The impact of FDI on child labor: Insights from an empirical analysis of sectoral FDI data and case studies

Children and Youth Services Review 47 (2014) 157–167

Contents lists available at ScienceDirect

Children and Youth Services Review

j ourna l homepage: www.e lsev ie r .com/ locate /ch i ldyouth

The impact of FDI on child labor: Insights from an empirical analysis ofsectoral FDI data and case studies☆

Nadia Doytch a,c,⁎, Nina Thelen b, Ronald U. Mendoza c

a City University of New York Brooklyn College, School of Business, Department of Economics, 217 Whitehead Hall, 2900 Bedford Avenue, Brooklyn, NY 11210, United Statesb United Nations Development Programme (UNDP), Bureau for Development Policy, United Statesc Asian Institute of Management (AIM), 123 Paseo de Roxas, Makati City 1229, Philippines

☆ We would like to thank the participants at the EasteYork, NY Feb. 25–27, 2011 and the Asian Institute oSeminar, Manila, The Philippines, Jan. 17, 2012 participacomments. We are also thankful to Christopher Milde for⁎ Corresponding author at: City University of New Yo

Business, Department of Economics, 217 WhiteheadBrooklyn, NY 11210, United States. Tel.: +1 718 951 5000

E-mail addresses: [email protected] (N. Do(N. Thelen), [email protected] (R.U. Mendoza).

URL:E-mail addresses:E-mail address: http://policy.aim1 See UNICEF Child Info Website (http://www.childin

drawn from DHS, MICS and other national surveys, 2000–

http://dx.doi.org/10.1016/j.childyouth.2014.09.0080190-7409/© 2014 Elsevier Ltd. All rights reserved.

a b s t r a c t
a r t i c l e i n f o
Article history:Received 29 January 2014Received in revised form 19 August 2014Accepted 15 September 2014Available online 8 October 2014

JEL classification:E24F15O15

Keywords:FDIChild laborIncome and substitution effectUN Convention on the Rights of the Child

Not all foreigndirect investment (FDI) is alike as far as its impact on various dimensions of humandevelopment isconcerned. This paper focuses, in particular, on child labor and it undertakes a cross-country empirical analysis ofthis issue, using data on100 countries spanning the period 1990–2009. Unlike earlier studies that focusmostly ontotal FDI, we also utilize data on disaggregated FDI, covering the main economic sectors of interest such asagriculture, mining, manufacturing, services, and finance. The empirical results suggest that different economicsectors generate varied effects on child labor. For instance, FDI in agriculture in Europe and Central Asia tendsto exacerbate child labor, whereas FDI in manufacturing in South and East Asia and FDI in mining in LatinAmerica appear negatively linked to child labor. Furthermore, signing on to the UN Convention on the Rightsof the Child (CRC) is positively associated with child labor. One possible explanation for the latter result is thatstronger anti-child labor laws could lead to multiple equilibria in labor markets, including the possibility ofincreasing child labor in certain sectors. Selected case studies help clarify the possible reasons behind this variedFDI impact on child labor, emphasizing among other factors supply chain management and the criticalimportance of policy implementation and coordination with the private sector.

© 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Protecting children and eradicating child labor are among the princi-pal objectives of over 190 governments that are signatories to the UNConvention on the Rights of the Child (CRC). Nevertheless, one outof every six children aged 5–14 are engaged in child labor in the devel-opingworld.1 Child labor results from a complex set of factors, and it hasbeen the subject of a growing body of literature examining the reasonsbehind it, as well as its various types and impacts (Edmonds, 2008).Empirical studies have also begun to examine the impact of increasedeconomic openness on the poor, and in particular, children. A handfulof these studies focus in particular on the impact of trade and foreign di-rect investment (FDI) on child labor. The inflow of FDI could help boost

rn Economic Association, Newf Management Policy Centernts for useful suggestions andhis helpful comments.rk Brooklyn College, School ofHall, 2900 Bedford Avenue,/2644; fax: +1 718 951 4384.ytch), [email protected]

.edu (R.U. Mendoza).fo.org/labour.html). Data was2010.

output and lead to an increase in income (an income effect), but also arise in demand for low wage labor, raising the relative returns fromchild labor (a substitution effect). Since these factors influence childlabor in opposite ways, the net effect is an area for empirical analysis.

This paper contributes to the literature in this area by undertaking across-country empirical analysis of the impact of FDI (disaggregated bysectors) on the incidence of child labor, using data on 100 countriesspanning the period 1990–2009. In addition to the more extensivescope of data, there are several other main innovations in this paperthat improve on earlier studies. First, in addition to total FDI, we alsoutilize data on disaggregated FDI, covering main economic sectors ofinterest such as agriculture, mining, manufacturing, services, and fi-nance. Earlier literature in this area has focused extensively on aggre-gate indicators, which leaves out much of the nuances across sectors.Child labor across sectors varies widely. We hypothesize that FDI acrossdifferent economic sectors would generate different outcomes —

including notably income and substitution effects — vis-à-vis childlabor. Second, and acknowledging the challenges faced by previousresearch, we also address the simultaneity issues arising in empiricalanalyses of FDI, income and child labor by utilizing a GMM estimationmethodology. This approach is possible because of the extensive paneldataset that this study utilizes. Finally, this study also uses an indicatorto test for a possible structural break in the model between the periodprior to the country signing on to the CRC and the period afterwards.

http://crossmark.crossref.org/dialog/?doi=10.1016/j.childyouth.2014.09.008&domain=pdf

http://dx.doi.org/10.1016/j.childyouth.2014.09.008

mailto:[email protected]



http://policy.aim.edu

http://www.childinfo.org/labour.html

Unlabelled image

http://dx.doi.org/10.1016/j.childyouth.2014.09.008

Unlabelled image

http://www.sciencedirect.com/science/journal/01907409

158 N. Doytch et al. / Children and Youth Services Review 47 (2014) 157–167

For many countries, this marked the beginning of stronger child protec-tion policies and institutional reforms to promote the welfare of chil-dren. We therefore expect that the period with the CRC in effectwould be associated with a reduction in child labor. To the best of ourknowledge, no empirical study has measured the possible impact onchild labor of signing on to the CRC.

The empirical results suggest that the aggregate indicator for FDI isnot linked to child labor, contradicting earlier research that found a neg-ative link. We interpret the difference in finding to be due to the im-provement in the empirical methodology in this study. Furthermore,when using the disaggregated indicators for FDI, the results show thatvarious sectors and regions are associated with child labor in differentways. For instance, we find evidence that FDI in agriculture in Europeand Central Asia seems to exacerbate child labor, whereas FDI inmanufacturing in South and East Asia and FDI in mining in LatinAmerica appear to be negatively linked to child labor.

A brief review of different cases suggests that further analysis of themicrofoundations of the FDI-child labor link is a rich area for future re-search, shedding further light on the broad linkages empirically exam-ined in this paper. Finally, the empirical results also suggest thatsigning on to the CRC is positively linked to child labor.We draw on ear-lier research by Basu (1999, 2001, 2005) to provide a possible explana-tion for this unintended effect.

In what follows, Section 1 briefly reviews the empirical literature onFDI and child labor. Section 2 discusses the data and empirical method-ology and Section 3 briefly presents the empirical results. Section 4draws on various case studies to shed light on the possible explanationsbehind the broad empirical relationships examined here. A brief con-cluding section reviews areas for future work.

2. Related literature

Myriad factors, spanning cultural, economic, political, social andothers, could help explain the child labor phenomenon.2 Our main at-tention in this paper is on one set of economic factors, with a specificfocus on FDI. Theoretically, child labor could be influenced by economicopenness, covering trade, FDI and other aspects, because of the relatedincome and substitution effects. First, increased economic opennesscould help to raise income by increasing productivity, creating jobsand boosting growth. This in turn could lead to lower incidence ofchild labor among the poor, as households no longer need to rely onchildren to help boost incomes above subsistence, while parents alsomay start to increase investment in their children's education andhealth (Basu & Van, 1998).3 Second, economic openness could alsostimulate the demand for cheap labor, which may also create a knock-on effect on child labor to the extent that children could performsome adult jobs. An indirect effect is also possible if the parents of

2 For a recent review of the empirical and theoretical literature, see Edmonds (2008). Inaddition, Basu (1999, 2001, 2005) provides an extensive elaboration of the pathology be-hind child labor. Basu and Van (1998) and Kitaura (2009) develop theoretical frameworksto explain the links across policy, wages and the possible income and substitution effectson child labor. The literature on the child labor phenomenon also reflects some disagree-ments, notably in what constitutes child labor. We acknowledge this as part of the policychallenge in this area. Nevertheless, we attempt to focus on approaches and definitionsthat are consistent and relevant for intergovernmental policies to advance the rights ofchildren, particularly in the context of the CRC.

