to go the other way.The English cotton mills of the last century maybe a case in point. D. A. Galbi (1997) has arguedthat the share of child labor in the mills fell duringthe early nineteenth century precisely because theearlier rise in child labor meant that, as thesechildren grew up, there would be a cohort of moreproductive adult workers.


later life is also assumed by Parsons andGoldin (1989), though they do not pur-sue the long-run consequence of thisassumption.

Let us use V– to denote the wage of

one unit of labor. If labor demand isperfectly elastic, this will be constantand so we take the wage rate to be fixedat V–. Hence, in period t, an adult who as achild, that is, in period t – 1, had workedet–1 amount, will have an income of

V–

L(1 − et−1) ≡ Wt. (5)As is the model of subsection 6.1, we

assume that it is the parent who decideswhether to send a child to work. Let usassume that there is a wage W__ suchthat, if adult income is below W__, theparent sends the child to work full time,that is e = 1. Let us also assume thatthere is a wage W— > W__ such that, if adultincome exceeds W—, he does not send hischild to work at all, that is, e = 0. Unlikein the model of subsection 6.1, we takeall adults to be identical and assumethat the child’s labor is not a 0 − 1 deci-sion but monotonically varies with theparent’s income. This may be summedup by assuming that the amount of workthat a child is made to do in period tis a function of the parent’s wage inperiod t:

et = e(Wt), (6)such that e(W) = 1, for all W ≤ W__; e(W) =0 for all W ≥ W—; and e′(W) < 0, for allW∈(W__,W—).

From (5) and (6) we get

et ≡ e(V– L(1 − et−1)) (7)which may be described in brief as

et ≡ Φ(et−1). (8)From the assumptions made above

we know that Φ is upward sloping andbounded above at e(V– L(0)) = e(V– ). In or-der to focus on the interesting case weconsider an Φ which intersects the 45°line more than once. In the case illus-

trated in Figure 3, there are threesteady-state equilibria, of which two arestable. These are depicted by points 0and E. At E, a poor parent makes thechild work full-time. The child acquiresno skills and so as an adult earns verylittle and has to send his child, in turn,to work full-time. This equilibrium de-picts what may be called a “child labortrap.”36 On the other hand 0 depicts anequilibrium where a child goes toschool, earns adequately as an adult andso can and does send his child to school.It is a virtuous cycle.37

There is scope for fruitful interven-tion by government. If an economy iscaught in a child labor trap, what is

36 A different kind of trap has been modeled byMukesh Eswaran (1996). In his framework the ex-istence of the institution of child labor biasesparents towards having more uneducated childrenrather than a few educated children. And thischoice in turn perpetuates the institution of childlabor.

37 It is possible to argue that in addition to thefact that an educated parent is able to educate herchild, the very fact of growing up in a householdin which the adults are educated confers knowl-edge and therefore increased productivity on achild. This could render the better equilibriumeven more dynamic than suggested by this model.For a discussion of the externalities of literacywithin a household, see Basu and James Foster(1998).


needed is a large effort to educate onegeneration, and this can get the econ-omy rolling towards the virtuous equi-librium without need for further action.D. P. Chaudhri (1997) has discussed sev-eral versions of “virtuous spirals” andinstances of how child labor tends todecline rapidly once a “tipping point” isreached.

As is shown by Parsons and Goldin(1989), the sub-optimality of invest-ment in education is closely-related tothe availability of efficient capital mar-kets (see, also, Jacoby and Skoufias1997). There is also evidence from thefield that the availability of credit ondecent terms can rescue many from theperils of child labor, since in developingcountries, such as India, a typical rea-son for a child to drop out of school isnot chronic poverty but a temporarymishap for the household, such as thefather losing a job or a sibling needingmedical support.

The credit question comes up starklyin the model of Baland and Robinson(1998). Theirs is a two-period model ofchild labor in which inefficiency occursdespite parental altruism, because par-ents may run out of resources needed toeducate the child. The only option thenis to borrow against the child’s futureincome, and this is typically not possi-ble. This shows that to solve the prob-lem of child labor one has to go beyondproviding perfect capital markets. Whatis needed is the enforcement of inter-generational contracts, whereby what iborrows for i’s child’s education is paidback by the child when she grows up,by when i may be dead. But it is notclear that creating institutions for en-forcing such contracts is desirable,since it will be open to the moral hazardof “inverse bequests,” whereby a personborrows funds in the name of the child’seducation and leaves the child with arepayment burden.

Another related feature, which I donot model here but is important, is thefertility decision. We have in this papertreated the number of children as ex-ogenously given. In reality, the numberof children a family has is partly voli-tional, and the decision may well de-pend on whether children can find work(Mead Cain 1977; Rosenzweig andEvenson 1977; Eswaran 1996; PranabBardhan and Udry 1999). Cain’s studyof Bangladesh shows that not only dochildren contribute to household in-comes, but a boy becomes a net pro-ducer by the age of twelve years, andafter the age of fifteen years his cumu-lative contribution exceeds their cumu-lative consumption. Plainly, these arefeatures that a proper dynamic modelshould be able to incorporate. There isneed for such a model and the aboveoutline should be viewed as motivationtowards it.

7. Sexual Harassment and Child Labor

Corporal punishment and sexual ex-ploitation have, at several stages of his-tory, been closely associated with childlabor. Commenting on Britain duringthe industrial revolution, Nardinelli(1982, p. 294) has observed how the useof violence was common against childlaborers, though “by the middle of the19th century it was seldom used forolder workers.”38 In a model on the useof violence with special reference tochild labor, Michael Chwe (1990) hastried to explain this asymmetric treat-ment of adults and children in terms ofthe lower reservation utility of the lat-ter. Evidence of violence also croppedup repeatedly during the interviews

38 Nardinelli argued that the use of violence wasa mechanism for raising productivity, a claim thathas not gone unchallenged (see Mary Mackinnonand Paul Johnson 1984, and Nardinelli 1984).For some contemporary accounts of coerced childlabor, see Jonathan Silvers 1996.


conducted by the British parliamentarycommittee investigating child labor in1832. For instance, a cotton spinner,Thomas Daniel, admitted that the childlaborers “are always in terror; and . . .the reason of their being in a state ofterror and fear is, that we are obliged tohave our work done, and we are com-pelled to use the strap . . . ” (BritishParliamentary Papers 1968, p. 325).

