ArticlePDF Available

The Effect of Domestic Work on Girls' Schooling: Evidence from Egypt

Authors:

Abstract and Figures

In Egypt, girls' work primarily takes the form of domestic tasks, which are not considered in many studies of child labor. This paper investigates the effect of girls' work on their school attendance. It uses a modified bivariate probit approach to estimate the effect of work on schooling while allowing for the simultaneous determination of the two outcomes. It presents evidence that the substantial burden of girls' domestic work leads to lower rates of school attendance. Policies that attempt to ban the labor-force work of children will have practically no effect on girls' education in Egypt, while interventions reducing the drudgery of household labor through, for example, improved water and sanitation infrastructure, have better prospects for success.
Content may be subject to copyright.
1
THE EFFECT OF DOMESTIC WORK ON GIRLS SCHOOLING:
EVIDENCE FROM EGYPT
Ragui Assaad, Deborah Levison, and Nadia Zibani
forthcoming in Feminist Economics (2010)
ABSTRACT
In Egypt, girls’ work primarily takes the form of domestic tasks, which are not
considered in many studies of child labor. This paper investigates the effect of girls’ work on
their school attendance. It presents evidence that the substantial burden of girls’ domestic work
leads to lower rates of school attendance. Policies that attempt to ban labor force work of
children will have practically no effect on girls’ education in Egypt, while interventions reducing
the drudgery of household labor through, for example, improved water and sanitation
infrastructure, have better prospects for success.
KEYWORDS
Child labor, schooling, domestic work, gender, Egypt, household economics
JEL CODES: D15, I21, J82, N35
2
THE EFFECT OF DOMESTIC WORK ON GIRLS’ SCHOOLING
INTRODUCTION
Human capital enhancement is a critical element of poverty reduction. Therefore,
children’s participation in primary and secondary schools plays an important role in the
development of societal infrastructure in poor countries. Although strong negative correlation
between children’s market work responsibilities and educational achievement or attainment has
been established (e.g., Christopher Heady 2000; David Post and Suet-ling Pong 2000), less
attention has been paid to the effect of girls’ domestic work on their schooling.
1
However, girls’
work falls in both of these domains, and in the case of Egypt, it is primarily domestic work.
Child labor and schooling are mutually determined in a complex web of social and
economic factors, including socio-cultural norms, economic shocks, living conditions and access
to services, and poverty in general. In this paper we attempt to investigate one of the possible
pathways in Egypt between the need for a girl to contribute to domestic work and her parents’
decision to either not send her to school or to have her drop out. The relationship between work
and schooling is complicated by the fact that there are factors that affect both a child’s ability to
go to school and her need to work and factors that affect one or the other. Moreover, working
affects the ability to attend and succeed in school, while failure in school may lead to a child
1
The possible exception is the literature on the effect of mother’s participation in the labor market on girl
children’s education, which implies that mothers’ and girls’ domestic work are potential substitutes (e.g., Benería
1992, Levison, Moe and Knaul 2001) .
3
being put to work. Infrastructure and access to services are an integral part of this web of
relationships. For example, it could very well be that for reasons associated with schooling
quality or accessibility, children fail in school (as found by Cynthia Lloyd, Sahar E. Tawila,
Wesley Clark, and Barbara Mensch 2003), which leads parents to put them to work. In this
situation, reducing the burden of children’s work without improving school quality would do
little to improving educational attainment. On the other hand, poor water and sanitation services
substantially increase the domestic work burden, leading parents to rely heavily on girls’ labor,
often at the expense of their schooling. This latter pathway, from quality of services to girls’
domestic work to schooling, is our focus in this paper.
According to recent survey data there have been substantial improvements in the school
enrollment rates of Egyptian children, and substantial reductions in the incidence of child work
in the 1990s (Nadia Zibani 2002). From 1988 to 1998, school enrollment of 614 year old girls
jumped from 75 to 86 percent, while their involvement in market and subsistence labor force
activities fell from 11.4 to 8.2 percent.
2
Still, in 1998 in Egypt approximately 886 thousand girls
ages 614 (14 percent) did not attend school, and approximately 513 thousand girls regularly
engaged in labor force work, while another 2.2 million girls (37.7 percent) spent time on
household chores and uncounted hours engaged in childcare (Zibani 2002).
The conventional measures of labor force work often ignore a variety of child activities
that could potentially jeopardize their schooling. This is especially true for girls, who must often
2
These figures are obtained from the 1988 Labor Force Sample Survey (LFSS) and the 1998 Egypt Labor
Market Survey (ELMS). The labor force participation rates reported here are based on an extended definition of
economic activity and a short reference period of one week with a one-hour per week minimum threshold. The 1988
survey did not measure the incidence of domestic work.
4
do domestic chores, which are not captured in the conventional definitions of work, for many
hours each day. When meal production and/or childcare are required, the times at which chores
take place may be less flexible than often assumed. In this paper we use a more inclusive
measure of work that includes all of girls’ work that is captured by our data. While girls’ work
time in Egypt is primarily spent in domestic tasks, our measure also captures market work and
some subsistence work.
We start from the position that all children should have the opportunity to attend school,
thus potentially reaping the benefits of increased human capital formation throughout the rest of
their lives.
3
Increasing enrollment and education attainment levels are also desirable as
mechanisms for reducing fertility and population growth (1.7 percent in Egypt in 20005). The
population of greatest interest to us includes those girls who are expected to be in school,
according to Egyptian law, and who are defined as ―children‖ under most international
conventions: those ages 6 to 14 years old. We focus this analysis on girls, who suffer from a
persistent gender gap in primary schooling in Egypt. The key question we seek to explore is the
following: when does girls’ work, broadly defined, put Egyptian girls at risk of not benefiting
from education to the extent possible?
3
This assumption also has its problems. When schools are of poor quality, children may benefit more from
other activities. When children are regularly beaten and verbally abused in schools, our assumption is again
problematic.
5
Domestic work
Girls’ domestic labor is also regularly ignored in analyses of children’s activities. In
particular, the child labor literature often overlooks the potential for housework and childcare
responsibilities to interfere with educational attainment. Exceptions include Pedro Goulart and
Arjun S. Bedi (2008), who distinguish between the impact of domestic work and market work on
educational outcomes for children in Portugal, and Gautam Hazarika and Arjun S. Bedi (2003),
who examine the effects of schooling costs on both types of child work in rural Pakistan.
Deborah Levison and Karine S. Moe (1998) and Deborah Levison, Karine S. Moe, and Felicia
Knaul (2001) document that an assessment of whether or not work impedes educational
attainment is sensitive to how one defines work, especially for girls. They also show that a
traditional definition of work misrepresents the gender differentials in the incidence and
determinants of work among children in Peru and Mexico. Although a distinction between
market work and domestic work is useful, the traditional definition of market work makes some
seemingly arbitrary distinctions between activities that are similar. Performing unpaid work in a
family enterprise and preparing food in a market stall are considered work, whereas similar
activities done for purposes of household consumption are not. While such distinctions may
make sense in the context of national accounts or labor force statistics, they may result in biases
when trying to understand the phenomena of child labor and schooling (Deborah Levison 2000).
Our data includes information on Egyptian womens and girls domestic activities, including the
three most important activities carried out in the reference week, by order of importance, as well
as the total amount of time spent on all domestic activities. While our measure of domestic work
is much more inclusive of the work burdens that girls bear, it probably understates that burden
because childcare is often understated even when household chores are reported.
6
Higher school enrollment and improved school attainment have been repeatedly shown to
be among the most effective ways to reduce both poverty and fertility. The benefits of schooling
accrue not only to the individual him or herself but to the entire society through a variety of
spillover effects. There is less consensus on the harmful effects of children’s work. Some argue
that if children learn important skills and discipline by working, the early onset of work could be
beneficial if it does not unduly affect schooling (Jo Boyden, Birgitta Ling, and William Myers
1998). On the other hand, if child work interferes with schooling or exposes children to harmful
and hazardous conditions, it clearly has detrimental effects. Because child market work is
strongly associated with not being in school and a small number of studies suggests this holds
for domestic work as well (Goulart and Bedi 2008, Hazarika and Bedi 2003, Levison and Moe
1998, Levison, Moe, and Knaul 2001) it is often assumed that child work causes school
dropout. This is not necessarily true, however. It could very well be that for other reasons, some
children are at risk of failing at school, and they engage in work because their schooling
prospects are poor. Having a better understanding of the pathways that link the social and
economic environment to child work and, through work to school, is crucial to implementing the
right policies. If work leads to school dropout, then policies to curtail child labor are justified.
However, if failure in school is the main reason children are put to work, then policy measures
need to focus on addressing the reasons for school failure as a first priority.
The majority of young girls who are not enrolled in school in Egypt have never even
started school. The evidence indicates that girls’ intensive participation in domestic work does
not start until they are about 10 years old. This raises an issue about whether work could be the
cause of girls not going to school in the first place. Our argument is two-fold. First, parents,
anticipating a girl’s contribution to domestic work in later years, may not feel that it is
7
worthwhile to enroll her in school for a few years. They thus avoid the costs associated with
schooling. In examining reported reasons why female adolescents in Egypt have never attended
school, Barbara Ibrahim, Sunny Sallam, Sahar El-Tawila, Omaima El-Gibaly, and Fikrat El-Sahn
(2000: 69) find that 47 percent of parents in Egypt place a low value on education, 44 percent
mention poverty, and 16 percent mention the need for the girl to help with household chores.
Second, while young girls may not be engaged in measurable domestic chores, it is likely that
they are caring for younger siblings. For example, Homa Hoodfar, writes that ―In lower-income
families in Cairo elder daughters participate actively in child care. Girls begin child care
activities as early as seven years. Some families keep the eldest daughter out of school in order to
help with child care and daily housework‖ (1986). In a later study, Homa Hoodfar documents
women’s observations that childcare ―was becoming increasingly demanding, owing to higher
expectations‖ (1996: 13). Not only did children remain relatively free of responsibilities for
many more years, until their late teens, but they could not simply be fed and ―sent off to play in
some field or street‖ (1996: 13). They required more supervision.
An enormous literature speaks to the enrollment and educational attainment of children in
developing countries, and a more recent and growing literature addresses child labor force work.
A number of studies from the last decade and a half explicitly recognize the necessity of
considering schooling in conjunction with children’s labor force employment and non-labor-
force work responsibilities.
4
This comprehensive type of approach is needed to attain an adequate
4
Examples include Deborah S. DeGraff and Richard Bilsborrow (2003) for the Philippines; Peter Jensen
and Helena Skyt Nielsen (1997) for Zambia; Sudharshan Canagarajah and Harold Coulombe (1998) for Ghana;
Felicia Knaul (1995) for Mexico and Colombia; Melissa Binder and David Scrogin (1999) for Mexico; Harry
Anthony Patrinos and George Psacharopoulos (1997) for Peru; George Psacharopoulos (1997) for Bolivia and
8
understanding of how to facilitate the educational success of children with multiple
responsibilities.
Few analyses, however, have taken account of the simultaneous nature of family (or
child) decisions regarding school and work activities due to various estimation difficulties. Some
authors use a multinomial logit approach to jointly consider categories: work, work and school,
school only, or neither (examples include Levison, Moe and Knaul [2001] and various chapters
in Christiaan Grootaert and Harry Anthony Patrinos [1999]). One of the problems of this
approach is the assumption of independence of irrelevant alternatives. Others have used ordered
probit models (authors of various chapters in Grootaert and Patrinos (1999) employ a shared
estimation strategy, including ordered probit models of children’s work and school
participation). The fundamental problem with such models is that they must assume parents and
children always rank order activities in a certain way. For example, analysts assume that school
only is preferred to combining school and work, which is preferred to work only. We find this
assumption inappropriate, not least because of the lack of empirical evidence to validate its use.
Others have used a bivariate probit approach that models work and school enrollment as two
interdependent binary decisions (Sudharshan Canagarajah and Harold Coulombe 1998; Samuel
Freije and Luis F. Lopez-Calva 2001; Jackline Wahba 2006). None of the above approaches
attempt to disentangle the causal effects of child work on school enrollment.
A number of papers have successfully used instrumental variables methods to identify the
causal impact of child labor on educational outcomes in the context of developing countries, but
in each case the focus has been on market work and generally on agricultural work. Michael A.
Venezuela; George Psacharopoulos and Ana Maria Arriagada (1989) for Brazil; Jose Canals-Cerda and Cristobal
Ridao-Cano (2004) for Bangladesh; and Emmanuel Skoufias (1994) for India.
9
Boozer and Tavneet K. Suri (2001), using regional variation in the pattern of rainfall to
instrument for monthly variations child work in Ghana, find a negative effect of child labor on
schooling. Kathleen Beegle, Rakeev Dehejia, and Roberta Gatti (2005) exploit rice prices and
commodity disasters as a source of exogenous variation in child labor in Vietnam. Using panel
data, they find statistically significant negative impacts of child labor on school participation and
educational attainment. Victoria Gunnarsson, Peter Orazem, and Mario A. Sanchez (2006) use
cross-country variation in truancy regulations to identify the (negative) effect of children’s labor
force work on test scores of third and fourth graders in nine Latin American countries.
There is a substantial literature on the relationship between time at work and school
attainment of high school or college students in the United States. These studies have generally
found little evidence that part-time work while attending school affects school achievement, with
a few exceptions. Ralph Stinebrickner and Todd R. Stinebrickner (2003) make use of unique data
from a college that has a mandatory work-study program in which jobs requiring varying levels
of effort are randomly assigned in the first semester. John H. Tyler’s (2003) identification
strategy relies on interstate variations in child labor laws. Goulart and Bedi (2008), mentioned
above, tackle the endogeneity of child labor using geographical variation in anti-child labor
programs and inspection regimes.
We use a modified bivariate probit approach that allows for the estimation of the effect of
work on schooling, while allowing for the simultaneous determination of the two outcomes. To
be implemented successfully, the approach requires the availability of instruments that determine
the probability of working but do not directly affect schooling decisions.
5
5
Canals-Cerda and Ridao-Cano (2004) use a similar approach to determine the effect of working while in
school on the probability of progressing to secondary school in rural Bangladesh.
10
Objectives
This paper has two main objectives. First, it examines the pathway between girls’ work
and school participation in Egypt. Second, it expands the definition of work for girls to include
unpaid domestic work done in the girls’ own homes. Girls’ involvement in domestic work is
extensive and undocumented in standard labor force statistics.