3 The literature on cash transfers also lends support to the hypothesis that increasedhousehold income among the poor may help to reduce child labor incidence and supportfurther investment in education. For instance, Edmonds and Schady (2009) examine theimpact of cash transfers on child labor in Ecuador. Randomly selected poor mothers re-ceived cash transfers amounting to about 7% of monthly expenditures. Poor families re-ceiving the transfer delayed their children's entry into the labor force. Students inbeneficiary households reduced their involvement inpaid employment by 78% andunpaideconomic activity inside their home by 32%.

children are drawn to take on more work hours, and they pass ontheir household and family enterprise work to their children.

For instance, Webbink, Smits, and de Jong (2012) examine the con-ditions of children's housework and family business work using data onover 150,000 children from 180 districts in 13 developing countries inAfrica and Asia (i.e. Ivory Coast, The Gambia, Ghana, Guinea Bissau,Sierra Leone, Togo, Malawi, Somalia, Syria, Yemen, Thailand, Vietnamand Bangladesh). Their study suggests that most children spend sometime on household chores, but in some countries, the majority of girlsspend well over 10 h a week on household work. More than half ofthe girls and about 30% of the boys in Somalia, for example, spendover 20 h a week on household work. These figures suggest that theseforms of child labor may be much more significant than commercialwork.

Aspects of economic openness — such as international trade orforeign investment inflows — could create economic opportunities forchildren and their families that would not otherwise have existed. Forinstance, firms in the export sector could transact with subcontractorsfrom the informal sector in the value chain, in turn boosting the demandfor unskilledworkers and creating opportunities for utilizing child labor.Essentially, the relative returns to child labor increase (due to higherequilibrium child wages), prompting a possible substitution to workin lieu of play, leisure and education. These income and substitution ef-fects could work in opposite directions, making the net final impact aquestion for empirical verification.

The empirical evidence on economic openness and child labor pro-vides some evidence of a negative relationship between indicators oftrade and FDI and child labor, suggesting that, contrary to argumentsthat globalization could increase the risk of child labor, in fact the latteris mitigated. For instance, Cigno, Rosati, and Guarcello (2002) utilizedata on all developing countries for the years 1980, 1990, 1995 and1998, and examine cross-country trade openness indicators — thetrade ratio (exports plus imports scaled by GDP) and the Sachs andWarner openness indicator4— and find that these are negatively relatedto proxy measures of child labor (e.g. child labor is alternatively mea-sured by the 10–14 labor participation rate, or by the primary schoolnon-attendance rate) in most model specifications that control alsofor skill composition. The variable for “skill composition” is proxied bythe share of the workforce aged 25 or over which completed only pri-mary education, as well as the additional group that completed second-ary or higher. Oneway to interpret this variable is that it controls for thecapability of the labor force to participate more successfully in theopportunities brought about by economic openness.

These authors do not consider this relationship a causal one. Insteadthey interpret the results as having provided some empirical supportthat trade exposure per se does not seem to be associated with in-creased child labor incidence. They further argue that trade opennesscould be resulting in higher skill premiums in countries with an abun-dance of educated workers; and that countries with a large number ofeducatedworkers are better able to integrate and compete in globalizedmarkets. The principal challenge with this study, however, is that thepossible endogeneity between trade and child labor has not been ade-quately addressed. This in turn led to a possible bias in the results astrade could influence child labor, whereas child labor could also createfeedback effects on trade.

In addition, Neumayer and de Soysa (2004) are among the first toempirically analyze the possible link between FDI and child labor.They study the relationship between indicators of economic openness,including trade and FDI, on child labor, using cross-sectional data for

4 Sachs andWarner (2001) considered countries “closed” if at least one of the followingconditions applied to the country: 1) average tariffs higher than 40%; 2) nontariff barrierscovering more than 40% of imports; 3) socialist economic system; 4) state monopoly ofmajor exports, and 5) black market premium in excess of 20%.

010

2030

40%

1960 1970 1980 1990 2000year

Latin America and the CaribbeanWorld

Europe and Central Asia South and East Asia and the Pacific

Sub-Saharan Africa

Labor force, children 10-14 (% of age group)

Fig. 1. Child labor rates by region (1960–2003).

159N. Doytch et al. / Children and Youth Services Review 47 (2014) 157–167

about 117 countries with indicators drawn from averaged figures in the1990s. In order to deal with endogeneity, they utilize instruments thatwould affect trade and the FDI penetration rate without affecting childlabor incidence: demographic, geographical and language instruments(e.g. population size, size of land area, a dummy for countries that arelandlocked, the minimum distance to New York, Brussels or Tokyo,and a dummy variable for countries, share of the same language witha developed country). They find evidence that trade and FDI penetrationare both negatively associated with child labor, suggesting that the in-come effect dominates the substitution effect.

Similarly, Davies and Voy (2009) empirically analyze the relation-ship between FDI and child labor, utilizing data for 145 countries forthe year 1995. In order to address the endogeneity between FDI andchild labor, since countries with high incidence of child labor also typi-cally have a low-skilled labor force, and this in turn is not necessarily astrong attraction for FDI, the authors utilize instrumental variablesdeveloped in earlier studies based on geographical determinants(Edmonds & Pavcnik, 2006; Frankel & Romer, 1999; Frankel & Rose,2002). These instruments would influence FDI but not necessarily theskill level or child labor incidence in the country. These authors find ev-idence that FDI is statistically negatively linked to child labor incidence,but the statistical significance of this link disappears when the incomelevel is included. Davies and Voy argue that this is evidence that themain effect of FDI on child labor runs through the income channel.5

In all, the findings in the empirical literature do not suggest that eco-nomic openness — proxied by increased trade and FDI — tends to in-crease child labor. It is important to consider the limitations of theseearlier studies which are dominated by cross-sectional data analyses(and therefore fail to cover variance over time), and despitemore recentefforts many do not adequately address issues such as endogeneity inthe variables of interest (and therefore could result in biased estimates).Earlier studies have also not disaggregated FDI according to differentsectors in the economy, thus failing to account for sector specific effectswhich may vary. These are among the main issues addressed by the in-novations in this paper, to which we turn to next.

100

200

300

400

Bill

ion

of U

SD

3. Data and methodology

3.1. Stylized facts

Child labor rates have been decreasing in all world regions since the1960s (see Fig. 1). According to theWorld Bank data6 in Eastern Europeand Central Asia some child labor was still in existence in the 1960s— a7% incidence. The rate was reduced in the next 5 decades down to 1%—

an average decrease of 80% for the entire period.In the Latin America and the Caribbean region the child labor rates

have declined from 13% to 6% in this period — a rate of decline of 53%in the period 1960–2003.7 And in Sub-Saharan Africa and South andEast Asia and the Pacific regions the declines have been respectively40% to 27%— an average rate of decrease of 33% and from 26% incidenceof child labor to 17%— a rate of decrease of 35%. For theworld as awholethe decline in child labor rates has been from 21.5% to 16.5%— an aver-age rate of decline of 33% for this period.

At the same time, aggregate FDI inflows have been increasing since1990s in all of the examined regions (see Fig. 2), with the world FDIflows almost doubling in the period 1990–2010. The sectoral FDI flowsto the examined regions, however, show more variability.

5 Similarly, Edmonds and Pavcnik (2005) examined the link of trade on child labor (i.e.the labor force participation of children 10–14) and found that openness is negatively as-sociated with child labor only if income is not included in the model specification. Theseauthors also concluded that the main effect of trade on child labor runs through the in-come channel.

6 World Development Indicators CDROM2005.7 The data set after 2003 becomes unbalanced.

3.2. Conceptual framework

Following the literature, wemodel the child labor rates as a functionof GDP per capita, quality of institutions, population density and the keyexplanatory variable FDI at the sector level. We also include a CRCdummy variable.

Log CLitð Þ ¼ Log CLit−1ð Þ þ Incomeit þ Institutionsit þ Densityitþ FDIit þ ρit þ εitε

� i:i:d 0;σð Þ ð1Þ

We hypothesize that income levels would have a negative associa-tion with incidence of child labor, since fewer families would face theneed to choose work for their children instead of schooling. In addition,with higher income levels average families have better ability to paytheir children's education. We expect that quality of institutions couldinfluence the incidence of child labor indirectly, as weaker institutions(e.g. the presence of corruption) could lead firms to hire more children.We expect that population density has a negative effect on child labor.We also expect that the CRC dummy ρit could be either positive or neg-ative, depending on multiple equilibria as we further explain.

