+ All Categories
Home > Documents > CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India,...

CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India,...

Date post: 06-Feb-2018
Category:
Upload: lenguyet
View: 213 times
Download: 0 times
Share this document with a friend
37
C HILD S CHOOLING AND W ORK D ECISIONS IN I NDIA :T HE R OLE OF H OUSEHOLD AND R EGIONAL G ENDER E QUITY Uma Sarada Kambhampati ABSTRACT This paper tests three hypotheses about how mothers’ autonomy in India affects their children’s participation in school and the labor market. To do so it extends the concept of mothers’ autonomy beyond the household to include the constraints imposed by the extent of gender equity in the regions in which these women live. This study began with the expectation that increased autonomy for Indian mothers living in heterosexual households would increase child schooling and decrease child work. However, the results are mixed, indicating that mother’s autonomy can be reinforced or constrained by the environment. The paper concludes that mothers and fathers in India make different decisions for girls vis-a `-vis boys and that the variables reflecting mothers’ autonomy vary in their impact, so that mothers’ level of education relative to fathers’ is not often statistically significant, while mothers’ increased contributions to household expenditure decrease the probability of schooling and girls’ work. KEYWORDS Child labor, gender roles, intrahousehold inequality JEL Codes: D13, J16 INTRODUCTION This paper considers whether increased autonomy for mothers in India improves child welfare, specifically in terms of whether children attend school or participate in the labor market. In this context, the factors used to determine how much autonomy a mother possesses are her education and employment status, her education and income contributions relative to her spouse, and the extent of gender equity that prevails in the region in which she lives. The paper asks whether mothers and fathers make symmetric decisions with regard to child work and schooling, whether mothers with greater autonomy make ‘‘better’’ decisions than those with less autonomy, and whether kinship systems are important in determining these decisions. Feminist Economics 15(4), October 2009, 77–112 Feminist Economics ISSN 1354-5701 print/ISSN 1466-4372 online Ó 2009 IAFFE http://www.tandf.co.uk/journals DOI: 10.1080/13545700903153997
Transcript
Page 1: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

CH I L D SC H O O L I N G A N D WO R K DE C I S I O N S I N

IN D I A: TH E RO L E OF HO U S E H O L D A N D

RE G I O N A L GE N D E R EQ U I TY

Uma Sarada Kambhampati

ABSTRACT

This paper tests three hypotheses about how mothers’ autonomy in Indiaaffects their children’s participation in school and the labor market. To do so itextends the concept of mothers’ autonomy beyond the household to includethe constraints imposed by the extent of gender equity in the regions inwhich these women live. This study began with the expectation that increasedautonomy for Indian mothers living in heterosexual households would increasechild schooling and decrease child work. However, the results are mixed,indicating that mother’s autonomy can be reinforced or constrained by theenvironment. The paper concludes that mothers and fathers in India makedifferent decisions for girls vis-a-vis boys and that the variables reflectingmothers’ autonomy vary in their impact, so that mothers’ level of educationrelative to fathers’ is not often statistically significant, while mothers’ increasedcontributions to household expenditure decrease the probability of schoolingand girls’ work.

KEYWORDSChild labor, gender roles, intrahousehold inequality

JEL Codes: D13, J16

INTRODUCTION

This paper considers whether increased autonomy for mothers in Indiaimproves child welfare, specifically in terms of whether children attendschool or participate in the labor market. In this context, the factors used todetermine how much autonomy a mother possesses are her education andemployment status, her education and income contributions relative to herspouse, and the extent of gender equity that prevails in the region in whichshe lives. The paper asks whether mothers and fathers make symmetricdecisions with regard to child work and schooling, whether mothers withgreater autonomy make ‘‘better’’ decisions than those with less autonomy,and whether kinship systems are important in determining these decisions.

Feminist Economics 15(4), October 2009, 77–112

Feminist Economics ISSN 1354-5701 print/ISSN 1466-4372 online � 2009 IAFFEhttp://www.tandf.co.uk/journals

DOI: 10.1080/13545700903153997

Page 2: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Analysis of the study data leads to the conclusion that mothers and fathersin India make different decisions for girls than they do for boys. Thevariables used to proxy for a mother’s autonomy vary in their impact on theprobability of children attending school and children working. Morespecifically, female autonomy measured by how much education a motherhas relative to the level the father has achieved is not often statisticallysignificant but when it is, higher autonomy of the mother measured by hereducation leads to increased child work and decreased schooling. Whenmothers’ contributions to household expenditures increase, especially inhouseholds with incomes below the Indian poverty line,1 the probability ofschooling for their children decreases. This surprising result might wellreflect the fact that mother’s contribution to household expenditure mightbe higher in poorer households. However, with mothers working andcontributing to household expenditure, daughters may not need to work.Once again, this is reflected in a decrease in the probability that daughtersyounger than 15 years will be working. Overall, this study finds that theeducation and employment characteristics (primary, secondary, andtertiary education and employment) of the mother and father matterindependently. Their positions relative to each other (mother’s expendi-ture contribution to the household and relative education) also matter asdoes the level of gender equity in the region.

Our analysis is undertaken in the social context of India, where genderequity varies considerably both across households and across regionsbecause kinship systems vary across castes, religions, and regions. NailaKabeer argues, for instance, that households based on patriarchy-patriliny-patrilocality are most common in the northern plains of India, amongMuslims, upper-caste Hindus, and landowning classes (2003: 116). TimDyson and Mick Moore (1983) divide the country into three separatekinship systems – the North Indian System, the South Indian system,and the East Indian System – based on their approaches to femaleindependence. In the North Indian system, spouses are unrelated in termsof kinship, men cooperate with and receive help only from those men whoare blood relatives, and women do not inherit property. These kinshipcharacteristics create a system in which groups of patrilineally related menrigidly control the household roles of women within their groups throughrestrictions like purdah (the physical seclusion of women), as a means ofmaintaining their honor, reputations, and power. In this environmentwomen have little freedom and are very carefully protected from outsideinfluences. In contrast, within the South Indian kinship system, spouses areoften cross-cousins (that is, the children of a parent’s opposite-sex sibling),close socioeconomic relations exist between men who are related by bloodand by marriage (see also Lupin Rehman and Vijayendra Rao 2004), andwomen may inherit property. Dyson and Moore (1983) argue that thissystem results in less rigid control of women’s movements. Relative to their

ARTICLES

78

Page 3: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

status in the north, daughters in southern families are more valued, botheconomically and socially. They are more likely to survive infancy andchildhood, to be educated, to work, to marry later, and to marry intohouseholds located closer to their natal homes, enabling them to maintainties with their parents after marriage (David E. Sopher 1980; Barbara D.Miller 1981; Patricia Jeffrey, Roger Jeffrey, and Andrew Lyon 1988).Revisiting the thesis that daughters in South India enjoy higher status andgreater autonomy that those in North India, Rehman and Rao (2004) drawsomewhat different conclusions. They find that village exogamy is commonin both North and South India and has mixed effects on female autonomy.Consanguinity, on the other hand, is seen to have negative effects ratherthan the positive ones identified by Dyson and Moore (1983).

In their earlier summing up of the debate, Jean Dreze and Amartya Sen(1996) argued that the highly unequal gender relations that exist in manyparts of the country are reflected in very low female labor-forceparticipation, a large gender gap in literacy rates, extremely restrictedfemale property rights, strong boy preference in fertility decisions,2 andwidespread neglect of female children (Dreze and Sen 1996: 142).

This paper makes three main contributions to the literature. First, while alarge and growing literature looks at the factors influencing the incidenceof child work including household poverty status, household asset owner-ship, and characteristics of the child (age and gender), much less has beensaid about the impact of female autonomy in the region on child work.Second, the literature on female autonomy has focused on its impact onfertility, infant mortality, child health, and household-expenditure patterns.However, few scholars have examined the impact of female autonomy onchild school attendance and work participation. The author knows of only ahandful of papers in this specific area (Kaushik Basu and Ranjan Ray 2002;Farzana Afridi 2006; Geoffrey Lancaster, Pushkar Maitra, and Ranjan Ray2006). Third, despite Agnes R. Quisumbing and John A. Maluccio’s (1999)conceptual broadening of female autonomy to include the characteristicsof the extended family and of the kinship system, most applied studies (withthe exception of Øystein Kravdal [2004]) of female autonomy have tendedto concentrate on autonomy within the household. Applied to India, wherelarge interregional differences in female autonomy exist, such an extensionis both interesting and fruitful, allowing this study to test the influence bothof individual autonomy characteristics and of the environments in whichthe women live. The paper assumes that regional differences in genderequity help establish norms that many households find difficult to ignore.3

DATA

The data analyzed here are from Round 50, Schedule 10 of the house-hold socioeconomic survey conducted by the National Sample Survey

CHILD SCHOOLING AND WORK

79

Page 4: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Organisation (NSSO) in India (1993). The data set is large and complex,covering all the states and union territories in India. It includes socio-economic information for 356,352 individuals belonging to 69,231 house-holds in rural India. While Round 50 of the survey focused on consumerexpenditure and employment, Schedule 10 within it concentrates oneducation and employment issues and offers detailed information on theeducational status and economic activity of members of each of thehouseholds (NSSO 1993). The data set thus provides exhaustive infor-mation on children’s activities and on the education of parents as well astheir current and usual employment including their occupation, hoursworked, and wages earned. This study also obtained information relatingto the Gender Equity Index of the various states from the HumanDevelopment Report for India (Planning Commission, Government ofIndia 2002).

This study defines children as those between 5–15 years of age, whichconforms to the decision put forward by the International Labour Organi-zation (ILO) and the United Nations Children’s Fund (ILO 2009).4 Sincethe paper focuses on child labor, the under-5 category is not considered.The current analysis concentrates on a sample of 93,825 children.Appendices A and B provide summary statistics of the binary variables(Appendix A) and continuous variables (Appendix B) used in the analysis.Although the data from the NSSO is rich and comprehensive, particularlyfor a household data source, some limitations with regard to the measuresfor child work need to be kept in mind. In rural areas child work is oftenhighly seasonal and may be misreported. If it occurs in conjunction withschooling, there is potential for ambiguity when the principal andsecondary activity statuses of children are recorded (Shakti Kak 2004: 50).

BACKGROUND TO SCHOOLING AND CHILD WORK ININDIA

While free public schools exist in most regions, parents whose childrenattend them incur considerable hidden costs for transport, uniforms,books, and tuition fees. Thus, Jandhyala B.G. Tilak (2002) found thathouseholds with children in school spent approximately 2.93 percent ofhousehold income on education, with the proportion being 3.16 percentfor boys and 2.57 percent for girls.

Turning to consider the school and work participation of children, wedefine work in this paper as including only market-based activities. Basedon this definition, Table 1 indicates that 59 percent of girls and 72 percentof boys only go to school, while 5 percent of girls and 7 percent of boys onlywork. These figures and those in the rest of this paper relate to a child’sprincipal activity. To identify the activities being undertaken by each child,we consider the Usual Principal Activity Status, which indicates the main

ARTICLES

80

Page 5: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

activity that the person is engaged in. According to the NSS, the usualactivity status relates to the activity of a person during the reference periodof 365 days preceding the date of survey. The activity in which a personspent more time during the year preceding the date of survey is the onethat is considered to be the primary activity status of the person. Inaddition, the dataset also provides the Usual Subsidiary Activity Status ofthe child. This variable indicates whether children are doing more thanone activity. A child whose principal activity is determined on the basis ofthe major time criterion may also have pursued some other economicactivity for thirty days or more during the reference period of 365 dayspreceding the date of survey. This is identified as the secondary activityof the child. In our dataset, a very small proportion of girls (0.86 percent)and boys (1.62 percent) were involved in more than one activity in 1993. Inthis paper, therefore we concentrate entirely on the main activity of thechild.