In what follows I want to raise a dif-ferent theoretical question concerningviolence and sexual harassment. Con-sider first the case of adult labor. Isthere a case for banning violence andsexual harassment when these occur asa consequence of voluntary contracting?I will argue that the answer is a condi-tional yes, and then try to show that theclaim may be carried over to the contextof child labor to construct a novel argu-ment for banning child labor in certainsituations. Though the theoretical claimapplies to all kinds of violence, I illus-trate it here with respect to sexual ha-rassment. This argument goes to thecore of the philosophical debate con-cerning the state’s right to ban volun-tary transactions between consentingindividuals, such as the signing of self-enslavement contracts or trade in hu-man organs or a child’s decision to sellhis labor. The best of the existing argu-ments are quite tenuous39 and so this isa matter that deserves closer scrutiny.

Economists usually take the line thata voluntary contract between two agentsthat does not have negative externalitieson uninvolved outsiders ought not to bebanned. Let me call this the “principleof free contract.” If we accept this prin-ciple, then it is not clear why sexualharassment in the work place should al-

ways be banned. Consider one kind ofharassment which is ex ante voluntary.Suppose an entrepreneur advertises forworkers, openly saying that he will pay awage above the market rate but he re-serves the right to sexually harass hisemployees. Therefore, if a worker joinsthis firm, it must be that she finds thatthe benefit outweighs the cost.

If we accept the principle of free con-tract, it now seems difficult to justifybanning open sexual harassment of thekind just described. But I believe thereis an argument that can be constructedwhich respects the principle of freecontract, and yet may justify a ban.

Note that under normal assumptionsthe following is true: If harassment isallowed, then those workers who are es-pecially strongly averse to harassmentwill be worse off because the marketwage that they will be able to commandwill be lower than what it would be ifno harassment were allowed by law. Icall this the “harassment lemma,” andbegin by proving it.

Suppose that we have a market withtwo kinds of workers, with type-1 work-ers having a stronger aversion to sexualharassment than type-2 workers. Forreasons of algebraic ease, let us assumethat type 1’s distaste for harassment isinfinite (so they would rather be unem-ployed than face harassment), whiletype 2’s distaste is zero. If the wage rateis W and there is no harassment, theaggregate supply of labor by type-iworkers is fi(W). As usual, fi′(W) > 0.

This is a competitive model; so work-ers and employers are wage-takers. Letme, for simplicity, assume that there isonly one employer and he gets a satis-faction of θ(>0) from harassing eachworker. The production function of theemployer is given by x = x(n), wherex′ > 0, x′′ < 0, n is the number of workersand x the total output.

Let WH be the wage for those who

39 For useful discussions of this debate seeDavid Ellerman (1992, chapter 9) and MichaelTrebilcock (1993, chapter 4). For an outline of thedebate on what constitutes sexual harassment, seeGillian Hadfield (1995).


sign the with-harassment contract andWN for those who sign the no-harass-ment contract. If the employer hires nHworkers under the H-contract and nNworkers under the N-contract, his totalprofit is x(nH + nN) − nHWH − nNWN +θnH. The first-order conditions frommaximizing this can be rearranged andwritten as follows:

x′(nH + nN) = WN (9)WH = WN + θ (10)

Clearly, type-1 workers will sign N-con-tracts and type-2 workers will sign H-contracts. Hence, the total supply ofworkers for N-contracts will be f1(WN)and the total supply for H-contracts willbe f2(WH). Therefore, using (9) and (10),we can say that WN∗ is an equilibrium if

x′(f1(WN∗) + f2(WN

∗ + θ)) = WN∗ .

Consider now a legal regime whereharassment is never allowed. Hence,there is only one wage in the market,W. The employer maximizes x(n) − nW.The total supply of labor is given byf1(W) + f2(W). W∗ is an equilibrium wageif x′(f1(W∗) + f2(W∗)) = W∗.

Since f2′ > 0 and x′′ < 0, it follows thatWN

∗ < W∗. This completes the proof.What the harassment lemma estab-

lishes is this. If we adhere to the princi-ple of free contract, then even thoughwe may have no reason for stopping fi-nite pairs of individuals (an employerand an employee) from getting into ha-rassment contracts, there may be goodreason for adopting the rule that no ha-rassment contracts should be allowed inthe workplace. This is made possibleby the fact that, in competition, the ag-gregate is not simply the sum of allatomistic acts.

Note that the harassment lemma doesnot provide sufficient reason for ban-ning harassment, but simply shows thatallowing contractual harassment cannotbe justified on grounds of the principle

of free contract, since allowing harass-ment typically has a negative externalityon uninvolved individuals. Clearly, anegative externality cannot be a suffi-cient reason for disallowing any action.Otherwise we would have to say that noone should be allowed to work on dayswhen there is a cricket match becausethose who like watching cricket will beadversely affected.

To establish a sufficiency criterion,we need to go beyond economics andidentify human preferences which are“fundamental” in the sense that no oneshould have to pay a price for havingsuch a preference. This may be a con-troversial list but very few people willdeny that there are such “fundamentalpreferences,” and also that not all pref-erences qualify as fundamental. In mostsocieties the preference not to be sexu-ally harassed would be considered fun-damental. No one should have to pay apenalty for having such a preference.On the other hand, the preference formissing work when a cricket match is onwould not be considered fundamentalby most of us. You may of course havesuch a preference but you should beprepared to pay some price for it.

It is the harassment lemma coupledwith the recognition that the preferencenot to be sexually harassed is funda-mental that clinches the case for anoutright ban on sexual harassment.

It is easy to see that the harassmentlemma, as an abstract idea, carries overfrom the domain of sexual harassmentto child labor, by assuming that house-holds have different degrees of aversionto sending their children to work. Forconcreteness turn to the model of Sec-tion 6.1 and assume that there are twotypes of households, 1 and 2, where atype-i household is one which wouldsend its children out to work if and onlyif adult wage drops to below wi. Let usassume that w1 = −∞ (that is, type 1


households never send their children towork), whereas w2 = ∞ (that is, type 2households always send their childrento work).