0

1970 1980 1990 2000 2010year

South and East Asia and the Pacific

Sub-Saharan Africa

Aggregate FDI Net Inflows

Eastern Europe and Central Asia

Latin america and the Caribbean

Fig. 2. Aggregate FDI inflows by region.

image of Fig.�1

image of Fig.�2


3.3. Empirical model

The main econometric model utilized in this study is as follows:

log CLitð Þ ¼ β0 þ β1 CLit−1ð Þ þ β2 log yitð Þ þ β3anticorr

þ β4 log densð Þ þ β5 fjit þ β6ρit þ β7D

t þ μ i þ εitμ i

� i:i:d 0;σμ i

� �εit � i:i:d 0;σεð Þ; E μ iεit½ � ¼ 0 ð2Þ

where i = 1, …, 100 and t = 1, …, 20.The variables μi and Dt are, respectively, a country-specific effect and

a time-specific effect represented by year dummies.

3.4. Data

The sources of data include: World Development Indicators (WDI)and International Labor Organization (ILO) for child labor rates; WorldDevelopment Indicators for income and demography data; Internation-al Country Risk Guide (ICRG) for institutions' quality data, where thevariable of choice is Anticorruption; andOECD, UNCTAD, national statis-tical agencies websites for sectoral FDI flows. Below are the definitionsof the key variables.

• Child labor or children 10–14 in the labor force is the share of that agegroup active in the labor force. Labor force comprises all people whomeet the International Labor Organization's definition of the econom-ically active population. The range for the child labor rates' data is1990–2009.

• Income is GDP per capita in constant 2005 prices. The range for theincome per capita data is 1990–2009.

• Population density is the people per sq km. The range for the popula-tion density data is 1990–2009.

• Anticorruption is defined by ICRG as ameasure of control of corruptionwithin the political system. Such corruption distorts the economic andfinancial environment and reduces the efficiency of government andbusiness. Actual or potential corruption may take the form of exces-sive patronage, nepotism, job reservations, ‘favor-for-favors’, secretparty funding, and suspiciously close ties between politics and busi-ness. The range of the “control of corruption” data is 1990–2009.

• f itj — Sectoral FDI flows. The superscript j refers to the different sectors:agriculture; mining and quarrying; manufacturing; and services FDI.Inward FDI is defined as net inflows, accounting for the purchasesand sales of domestic assets by foreigners in the corresponding year.

Table 1Regression results for all countries in the sample (up to total 88).a

Variables Total FDI/GDP Agricultural FDI/GD

Log of lagged working children 10–14 1.029***(49.28)

1.059***(92.16)

GDP per capita in 2000 USD 4.77e−06(0.91)

.00001***(3.47)

FDI/GDP variable .037(1.09)

2.199(1.61)

Control of corruption − .010*(−1.87)

− .004(−0.72)

Log of population density − .004(−0.70)

− .019**(−2.29)

Child rights convention dummy .080*(1.78)

.031(0.94)

Constant − .208**(−1.96)

− .091(−1.52)

Observations 1470 293Countries 88 37AR(2) test 0.959 0.555

a Figures in parentheses are z-statistics. The coefficients and the z-statistics are robust to heteconstructed for the lagged level of FDI, GDP per capita, and the respective institutional variable

The sectoral distribution is based on definitions in the InternationalStandard Industrial Classification (ISIC), revision 3, divisions 15–37.Services correspond to ISIC divisions 50–99. Services include valueadded in wholesale and retail trade (including hotels and restau-rants), transport, and government, financial, professional, and person-al services such as education, health care, and real estate services. Thedata availability of the sectoral FDI data is presented in an additionaltable (Table 6).

• UN Convention on the Rights of the Child (CRC) is a dummy variable,reflecting whether the country has signed the CRC at the time whenthe child labor data was reported. The range of the dummy variableis 1990–2009.

3.5. Empirical methodology

The method of fixed effects is designed to control for the unobservedcountry-specific time-invariant effects in the data. However, two tech-nical consequences of the within transformation are that the methodis not informativewhenwe dealwith variables with little time variationand that it does not address the problem of endogeneity. For these rea-sons, we choose the method of Blundell–Bond System GMM.

The Blundell–Bond GMM methodology handles this issue byallowing us to create instruments for the key explanatory variable —

FDI/GDP, any other explanatory variables, such as real per capita GDP,in addition to an automatically created instruments for the lagged de-pendent variable, which necessarily suffers from endogeneity. Thus, inthe case of our regressions, instruments have been created for allthree variables listed above — FDI/GDP; real per capita GDP and laggedlevel of child labor rates. In all three cases the instruments are struc-tured as matrices stacked with both the lagged levels and lagged differ-ences of the instrumented variable. The number of lags is limited tothree by the use of an option in the code, which is run in Stata. Sincethe instruments are created out of lags, the instrumental matrices areexogenous.

The Blundell–Bond System GMM methodology requires a set ofconditions to be met. First, even if the unobserved country-specificeffect is correlated with the regressors' levels, it is not correlated withtheir differences. The condition also means that the deviations of theinitial values of the independent variables from their long-run valuesare not systematically related to the country-specific effects. Thesesets of conditions can be written as follows.

P Mining FDI/DP Manufacturing FDI/GDP Services FDI/GDP

1.065***(80.28)

1.056***(46.74)

1.055***(62.19)

.00001***(3.18)

9.25e−06(1.51)

9.55e−06*(1.91)

−1.523(−1.64)

−1.033(−0.78)

− .242(−0.45)

− .010(−1.01)

− .006(−0.69)

− .007(−0.85)

− .015*(−1.74)

− .015**(−2.12)

− .012(−1.62)

.030(1.17)

.064*(1.79)

.064*(1.81)

− .184***(−2.62)

− .202(−2.67)

− .208**(−2.46)

333 407 40538 43 430.552 0.880 0.754

roscedasticity and obtained from one-step Blundell–Bond SystemGMMwith instruments. ***, ** and * denote significance at 1%, 5%, and 10%, respectively.

Table 2Regression results for Latin America and Caribbean countries (up to total 23).a

Variables Total FDI/GDP Agricultural FDI/GDP Mining FDI/DP Manufacturing FDI/GDP Services FDI/GDP


1.057***(68.09)

1.062***(77.81)

1.064***(81.99)

1.067***(78.35)


4.76e−06(0.45)

−9.27e−07(−0.13)

2.21e−06(0.33)

9.19e−07(0.13)

FDI/GDP variable − .371***(−2.87)

1.819(0.65)

−1.997**(−2.01)

−2.009(−1.38)

− .077(−0.13)

Control of corruption .007(1.03)

.010(1.10)

.009(0.81)

.009(1.15)

.003(0.38)


− .035(−1.59)

− .024*(−1.86)

− .013(−1.29)

− .013(−1.12)

Child rights convention dummy − .041(−1.43)

.172***(2.92)

.043(0.38)

.045***(2.88)

.033***(2.71)

Constant − .132(−1.16)

− .145(−1.30)

− .189*(−1.91)

− .247**(−2.48)

− .228**(−2.56)

Observations 411 144 167 207 207Countries 23 14 15 17 17AR(2) test 0.590 0.392 0.679 0.579 0.634

a Figures in parentheses are z-statistics. The coefficients and the z-statistics are robust to heteroscedasticity and obtained from one-step Blundell–Bond SystemGMMwith instrumentsconstructed for the lagged level of FDI, GDP per capita, and the respective institutional variable. ***, ** and * denote significance at 1%, 5%, and 10%, respectively.