Admittedly, girls’ work could have implications for schooling beyond determining
current school attendance. It could affect the regularity of school attendance as well as school
performance and grade advancement. Ultimately, any negative effects are bound to increase
school dropout and thus affect enrollment. Given the limitations of our data sources we are
unable to consider these other dimensions of schooling.
DATA
The data for this study are obtained from the Egypt Labor Market Survey (1998 ELMS)
(Economic Research Forum 1998), which is a nationally representative household survey carried
out on a sample of 5,000 households in late 1998.
6
The survey was conducted under the auspices
of the Economic Research Forum, in collaboration with the Central Agency for Public
Mobilization and Statistics (CAPMAS), the Government of Egypt's central statistical agency.
The 1998 ELMS was designed to be comparable to a special round of the Egyptian Labor Force
Sample Survey (LFSS) conducted exactly 10 years earlier in October 1988 (CAPMAS 1988), but
6
Ragui Assaad was principal investigator, responsible for the sampling design, instrument, training of
enumerators, and fielding of the survey.
11
the 1998 survey included significantly more information on a variety of topics including
schooling and domestic work for women and girls. Completed questionnaires were obtained for
4,816 households and 23,997 individuals, of whom 2,458 were girls between the ages of 6 and
14. Due to missing data on some variables, our final sample includes 2,442 girls ages 614.
WORK AND SCHOOL IN THE EGYPTIAN CONTEXT
Basic education
There are currently (in 2009) nine years of mandatory basic education in Egypt, made up
of six years of primary education and three years of lower secondary education. For a period of
nine years, from 1990 to 1999, the years of primary education were reduced to 5 years to allow
for the absorption of a larger number of children in the school system.
7
Since all of the girls in
our sample would have been below the fifth grade in 1990, they would have been required to
take only 5 years of primary schooling and three years of lower secondary schooling. Despite the
fact that basic schooling up to and including the lower secondary stage has been mandatory in
Egypt since 1991, the law is not strictly enforced and non-enrollment or school dropout before
the mandated requirement is not uncommon. Prior to 1984, mandatory schooling was confined to
the primary stage.
Children typically enter the education system at age 6. They are generally not allowed to
enter before age 6, and some start late at age 7 or 8. By age 14, they should be in their last year
of basic education. Thus, all the 614 year olds in our sample should be enrolled. Still, we find
that 14.1 percent of girls in that age group were out of school in 1998, and of these79 percent had
7
The sixth year was phased back in starting with the children who entered the first grade in 2000.
12
never been to school. Among 124 year olds, 19.4 percent were out of school, and two-thirds of
these had never attended school. In contrast, only 12.3 percent of boys 124 were not in school,
and less than half of these had never attended school. Thus non-enrollment appears to be a more
important reason than dropout for girls’ lack of school attendance, and it appears to be a more
common reason for girls than for boys.
According to the 1998 ELMS, most enrolled 614 year old girls and boys attend public
schools (89.6 percent in 1998), with the remainder split between private schools (7.6 percent)
and religious schools (2.8 percent). On average, 614 year olds spend 5.9 hours per day in
school. Many schools work in shifts: 46 percent of children are in schools with more than one
shift, with shifts typically being held in the morning and afternoon. In a school with a given
number of shifts, school hours are not flexible, so that school attendance is essentially a zero-one
variable.
Enrollment increased substantially between 1988 and 1998. From a comparison of the
1988 LFSS and the 1998 ELMS, we see that rural girls the group with the lowest enrollment
rates benefited disproportionately. In 1988, 65 percent of rural girls ages 6 to 14 were enrolled,
and by 1998 this proportion had risen to 81 percent. In contrast, rural boys’ enrollment increased
from 88 to 91 percent. Urban children had substantial increases in enrollment as well, albeit from
higher initial levels. Urban girls' enrollment rates went from 89 to 93 percent and boys from 93
to 95 percent. Detailed analysis of the determinants of school attainment in Egypt shows that
although the proportion of children who have ever enrolled and the retention of enrollees have
both increased in Egypt, and the gender gap in these variables was reduced from 1988 to 2005,
failure to ever enroll continues to be a serious problem for girls. Analyzing data from
Demographic and Health Surveys, Ray Langsten and Tahra Hassan (2007) show that failure to
13
enroll accounts for 75 percent of girls’ failure to complete basic education in Egypt. Retention
rates for those who do enroll are similar for boys and girls and exceed 95 percent. Langsten and
Hansen also show that failure to enroll is strongly concentrated among poor girls in rural Upper
Egypt.
That failure to ever enroll is so clearly associated with poverty raises questions about the
affordability of schooling for poor rural families, both in terms of out-of-pocket expenses and the
opportunity costs of a girl’s time. Although education is ostensibly free at all levels, families
must pay school fees that amount to about LE 20 per month (US $3.50), as well as bear the costs
of uniforms. Moreover, a majority of parents in Egypt believe that in order for their children to
succeed in school, they must provide them with supplemental tutoring. According to data from
the 1998 ELMS, 53 percent of students 614 take some form of supplementary tutoring, most of
which is in the form of private tutoring either in groups or as individuals. Other forms of
supplemental tutoring include after-school programs and programs organized by local mosques
or voluntary organizations. The average expenditure on tutoring in 1998 was LE 15/month (US
$2.80/month). More in-depth analysis of the 1998 ELMS data shows that there were no gender
differences in either the incidence of tutoring or expenditure on it (Asmaa Elbadawy, Ragui
Assaad, Dennis Ahlburg, and Deborah Levison 2004). One of the main contentions of this paper,
however, is that the opportunity cost of a girl’s time at home is also a major impediment to girls’
schooling. If parents believe that the most valuable use of a girl’s time is in the home, they may
never enroll her in school, even if her actual involvement in household chores would only start
several years later.
Physical access to school could also be a major impediment to ever-enrollment for poor
rural girls, especially those living in small hamlets and satellite villages (Ray Langsten and Tahra
14
Hassan 2009: 35). Although 99 percent of Egyptian villages have primary schools, many girls
living in hamlets and satellite villages must walk through fields to get to school, raising parents’
concerns about their safety. Girls’ access to schooling is one area where considerable progress
has been made in Egypt in recent years. Besides the massive school building campaign by the
Ministry of Education, which managed to build 11,000 primary and preparatory schools between
1992 and 2001, there have been a large number of special initiatives directed expressly at girls’
schooling (Egypt, Ministry of Education 2003).
Work
Until 1996, children were allowed to begin working outside the home at age 12 under
certain conditions. The minimum age of work was increased to 15 in 1996 to bring it into line
with the age of mandatory schooling. Despite these laws, a sizeable proportion of children
continue to work.
8
The interviewers conducting the 1998 ELMS survey were carefully trained to
inquire about childrens work with a great deal of sensitivity. Instead of simply asking whether
an individual was working, a series of screening questions about specific activities, such as
participating in family projects or learning a trade in a workshop, were asked. Although the
survey instructions generally required that the individual respond to the questions in the
individual questionnaires him or herself, an exception was made for children under 15, on whose
behalf an adult member of the household could respond. This exception was made to ensure that
the respondents understood their right for informed consent. Parents opted to respond on behalf
of their children in 80 percent of cases and allowed children to respond for themselves in 20
8
For an overview of child labor in Egypt, see Kawther Abu Gazaleh, Lamia Bulbul, Soheir Hewala, and
Suadad Najim (2004) and Zibani (2002).
15
percent of cases. When the proportion of children 6 to 14 reported to be working is tabulated
against a variable indicating whether there was a proxy respondent, we find that there is no
significant difference in activity rates by type of respondent for either boys or girls when we use
the market definition of work (defined below). This suggests that proxy respondents were not
systematically underreporting child market work. However, there is a substantial difference in
reporting of girls’ domestic work. Parents’ underreporting of this kind of work can probably be
attributed to different perceptions between parents and children about what constitutes domestic
work (Levison 2000).
In our inclusive definition of work, we incorporate three categories of activities: labor
market work, subsistence work, and domestic chores performed by women and girls at home.
9
Although the last is not considered employment according to international definitions of
economic activity (International Labour Organization [ILO] 1982)), this kind of work can
interfere with school attendance and performance. Labor market work includes productive
activities for the purpose of market exchange. Only 106 thousand, or 1.7 percent of girls ages 6
14, performed market work in 1998. Subsistence work includes activities involving the
production and processing of primary goods for purposes of household consumption, such as
feeding and caring for livestock and poultry, or making butter or cheese. Domestic work includes
cooking, errands, house cleaning, collecting water, laundry, and childcare. We suspect, however,
that some chores, such as childcare, are underreported. In the most comprehensive study to date
measuring child work, Pamela Reynolds (1991) illustrates the difficulties of capturing these
kinds of tasks. Indeed, as shown above, the greatest discrepancy between children and their
9
Similarly, Marcel Fafchamps and Jackline Wahba (2006) examine three types of work, including market
work, subsistence work, and household chores.
16
parents in reporting child work had to do with domestic work. Overall, we find that 2.7 million
girls (42.8 percent) are working under the inclusive market, plus subsistence, plus domestic
work definition of work.
10
Reference period
Labor statistics including average weekly hours of work are collected for two reference
periods: a short reference period of one week and a long reference period of 3 months or one
year (3 months in the 1998 ELMS). In light of Deborah Levison, Jasper Hoek, David Lam, and
Suzanne Duryea’s (2007) findings that short reference periods can result in intermittent child
workers being misidentified as being out of the labor force, we opted to use the more inclusive
long reference period even though it may include some summer work. This 3-month reference
period includes the months of August, September, and October 1998. It turns out that the change
in reference period makes very little difference for girls ages 614. It appears that if a girl works,
she works all year round.
Hours worked cut-off
A further issue in the detection of work among children is the number of hours per week
that such work is undertaken. The international recommendations are to consider an individual
who is engaged in an economic activity for at least 1 hour per week as employed. Since our
interest is in detecting the kind of work that can potentially interfere with a childs schooling, we
use a higher threshold of 14 hours per week to identify a working child. A girl is considered
10
These figures are based on a one-week reference period with a one-hour minimum threshold. The figure
is 2 million (32 percent) if a 14-hour per week threshold is used instead.
17
working according to the inclusive definition if she participates in either domestic, subsistence,
or market work for 14 or more hours per week for all the activities combined.
11
We experimented
with different hour cut-offs, and results are shown in Figure 1 and Table 1, with a discussion
below. Predictably, as the requirement for being considered ―working‖ becomes more
demanding, fewer girls fall into the working category. We chose a 14-hour cut-off based on
evidence presented below plus the ad hoc assumption that while two hours of work per day (on
average) might not greatly hinder schoolwork, more than that could become problematic.
12
In the
sensitivity analysis section below, we test the robustness of our regression and simulation results
to changes in the hour cut-off.
Interrelated patterns of work and school
Table 1 provides the means and standard deviations for our dependent variables,
weighted by the appropriate sampling weights, while Figure 1 shows how school and work
patterns change with age. For purposes of comparison, Figure 1 includes school attendance and
labor market work of boys as well as girls, and it shows ages 617 to provide a broader
11
Because of the way the questionnaire is designed, we do not observe the number of hours in subsistence
work for girls who are engaged in market work; thus, our inclusive measure will understate the total hours of work
for the small number of girls engaged in both market work and subsistence work.
12
Our reasoning is not completely ad hoc. Out of 24 hours, if a child sleeps for 9 hours, spends 2 hours on
meals and self-care, 6 hours in school, 1 hour going to and from school, 1 hour studying, 0.5 hours in tutoring
sessions (about 53 percent of in-school 614 year-olds are tutored in 2006), and 2 hours in various kinds of work
activities, that leaves only 2.5 hours for TV-watching, play, and other social activities. William Kandel and David
Post (2003) also used a 14-hour cut-off, with similar justifications.
18
overview of time use patterns in adolescence.
13
Among girls 614, 85.8 percent attend school,
and 14.2 percent do not (Table 1). In contrast, half as many boys (7.6 percent) are out of school.
Girls’ work responsibilities increase steadily with age, with 60 percent of 14-year-olds engaged
in at least 14 hours of inclusive work. Table 1 and Figure 1 show that the hour cut-off used does
make a difference in the measurement of inclusive work. The broadest definition, inclusive work
for at least one hour per week, leads to a steep increase in the percentage of girls working as age
rises, so that by age 14, 70 percent of girls are included in this group of workers. In contrast,
applying higher cut-offs systematically reduces the percentage of girls counted as workers,
resulting in a fan-shaped set of trend lines. At age 14, the biggest percentage difference (20
percent) appears when we contrast a 14-hour cut-off (2 hours of work per day on average) with a
21-hour cut-off (3 hours of work per day on average).
Figure 1 shows that as children become adolescents, they take on more adult roles and
responsibilities, which are reflected in their activities. School attendance decreases and is
replaced by substantial increases in the proportion of children engaged in market work (for boys)
and domestic work (for girls); the proportion doing both kinds of work rises as well (not shown).
Since our primary concern is to study how schooling and work interact, Table 1 presents
the cross-classification of the schooling and work variables with the 14-hour per week cutoff
used throughout the rest of this paper. As part of a sensitivity analysis, we also show these with a
number of other possible cut-offs. First, however, we separate out market work. Very few girls
participate in market work (1.7 percent), and among those who do, few manage to combine work
13
The slight dip in school attendance at age 10 is most likely due to age heaping. Lack of precision about
age (and, thus, heaping on 10) is more likely for those children who do not go to school because parents of children
who are out of school are likely to be illiterate and therefore not sure of the ages of their children.
19
with school. Domestic work, on the other hand, appears to be compatible with schooling to some
extent, as captured by the inclusive work variable. This can be seen by examining the proportion
of girls both at work and in school, which is never greater than 31 percent but remains above 20
percent when the working group includes those working at least 14 hours per week. It declines
substantially (to 9.5 percent) when the cut-off is at 21 hours per week.
According to the market work definition, about 13 percent of girls ages 614 are neither
in school nor working (so-called ―idle‖); when the inclusive definition of work is used, it
becomes clear that few girls are ―idle‖ (see Table 1). Thus the persistently higher proportion of
girls who are typically reported to be neither in school nor at work hinges crucially on how work
is defined for girls.
Weekly hours worked are high on average 49.4 hours for girls 614 engaged in at
least one hour of market work. For girls doing at least one hour per week of inclusive work, the
average is 12.3 hours worked per week. Applying the 14-hour cut-off to our sample eliminates a
number of girls working low hours doing subsistence and household work, and it raises the mean
to 25.7 weekly hours. The 28-hour cut-off reduces the sample to those girls who are essentially
full-time workers, as indicated by their mean of 41 work hours per week.