Appendix A provides summary statistics for the levels of schooling andwork in regions with different levels of gender equity. It shows that inregions with high gender equity, on average, 74 percent of children haveschooling as their primary activity, and 9 percent have work as their primaryactivity; in areas with low gender equity the corresponding statistics are 65percent (in school) and 5 percent (working). Clearly, higher proportionsof children work and go to school in regions with higher gender equitythan in regions with lower gender equity. While the schooling statistics areas expected, that is, more children go to school in regions with greatergender equity, the work statistics are unexpected. They indicate that morechildren also work in more equitable regions. To consider whether thesedifferences in percentages of working children are significant, we testwhether these patterns hold up after controlling for household character-istics. Even greater differences exist between families living above thepoverty line and those below, with 59 percent of children in the poorer

Table 1 School, work, and household chores done by children in rural households inIndia (for age group 5–15 yrs)

No. of girls Girls (%) No. of boys Boys (%)

School 25858 59.2 36208 72.2Work 2310 5.3 3628 7.2Chores 4684 10.7 358 0.7More than 1 principal activitya 605 1.4 894 1.8None 10215 23.4 9082 18.1TOTAL 43672 100 50170 100

Notes: aThis variable denotes children who do more than one principal activity where the principalactivity accounts for a certain number of hours of a child’s time during the week. This classification isdifferent from that of subsidiary activities used in Table 3.

CHILD SCHOOLING AND WORK

81

Page 6: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

families attending school while 81 percent of children from families abovethe poverty line do so. Similarly, 8 percent of children in the poorerfamilies work compared with 5 percent in the families that are not poor.Further details about the data underlying these statistics are provided in thesection on Empirical Estimation. Thus both regional gender equity andhousehold income/expenditure have statistically significant effects onpercentage of children participating in school or the labor market in acertain region. As indicated above, we will consider whether this resultholds up once we control for parental (and other family) characteristics.

In considering these figures, it is important to note there are incentives tounderreport child work in India. First, many types of child work are illegalin India. The Indian Constitution prohibits child work in certain sectorsand in many hazardous industries (the Indian Child Labour Prohibitionand Regulation Act [Government of India 1986]). The Act also regulatesthe number of hours worked by children and the conditions in which theywork. Thus, children are not allowed to work in two establishements on thesame day; they are not permitted to work more than three hours without abreak; and employing children at night (between 7 pm and 8 am) is notpermitted. However, the Indian government has not attempted to abolishlabor by children under the age of 14 years and most laws rarely extend tothe rural informal sector where children are employed on farms, oftenunder parental supervision. Therefore, while state attempts to regulatechild labor might cause some underreporting in the current sample, itwould be surprising if the effect were marked. Second, some under-reporting may be the result of a household’s attempt to take advantage ofthe midday meal scheme in schools (Kak 2004). Thus, households maysend children to school for only part of the day and keep them at work forthe rest of the day. Third, in these statistics children who are engaged inhousehold chores are not reported as working. Instead, they are reportedunder a separate category of household chores. Thus, in Table 1, we cansee that 10 percent of girls and 0.7 percent of boys indicate that theirprimary activity is doing household chores. Finally, many children (23percent of girls and 18 percent of boys) are reported neither as going toschool nor as working. Instead, this category of children may well be thosewho would work if employment existed but are not able to do so becauseof labor market conditions (Uma S. Kambhampati and Raji Rajan 2006,2008).

THEORETICAL BACKGROUND AND ESTIMATION ISSUES

Traditionally, economists undertook analyses of household behavior withinthe unitary model of the household (Gary S. Becker 1965), which saw thehousehold as a single altruistic unit in which decisions were made by thehousehold head. In essence, it assumed the congruence of decisions made

ARTICLES

82

Page 7: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

by different members of a household. By the late 1980s, however, theunitary model was overtaken by models that argued that the decisions madewithin households varied according to whether a father or a mother madethem. This literature developed within a game theoretic framework inwhich household members could be seen as playing a bargaining game(Marilyn Manser and Murray Brown 1980; Marjorie B. McElroy and MaryJean Horney 1981) or as negotiating to achieve some form of efficiency(Pierre-Andre Chiappori 1988, 1992). This paper also tests the hypothesisthat mothers’ and fathers’ wages have different impacts on householddecisions (Hypothesis 1). If proved, this hypothesis will allow for therejection of a unitary household model of decision making and confirmthat some bargaining is occurring within the households.

A recent advance in this literature has been the recognition that thebargaining power held by different household members is itself endogen-ous. Thus, Kaushik Basu (2006) argues that female labor supply is both afactor in household decision making and a determinant of the householdbalance of power. In a hypothetical, heterosexual, nuclear household, witha woman who is only interested in spending on one good (he calls it milk)and a man on another (alcohol), Basu assumes that both the man and thewoman would find it painful to send their child out to work. Maximizingthe household’s utility function subject to a budget constraint that includesthe income earned by the child, Basu finds that as the woman’s powerincreases, the household will spend more and more of its income on thegood for which she has a preference (milk). The opposite will be true as theman’s power within the household increases. This conclusion providesthe intuition behind Basu’s results. When all the power in the householdrests with one agent (whether the man or the woman), the child present ismore likely to work, because this single agent reaps all the benefits of theadded income (in terms of increases in the goods for which that individualhas a preference).5 However, when the power is equally divided betweenthe man and the woman, a single agent does not reap all the benefits of anincrease in household income, and therefore the child is less likely to work.This reasoning leads to the second hypothesis (Hypothesis 2) tested in thispaper: that when the household balance of power in terms of relative wagesof the spouses and their relative education levels tilt in favor of the mother,there will be a decrease in children working and an increase in theprobability of children going to school. In this context, this paper tests forthe possibility that the impact of the woman’s contribution to householdexpenditure is not linear.

Finally, most studies have concentrated on female autonomy withinhouseholds. As noted earlier, Quisumbing and Maluccio (1999) broadenthe notion of autonomy, arguing that bargaining power within a householdis determined by control over resources, influences over the bargainingprocess, mobilization of interpersonal networks, and basic attitudinal

CHILD SCHOOLING AND WORK

83

Page 8: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

attributes. They also argue that legal rights, education levels, or bargainingskills may influence the bargaining process and that ‘‘in societies wherethe extended family is a key player in intra-household allocation, such asthose in South Asia, the characteristics of the extended family may affectintra-household allocation outcomes’’ (Quisumbing and Maluccio 1999:10). In a study of child mortality in India, Kravdal (2004) finds that notonly the education level of mothers but also the average education levelof women in the area have a statistically significant impact on childmortality. In this paper, we hypothesize that the extent of female autonomyin the region will also influence child work and school participation(Hypothesis 3).

METHODOLOGY

To summarize, this paper aims to test three hypotheses arising from theliterature:

Hypothesis 1: Fathers and mothers make similar decisions aboutchild welfare (as reflected in child schooling and work in thisstudy). This would be similar to arguing that the household is aunitary one where all incomes are pooled and all decisions are jointlymade.

Hypothesis 2: Mothers with greater autonomy within the householdmake decisions that will increase child schooling and decrease childwork.

Hypothesis 3: The gender equity conditions that exist in a region playan important role in determining the probability of child schoolingand work.

To address these issues, this study estimates a bivariate probit model ofchild work and schooling in India. The model is estimated separately forboys and for girls and for children living in households above and below thepoverty line.6

The mother’s autonomy within the household is proxied by includingher relative monetary contribution to the household as well as her edu-cation relative to the father’s. The former makes a good proxy for themother’s influence because there might well be many sources of income(both wage and non-wage), and it is the mother’s monetary contribution tooverall household expenditure that is likely to determine how much powershe weilds in decision making.7 We also include the mother’s absoluteeducation level as well as her education relative to that of her spouse. Anyinfluence she derives from the kinship system in the region or from thesociocultural environment is captured by the inclusion of a state-level

ARTICLES

84

Page 9: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

gender equity index. While this index allows for variation in gender equityacross states, its inclusion makes the implicit assumption that gender equitydoes not vary within states. This is, of course, not entirely appropriate.However, it is the most disaggregated level at which the index is currentlyavailable (Planning Commission, Government of India 2002).8 Themeasurement of female autonomy in the study is therefore quite limited.It is possible, for instance, that the presence of other household members,especially a mother-in-law as well as the characteristics of the mother-in-law(her education and employment), may influence the mother’s autonomyin the household. Similarly, the education of fathers may increase femaleautonomy in the household. Our measures are therefore only a firstapproximation to the extent of female autonomy that exists within thehousehold.

Empirical estimation

To consider the impact of female autonomy on child work and schooling,this study estimates a standard bivariate probit model in which school andwork are two binary dependent variables specified according to theprincipal activity status of the child (see Appendix A). A child can only haveone principal activity (unless the child spends exactly half the time in eachactivity). As Table 1 indicates, there is a very small proportion of children inthis category (1.8 percent of boys and 1.4 percent of girls). The child canalso be engaged in one or more subsidiary activities but again, there are fewchildren who do this. 0.86 percent of girls and 1.62 percent of boys areengaged in more than one activity. The vast majority of children, therefore,are engaged in only one activity. For the purposes of this paper, we areinterested in the probability of this activity. Child work is said to occurwhen the principal activity of the child refers to any one of those activitiescategorized as ‘‘employment’’ within the data. Here the dependent vari-able, Work, is coded 1 if the child is working and 0 otherwise. When theprincipal activity of the child refers to attending educational institutions thechild is categorized as attending School (School¼ 1). This classification isbased on parents reporting children’s activities.9

This paper divides the sample by gender as well as by poverty status. Forthe second category it uses the poverty line set by the Indian government in1992 of Rs.296 per capita per month in urban areas and Rs.276 per capitaper month in rural areas. The poverty line is a per capita figure. Since ourdata in this paper relates to the rural sample in 1992/3, it is the poverty linefor the rural sector in 1992 that is the appropriate one. The current studyalso uses per-capita expenditure rather than income, as is the norm in theliterature, because the expenditure figure takes into account informalincome sources and provides a longer-term income profile, one notaffected by short-term changes in income levels. Households with monthly

CHILD SCHOOLING AND WORK

85

Page 10: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

per-capita expenditures above the official poverty line are in the abovepoverty line sample. These divisions result in four subsamples: girls inhouseholds above the poverty line, girls in households below the povertyline, boys in households above the poverty line, and boys in householdsbelow the poverty line. Separate estimations for each subsample allow thisstudy to determine whether the impact of the female autonomy variablesvaries according to the gender of the children as well as across the povertyclasses – for example, might women among the poorer groups have greaterautonomy or might women have greater autonomy over decisions relating todaughters rather than sons?