Table 3Regression results for Europe and Central Asia countries (up to total 10).a



1.174***(412.00)

1.157***(617.07)

1.156***(66.26)

1.146***(178.96)

GDP per capita in 2000 USD 8.45e−06***(3.23)

.00002***(31.45)

.00002***(73.66)

.00003***(8.29)

.00003***(15.39)

FDI/GDP variable − .730(−1.31)

107.221***(10.42)

−4.485(−1.31)

2.333(0.68)

− .350(−1.01)

Control of corruption − .007(−0.48)

.021***(5.23)

.023***(15.78)

.022***(2.82)

.023***(3.72)

Log of population density − .012**(−2.27)

.242***(11.61)

.238***(29.14)

.034(1.11)

.036(1.19)

Child rights convention dummy .105**(2.01)

.019***(7.10)

.264***(52.84)

.072***(8.90)

.072***(9.81)

Constant − .165***(−2.84)

−1.682***(−15.88)

−1.862***(−45.69)

− .836***(−5.81)

− .804***(−6.25)




(i) Country-specific effects are not correlated with first-differenceddependent or independent variables8:

E CLi;t−1−CLi;t−2

� �μ i þ εitð Þ

h i¼ 0

E yi;t−1−yi;t−2


h i¼ 0

E f ji;t−1− f ji;t−2


h i¼ 0:

(ii) And the standard “no second order autocorrelation” in the errorterm conditions9:

E CLi;t−s εit−εi;t−1

� �h i¼ 0 for s≥2 and t ¼ 3;…T

8 The country specific term μi is estimated by the software in the background as well asthe validity of this set of conditions.

9 The AR(2) statistic is reported in each regression result table.

E yi;t−s εit−εi;t−1

� �h i¼ 0 for s≥2 and t ¼ 3;…T

E f ji;t−s εit−εi;t−1

� �h i¼ 0 for s≥2 and t ¼ 3;…T:

4. Empirical results and analysis

Perhaps due to the innovations in the data and methodology, wefind empirical results that run counter to earlier studies. The coefficientof the total FDI indicator was positive and not statistically significant,which runs counter to earlier findings of a negative and statistically sig-nificant link between FDI and child labor (Table 1). It is possible that thisis due to the empirical methodology in this paper, i.e. the GMM estima-tor, which allows for a more robust treatment of the endogeneity of FDIand child labor. As noted earlier in Section 2, a key drawback of earlierstudies had to do with the failure to address this issue, leading to possi-bly biased estimates.

As expected, the disaggregated FDI indicators revealed varyingresults (see Tables 1–5), depending on the sector and the countries

Table 4Regression results for South and East Asia countries (up to total 12).a



1.049***(70.34)

1.044***(106.62)

1.049***(126.86)

1.052***(137.76)

GDP per capita in 2000 USD − .00001(−0.81)

− .00001(−1.23)

−3.21e−06(−0.39)

−8.17e−06(−1.08)

−6.54e−06(−0.94)

FDI/GDP variable − .222(−0.85)

4.180**(2.49)

− .260(−0.13)

− .833**(−2.06)

−1.487*(−1.85)

Control of corruption .004(0.62)

.007(0.74)

.0005(0.14)

.011*(1.88)

.012*(2.33)


− .011(−0.38)

− .017(−1.34)

− .023**(−2.23)

− .019**(−2.02)

Child rights convention dummy − .020*(−1.95)

.031**(2.30)

.023**(2.09)

.014(1.33)

.011*(1.65)

Constant − .021(−0.33)

− .152(−0.78)

− .018(−0.18)

− .007(−0.11)

− .002(−0.04)




(regions) examined.When all countries are considered together, we arenot able to isolate any significant effects of FDI on child labor (Table 1).In the Latin America and Caribbean regions, there is evidence that min-ing FDI is negatively associated with child labor (Table 2). This could bepotentially attributed to anoverall economic development that spurs upfrom the mining sector. Such development may improve the economicsituation of children by allowingmore families to send their children toschool rather than engage them in contributing to the family budget.

The agricultural sector is a sector where children traditionally weremore engaged in helping their families. In Europe and Central Asiachild labor is positively linked to agricultural FDI (Table 3). Agriculturalwork in this region has seasonal frequency and older children tend to beengaged in helping to collect agricultural produce mostly in summer.Thus, an increase of incidence of child labor in Europe and Central Asiadoes not necessarily result in a decrease of school attendance in this re-gion (Table 3).

The situationwith child labor in South and East Asia is similar, whereincidence of child labor increaseswithmore FDI in the agriculture sector(Table 4). However, child labor tends to decrease with increases of FDIin manufacturing and services (Table 4). Manufacturing and servicesFDI in this region has been known to contribute to economic growth

Table 5Regression results for African countries (up to total 31).a

Variables Total FDI/GDP Agricultural FDI/GD


1.026***(97.49)


.00001(0.44)

FDI/GDP variable .024(1.29)

.615**(2.36)

Control of corruption − .001(−0.76)

.002(0.47)


− .002(−1.34)

Child rights convention dummyb − .002(−0.36)

Constant − .226(−2.59)

− .104**(−2.40)

Observations 570 24Countries 31 6AR(2) test 0.256 0.996

a Figures in parentheses are z-statistics. The coefficients and the z-statistics are robust to heteconstructed for the lagged level of FDI, GDP per capita, and the respective institutional variable

b The child rights convention dummy dropped in the case of regressions with agricultural, m

(Doytch & Uctum, 2011). As such these sectors are indirectly expectedto contribute to schooling in this region as well.

In Sub-SaharanAfrica,where the situation of children is known to bevery difficult, with high rates of infant mortality, low rates of schoolenrolment and high rates of children becoming victims of military con-flicts, the overall rates of child labor in all sectors are also quite high(Edmonds, 2008). The negative relation between services FDI andchild labor could be explained with the role of financial FDI in economicdevelopment. Financial FDI, as part of financial development, has beenproven an important driver of economic growth (Table 5). Availabilityand access to credit can provide entrepreneurial opportunities, includ-ing for micro, small and medium enterprises and ultimately providean alternative for some families that would otherwise have no otherchoice but sending their children to work from an early age.

These context-varying empirical relationships are to be expected inlight of the more recent FDI literature that stresses their varied impacton different sectors and on different groups in society (e.g. Lipsey &Sjöholm, 2005; Te Velde, 2003). Among the key channels throughwhich the impact of FDI manifests itself, is the cooperation between af-filiates and local firms. It is possible that these foreign investments con-tribute to economic and human development outcomes by requiring

P Mining FDI/DP Manufacturing FDI/GDP Services FDI/GDP

.991***(794.53)

1.012***(202.81)

1.009***(187.81)

− .00005***(−11.93)

2.18e−06(0.13)

.00002*(1.85)

.023**(2.44)

.184***(3.39)

− .306***(−3.47)

− .001(−1.19)

− .0007(−0.35)

− .0003(−0.15)

− .001**(−2.20)

− .002(−1.33)

− .002(−1.05)

− .010***(−13.60).050***(7.12)

− .051***(−3.80)

− .041***(−3.14)

28 31 305 6 60.924 0.310 0.029

roscedasticity and obtained from one-step Blundell–Bond SystemGMMwith instruments. ***, ** and * denote significance at 1%, 5%, and 10%, respectively.anufacturing and services FDI due to multicollinearity.

Table 6Sectoral FDI data coverage.

Countries Data coverage Countries Data coverage Countries Data coverage Countries Data coverage

Albania 2004–2010 Germany 1985–2010 Morocco 1996–2010 Tunisia 1980–2010Argentina 1978–2010 Greece 2001–2010 Mozambique 2002–2010 Turkey 1992–2010Armenia 1999–2009 Guatemala 2005–2010 Myanmar 1990–2010 United Kingdom 1980–2010Austria 1998–2010 Honduras 1993–2010 Netherlands 1980–2010 United States 1980–2010Australia 1985–2010 Hong Kong 1998–2009 Nicaragua 1991–2010 Uruguay 2001–2010Azerbaijan 1995–2008 Hungary 1999–2010 Norway 1994–2010 Venezuela 1976–2010Bangladesh 1995–2010 Iceland 1988–2010 Oman 2004–2010 Vietnam 1999–2010Belgium 2002–2010 India 1995–2010 Pakistan 1985–2010Belize 1999–2010 Indonesia 1999–2010 Panama 1990–2010Bolivia 1974–2008 Ireland 1983–2010 Paraguay 1990–2010Brazil 1996–2010 Israel 1996–2010 Portugal 1980–2010Brunei 1999–2010 Italy 1980–2010 Peru 1974–2010Bulgaria 1998–2010 Jamaica 1998–2009 Philippines 1974–2010Canada 1980–2010 Japan 1980–2010 Poland 1994–2010Chile 1974–2010 Kazakhstan 1993–2010 Romania 2003–2010China 1997–2010 Korea, Rep. 1979–2010 Russian Fed. 1998–2010Colombia 1990–2010 Kyrgyz Rep 1995–2009 Saudi Arabia 1990–2010Costa Rica 1974–2010 Lao PDR 1999–2010 Serbia 2004–2010Croatia 1993–2010 Latvia 1995–2010 Singapore 1999–2010Cyprus 1997–2010 Lithuania 1997–2010 Slovak Rep 2000–2010Czech Rep. 1993–2010 Luxembourg 2005–2010 Slovenia 2006–2010Denmark 1982–2010 Macao 2001–2009 Spain 1980–2010Dominican Rep. 1993–2010 Macedonia, FYR 1997–2010 Sweden 1989–2010Ecuador 1986–2010 Madagascar 2005–2010 Switzerland 1993–2010El Salvador 1998–2010 Malaysia 1999–2010 Syrian Arab Rep. 2004–2008Estonia 1994–2010 Mauritius 1990–2010 Tanzania 1999–2008Finland 1980–2010 Mexico 1974–2010 Thailand 1970–2010France 1980–2010 Moldova 2004–2008 Trinidad and Tobago 1974–2010


suppliers to meet higher standards of quality and also responsible busi-ness practices (i.e. corporate social responsibility or CSR in the businesslingua). The latter may include more stringent monitoring and stan-dards to promote child rights and prevent child labor. Nevertheless, itis well known that not all multinational corporations are able to pro-mote these higher standards.