Overall, the most striking feature of Table 1 is the extent to which official definitions of
work (―market work‖) understate the work of girls. When work is broadly defined to include
domestic and subsistence work, nearly one-third of girls ages 614 work for at least 14 hours per
week. As seen in Figure 1, there is a shift upward in girls’ workload between ages 11–3,
simultaneously with a decline in school attendance; this may correspond with the greater
restrictions on mobility placed on Egyptian girls at puberty (Barbara S. Mensch, Barbara L.
Ibrahim, Susan M. Lee, and Omaima El-Gibaly 2003). Girls’ workload increases again at 156.
20
This is likely explained by evidence collected by Mensch, et al. (2003), who document that about
half of 102-year old boys participate in domestic chores, compared to only about 20 percent of
169 year old boys. They write:
The activity profile that emerges by late adolescence reflects the expected patterns for
Middle Eastern societies: boys have considerably more free time than girls, and they tend
to spend that time outside the home, with friends, engaged in sports, or visiting friends’
homes. Work roles become gradually segregated, with boys more likely to participate in
paid labor while girls participate in domestic work within the household (14).
The cultural expectation is that girls are being prepared for roles as homemakers and
mothers, and their early involvement in domestic work is seen as part of their preparation for
these adult roles. Especially for poor women, this makes a lot of sense, as these women have
extremely limited access to the paid labor market in Egypt (Ragui Assaad and Melanie Arntz
2005).
METHODOLOGY
Our econometric model, described below, estimates the effect of work on school
enrollment, while allowing for endogenous work status. It also allows for schooling to affect
work, but since the measurement of that effect is not the focus of this paper, we specify the work
equation as a reduced form.
The two binary dependent variables are described in detail above. In brief, one of the
dependent variables indicates whether or not the girl is in school at the time of the survey. (Our
schooling variable is based on a question about whether or not the child is currently in school,
21
which was interpreted as whether or not the child was attending school, and not merely enrolled.)
While it would be ideal to have measures of educational outcomes such as grades received, test
scores, and grade promotion and repetition rates, these are not available for our sample. Girls
who are not currently enrolled have either never been to school or have attended for some time
and then dropped out. In the sensitivity analysis we examine alternative specifications of our
schooling dependent variable that capture whether a girl had ever been enrolled and, if so, if she
had dropped out (current enrollment conditional on having ever been enrolled). The other
dependent variable indicates whether or not the girl is engaged in work for at least 14 hours per
week on average in the three-month reference period.
Control variables
The control variables discussed here are common to both the work and schooling
equations. Additional instruments included in the work equation are discussed in the
Identification Strategy section below. We estimate a sequential set of models, adding in each
subsequent model control variables that might be argued to be endogenous. Model 1, the most
basic model, includes characteristics of the girl (age and her relationship to the head of the
household); parental age and education; whether or not they are present in the household, and if a
father is absent whether or not this absence is temporary; region; and the household’s wealth
quintile, separately for urban and rural areas. Model 1 also includes community-level controls:
the proportion of male and female populations in the locality with secondary education or
22
above; and the proportion of girls and boys aged 611 and 124 enrolled in school.
14
Mindful of
the fact that father’s employment status, the presence of a family business, and household
composition are sometimes considered endogenous to other household decisions, we estimate
two additional models. Model 2 adds variables related to the father’s employment and the
presence of household enterprises. Household composition variables are included only in Model
3. Descriptive statistics for the explanatory variables are presented in Table 2.
As a result, Model 1 is our most defensible model. We include Models 2 and 3 despite
some misgivings about the possible endogeneity of the father’s employment status and
household composition variables because of our conviction that these are potentially important
factors in household decision making.
Controls were included because they have been shown to be relevant to child work and
school in either theory or previous empirical research, or both. Virtually all empirical work on
child labor has indicated that the age and gender of the child are important determinants of the
child’s educational and work activities. The child’s relationship to the household head might also
have an effect. In many countries, children who are not sons or daughters of the reference person
are distant relatives who are filling the role of domestic servants. In the Egyptian case, most of
them are grandchildren of the head; in fact, we find that they tend to work less than girls who are
daughters of the head of household.
We hypothesize that parents’ ages affect child activities. Younger parents are likely to be
at a more resource-constrained point in their lifecycle and may have less ability to pay school-
14
We include a cluster correction to take into account that these variables are not independent across
observations in the same primary sampling unit (PSU). There are 200 PSUs in the sample. A village or
neighborhood was typically represented by at most one PSU, with the exception of one or two that had two PSUs.
23
related fees, as well as a greater need for their children’s labor. We include measures of the
father’s age and the mother’s age when the child was age 6 to capture this effect.
There is ample empirical evidence in the literature that the education of the parents
decreases the probability of working and increases the probability of schooling for children in
developing countries (Christiaan Grootaert 1998; Christiaan Grootaert and Ravi Kanbur 1999;
Sonia Bhalotra and Christopher Heady 2003; Zafiris Tzannatos 2003). In our models, father’s
and mother’s education are specified as continuous variables; sets of dummy variables were tried
as well but were found to provide similar results.
When fathers are absent from the household in Egypt, it often implies that they have
migrated to an oil-rich Arab country to work; such fathers generally are in contact with their
families and may send remittances to them. We therefore distinguish between the temporary
absence of the father and his permanent absence, as would be the case for widowed and perhaps
also divorced mothers; our expectation is that a father’s permanent absence has more negative
consequences for his children than his temporary absence. If a mother is found to be married and
her spouse not present, we deem the father’s absence temporary. We suspect that children living
with their father and a stepmother or with a single father may be treated differently than children
living with their father and their birth mother; four dummy variables describe these various
possibilities. Because child fostering is not common in Egypt, the presence or absence of parents
is unlikely to be endogenous to decisions about child schooling.
Mark R. Rosenzweig (1977) and others argue that the substitutability between the work
of girl children and that of her mother makes the mothers employment status endogenous. When
mothers work outside the home, girls may stay home to take over their duties, and a mother who
24
has a daughter who is old enough to care for her siblings is more able to engage in labor force
work. Because of this we omit the mothers employment status from the explanatory variables.
Father’s employment, however, is arguably exogenous to decisions about child activities.
Whether the father is employed or not and the nature of the father’s employment both matter. If
the father is unemployed or in irregular employment, income from a girls’ market work may be
considered a substitute for the father’s labor income, increasing the probability that she does not
go to school. Furthermore, the effect of the father being an employer or self-employed as
opposed to an employee is important because it raises the probability that the girl will be an
unpaid family worker. To allay concerns about endogeneity, we capture the fathers employment
status at the time the girl was 6 years of age rather than at the time of the survey. A series of
dummy variables describes the father’s sector, type of employment, and employment status.
15
To measure wealth, we use an asset index, constructed separately for urban and rural
areas to determine how the position of the girl’s household in the wealth distribution affects her
work and schooling status. The construction of these wealth indices is described in the appendix.
15
Irregular private sector work is the omitted category. ―Regular private sector‖ jobs
consist of permanent and temporary but continuous jobs in the private sector, while ―irregular
private sector‖ jobs consist of intermittent and seasonal jobs. Public sector work is typically
regular. Non-wage workers are either employers, self-employed workers, or, in some rare cases,
unpaid workers for a family enterprise. Non-working fathers are either unemployed or out of the
labor force. We expect that fathers in some types of positions are more likely to be able or
willing to take their sons (but not daughters) to work with them. Non-working fathers may stay
home and generate more household work for daughters.
25
Since child labor is a phenomenon that primarily affects poor children, the top three quintiles of
wealth were combined into a single category.
For purposes of this analysis, we identify six regions in Egypt. The Greater Cairo region,
which includes the entire Cairo urban agglomeration, is the reference region. Alexandria and the
Suez Canal cities are lumped together as the other metropolitan region. The non-metropolitan
urban regions include urban Lower Egypt, which comprises the cities of the Nile Delta, and
urban Upper Egypt, which comprises the cities of the Nile Valley south of Cairo. The rural
components of Lower and Upper Egypt are the fifth and sixth regions we consider. Upper Egypt,
and especially its rural component, is generally considered to be the poorest and most socially
conservative region in Egypt. Community-level controls are derived from 1996 population
census data at the village and neighborhood level. They include the percentages of the adult male
and female populations with secondary schooling and above and the percentages of children of
both sexes in the age groups 611 and 124, who are enrolled in school.
Detailed age/sex categories for household members were included in Model 3 to examine
the effects of household composition on children’s work and schooling. Although these variables
are potentially endogenous, Janet S. Netz and Jon D. Haveman (1999) argue strongly for their
inclusion in labor force models. Many other researchers have found that children and
adolescents’ responsibilities depend on who else is available in the household to do labor force
work, household tasks, and childcare (e.g., Freije and Lopez-Calva 2001; Deborah S. DeGraff
and Richard Bilsborrow 2003). Child activities may act as complements or substitutes for the
activities of these other household members.
In preliminary analyses, we included two variables to control for the accessibility of
schooling in the child’s community: walking time to the nearest age-appropriate school and a
26
dummy indicating that there is no school within walking distance. In addition, a variable
indicating whether or not the child attended a school with multiple shifts was included to proxy
for school quality. The walking time to school variable and the school shifts variables were never
statistically significant in any of the model specifications. The dummy for ―school not in walking
distance‖ seems to have been mis-measured. These three variables were dropped from all
models.
Estimation
Our full-information maximum-likelihood estimation approach relies on a model of
binary choice with work as a binary endogenous regressor. The general framework for
examining the effect of endogenous treatments on discrete outcomes is laid out in Arild Aakvik,
James Heckman, and Edward Vytlacil (2005). They develop a model to study the impact of
interventions on discrete outcomes when responses to treatment vary among observationally
identical individuals. The model offers a way to control for selection in determining the effect of
the treatment and to estimate an average treatment effect, as well as an effect of the treatment on
the treated. The model potentially allows the effect of observables and unobservables on the
outcome to differ for the treated and untreated sub-samples. In our case, the ―treatment‖ is child
work and the ―outcome‖ is school attendance. A similar model is applied to the effect of working
while in school on school progress in a paper on Bangladesh by Jose Canals-Cerda and Cristobal
Ridao-Cano (2004). Canals-Cerda and Ridao-Cano estimates a switching probit model, where a
separate schooling equation is estimated for working and non-working children. The switching
probit model was also applied in a study of the effect of fertility on women’s labor force
participation in the US by Raquel Carrasco (2001).
27
The model
We assume there are two potential binary schooling outcomes (S1 and S0) for the working
(W=1) and non-working (W=0) states, respectively, where Sk=1 if the child is attending school
and Sk=0 if s/he is not (k = 0, 1). The observed schooling outcome is given by S= W S1 + (1-
W)S0; the subscript indexing an individual child is suppressed. The observed binary outcomes
are generated according to underlying latent index structure as follows:
0 iff )0()0(
1 iff )0()0(
)0()0(
00
*
00
11
*
11
*
WXSS
WXSS
ZWW ww
11
11
11
where
*
1
*
0
* and , SSW
are latent variables indicating the difference in the households
utility between putting and not putting the child to work and sending and not sending the child to
school, respectively. These decisions are not necessarily sequential; the methodology used
allows for the two decisions to be simultaneous. Z and X are vectors of regressors, with the need
to have at least one regressor in Z that is not in X for purposes of identification. Given the
relatively small number of working children in our sample, we refrain from estimating a full set
of schooling equation parameters for the working subsample by imposing the restriction
01
XX
, so that the effect of work on schooling introduces an additive shift in the
schooling equation. The disturbances
are assumed to be normally distributed with zero
mean and covariance matrix as follows
1
1
1
0
110
w
w
28
Since S1 and S0 are never jointly observed,
10
is not identified. We started by estimating
models that did not impose the restriction that
ww
01
and tested whether the restriction can
be upheld. Since the restriction was upheld in all cases where we obtained convergence, we only
present results for models where the restriction is imposed.
16
The restriction
ww
01
implies
that pooling working and non-working girls into a single schooling equation does not introduce
selectivity bias. Since the restriction is upheld, the model we estimate reduces to a bivariate
probit model with work as a dummy endogenous variable in the schooling equation.
Consistent estimates of the parameters
www
010 ,,,,
can be obtained using full
information maximum likelihood methods. The log-likelihood function is given by:
0,1 1,0 0,0 000110
1,1 11 loglogloglog
SW SW SWSW PPPPL
The joint probabilities of working and attending school are given by:
,,]0,0Pr[
,,]1,0Pr[
,,]0,1Pr[
,,]1,1Pr[
200
201
210
211
sw
sw
sw
sw
XZSWP
XZSWP
XZSWP
XZSWP
where
2
is the bivariate normal distribution function and the individual subscripts i are
suppressed for clarity.
16
We obtained the following test statistics for a Wald test of the equality of the two correlation
coefficients: Model 1 had 2(1) = 0.05, p-value=0.821; Model 2 had 2(1) = 0.07, p-value=0.793; and the
unrestricted Model 3 did not converge. An insignificant test statistic upheld the restriction.
29
Identification strategy
To identify the structural schooling equation we need instruments that can be excluded
from the schooling equation but that have some explanatory power in the work equation. Since
we are examining both market work and household work, our instruments need to proxy for the
demand for such work and be exogenous to household decisions on child work and schooling.
17
To proxy for demand for market work, we use instruments that indicate the prevalence in the
local community (village or neighborhood) of the occupations in which children are most often
found. We conjecture here that most girls work close to home, so that it is local labor market
conditions that will determine the demand for their market labor. Given the structure of labor
markets for educated workers in Egypt, we suggest that decisions on schooling are made on the
basis of returns to schooling in a much broader regional or national labor market and will thus
not be affected by local labor market conditions. The instruments we use to proxy for the
demand for market work are the percentages, among the working age population of the locality,
of workers in service and trade occupations, and in agriculture. These percentages are obtained
from the 1996 population census for the village or neighborhood in which the girl lives.
To proxy for the demand for domestic work, we use the household’s access to basic
public services: piped water, piped sewage disposal, and garbage collection. The absence of such
services is expected to substantially increase the domestic burden of women and girls. It is often
the job of girls and young women to go fetch water from the closest source; and children
typically are in charge of carrying waste to the closest dumping area for disposal (Kawther Abu
17
Victor Levy (1985) relies on changes in cropping patterns to identify the effect of child labor in
agriculture on the total fertility rate in Egypt.
30
Gazaleh, Limia Bulbul, Soheir Hewala, and Suadad Najim 2004: 910). A frequent image in
rural Egypt is one of girls and women washing dishes and clothing along canals and waterways;
the absence of running water also makes these regular tasks more onerous and time-consuming.
