the impact of household income on child labour in urban turkey

This article was downloaded by: [Aston University]On: 03 September 2014, At: 09:03Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK

The Journal of DevelopmentStudiesPublication details, including instructions for authorsand subscription information:http://www.tandfonline.com/loi/fjds20

The impact of household incomeon child labour in urban TurkeyMeltem Dayioğlu a

a Middle East Technical University , TurkeyPublished online: 24 Jan 2007.

To cite this article: Meltem Dayioğlu (2006) The impact of household income on childlabour in urban Turkey, The Journal of Development Studies, 42:6, 939-956, DOI:10.1080/00220380600774723

To link to this article: http://dx.doi.org/10.1080/00220380600774723

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all theinformation (the “Content”) contained in the publications on our platform.However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, orsuitability for any purpose of the Content. Any opinions and views expressedin this publication are the opinions and views of the authors, and are not theviews of or endorsed by Taylor & Francis. The accuracy of the Content shouldnot be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions,claims, proceedings, demands, costs, expenses, damages, and other liabilitieswhatsoever or howsoever caused arising directly or indirectly in connectionwith, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly

http://www.tandfonline.com/loi/fjds20

http://www.tandfonline.com/action/showCitFormats?doi=10.1080/00220380600774723

http://dx.doi.org/10.1080/00220380600774723

forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

http://www.tandfonline.com/page/terms-and-conditions

http://www.tandfonline.com/page/terms-and-conditions

The Impact of Household Income onChild Labour in Urban Turkey

MELTEM DAYIOGLUMiddle East Technical University, Turkey

Final version received June 2005

ABSTRACT The aim of this study is to investigate the determinants of child labour in urbanTurkey with a special reference to low household income or poverty as one of its root causes.Studies done elsewhere have produced mixed results which necessitate the relationship to bestudied at country-level. The data from urban Turkey indicate that children from poorer familiesstand at a higher risk of employment. This finding is confirmed using various measures ofhousehold material well-being. Simulation results have further pointed out that currentinterventions are not likely to produce a sizeable impact on the child labour problem.

I. Introduction

Employment of children is a common phenomenon in many developing countriesincluding Turkey. One need not to go to rural areas to see children engaged ineconomic activities. In cities, they work as street vendors, as apprentices and blue-collar workers in small (and sometimes large) establishments, and as service sectorworkers in restaurants, coffee houses and the like. Many more work in familyestablishments as unpaid family workers.

Official estimates of child workers stand at little over 1.6 million among 6- to17-year-olds, putting the child employment rate at 10.2 per cent (SIS, 2001). Childlabour has moved onto the agenda of the Turkish government and the general publicwith the implementation in 1992 of the International Program on the Elimination ofChild Labour (IPEC), an ILO initiative. Though certain sections of the Turkishsociety regarded the initiative as yet another ‘protectionist’ act by the westerncountries, by and large the program received wide acceptance and increased publicawareness and concern toward child labour. Partly as a measure to curb childlabour, in 1997 the government of Turkey raised compulsory years of schooling fromfive to eight years, and in 1998 signed ILO Convention 138 which in effect raised theminimum age of employment to 15 years.1 Furthermore, in 2001 Turkey ratified ILO

Correspondence Address: Meltem Dayıoglu, Department of Economics, Middle East Technical University,

06531 Ankara, Turkey. Email: [email protected]

Journal of Development Studies,Vol. 42, No. 6, 939–956, August 2006

ISSN 0022-0388 Print/1743-9140 Online/06/060939-18 ª 2006 Taylor & Francis

DOI: 10.1080/00220380600774723

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

Convention 182 that calls for the elimination of the worst forms of child labourcovering all children under 18 years of age.Employment of children is a concern since it may negatively affect their welfare.

The work they do and/or the employment relation they are in may impede theirmental, physical and psychological development. In fact, as will be illustrated in thepaper, many work for very long hours with little pay. However, light work might bebeneficial for children to some extent, as it is a form of socialisation. In fact undercertain conditions, it might very well increase their welfare. A number of studiespoint out that many child workers also attend school and in some cases it is theearnings from work that makes their schooling possible (Patrinos and Pscharopou-los, 1997; Myers, 1989). Understanding the work patterns of children and the factorsthat lead them to work is important precisely because of the multifaceted nature ofchild labour. Being categorically against all forms of child labour would beequivalent to closing our eyes to the very often impoverished state of children andtheir families and is unlikely to do them any good. In an effort to distinguish workthat may potentially impede children’s development and therefore might be deemedundesirable, we have chosen to consider as ‘child workers’ only those children whowork on average for 14 or more hours per week. The choice of the cut-off point is inline with the ILO definition of light work for children above 12 years of age, which ispermissible.2 Children in light work constitute 1.5 per cent of the working children inour data.Despite the growing abundance of research elsewhere, the number of studies that

attempt to establish causality between the employment of children and their socio-economic background in Turkey is very limited. Tunalı (1996), using household-levelmicro-data, investigates the work and schooling outcomes of male and femalechildren aged 6- to 14-years. He finds child’s age and gender, parental education andthe region of residence to be important determinants of child labour. Older malechildren and those with lower parental education have a higher likelihood ofemployment.3 Due to the nature of the data set employed, Tunalı does notinvestigate the relationship between child labour and household income which, asdiscussed below, constitute the main area of inquiry for this paper.The present study examines the impact on child labour of various household

characteristics with a special reference to household material well-being. Usinghousehold income as well as various other indicators, the study attempts tounearth the relationship between the economic status of the household and theincidence of child labour. Based on the established relationship, the paper thenattempts to evaluate the effectiveness of the current programs used to curb childlabour.The rest of the paper is organised as follows. Section II describes the data set

employed and gives a brief account of the household characteristics and theemployment patterns of children. Section III discusses the theoretical approaches putforward to explain child labour and outlines the empirical specification of the modelemployed. Section IV presents the results on the determinants of child labour andinvestigates the link between household economic status and child labour. Section Vcarries out a series of simulation exercises to assess the impact of currentinterventions on child labour. Section VI concludes the paper.