Variables included

Although the primary concern here is the influence mothers may have onthe probability of child work, this study also controls for personal charac-teristics of the child (including age and sex), for household traits (such asreligion, social status, illiteracy rates of both male and female residents,number of adult dependants, land ownership, and debt status), and forregional characteristics (average village wages and regional dummies).Included also are those variables that reflect maternal autonomy at twolevels: the autonomy of women in general in the region and the autonomyof mothers within their households (see also Appendix A and 1b). Theformer is reflected in the Gender Equity Index, a measure of femaleautonomy devised by the United Nations Development Programme(UNDP) and measured across Indian states, while the latter is proxied byincluding the mother’s own education and employment characteristics.Thus, mothers’ education levels (primary, secondary, and tertiary) andmothers’ wages are both included, as are mothers’ contributions tohousehold expenditure and mothers’ education levels relative to fathers’.Finally, each of these variables is interacted with the Gender Equity Indexto capture whether educated mothers who live in regions with greaterfemale autonomy have different impacts on child work and schooling thaneducated mothers in areas where women have limited autonomy. Therationales underlying these variables are discussed in detail below.

Autonomy of women in the region

There are great differences in the levels of autonomy women enjoy indifferent parts of India, as reflected by the fact that their literacy andemployment levels vary according to region. The Gender Equity Indexcaptures the disparities between men and women in education, health,employment, and income: the higher the index, the more equitable aregender relations. The contrast between Kerala, a state in the country’ssouthwestern tip, which in 1991 had a Gender Equity Index of 0.825, and

ARTICLES

86

Page 11: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Bihar, in the northeast, with a 1991 index of 0.469 epitomizes these regionaldifferences in Indian women’s autonomy. If female autonomy in a regionincreases the welfare of children within individual households, then wewould expect that rates of child work are lower and rates of child schoolingare higher in regions where the gender equity index is high. However,female autonomy in the region may not have such a straightforward impacton child work and schooling, particularly because the definition of childwelfare this study employs (more school and less work) might not match theneeds of individual households. Households functioning under an incomeconstraint may not be able to afford to keep children out of work. Thus, inhigh Gender Equity Index regions, there is on average more child schoolingbut also more child work. This is because these are regions where adultwomen are better educated and also more likely to work. They are thereforelikely to be aware of the importance of education for their children and toreinforce this. However, given the household’s income constraint, and giventhat these women are also working, they are better placed to introduce theirchildren to the labor market. Therefore, while at first glance one mightexpect increased schooling and decreased work, this may not be the finaloutcome. The impact of the Gender Equity Index variable may depend onthe characteristics of the individual mothers (their education and employ-ment) and of the households (spouse education, employment, socialgroupings, number of dependents, etc.) they operate in. To allow for theseeffects, this study interacts the index with the mother’s education andemployment variables.

Individual characteristics of parents: Mother’s education

In all countries, better-educated parents are generally assumed to havegreater abilities and incentives than less-educated ones to improve theirchildren’s educations. They are also considered more likely to valueeducation. Mark R. Rosenzweig and Kenneth I. Wolpin (1982) argue thatthere is a strong intergenerational transfer of educational achievementfrom parents to children. To allow for this, both father’s and mother’s levelsof education are included in the model. They are included as three sepa-rate binary categorical variables (mother’s primary education, mother’ssecondary education, and mother’s tertiary education; and father’s primaryeducation, father’s secondary education, and father’s tertiary education)that identify primary, secondary, and tertiary (higher) education,10 withuneducated mothers and fathers being the excluded category. We expect toshow that higher levels of parent education result in increased childschooling and decreased child work because we assume that educatedparents place an intrinsic value on their children’s educations.

In the high Gender Equity Index regions, the proportion of mothers withany education is higher (see Appendix A). This is also true in households

CHILD SCHOOLING AND WORK

87

Page 12: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

above the poverty line compared to those below the poverty line. Since thisstudy also considers whether the impact of the education variable varies inregions where female autonomy exists relative to those where it does not, itinteracts this variable with the Gender Equity Index of the region byincluding three related variables as follows:

Female autonomy within a household¼Gender Equity Index6mother’seducation, where mother’s education is defined as primary, secondary, ortertiary education.

This study proposed that the impact of gender equity would reinforcethat of mothers’ education. Thus, we expect that if a mother’s educationinclined her toward more education for her children, when this occurredtogether with regional gender equity, the latter would empower the motherto work toward fulfilling her preference for better-educated children.

Individual characteristics of parents: Employment and wages

Mothers’ wages increase household incomes and could decrease the needto send children out to work. This variable is likely to be endogenous andhas therefore been instrumented (see Table 2).11 The higher a mother’swage, given all other wages in the household, the higher child schoolingwould be expected to be and the lower child work would be expected to be,if India is like other countries.

The inclusion of mothers’ wages also allows this study to considerwhether the source of household income has an impact on child schoolingand child work: that is, do wages earned by fathers have the same impact onthe probabilities of work and schooling for their children as those earnedby mothers? Many writers studying developing countries argue that a higherproportion of mothers’ wages is spent on goods for children and a higherproportion of fathers’ wages on so-called adult goods like alcohol andcigarettes (John Hoddinott 1992; see also Cheryl R. Doss [1996a] for therole played by assets in determining female autonomy and householdexpenditure patterns). In the current analysis, this claim implies that ahigher proportion of mothers’ wages than of fathers’ will be spent onschooling and preventing child work. This study tests whether this is thecase by formally testing in its model whether the coefficient of mother’swage is equal to that of father’s wage. A rejection of this hypothesis wouldimply a rejection of the unitary household model.

Autonomy within the household

This study uses two variables to capture the autonomy mothers have indecision making compared with that of fathers in the household. Thesevariables relate to the mothers’ own education and employment charac-teristics relative to those of fathers.

ARTICLES

88

Page 13: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Tab

le2

To

bit

and

sam

ple

sele

ctio

nes

tim

atio

no

fin

stru

men

tsfo

rfa

ther

’san

dm

oth

er’s

wag

es

Tob

itSa

mpl

ese

lect

Mot

her’

sw

age

Fath

er’s

wag

eM

othe

r’s

wag

eFa

ther

’sw

age

Var

iabl

eC

oeffi

cien

tSt

anda

rder

ror

Coe

ffici

ent

Stan

dard

erro

rV

aria

ble

Coe

ffici

ent

Stan

dard

erro

rC

oeffi

cien

tSt

anda

rder

ror

Co

nst

ant

-384

.74*

**11

.71

-122

3.85

***

22.6

8C

on

stan

t-6

0.34

***

8.43

-230

.79*

**67

.26

Age

12.8

4***

0.57

32.2

8***

0.95

Age

2.56

***

0.39

8.15

**3.

14A

gesq

uar

ed-0

.22*

**0.

01-0

.43*

**0.

01A

gesq

uar

ed-0

.03*

**0.

01-0

.09*

**0.

03P

rim

ary

edu

cati

on

-95.

62**

*4.

01-5

9.32

***

3.57

Pri

mar

yed

uca

tio

n-1

1.66

***

1.69

-0.2

14.

42

Seco

nd

ary

edu

cati

on

-22.

09**

*4.

9232

.79*

**4.

00Se

con

dar

yed

uca

tio

n64

.87*

**2.

5567

.23*

**6.

19

Ter

tiar

yed

uca

tio

n36

.85*

**7.

3937

3.89

***

5.64

Ter

tiar

yed

uca

tio

n13

9.57

***

3.94

296.

55**

*14

.09

Vil

lage

wag

e0.

64**

*0.

0610

2.71

***

1.54

Vil

lage

wag

e14

.01*

**0.

2725

.71*

**1.

12L

and

-0.0

5***

0.00

2-0

.12*

**0.

003

Lan

d-0

.01*

**0.

001

-0.0

4***

0.01

Sigm

a22

8.71

***

1.84

351.

85**

*1.

59L

AM

BD

A2.

132.

260.

010.

80L

Mte

st[d

f]fo

rT

ob

it35

09.3

8(8

)22

679.

80(8

)R

squ

ared

0.22

0.19

An

ova

bas

edfi

tm

easu

re10

.22

0.30

Ad

j-Rsq

uar

ed0.

220.

19

Dec

om

pb

ased

fit

mea

sure

0.49

0.36

Fte

st26

51.4

(0.0

)75

1.32

(0.0

)

Lo

gli

keli

ho

od

chi-s

qu

are

test

v.h

igh

(0.0

)54

04.2

2(0

.000

)

Not

es:*

**d

eno

tes

sign

ifica

nce

at1

per

cen

t,**

at5

per

cen

t,an

d*

at10

per

cen

t.In

the

fath

ereq

uat

ion

,age

and

edu

cati

on

refe

rto

the

fath

er,t

he

vill

age

wag

eis

the

aver

age

mal

evi

llag

ew

age,

and

lan

dis

ho

use

ho

ldla

nd

ho

ldin

gs.

Inth

em

oth

er’s

wag

eeq

uat

ion

,ag

ean

ded

uca

tio

nre

fer

toth

em

oth

er,

the

vill

age

wag

eis

the

aver

age

fem

ale

vill

age

wag

e,an

dla

nd

ish

ou

seh

old

lan

dh

old

ings

.

CHILD SCHOOLING AND WORK

89

Page 14: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Education relative to spouse

This variable is included as a measure of female autonomy within thehousehold. While mother’s primary education, mother’s secondaryeducation, and mother’s tertiary education proxy a mother’s preferenceswith respect to education, the mother’s education relative to father’s isassumed to influence her ability to negotiate in defense of her preferenceswith respect to her children’s work and schooling. Appendix A shows thaton average, even mothers in high gender equity regions are less educatedthan fathers, with 0.26 years of education for every year of education thefather has. In fact, this study finds that only 1.3 percent of mothers in oursample have more education than fathers do; 12.3 percent of mothers haveless. As expected, the number of years of education the mother has relativeto her husband is higher in the high Gender Equity Index regions (0.26)and in families living above the poverty line (0.35). Once again, to allow forthe possibility that a mother’s education relative to a father’s may have agreater impact on children’s welfare in regions with more female autonomythan in those with less, this study interacts this variable with the GenderEquity Index. The interacted variable captures the above possibility.

Mother’s contribution to household expenditure

While the impact mothers’ incomes may have is tested by looking atmothers’ wages, holding fathers’ wages constant, looking at mothers’contribution to household expenditure also allows this study to considerwhat happens as mothers increase their contributions to householdexpenditures. Researchers have generally argued that the more a mothercontributes to the household budget, the more bargaining power she willhave within the household. However, this premise is not often tested incountries like India, where traditionally women do not work. Summarystatistics for our sample, for example, reveal that mothers contribute moreto household expenditure in high Gender Equity Index regions (0.19) andin households living below the poverty line (0.19) than in low GenderEquity Index regions (0.13) or more prosperous households (0.096). Thus,both a household’s need to survive and an environment of higher genderequity increase a mother’s contribution to household expenditure. A rise ina mother’s contributions may be tied to a decrease in the father’s; that is,the mother works because she must. In this case, a mother’s contributionscould be interpreted as symptomatic of the marginality of the household.Becoming a breadwinner may increase a mother’s autonomy within thehousehold, but the autonomy of the family in general (and of the motherin particular) outside the household may decrease owing to its reducedcircumstances. Alternatively, the mother’s contributions may increase inthe context of a relatively prosperous household. In such a case, on the

ARTICLES

90

Page 15: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

other hand, a mother’s contributions to expenditure might be interpretedas reflecting increased female autonomy both within the household andin the community. If these two assumptions are true, one might expectthis variable to have a different impact on households above and belowthe poverty line. This study also tests whether this variable (mother’scontribution to household expenditure) has a nonlinear impact on childwelfare, as Basu and Ray (2002) hypothesize by including a quadratic termin it. Since this variable is also likely to be endogenous, just as mothers’ andfathers’ wages are, it is derived from the instrumented mother’s wage.