To ensure that the absence of such services is not affecting the girl’s schooling status through
some correlation with the quality of schooling in the community, we include in both the work
and schooling equations direct measures of access to schooling in the girl’s community: the
percentages of the adult male and female populations with secondary schooling and above and
the percentages of children of both sexes in the age groups 611 and 124, who are enrolled in
school. We also correct for the urban and rural status of the community through the region
dummies, which are for the most part highly statistically significant. Having done this, we find
that the access to service variables are jointly insignificant in the schooling equation, in an
unrestricted version of the model. (See the overidentification test in the section entitled
Simulation results and sensitivity analysis.) This implies that variations in home infrastructure
are independent of the availability and quality of schools once we control for location and
community-level enrollment rates. We maintain that variables indicating access to urban services
in Egypt are for the most part exogenous to household decisions since they are a function of
where the household resides rather than the result of a separate decision-making process about
whether or not to purchase the service in question. Given the rigidity of the housing market in
Egypt, and the resulting relative immobility of households, decisions about where to reside are at
the very least pre-determined if not completely exogenous.
18
18
According to the 1998 ELMS, only 6.3 percent of adults who ever worked changed their place of
residence in the ten years previous to the survey.
31
ESTIMATION RESULTS
The endogenous work and schooling variables take on the value of one when a girl is in
school and when a girl is working, respectively. Girls are defined as working if they work at least
14 hours per week in the labor force and/or on subsistence production and/or on domestic tasks.
The results are shown in Table 3. Recall that we estimate a sequential set of models, adding in
each subsequent model variables that might be argued to be endogenous (see Variables section).
As a general rule, our results on the coefficients of variables entered in Model 1 are robust to the
inclusion of variables entered in Models 2 and 3, implying that if there is a simultaneity problem,
it does not bias the estimates of other statistically significant explanatory variables. For binary
outcome models, marginal effects are necessary to allow us to speak to the magnitudes of
particular effects. In bivariate probit models, marginal effects need to be evaluated in relation to
the predicted probability of each work/school state for the reference girl. The marginal effects
shown in Table 4 are based on the Model 1 specification.
The effect of work on schooling
Perhaps the most significant finding of this research is that girls’ inclusive work has a
substantial direct effect on schooling. As shown in Table 3, the coefficient of the ―currently
working‖ variable, which indicates the effect of work on the probability of schooling, is negative
and highly statistically significant for all models. As shown in Table 4, the predicted
unconditional probability of work for the 14-year-old reference girl is 46.1 percent and the
32
unconditional probability of schooling is 83.3 percent.
19
Because many girls are responsible for
household work, the probability of going to school and not working is only 37.6 percent. She is
more likely to combine work and schooling, with a probability of 45.7 percent of doing so, but
she also has a very small probability 0.4 percent of working and not going to school. The
reference girl has a relatively high probability 16.3 percent of neither working nor going to
school.
A 10 percentage point increase in the probability of work reduces the marginal
probability of schooling by six percentage points. This results in an estimate of -0.33 for the
elasticity of the probability of school attendance with respect to work for the reference girl. It
also reduces the joint probability of being in school and not working by 4.6 percentage points
and increases the joint probability of being in school and working by 1.5 percentage points.
Large positive and statistically significant estimates of indicate that unobservables that
raise the probability of work also raise the probability of schooling, after the impact of work
itself is controlled for. Girls who work (mostly in their own homes) appear to be those who
would otherwise be more likely to remain in school.
19
Because work is endogenous to the schooling decision, some value has to be set for the probability of
work to calculate the marginal probability of schooling and the joint probabilities of work and school. We use a
value of 0.46, the unconditional probability of work of the reference girl. For the other characteristics of the
reference girl, see note to Table 4.
33
Simulation results and sensitivity analysis
In this section, we conduct simulations for girls of various profiles while examining the
sensitivity of our results to model specification, the use of different hour thresholds for defining
work, and an alternative way of defining the schooling variable. In each case, to estimate the
effect of work on schooling, we use the estimated coefficients from the models to predict the
conditional probability of school attendance for a girl of given characteristics if she works or
does not work. The effect is evaluated at the values of the explanatory variables that correspond
to the characteristics of a reference girl, a ―most-vulnerable‖ girl, and a ―least-vulnerable‖ girl.
The reference girl is 14 years old, has both parents living at home and with mean years of
schooling. She lives in Greater Cairo and belongs to a household in the lowest urban wealth
quintile in a neighborhood with the mean percentage of service, trade, and agricultural workers,
the mean percentage of men/women with secondary education and above, and the mean
enrollment rates for children 611 and 124. In addition, in Model 2, her father is an irregular
private sector worker, and her household has no farm or non-farm enterprises. In Model 3, she is
in a household that has no other members besides parents. The most vulnerable girl is similar to
the reference girl but her father has no formal schooling and her mother is absent. In Model 3,
she lives in a household with one child under two, another between 6 and 9, and a sister between
10 and 14. The least vulnerable girl lives in a household in the top three urban wealth quintiles,
has a father with 12 years of schooling and a mother with 9, and, in Models 2 and 3, her father is
a public sector worker. In other respects she is similar to the reference girl.
The simulation results shown in Table 5 are consistent with the scenarios corresponding
with the assigned characteristics of the reference girl, most vulnerable girl, and least vulnerable
girl. In each column, the probabilities of being in school conditional on not working and
34
conditional on working are generally highest for the least vulnerable girl and lowest for the most
vulnerable girl. In addition, the relative effect of work on schooling is generally smallest for the
least vulnerable girl and largest for the most vulnerable girl, with the reference child somewhere
in between. These patterns hold for Models 1, 2, and 3. Not only are the most vulnerable girls
least likely to go to school irrespective of work responsibilities, but work also places them at an
even greater disadvantage than it does other children. Even when she does not work, the
probability of schooling for the most vulnerable girl is substantially lower than that of the
reference girl, and the impact of work on school attendance for the most vulnerable girl is
immense: the probability of attending school is reduced to almost zero.
Shifting the hour threshold from 7 to 28 hours per week affects the magnitude of the
effect of work on schooling in predictable ways. Because higher hour thresholds imply
increasing intensity of work among the sample of working girls, the impact of work on their
schooling is consistently larger. For example, a reference girl who works under the 7-hour cut-
off (Model 1) has a 57 percent decline in her likelihood of attending school, whereas under the
28 hour cut-off the decline is 90 percent. Note that, for the most vulnerable girl, the probability
of schooling declines with increasing hour cut-offs even when the girl is not working. This is
because the non-working sample includes an increasing share of girls who are working, although
their work hours are below the threshold. This provides additional support for our decision to use
a lower, 14-hour, cut-off.
We ran additional specifications of the model (regression results not shown) to determine
the mechanism by which working reduces girls’ school attendance. There are two routes
whereby girls’ work may affect their school attendance. First, parents who intend their daughters
to contribute substantially to household production and their sons to engage in market work may
35
decide that it is not worthwhile to send their daughters to school at all. Second, girls who are in
school may find work interfering with their ability to maintain their studies or may find school
impeding their work and then drop out of school. Our main model does not distinguish
between these reasons.
We examine the effect of work on initial school entry by changing the dependent
schooling variable from ―currently enrolled or not‖ to ―ever enrolled in school or not.‖ We also
try a model that captures school dropout among those who have ever enrolled, using the same
dependent variables as in our main model but with the sample limited to those children who have
ever been in school. Due to the small number of 614 year-olds in the sample who started school
then dropped out, the school dropout model did not converge. In the ever-in-school model, the
effect of work is significant under all three model specifications. Thus, the future prospect of
work affects school entry. As shown in the last column of Table 5, the effect of work on the
probability of ever attending school is smaller than its effect on current school attendance. For
the reference girl (Model 1), the relative decline in the probability of ever-attending school due
to work is 35 percent, as compared to a 62 percent decline in the probability of currently
attending school. This result suggests that some parents see the role of girls as being limited to
the home, and they will not send a girl to school, even if her work responsibilities begin well
beyond the age of school entrance.
Our final simulations are based on varying the level of service infrastructure available to
girls and their families, thus endogenously shifting the girls’ work burdens and, in turn, affecting
their likelihood of attending school. In these simulations, we continue to examine the impact of
using different hour cut-offs to define work. As shown in Figure 2, eight different combinations
of the access to service variables (piped water, sewer, garbage collection) ranging from all three
36
services present (on the left side) to all three services absent (on the right side) are used to
exogenously shift the probability of work. Clearly, the variable with the largest impact is the
availability of piped water, which seems to particularly affect the likelihood of work when work
is defined using the higher hour thresholds.
In Figures 3 through 5, the different bundles of services provide the variation that drives
the predicted probability of work plotted on the x-axis. The left-most end of each curve
corresponds to the left-most bundle in Figure 2 (all services present). Each additional movement
to the right corresponds with the next bundle in Figure 2, with the right-most being no services
present. The y-axis shows the resulting change in the predicted probability of attending school. In
Figure 3, each of the curves represents the relationship between the predicted probabilities of
work and schooling when work is defined using a particular hours cut-off. Although the level of
work implied by each cut-off is quite different, the slopes of the curves are very similar,
implying that the effect of work on schooling is robust to changes in definition.
Effects of control variables on work and schooling
The most interesting effects of control variables on work and schooling outcomes are as
follows:
20
Parental education is typically regarded as a primary determinant of child activities.
We find that a father’s but not a mother’s education has a negative effect on whether a girl
works. Both parents’ education has a positive effect on a girl’s schooling. For a 14-year-old
reference girl, parental education levels increase the predicted probability of schooling by 0.8
20
A more detailed discussion of the control variables can be found in Ragui Assaad, Deborah Levison, and
Nadia Zibani (2007).
37
percentage points per year of her father’s schooling and by 1.3 percentage points for each
additional year of her mother’s schooling (Table 4).
21
The absence of parents appears to substantially increase a girl’s likelihood of working.
As shown in Table 4, the unconditional probability of work for a 14-year-old reference girl
increases by 39.7 percentage points when both her parents are absent. The estimated effects of
the mother-absent dummy variable is indeed very interesting. Girls living with their fathers but
not with their birth mothers have a lower participation in school: the unconditional probability of
school for a 14-year-old reference girl falls by 44.3 percentage points (Table 4).
In Figure 4, we simulate the effect of work on girls’ schooling for different types of
parental absence, again with the eight different bundles of services underlying the variation in
the probability of work. Girls with absent mothers whether or not a stepmother is present
have much lower predicted probabilities of both work and schooling, compared to girls with both
parents present. This suggests that a third unmeasured activity might be interfering with school
attendance for these girls. Our conjecture is that childcare as a domestic work responsibility is
substantially underestimated. Girls with absent mothers might be engaging in substantial
amounts of unmeasured childcare, leading to this result. Girls with both parents absent have
much higher levels of work and, as a result of that work, lower school attendance. Whether these
girls are living with foster parents or other relatives, they are clearly expected to work no matter
how many services are available to the household.
21
In the absence of controls for the father’s income, the father’s education variable may partially capture
the effect of income on a girl’s work and schooling. We address this to some extent by including the wealth
quintiles, which should be highly correlated with family income.
38
With respect to region, it appears that the large schooling disadvantage of girls in rural
Upper Egypt compared to the metropolitan regions that showed up in the descriptive statistics
disappears once other variables, including work, are controlled for. This suggests that the low
rates of girls’ schooling in Upper Egypt can be attributed more to the fact that they are more
likely to work rather than to the social conservatism of the region, as is often assumed.
Wealth has the expected positive effect on schooling of girls, but it is only statistically
significant in urban areas. Girls’ work is quite responsive to wealth: declines relative to the
lowest quintiles are found in all but the second lowest urban quintile. Still, even in the top three
urban quintiles, a 14-year-old reference girl has a 36.8 percent unconditional probability of
working, compared to 46.1 percent in the lowest urban quintile (Table 4). Thus, by the time the
top three quintiles of wealth are reached, a 14-year-old reference girl has had a fall in her
probability of working of only 9.3 percentage points in urban areas. In rural areas, the
corresponding decline is 13.4 percentage points. Domestic work appears to be part of girls’
responsibilities even in middle class households.
22
Effects of household wealth and region are simulated in Figure 5. We contrast Greater
Cairo, an urban area, with rural Upper Egypt, the most disadvantaged region in Egypt. Girls in
the rural region clearly work more, regardless of services available, but they are not intrinsically
less likely to go to school than their urban counterparts. To the extent that their schooling levels
are observed to be lower, it is because they work more. The absence of services appears to have a
bigger impact on increasing work burdens for girls in urban areas: the poorest urban girls with no
services fare worse than the poorest rural girls with no services, in terms of school attendance.
22
Eva Mueller (1984) documents that rural children in Botswana are more likely to work if their families
are wealthy enough to own complementary assets, such as land, farming implements, and livestock.
39
Girls’ work and schooling are affected by the composition of their households. Their
school attendance is hindered by the presence of children under the age of 2 and between the
ages of 6 and 9, as well as by the presence of other girls ages 104. They are more likely to
attend school the more women above age 18 there are in the household. Perhaps because
childcare is not captured very well even by our inclusive definition of work, girls’ work is not
affected by the presence of young children. We are encouraged by the fact that the statistically
significant coefficients of other variables in the model remain significant and their magnitudes
are robust to within a standard error to the inclusion of the household composition variables. The
presence of some household members seems to matter most to the extent that they generate
household work to be accomplished.
Specification tests
In addition to the test for whether the working and non-working girls could be pooled
into a single schooling equation described above, we conducted other tests to examine the
robustness of our model specification. These include (i) an overidentification test to assess the
validity of the exclusion of the instruments from the schooling equation; (ii) a test of the joint
significance of the instruments in the work equation; (iii) a comparison of our model results with
a model that treats work as exogenous; and (iv) a comparison of our results with a reduced form
model that does not explicitly include work as a determinant of schooling. Additionally,
sensitivity analyses described above include three alternative sets of explanatory variables;
alternative definitions for the schooling variable; and simulations for girls with different profiles.
We test the exclusion of the instruments from the schooling equation using an
overidentification test appropriate for non-linear simultaneous equation models. This is a
40
likelihood ratio test that is valid under the maintained assumption of normality of the error terms.
The test consists of not imposing any exclusion restrictions and relying on the non-linear
functional form to obtain a just-identified model.