940 M. Dayıoglu

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

II. The Data and Employment Patterns of Children

The data for this study come from the 1994 Household Income Distribution Survey(HIDS) conducted by the State Institute of Statistics (SIS) and predates some of thefavourable legislative developments discussed earlier. It covers 80,380 individualsfrom 18,264 urban households. From this, 11,675 children in the 12- to 17-year agecategory, who were single at the time of the survey and could be linked to theirmother’s and father’s in the household, are drawn. The strength of the present dataset is that it provides detailed information on household income, which was collectedretrospectively for the previous year, as well as household assets and dwellingfacilities. Despite the richness of the data set, it has a potentially importantdrawback. It provides labour market information only for those 12-years of age andabove. However, this might not prove to be a serious problem for the current studysince child labour in urban areas becomes especially an issue beyond age 12. Officialestimates indicate that children in the 6- to 11-year age group only constitute 2.5 percent of all working children. They have an overall employment rate of 1.3 per cent(SIS, 2001).

The mean employment rate of children age 12–17 in our data set is 13.2 per cent.Male children in particular stand at a higher risk of employment making up almostthree quarters of the child workforce (Table 1).4 Wage employment constitutes themost common form of child employment in urban Turkey. A little over threequarters of working children are employed as wage earners (Table 2). Of theremaining quarter, over 90 per cent work as unpaid family workers and the rest areself-employed. The most striking aspect about working children is their long hours ofwork; the mean rate per week being 50.5 hours. Wage earners, in particular, put invery long hours. Another important characteristic of working children is theirrelative educational attainment vis-a-vis non-working children. In terms of thehighest diploma obtained, correcting for age, working children have, on average,fewer years of schooling. However, the schooling gap between the two groups lookslower than it actually is due to censoring. To put it differently, taking a simpleaverage of the completed years of schooling underestimates the school attainment ofnon-working children since the majority of them (78.4 per cent) are still in school.Underestimation is less of an issue for working children, the majority (84.1 per cent)of whom are not in school.5 Indeed, the survival graph produced in Figure 1 showsthat correcting for non-enrolment and age, there is clearly a trade-off between workand school; the probability of dropping out of the schooling system is much higherfor working children, which necessarily means that fewer of them will be attaininghigher levels of schooling. The trade-off seems to be particularly high for wageearners, who have lower enrolment rates than non-wage earners (Table 2).

The earnings of children are also of prime interest. Table 1 reports the averageannual household income as well as children’s earnings by taking into account onlythose who work for pay. The figures indicate that working children make a sizeablecontribution to household income; the annual earnings of children make up 21.6 percent of the total earnings of the child and his/her parents and 13.3 per cent ofhousehold income (Table 3). In about one fifth of households there is more than onechild working for pay. Among such households, the contribution of children to the

Household Income on Child Labour 941

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

Table 1. Summary statistics for children (12–17 years)

All childrenWorkingchildrena

Non-workingchildren

Child characteristics

Age 14.45 15.48 14.30(1.71) (1.41) (1.70)

Years of schooling 5.80 5.57 5.84(2.05) (1.72) (2.09)

Enrolled in school (%) 70.17 15.91 78.43Female (%) 47.81 26.86 51.0Child’s annual earningsb 20.46

(17.53)

Parental characteristics

Mother’s years of schooling 4.02 2.83 4.20(3.54) (2.58) (3.63)

Mother’s age 39.46 40.37 39.33(5.99) (6.40) (5.91)

Mother absent (%) 0.67 1.36 0.57Father’s years of schooling 6.24 4.65 6.47

(3.66) (2.34) (3.76)Father’s age 43.77 44.57 43.65

(6.77) (7.38) (6.67)Father public sector employee (%) 26.38 13.47 28.30Father private sector wage earner (%) 32.52 36.74 31.86Father’s annual earningsc 111.59 93.81 114.25(million TL) (161.96) (148.79) (163.68)Father absent (%) 5.70 7.33 5.45

Household characteristics

No. of children ages 0–6 0.34 0.34 0.34(0.67) (0.69) (0.67)



No. of adult members 2.72 2.84 2.70(1.10) (1.12) (1.10)

Total household incomec (exc. children’s) 162.45 141.19 165.68(annual: million TL) (184.46) (165.32) (187.0)Income from financial assetsc 3.71 1.52 4.05(annual: million TL) (24.29) (7.20) (25.91)Income from propertyc 21.46 17.81 22.01(annual: million TL) (31.30) (24.76) (32.15)Transfer incomec 13.01 9.04 13.61(annual: million TL) (31.84) (25.20) (32.70)Land owned (in hectare)d 1.97 3.51 1.73

(16.42) (19.45) (15.90)Household enterprise (%) 31.37 35.85 30.68

Regions (%)

Marmara 32.20 39.79 31.05Aegean 11.04 13.36 10.69

(continued)

942 M. Dayıoglu

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

Table 1. (Continued)

All childrenWorkingchildrena

Non-workingchildren

Mediterranean 13.39 10.29 13.87Central Anatolia 17.96 11.74 18.91Black Sea 8.42 14.44 7.50East Anatolia 5.75 2.42 6.26Southeast Anatolia 11.23 7.97 11.72Number of observations 11,675 1,296 10,379

Notes: Figures in parentheses are standard deviations. All income figures are deflated to allowfor regional price variations using 1987 CPI. aexcludes children whose actual hours of workare less than 14hrs per week; bexcludes children with no income; cincludes those with zeroincome; dincludes those with no land.