By using a diagrammatic techniquesimilar to the one used in Figure 2, it isimmediately obvious that the marketwage rises if a ban on child labor is im-posed. Conversely, if a ban on child la-bor is removed, wages must fall; and sotype-1 households become worse off.This establishes the counterpart of theharassment lemma.

Therefore, not having a ban on childlabor penalizes households that have astronger preference not to send theirchildren to work. Following the claimsof the above subsection, this leads to acase for a ban on child labor if we con-sider a household’s preference not tosend its children to work to be a funda-mental preference. To my mind this isnot unambiguously fundamental or non-fundamental. It is plainly more funda-mental than a person’s preference formissing work during cricket matchesand probably less so than a preferencefor not submitting to harassment. So,while the case is not automatic, this isa possible avenue for justifying legalaction against child labor.

8. International Labor Standards

A topic closely related to the subjectof child labor is that of “internationallabor standards” and the use of “socialclauses” as a prerequisite for trade. Formany years, overtly or covertly, peoplehave argued the need to enforce someminimal standards of labor, concerningworking conditions, duration of work,the level of wages, and so on.40

Laborers have often had to work in

appalling conditions for meager wages.So a part of the agitation for minimalinternational labor standards is indeedinspired by a genuine concern for thewell-being of workers. But as is only tobe expected in such situations, forces ofprotection and lobbies with their owninterest in mind have taken up positionsbehind this banner. The presence of in-ternational organizations makes it nowpossible, what earlier would be infeasi-ble, to bring multinational pressure toindividual nations to comply with someminimal labor standards. In forums,such as the ILO, the GATT and now theWTO, the subject of labor standards orsocial clauses has been a live one (Jag-dish Bhagwati 1995). Some industrial-ized nations have campaigned for theinclusion of social clauses in the WTOwhich would either deny nations that donot fulfil minimal labor standards mem-bership in the WTO, or enable othernations to place a trade embargo on anynation that violates the standards.

One major area for setting interna-tional labor standards concerns child la-bor (see Gary Fields 1995; David Bloomand Waseem Noor 1996; Keith Maskusand Jill Holman 1996; Stephen Golub1997; Drusilla Brown 1998). As an ulti-mate objective everybody agrees thatchildren should not work, just as adultsshould not overwork and get underpaid.Given these objectives, what is the rightpolicy response? Here answers have dif-fered widely. One answer is that sinceone country’s acquiescence to lower la-bor standards gives it trading advan-tages in labor-intensive goods, thereshould be multilateral sanctions againstsuch a country; and social clausesshould be used to deter such “illegiti-mate advantages” (see, for instance,Terry Collingsworth, William Goold,and Pharis Harvey 1994; and FrankWilkinson 1994).

At the other end of the spectrum is

40 See Steve Charnovitz (1987) and Srinivasan(1996) for discussion of the roots of the labor-stan-dards movement. For a discussion of alternativeinterpretations of “labor standards” see WernerSengenberger and Duncan Campbell (1994).


the argument that countries trade onthe basis of their relative advantages,and for some countries the advantagelies in their cheap labor. To try to levelthese out through the use of multilat-eral threats is to practice protectionism,which is likely to hurt not just workersin the Third World but consumers inthe developed nations as well. In par-ticular, several economists have arguedthat a social clause in the WTO is notthe right response to child labor andother problems of labor standards(Bhagwati 1995; Srinivasan 1996). In-stead, Bhagwati (1995, p. 757) arguesfor “methods of suasion” and for theILO to be the main international agencyto strive towards better standards.

A variety of positions have also beentaken between the two described above.Rodrik (1996), for instance, argues that,while one has to be aware that the “up-ward harmonization” of labor standardscan rob poor nations of their compara-tive advantage in labor-intensive goods,it may be all right to use trade restric-tions against nations that violate awidely held moral code in the importingcountry so as to safeguard domestic la-bor standards. He calls this a “socialsafeguard clause,” and suggests (1) in-stitutional mechanisms for ensuringthat these are not misused by one na-tion against another and (2) a schemefor providing compensation (under cer-tain circumstances) to developing coun-tries that lose out as a consequence ofsuch sanctions.41

To check on the impact of imposinginternational labor standards on devel-oping nations (and to deduce whichones are worthwhile) it is important tohave a suitable theoretical model. Myinterest here is in child labor standards.

The theoretical modeling of labor stan-dards is still in its nascency (see Bloomand Noor 1996; Maskus and Holman1996; Srinivasan 1996). Of these,Maskus and Holman are most directlyconcerned with child labor standards.Their model is based on introducing the“demand for child labor standards” asan argument in the representativeagent’s Cobb-Douglas utility function.This makes some of the conclusions,such as the need for child labor stan-dards, too direct a consequence of theassumptions of the model; and it alsoleaves no margin for the fact that peo-ple’s attitude to child labor may embodya misunderstanding of the implicationsof child labor for child welfare.

It seems to me more valuable to ana-lyze policy starting from a standard util-ity function, instead of one chosen spe-cially for the occasion. In what followswe derive lessons for labor standardsfrom a model which (i) does not intro-duce “labor standards” as a direct argu-ment in the utility function and (ii)recognizes the interdependence ofdecisions between nations.

Consider the model of Section 6.1.Let us simplify it a little by supposingthat all households are identical and atadult-wage W— they draw their childrenout of the labor force. Let us also as-sume γ = 1. On the other hand, we com-plicate the model a little by assumingthat while the aggregate demand for la-bor is given by AD in Figure 4 and ag-gregate supply by BCFG (its strangeshape is natural and was explained inSection 6.1), the demand and supplyare evenly distributed over t regions,R1, . . ., Rt, of the economy. In otherwords, if there are aggregates of N con-sumers and n firms, each region has N/tconsumers and n/t firms.

In other words, the demand curve ADis a horizontal summation of the de-mand of n identical firms and also the

41 The idea of compensating developing nationsfor adopting higher labor standards was originallycontained in Ron Ehrenberg (1994). See Krueger(1997) for discussion.


horizontal summation of the demandcurves in the t regions.