23
We then carry out a likelihood ratio test of
this just-identified version of the model against the over-identified version, which imposes the
exclusion restrictions on the schooling equation. A statistically insignificant test statistic
indicates that the instruments can be safely excluded from the schooling equation. Our
instruments passed the over-identification test in all the specifications we estimate.
24
Our second specification test assesses the joint significance of the five instrumental
variables in the work equation. The tests indicate that the instruments are jointly strongly
significant.
25
Among our instruments, those that are designed to measure the burden of domestic
work are jointly statistically significant at the 1 percent level, whereas those designed to measure
market work are not, presumably because of the very low incidence of market work among girls.
23
Brian McCall (1992) shows that a binary choice model with a binary endogenous regressor is just
identified without exclusion restrictions if the underlying dependent variable is linearly related to the explanatory
variables and if two of the explanatory variables are continuous. The error terms are also assumed to be continuous.
Lung Fei Lee (1992) shows that this overidentification test statistic is distributed as 2 with degrees of freedom
equal to the number of excluded instruments.
24
The overidentification test has 5 degrees of freedom, one for each of the five instruments. We obtain the
following test statistics: Model 1 had 2(5) = 3.49, p-value = 0.625; Model 2 had 2(5) = 3.65, p-value = 0.601; and
Model 3 had 2(5) = 4.07, p-value = 0.539. Thus, the results are as we had hoped, and exclusion of the instruments
from the schooling equation is upheld. Since the likelihood ratio test does not allow for a cluster correction, this test
was carried out without it.
25
We obtain the following test statistics for a Wald test of joint significance: Model 1 had 2(5) = 28.46, p-
value = 0.0001; Model 2 had 2(5) = 25.83, p-value = 0.0001; and Model 3 had 2(5) = 22.33, p-value = 0.0005.
41
To test what difference it makes to treat work as endogenous, we used an alternative
model specification where work is considered exogenous to the schooling decision and simply
entered as a dummy variable in a probit equation for schooling. In this specification, the effect of
work on schooling is negative and statistically significant at the 1 percent level, as expected, but
the magnitude of the effect is different from that obtained when work is considered endogenous.
When work is not endogenized, the estimate of the impact of work on schooling for girls is
understated. For a 14-year-old reference girl, a 10 percent increase in the probability of work
reduces the probability of being in school by 6 percentage points in the endogenous work model,
compared to only 2 percentage points in the exogenous work model (Model 1). The fact that the
effect strengthens when work is endogenized suggests that there is positive selection, meaning
that girls who work are those who are more likely to succeed in school. That is, the frequent
assumption that academically less-able children are more likely to work is not supported for
girls’ domestic work in Egypt. Tyler (2003) and Stinebrickner and Stinebricker (2003), like us,
also find positive selection into work, but Goulart and Bedi (2008) find negative selection. When
we run the same models for boys’ market work, we also find mild evidence of negative selection,
suggesting that boys who are more likely to fail in school are more likely to work. Thus, the
issue of positive versus negative selection is context specific and probably depends on both
gender and the type of work. In our case, the majority of girls who are not in school never
attended school, so their parents did not observe their academic abilities. In contrast, boys who
are not in school are, for the most part, drop-outs. Presumably they did less well in school than
boys who did not drop out, thus explaining the observed negative selection.
How does our approach compare with the approach common to this literature, which
acknowledges that work and schooling are interrelated but does not attempt to identify the effect
42
of work on schooling? This approach is essentially the reduced form version of our model, where
the work variable is excluded from the schooling equation but the correlation between work and
schooling is captured through the correlation coefficient of the two error terms ( ) in the
bivariate probit model. In the reduced form, is negative and statistically significant at the 1
percent level, reflecting the known negative association between work and school. When the
effect of work on schooling is explicitly modeled, turns positive, suggesting further evidence
of positive selection.
CONCLUSION
Our primary objective in this study was to ascertain the effect of girls’ work on their
school attendance. Since most of girls’ work in Egypt takes the form of domestic work, we
highlight this particular relationship in this paper. We find a substantial effect of work (broadly
defined, with a 14-hour threshold) on girls’ school attendance: a 10 percentage point increase in
the probability of work results in a 6 percentage point decrease in the probability of school
attendance. Our estimation method allows us to determine the net impact of work on schooling,
correcting for both observable and unobservable characteristics of the child and her household.
The results indicate that many girls who work would have been in school had they not been
expected to work 14+ hours per week. Thus work seems to have a direct and detrimental effect
on girls’ schooling.
Is it believable that domestic tasks affect schooling? In fact, households’ need for
―chores‖ – including chores accomplished at particular times of day can be large and relatively
inelastic. Water and/or fuel may have to be gathered from great distances, wastewater and
garbage may require disposal, the day’s primary meal may have to be produced from
43
unprocessed materials and made ready in the middle of the day, the household may include
infants or young children requiring constant watching and regular tending, and laundry may need
to be washed in the morning at the closest waterway so that it will dry before nightfall. The
drudgery of everyday life, especially among the poor, should not be underestimated. Kathy R.
Kamphoefner, who conducted interviews among low-income women in Cairo, writes,
Without home appliances, supermarkets, and the prepared foods that are taken for granted
in the developed world, basic housework, laundry, and food purchase and preparation make for a
full-time occupation. Labor-saving devices, such as major household appliances, are rare in
lower-class households. Cairo’s winds blow in a new layer of dust every day…. Shopping is very
time-consuming, as items must be purchased from many different small shops…. All cooking is
done from scratch. Vegetables have to be scrubbed and rice has to be sifted through by hand to
pick out dirt and small stones. Dishes are commonly washed in cold water and hauled upstairs to
the apartment from the common water faucet. Clothes are laboriously washed by hand in
washtubs and hung off balconies to dry. Basic housework is far more than an exhausting day’s
work. (1996: 95)
The typical image evoked by the words ―child labor‖ is one of a factory worker, toiling
long hours, day after day. In fact, such children are extremely rare in Egypt; according to one
estimate, based on surveys in over 30 low-income countries, only 2.4 percent of children ages 5
14 work outside of their household for pay (Eric Edmonds and Nina Pavcnik 2005). The majority
of child labor force workers are engaged in agricultural production, which may or may not
involve long hours or hours likely to conflict with school, depending on the type of production
involved weeding versus harvesting, for example. Neither weeding nor fetching water has
anything inherently wrong with it; most parents will assert that doing small jobs is good for
44
children. The problem, in our view, is when children’s jobs, be they officially labor force work
or not, interfere substantially with human capital production that will benefit both the children
themselves and their society. Our findings indicate the importance of considering non-labor-
force work alongside labor force work in studies of child labor, keeping in mind the differences
that may exist between them in terms of exposing the child to abuse and health hazards.
Our conclusions on the effect of work on schooling rest on how well we are able to
identify the structural schooling equation through appropriate exclusion restrictions. Because
girls are involved mostly in domestic work, we rely upon instruments that proxy for the demand
for domestic work by indicating household access to basic services, namely access to piped
water, piped sewarage, and garbage collection. These infrastructure variables are therefore
assumed to affect a girl’s domestic work burden, but not to have a direct effect on her school
attendance. The fact that poor infrastructure is often associated with poor access to schools at the
community level is addressed by community and regional level variables that explicitly control
for access to schooling. While these instruments are not ideal, they are effective and withstand
the standard empirical tests. We view this analysis as one contribution to a body of evidence in
which different approaches are taken, with the goal of eventually shedding light on these larger
social science questions (Robert Moffitt 2005).
Why is it interesting to know whether child work leads to a reduction in school
attendance? It matters because it is important to know whether interventions should target a
reduction in the domestic work burden of girls or focus exclusively on improving the
attractiveness of schooling. Our approach does not investigate the pathway that goes from
schooling to work, so we cannot address the relative merits of these two approaches, although we
suspect that both are important. Lloyd, et al. (2003) provide convincing evidence that improving
45
school quality would result in better school retention and grade attainment. Our results provide
complementary evidence that interventions to reduce girls’ work would also increase school
attendance. The question is: how should such a reduction be achieved?
An approach that bans labor force work, and attempts to enforce such a ban, will have
practically no effect for girls in Egypt. Since girls doing subsistence work and market work tend
to work on family farms or in family enterprises, a legal approach is unlikely to affect them in
general, children are legally allowed to work in family businesses, and even if not, enforceability
would be difficult in such circumstances. Families resort to hiring their own children when they
face high transaction costs in hiring outside labor (Yoram Ben-Porath 1980). Policy changes that
reduce such transaction costs may reduce the use of children’s labor in family enterprises. For
example, the recent changes in Egypt’s labor law that reduced barriers to hiring and firing
workers may have implications with respect to children’s family labor by making it easier to hire
outside workers and therefore reducing the reliance of small enterprise on family labor.
An approach targeting labor market work, even if successful in reducing unpaid family
work for children, is unlikely to do much for girls’ school attendance given that only a tiny
fraction of girls participate in such work. Approaches that reduce the drudgery of household
work or that compensate families for the opportunity cost of a girl’s time at home would be
much more effective in increasing their school attendance. Some programs currently being
implemented in Egypt pursue the compensation approach. For instance, the Girl’s Education
Initiative includes a conditional food aid component sponsored by the World Food Program
(WFP). Under this program, children (mostly girls) attending girl-friendly schools receive a daily
dry meal made up of commercially produced snacks as well as a take-home ration for their
families (Ronald G. Sultana 2007: 51). Another program along the same lines is the Girls’
46
Scholarship Program implemented by the Center for Development and Population Activities
with funding from USAID. This is essentially a conditional cash transfer (CCT) program
targeting girls from low-income families who were not attending school or were at risk of
dropping out in four Upper Egyptian governorates (Center for Development and Population
Activities [CEDPA] 2005).
While these programs clearly attempt to change parental calculus regarding whether or
not to send a girl to school, they do not directly contribute to reduce the domestic work burden.
Our research suggests that access to better infrastructural services, such as piped water,
sewerage, and garbage collection would go a long way toward reducing girls’ domestic work
burdens. Among these, lack of access to piped water imposes the greatest work burden on
women and girls by forcing them to use canals and other waterways to wash dishes and clothes
and to spend time and effort fetching drinking water from public taps. We find that poor urban
girls living in unserviced informal settlements might even be more vulnerable to not attending
school than poor girls from rural Upper Egypt, who have traditionally been flagged as the most
vulnerable (Sahar El-Tawila, Omaima El-Gibaly, Barbara Ibrahim, Fikrat El Sahn, Sunny
Sallam, Susan M. Lee, Barbara S. Mensch, Hind Wassef, Sarah Bukhari, and Osman Galal 1999;
Langsten and Hassan 2007). Families in these informal settlements have fewer options to fetch
drinking water, get rid of wastewater and garbage, and wash their dishes and clothing, leading to
a heavier burden on girls. Other interventions to reduce girls’ domestic work burdens include the
provision of adequate energy sources for cooking, reducing marketing time via refrigeration, and
providing solid waste and sewage disposal services.
Another implication of our work is the importance of parental presence in protecting girls
from non-attendance in school and involvement in domestic work. Girls with absent mothers
47
were particularly vulnerable to not being in school, although their visible involvement in work
was lower, presumably because they are engaged in more childcare as substitute care takers for
their siblings, an activity that is not adequately captured even by our expanded work measure.
Girls with both parents absent, that is, girls in fostering situations, are highly vulnerable to both
increased work and decreased school attendance because of their high work burdens. While there
is only a limited amount government policies can do to alter the pattern of parental presence,
these parenting situations can be used to target girls at risk.
While girls’ work reduces their schooling via both non-entry and dropout, it may be that
reducing non-entry would be an easier initial policy objective than reducing dropout, given the
fact that the majority of girls not currently in school in Egypt have actually never been to school.
Many efforts to reduce non-entry are currently underway in Egypt, and a massive school
building campaign of the Ministry of Education has made schools more physically accessible to
girls. These interventions have clearly had some positive effects as indicated by the decline in the
gender gap in ever enrollment in recent years. Other policy levers to reduce non-entry include
linking birth registrations to local school entry lists, subsidization of school-entry-related
expenses, and positive, persuasive outreach to parents and stepparents. Young girls, of school-
entry age, are relatively unproductive workers, and parents face a small opportunity cost of
schooling. Once girls are enrolled, it is then easier to keep track of them as part of the process of
reducing dropouts via monitoring school attendance and performance.
Ragui Assaad, Hubert H. Humphrey Institute of Public Affairs, University of Minnesota
301 19th Avenue South, Minneapolis, MN 55455, USA
e-mail: assaad@umn.edu
48
Deborah Levison, Hubert H. Humphrey Institute of Public Affairs, University of
Minnesota
301 19th Avenue South, Minneapolis, MN 55455, USA
e-mail: dlevison@umn.edu
Nadia Zibani, Poverty, Gender, and Youth Program, Population Council
59, Misr Helwan Agricultural Road, Cairo, Egypt
e-mail: nzibani@popcouncil.org
BIOGRAPHIES
Ragui Assaad is Professor at the Hubert H. Humphrey Institute of Public Affairs at the
University of Minnesota. On leave, he recently spent three years heading the Population
Council’s regional office in Cairo. He studies and advises on labor markets in Egypt, the Middle
East, and North Africa, with a focus on women and youth.
Deborah Levison is Professor at the Hubert H. Humphrey Institute of Public Affairs at the
University of Minnesota. She is an economist with a BA from Smith College and a PhD from the
University of Michigan. Her recent research focuses on children’s work and education in poor
countries, with an emphasis on gender, family, and labor market interactions.
49
Nadia Zibani is a technical advisor at the Poverty, Gender and Youth program at the Population
Council’s Middle East and North Africa regional office in Cairo, Egypt. She is currently
involved in projects related to out-of-school adolescents girls in Egypt and child labor in small
and micro enterprises).
.
ACKNOWLEDGMENTS
We thank Cristobal Ridao-Cano for allowing us to adapt his Stata program for a
switching probit model. We are grateful to Dennis Ahlburg, Ray Langsten, Brian McCall, David
Post, and Insan Tunali for detailed and valuable comments. Sara Kaufman, Pingkang Yu, Yi
Zheng, and Hongliang Zhang provided competent research assistance. We gratefully
acknowledge financial assistance and technical support from the International Center for
Economic Growth (ICEG) through its Economic Policy Initiative Consortium project in Egypt,
which is funded by USAID.