The abovementioned empirical results also suggest that each indus-try in each region/country may have distinct net income and subs-titution effects as far as child labor and FDI are concerned. A briefanalysis of specific region/country cases may shed further light onthese contexts.

5. Cases of child labor in different regions/countries

The following cases help illustrate how investments and trade in dif-ferent sectors of the economy in different regions/countries bring aboutwidely differing results as far as their possible child labor impact is con-cerned. These cases help to establish possible explanations for themuchbroader aggregate empirical results in Section 3.10 The first two casestudies on gold mining in Kenya and Mali confirm the positive link be-tween FDI and child labor in the mining sector in Africa and suggestthat raising standards and requirements for FDI and transparent supplychains could help decrease the exploitation of children in mining. Thethird case study on cocoa production in West Africa (Ivory Coast andGhana) finds a positive link between FDI and child labor in the agricul-tural sector in Africa and points to the importance for largemultination-al companies to comply with commitments and to trace their supplychains. Confirming the empirical evidence of this paper for South andEast Asia, the fourth case study on services and manufacturing inVietnam finds a negative link between FDI and child labor, suggestingthat FDI to sectors that require a higher educated labor force might

10 Please note that the empirical results necessarily represent an average effect whichdoes not preclude that there are singular caseswhich point in the other directions. In orderto present the evidence as objectively as possible, in this section, we also include a casestudy that does not fully support the argument of the paper.

provide advantages for children in terms of reduced child labor andin terms of investment in human capital. The fifth case study onKazakhstan confirms the positive link of FDI and child labor in the agri-cultural sector in Europe and Central Asia and demonstrates how childlabor in certain agricultural activities, like FDI heavy tobacco production,remains a serious concern. It suggests that multinational companieshave to improve the working conditions all along their productionchain. It also demonstrates that the collaboration of the host govern-ment in protecting children is crucial. The sixth case study on carpetmaking in India and Nepal finds a negative link between FDI and childlabor in the manufacturing sector in South and East Asia and suggeststhat fair and ethical trade initiatives might have contributed to a de-crease in child labor in Nepal's and India's handmade carpet industry.The seventh and last case study on the production of hybrid seeds inIndia does not entirely support the empirical evidence of this papersince it finds that child labor is alsomuch of an issue in farms supplyingnational hybrid seed producing companies. This case study once againhighlights the importance for multinational companies' to implementhigher standards along their entire supply chains. Essentially, the lackof transparency in supply chains leaves these prone to sourcing mate-rials and products that are linked to child labor.

5.1. FDI and child labor in mining in Africa — gold mining in Kenya11

In Kenya, the lack of sufficient formal job opportunities discourageschildren (and their parents) from investing in a school education.Considered a better alternative for making ends meet, the gold minesin western Kenya's Nyanza province attract children. Children thereoften work in actual extraction or in ancillary services such as sellingfood. The local Children's Welfare Office estimates that about 15,000children areworking in goldmines in the districts of Nyatike andMigori(UNOCHA, 2012). The number is said to be even higher during week-ends and school holidays.

11 This case study is based on UNOCHA (2012).


The work in gold mines also poses severe risks to children's health.For instance, with its high toxicity and detrimental impact on children'sdevelopment the use of mercury for amalgamating small particlesof gold is a source of great concern. According to ILO (2011), gold-mining areas hold a number of mercury-related health risks for chil-dren. First, mercury can penetrate through the child's skin when han-dling it. In addition, the child can inhale mercury fumes when it isburned off over the fire. Children can also ingest particles that remainon their hands when eating or when consuming contaminated foodgrown in the surrounding area. Available evidence suggests that respi-ratory infections in Migori are high — at 37%, Migori has the province'shighest prevalence of TB, as well as high rates of respiratory tract infec-tions (UNOCHA, 2012). Dusty conditions in the mines may have alsocontributed to these health risks.

Furthermore, the engagement in work in gold mining is leading tofood insecurity since many people abandon farm work for jobs in thegold mines. Agriculture officials of the district of Nyatike inform thatat least 69% of the population is affected by food insecurity (UNOCHA,2012).

School drop-outs are reported to increase with the proximity of theschool to a goldmine. Childrenworking in the goldmines in the areas ofNyatike and Migori generally attend school for just two days a weekwhile they work on the remaining days (i.e. school attendance inthese areas is about 35% of the total school days) (UNOCHA, 2012).

The Department of Mines and Geology Pure estimates the golddeposits in Migori at a value of around KSh64 billion (about US$764million) (UNOCHA, 2012). Despite the richness in resources, however,the local residents are poor and middlemen take advantage of theoften low level of education of these people who tend to settle for anyprice they are given. Most gold mines are now run by small-scale com-panies and artisanal miners.

Lessons from this case suggest that raising standards and require-ments for FDI and a transparent supply chain could help decrease theexploitation of children in mining.

13 Commonly called the “Harkin–Engel Protocol” the official name of the protocol is “Pro-tocol for the Growing and Processing of Cocoa Beans and Their Derivative Products In aManner that Complies with ILO Convention 182 Concerning the Prohibition and Immedi-ate Action for the Elimination of the Worst Forms of Child Labor” For more information,see www.cocoainitiative.org/images/stories/pdf/harkin%20engel%20protocol.pdf.14 This case study is based on Mashayekhi et al. (2011).15 According to Vietnamese law, employment of children younger than 15 years of age isillegal. However, some exceptions apply for 12 and 15 year olds as defined by theVietnamese government (Mashayekhi et al., 2011).16 The two data sources are the annual enterprise survey of the General Statistical Office

5.2. FDI and child labor in mining in Africa — gold mining in Mali12

Mali is Africa's third largest gold producer. Mali's artisanal goldminers often rely on low-tech methods. HRW (Human Rights Watch)(2011, p. 6) informs that between 20,000 and 40,000 children arework-ing in Mali's artisanal gold mining sector. Children working in thesemines are often exposed to extremely harsh and hazardous conditions;children as young as six years old are taskedwith diggingmining shafts,pull up heavy weights of ore, working underground, etc. The work withmercury, a toxic substance used to separate the gold from the ore, is det-rimental to children's health.

Child labor in Mali involves migration by the children (childrentravel to themines from other parts of Mali as well as from other neigh-boring countries)whooften travel to themines by themselves being ex-posed to risks like exploitation and abuse or robbery. While Mali hasstrong laws on child labor and on compulsory and free education, im-plementation and enforcement of these policies remain key challenges.

HRW (2011, p. 75) notes that Mali's artisanally mined gold exportsof around four metric tons per year amount to around US$230 million(at September 2011 prices). The major share of this gold is exportedto Switzerland and the United Arab Emirates. Human Rights Watch(HRW, 2011) tried to get in touch with a number of companies thathave bought gold from Mali's artisanal mines to inform them aboutthe findings of their report. One out of the three international compa-nies (Kaloti Jewellery International, based in Dubai, a Belgian company,Tony Goetz, and the Swiss company Decafin) that Human RightsWatchwas able to contact, Kaloti, ceased to buy gold from Mali's artisanalmines after learning about the report's findings.

12 This case study is based on HRW (2011).

Large and non-transparent supply chains with many actors benefit-ing on the way seem to be at the core of continued child labor. The goldmoney travels a long way since artisanal miners usually sell the gold tolocal traderswho supplymiddlemen and trading houses in Bamako, thecapital of Mali. Human Rights Watch (HRW, 2011) reports that most ofthe Malian traders it interviewed expressed little concern about childlabor and health risks from mercury use.