While we can think of R1, . . . , Rt asdifferent regions of a country, we can alsothink of these as t countries belongingto a global economy. We could think ofthese together as belonging to a “region”of the world. I shall assume that the worldprice of the good produced by this re-gion is one and, thanks to free trade,nothing that happens in the region canchange this price. Labor, on the otherhand, is region specific. Workers cannotmigrate from one country to another.

What we shall do is to study the im-pact of child labor laws on welfare withand without capital mobility. In thismodel I interpret capital mobility en-tirely in terms of the mobility of firms.In other words we shall consider two al-ternative scenarios: one where firms canfreely move between the t countries (orregions), and another in which the firmsare entirely country specific.

Suppose the economy is at the badequilibrium, EB. Wage is WL; the totalemployment is OG. The left-hand panelof Figure 4 depicts what happens ineach region. In each region employmentis equal to O′G′ = (OG) ⁄ t. The figure isdrawn as if t = 2.

Consider first the case where firmsare region specific. If a ban is imposedon child labor in only one country (theone shown in the left panel of Figure4), then, for reasons given in Section6.1, equilibrium will drift from EB′ to EG′

(in the left panel of Figure 4). In thecase illustrated in Figure 4, workers inthis region will be better off by the ban.

Now suppose firms are fully mobile.Once again consider a ban on child la-bor imposed only in one country (de-picted in the left panel of Figure 4).This will imply that the total labor sup-ply in these t countries will be BCHJ,where HF = CF/t, since child labor fromone country is no longer available.42

Hence equilibrium wage rises to W∗.Let us see what the equilibrium looks

like for each individual nation. Clearly,the nation that imposes the ban will beat a point like K, where wage is W∗ andtotal employment is O′J′ (i.e. only adultsare employed). All other nations are ata point such as M. This is caused by theflight of capital (in this case, firms)from the nation imposing the ban.

In this new equilibrium, the countryimposing the ban will have to incur

42 Note also that JG must be equal to J′G′.


large costs monitoring the ban becauseit is not in the self-interest of workersand employers to comply with the ban.In addition, the workers of this nationare likely to be worse off with the ban.This will be especially true if t is large(which will imply that W∗ will be closeto WL).

Now suppose that through an interna-tional labor standards agreement all na-tions agree to impose a ban on child la-bor. There is now no flight of capitalsince one nation is as good as another.So the aggregate equilibrium shifts toEG and each nation shifts from EB′ to EG′ .All workers are better off and the bandoes not need to be monitored since inthis new equilibrium wage is so high (atWH) that parents do not want to sendtheir children to work.

It is true that we get this extreme re-sult by focusing on the case where theaggregate demand curve AD intersectsthe aggregate supply curve more thanonce. Nevertheless, what it illustrates isan important implication of the modelof Section 6.1 for international laborstandards.

There are contexts where, even thoughlabor standards are undesirable for eachnation behaving atomistically, they areworthwhile if the standards are coordi-nated across a large number of nations.Indeed, once this is achieved it may bein the self-interest of everybody to liveby the standards.

One must add some notes of cautionto the above proposition. First, theproposition must not be taken to de-tract from the claim that labor stan-dards are often a front for protection-ism. The proposition claims that it ispossible for coordinated effort to bene-fit all. It does not say that this will al-ways be the case. What the propositionhighlights is the fact that coordinatedeffort is better than individual effort.Second, it is conceivable that once we

allow for the heterogeneity of nations,the case will be (when there is a case atall) for labor standards which varyacross nations. Third, if we worked witha demand curve that was sufficientlylow so that it intersects the aggregatesupply curve of labor once and on thesegment FG, we would find that evencoordinated labor standards imposed onall nations will need monitoring (com-pliance being not in the self-interest ofall individuals), though it will continueto be the case that it is better to have acoordinated effort rather than one im-posed idiosyncratically on some nations.

9. Conclusion

The literature on child labor is an il-lustration of abundance and anarchy.Theoretical writings on the subject arerelatively few, though one finds theo-retical insights in many unexpected pa-pers and books which may be otherwisepurely empirical or descriptive. Theempirical writings on child labor arenumerous but they are usually notfounded on any theory. By bringing to-gether the main theoretical ideas, thissurvey hopes to encourage not just fur-ther theoretical research but empiricalwork which is analytically better founded.

Also evident from this survey is thefact that there is no unique prescrip-tion. Should child labor be banned out-right? Should the WTO be given the re-sponsibility of enforcing restrictions onchild labor through the use of tradesanctions? Should there be a legal mini-mum wage for adults so as to make itunnecessary for parents to send theirchildren to work? The answer dependson the context.

It was argued in this paper that thereis much that can and ought to be done,but the precise policy to be followeddepends on the economic milieu forwhich the prescription is being sought.


The main policy divide is between legalinterventions and what may be calledcollaborative interventions, that is, pub-lic action which alters the economic en-vironment such that parents of theirown accord prefer to withdraw the chil-dren from the labor force. The avail-ability of good schools, the provisionof free meals, and efforts to bolsteradult wages are examples of collabora-tive interventions. We have discussedexamples and given arguments to showthat such interventions are, in general,a desirable way of curbing child labor.However, many of these actions maynot be feasible. There may not bemoney enough in the government’s cof-fers to run better schools or to improvethe infrastructure, which would resultin higher adult wages.

In such circumstances, should gov-ernment resort to legal action to restrictchild labor? There seems to be someagreement that some minimal restric-tions, such as children being preventedfrom working in hazardous occupationsor under bonded labor conditions, areworth enforcing legally. It is true thatone can always think of some circum-stance where even such a minimal lawwill work to the detriment of the child.But, by and large, children being madeto work in hazardous conditions iseither an act of child abuse or igno-rance on the part of the parents and thechild, and so, in general, it is better todeclare such action as illegal.

But what about a more general banon child labor per se? The evidence andthe theory that we studied in this papersuggest that there is no unconditionalanswer to this. There are circumstanceswhere, even if such a total ban werefeasible and costless to implement, itought not to be implemented. To un-derstand this one must realize thatthere are worse things that can happento children than having to work. In very

poor regions, the alternative to workmay be to suffer acute hunger or starva-tion. Indeed, when child labor occurs asa mass phenomenon, it is likely that thealternative to work is very harsh, be-cause even poor parents do not in gen-eral like to send their children to workif they can help it.