50
Table 1: Weighted Proportions of Girls (6-14) Working and Attending School and Average Hours Worked, Egypt 1998
Proportion
attending
school
Proportion
working
Proportion
in school
only
Proportion
both at
work and
in school
Proportion
at work
only
Proprotion
neither at
work nor in
school
Mean Std. Dev
Market Work (>=1 hour/week) 0.858 0.017 0.856 0.002 0.016 0.126 49.4 22.4
Inclusive Work
>=1 hour/week 0.858 0.426 0.549 0.309 0.117 0.025 21.3 16.2
>=7 hours/week 0.858 0.417 0.554 0.304 0.113 0.029 21.4 16.2
>=14 hours/week 0.858 0.319 0.646 0.212 0.107 0.035 25.7 16.1
>=21 hours/week 0.858 0.192 0.763 0.095 0.097 0.045 33.4 16.9
>=28 hours week 0.858 0.116 0.820 0.038 0.078 0.064 41.4 17.6
Hours worked/week
51
All
School
only
Work and
School**
Work
Only**
Neither
10.379 9.807 11.818 11.650 9.014
(2.471) (2.415) (1.883) (2.037) (2.805)
1.138 1.020 1.432 1.399 0.890
(0.503) (0.480) (0.420) (0.445) (0.535)
0.087 0.093 0.071 0.087 0.071
(0.282) (0.291) (0.258) (0.283) (0.259)
36.149 36.239 35.690 36.034 37.900
(14.636) (14.377) (14.965) (15.767) (14.780)
32.470 32.856 32.278 30.998 31.017
(8.667) (7.974) (8.157) (11.326) (13.018)
7.249 8.241 6.696 1.903 4.043
(5.862) (5.884) (5.444) (2.792) (4.880)
5.434 6.394 4.822 0.641 1.814
(5.715) (5.870) (5.211) (1.707) (3.987)
0.015 0.007 0.015 0.063 0.043
(0.121) (0.085) (0.122) (0.244) (0.204)
0.052 0.051 0.060 0.044 0.014
(0.221) (0.221) (0.238) (0.205) (0.120)
0.045 0.049 0.039 0.024 0.043
(0.207) (0.216) (0.195) (0.154) (0.204)
mother absent * 0.030 0.020 0.026 0.083 0.143
(0.170) (0.139) (0.160) (0.276) (0.352)
0.139 0.111 0.133 0.359 0.171
(0.346) (0.315) (0.340) (0.480) (0.377)
0.422 0.465 0.417 0.150 0.271
(0.494) (0.499) (0.493) (0.358) (0.448)
0.177 0.173 0.182 0.189 0.200
(0.382) (0.379) (0.386) (0.393) (0.403)
0.226 0.214 0.235 0.277 0.314
(0.419) (0.410) (0.424) (0.448) (0.468)
0.035 0.037 0.034 0.024 0.043
(0.184) (0.188) (0.181) (0.154) (0.204)
0.131 0.113 0.137 0.233 0.214
(0.338) (0.316) (0.344) (0.424) (0.413)
0.226 0.244 0.225 0.112 0.143
(0.418) (0.429) (0.418) (0.316) (0.352)
0.157 0.142 0.160 0.228 0.286
(0.364) (0.349) (0.366) (0.420) (0.452)
0.127 0.130 0.139 0.083 0.086
(0.333) (0.336) (0.346) (0.276) (0.282)
0.333 0.400 0.272 0.039 0.100
(0.471) (0.490) (0.445) (0.194) (0.302)
0.105 0.065 0.079 0.315 0.345
(0.307) (0.247) (0.271) (0.466) (0.479)
0.077 0.064 0.073 0.180 0.129
(0.267) (0.244) (0.261) (0.385) (0.337)
0.246 0.231 0.306 0.214 0.214
(0.431) (0.421) (0.461) (0.411) (0.413)
Table 2: Means and Standard Deviations of Explanatory Variables, Girls Ages 6-14
HH in 2nd lowerst rural quintile*
HH in top three rural quintiles*
HH in the lowest urban quintile (reference) *
HH in the lowest rural quintile (reference)*
HH has nonfarm enterprise*
father absent permanently*
father absent temporarily*
HH in 2nd lowest urban quintile*
age squared/100
father regular private sector wage worker when girl was age 6*
HH in top three urban quintiles*
father's years of schooling
both parents absent*
father irregular private wage worker (reference)*
not daughter of household head*
father nonwage worker when girl was age 6*
father not working or work data missing when girl was age 6*
HH has farm enterprise*
father public sector worker when girl was age 6*
mother's years of schooling
father's age when girl was age 6
mother's age when girl was age 6
Child Characteristics
Parental and HH Characteristics
Wealth
age
52
All
School
only
Work and
School**
Work
Only**
Neither
0.169 0.198 0.124 0.087 0.071
(0.375) (0.399) (0.329) (0.282) (0.258)
0.120 0.141 0.079 0.063 0.100
(0.325) (0.349) (0.270) (0.244) (0.302)
0.146 0.141 0.188 0.073 0.157
(0.353) (0.349) (0.391) (0.260) (0.367)
0.181 0.190 0.180 0.126 0.143
(0.385) (0.393) (0.385) (0.333) (0.352)
0.199 0.178 0.257 0.223 0.186
(0.399) (0.382) (0.437) (0.417) (0.392)
0.184 0.151 0.173 0.427 0.343
(0.388) (0.358) (0.378) (0.496) (0.478)
20.98 22.47 20.32 13.74 12.70
(12.12) (12.09) (10.01) (13.72) (11.13)
30.08 31.62 28.77 23.78 22.76
(11.55) (11.61) (9.36) (13.06) (10.43)
89.90 91.30 91.40 78.60 78.90
(11.50) (9.90) (8.80) (17.50) (17.30)
94.70 95.20 95.10 90.30 90.90
(5.90) (5.00) (4.90) (9.40) (10.70)
78.10 80.60 79.40 60.60 60.80
(17.60) (15.60) (15.80) (21.20) (22.50)
85.10 86.20 84.90 79.40 79.00
(9.40) (8.90) (8.50) (11.90) (12.40)
Instruments
9.910 8.626 10.240 17.045 16.354
(10.294) (9.750) (9.701) (11.632) (11.758)
4.233 4.306 4.181 3.921 3.849
(1.525) (1.527) (1.454) (1.582) (1.666)
HH without piped water* 0.090 0.050 0.099 0.340 0.229
(0.287) (0.218) (0.300) (0.475) (0.423)
HH without piped sewage disposal* 0.438 0.381 0.478 0.723 0.629
(0.496) (0.486) (0.500) (0.448) (0.487)
HH without garbage collection* 0.557 0.499 0.612 0.811 0.757
(0.497) (0.500) (0.488) (0.393) (0.432)
0.327 0.310 0.304 0.379 0.757
(0.561) (0.549) (0.557) (0.525) (0.788)
0.520 0.527 0.445 0.641 0.571
(0.677) (0.679) (0.622) (0.757) (0.714)
0.552 0.522 0.525 0.786 0.757
(0.706) (0.694) (0.662) (0.840) (0.731)
0.457 0.449 0.394 0.621 0.629
(0.636) (0.637) (0.574) (0.700) (0.765)
0.458 0.440 0.482 0.597 0.300
(0.641) (0.631) (0.665) (0.631) (0.645)
0.274 0.255 0.296 0.325 0.371
(0.528) (0.506) (0.561) (0.606) (0.516)
0.257 0.237 0.285 0.286 0.414
(0.489) (0.473) (0.514) (0.494) (0.602)
0.429 0.419 0.422 0.519 0.457
(0.809) (0.815) (0.802) (0.795) (0.755)
0.537 0.481 0.615 0.743 0.657
(0.886) (0.853) (0.932) (0.951) (0.946)
0.132 0.135 0.131 0.121 0.100
(0.346) (0.349) (0.344) (0.342) (0.302)
0.064 0.066 0.062 0.058 0.057
(0.245) (0.249) (0.241) (0.235) (0.234)
2442 1633 533 206 70
Standard deviations in parentheses
* Denotes a dummy variable
**
Table 2: Means and Standard Deviations of Explanatory Variables, Girls Ages 6-14 (Contn'd)
Work is defined using the 14 hour/week threshold
# of females ages 18-59 in HH
# of females ages 15-17 in HH
# of males ages 15-17 in HH
# of males age 60 and over in HH
% of girls ages 6-11 enrolled in school in locality
% boy ages 6-11 enrolled in school in locality
% girls ages 12-14 enrolled in school in locality
% boys ages 12-14 enrolled in school in locality
# of females age 60 and over in HH
# of males ages 18-59 in HH
Region of Residence & Community Characteristics
% agricultural workers in locality
Urban Upper Egypt*
% service & trade workers in locality
# of children ages 3-5 in HH
# of children ages 6-9 in HH
# of girls ages 10-14 in HH
# of boys ages 10-14 in HH
Alexandria and Suez Canal*
Urban Lower Egypt*
% male w/ secondary school or above in locality
Rural Lower Egypt*
# of children ages 0-2 in HH
% female w/ secondary school or above in locality
Greater Cairo (reference) *
Rural Upper Egypt*
Number of Observations
Household Composition
53
Variables work school work school work school
working -2.378*** -2.367*** -2.473***
(0.164) (0.156) (0.144)
Child Characteristics
age 0.460*** 0.561*** 0.454*** 0.569*** 0.470*** 0.616***
(0.121) (0.135) (0.122) (0.135) (0.121) (0.137)
age squared/100 -0.980+ -2.125** -0.951 -2.160*** -1.000+ -2.389***
(0.579) (0.654) (0.582) (0.651) (0.575) (0.667)
not daughter of household head -0.236+ 0.104 -0.246+ 0.108 -0.347* 0.02
(0.124) (0.128) (0.128) (0.132) (0.173) (0.180)
Parental and HH Characteristics
father's age when girl was age 6 0.004 0.006 0.005 0.006 0.006 0.005
(0.006) (0.007) (0.006) (0.008) (0.006) (0.007)
mother's age when girl was age 6 -0.013+ -0.017+ -0.013+ -0.017+ -0.012 -0.023*
(0.008) (0.009) (0.008) (0.010) (0.008) (0.010)
father's years of schooling -0.025** 0.031* -0.023* 0.030* -0.025* 0.033**
(0.009) (0.013) (0.010) (0.013) (0.010) (0.012)
mother's years of schooling -0.005 0.050** -0.004 0.050** -0.005 0.052**
(0.009) (0.017) (0.009) (0.017) (0.009) (0.016)
both parents absent 1.171** 0.292 1.159** 0.299 1.244** 0.006
(0.411) (0.496) (0.414) (0.485) (0.426) (0.480)
father absent permanently 0.173 0.702+ 0.178 0.690+ 0.236 0.599
(0.293) (0.402) (0.293) (0.410) (0.307) (0.405)
father absent temporarily 0.11 0.238 0.08 0.262 0.141 0.205
(0.276) (0.358) (0.281) (0.363) (0.285) (0.353)
mother absent (w/ or w/o stepmother present) -0.635+ -1.246** -0.656+ -1.235** -0.553 -1.363**
(0.353) (0.472) (0.358) (0.469) (0.359) (0.467)
father public sector worker when girl was 6 -0.146 0.067 -0.158 0.044
(0.110) (0.148) (0.110) (0.149)
father regular private sector wage worker when -0.056 -0.051 -0.063 -0.058
girl was 6 (0.108) (0.133) (0.108) (0.136)
father nonwage workers when girl was 6 -0.092 -0.048 -0.096 -0.001
(0.109) (0.141) (0.110) (0.145)
father not working or work data missing when -0.095 0.053 -0.104 0.098
girl was 6 (0.156) (0.222) (0.157) (0.229)
farm enterprise 0.039 0.062 0.049 0.056
(0.116) (0.138) (0.117) (0.139)
non-farm enterprise -0.149 0.169 -0.153 0.098
(0.091) (0.133) (0.093) (0.128)
Wealth
HH in 2nd lowest urban wealth quintile 0.078 0.218 0.087 0.225 0.066 0.204
(0.131) (0.147) (0.132) (0.152) (0.131) (0.141)
HH in top three urban wealth quintiles -0.239* 0.378* -0.219+ 0.364* -0.222+ 0.24
(0.118) (0.180) (0.121) (0.180) (0.121) (0.154)
HH in 2nd lowest rural wealth quintile -0.442** 0.07 -0.417** 0.059 -0.420** 0.056
(0.141) (0.184) (0.143) (0.182) (0.143) (0.172)
HH in top three rural wealth quintiles -0.349* 0.17 -0.304* 0.144 -0.306* 0.158
(0.143) (0.147) (0.144) (0.157) (0.144) (0.149)
Region of Residence and Community Characteristics
Alexandria and Suez Canal -0.302+ -0.268 -0.299+ -0.253 -0.307* -0.304
(0.156) (0.255) (0.154) (0.255) (0.156) (0.245)
Urban Lower Egypt 0.16 0.229 0.189 0.243 0.186 0.263
(0.169) (0.233) (0.170) (0.227) (0.168) (0.213)
Urban Upper Egypt 0.238+ 0.651* 0.248+ 0.671* 0.240+ 0.683**
(0.138) (0.258) (0.138) (0.265) (0.137) (0.245)
Rural Lower Egypt 0.541* 0.570* 0.514* 0.610* 0.495* 0.549*
(0.246) (0.275) (0.248) (0.286) (0.244) (0.267)
Rural Upper Egypt 0.788*** 0.681* 0.756*** 0.710* 0.747*** 0.681*
(0.221) (0.292) (0.220) (0.306) (0.220) (0.272)
Table 3: Parameter Estimates from Bivariate Probit Model, Girls 6-14.