Finally, we interact the mother’s contribution to household expenditurevariable with the Gender Equity Index variable to detect its impact on childwelfare when other women in the region also have some autonomy (that is,gender equity is high).

RESULTS

We began by estimating two models, one with the Gender Equity Indexalone and the other with the Gender Equity Index interacted with maternalcharacteristics such as education and income. Since a number of theinteraction terms were statistically significant, this study presents only theresults for the latter model.12 Before discussing the results, I will brieflyexplain the instruments estimated for mothers’ and fathers’ wages.

Instruments for father’s and mother’s wages

A major problem for any study of female autonomy is the endogeneity ofwages (see Doss [1996a] for a discussion of this problem and of possiblesolutions). This study corrects for this problem by instrumenting mothers’wages using mother’s age, average village female wages, mother’s edu-cation, and household landholdings (see Table 2 for results of theseestimations). However, the wage data are plagued by relatively large num-bers of zero values. These might arise because the subjects are unemployed,not looking for work, or working in a family enterprise or in a subsistencemanner on the family farm. In all these cases, their wage entry may wellshow a zero value. Estimation using Ordinary Least Squares will result inbiased estimates. I correct for this using both the Sample Selection andTobit methods. The Tobit method corrects for the left truncation of thedata by having a likelihood function with two parts: the first being the LogLikelihood summed over uncensored observations (identical to the loglikelihood for OLS) and the second being the likelihood for the censoredobservations. The second method is the Heckman sample selection model,which models the probability of a variable being zero explicitly and thenincludes the Inverse Mills Ratio from this estimation as an independentregressor into the wage model. This study also instruments fathers’ wages

CHILD SCHOOLING AND WORK

91

Page 16: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

(which are likely to be endogenous) in a similar manner, using the relevantvariables (a father’s age, village male wages, a father’s education, andhousehold landholdings). Table 2 presents the results for both estimations.

The diagnostics (see LM Test for Tobit) reject the Tobit model for bothfathers and mothers. This test considers whether the log likelihood of theTobit model is significantly different from the sum of the log likelihood forthe constituent Probit and truncated regressions. The result (a chi-squarevalue of 3,509 with 8 degrees of freedom) indicates that we can reject thehypothesis that a Tobit model fits with 99 percent probability (the criticalchi-square value for 95 percent probability being 15.51). We therefore usethe sample selection predictions as instruments for wage equations in therest of the paper.

The results of the sample selection estimation indicate that mothers’wages increase with age but the rate of growth tapers off. While primaryeducation decreases the wage earned by mothers, secondary and tertiaryeducation have the expected positive effects. The higher the average villagefemale wage, the higher a mother’s wage is; however, in households thatown land, the mother’s wage is lower. Fathers’ wages, too, increase with ageand with average village wage. Village wage affects child schooling andlabor largely via its impact on adult wages, and including it as a determinantmakes this channel of causation explicit. While primary education has nostatistically significant impact on fathers’ wages, both secondary and tertiaryeducation have a positive and statistically significant impact on this variable.Finally, fathers’ wages also decrease when the household owns land. Theseresults are all highly statistically significant and are as expected. Theinsignificance of lambda implies that selection is not a significantdeterminant of wages. The Log Likelihood Chi Square test for the modelconfirms that it is highly statistically significant, as does the F test of the jointsignificance of the coefficients. We therefore conclude that our instru-ments for mother’s wage and father’s wage are good.

Results for child schooling and work

The estimated instruments for fathers’ and mothers’ wages are included inthe bivariate probit model for school and work. The marginal effects fromthis model are shown in Table 4, which presents only the coefficients thatare pertinent to the hypotheses in this paper. The full set of results(including the controls) is given in Appendix 2. Since many of the variablesof interest have both a direct and an indirect effect (through the GenderEquity Index) this paper considers the size of the net effect separately inTable 6 and in the section ‘‘Regional and household autonomy: The netimpact.’’

We test the division of the sample into four sub-groups – girls below thepoverty line, girls above the poverty line, boys below the poverty line, and

ARTICLES

92

Page 17: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

boys above the poverty line – using the Likelihood Ratio test (see Table 3).The results confirm that the separation of the sample by gender as wellas by poverty status is appropriate. The Chi-Square test shows that theestimation is significantly different statistically for boys and for girls and alsofor children in households above and below the poverty line.

I will discuss the results of the four subsamples in the context of thehypotheses set out previously.

Hypothesis 1: Fathers and mothers make symmetric decisions with regard to childwork and schooling

The influence of fathers and mothers on these decisions is captured in thevariables relating to their wages and to their individual education levels.The results (Table 4) indicate that fathers’ and mothers’ wages have verydifferent impacts. While a higher mother’s wage significantly increases theprobability of schooling for both boys and girls below the poverty line, thesize of the coefficient is very small. Mothers’ wages would have to rise byRs.100 (from an average of less than Rs.50 for all four subsamples) toincrease the probability of schooling by 0.3 percent. Fathers’ wages do nothave a statistically significant influence on schooling in any of the four sub-samples.

On the other hand, the results indicate that a rise in mothers’ wagesincreases the probability that girls will work in households both above andbelow the poverty line, while an increase in fathers’ wages increases theprobability of work only for boys in households below the poverty line.

Table 3 Testing the division of the sample into subgroups: likelihood ratio tests

URSS RRSS RRSS-URSSLR¼ 2

(RRSS-URSS)Probability(w25LR)

Model for boys: above andbelow poverty line beingequal

-19515.10 -10694.30 8820.77 17641.54 0.999

Model for girls: above andbelow poverty line beingequal

-20510.50 -11476.20 9034.28 18068.56 0.999

Model for children abovepoverty line: boys andgirls equal

-12357.00 -6386.46 5970.56 11941.12 0.999

Model for children belowpoverty line: boys andgirls equal

-27673.60 -15829.30 11844.31 23688.62 0.999

Notes: The null hypothesis is that running two separate models is equivalent to running a modelacross the two subsamples. URSS¼unrestricted sum of squares of the two separate models.RRSS¼Restricted sum of squares of a model in which all coefficients are constrained to be equal inthe two subsamples.

CHILD SCHOOLING AND WORK

93

Page 18: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Tab

le4

Th

eim

pac

to

fm

oth

er’s

edu

cati

on

and

emp

loym

ent:

mar

gin

alef

fect

s(w

ith

inte

ract

ive

term

sw

ith

GE

I)–

sub

set

of

resu

lts

Gir

lsbe

low

pove

rty

lin

eB

oys

belo

wpo

vert

yli

ne

Gir

lsab

ove

pove

rty

lin

eB

oys

abov

epo

vert

yli

ne

Var

iabl

eC

oeffi

cien

tSt

anda

rder

ror

Coe

ffici

ent

Stan

dard

erro

rC

oeffi

cien

tSt

anda

rder

ror

Coe

ffici

ent

Stan

dard

erro

r

SCH

OO

LSC

HO

OL

SCH

OO

LSC

HO

OL

Mo

ther

’sp

rim

ary

edu

cati

on

0.11

0.13

-0.0

10.

14-0

.11

0.15

-0.4

3**

0.16

Mo

ther

’sse

con

dar

yed

uca

tio

n0.

88**

*0.

240.

62**

*0.

260.

30.

240.

20.

24M

oth

er’s

tert

iary

edu

cati

on

2.11

***

0.67

5.16

***

2.42

-0.1

70.

410.

611.

06F

ath

er’s

emp

loym

ent

0.02

0.04

0.09

***

0.04

-0.0

90.

090.

20**

*0.

08M

oth

er’s

wag

e0.

003*

**0.

001

0.00

3***

0.00

1-0

.001

0.00

20.

001

0.00

2F

ath

er’s

wag

e-0

.001

0.00

1-0

.002

***

0.00

10.

000

0.00

10.

001

0.00

1M

oth

er’s

exp

end

itu

reco

ntr

ibu

tio

n-2

.48*

**0.

35-1

.512

***

0.3

-0.3

40.

95-2

.76*

**0.

98

Mo

ther

’sex

pen

dit

ure

con

trib

uti

on

squ

ared

-0.3

10.

43-1

.43*

**0.

461.

652.

043.

792.

79

Fat

her

’sp

rim

ary

edu

cati

on

0.43

***

0.04

0.43

***

0.04

0.17

***

0.05

0.18

***

0.06

Fat

her

’sse

con

dar

yed

uca

tio

n0.

68**

*0.

080.

70**

*0.

080.

40**

*0.

110.

19*

0.12

Fat

her

’ste

rtia

ryed

uca

tio

n0.

91**

*0.

281.

46**

*0.

30.

430.

38-0

.18

0.41

Rel

ativ

eed

uca

tio

n-0

.10.

09-0

.24*

**0.

11-0

.2*

0.12

0.05

0.13

Gen

der

Eq

uit

yIn

dex

-0.5

8***

0.08

-0.2

1***

0.09

-0.6

2***

0.11

-0.3

6***

0.12

Gen

der

Eq

uit

yIn

dex

*mo

ther

’sp

rim

ary

edu

cati

on

0.56

***

0.16

0.69

***

0.17

0.66

***

0.18

0.74

***

0.19

Gen

der

Eq

uit

yIn

dex

*mo

ther

’sse

con

dar

yed

uca

tio

n-0

.53

0.35

-0.0

80.

370.

50*

0.31

0.17

0.33

Gen

der

Eq

uit

yIn

dex

*mo

ther

’ste

rtia

ryed

uca

tio

n-2

.34*

**1.

02-7

.03*

*3.

780.

800.

59-0

.49

1.56

Gen

der

Eq

uit

yIn

dex

*mo

ther

’sex

pen

dit

ure

con

trib

uti

on

3.03

***

0.38

0.93

***

0.43

1.39

1.00

2.29

***

1.13

(con

tin

ued

)

ARTICLES

94

Page 19: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Tab

le4(C

onti

nu

ed)

Gir

lsbe

low

pove

rty

lin

eB

oys

belo

wpo

vert

yli

ne

Gir

lsab

ove

pove

rty

lin

eB

oys

abov

epo

vert

yli

ne

Var

iabl

eC

oeffi

cien

tSt

anda

rder

ror

Coe

ffici

ent

Stan

dard

erro

rC

oeffi

cien

tSt

anda

rder

ror

Coe

ffici

ent

Stan

dard

erro

r

SCH

OO

LSC

HO

OL

SCH

OO

LSC

HO

OL

Gen

der

Eq

uit

yIn

dex

*mo

ther

’sex

pen

dit

ure

con

trib

uti

on

squ

ared

0.44

0.54

2.14

***

0.71

-4.2

33.

12-3

.75

4.89

Gen

der

Eq

uit

yIn

dex

*rel

ativ

eed

uca

tio

n0.

130.

150.

090.

180.

030.