Curiously enough, despite this, thereare circumstances where a total banmay be desirable from the point of viewof the well-being of the children. This isbecause, whereas a single parent with-drawing the child from work cannot in-fluence equilibrium wages, a large scalewithdrawal of child labor can causeadult wages to rise so much that theworking class household is better off. Itwas argued in this paper that this is un-likely to be true for very poor econo-mies but may be valid for relatively bet-ter-off countries. Even so, one wouldneed to do detailed empirical work todecide whether such a total ban isworthwhile. The interesting insight thattheory gives us here is to tell us that itmay be so and to give some hints aboutthe type of economy where this is likely.

Another thing that we know is that, ifa ban is deemed desirable, a good wayto implement it is by making schoolingcompulsory. This is because a child’spresence in school is easier to monitorthan a child’s abstention from work. Itis true that schooling is compatible witha certain amount of work, since chil-dren can work before and after school;but it is a good way to prevent full-timework and of course is desirable in itself.It has, in fact, been argued by somethat, in very poor countries, we shouldmake it possible for children to combineschool with work, instead of thinking ofthese as mutually exclusive activities.

The case for a ban on child labor inthe export goods sector alone, whichwould be a natural concomitant of ef-fort in industrialized nations to boycott


the import of goods made with childlabor, is weaker, since this could resultin children being diverted to less desir-able or more hazardous work. In gen-eral, it is better to take economy-widemeasures against child labor and, ifthere is to be a sector-specific ban, thisshould be based on the working con-ditions of that sector, rather than thedestination of the goods.

This reservation carries over to cer-tain kinds of international action, suchas the imposition of minimal labor stan-dards as a prerequisite for trade, sincethis results in the maintenance of stan-dards only in the exports sector. In ad-dition, international labor standards cancause trade distortions by failing to rec-ognize the comparative advantages ofdifferent nations. The paper used someof the models of child labor to investi-gate the impact of international stan-dards. It was found that one beneficialeffect of such standards could be thehelp that they can provide to develop-ing nations to make a coordinated im-provement in their working conditions,without causing a flight of capital. Theidea of international labor standards asan instrument of collusion among devel-oping countries suggests news ways tothink about this problem. The discus-sion here was brief and preliminary andsimply points to directions for futureresearch.

REFERENCES

Addison, Tony; Sonia Bhalotra, Fiona Coulter, andChristopher Heady. 1997. “Child Labor in Paki-stan and Ghana: A Comparative Study,” mimeo,U. Warwick.

Agarwal, Bina. 1997. “‘Bargaining’ and Gender Re-lations: Within and Beyond the Household,”Feminist Econ., 3:1, pp. 1–51.

Anderson, Michael. 1971. Family Structure inNineteenth Century Lancashire. Cambridge:Cambridge U. Press.

Angrist, Joshua and Alan Krueger. 1991. “DoesCompulsory School Attendance Affect School-ing and Attendance?” Quart. J. Econ., 106:4,pp. 979–1014.

Ashagrie, Kebebew. 1993. “Statistics on Child La-

bor,” Bul. Labor Statist., 3, International LaborOrganization (ILO), Geneva.

———. 1998. “Statistics on Child Labor and Haz-ardous Child Labor in Brief,” mimeo, Bureau ofLabor Statistics, ILO, Geneva.

Bachman, Sarah L. 1995. “Children at Work,”Commentary, San Jose Mercury News, July 16.

Baland, Jean-Marie and James Robinson. 1998. “AModel of Child Labor,” mimeo, U. Namur, Bel-gium.

Banerjee, Abhijit and Andrew Newman. 1993.“Occupational Choice and the Process of Devel-opment,” J. Polit. Econ., 101:2, pp. 274–98.

Bardhan, Pranab and Christopher Udry. 1999. De-velopment Microeconomics. Oxford: Oxford U.Press.

Basu, Alaka Malwade. 1993. “Family Size andChild Welfare in an Urban Slum: Some Disad-vantages of Being Poor but Modern,” in Fertil-ity, Family Size and Structure. Cynthia B.Lloyd, ed. NY: Population Council.

Basu, Kaushik. 1994. “The Poor Need Child La-bor,” New York Times, Nov. 29.

———. 1998. “Social Norms and the Law,” in TheNew Palgrave Dictionary of Economics and theLaw. Peter Newman, ed. London: Macmillan.

Basu, Kaushik and James Foster. 1998. “On Mea-suring Literacy,” Econ. J., 108:451, pp. 1733–49.

Basu, Kaushik and Pham Hoang Van. 1998. “TheEconomics of Child Labor,” Amer. Econ. Rev.,88:3, pp. 412–27 .

Becker, Gary S. 1964. Human Capital. NY: Co-lumbia U. Press.

Besley, Timothy and Stephen Coate. 1992. “Un-derstanding Welfare Stigma: Taxpayer Resent-ment and Statistical Discrimination,” J. PublicEcon., 48:2, pp. 165–83.

Bhagwati, Jagdish. 1995. “Trade Liberalization and‘Fair Trade’ Demands: Addressing the Environ-ment and Labor Standards Issues,” WorldEcon., 18:6, pp. 745–59.

Bhatty, Kiran et al. 1997. “Class Struggle,” IndiaToday, 22, Oct. 13, pp. 69–73.

Bloom, David and Waseem Noor. 1996. “LaborStandards and the Emerging World Economy,”mimeo, Harvard U.

Bolin-Hort, Per. 1989. Work, Family, and theState: Child Labor and the Organization of Pro-duction in the British Cotton Industry, 1780–1920. Lund: Lund U. Press.

Bonnet, Michel. 1993. “Child Labor in Africa,”Int. Labor Rev., 132:3, pp. 371–89.

Bourguignon, Francois and Pierre-Andre Chiap-pori. 1994. “The Collective Approach to House-hold Behavior,” in The Measurement of House-hold Welfare. Richard Blundell, Ian Preston,and Ian Walker, eds. Cambridge, UK: Cam-bridge U. Press.