Model 3
Model 2
Model 1
54
Variables work school work school work school
0.058*** 0.012 0.057** 0.013 0.057*** 0.011
(0.017) (0.018) (0.017) (0.018) (0.017) (0.017)
% male w/ secondary school or above -0.055*** -0.024 -0.054** -0.025 -0.054*** -0.025
(0.016) (0.017) (0.016) (0.017) (0.016) (0.016)
-0.003 0.023* -0.002 0.022* -0.001 0.026**
(0.010) (0.010) (0.010) (0.010) (0.010) (0.009)
0.028* -0.016 0.027* -0.017 0.026* -0.018
(0.013) (0.013) (0.013) (0.013) (0.013) (0.014)
-0.005 0.020** -0.005 0.020** -0.005 0.018**
(0.006) (0.007) (0.006) (0.007) (0.006) (0.006)
-0.004 -0.011 -0.005 -0.009 -0.006 -0.013
(0.008) (0.009) (0.008) (0.009) (0.008) (0.010)
Instruments
% agricultural workers in locality (X) -0.005 -0.005 -0.004
(0.007) (0.007) (0.007)
% service & trade workers in locality (X) 0.001 0.007 0.005
(0.027) (0.028) (0.028)
households without piped water (X) 0.528*** 0.514*** 0.475***
(0.112) (0.111) (0.113)
households without piped sewage disposal (X) 0.154 0.146 0.152
(0.100) (0.102) (0.102)
households without garbage collection (X) 0.128 0.131 0.121
(0.087) (0.088) (0.084)
Household Composition
# of children age 0-2 in HH 0.058 -0.204*
(0.067) (0.084)
# of children age 3-5 in HH 0.021 -0.012
(0.046) (0.057)
# of children age 6-9 in HH -0.048 -0.193***
(0.046) (0.057)
# of girls age 10-14 in HH 0.022 -0.130+
(0.060) (0.071)
# of boys age 10-14 in HH 0.097+ 0.055
(0.051) (0.069)
# of females age 15-17 in HH 0.014 -0.08
(0.067) (0.064)
# of males age 15-17 in HH -0.062 -0.088
(0.064) (0.088)
# of females age 18-59 in HH -0.055 0.111+
(0.043) (0.062)
# of males age 18-59 in HH 0.014 0.008
(0.047) (0.057)
# of females age 60 and over in HH -0.092 0.220+
(0.103) (0.122)
# of males age 60 and over in HH 0.258 -0.002
(0.176) (0.194)
Constant -5.252*** -2.504* -5.064*** -2.607** -5.288*** -1.980*
(0.976) (0.973) (0.988) (0.949) (0.984) (0.951)
Correlation of errors (rho)
Chi2-test (1) for rho=0
p-value for likelihood ratio test of rho=0
Log Likelihood
Log likelihood without regressors
Number of observations
Standard errors in parentheses. (X) indicates that the variable
Significant coefficients at 0.1% (***) ,1% (**), 5% (*) and
- A cluster correction was used to adjust standard errors for the fact that some observations on community-level variables are not
independent across girls in the same PSU.
40.583
0.000
-1643.94
31.515
2442
-1676.10
0.000
-2226.39
2442
-2226.39
Table 3: Parameter Estimates from Bivariate Probit Model, Girls 6-14 (Contn'd)
Model 3
Model 2
Model 1
0.838
% of girls ages 6-11 enrolled in school in locality
% of boys ages 6-11 enrolled in school in locality
0.799
% of girls ages 12-14 enrolled in school in locality
% of boys ages 12-14 enrolled in school inlocality
35.708
0.000
-1669.35
-2226.39
2442
% female w/ secondary school or above
0.805
55
In School (S=1) or not in school (S=0)
Working (W=1) or not working (W=0)
W=1 S=1
S=1 &
W=0
S=1 &
W=1
S=0 &
W=1
S=0 &
W=0
Probability for reference individual 0.461 0.833 0.376 0.457 0.004 0.163
Change in probability due to change in :
probability of work by 10 percentage points -0.060 *** -0.046 *** -0.015 ** 0.015 ** 0.046 ***
(0.08) (0.061) (0.053) (0.053) (0.061)
Child Characteristics
age 0.183 *** 0.141 *** -0.045 0.185 *** -0.003 -0.138 ***
(0.047) (0.044) (0.055) (0.046) (0.004) (0.042)
age squared/100 -0.389 + -0.532 ** -0.121 -0.411 0.022 0.510 **
(0.228) (0.200) (0.246) (0.221) (0.029) (0.184)
not daughter of household head -0.0919 + 0.025 0.113 * -0.089 + -0.003 -0.021
(0.047) (0.030) (0.048) (0.047) (0.004) (0.029)
Parental and HH Characteristics
father's age when age 6 0.002 0.002 0.000 0.002 0.000 -0.002
(0.002) (0.002) (0.002) (0.002) (0.000) (0.002)
mother's age when age 6 -0.005 + -0.004 0.001 -0.005 + 0.000 0.004
(0.004) (0.003) (0.004) (0.003) (0.000) (0.003)
father's years of schooling -0.010 ** 0.008 * 0.017 *** -0.009 * -0.001 -0.007 *
(0.004) (0.003) (0.004) (0.004) (0.001) (0.003)
mother's years of schooling -0.002 0.013 ** 0.014 * -0.001 -0.001 -0.012 **
(0.003) (0.004) (0.005) (0.004) (0.001) (0.004)
both parents absent 0.397 *** 0.063 -0.305 *** 0.368 *** 0.029 -0.092
(0.098) (0.094) (0.077) (0.103) (0.045) (0.069)
father absent permanently 0.069 0.119 * 0.046 0.073 -0.004 -0.115 *
(0.116) (0.059) (0.12) (0.116) (0.005) (0.057)
father absent temporarily 0.0438 0.053 0.007 0.046 -0.002 -0.051
(0.110) (0.072) (0.111) (0.109) (0.005) (0.069)
mother absent -0.229 * -0.443 * -0.188 -0.256 ** 0.026 0.417 *
(0.108) (0.194) (0.152) (0.094) (0.047) (0.169)
Wealth
HH in 2nd lowest urban wealth quintile 0.031 0.049 0.016 0.033 -0.002 -0.047
(0.052) (0.034) (0.052) (0.052) (0.003) (0.033)
HH in top three urban wealth quintiles -0.093 * 0.078 0.166 *** -0.089 + -0.004 -0.074 +
(0.047) (0.045) (0.046) (0.046) (0.005) (0.042)
HH in 2nd lowest rural wealth quintile -0.167 *** 0.017 0.180 *** -0.163
***\
-0.004 -0.013
(0.049) (0.043) (0.051) (0.050) (0.005) (0.043)
HH in top three rural wealth quintiles -0.134 ** 0.039 0.169 *** -0.13 * -0.004 -0.035
(0.052) (0.033) (0.047) (0.053) (0.005) (0.031)
Region of Residence and Community Characteristics
Alexandria and Suez Canal -0.116 * -0.075 0.041 -0.116 * 0.000 0.076
(0.059) (0.073) (0.072) (0.059) (0.004) (0.070)
Urban Lower Egypt 0.064 0.051 -0.014 0.065 -0.001 -0.050
(0.067) (0.049) (0.066) (0.067) (0.004) (0.05)
Urban Upper Egypt 0.09 5 + 0.114 * 0.016 0.099 + -0.004 -0.110 *
(0.055) (0.055) (0.069) (0.054) (0.005) (0.052)
Rural Lower Egypt 0.210 * 0.105 + -0.107 0.212 * -0.002 -0.103 +
(0.092) (0.057) (0.089) (0.090) (0.004) (0.054)
Rural Upper Egypt 0.29 4 *** 0.117 * -0.178 * 0.295 *** -0.001 -0.116 *
(0.078) (0.059) (0.078) (0.077) (0.005) (0.055)
0.023 *** 0.003 -0.019 *** 0.022 ** 0.001 -0.004
(0.007) (0.005) (0.006) (0.007) (0.001) (0.005)
% male w/ secondary school or above -0.022 *** -0.006 0.015 ** -0.021 *** 0.000 0.007
(0.006) (0.003) (0.005) (0.006) (0.001) (0.004)
-0.001 0.006 * 0.007 -0.001 0.000 -0.005 *
(0.004) (0.003) (0.004) (0.004) (0.001) (0.003)
0.011 * -0.004 -0.014 *** 0.010 * 0.001 0.003
(0.005) (0.003) (0.004) (0.005) (0.001) (0.003)
-0.002 0.005 * 0.007 ** -0.002 0.000 -0.005 *
(0.002) (0.002) (0.002) (0.002) (0.000) (0.002)
-0.002 -0.003 -0.001 -0.002 0.000 0.003
(0.003) (0.002) (0.003) (0.003) (0.000) (0.002)
Instruments
% agricultural workers in locality -0.002 0.002 -0.002 0.000 0.000
(0.003) . (0.003) (0.003) (0.000) (0.000)
% service & trade workers in locality 0.001 0.0 00 0.000 0.000 0.000
(0.011) . (0.010) (0.010) (0.000) (0.000)
household without piped water 0.205 *** -0.188 *** 0.188 *** 0.017 -0.017
(0.041) . (0.038) (0.038) (0.014) (0.014)
household without piped sewage disposal 0.061 -0.058 0.058 0.003 -0.003
(0.040) . (0.038) (0.038) (0.004) (0.004)
household without garbage collection 0.051 -0.048 0.048 0.002 -0.002
(0.035) . (0.032) (0.033) (0.003) (0.003)
Table 4: Marginal Effects on the Joint Probabilities of Work and School Based on Model (1), Reference Girl
% female w/ secondary school or above
% female children ages 6-11enrolled in school in locality
% male children ages 6-11 enrolled in school in locality
% of female children ages 12-14 enrolled in school in
locality
% male children ages 12-14 enrolled in school in locality
*Based on marginal change for continuous variables and change from 0 to 1 for dummy variables
**A cluster correction was used to adjust standard errors for the fact that some observations on community level
variables are not independent across girls in the same PSU.
^The reference individual is a 14 year old girl whose father and mother are present and have mean years of schooling. She lives in Greater Cairo and
belongs to a household in the lowest urban wealth quintile that has no household enterprise. She lives in a neighborhood with the mean percentage of
agricultural, service and trade workers, the mean percentage of males/females with secondary education and above and the mean school enrollment rates.
*** significant at 0.1%,** significant at 1%; *significant at 5%, + significant at 10%
56
Ever in
school or
not
7
hours
Base Run ≥
14 hours
≥ 21
hours
≥ 28
hours
≥ 14
hours
Reference girl*
Model 1 Pr[S=1 | W=0] 0.985 0.980 0.962 0.932 0.998
Pr[S=1 | W=1] 0.426 0.376 0.180 0.097 0.651
Abs Change due to work -0.559 -0.604 -0.782 -0.835 -0.347
Rel Chg due to work -57% -62% -81% -90% -35%
Model 2 Pr[S=1 | W=0] 0.986 0.977 0.961 0.921 0.997
Pr[S=1 | W=1] 0.447 0.358 0.184 0.095 0.629
Abs Change due to work -0.539 -0.619 -0.777 -0.826 -0.368
Rel Chg due to work -55% -63% -81% -90% -37%
Model 3 Pr[S=1 | W=0] 0.995 0.990 0.976 0.944 0.999
Pr[S=1 | W=1] 0.544 0.436 0.224 0.103 0.688
Abs Change due to work -0.451 -0.553 -0.752 -0.840 -0.311
Rel Chg due to work -45% -56% -77% -89% -31%
Most vulnerable girl**
Model 1 Pr[S=1 | W=0] 0.783 0.624 0.539 0.383 0.718
Pr[S=1 | W=1] 0.059 0.020 0.005 0.001 0.030
Abs Change due to work -0.724 -0.604 -0.534 -0.382 -0.688
Rel Chg due to work -92% -97% -99% -100% -96%
Model 2 Pr[S=1 | W=0] 0.790 0.610 0.523 0.355 0.685
Pr[S=1 | W=1] 0.064 0.018 0.005 0.001 0.025
Abs Change due to work -0.726 -0.592 -0.518 -0.354 -0.660
Rel Chg due to work -92% -97% -99% -100% -96%
Model 3 Pr[S=1 | W=0] 0.719 0.461 0.298 0.157 0.770
Pr[S=1 | W=1] 0.032 0.005 0.001 0.000 0.018
Abs Change due to work -0.687 -0.456 -0.297 -0.157 -0.651
Rel Chg due to work -96% -99% -100% -100% -85%
Least vulnerable girl***
Model 1 Pr[S=1 | W=0] 0.998 0.997 0.994 0.992 1.000
Pr[S=1 | W=1] 0.702 0.652 0.422 0.344 0.839
Abs Change due to work -0.296 -0.346 -0.572 -0.647 -0.160
Rel Chg due to work -30% -35% -58% -65% -16%
Model 2 Pr[S=1 | W=0] 0.998 0.997 0.993 0.991 1.000
Pr[S=1 | W=1] 0.704 0.651 0.421 0.357 0.853
Abs Change due to work -0.294 -0.346 -0.572 -0.634 -0.147
Rel Chg due to work -29% -35% -58% -64% -15%
Model 3 Pr[S=1 | W=0] 0.999 0.998 0.995 0.992 1.000
Pr[S=1 | W=1] 0.736 0.679 0.442 0.334 0.863
Abs Change due to work -0.263 -0.319 -0.554 -0.658 -0.137
Rel Chg due to work -26% -32% -56% -66% -14%
*** The least vulnerable girl is similar to the reference girl except for the following: she lives in a household in the top 3 urban
wealth quintiles, her father has 12 years of schooling and her mother has 9 years of schooling, and her father is a public sector
worker in Models 2 and 3.
Sensitivity to work hours cut-off
Table 5: Simulation of Effect of Inclusive Work on Schooling, with Sensitivity Analysis Regarding Hours
Cut-Off and Specification of Schooling Variable
Specification of Schooling Variable
Notes:
Currently in school or not
* The reference girl is 14 years old, has both parents living at home and has parents with the mean years of schooling. She lives
in Greater Cairo and belongs to a household in the lowest urban wealth quintile. She lives in a neighborhood with the mean
percentage of service, trade and agricultural workers, the mean percentage of males/females with secondary education and above,
and the mean enrollment rates for children 6-11 and 12-14. In addition, in Model 2, her father is an irregular private sector
worker and her household has no farm or non-farm enterprises. In model 3, she is in a household that has no other members
besides her parents.
** The most vulnerable girl is similar to the reference girl but has a father with no schooling, and her mother is absent. In Model
3 she lives in a household with additional children: one child under 2, another between 6 and 9, and a sister age 10-14.
57
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
6 7 8 9 10 11 12 13 14 15 16 17 18
Percentage
Age
Figure 1
Percentage of Girls and Boys Attending School and Engaged in
Work According to Various Definitions by Age, 1998
girls'market
work
boys'market
work
girls' incl. work
>=28h/w
girls' incl. work
>=21h/w
girls' incl. work
>=14h/w
girls' incl. work
>=7h/w
girls' incl. work
>=1h/w
girls' schooling
boys'schooling
58
Figure 2
0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1
Pred. Probability of Work
all 3 srvcs pres.
no sewerage
no garbage
no sew‘rage & garbage
no water
no water & garbage
no water & sewerage
all 3 srvcs abs.