Table 2. Statistics on working children

Wage earners Non-wage earners

Age 15.52(1.39)

15.35(1.47)

Female (%) 24.41 34.44Years of schooling 5.48

(1.63)5.84(1.93)

Enrolled in school (%) 11.29 30.40Hours of work 52.67

(12.80)46.33(18.77)

Proportion of working children (%) 75.83 24.17Number of observations 882 415

Note: Figures in parentheses are standard deviations.

Figure 1. Probability of school enrolment by completed schooling and work status


Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

household budget is close to 25 per cent. Ignoring any possible labour supplyadjustments on the part of the other household members, Figure 2 illustrates thatwithdrawing children from the labour market reduces the incomes of especially thelower income households in a significant way as illustrated by an increase in theproportion of households at lower income levels. The vertical line in Figure 2indicates the poverty line, which is taken to be half the median income. While only 14per cent of households fall below the poverty line when children’s earnings areincluded, the incidence of poverty would increase to 26 per cent if children were tocease being a source of income for the household.The parents of working children are relatively less educated (see Table 1), which

help explain their lower earnings and consequently, the relatively high share ofworking children’s incomes in the household budget. These observations naturallylead to the conclusion that the incidence of child labour must be higher among poorerhouseholds. On the basis of per capita household income (excluding the earnings of

Table 3. Children’s contributions to household income

Children’s earnings as a proportion of:

parental and child earnings household income

Households with at least oneworking child for pay

21.22 13.31

Households with multipleworking children for paya

38.69 24.81

Note: Only considers children who work for pay. aearnings of all children within the householdare summed up.

Figure 2. Household’s position in income distribution

944 M. Dayıoglu

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

children) and household assets and dwelling facilities, income and wealth quintiles areformed. As shown in Table 4, the employment rate of children in lower quintiles isindeed much higher than in upper quintiles. It remains to be seen whether thisrelationship will persist when other factors affecting child labour are controlled for.

III. Theoretical Framework and the Model Employed

In this study we employ the household production model to analyse the intra-household resource allocation problem. The standard household production modelconsiders the household as maximising a common utility function subject to the full-income constraint (Becker, 1965, 1981; Gronau, 1973). In this framework, thehousehold combines the home time of its members with market goods that areacquired through market labour time to produce utility yielding commodities. Theoptimal time allocation between market work and non-market activities results fromthis optimisation process. The resulting reduced form demand equations forcommodities and the labour supply function depend on such exogenous factors asprices, wages and unearned income. The household production model has been usedwidely in analysing the issue of child labour (Rosenzweig and Evenson, 1977).6

In the literature, low household income or poverty is often cited as the mostimportant cause of child labour (Grootaert and Kanbur, 1995). In a recenttheoretical paper Basu and Van (1998) dwell on this issue and consider thepossibility of multiple equilibria emerging in situations where a potential for childlabour exists; a ‘good’ equilibrium where no children work and a ‘bad’ equilibriumwhere children work. Basu and Van build their model around altruistic parents whosend their children to work if and only if household income excluding children’scontributions fall short of some exogenously determined minimum. Furthermore,they assume that adult labour can be substituted for child labour. They argue that asufficient increase in adult wages can have the effect of reducing the incidence ofchild labour.

Many of the recent studies employ the luxury axiom of Basu and Van, where notsending children to work is viewed as a luxury that only relatively better offhouseholds could possibly afford (Canagarajah and Coulombe, 1997; Blunch andVerner, 2000; Borooah, 2000; Ray, 2000). To test the validity of the luxury axiom,various indicators of household material well-being have been employed rangingfrom (per capita) household income, household expenditures, non-wage income to

Table 4. Employment rate of children by income and wealth quintiles

Income quintilesa Wealth quintilesb

Bottom 20 per cent 17.77 19.26Second 20 per cent 14.08 14.10Third 20 per cent 12.67 12.82Fourth 20 per cent 11.38 11.82Top 20 per cent 7.57 5.53

Note: abased on per capita non-child annual household income; bbased on households assetsand dwelling facilities.


Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

paternal wages. The resulting empirical findings on the link between householdwelfare and child labour have been mixed. To shed light on the way in whichhousehold income is related to child labour in urban Turkey we rely on a number ofindicators including annual household income (excluding the children’s contribu-tions), father’s annual earnings and non-wage income, and household assets.7 Thegeneral approach adopted in the literature is to treat the labour supply decision ofadult males to be independent of the labour supply decision of adult females andchildren in the household, but not vice versa. In this sense, household income(inclusive of non-wage income) might be endogenous to the labour supply decisionof children as it will also include mother’s (and possibly other adult women’s)earnings. As an alternative measure, we consider only the father’s earnings and thenon-wage income accruing to the household members. The latter information isavailable in its sub-components in the form of income from financial assets, propertyand transfers. We utilise this information primarily to see how far transfers, apossible form of intervention, can have an impact on the employment of children.The third indicator is a wealth index constructed on the basis of household assetsand dwelling facilities using principal components analysis. On the basis of thewealth index, which can be taken as a long-term indicator of household welfare,households are divided into five equal groups. The lowest 20 per cent are consideredto constitute the poor.8