20-0

.01

0.21

Co

ntr

ols

Yes

Yes

Yes

Yes

WO

RK

WO

RK

WO

RK

WO

RK

Mo

ther

’sp

rim

ary

edu

cati

on

-0.3

70.

250.

79**

*0.

270.

250.

290.

46**

0.24

Mo

ther

’sse

con

dar

yed

uca

tio

n0.

60.

640.

710.

630.

130.

43-0

.41

0.53

Mo

ther

’ste

rtia

ryed

uca

tio

n0.

880.

75-2

.99

1705

06.0

6-0

.26

0.75

-0.3

41.

47F

ath

er’s

emp

loym

ent

0.16

***

0.06

0.03

0.05

0.01

0.12

-0.2

***

0.1

Mo

ther

’sw

age

0.00

4***

0.00

1-0

.002

0.00

10.

01**

*0.

003

0.00

10.

003

Fat

her

’sw

age

-0.0

010.

001

0.00

2**

0.00

1-0

.001

0.00

2-0

.001

0.00

2M

oth

er’s

exp

end

itu

reco

ntr

ibu

tio

n-1

.95*

**0.

360.

010.

65-3

.50*

**1.

722.

181.

68

Mo

ther

’sex

pen

dit

ure

con

trib

uti

on

squ

ared

-0.3

8*0.

231.

621.

673.

053.

25-3

.66

6.02

Fat

her

’sp

rim

ary

edu

cati

on

-0.1

3***

0.05

-0.4

2***

0.05

-0.1

0.08

-0.0

90.

08F

ath

er’s

seco

nd

ary

edu

cati

on

-0.2

5***

0.12

-0.6

7***

0.10

-0.3

3**

0.17

-0.2

30.

17F

ath

er’s

tert

iary

edu

cati

on

-0.1

00.

49-1

.31*

**0.

44-0

.16

0.63

-0.2

10.

62R

elat

ive

edu

cati

on

0.11

0.14

-0.0

10.

190.

42**

*0.

190.

090.

21G

end

erE

qu

ity

Ind

ex0.

66**

*0.

121.

120*

**0.

140.

52**

*0.

210.

64**

*0.

17

(con

tin

ued

)

CHILD SCHOOLING AND WORK

95

Page 20: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Tab

le4(C

onti

nu

ed)

Gir

lsbe

low

pove

rty

lin

eB

oys

belo

wpo

vert

yli

ne

Gir

lsab

ove

pove

rty

lin

eB

oys

abov

epo

vert

yli

ne

Var

iabl

eC

oeffi

cien

tSt

anda

rder

ror

Coe

ffici

ent

Stan

dard

erro

rC

oeffi

cien

tSt

anda

rder

ror

Coe

ffici

ent

Stan

dard

erro

r

WO

RK

WO

RK

WO

RK

WO

RK

Gen

der

Eq

uit

yIn

dex

*m

oth

er’s

pri

mar

yed

uca

tio

n-0

.35

0.30

-1.5

9***

0.29

-0.9

8***

0.35

-0.8

9***

0.26

Gen

der

Eq

uit

yIn

dex

*m

oth

er’s

seco

nd

ary

edu

cati

on

-1.8

6***

0.87

-1.2

60.

88-1

.70*

**0.

670.

210.

66

Gen

der

Eq

uit

yIn

dex

*m

oth

er’s

tert

iary

edu

cati

on

-13.

0611

9105

.19

0.27

3635

68.6

3-1

.42*

0.89

0.29

2.00

Gen

der

Eq

uit

yIn

dex

*m

oth

er’s

exp

end

itu

reco

ntr

ibu

tio

n2.

36**

*0.

450.

580.

952.

751.

83-2

.22

1.67

Gen

der

Eq

uit

yIn

dex

*m

oth

er’s

exp

end

itu

reco

ntr

ibu

tio

nsq

uar

ed

0.56

*0.

32-2

.49

2.48

-0.7

15.

170.

618.

76

Gen

der

Eq

uit

yIn

dex

*re

lati

veed

uca

tio

n0.

35*

0.21

-0.1

40.

230.

140.

3-0

.21

0.31

Co

ntr

ols

Yes

Yes

Yes

Yes

Not

es:*

**d

eno

tes

1p

erce

nt

stat

isti

cals

ign

ifica

nce

,**

for

5p

erce

nt,

and

*fo

r10

per

cen

t.R

esu

lts

rela

tin

gto

the

con

tro

lvar

iab

les

hav

eb

een

excl

ud

edfr

om

the

tab

leto

mak

eit

less

un

wie

ldy.

Th

eva

riab

les

incl

ud

eag

e,ag

esq

uar

ed,

sex,

sex

of

ho

use

ho

ldh

ead

,b

irth

ord

er,

nu

mb

ero

fsi

bli

ngs

,H

ind

u,

Mu

slim

,sc

hed

ule

dca

ste

and

trib

ed

um

my,

deb

tst

atu

so

fh

ou

seh

old

,fem

ale

and

mal

eil

lite

racy

leve

lsw

ith

inh

ou

seh

old

s,am

ou

nt

of

lan

dh

eld

,nu

mb

ero

fd

epen

dan

tso

lder

than

60ye

ars,

vill

age

wag

e,an

da

du

mm

yto

ind

icat

ere

gio

n(S

ou

tho

rN

ort

h).

ARTICLES

96

Page 21: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Again, the magnitude of the coefficients is relatively small. These resultsconfirm the findings of other studies that show mothers’ employment iscomplementary with daughters’ employment (Olga Nieuwenhuys 1996).Complementarity also seems to exist between the employment of boys andtheir fathers in households below the poverty line.

To formally test whether mothers’ and fathers’ wages have differentimpacts on the probability of child work and schooling, this study uses aWald test of the restrictions that the coefficients of mothers’ and fathers’wages are insignificantly different in the various subsamples. Our results(see Table 5) indicate that fathers’ and mothers’ wages have a significantlydifferent impact in statistical terms on boys in households both aboveand below the poverty line but a very similar impact on girls. Thus, thehousehold is clearly not unitary in all dimensions. While parents’ wagesseem to have a symmetric impact on girls, they are not symmetric in theirimpact on boys.

Hypothesis 2: Mother’s autonomy within the household increases childschooling and decreases child work

As mentioned, this study measures female autonomy using a mother’scontribution to household expenditure and her education level relative tothat of a father in the same household. Turning first to consider whethermothers’ education relative to fathers’ has a statistically significant effect,this study finds that, contrary to expectations, increases in mothers’education relative to fathers’ significantly decreases the probability ofschooling of girls in households above the poverty line (by 20 percent) andof boys in households below this line (by 24 percent). Thus, while positivechanges in a mother’s absolute education have a positive impact on childschooling and education in households below the poverty line, increases inmothers’ education relative to fathers’ have a negative impact on childschooling and education. Mother’s relative education has no statisticallysignificant impact on work probabilities except in the subsample of girls in

Table 5 Wald test for the equality of the coefficient of mother’s and father’s wages

SampleWald Statistic for

b(Mother’s Wage) - b(Father’s Wage)¼ 0Probability fromChi-Squared [1]

Girls below poverty line 0.10 0.75Boys below poverty line 4.82 0.03Girls above poverty line 0.95 0.33Boys above poverty line 5.66 0.02All below poverty line 2.47 0.12All above poverty line 0.77 0.38All girls 0.001 0.98All boys 10.20 0.001

CHILD SCHOOLING AND WORK

97

Page 22: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

households above the poverty line. In this subsample, the higher the levelof education a mother has attained compared with that of the father, themore likely their daughter is to work. Looking next at whether the impactof this variable varies by the level of gender equity prevalent in the state,this study finds that the interaction term is statistically insignificant in allcases (except one) for both school and work. Thus, while the impact of themother’s absolute education level varies with the gender equity of the state(see discussion on Hypothesis 3), the impact of her education level relativeto her spouse’s does not. In the only case where the latter factor isstatistically significant (girls in households above the poverty line), itactually decreases schooling probabilities and increases work probabilities,though this variable is statistically insignificant in all other cases.

The other proxy for female autonomy that we include in our model ismother’s expenditure contribution. The results show that as mothers’ con-tributions to household expenditure increase, the probability of schoolingin three out of four subsamples decreases. The exception is for girls abovethe poverty line. This study also finds that while a rise in mothers’ wagesalone increases the probability of school for both girls and boys below thepoverty line, an increase in mothers’ contributions to household expendi-ture has the opposite effect: it decreases the probability that boys and girlswill attend school. However, as mothers’ expenditure contribution increa-ses, the probability of work for girls in households both above and below thepoverty line decreases. The results therefore indicate that greater autonomyfor the mother as reflected in higher household expenditure contributionsactually decreases schooling in the cases where such contributions have astatistically significant impact; on the other hand, greater female autonomydecreases the probability of work for girls. Note also that the quadratic termis statistically insignificant in six of eight estimations and in both cases whereit is significant, the impact of the quadratic term is to reinforce rather thanmitigate the effect of the linear term.

In examining whether the impact of this variable varies across stateswith different levels of gender equity, this study finds that, in three of thefour subsamples, when both gender equity and increases in a mother’scontribution to household expenditure are high, the probability ofschooling for both boys and girls is high. Thus, the autonomy womenderive from their household expenditure contributions is reinforced byregional equity levels. This study also finds that this variable does not have astatistically significant impact on the probability of work for any subsampleof children, except for girls below the poverty line. For this group, theprobability of work increases when mother’s contribution to householdexpenditure and the Gender Equity Index are both high.

Thus, one cannot straightforwardly accept or reject Hypothesis 2. Thepattern depends on the subsample – girls/boys, children living in house-holds above and below the poverty line, etc. – and on mothers’ autonomy

ARTICLES

98

Page 23: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

within the household. Note, however, that mothers’ autonomy within thehousehold does not automatically incline mothers toward seeking moreeducation and less work for their children. One possible interpretation ofthis finding might be that mothers who seemingly have greater autonomywithin the household may actually be highly constrained externally. Underthe constraints posed by their economic circumstances, both mothers andfathers make similar decisions regarding child work and schooling.

Hypothesis 3: Regional gender equity increases child schooling anddecreases child work

The results indicate that holding all other factors constant, an increase inthe Gender Equity Index decreases the probability of child schooling andincreases the probability of child work in all four subsamples. Sensitivityanalysis estimating two versions of the bivariate school and work model –first with the Gender Equity Index as the only variable in the model andsecond, with this index in the model together with all other variables butexcluding the interaction terms – confirms that this is a robust result. Ofthe six subsamples for which we estimated the impact of the Gender EquityIndex on school and work, the effect was negative in five cases. Theseresults therefore confirm that even relatively empowered mothers mayprefer to send their children to work rather than to school. The reasons forthis choice are beyond the scope of the current study. However, to testwhether it is caused by a lack of schools in the region, this study considersthe correlation between school availability (number of schools per 1,000children) and the Gender Equity Index. The results indicate a correlationof 0.568 for primary schools and 0.523 for upper schools. Though thiscorrelation is not very high, it is clearly positive. There is therefore noindication that the lower levels of child schooling in high Gender EquityIndex regions might arise from a lack of schools in these regions. Itmight therefore simply be that mothers see better opportunities for theirchildren in employment than in schooling – that is, the education availableis of poor quality or gives poor returns (Jean Dreze and Haris Gazdar 1997).