British Parliamentary Papers. 1968. “Report fromthe Select Committee [1831–32] on the ‘Bill toRegulate the Labor of Children in the Mills andFactories of the United Kingdom’,” IndustrialRevolution: Children’s Employment, vol. 2.Shannon: Irish U. Press.


Brown, Drusilla. 1998. “A Transactions CostPolitics Analysis of International Child LaborStandards,” mimeo, Tufts U.

Brown, Martin; Jens Christiansen, and PeterPhilips. 1992. “The Decline of Child Labor inthe US Fruit and Vegetable Canning Industry:Law or Economics,” Bus. Hist. Rev., 66:4, pp.723–70.

Browning, Martin; Francois Bourguignon, Pierre-André Chiappori, and Valérie Lechene. 1994.“Income and Outcomes: A Structural Model ofIntrahousehold Allocation,” J. Polit. Econ.,102:6, pp. 1067–96.

Burra, Neera. 1995. Born to Work: Child Labor inIndia. Delhi: Oxford U. Press.

Cain, Mead T. 1977. “The Economic Activities ofChildren in a Village in Bangladesh,” Popula-tion Devel. Rev., 3:3, pp. 201–27.

Cartwright, Kimberly and Harry A. Patrinos. 1998.“Child Labor in Urban Bolivia,” in The PolicyAnalysis of Child Labor: A Comparative Study.Christiaan Grootaert and Harry A. Patrinos,eds. Manuscript, World Bank.

Chandrasekhar, C.P. 1997. “The Economic Conse-quences of the Abolition of Child Labor,” J.Peasant Stud., 24:3, pp. 137–79.

Charnovitz, Steve. 1987. “The Influence of Inter-national Labour Standards on the World Trad-ing Regime: A Historical Review,” Int. Lab.Rev., 126:5, pp. 565–84.

Chaudhri, D. P. 1997. “A Policy Perspective onChild Labour in India with Pervasive Genderand Urban Bias in School Education,” Indian J.Labour Econ., 40:4, pp. 789–808.

Chwe, Michael Suk-Young. 1990. “Why WereWorkers Whipped? Pain in a Principal-AgentModel,” Econ. J., 100:403, pp. 1109–21.

Collingsworth, Terry; William J. Goold, and PharisJ. Harvey. 1994. “Time for a New Global Deal,”Foreign Aff., 73:1, pp. 8–13.

Cunningham, Hugh. 1990. “The Employment andUnemployment of Children in England c.1680–1851,” Past and Present, 126, pp. 115–50.

———. 1996. “Combating Child Labor: The Brit-ish Experience,” in Cunningham and Viazzo1996.

Cunningham, Hugh and Pier Paolo Viazzo, eds.1996. Child Labor in Historical Perspective,1800–1985: Case Studies from Europe, Japanand Colombia. Florence: UNICEF.

———. 1996a. “Some Issues in the HistoricalStudy of Child Labor,” in Child Labor in His-torical Perspective, 1800–1985: Case Studiesfrom Europe, Japan and Colombia. Hugh Cun-ningham and Pier Paolo Viazzo, eds. Florence:UNICEF.

Dasgupta, Partha. 1993. An Inquiry into Well- Being and Destitution. Oxford: Oxford U. Press.

De Herdt, Rene. 1996. “Child Labor in Belgium,1800–1914,” in Child Labor in Historical Per-spective, 1800–1985: Case Studies from Europe,Japan and Colombia. Hugh Cunningham andPier Paolo Viazzo, eds.

Dixit, Avinash. 1998. “Discussion of ‘A Transac-

tions Cost Politics Analysis of InternationalChild Labor Standards’,” mimeo, Princeton U.

Dreze, Jean and Haris Gazdar. 1996. “UttarPradesh: The Burden of Inertia,” in Indian De-velopment: Selected Regional Perspectives. JeanDreze and Amartya Sen, eds. Oxford: Oxford U.Press.

Ehrenberg, Ronald G. 1994. Labor Markets andIntegrating National Economies. Washington,DC: Brookings Institution.

Ellerman, David. 1992. Property and Contract inEconomics. Cambridge, MA: Blackwell.

Eswaran, Mukesh. 1996. “Fertility, Literacy andthe Institution of Child Labor,” mimeo, U. Brit-ish Columbia, Vancouver.

Fields, Gary S. 1995. “Labor Standards and Inter-national Trade,” mimeo, OECD, Paris.

Folbre, Nancy. 1986. “Hearts and Spades: Para-digms of Household Economics,” World Devel.,14:2, pp. 245–55.

Francois, Patrick. 1998. “Gender DiscriminationWithout Gender Difference: Theory and PolicyResponses,” J. Public Econ., 68:1, pp. 1–32.

Galbi, Douglas A. 1997. “Child Labor and the Di-vision of Labor in the Early English CottonMills,” J. Population Econ., 10:4, pp. 357–75.

Galor, Oded and Joseph Zeira. 1993. “Income Dis-tribution and Macroeconomics,” Rev. Econ.Stud., 60:1, pp. 35–52.

Genicot, Garance. 1998. “An Efficiency WageTheory of Child Labor: Exploring the Implica-tions of Some Ideas of Leibenstein and Marx.”Econ. Dept. Working Paper #463, Cornell U.

Glomm, Gerhard. 1997. “Parental Choice of Hu-man Capital Investment,” J. Devel. Econ., 53:1,pp. 99–114.

Goldin, Claudia D. 1979. “Household and MarketProduction of Families in a Late NineteenthCentury American Town,” Explor. Econ. Hist.,16:2, pp. 111–31.

Golub, Stephen S. 1997. “Are International LaborStandards Needed to Prevent Social Dump-ing?” Finance Develop., 34:4, pp. 20–25.

Granovetter, Mark. 1978. “Threshold Models ofCollective Behavior,” Amer. J. Soc., 83:6, pp.1420–43.

Grootaert, Christiaan. 1998. “Child Labor in Coted’Ivoire: Incidence and Determinants,” in ThePolicy Analysis of Child Labor: A ComparativeStudy. Christiaan Grootaert and Harry Patrinos,eds. Manuscript, World Bank, Washington.