The dependent variable is the employment of children at any time over the12-month period in 1994.9 Incidentally, we might be including in here children whowork only during the summer months. We opted not to exclude such children fromthe analysis for the main reason that they work for very long hours (on average 41hours per week). Even though such summer work does not compete with schoolingfor the child’s time, the long hours may take their toll in various ways. It is also quitelikely that the summer work will be extended into the school year and therebyrendering the child unable to start school on time or to attend school on a regularbasis. We later undertake sensitivity analysis to evaluate the impact on the results ofremoving such children from the sub-group of child workers.Besides the household material well-being, other independent variables include the

age and the sex of the child, and his/her place of residence (seven regions of thecountry) which will determine his/her potential wages, the age and the educationlevel of the parents and the sector of employment of the father which will determineparental opportunity cost of time, the number of children in the household whichwill reflect the pressure on household resources, the number of adults, the amount ofland owned and the existence of a household enterprise as proxied by presence ofself-employed adult males in the household. Holding all other factors constant,having a household establishment will increase the chances that the child will beemployed as work is readily available. The amount of land owned, while increasingthe possibility of having a household-based establishment, will also signify a greaterhousehold wealth and can therefore have a negative effect on child labour.10

Bhalotra and Heady (2003) argue that in the face of land and labour marketimperfections, greater amounts of land holdings may lead to higher incidence ofchild labour. There are a small group of children in the data set without a mother ora father. We control for absent fathers and mothers in the model by including absentfather and mother dummies.

946 M. Dayıoglu

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

IV. Determinants of Child Labour

Household Welfare

The estimation results presented in Table 5 indicate that children’s employment isindeed responsive to household material well-being. This result holds regardless ofthe way in which household welfare is defined. The first set of results in Table 5shows that child employment goes down with household income. However, it shouldalso be mentioned that the effect of household income on child employment is not allthat large. The marginal effects indicate that one standard deviation increase inhousehold income around the mean (or doubling the household income) decreasesthe probability of child employment by 1.2 percentage points (see, in the appendix,Table A1).11

The results of the alternative specification given in the second panel of Table 5show that an increase in father’s earnings is instrumental in withdrawing childrenfrom the labour market, supporting the claim in the literature that an improvementin adult wages, in our case father’s wages, will result in a decline in the incidence ofchild labour.12 However, the marginal effects indicate again that one standarddeviation increase in father’s wages (or increasing paternal wages 2.5 fold) reducesthe employment probability of children by a single percentage point (Table A1).

The coefficients on the non-wage income components are negative but significantonly in the case of income from property and transfers.13 As a group they exert asignificant negative effect on child employment which is in line with theory andempirical findings. Comparing the marginal effects of the two significant non-wageincome components reveals no significant differences between them. It is alsointeresting to note that children’s employment is more responsive to a givenchange in non-wage income than to a change in paternal earnings; the marginaleffect of non-wage income being in the order of 70.04 (s.e. 0.01). Consequently, onestandard deviation increase in non-wage income around the mean brings about 1.8percentage points decline in child labour, which is still not that large considering thatthe required increase corresponds to doubling the mean non-wage income. It shouldalso be noted that non-wage income makes up only 23.3 per cent of the totalhousehold income (excluding children’s earnings). Income from property, inclusiveof imputed rent from owner occupied dwellings, constitutes 12.9 per cent of thisfigure, while transfers, the biggest component of which is retirement income, makeup 8 per cent. Income from financial assets is even smaller recorded at 2.3 per cent ofhousehold income. What these figures imply is that the livelihoods of the majority ofhouseholds depend on their labour earnings so that unless deliberate action is takenquite modest improvements should be expected in child labour as the labour marketrewards improve.

The wealth index also indicates that the economic standing of the household isimportant in determining the employment status of the child. The likelihood ofchildren’s employment is found to be significantly higher in the bottom three wealthquintiles as opposed to the top quintile. Measured at mean characteristics, a childcoming from a household in the bottom quintile as opposed to the top quintile has7.8 percentage points higher probability of being in the labour force. Being in thesecond quintile (third quintile) as opposed to the top quintile, on the other hand,increases the likelihood of employment by 4.4 (2.7) percentage points.14


Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

Table 5. Probit coefficient estimates for probability of work

Spec I:householdincome

Spec II:paternalearnings

Spec III:wealthquintiles

Spec IV:incomequintiles

Child characteristics

Child’s age 1.155*** 1.148*** 1.167*** 1.165***[0.210] [0.212] [0.211] [0.210]

Child’s age squared 73.138*** 73.109*** 73.164*** 73.170***(61072) [0.708] [0.715] [0.712] [0.707]

Female child 70.574*** 70.574*** 70.582*** 70.574***[0.039] [0.039] [0.040] [0.039]

Parental characteristics

Mother’s education 70.052*** 70.047*** 70.038*** 70.048***[0.009] [0.009] [0.009] [0.009]

Mother’s age 0.01 0.016 0.014 0.011[0.034] [0.034] [0.035] [0.034]

Mother’s age squared 70.01 70.017 70.018 70.011(61072) [0.039] [0.039] [0.040] [0.039]

No mother 70.125 0.012 70.071 70.063[0.784] [0.786] [0.796] [0.784]

Father’s education 70.059*** 70.055*** 70.044*** 70.057***[0.009] [0.009] [0.009] [0.009]