Of course, the Gender Equity Index has more than a stand-alone impact.Its effect is mediated through the level of education and the employmentcharacteristics of mothers within the household. Thus, a mother’s edu-cation may have a different impact on regions where few women have goneto school than it does where the vast majority have received someeducation. In the latter case, the mother’s autonomy is reinforced by theautonomy of other women in the region. While the stand-alone impact of amother’s education on child school probabilities was positive in householdsbelow the poverty line, its impact as mediated through the Gender EquityIndex of a given state is more complicated. The coefficients of theinteraction variables [Gender Equity Index * mother’s primary education;

CHILD SCHOOLING AND WORK

99

Page 24: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Gender Equity Index * mother’s secondary education; Gender EquityIndex * mother’s tertiary education] indicate that mother’s primary andtertiary education have a statistically significant impact on child schoolingbut mother’s secondary education does not. Our results indicate thatmothers with primary education living in regions with higher gender equityincrease the probability of children going to school. When mothers havetertiary education in regions with higher gender equity, then theprobability of schooling for boys and girls in households below the povertyline is significantly lower than for mothers with tertiary education in regionswith lower gender equity.13 Thus, mothers with tertiary education havemore impact on child schooling when few women in the neighborhood arehighly educated, while less educated mothers (with primary educationalone) have greater impact when they live in regions or communities wheremore equitable gender relations prevail.

Examining the impact of the interaction between mother’s educationand regional gender equity on the probability of child work, this studyfinds that when mothers with primary education live in regions with highgender equity, then the probability that their children will work decreasesin all subsamples except for girls below the poverty line. Thus, in stateswhere there is greater gender equity, mothers with primary educationdecrease child work. Mothers with primary education living in states withlower gender equity have a smaller impact. Moreover, mothers withsecondary education who live in states with greater gender equity have agreater impact than they would in states with low gender equity on decrea-sing the probability of girls’ employment, although they have no statisticallysignificant impact on boys’ employment. Overall, therefore, regionalgender equity is extremely important in determining the effect mothers’education (primary, secondary, and tertiary education) may have on thework and school probabilities of both boys and girls.

Regional and household autonomy: The net impact

In the estimated model, the net impact of a mother’s income andeducation depends on the coefficient of the variable itself as well as thecoefficient of the interaction term with the Gender Equity Index. For asingle variable (say, Mother’s Primary Education), the final effect willtherefore be as follows:

Schooli ¼ aþ b (mother’s primary education)i þ g½Gender Equity Index

� (mother’s primary education)i� þ ZZi þ ei

¼ aþ (mother’s primary education)iðbþ g Gender Equity Index)

þ ZZi þ ei

ARTICLES

100

Page 25: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

where Z denotes all the other variables in the model and i denotes theindividual.

Thus, the net coefficient of mother’s primary education is not aconstant but depends on the Gender Equity Index of the state. Sinceconceptualizing the size of this impact is not straightforward, this studyconsiders the range within which the effect falls by calculating the size ofthe coefficient of mother’s primary education (and of other relevantvariables) in the state with the lowest Gender Equity Index (Bihar with0.469) and the state with the highest Gender Equity Index (Kerala with0.825). Table 6 presents the results of this calculation and of the othervariables of interest.

To interpret these results, in all cases, this study considers whether thenet coefficient (the value in each cell) increases or decreases betweenBihar and Kerala. Since all other states in the sample are ranged betweenBihar and Kerala in terms of their Gender Equity Index, their coefficientsmust also fall between those of these two states.

Mother’s education (primary, secondary, and tertiary) andstate Gender Equity Index

Thus, Table 6 shows that the impact of mother’s primary education onchildren’s schooling is increasing in all samples (girls and boys aboveand below the poverty line). While the net impact of mother’s primaryeducation on girls’ schooling in households below the poverty line is 0.367in Bihar, it increases to 0.565 in Kerala. Similarly, the net probability ofmother’s primary education on boys’ schooling below the poverty line is0.314 in Bihar, but it increases to 0.561 in Kerala. Thus, while mothers withprimary educations have a positive impact on their children’s schoolingin both states, they have a larger positive effect in Kerala, the state withthe highest Gender Equity Index in India. This finding confirms thepattern from the marginal effects, which indicated that women (includingmothers) with primary educations or low levels of education gain greaterautonomy by living in regions with gender equity. Mother’s primaryeducation is more effective when it occurs in states with greater genderequity.

The negative impact of mother’s primary education on the probability ofchild work also increases with the Gender Equity Index of the state, exceptfor the sample of girls below the poverty line. Thus, the net probability ofwork for girls in households above the poverty line decreases with mother’sprimary education (the coefficient is always negative), but this variable hasa greater impact in Kerala (-0.810) than in Bihar (-0.461). This is true forthree of the four subsamples (except for girls below the poverty line). Thus,mothers with primary educations living in states with gender equity are lesslikely to send their children out to work.

CHILD SCHOOLING AND WORK

101

Page 26: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Tab

le6

Net

imp

act

on

the

pro

bab

ilit

yo

fch

ild

wo

rkan

dsc

ho

oli

ng

of

Gen

der

Eq

uit

yIn

dex

and

mo

ther

’sed

uca

tio

nan

din

com

eva

riab

les

intw

ost

ates

–B

ihar

and

Ker

ala

Gen

eral

mea

npr

obab

ilit

y

Mot

her’

spr

imar

yed

uca

tion

Mot

her’

sse

con

dary

edu

cati

onM

othe

r’s

tert

iary

edu

cati

onM

othe

r’s

inco

me

con

trib

uti

on

Mot

her’

sin

com

eco

ntr

ibu

tion

squ

ared

Rel

ativ

eed

uca

tion

Bih

arK

eral

aB

ihar

Ker

ala

Bih

arK

eral

aB

ihar

Ker

ala

Bih

arK

eral

aB

ihar

Ker

ala

Bih

arK

eral

a

Gir

lsb

elo

wsc

ho

ol

0.47

0.83

0.37

0.57

0.64

0.45

1.02

0.18

-1.0

60.

02-0

.11

0.05

-0.0

40.

01G

irls

abo

vesc

ho

ol

0.47

0.83

0.20

0.43

0.53

0.71

0.21

0.5

0.32

0.81

-0.3

3-1

.84

-0.1

8-0

.17

Bo

ysb

elo

wsc

ho

ol

0.47

0.83

0.31

0.56

0.59

0.56

1.86

-0.6

4-1

.08

-0.7

5-0

.43

0.33

-0.1

9-0

.16

Bo

ysab

ove

sch

oo

l0.

470.

83-0

.08

0.18

0.28

0.34

0.38

0.20

-1.6

8-0

.87

2.03

0.7

0.04

0.04

Gir

lsb

elo

ww

ork

0.47

0.83

-0.3

7-0

.37

-0.8

7-1

.53

0.00

0.00

-0.8

5-0

.01

-0.1

20.

080.

160.

29G

irls

abo

vew

ork

0.47

0.83

-0.4

6-0

.81

-0.8

0-1

.40

-0.6

7-1

.18

-2.2

1-1

.23

0.00

0.00

0.42

0.42

Bo

ysb

elo

ww

ork

0.47

0.83

0.06

-0.5

10.

000.

000.

000.

000.

000.

000.

000.

000.

000.

00B

oys

abo

vew

ork

0.47

0.83

0.04

-0.2

80.

000.

000.

000.

000.

000.

000.

000.

000.

000.

00

Not

es:T

he

valu

esin

the

cell

sin

this

tab

lep

rovi

de

the

effe

cto

fea

chva

riab

lean

dit

sin

tera

ctio

nw

ith

Gen

der

Eq

uit

yIn

dex

fro

mth

efo

llo

win

gaþb(

mo

ther

’sp

rim

ary

edu

cati

on

) iþg

Gen

der

Eq

uit

yIn

dex

*(m

oth

er’s

pri

mar

yed

uca

tio

n) iþZZ

iþe i

.W

her

eth

em

argi

nal

effe

cts

wer

ein

sign

ifica

ntl

yd

iffe

ren

tfr

om

zero

,th

eyw

ere

rest

rict

edto

zero

,gi

vin

gu

s,in

som

eca

ses,

no

imp

act

of

the

vari

able

inei

ther

stat

e.

ARTICLES

102

Page 27: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

On the other hand, the impact of mothers with tertiary education onchild schooling is higher in Bihar than in Kerala in all subsamples exceptgirls above the poverty line, and mothers’ tertiary education levels have noeffect on the probability of child work.14 This result might be interpreted asarguing that because there are fewer women with tertiary education in lowGender Equity Index states like Bihar, women with tertiary education inthese states might exert greater influence within their households at least asfar as child schooling is concerned. They remain ineffective in decreasingchild work, however, and this might be because very few children are likelyto work in households where mothers have tertiary education. However,with the small number of women with tertiary education, it is not clear thatwe should give much weight to this result. While mothers’ with secondaryeducation have a positive impact on child schooling in all subsamplesregardless of the Gender Equity Index of the state, the impact of mother’ssecondary education is larger in Bihar than in other states in two of the foursubsamples. Having a mother who has achieved a secondary educationdecreases the probability of work for girls in all states but the effect is largestin Kerala, that is the magnitude of the effect increases with the GenderEquity Index of the state. Mother’s secondary education has no impact onboys in any state.

Mother’s contribution to household expenditure and Gender Equity Index

Turning to the net impact of mother’s contribution to householdexpenditure, this study finds that the effect of this variable increases withthe Gender Equity Index for both school and work for girls in householdsbelow the poverty line. Thus, as mothers’ expenditure contributionsincrease, the probability of girls’ schooling and work will increase as theGender Equity Index increases. The impact on girls’ schooling is negative inBihar (-1.057) and positive in Kerala (0.022) for girls below the povertyline. Across all states, between these two extremes, the probability ofschooling increases as the Gender Equity Index increases. Similarly, thoughmother’s contribution to household expenditure decreases the probabilityof schooling for boys living in households above and below the poverty linein both Bihar and Kerala, its negative impact is smaller in Kerala (-0.746)than in Bihar (-1.078). This could be because in high gender equity stateslike Kerala, female employment and earnings are likely to reflect femalechoice and autonomy, while in low gender equity states like Bihar, they maymerely reflect the financial constraints of the household concerned.

Mothers’ expenditure contributions also decrease the probability ofwork for girls below the poverty line in Bihar by -0.845. The net effect of thisvariable is very close to zero (-0.005). Thus, in states with low gender equity,when mothers’ contribution to household income increases, the probabilityof work for girls decreases.

CHILD SCHOOLING AND WORK

103

Page 28: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Relative education and Gender Equity Index

The impact of mothers’ educations relative to fathers’ increases with theGender Equity Index across the two states, except for the subsample of boysin households above the poverty line. Thus, an increase in mothers’education levels relative to those of fathers decreases schooling for girlsliving below the poverty line in Bihar (-0.04), while it increases schoolingprobability for girls living below the poverty line in Kerala (0.005). It alsoincreases the probability of work for girls living in households below thepoverty line in both Bihar (0.163) and Kerala (0.287), but the effect isgreater in Kerala. Thus, in the subset of households below the poverty line,girls living in Kerala are more likely to work than are those in Bihar, but theyare also more likely to go to school as their mothers’ levels of educationrelative to their fathers’ increases. In households above the poverty line inthese two states, the difference in the impacts of this variable on childschooling and work is very small. Thus, the results indicate that the GenderEquity Index has an impact on the outcome, often overshadowing the effectof the mother’s autonomy variable. Also, the impact varies in householdsabove and below the poverty line, largely because in households below thepoverty line, the constraints of household finances are likely to be tighterand the role for female choices and autonomy more circumscribed.