The above description might reflect the reality in many more goldexporting countries in West Africa's copper belt. Human Rights Watch(HRW, 2011) reports that child labor in artisanal gold mining is partic-ularly common in West Africa's gold belt, which comprises BurkinaFaso, Ivory Coast, Ghana, Guinea, Mali, Niger, Nigeria, and Senegal.

5.3. FDI and child labor in agriculture in Africa— cocoa production inWestAfrica (Ivory Coast and Ghana)

The West African cocoa belt is an area where companies like Nestlébuy most of their cocoa. Long and complex supply chains make it diffi-cult for large companies to trace the supply back to the people that har-vest the produce they buy Hawksley (2011).

While the cocoa industry signedan international protocol13 in2001ban-ning dangerous child labor in cocoa production and promising resources forcombating it, child labor is far from being history inWest Africa.

A US government-backed report by Tulane University (2011, p. 7)shows that over 50% of agricultural households' children in the cocoabelt in Ivory Coast and Ghana work in agriculture, with 25 to 50%working in cocoa. In the observed period of the survey, a projectedtotal of 819,921 children in Ivory Coast and 997,357 children inGhana, in totalmore than 1.8million,worked in cocoa-related activities.Frequently involved in hazardous work like child trafficking, forcedlabor etc., about 5% of children in agricultural households in the cocoa-belt in Ivory Coast and more than 10% in Ghana are paid for theirwork (Tulane University, 2011, p. 7). The preceding implies the urgentneed of large multinational companies to comply with commitmentsand to meticulously trace their supply chains to avoid the use of childlabor in their production chains.

5.4. FDI and child labor in services and manufacturing in South and EastAsia— services and manufacturing in Vietnam14

Child labor of children under age 15 is still widespread in Vietnam,despite the fact of it being illegal by law. UNICEF (2011, Table 9,p. 123) estimates that around 16% of 5 to 14 year old children wereworking in Vietnam between 2000 and 2009. This rate was above theaverage rate of child labor for East Asia and the Pacific, which stood at11% in the same period.

A study published by the United Nations (Mashayekhi, Olarreaga, &Porto, 2011) analyzed the impact of FDI in the services andmanufactur-ing sectors on child labor in Vietnam. The statistics in the study refer tochildren between 10 and 14 years of age who are not allowed to workunder Vietnamese law.15 Based on repeated household surveys andenterprise surveys with information for the years 2002, 2004 and2006,16 the study found that FDI in the services and in the manufactur-ing sectors is negatively correlated with child labor, i.e. that an increase

of Vietnam and the Vietnam Household Living Standard Surveys (VHLSS) (Mashayekhiet al., 2011).

http://www.cocoainitiative.org/images/stories/pdf/harkin%20engel%20protocol.pdf


in the entry of FDI is associated with a reduction in children's labor sup-ply in both sectors.

Mashayekhi et al. (2011) find that the marginal effect of FDI on thereduction of child labor in the services sector is larger than in themanufacturing sector, demonstrating the great potential of FDI for aneven stronger reduction of child labor in the services sector. However,to date, manufacturing FDI in Vietnam is much larger than servicesFDI and it has so far had a greater effect than services FDI. The studyalso reveals that services FDI contributes to an increase in school enrol-ment rates of children aged 6 to 19 years. The authors link this increasein school enrolment to the fact that FDI in the services sector tends torequire more educated labor. This suggests that FDI that is channeledto sectors that require a higher educated labor force might provide ad-vantages for children — not only in terms of reduced child labor butalso in terms of investment in human capital.

18 While theGovernment of Nepal has ratified relevant conventions to regulate age stan-

5.5. FDI and child labor in agriculture in Europe and Central Asia— tobaccofarming in Kazakhstan17

Tobacco farming in Kazakhstan attracts Kazakhs and migrantworkers often from neighboring Kyrgyzstan, for seasonal work (HRW,2010). Kazakh farm owners employ the workers. Many of the farmowners in turn contract with and supply tobacco leaf to internationaltobacco companies.

One of the worst existing forms of child labor worldwide, tobaccofarming exposes children not only to physical hard work, long workinghours under high heat and sun during harvest times but also to hazard-ous pesticides and other health risks associated with tobacco plant har-vest (e.g. when harvesting wet tobacco leaves, nicotine is absorbedthrough the skin) (ILO, 2011). These health hazards are particularlydangerous for children whose bodies are still developing and hencemore vulnerable. Poor access to water, sanitation and nutrition arealso common factors for children in this field (HRW, 2010).

Migrant children are reported tomiss at least two to threemonths ofschool in Kyrgyzstan in order to accompany their families to Kazakhstanfor work (HRW, 2010). In the absence of permanent residence inKazakhstan, many migrant worker families are prohibited to enrolltheir children in school in Kazakhstan.

An ILO (2006, p. 4) study on child labor in agriculture informs thatwith a share of 70% of the domestic tobacco market, Philip MorrisKazakhstan (PMK), a subsidiary to Philip Morris International (PMI), isthe leading tobacco producer in Kazakhstan. The company is an impor-tant buyer and employer for local tobacco farmers in Kazakhstan.

ILO (2006, p. viii) found that child labor in tobacco farming inKazakhstan was prevalent with both Kazakh and migrant childrenaged 5–17. While interviewees observed a reduction in child labor intobacco farming between 1994 and 2004 which might be linked toPMK's policy prohibiting child labor ILO (ILO, 2006, p. 10) reports thatthese policies were violated in many cases. A Human Rights Watch(HRW, 2010, p. 3) study based on testimony of 68 migrant tobaccofarm workers in 2009 and early 2010, reported 72 cases of children,one as young as 10 years old, working in tobacco farming in 2009.

In response to the HRW study, PMK, increased protections formigrant workers on tobacco farms by requiringwritten contracts to en-sure migrant workers receive regular payments and other protections(HRW, 2012; Verité, 2011). PMI and PMK also promised improvementsin working conditions, expanding training with regard to labor rightsand hazards of child labor for workers, farmers, and PMK employeesand to collaborate with the government for improving access to schoolsfor migrant workers' children to prevent child labor.

Despite PMI and PMK's efforts of having implemented programsdesigned to detect and prevent child labor in tobacco in Kazakhstan,

17 This case study is based on HRW (2010).

child labor remains a serious concern in tobacco farming (HRW,2012). HRW (2012) informs that the Kazakh government continues tohinder migrant workers' children from registering in schools.

This case study underscores how child labor in certain agriculturalactivities, like FDI heavy tobacco production, remains a serious concern.Migration for seasonal agricultural work is a common phenomenon inmuch of Europe and Central Asia. Children are often forced to leavetheir homes and work in the fields. While it certainly shows that multi-national companies have to improve the working conditions all alongtheir production chain, it also demonstrates that the collaboration ofthe host government is crucial.

5.6. FDI and child labor in manufacturing in South and East Asia — carpetmaking in Nepal and India

The use of child labor (including bonded child labor) in the hand-made carpet industry is still subsistent in a number of countries, includ-ing Nepal (Office of the United States Trade Representative, 2012).Children's tasks related to the production of carpets include wooldying and spinning, thread rolling, carpet weaving, and carpet washing.Children working in carpet making are exposed to damaging chemicals,and harmful dust, often working long hours in unsafe conditions(United States Department of Labor, 2011).

While the Nepal Labor Force Survey 2008 finds that overall 1.6 mil-lion children aged 5–17, mostly girls, in Nepal are engaged in child labor(ILO (International Labour Organization) & CBS (Central Bureau of Sta-tistics) of Nepal, 2012, p. xi), it also reports that child labor has declinedin some sectors, one of them being carpet making (ILO (InternationalLabour Organization) Country Office for Nepal, 2012).

The Government of Nepal is reportedly taking action to fight childlabor, including bonded child labor (Office of the United States TradeRepresentative, 2012) but much remains to be done, includingstrengthening the enforcement of existing laws.18 Children under age16 are by law prohibited from engaging in hazardous work (UnitedStates Department of Labor, 2011) but evidence shows that compliancewith this law is lacking (ILO (International Labour Organization) & CBS(Central Bureau of Statistics) of Nepal, 2012).