Father’s age 70.05 70.047 70.037 70.047[0.035] [0.035] [0.035] [0.035]

Father’s age squared 0.05 0.049 0.037 0.047(61072) [0.037] [0.037] [0.037] [0.037]

Father public sector employee 70.187** 70.301*** 70.190** 70.139*[0.080] [0.091] [0.081] [0.082]

Father private sector wage 0.099 70.02 0.043 0.107earner [0.074] [0.083] [0.075] [0.074]

No father 71.519* 71.418* 71.187 71.436*[0.839] [0.840] [0.846] [0.840]

Father’s annual earnings 70.054**(61078) [0.023]

Household characteristics

No. of children aged 0.035 0.033 0.018 0.0150–6 years [0.032] [0.032] [0.032] [0.032]

No. of children aged 0.072** 0.069** 0.05 0.0497–11 years [0.031] [0.031] [0.031] [0.032]

No. of children aged 0.092*** 0.083*** 0.075*** 0.064**12–17 years [0.026] [0.026] [0.026] [0.027]

No. of adult members 70.014 70.021 70.018 70.027[0.022] [0.022] [0.022] [0.022]

Land owned (hectares) 0.006*** 0.006*** 0.006*** 0.006***[0.001] [0.001] [0.001] [0.001]

Household establishment 0.246*** 0.137* 0.214*** 0.294***[0.072] [0.083] [0.071] [0.073]

Annual household income 70.053***(61078) [0.020]

Annual income from property 70.208*(61078) [0.121]

(continued)

948 M. Dayıoglu

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

To be able to compare these results with those reported above, we haveconstructed income quintiles based on per capita household income as well. The twosets of results obtained on the basis of wealth and income quintiles are in agreementwith each other. As noted earlier, being in the bottom three quintiles increases theprobability of child employment, though to a lesser extent than what we found onthe basis of wealth quintiles. As opposed to the top income quintile, being in thebottom quintile increases the probability of employment by 4.5 percentage points,while being in the second and third income quintiles increases the probability of childemployment by 3.2 and 1.8 percentage points respectively. The relatively biggerresponse obtained on the basis of wealth indicators can be interpreted to indicatethat households draw upon their accumulated wealth in times of hardship beforethey put their children to work, which is consistent with the assumption of Basu andVan (1998) that parents are altruistic toward their children.

Table 5. (Continued)





Annual income from financial 70.567assets (61078) [0.348]

Annual income from transfers 70.414**(61078) [0.166]

Bottom 20 per cent 0.637*** 0.376***[0.086] [0.085]

Second 20 per cent 0.429*** 0.284***[0.083] [0.080]

Third 20 per cent 0.294*** 0.172**[0.083] [0.080]

Fourth 20 per cent 0.117 0.115[0.088] [0.083]

Regions (ref. S.E. Anatolia)

Marmara 0.675*** 0.691*** 0.764*** 0.703***[0.083] [0.083] [0.084] [0.084]

Aegean 0.852*** 0.865*** 0.915*** 0.872***[0.086] [0.086] [0.086] [0.086]

Mediterranean 0.376*** 0.387*** 0.377*** 0.387***[0.079] [0.079] [0.078] [0.079]

Central Anatolia 0.440*** 0.439*** 0.455*** 0.436***[0.083] [0.083] [0.082] [0.082]

Black Sea 0.841*** 0.848*** 0.909*** 0.852***[0.080] [0.081] [0.080] [0.080]

East Anatolia 0.122 0.128 0.140* 0.131[0.085] [0.085] [0.085] [0.085]

Constant 710.374*** 710.411*** 711.368*** 710.801***[1.712] [1.729] [1.726] [1.712]

No. of observations 11675 11675 11675 11675Log likelihood 73323.561 73307.990 73279.770 73313.999

Note: Figures in parentheses are standard errors. *significant at 10 per cent; **significant at5 per cent; ***significant at 1 per cent.


Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

Other Determinants of Child Labour

The individual and household characteristics determining children’s employment isin line with the general findings in the literature. The child’s age is an importantdeterminant of his/her likelihood of employment. As the child grows older, the riskthat s/he will be employed rises though at a decreasing rate. Male children stand at ahigher risk of labour market employment.15 Parental schooling is also an importantfactor affecting child employment. An additional year of maternal and paternalschooling is expected to reduce children’s employment by about 0.5–0.7 percentagepoints (see Table A1) – the difference in the marginal effects of the two variablesbeing statistically insignificant. Father’s employment in the public sector exerts anegative effect on child employment. Since we are controlling for father’s earnings,this effect must stem from the social environment governing civil employment and/orlack of social networks to place the child at work. Larger numbers of children in thefamily also increase the likelihood of children’s employment. Young childrenconstitute an exception to this observation possibly because they do not exert asmuch pressure on family finances as older children. The greater work availabilitymade possible by the presence of a household-based establishment increases thelikelihood of children’s employment. So do the higher amounts of land owned by thehousehold, which is consistent with the findings of Bhalotra and Heady (2003).Quite curiously, the absent father dummy, when significant, carries the wrong