DISCUSSION AND CONCLUSION

This paper set out to test three hypotheses relating to the impact ofmother’s autonomy on particular measures of child welfare: participationin school and in the labor market. To do this, it extended the concept offemale autonomy beyond the household to include the constraints imposedby the levels of gender equity prevalent in the regions that the women livein. It began with the expectation that increased autonomy for motherswould increase child schooling and decrease child work. This resulted inthree hypotheses that we tested in this paper but which yielded mixedresults.

First, we tested whether fathers’ and mothers’ wages yield similaroutcomes with respect to the schooling and work of their children. Ourresults indicate this is not the case: mother’s wages increase the probabilityof schooling but also increase the probability of work especially for girls.Father’s wages have less impact. These findings reinforce the results ofprevious studies (Kaushik Basu and Ranjan Ray 2002; Patrick M. Emersonand Andre Portela Souza 2002; Afridi 2006).

Second, we hypothesized that in households where mothers have greaterautonomy, the probability of child schooling would be higher and that ofchild work would be lower. Our results indicate that reality is more complexthan this hypothesis would indicate. Thus, we find that when mothers have

ARTICLES

104

Page 29: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

greater autonomy, the impact depends upon the subsample we areconsidering (boys, girls, high or low income) and also upon whether we areconsidering school or work. We can, however, conclude that mother’sautonomy in itself does not imply better outcomes (more schooling and lesswork) for children. Mother’s relative education, for instance, decreases theprobability of schooling for both boys and girls but has no significantimpact on work probabilities. Mother’s expenditure contributions, onthe other hand, decrease the probability of work for girls and also theprobability of schooling. The impact of both these variables also changeswith the gender equity of the state that the mother lives in.

Finally, our third hypothesis tests whether the extent of gender equityprevalent in a state increases schooling and decreases work. Our resultsindicate that while it increases child schooling as expected, it also increasesthe probability of child work. Thus, our results in the previous subsectionindicate that the net effect of gender equity and mother’s contribution tohousehold expenditure on girl’s schooling and work probabilities andon boy’s work probabilities increases with the gender equity of the state.Together, our results seem to indicate that when mothers contribute tohousehold expenditure in states with gender equity, then the results aremore benign than when they contribute in states without gender equity.This is not surprising because in the former regions, mothers’ contributionsare associated with the availability of choices and therefore more closelyreflect autonomy than in the latter (low gender equity) states, where amother’s contribution is more likely to reflect financial constraints and alack of choices.

Overall, our results indicate that mother’s education (on its own ratherthan relative to the father’s) is an important determinant of the probabilityof child work and schooling. So also is mother’s contribution to householdexpenditure. In both cases, however, the impact depends significantly uponthe gender equity of the state that the child lives in. In most cases, higherlevels of gender equity reinforce mothers’ autonomy, while lower levels ofregional gender equity offset any autonomy the mother would otherwisehave.

Uma Sarada Kambhampati, School of Economics andCentre for Institutional Performance, University of Reading,

PO Box 218, Whiteknights, Reading RG6 6AA, UKe-mail: [email protected]

ACKNOWLEDGMENTS

I am grateful to the Department for International Development, UK, forfunding the project that made this research possible. I am also grateful toparticipants in the Econometrics Society Australasian Meeting 2006, and in

CHILD SCHOOLING AND WORK

105

Page 30: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

the International Association of Feminist Economists Conference inSydney, 2006, for extremely useful comments. Any errors that remain aremine alone.

NOTES1 The poverty line in India is set by the Indian government separately for rual and

urban areas. It indicates the income level required to survive at subsistence level andis used to calculate the level of poverty in the country. In 1992, it was Rs.296 per capitaper month in urban areas and Rs.276 per capita per month in rural areas.

2 Figures in Appendix A appear to bear this out with the proportion of girls in thesample being 48.2 percent in high gender equity regions, while in low gender equityregions it is 45.5 percent.

3 Thus, if the norm in a region is to educate girls, then even a household thattraditionally would not educate its girl children may succumb to societal pressures.Conversely, if regional norms dictate that women do not go out unless accompaniedby someone from their household, this would increase the obstacles to the educationof daughters.

4 ILO conventions recommend a minimum age for admission to employment or workthat must not be less than the age for completing compulsory schooling, and in anycase not less than 15 years. Lower ages are permitted – generally in countries whereeconomic and educational facilities are less well developed. The minimum age forthose countries is 14 years, with 13 years permitted if the child is engaging in ‘‘lightwork.’’ The minimum age for ‘‘hazardous work,’’ however, is 18 years.

5 Note, however, that this result depends on both father and mother being unwilling tosend the child out to work.

6 Households in this dataset are defined as all people living and eating under one roofand cooking in one kitchen. However, in the study, we are concerned with the kindsof decisions that fathers and mothers make with regard to child schooling and work.Our concern in this context is with the nuclear family but it is likely to be affected byother family conditions, including the existence of wider family members. We attemptto allow for this by including the number of older dependents in the household andalso the overall family income (so that mother’s income is a proportion of totalhousehold expenditure). We also allow for the possibility that some children live infemale headed households by including a dummy to indicate these households.

7 While including the mother’s wage relative to the father’s wage seems to be anobvious choice here, this study included mother’s wage relative to householdexpenditure for several reasons. First, if father’s wage is zero, the resultant variable isindeterminate, even though it is clear that in this case, the mother’s wage mightincrease her power within the household. Also, since household expenditure is thefinal variable to which all wages are contributing, the mother’s contribution toexpenditures might be expected to determine her power. In cases where the house-hold has no nonwage sources of income, this variable collapses to being simplymother’s wage relative to father’s wage. However, in households where there arenonwage income sources, such as income from goods sold at market or rentalincome, then this variable provides more information than simply mother’s wage as aproportion of father’s wage.

8 For this measure I am using the Gender Disparity Index produced by the UNDP.I have renamed it the Gender Equity Index because the higher the index, the moreequitable are gender relations and therefore the name change helps clarify discussionof the study results.

ARTICLES

106

Page 31: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

9 Children who do two activities – work and study, for instance – are classified eitherwithin school or within work, depending on which activity they spend more timedoing. This either-or classification is useful because it considers the child’s primaryactivities in binary terms. However, it does not allow us to consider children who aredoing another activity as a secondary activity. This possibility does not seem to presenta major problem in the sample because summary statistics indicate that a majority ofthe children (85 percent of boys and 71 percent of girls) who did some work workedfull time, that is, 7 days a week.

10 As this is the rural sector, there are very few parents with tertiary education, especiallyamong mothers in this sample (only 0.7 percent of the mothers and 3 percent of thefathers in the entire sample have tertiary education).

11 It is endogenous because it is partly determined by the level of household income.This can bias the estimates and therefore we need to use instruments for theseendogenous variables.

12 Results for the former model are available upon request from the author.13 Since there are so few mothers with tertiary education in the below poverty line

households, the standard errors of this variable are very high in the work equation.This study therefore interprets this result with caution.

14 Note that it is in the work equations that the coefficient of mothers’ tertiary educa-tion has very high standard errors. So, these probabilities must be interpreted withcaution.

REFERENCES

Afridi, Farzana. 2006. ‘‘Female Empowerment and the Gender Gap in Schooling inIndia.’’ Working paper, Syracuse University, New York.

Basu, Kaushik. 2006. ‘‘Gender and Say: a Model of Household Behaviour withEndogenously Determined Balance of Power.’’ The Economic Journal 116(511): 558–80.

Basu, Kaushik and Ranjan Ray. 2002. ‘‘The Collective Model of the Household and anUnexpected Implication for Child Labour; Hypothesis and an Empirical Test.’’ WorldBank Policy Research Working Paper 2813.

Becker, Gary S. 1965. ‘‘A Theory of Allocation of Time.’’ The Economic Journal 75(299):493–517.

Chiappori, Pierre-Andre. 1988. ‘‘Rational Household Labor Supply.’’ Econometrica 56(1):63–89.

———. 1992. ‘‘Collective Labor Supply and Welfare.’’ Journal of Political Economy 100(3):437–67.

Doss, Cheryl R. 1996a. ‘‘Women’s Bargaining Power in Household Economic Decisions:Evidence from Ghana.’’ Department of Applied Economics Staff Paper P96-11,University of Minnesota.

Dreze, Jean and Amartya Sen. 1996. India: Economic Development and Social Opportunity.New Delhi: Oxford University Press.

Dreze, Jean and Haris Gazdar. 1997. ‘‘Uttar Pradesh: The Burden of Inertia,’’ in JeanDreze and Amartya Sen, eds. 1997. Indian Development: Selected Regional Perspectives(UNU/Wider Studies in Development Economics), pp. 33–108. New York: Oxford UniversityPress.

Dyson, Tim and Mick Moore. 1983. ‘‘On Kinship Structure, Female Autonomy, andDemographic Behavior in India.’’ Population and Development Review 9(1): 35–60.

Emerson, Patrick M. and Andre Portela Souza. 2002. ‘‘Bargaining Over Sons andDaughters: Child Labour, School Attendance and Intra–Household Gender Bias inBrazil.’’ Working Paper 02–W13, Vanderbilt University.

CHILD SCHOOLING AND WORK

107

Page 32: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Hoddinott, John. 1992. ‘‘Rotten Kids or Manipulative Parents: Are Children Old AgeSecurity in Western Kenya?’’ Economic Development and Cultural Change 40(3): 545–65.

Government of India. 1986. Indian Child Labour Prohibition and Regulation Act. 1986.http://indiacode.nic.in/fullact1.asp?tfnm¼198661

International Labour Organization (ILO). 2009. ILO Convention No. 138 Minimum AgeConvention, 1973. http://www.ilocarib.org.tt/projects/cariblex/conventions_6.shtml(accessed June 2009).

Jeffrey, Patricia, Roger Jeffrey, and Andrew Lyon. 1988. ‘‘When Did You Last See YourMother? Aspects of Female Autonomy in Rural North India,’’ in John C. Caldwell,Allan G. Hill and Valerie J. Hull, eds. Micro-Approaches to Demographic Research, pp. 321–33. London: Kegan Paul International.

Kabeer, Naila. 2003. Reversed Realities: Gender Hierarchies in Development Thought. London:Verso.

Kak, Shakti. 2004. ‘‘Magnitude and Profile of Child Labour in the 1999s: Evidence fromthe NSS Data.’’ Social Scientist 32(1–2): 45–73.

Kambhampati, Uma S. and Raji Rajan. 2006. ‘‘Economic Growth: A Panacea for ChildLabour?’’ World Development 34(3): 426–45.

———. 2008. ‘‘The ‘Nowhere’ Children: Girls in India’s Rural Economy.’’ Journal ofDevelopment Studies 44(9): 1309–41.

Kravdal, Øystein. 2004. ‘‘Child Mortality in India: The Community Level Effect ofEducation.’’ Population Studies 58(2): 177–92.