Grootaert, Christiaan and Ravi Kanbur. 1995.“Child Labour: An Economic Perspective,” Int.Lab. Rev., 134:2, 187–203.

Gupta, Manash R. 1997. “Unemployment of AdultLabor and the Supply of Child Labor: ATheoretical Analysis,” mimeo, Jadavpur U., Cal-cutta.

———. Forthcoming. “Wage Determination of aChild Worker: A Theoretical Analysis,” Rev. De-vel. Econ.

Hadfield, Gillian K. 1995. “Rational Women: ATest for Sex-Based Harassment,” Cal. LawRev., 83, pp. 1151–89, October.


Harkin, Sen. Tom. 1994. “Now Act to Ban theProducts of Child Labor,” letter to the editor,New York Times, December 5.

Horan, Patrick M. and Peggy G. Hargis. 1991.“Children’s Work and Schooling in the LateNineteenth Century Family Economy,” Amer.Soc. Rev., 56:5, pp. 583–96.

Horrell, Sara and Jane Humphries. 1995. “‘TheExploitation of Little Children’: Child Laborand the Family Economy in the IndustrialRevolution,” Explor. Econ. Hist., 32:4, 485–516.

———. 1995a. “Women’s Labor Force Participa-tion and the Transition to the Male-Breadwin-ner Family, 1790–1865,” Econ. Hist. Rev., 48:1,pp. 89–117.

Hutchins, B.L. and A. Harrison. 1903. A Historyof Factory Legislation. London: P.S. King andSon. [References to 1966 ed. by London: Cass.]

ILO. 1995. Child Labor in Nepal: An Overviewand a Proposed Plan of Action. Geneva: ILO.

———. 1996. Child Labor: Targeting the Intoler-able. Geneva: ILO.

———. 1996a. Economically Active Populations:Estimates and Projections, 1950–2010. Geneva:ILO.

Jacoby, Hanan G. and Emmanuel Skoufias. 1997.“Risk, Financial Markets, and Human Capital ina Developing Country,” Rev. Econ. Stud., 64:3,pp. 311–35.

Jayaraj, D. and S. Subramanian. 1997. “Child La-bor in Tamilnadu: A Preliminary Account of itsNature, Extent and Distribution,” WP 151, Ma-dras Institute of Development Studies, Chen-nai.

Jensen, Peter and Helena Skyt Nielsen. 1997.“Child Labor or School Attendance? Evidencefrom Zambia,” J. Population Econ., 10:4, 407–24.

Knutsson, Karl Eric. 1997. Children: Noble Causesor Worthy Citizens? Aldershot, UK: AshgatePublishing.

Kothari, Smitu. 1983. “There’s Blood on thoseMatchsticks: Child Labor in Sivakasi,” Econ.Polit. Weekly, 18:27, pp. 1191–202.

Krueger, Alan. 1997. “International Labor Stan-dards and Trade,” in Proceedings Annual WorldBank Conference on Development Economics1996. Washington: World Bank.

Kruse, Douglas and Douglas Mahoney. 1998. “Ille-gal Child Labor in the United States: Preva-lence and Characteristics,” NBER, WP 6479.

Kulkarni, Manu. 1983. “Matchmaking Children ofSivakasi,” Econ. Polit. Weekly, 18:43, pp. 1855–56.

Labenne, Sophie. 1995. “Determinants of ChildLabor in India,” Mimeo, U. Namur, Belgium.

Levison, Deborah et al. Forthcoming. “Is ChildLabor Really Necessary in India’s Carpet Indus-try?” in Economics of Child Labor in SelectedIndustries of India. Richard Anker et al., eds.New Delhi: Hindustan Publishers.

Levy, Victor. 1985. “Cropping Pattern, Mechaniza-tion, Child Labor, and Fertility Behavior in aFarming Economy: Rural Egypt,” Econ. Devel.Cult. Change, 33:4, pp. 777–91.

Lindbeck, Assar; Stan Nyberg, and JorgenWeibull. 1999. “Social Norms and Economic In-centives in the Welfare State,” Quart. J. Econ.,114:1, pp. 1–35.

Ljungqvist, Lars. 1993. “Economic Underdevelop-ment: The Case of a Missing Market for HumanCapital,” J. Devel. Econ., 40:2, pp. 219–39.

Lyons, John. 1989. “Family Response to EconomicDecline: Handloom Weavers in Early Nine-teenth-Century Lancashire,” Res. Econ. Hist.,12, pp. 45–91.

Mackinnon, Mary and Paul Johnson. 1984. “TheCase Against Productive Whipping,” Explor.Econ. Hist., 21:2, pp. 218–23.

Maharatna, Arup. 1997. “Children’s Work Activi-ties, Surplus Labor and Fertility: Case Study ofSix Villages in Birbhum,” Econ. Polit. Weekly,32:7, pp. 363–69.

Manser, Marilyn and Murray Brown. 1980. “Mar-riage and Household Decision-Making: A Bar-gaining Analysis,” Int. Econ. Rev., 21:1, pp. 31–44.

Margo, Robert A. and T. Aldrich Finegan. 1996.“Compulsory Schooling Legislation and SchoolAttendance in Turn of the Century America,”Econ. Letters, 53:1, pp. 103–10.

Marshall, Alfred. 1920. Principles of Economics,8th ed. London: Macmillan [Original pub.1890].

Marx, Karl. 1867. Capital: A Critique of PoliticalEconomy. [References to 1954 ed., Moscow:Progress Publishers.]

———. 1875. Critique of the Gotha Pro-gramme.[References to 1938 English transla-tion in Marx and Frederick Engels, SelectedWorks, Vol. II. NY: International Publishers.]

Maskus, Keith E. and Jill A. Holman. 1996. “TheEconomics of Child Labor Standards,” mimeo,Econ. Dept., U. Colorado Boulder.

McElroy, Marjorie and Mary Jean Horney. 1981.“Nash-Bargained Household Decisions: To-wards a Generalization of the Theory of De-mand,” Int. Econ. Rev., 22:2, pp. 333–49.

McKelway, A. J. 1906. “Child Labor in the South-ern Cotton Mills,” Annals Amer. Acad. Polit.Soc. Sci., 27:2, pp. 1–11.