sign. In other words, in households where the father is absent, children seem to havea lower likelihood of employment. This rather surprising observation is noted by anumber of researchers, some offering the explanation that the presence of a fathermight be important for networking and placing the child at a job (Levison, 1997;Binder and Scrogin, 1999). Though this explanation might be important to someextent, it seems to us that another conjecture worth considering is the possible under-reporting of non-wage income. Remittances sent by absent fathers working oversees(mostly western Europe) or in larger cities in Turkey, or the transfers received bywidowed mothers could be an important income source for female headedhouseholds.16 The finding that absent father dummy becomes insignificant whenhousehold economic status is represented by the wealth index gives support to ourconjecture. If, indeed, under-reporting is the case, then we would be overestimatingthe impact of transfers. However, considering that only 5 per cent of the samplehouseholds are female headed, the bias might not be significant. Re-running themodel on two-parent households only, shows that there is a slight drop in thetransfer income coefficient while the coefficients on the other non-wage componentsremain practically the same.Being in different parts of the country also affects the likelihood of children’s

employment. In particular, children who live in the Aegean and the Marmararegions and the Black Sea Coast have a much higher likelihood of employment asopposed to those living in southeast Anatolia, which is the most underdevelopedregion of the country. The higher likelihood of employment in the Black Sea regioncan be attributed to the relatively more important role agriculture plays in theregion’s economy. The Aegean and Marmara regions, on the other hand, are veryindustrial with the manufacturing sector constituting an important source ofemployment for both the adults and children.

950 M. Dayıoglu

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

Sensitivity Analysis

As mentioned earlier, the working sample might include school children who mighthave worked only during the summer months. Since the data do not provideinformation on which months of the year the children were employed, we make theassumption that they are comprised of those who work less than the summerholidays, which is three months in Turkey. Children whom we suspect only workover the summer holidays constitute 9.1 per cent of the working children in oursample. To see whether the results discussed earlier change when these children arere-grouped among non-workers we run a series of estimations. The results (notpresented here) indicate that the findings remain intact.

V. Simulations Under Different Scenarios

Under the assumption that low household income/poverty is the main culprit ofchild labour, a number of projects, mostly carried out within the framework of IPEC(International Program on the Elimination of Child Labour), have aimed atincreasing household incomes in localities where a high incidence of child labour isobserved. While some of these projects tried to increase household income throughin-cash and/or in-kind transfers, others opted to encourage households to start theirown business through micro-credit programs. Using the mean characteristics ofchildren from households at the bottom quintile and coefficients estimated for theentire sample, we ran a series of simulations under different scenarios to establish theexpected impact of the implemented policies on the incidence of child labour.

The predicted probability of child employment among poor households, definedas those in the bottom 20 per cent in terms of household income, is found to be 10.3per cent.17 Programs that aim to reduce child labour by encouraging poorhouseholds to set up their own businesses are likely not to realise their goal offewer working children.18 Assuming that before the implementation of the programa household establishment does not exist, its creation is predicted to drasticallyincrease the incidence of child labour among poor households; the change beingfrom 9.3 per cent to 15.2 per cent. This result follows from the fact that child labouris more common in households where a household enterprise exists. Grootaert (1999:56) arriving at a similar result warns that ‘household enterprises are a double-edgedsword’. For poverty alleviation programs not to have the adverse effect of increasingchild labour at the expense of child schooling, he advocates that proper incentives becreated within the poverty alleviation programs to encourage child schooling.

The figures cited above are generated under the scenario that the household is notable to generate enough income through its enterprise to push itself to the nextincome bracket. Even if the resulting increase in income is such that the householdfinds itself in the second quintile, the incidence of child labour in the household withenterprises is still higher (13.2 per cent) than what we started out with. However, itshould also be mentioned that often the amount of credit provided is too small toenable families to generate a substantial amount of income to allow for upwardmobility so that we should expect higher rather than lower incidence of child labour.These exercises clearly indicate that the form of assistance provided to the poorhouseholds carries great importance for the incidence of child labour.


Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

Notwithstanding the predicted adverse effect of income generation programs inthe form of higher child labour, they are highly favoured primarily because of theassumption that children’s working conditions improve when they work for theirfamily members rather than for an unrelated employer. In many of the implementedprojects reduction in child labour has not been the sole objective, they have alsoaimed to improve the working conditions of children. Despite the fact that the twoobjectives can be in contradiction at times, if the latter is the primary objective, thenmicro-credit type programs might help to increase child welfare. Although such aninvestigation is beyond the scope of the current study, simply judged from theperspective of hours of work, children employed in household-based establishmentsare found to work for substantially fewer hours per week and number of weeks peryear. If indeed, children fare better in the employment of their kin, what policymakers need to decide is how much more child labour can be tolerated in the interestof employing children under better conditions.Transfers can also be used in withdrawing children from the labour market and in

re-orienting them toward school. Admittedly they currently constitute a relativelyless favoured tool in combating child labour and in empowering the poor because ofthe limited resource base of implementing institutions, among other reasons. Ourresults point out that, transfers can be of use in withdrawing children from thelabour market but that quite large sums are needed to produce a visible impact onchild labour. For instance, an income transfer to the poor households of onestandard deviation of what they on average receive would lead to 1.2 percentagepoints decline in child labour. Levison (1997) arrives at a similar conclusion forBrazilian children.The relatively smaller impact of transfers can in part be attributed to the fact that

we employ a relatively long-term definition of income, whereas the true role of non-wage income might be felt in poor households in the short-run by way of bufferingchildren against financial crisis. If this is so, in an environment of imperfect capitalmarkets, transfers can be used to mitigate the impact of unforeseen events thatchallenge the welfare of the household and therefore, necessitate the employment ofchildren even if for short durations. Currently, apart from sporadic assistanceprovided by a handful of NGOs, the most important institutions that provide regularcash and in-kind transfers to needy families on behalf of the State are the SocialServices and Child Protection Agency and the Social Solidarity Foundations (SSF).However, due to their bureaucratic structures it is doubtful that they can be flexibleenough to provide funds to help families overcome short-term shocks.The effectiveness of transfers can potentially be increased if they can be linked to

programs that reduce the time available for children to work. For instance, the‘conditional cash transfer’ program of the SSF launched in 2002 requires the regularschool attendance of children for continued State assistance. Again, provided thattransfers are high enough to convince the poor families to take part in the program,increased school enrolment might be achieved along with a reduction in child labour.19

VI. Conclusion

Employment of children is found to be responsive to the economic status of thehousehold. An increase in household income whether in the form of an improvement

952 M. Dayıoglu

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

in paternal earnings or non-wage income, is expected to bring about a reduction inchild labour. However, we have also established that rather high income increasesare required to bring about a sizeable reduction in child labour.