FDI in Nepal is highly concentrated in the manufacturing sectorwithin which the textile and garment industries are strong recipients(Khanal & Shrestha, 2008). Aimed at preventing an international banon carpets that were associated with child labor, in the early 1990s,NGOs (non-governmental organizations), the government and the pri-vate sector joined forces to create a child labor free certification for car-pets produced in Nepal.19 A study on the impact of social labeling byNGOs on the incidence of child labor in Nepal finds that the probabilityof child labor is negatively correlated with the implementation of a la-beling program by the carpet industry (Chakrabarty, Grote, &Lüchters, 2006). It is also found to be positively correlated withchildren's school attendance (Chakrabarty et al., 2006).

Similarly, studying the impact of the labeling of child labor freecarpets on the welfare of children and their families in India andNepal, Chakrabarty and Grote (2009) find a reduction of child laborfor relatively better off households (those living above the subsistencelevel) but no significant impact for those living below this level.

While more research is needed, it can be assumed that fair andethical trade initiatives might have contributed to a decrease in childlabor in Nepal's and India's handmade carpet industry. The decrease inproduction in handmade carpets might also have contributed to the

dards for children's entry to the labor force (including ILO Conventions 138 and 182)moreefforts are needed to address the issue (ILO and CBS of Nepal (2012).19 For a review of social labeling and related initiatives to decrease child labor in Nepal'scarpet industry, please refer to World Education (2009).


decline in the use of child labor in India's manufacturing sector(Venkateswarlu, Ramakrishna, & Moid, 2006). Nepal's and India'scases demonstrate that the relationship between FDI and child labormight be more complex than it appears at first sight.

5.7. FDI and child labor in agriculture in South and East Asia — productionof hybrid seeds in India

Themajority of children that are engaged in theworst forms of childlabor in India work in agriculture (close to 70%), particularly in the pro-duction of rice and hybrid seeds (United States Department of Labor,2011, p. 367).20 This kind of work often involves heavy lifting and theexposure to pesticides that can be harmful to children's health. Childrenworking in the production of hybrid seeds are also often found to workunder forced conditions. Without a minimum age for employment, anda minimum age for hazardous work that does not comply with interna-tional standards, chances in India are high that even very young chil-dren may take up work that might be detrimental to their health(United States Department of Labor, 2011).

Commodities that are produced using hybrid seeds in India includecotton and vegetables like tomato, sweet and hot pepper. Reported tobe more labor- and capital-intensive than conventional cotton produc-tion, children are often considered a low-cost option for hybrid seedproduction (EJF, 2007).While several studies point to large involvementof child labor in the production of cotton,21 children are also found towork in the production of several hybrid vegetable seeds, one of thefastest growing industries in India.

Analyzing the involvement of child labor in five important cropsnamely tomato, sweet and hot pepper, okra and brinjal22 which arehighly labor-intensive, a 2010 study (Venkateswarlu, 2010) informsthat the recent increase in the demand for hybrid seeds has led to an in-crease of private hybrid seeds producing and selling companies. The top10 companies control more than 80% of the vegetable seed market(Venkateswarlu, 2010, p. 6).23 Out of 490 sample farms, in three states,namely Karnataka, Maharashtra and Gujarat, 58.5% produce seeds formultinational companies or their joint venture companies and 41.5%for Indian companies (Venkateswarlu, 2010, p. 4).24

Child labor involvement in multinational companies25 that havestarted to address the problem of child labor on farms producing seedfor them is lower compared to other seed companies that have not.26

However, the situation on the farms supplying other multinationalcompanies that have not yet initiated taking measures to combat childlabor27 is no different from Indian companies where child labor isoften still prevalent in large numbers.28 In this case, the implementationof higher standards in multinational companies' entire supply chainsmight help to reduce child labor in agriculture, more specifically, in

20 It should be noted that this report also points out the hazardouswork that is still beingdone by children in the manufacturing sector.21 See, for instance, EJF (2007) and Iram and Fatima (2008).22 Eggplant.23 The top 10 companies that controlmore than80% of the vegetable seedmarket are themultinational companies Syngenta, Nunhems, Bejo Sheetal, Seminis, Advanta and US Agriand the leading Indian companies Namdhari, Ankur and Vibha; the tenth, Mahyco, is in ajoint venture with Monsanto.24 The study is based on field visits that were conducted during the cross-pollination pe-riod, hence, the data of the work force composition in the study restricted to this activityonly.25 Syngenta, Nunhems and Seminis, in particular.26 For instance, the proportion of child labor (below 14 years) to the total work force onhot pepper farms in Karnataka varied between zero and 16% on Syngenta, Nunhems andSeminis farms, which is low compared to 24 to 38% foundon the farms of other companies(Venkateswarlu, 2010, p. 23).27 Bejo Sheetal, Advanta US Agri and East West Seeds (Venkateswarlu, 2010).28 For instance, the proportion of children below 14 years of age to the total work forceon hot pepper farms in Karnataka varied between 22 and 33% on Bejo Sheetal, Advantaand US Agri farms (Venkateswarlu, 2010, p. 23).

the production of hybrid seeds. For multinational companies, this isgoing to imply a number of challenges, including stricter monitoringand internal policing of their entire supply chain, possibly implying ad-ditional costs. In the longer run however, this is likely to be consistentwithmostmultinational companies'marketing and branding strategies,to the extent that consumers are becomingmuchmore sensitive to theiradherence to corporate social responsibilities in combating (or at leastnot exacerbating) child labor.

6. Conclusion

All of the foregoing suggests that the combined income and subs-titution effects of FDI appear to vary across sectors and countries (orregions). These are important nuances that earlier studies failed tocapture in assessing the link between FDI and child labor. Moreover,the coefficient of the variable indicating that the country had signedon to the CRC was positive and statistically significant for the modelspecification analyzing total FDI, manufacturing FDI and services FDI.With the exception of mining FDI among the African countries in thesample, all other model specifications for other regions appear toshow either a positive link or a coefficient that is not statistically differ-ent from zero.

These results run counter to the intended effect of signing on to theCRC. Possible explanations could be developed based on earlier work byBasu (1999, 2001, 2005) who wrote about possible unintended con-sequences from a ban on child labor, including possibly increasing it.Essentially, stronger laws and policies against child labor could lead tomultiple equilibria in the labormarket. A suddenwithdrawal of childrenfrom the labor force could increase thewage rates in different industrieswhere adult and child labor are somewhat substitutable. Since the im-pact across industrieswould differ— depending on the degree of substi-tutability of adult and child labor, as well as other factors such as theincome effect through higher adult wages — it is possible to see strongpecuniary incentives for children to work in some sectors. Hence, thenet impact would depend on the income and substitution effects onceagain; and a ban on child labor will not necessarily lead to a reductionin child labor.29

Along with the above conclusions, however, we do acknowledgethat this study presented one empirical approach to address the issues,and that there are limitations to the methodology utilized here. Withthe present data and information that we have, we recognize the needfor further fine tuning of the data collected on child labor. Future studiescould perhaps utilize micro-data and more robust impact evaluationmethods which could complement and validate the findings of ourcross-country study.

Finally, as the cases in this paper help illustrate, supplier arrange-ments could be an important avenue for mitigating child labor, as mul-tinational companies may help elevate the standards to which localbusinesses are held.

29 The CRC variable has been included to help illustrate the possible structural break be-tween the period before signing on to the CRC and afterwards. There are few (if any) stud-ies that have utilized this variable, and so this is quite novel for the child rights and childpolicies literature. (A possible exception is the work of Neumayer (2005) pertaining tothe effects of human rights treaties on respect for human rights). To helpmitigate the pos-sibility of other factors captured by the CRC variable, we have introduced other RHS vari-ables as noted earlier. Nevertheless, we do acknowledge that the CRC variable remains arough proxy variable seeking to capture complex changes occurring across countries inthepromotion of child rights. The empirical analysis ismeant to be complimented by qual-itative analyses and case studieswhichbetter capture those complex nuances; and our pa-per has been structured accordingly. Ultimately, the empirical literature on child laborcould be dramatically enhanced through better metrics that seek to capture the complexnuances on when and where this phenomenon could be observed.


References

Basu, K. (1999). Child labor: Cause, consequence, and cure, with remarks on internationallabor standards. Journal of Economic Literature, 37(3), 1083–1119.

Basu, K. (2001). A note on the multiple general equilibria with child labor. EconomicsLetters, 74(3), 301–308.

Basu, K. (2005). Child labor and the law: Notes on possible pathologies. Economics Letters,87(2), 169–174.

Basu, K., & Van, P. (1998). The economics of child labor. American Economic Review, 88(3),412–427.

Chakrabarty, Sayan, & Grote, Ulrike (2009). Child labor in carpetweaving: Impact of sociallabeling in India and Nepal. World Development, 37(10), 1683–1693.