Mill, John Stuart. 1848. Principles of PoliticalEconomy. [References to 1970 ed., Har-mondsworth, UK: Penguin.]

Moehling, Carolyn M. 1995. “The IntrahouseholdAllocation of Resources and the Participation ofChildren in Household Decision-Making: Evi-dence from Early Twentieth Century America,”mimeo, Northwestern U.

———. 1998. “State Child Labor Laws and theDecline of Child Labor,” mimeo, Ohio State U.

Myers, William. 1988. “Alternative Services forStreet Children: The Brazilian Approach,” inCombating Child Labour. Assefa Bequele andJo Boyden, eds. Geneva: ILO.

Nardinelli, Clark. 1982. “Corporal Punishmentand Children’s Wages in Nineteenth-CenturyBritain,” Explor. Econ. Hist., 19:3, pp. 283–95.

———. 1984. “The Productivity of Corporal


Punishment: A Reply to Mackinnon andJohnson,” Explor. Econ. Hist., 21:2, pp. 224–28.

———. 1990. Child Labor and the IndustrialRevolution. Bloomington: Indiana U. Press.

Parsons, Donald O. and Claudia Goldin. 1989.“Parental Altruism and Self-Interest: Child La-bor Among Late Nineteenth-Century AmericanFamilies,” Econ. Inquiry, 27:4, pp. 637–59.

Patrinos, Harry A. and George Psacharopoulos.1997. “Family Size, Schooling and Child Laborin Peru,” J. Population Econ., 10:4, pp. 387–405.

Pigou, Arthur C. 1920. The Economics of Welfare.[References to 1962 ed. London: Macmillan.]

Psacharopoulos, George. 1997. “Child Labor Ver-sus Educational Attainment: Some Evidencefrom Latin America,” J. Population Econ., 10:4,pp. 377–86.

Ray, Ranjan. Forthcoming. “Analysis of Child La-bor in Peru and Pakistan: A ComparativeStudy,” J. Population Econ.

Riley, Nancy E. 1997. “Gender, Power, and Popu-lation Change,” Population Bul., 52:1, pp. 2–48.

Rodgers, Gerry and Guy Standing, eds. 1981.Child Work, Poverty, and Underdevelopment.Geneva: ILO.

Rodrik, Dani. 1996. “Labor Standards in Interna-tional Trade: Do They Matter and What Do WeDo about Them?” Policy Essay No. 20, Over-seas Development Council, Washington, DC.

———. 1997. Has Globalization Gone Too Far?Washington, DC: Institute for Int. Economics.

Rosenzweig, Mark R. and Robert Evenson. 1977.“Fertility, Schooling and the Economic Contri-bution of Children in Rural India: An Econo-metric Analysis,” Econometrica, 45:5, pp. 1065–79.

Saito, Osamu. 1996. “Children’s Work, Industriali-zation and the Family Economy in Japan, 1872–1926,” in Child Labor in Historical Perspective,1800–1985: Case Studies from Europe, Japanand Colombia. Hugh Cunningham and PierPaolo Viazzo, eds. Florence: UNICEF.

Scholliers, Peter. 1995. “Grown-ups, Boys andGirls in the Ghent Cotton Industry: The Voort-man Mills, 1835–1914,” Social Hist., 20:2, pp.201–18.

Sen, Amartya K. 1990. “Gender and CooperativeConflict,” in Persistent Inequalities. IreneTinker, ed. NY: Oxford U. Press.

Sengenberger, Werner and Duncan Campbell,eds. 1994. International Labor Standards andEconomic Interdependence. Geneva: Int. Insti-tute for Labor Studies.

Silvers, Jonathan. 1996. “Child Labor in Pakistan,”The Atlantic Monthly, February, pp. 79–92.

Srinivasan, T. N. 1996. “International Trade andLabor Standards from an Economic Perspec-tive,” in Challenges to the New World TradeOrganization. Pitou van Dijck and Gerrit Fa-ber, eds. The Hague: Kluwer Law Interna-tional.

Strauss, John and Duncan Thomas. 1995. “HumanResources: Empirical Modeling of Householdand Family Decisions,” in Handbook of Devel-opment Economics, Vol. IIIA. Jere Behrmanand T. N. Srinivasan, eds. Amsterdam: Elsevier.

Swaminathan, Madhura. 1997. “Do Child WorkersAcquire Specialized Skills? A Case Study ofTeenage Workers in Bhavnagar,” Indian J. La-bor Econ., 40:4, pp. 829–39.

Swinnerton, Kenneth and Carol Ann Rogers.Forthcoming. “The Economics of Child Labor:Comment.” Amer. Econ. Rev.

Thomas, Duncan. 1990. “Intra-Household Re-source Allocation: An Inferential Approach,” J.Human Res., 25:4, pp. 635–64.

Trebilcock, Michael. 1993. The Limits of Freedomof Contract. Cambridge, MA: Harvard U. Press.

Udry, Christopher. 1996. “Gender, AgriculturalProduction and the Theory of the Household,”J. Polit. Econ., 104:5, pp. 1010–46.

UNICEF. 1994. Children at Work. Bangkok:UNICEF East Asia and Pacific Regional Office.

US Department of Labor. 1993. Foreign LaborTrends: International Child Labor Problems.Washington, DC: Bureau Int. Labor Affairs.

Vincent, David. 1981. Bread, Knowledge and Free-dom: A Study of Nineteenth-Century WorkingClass Autobiography. London: Europa Publica-tions.

Weiner, Myron. 1991. The Child and the State inIndia: Child Labor and Education Policy inComparative Perspective. Princeton: PrincetonU. Press.

———. 1996. “Child Labor in India: Putting Com-pulsory Primary Education on the PoliticalAgenda,” Econ. Polit. Weekly, 31:45–46, pp.3007–14.

Wilkinson, Frank. 1994. “Equality, Efficiency andEconomic Progress: The Case for UniversallyApplied Equitable Standards for Wages andConditions of Work,” in International LaborStandards and Economic Interdependence.Werner Sengenberger and Duncan Campbell,eds. Geneva: International Institute for LaborStudies.


child labor: cause, consequence, and cure, with remarks on

Documents