Our results have also pointed out that the incidence of child labour is higher amonghouseholds that are asset poor as opposed to being income poor. The implication isthat households use their asset base as a buffer to keep their children out of the labourmarket. In times of crisis, the household might respond to the deteriorating marketconditions by first liquidating the household assets and then, putting the child towork. In this sense, minor economic downturns might not be as alarming, thoughmajor economic crisis like that of 2001 might very well push more children to thelabour market as households move from being income poor to asset poor.

As the simulation exercises indicate certain interventions such as micro-creditprograms geared toward alleviating the position of poor households through theestablishment of a household enterprise might actually lead to an increase in childlabour. While effective program development requires the careful evaluation of theresponses of children and their families to various socio-economic variables, themerits of following a holistic approach are also clear. The goal of increasinghousehold income through transfers or the establishment of a household enterpriseshould be considered as part of an integrated poverty alleviation strategy, and not asstand-alone programs. Taking into account our finding that higher parentaleducation reduces the risk of child employment, and the important links educationhas with earnings potential of parents and various dimensions of child welfare, acrucial component of any interventions strategy would include increasing theknowledge base of the household on various matters ranging from businessmanagement to child welfare.

Acknowledgements

I wish to thank Ragui Assaad, Erol Taymaz and two anonymous referees for theirvaluable comments and suggestions. I assume responsibility for the remaining errors.

Notes

1. ILO Convention 138 stipulates age 13 as the minimum age for light work. Prior to the ratification of

the Convention, minimum age of employment was 12-years in Turkey, though an age limit did not

exist for children engaged in agricultural activities or for those employed in household enterprises.

2. Another dimension of light work is the nature of work itself. Children are not permitted to engage in

work that is unsuitable for their capacity even for few hours.

3. Tunalı (1996) does not correct for household income so that it is not clear whether it is the inferior

schooling of children’s parents per se that leads them to work or the low household income due to low

schooling.

4. Sampling weights are applied to all summary tables.

5. For children who work all year round, the survey does not provide information on school enrolment.

The enrolment figures in Tables 1 and 2 are estimated under the assumption that children who work

for 12 months for at least 14 hours per week have dropped out of the schooling system. This is a

plausible assumption considering that such children work 52.8 hours, on average, per week.

6. Many others that followed Rosenzweig and Evenson (1977) analysed the time allocation of children

between work, leisure and schooling. Apart from the work cited in the text see for instance, Levy

(1985), Skoufias (1994), Binder and Scrogin (1999), Levison et al. (2001).


Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

7. We do not express income in per capita or per adult equivalent since we include in the estimation a

series of regressors reflecting household composition.

8. We use the first component in the principal components analysis to construct the asset weights. The

variance accounted for by the first principal component is 23 per cent. Each household is assigned an

index value based on the assets they have weighted by the associated assets weights. The wealth

quintiles are then constructed by ordering the households with respect to their index value and then

dividing them into five equal groups. A similar procedure has been used by Filmer and Pritchet (2001)

and Assaad et al. (2004).

9. We consider the child as the unit of analysis. Since more than one child may come from the same

household, we use the Huber correction by taking households as clusters.

10. The amount of land owned is assumed to be a long-term decision variable and that it is unlikely to be

endogenous to the decision to put the child to work.

11. As argued by Bhalotra and Heady (2003) the rather small income effect may have been due to

simultaneity. However, this should not be an issue for paternal earnings.

12. This result does not change when we replace paternal earnings with total household earnings or

earnings accruing to male adults in the household.

13. Excluding the amount of land owned does not change the coefficient on the property income in a

significant way.

14. Marginal effects are calculated by assuming in turn that the child is in the first, second and third

quintile. The marginal effects calculated in this way differ from what is presented in Table A1 where

the dummies are switched on and off while the rest of the variables, including the other dummies, are

held at their means.

15. We have opted not to disaggregate the analysis by sex since income effects do not to differ between

male and female children.

16. Female headship is observed only when the husband is absent from the household. Only in 0.3 per cent

of the households do we see female headship though the husband is present in the household.

17. In terns of the wealth index this figure is slightly higher estimated at 12.8 per cent.

18. Since the micro-credit programs hardly ever provide funds large enough to enable a family to buy their

own plot of land, we leave the land endowment of the household unchanged.

19. The impact of the program on child labour and schooling is not clear since program evaluation has

not been carried out yet.

References

Assaad, R., Levison, D. and Zibani, N. (2004) The effect of child work on schooling: evidence from Egypt,

mimeograph, University of Minnesota.

Basu, K. and Van, P. H. (1998) The economics of child labour, American Economic Review, 88(3), pp. 412–

27.

Becker, G. S. (1965) A theory of the allocation of time, Economic Journal, 75, pp. 493–517.

Becker, G. S. (1981) A Treatise on the Family (Cambridge, MA: Harvard University Press).