Chakrabarty, Sayan, Grote, Ulrike, & Lüchters, Guido (2006). The trade-off between childlabor and schooling: Influence of social labeling NGOs in Nepal. ZEF — Discussion pa-pers on development policy no. 102, Center for Development Research. Zentrum fürEntwicklungsforschung, Bonn (www.zef.de/fileadmin/webfiles/downloads/zef_dp/zef_dp102.pdf).

Cigno, A., Rosati, F., & Guarcello, A. (2002). Does globalization increase child labor?WorldDevelopment, 30(9), 1579–1589.

Davies, R., & Voy, A. (2009). The effect of FDI on child labor. Journal of DevelopmentEconomics, 88(1), 59–66.

Doytch, N., & Uctum, M. (2011). Does the worldwide shift of FDI from manufacturing toservices accelerate economic growth: A GMM estimation study. Journal ofInternational Money and Finance, 30, 410–427.

Edmonds, E. V. (2008). Child labor. In T. P. Schultz, & J. Strauss (Eds.), Handbook ofDevelopment Economics, Vol. 4, North-Holland: Elsevier (Chapter 57).

Edmonds, E. V., & Pavcnik, N. (2005). Child labor in the global economy. Journal ofEconomic Perspectives, 19(1), 199–220.

Edmonds, E. V., & Pavcnik, N. (2006). International trade and child labor: Cross-countryevidence. Journal of International Economics, 68(1), 115–140.

Edmonds, E. V., & Schady, N. (2009). Poverty alleviation and child labor. National Bureau ofEconomic Research (No. w15345).

EJF (Environmental Justice Foundation) (2007). The children behind our cotton.London: Environmental Justice Foundation (www.ejfoundation.org/pdf/The%20Children%20behind%20Our%20Cotton%20FINAL.pdf).

Frankel, J. A., & Romer, D. (1999). Does trade cause growth? American Economic Review,89(3), 379–399.

Frankel, J. A., & Rose, A. (2002). An estimate of the effect of common currencies on tradeand income. Quarterly Journal of Economics, 117(2), 437–466.

Hawksley, Humphrey (2011). Nestle ‘to act over child labour in cocoa industry’. BBC News(November 28. [www.bbc.co.uk/news/world-africa-15917164]).

HRW (Human Rights Watch) (2010). “Hellish work”: Exploitation of migrant tobaccoworkers in Kazakhstan. (New York. [www.hrw.org/reports/2010/07/14/hellish-work]).

HRW (Human Rights Watch) (2011). A poisonous mix: Child labor, mercury, and artisanalgold mining in Mali. (New York. [www.hrw.org/reports/2011/12/06/poisonous-mix]).

HRW (Human Rights Watch) (2012). World report 2012: Events of 2011. (New York.[www.hrw.org/world-report-2012/world-report-2012-kazakhstan]).

ILO (International Labour Office) (2006). Child labour in tobacco and cotton growing inKazakhstan. Rapid assessment report. Almaty, International Labour Organization(www.ilo.org/ipecinfo/product/viewProduct.do?productId=8150).

ILO (International Labour Organization) (2011). Children in hazardous work: What weknow, what we need to do. Geneva: International Programme on the Elimination ofChild Labour (IPEC) (www.ilo.org/wcmsp5/groups/public/@dgreports/@dcomm/@publ/documents/publication/wcms_155428.pdf).

ILO (International Labour Organization) Country Office for Nepal (2012). Eliminating childlabor in Nepal: Facts, figures, commitment and action. (Kathmandu. [www.ilo.org/wcmsp5/groups/public/@asia/@ro-bangkok/@ilo-kathmandu/documents/projectdocumentation/wcms_182777.pdf]).

ILO (International Labour Organization), & CBS (Central Bureau of Statistics) of Nepal(2012). Nepal child labour report: Based on data drawn from the Nepal Labour

Force Survey 2008. (Kathmandu. [www.ilo.org/ipec/Informationresources/WCMS_182002/lang-en/index.htm]).

Iram, Uzma, & Fatima, Ambreen (2008). International trade, foreign direct investmentand the phenomenon of child labor: The case of Pakistan. International Journal ofSocial Economics, 35(11), 809–822 (www.emeraldinsight.com/journals.htm?articleid=1747069&show=html).

Khanal, D. R., & Shrestha, P. K. (2008). Trade and investment linkages and coordination inNepal: Impact on productivity and exports and business perceptions. (No. 5208).

Kitaura, K. (2009). Child labor, education aid and economic growth. Journal ofMacroeconomics, 31(4), 614–620.

Lipsey, R., & Sjöholm, F. (2005). In T. Moran, E. Graham, & M. Blomström (Eds.), Does for-eign direct investment promote development? (pp. 23–44). Washington, D.C.: Institutefor International Economics (Chapter 2).

Mashayekhi, Mina, Olarreaga, Marcelo, & Porto, Guido (2011). Services, trade and de-velopment. New York and Geneva: United Nations (http://unctad.org/en/docs/ditctncd2010d5_en.pdf).

Neumayer, E. (2005). Do international human rights treaties improve respect for humanrights? Journal of Conflict Resolution, 49(6), 925–953.

Neumayer, A., & de Soysa, I. (2004). Trade openness, foreign direct investment and childlabor. World Development, 33(1), 43–63.

Office of the United States Trade Representative (2012). Nepal. Washington, D.C.: Execu-tive Office of the President of the United States [www.ustr.gov/countries-regions/south-central-asia/Nepal]. Accessed on 30 July 2012).

Sachs, J. D., & Warner, A. M. (2001). The curse of natural resources. European EconomicReview, 45(4), 827–838.

Te Velde, D. (2003). Foreign direct investment and income inequality in Latin America.London: Overseas Development Institute (http://www.odi.org.uk/resources/docs/1928.pdf).

Tulane University (2011). Oversight of public and private initiatives to eliminate the worstforms of child labor in the cocoa sector in Cote d'Ivoire and Ghana. New Orleans, LA:Payson Center for International Development and Technology Transfer, Tulane Uni-versity (www.childlabor-payson.org/Tulane%20Final%20Report.pdf).

UNICEF (United Nations Children's Fund) (2011). The state of the world's children 2011.New York: United Nations Children’s Fund (UNICEF) [www.unicef.org/sowc2011/pdfs/SOWC-2011-Main-Report_EN_02092011.pdf].

United States Department of Labor (2011). 2010 findings on the worst forms of child labor.(Washington, D.C. [http://www.dol.gov/ilab/programs/ocft/PDF/2010TDA.pdf]).

UNOCHA (United Nations Office for the Coordination of Humanitarian Affairs) (2012).Kenya: ‘Goldmining beats school any day’. IRIN Humanitarian News and Analysis (Feb-ruary 9. [http://irinnews.org/Report/94822/KENYA-Gold-mining-beats-school-any-day]).

Venkateswarlu, Davuluri (2010). Growing up in the danger fields: Child and adult labor invegetable seed production in India. Washington, D.C., Utrecht, The Hague: ILRF (Inter-national Labour Rights Forum), ICN (India Committee of the Netherlands), SCL (StopChild Labour— School is the Best Place to Work) (www.indianet.nl/pdf/dangerfields.pdf).

Venkateswarlu, D., Ramakrishna, R. V. S. S., & Moid, M. A. (2006). Child labour in carpet in-dustry in India: Recent developments. International Labour Rights Fund [www.laborrights.org/sites/default/files/publications-and-resources/child%20labor%20in%20carpet%20industry%20122706.pdf].

Verité (2011). Farm level assessment of adherence to PMI GAP standards in Kazakhstan. Am-herst: MA (www.pmi.com/eng/media_center/company_statements/documents/farm_level_assessment_of_adherence_to_pmi_gap_standards_in_kazakhstan%20%E2%80%93%20may_2011.pdf).

Webbink, E., Smits, J., & de Jong, E. (2012). Hidden Child Labor: Determinants of house-work and family business work of children in 16 developing countries. WorldDevelopment, Elsevier, Vol., 40(3), 631–642.

World Education (2009). Children working in carpet industry: Child labor status report2009. Boston, MA: World Education, Inc [www.worlded.org/docs/Publications/ChildLabor/bfp-children-in-carpet-factories-report.pdf].

http://refhub.elsevier.com/S0190-7409(14)00332-6/rf0005





















































































































Documents

The impact of FDI on child labor: Insights from an empirical analysis of sectoral FDI data and case studies