Lancaster, Geoffrey, Pushkar Maitra, and Ranjan Ray. 2006. ‘‘Endogenous Intra-household Balance of Power and its Impact on Expenditure Patterns: Evidence fromIndia.’’ Economica 73(291): 435–60.

Manser, Marilyn and Murray Brown. 1980. ‘‘Marriage and Household Decision-Making:A Bargaining Analysis.’’ International Economic Review 21(1): 31–44.

McElroy, Marjorie B. and Mary Jean Horney. 1981. ‘‘Nash-Bargained HouseholdDecisions: Toward a generalization of the theory of demand.’’ International EconomicReview 22(2): 333–49.

Miller, Barbara D. 1981. The Endangered Sex: Neglect of Female Children in Rural India.Ithaca: Cornell University Press.

National Sample Survey Organisation (NSSO). 1993. National Sample Survey, 50thRound, Schedule 10. CD-ROM.

Nieuwenhuys, Olga. 1996. ‘‘The Paradox of Child Labor and Anthropology.’’ AnnualReview of Anthropology 25: 237–51.

Planning Commission, Government of India. 2002. National Human Development Report2001. New York: Oxford University Press.

Quisumbing, Agnes R. and John A. Maluccio. 1999. ‘‘Intra-Household Allocation andGender Relations: New Empirical Evidence.’’ Policy Research Report on Gender andDevelopment Working Paper Series, No. 2, World Bank Development Research Group.

Rehman, Lupin and Vijayendra Rao. 2004. ‘‘The Determinants of Gender Equity inIndia: Examining Dyson and Moore’s Thesis with New Data.’’ Population andDevelopment Review 30(2): 239–68.

Rosenzweig, Mark R. and Kenneth I. Wolpin. 1982. ‘‘Government Interventions andHousehold Behaviour: Anticipating the Unanticipated Consequences of SocialProgrammes.’’ Journal of Development Economics 10(2): 209–25.

Sopher, David E. 1980. ‘‘The Geographical Patterning of Culture in India,’’ in Davod E.Sopher, ed. An Exploration of India: Geographic Perspectives on Society and Culture. Ithaca,NY: Cornell University Press.

Tilak, Jandhyala B.G. 2002. ‘‘Determinants of Household Expenditure on Education inRural India.’’ NCAER Working Paper 88, National Council of Applied EconomicResearch.

ARTICLES

108

Page 33: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

App

endi

xA

Dat

ad

escr

ipti

on

of

du

mm

yva

riab

les

dis

cuss

edin

the

pap

er

Hig

hge

nde

req

uit

yL

owge

nde

req

uit

yB

elow

pove

rty

lin

eA

bove

pove

rty

lin

e

Du

mm

yva

riab

les

Des

crip

tion

No.

of1s

Pro

p.of

1s(%

)N

o.of

1sP

rop.

of1s

(%)

No.

of1s

Pro

p.of

1s(%

)N

o.of

1sP

rop.

of1s

(%)

Sch

oo

l(¼

1)Id

enti

fies

the

pri

mar

yac

tivi

tyo

fth

ech

ild

;co

ded

1if

the

pri

mar

yac

tivi

tyis

stat

edto

be

wo

rk,

else

cod

ed0

(1¼

pri

mar

yac

tivi

tyis

atte

nd

ing

sch

oo

l)

1826

573

.85

3537

564

.60

2842

459

.01

2521

680

.50

Wo

rk(¼

1)Id

enti

fies

the

pri

mar

yac

tivi

tyo

fth

ech

ild

,co

ded

1if

the

pri

mar

yac

tivi

tyis

stat

edto

be

atte

nd

ing

sch

oo

l,el

seco

ded

0(1¼

pri

mar

yac

tivi

tyis

goin

gfo

rw

ork

)

2113

8.54

2939

5.37

3639

7.56

1413

4.51

Sex

(1¼

girl

s;0¼

bo

ys)

Gen

der

of

the

chil

d,

cod

ed1¼

girl

s0¼

bo

ys;

(1¼

girl

)11

918

48.1

924

942

45.5

422

349

46.4

014

511

46.3

0

Mo

ther

’sp

rim

ary

edu

cati

on

(¼1)

Mo

ther

’sp

rim

ary

edu

cati

on

;co

ded

pri

mar

yed

uca

tio

n,

else¼

0

4621

18.6

891

6616

.70

6233

12.9

475

5424

.10

Mo

ther

’sse

con

dar

yed

uca

tio

n(¼

1)M

oth

er’s

seco

nd

ary

edu

cati

on

;co

ded

seco

nd

ary

edu

cati

on

,el

seco

ded

0

2708

10.9

537

026.

7618

073.

7546

0314

.70

Mo

ther

’ste

rtia

ryed

uca

tio

n(¼

1)M

oth

er’s

tert

iary

edu

cati

on

;co

ded

tert

iary

edu

cati

on

,el

seco

ded

0

716

2.90

1849

3.38

565

1.12

2000

6.40

(con

tin

ued

)

CHILD SCHOOLING AND WORK

109

Page 34: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

App

endi

xA(C

onti

nu

ed)

Hig

hge

nde

req

uit

yL

owge

nde

req

uit

yB

elow

pove

rty

lin

eA

bove

pove

rty

lin

e

Du

mm

yva

riab

les

Des

crip

tion

No.

of1s

Pro

p.of

1s(%

)N

o.of

1sP

rop.

of1s

(%)

No.

of1s

Pro

p.of

1s(%

)N

o.of

1sP

rop.

of1s

(%)

Fat

her

’sem

plo

ymen

tF

ath

er’s

emp

loym

ent;

bin

ary

vari

able

cod

ed0¼

no

wo

rk,

else

1(e

mp

loye

1)Q

:If

the

pri

nci

pal

acti

vity

of

the

fath

erw

asw

ork

ing,

did

he

wo

rkm

ore

or

less

regu

larl

yin

the

last

365

day

s?

2149

286

.90

5110

893

.30

4275

788

.80

2984

395

.22

Fat

her

’sp

rim

ary

edu

cati

on

(¼1)

Fat

her

’sp

rim

ary

edu

cati

on

;co

ded

pri

mar

yed

uca

tio

n,

else¼

0

7384

29.8

615

981

29.1

813

743

28.5

396

2230

.70

Fat

her

’sse

con

dar

yed

uca

tio

n(¼

1)F

ath

er’s

seco

nd

ary

edu

cati

on

;co

ded

seco

nd

ary

edu

cati

on

,el

seco

ded

0

5303

21.4

410

195

18.6

066

5813

.80

8840

28.2

0

Fat

her

’ste

rtia

ryed

uca

tio

n(¼

1)F

ath

er’s

tert

iary

edu

cati

on

;co

ded

tert

iary

edu

cati

on

,el

seco

ded

0

1224

4.95

3910

7.14

1340

2.78

3794

12.1

0

ARTICLES

110

Page 35: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

App

endi

xB

Dat

ad

escr

ipti

on

of

con

tin

uo

us

vari

able

sd

iscu

ssed

inth

ep

aper

Con

tin

uou

sva

riab

les

Des

crip

tion

Hig

hge

nde

req

uit

yL

owge

nde

req

uit

yB

elow

pove

rty

lin

eA

bove

pove

rty

lin

e

Age

Age

of

the

chil

d9.

894

3.13

89.

711

3.13

49.

532

3.10

010

.128

3.15

7A

gesq

uar

edA

ge*A

ge10

7.73

063

.283

104.

136

62.9

8810

0.46

661

.796

112.

543

64.3

55M

oth

er’s

wag

eM

oth

er’s

wag

ein

stru

men

ted

usi

ng

mo

ther

age,

mo

ther

edu

cati

on

,vi

llag

efe

mal

ew

ages

,an

dvi

llag

eu

nem

plo

ymen

t.T

he

met

ho

do

fin

stru

men

tati

on

isp

red

icti

on

fro

ma

Sam

ple

Sele

ctio

nm

od

el.

43.9

5738

.664

28.3

3544

.800

32.4

9133

.516

38.5

7956

.070

Fat

her

’sw

age

Fat

her

’sw

age

inst

rum

ente

du

sin

gfa

ther

age,

fath

ered

uca

tio

n,

vill

age

mal

ew

ages

,an

dvi

llag

eu

nem

plo

ymen

t.T

he

met

ho

do

fin

stru

men

tati

on

isp

red

icti

on

fro

ma

Sam

ple

Sele

ctio

nm

od

el.

93.6

0066

.582

90.7

9478

.021

81.3

6355

.235

107.

145

94.9

09

Mo

ther

’sex

pen

dit

ure

con

trib

uti

on

Mo

ther

’sw

age

asa

pro

po

rtio

no

fh

ou

seh

old

exp

end

itu

re

0.19

00.

194

0.13

20.

255

0.19

00.

267

0.09

60.

140

Rel

ativ

eed

uca

tio

nN

um

ber

of

year

so

fed

uca

tio

no

fm

oth

erd

ivid

edb

yfa

ther

’sed

uca

tio

n(i

nye

ars)

0.26

30.

444

0.20

60.

386

0.14

10.

344

0.35

10.

458

Fat

her

’sag

eA

geo

ffa

ther

44.0

8811

.795

44.7

7012

.096

43.9

3311

.856

45.4

9912

.171

Fat

her

’sag

esq

uar

edSq

uar

eo

fag

eo

ffa

ther

2082

.921

1182

.507

2150

.618

1215

.868

2070

.706

1178

.504

2218

.309

1241

.022

(con

tin

ued

)

CHILD SCHOOLING AND WORK

111

Page 36: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

App

endi

xB(C

onti

nu

ed)

Con

tin

uou

sva

riab

les

Des

crip

tion

Hig

hge

nde

req

uit

yL

owge

nde

req

uit

yB

elow

pove

rty

lin

eA

bove

pove

rty

lin

e

Mo

ther

’sag

eA

geo

fm

oth

er37

.787

10.2

6339

.162

10.5

6238

.199

10.3

2839

.545

10.6

77M

oth

er’s

age

squ

ared

Squ

are

of

age

of

mo

ther

1533

.136

900.

594

1645

.235

941.

074

1565

.820

906.

421

1677

.788

961.

112

Vil

lage

fem

ale

wag

eL

og

of

aver

age

vill

age

fem

ale

wag

e3.

648

1.82

82.

608

2.36

53.

005

2.15

12.

894

2.39

7

Vil

lage

mal

ew

age

Lo

go

fav

erag

evi

llag

em

ale

wag

e5.

069

1.03

14.

719

1.66

74.

715

1.46

55.

003

1.54

9

Gen

der

Eq

uit

yIn

dex

Gen

der

Eq

uit

yIn

dex

(pu

bli

shed

asth

eG

end

erD

isp

arit

yIn

dex

by

UN

DP

):m

easu

res

dis

par

ity

ined

uca

tio

n,

hea

lth

,an

dem

plo

ymen

tac

ross

the

gen

der

s.

0.77

20.

041

0.46

80.

234

0.58

00.

212

0.53

90.

277

ARTICLES

112

Page 37: CHILD CHOOLING ANDW D I :T R H R G E · PDF filegreater autonomy that those in North India, Rehman and Rao (2004) draw somewhat different conclusions. They find that village exogamy

Recommended