Bhalotra, S. and Heady, C. (2003) Child farm labour: the wealth paradox, World Bank Economic Review,

17(2), pp. 197–227.

Binder, M. and Scrogin, D. (1999) Labour force participation and household work of urban

schoolchildren in Mexico: characteristics and consequences, Economic Development & Cultural

Change, 48(1), pp. 123–54.

Blunch N. and Verner, D. (2000) Revisiting the link between poverty and child labour: the Ghanaian

experience, World Bank Policy Research Working Paper 2488.

Borooah, V. K. (2000) The welfare of children in central India: econometric analysis and policy

simulation, Oxford Development Studies, 28(3), pp. 263–87.

Canagarajah, S. and Coulombe, H. (1997) Child labour and schooling in Ghana, Policy Research

Working Paper 1844, World Bank.

Filmer, D. and Pritchett, L. (2001) Estimating wealth effects without income or expenditure data—or

tears: education enrolment in India’, Demography, 38(1), pp. 115–32.

Gronau, R. (1973) The intrafamily allocation of time: the value of the household wifes’ time, American

Economic Review, 63(4), pp. 634–51.

954 M. Dayıoglu

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

Grootaert, C. and Kanbur, R. (1995) Child labour: an economic perspective, International Labour Review,

134(2), pp. 187–203.

Grootaert, C. (1999) Child labour in Cote d’Ivoire, in C. Grootaert and H. Patrinos (eds), The Policy

Analysis of Child Labour: A Comparative Study (New York: St Martin’s Press).

Levison, D. (1997) Household composition and early human capital formation: evidence from Brazil on

children’s labour force work and schooling, mimeograph, University of Minnesota.

Levison, D., Moe K. S. and Knaul, F. (2001) Youth education and work in Mexico, World Development,

29, pp. 167–88.

Levy, V. (1985) Cropping pattern, mechanisation, child labour, and fertility behaviour in a farming

economy: rural Egypt, Economic Development & Cultural Change, 33(4), pp. 777–91.

Myers, W. (1989) Urban working children: a comparison of four surveys from South America,

International Labour Review, 128(3), pp. 321–35.

Patrinos, H. and Psacharopoulos, G. (1997) Family size, schooling and child labour in Peru: an empirical

analysis, Journal of Population Economics, 10, pp. 387–405.

Ray, R. (2000) Analysis of child labour in Peru and Pakistan: a comparative study, Journal of Population

Economics, 13(1), pp. 3–19.

Rosenzweig, M. R. and Evenson, R. E. (1977) Fertility, schooling, and economic contribution of children

in rural India: an econometric analysis, Econometrica, 45, pp. 1065–79.

Skoufias, E. (1994) Market wages, family composition and the time allocation of children in agricultural

households, Journal of Development Studies, 30(2), pp. 335–60.

State Institute of Statistics (SIS) (2001) Turkiye’de Calısan Cocuklar (Child labour in Turkey) (Ankara:

SIS).

Tunalı, _I. (1996) Education and work: experiences of 6- to14-year-old children in Turkey, in T. Bulutay

(ed.), Education and the Labour Market in Turkey, pp. 106–43 (Ankara: SIS).

Appendix

Table A1. Marginal effects on the estimated probit coefficients





Child’s age 0.146 0.143 0.144 0.147Child’s age squared (61072) 70.397 70.387 70.390 70.400Female child 70.073 70.072 70.072 70.073Mother’s education 70.007 70.006 70.005 70.006Mother’s age 0.001 0.002 0.002 0.001Mother’s age squared (61072) 70.001 70.002 70.002 70.001No mother 70.014 0.002 70.008 70.008Father’s education 70.007 70.007 70.005 70.007Father’s age 70.006 70.006 70.005 70.006Father’s age squared (61072) 0.006 0.006 0.005 0.006Father public sector employee 70.022 70.034 70.022 70.017Father in private sector 0.013 70.003 0.005 0.014No father 70.073 70.071 70.067 70.072Father’s earnings (61078) 70.007No. of children aged 0–6 years 0.004 0.004 0.002 0.002No. of children aged 7–11 years 0.009 0.009 0.006 0.006No. of children aged 12–17 years 0.012 0.010 0.009 0.008No. of adult members 70.002 70.003 70.002 70.003Land owned (hectares) 0.001 0.001 0.001 0.001Household establishment 0.034 0.018 0.028 0.041

(continued)


Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

Table A1. (Continued)





Household income (61078) 70.007Income from property (61078) 70.026Income from financial assets (61078) 70.071Income from transfers (61078) 70.052Bottom 20 per cent 0.100 0.054Second 20 per cent 0.063 0.040Third 20 per cent 0.042 0.023Fourth 20 per cent 0.015 0.015Marmara 0.124 0.126 0.144 0.130Aegean 0.174 0.176 0.190 0.180Mediterranean 0.058 0.059 0.057 0.060Central Anatolia 0.069 0.068 0.071 0.068Black Sea 0.158 0.158 0.173 0.161East Anatolia 0.016 0.017 0.019 0.018Estimated prob. of work at mean 0.065 0.063 0.063 0.065

Notes: For dummy variables, marginal effects are calculated by comparing the probabilitywhen the dummy variable is 1 and when it is 0. For continuous variables, marginal effects arecalculated at the means.

956 M. Dayıoglu

Dow

nloa

ded

by [

Ast

on U

nive

rsity

] at

09:

03 0

3 Se

ptem

ber

2014

the impact of household income on child labour in urban turkey

Documents