Availability of Services
>=7h/w >=14h/w
>=21h/w >=28h/w
Predicted Probability of Work by Eight Bundles of Services
for Different Hours Thresholds
59
Figure 3
.3
.35
.4
.45
.5
.55
.6
.65
.7
.75
.8
.85
.9
.95
1
Pred. Probability of School
.1 .15 .2 .25 .3 .35 .4 .45 .5 .55 .6 .65 .7 .75 .8 .85 .9 .95 1
Pred. Probability of Work
>=7h/w >=14h/w
>=21h/w >=28h/w
Predicted Probability of School vs. Predicted Probability of Work
for Different Hours Thresholds at Different Bundles of Services
60
Figure 4
.3
.35
.4
.45
.5
.55
.6
.65
.7
.75
.8
.85
.9
.95
1
Pred. Probability of School
.2 .25 .3 .35 .4 .45 .5 .55 .6 .65 .7 .75 .8 .85 .9 .95 1
Pred. Probability of Work
both parents present
mother absent
both parents absent
Predicted Probability of School vs. Predicted Probability of Work (>=14h/w)
by Parental Absence Status at Different Bundles of Serivces
61
Figure 5
.3
.35
.4
.45
.5
.55
.6
.65
.7
.75
.8
.85
.9
.95
1
Pred. Probability of School
.3 .35 .4 .45 .5 .55 .6 .65 .7 .75 .8 .85 .9 .95 1
Pred. Prob. of Work
Poorest quint.,Gr.Cairo Poorest quint.,Rur.Upp.Eg.
2nd poorest quint.,Gr.Cairo 2nd poorest quint.,Rur.Upp.Eg.
Top 3 quint.,Gr.Cairo Top 3 quint.,Rur.Upp.Eg.
Predicted Probability of School vs. Predicted Probability of Work (>=14h/w)
by Wealth Quintile and Region at Different Levels of Services
62
REFERENCES
Aakvik, Arild, James Heckman, and Edward Vytlacil. 2005. ―Estimating Treatment Effects for
Discrete Outcomes When Responses to Treatment Vary: An Application to Norwegian
Vocational Rehabilitation Programs.‖ Journal of Econometrics 125(12): 1551.
Abu Gazaleh, Kawther, Lamia Bulbul, Soheir Hewala, and Suadad Najim. 2004. Gender,
Education and Child Labour in Egypt. Geneva, Switzerland: International Labour
Organization (ILO).
Assaad, Ragui and Melanie Arntz. 2005. ―Constrained Geographical Mobility and Gendered
Labor Market Outcomes under Structural Adjustment: Evidence from Egypt.‖ World
Development 33(3): 43154.
Assaad, Ragui, Deborah Levison and Nadia Zibani. 2007. ―The Effect of Child Work on
Schooling: Evidence from Egypt.‖ Minnesota Population Center Working Paper 2007-4,
University of Minnesota. http://www.pop.umn.edu/research/mpc-working-papers-
series/2007-working-papers-1 (accessed August 2009)
Beegle, Kathleen, Rakeev Dehejia, and Roberta Gatti. 2005. ―Why Should We Care About Child
Labor? The Education, Labor Market, and Health Consequences of Child Labor
Working Paper 3479, National Bureau of Economic Research, Cambridge MA.
Benería, Lourdes. 1992. ―The Mexican Debt Crisis: Restructuring the Economy and the
Household,‖ Chapter 4 in Unequal Burden: Economic Crisis, Persistent Poverty, and
Women’s Work, ed. Lourdes Benería and Shelly Feldman, Westview: Boulder CO.
63
Ben-Porath, Yoram. 1980. ―The F-Connection: Families, Friends, and Firms and the
Organization of Exchange.‖ Population and Development Review 6(1): 130.
Bhalotra, Sonia and Christopher Heady. 2003. ―Child Farm Labor: The Wealth Paradox. World
Bank Economic Review 17(2): 197222.
Binder, Melissa and David Scrogin. 1999. ―Labor Force Participation and Household Work of
Urban Schoolchildren in Mexico: Characteristics and Consequences.‖ Economic
Development and Cultural Change 48(1): 12354.
Boozer, Michael A. and Tavneet K. Suri. 2001. ―Child Labor and Schooling Decisions in
Ghana.‖ Labor and Population Workshop Paper, Yale University, New Haven,
Connecticut.
Boyden, Jo, Birgitta Ling, and William Myers. 1998. What Works for Working Children.
Stockholm: Radda Barnen and UNICEF.
Canagarajah, Sudharshan and Harold Coulombe. 1998. ―Child Labor and Schooling in Ghana.‖
Policy Research Working Paper 1844. The World Bank, Washington, D.C.
Canals-Cerda, Jose and Cristobal Ridao-Cano. 2004. The Dynamics of School and Work in
Rural Bangladesh.‖ World Bank Policy Research Working Paper No. 3330. The World
Bank, Washington D.C.[ http://www-
wds.worldbank.org/servlet/WDSContentServer/WDSP/IB/2004/07/06/000090341_20040
706151558/Rendered/PDF/WPS3330.pdf] (Accessed September 2009).
CAPMAS 1988. The Ocotber 1988 Round of the Labour Force Sample Survey (LFSS). Cairo:
Central Agency for Public Mobilization and Statistics.
64
Carrasco, Raquel. 2001. ―Binary Choice with Binary Endogenous Regressors in Panel Data:
Estimating the Effect of Fertility on Female Labor Participation.‖ Journal of Business
and Economic Statistics 19(4): 38594.
Center for Development and Population Activities (CEDPA). 2005. ―Girls’ Access to
Education.‖ New Horizons and New Visions: Mobilizing Communities for Girls’
Education in Egypt. , Washington, D.C.: CEDPA.
www.cedpa.org/files/715_file_FACT_SHEET_Overview_Improving_Girls_Lives_in_Eg
ypt.pdf (Accessed July 2007).
DeGraff, Deborah S. and Richard Bilsborrow. 2003. ―Children’s School Enrollment and Time at
Work in the Philippines.Journal of Developing Areas 37(1): 12758.
Economic Research Forum (ERF) 1998. Egypt Labor Market Survey of 1998. Data available
from http://www.erf.org.eg.
Edmonds, Eric and Nina Pavcnik. 2005. ―Child Labor in the Global Economy.Journal of
Economic Perspectives 19(1): 199220.
El-Tawila, Sahar, Omaima El-Gibaly, Barbara Ibrahim, Fikrat El-Sahn, Sunny Sallam, Susan M.
Lee, Barbara S. Mensch, Hind Wassef, Sarah Bukhari, and Osman Galal. 1999.
Transitions to Adulthood: A National Survey of Egyptian Adolescents. Cairo: Population
Council.
Elbadawy, Asmaa, Ragui Assaad, Dennis Ahlburg and Deborah Levison. 2004. ―Private and
Group Tutoring in Egypt: Where is the Gender Inequality?‖ Economic Research Forum
Working Paper No. 200429, Economic Research Forum, Cairo, Egypt [www.erf.org.eg]
65
Fafchamps, Marcel and Jackline Wahba. 2006. ―Child Labor, Urban Proximity and Household
Composition.‖ IZA Discussion Paper 1966, Institute for the Study of Labor (IZA) and
University of Bonn.
Filmer, Deon and Lant Pritchett. 2001. ―Estimating Wealth Effects without Income or
Expenditure Data or Tears: Education Enrollment in India.Demography 38 1: 11532.
Freije, Samuel and Luis F. Lopez-Calva. 2001. Child Labor, School Attendance, and Poverty in
Mexico and Venezuela. Mexico City: Centro de Estudios Económicos, El Colegio de
México.
Goulart, Pedro and Arjun S. Bedi. 2008. Child Labour and Educational Success in Portugal.
Economics of Education Review 27(5): 57587.
Grootaert Christiaan. 1999. ―Child Labor in Cote-d’Ivoire,‖ in Christiaan Grootaert and Harry
Anthony Patrinos, eds. The Policy of Child Labor: A Comparative Study, pp. 2362. New
York: St. Martins Press.
Grootaert, Christiaan and Ravi Kanbur. 1999. ―Child Labor: A Review.‖ Policy Research
Working Paper 1454, The World Bank, Washington, D.C.
Grootaert, Christiaan, and Harry Anthony Patrinos, eds. 1999. The Policy Analysis of Child
Labor: A Comparative Study. New York: St. Martins Press.
Gunnarsson, Victoria, Peter Orazem, and Mario A. Sánchez. 2006. ―Child Labor and School
Achievement in Latin America.World Bank Economic Review 20(1): 3154.
Hazarika, Gautam and Arjun S. Bedi. 2003. Schooling Costs and Child Work in Rural
Pakistan.‖ Journal of Development Studies 39(5): 2964.
66
Heady, Christopher. 2000. ―What is the Effect of Child Labour on Learning Achievement?
Evidence from Ghana.‖ Innocenti Working Paper 79, UNICEF Innocenti Research
Centre, Florence.
Hoodfar, Homa. 1986. Child Care and Child Survival in Low-Income Neighborhoods of Cairo.
MEAwards Working Paper, No. 29. Cairo: The Population Council.
———. 1996. ―Survival Strategies and the Political Economy of Low-Income Households in
Cairo,‖ in Diane Singerman and Homa Hoodfar, eds. Development, Change, and Gender
in Cairo: A view from the Household, pp. 126. Bloomington IN: Indiana University
Press.
Ibrahim, Barbara, Sunny Sallam, Sahar El-Tawila, Omaima El-Gibaly, and Fikrat El-Sahn. 2000.
Transitions to Adulthood: A National Survey of Egyptian Adolescents. New York: The
Population Council.
International Labour Organization (ILO). 1982. Resolution concerning statistics of the economically
active population, employment, unemployment and underemployment, adopted by the Thirteenth
International Conference of Labour Statisticians (October 1982).
[http://www.ilo.org/public/english/bureau/stat/download/res/ecacpop.pdf] Accessed August 12, 2009.
International Labour Organization (ILO). 2002. Current International Recommendations on
Labour Standards. Geneva, Switzerland: International Labour Office.
Jensen, Peter and Helena Skyt Nielson. 1997. ―Child Labor or School Attendance? Evidence
from Zambia.‖ Journal of Population Economics 10: 40724.
67
Kamphoefner, Kathy R. 1996. ―What’s the Use? The Household, Low-Income Women, and
Literacy, in Diane Singerman and Homa Hoodfar, eds. Development, Change, and
Gender in Cairo, pp. 80109. Bloomington IN: Indiana University Press.
Kandel, William and David Post. 2003. ―After School Work in Mexico: Competing for
Children’s Time Success.‖ International Journal of Educational Development 23(3):
299314.
Knaul, Felicia. 1995. Young Workers, Street Life and Gender: The Effect of Education and
Work Experience on Earnings in Colombia.‖, PhD diss., Harvard University.
Langsten, Ray and Tahra Hassan. 2007. ―Basic Education Attainment in Egypt: Trends and
Determinants.‖Paper presented at the 2007 Annual Meeting of the Population
Association of America (PAA). New York, March 29-31.
Langsten, Ray and Tahra Hassan. 2009. ―Education Transitions in Egypt.Poster presented at
the 2009 annual Meeting of the Population Association of America (PAA). Detroit, MI,
April 30-May 2.
Lee, Lung Fei. 1992. ―Amemiya’s Generalized Least Squares and Tests of Overidentification in
Simultaneous Equation Models with Qualitative or Limited Dependent Variables.‖
Econometric Reviews 11 (3): 31928.
Levison, Deborah. 2000. ―Children as Economic Agents.‖ Feminist Economics 6(1): 12534.
68
Levison, Deborah, Jasper Hoek, David Lam, and Suzanne Duryea. 2007. ―Intermittent Child
Employment and Its Implications for Child Labour Estimates.‖ International Labour
Review 146(34): 21751.
Levison, Deborah and Karine S. Moe. 1998. ―Household Work as a Deterrent to Schooling: An
Analysis of Adolescent Girls in Peru.‖ Journal of Developing Areas 32(3): 33956.
Levison, Deborah, Karine S. Moe, and Felicia Knaul. 2001. ―Youth Education and Work in
Mexico.‖ World Development 29(1): 16788.
Levy, Victor. 1985. ―Cropping Pattern, Mechanization, Child Labor, and Fertility Behavior in a
Farming Economy: Rural Egypt.‖ Economic Development and Cultural Change 33(4):
77791.
Lloyd, Cynthia, Sahar E. Tawila, Wesley Clark, and Barbara Mensch. 2003. ―The Impact of
Educational Quality on School Exit in Egypt.‖ Comparative Education Review 47(4):
44467.
McCall, Brian. 1992. ―A Note on the Identifiability of a Dynamic Binary Choice Model with
State Dependence.‖ Economic Letters 40(3): 27380.
Mensch, Barbara S., Barbara L. Ibrahim, Susan M. Lee, and Omaima El-Gibaly. 2003. ―Gender-
role Attitudes among Egyptian Adolescents.Studies in Family Planning 34(1): 818.
Moffitt, Robert. 2005. ―Remarks on the Analysis of Causal Relationships in Population
Research.‖ Demography 42(1): 91108.
69
Mueller, Eva. 1984. The Value and Allocation of Time in Rural Botswana.‖ Journal of
Development Economics 15(13): 32960.
Netz, Janet S. and Jon D. Haveman. 1999. ―All in the Family: Family, Income, and Labor Force
Attachment.Feminist Economics 5(3): 85106.
Egypt Ministry of Education. 2003. ―The National Plan for Education for All (2002/2003–
2015/2016).‖ Arab Republic of Egypt Ministry of Education.
Patrinos, Harry Anthony and George Psacharopoulos. 1997. ―Family Size, Schooling and Child
Labor in Peru: An Empirical Analysis,‖ Journal of Population Economics 10(4): 387
405.
Post, David and Suet-ling Pong. 2000. ―Employment During Middle School: The Effects on
Academic Achievement in the U.S. and Abroad.Educational Evaluation and Policy
Analysis 22(3): 27398.
Psacharopoulos, George. 1997. ―Child Labor versus Educational Attainment: Some Evidence
from Latin America.‖ Journal of Population Economics 10(4): 37786.
Psacharopoulos, George and Ana Maria Arriagada. 1989. ―The Determinants of Early Age
Human Capital Formation: Evidence from Brazil.‖ Economic Development and Cultural
Change 37(4): 683704.
Reynolds, Pamela. 1991. Dance Civet Cat: Child Labour in the Zambezi Valley. Athens, OH:
Ohio University Press.
70
Rosenzweig, Mark R. 1977. ―Farm-Family Schooling Decisions: Determinants of the Quantity
and Quality of Education in Agricultural Populations.‖ Journal of Human Resources
12:7192.
Skoufias, Emmanuel. 1994. ―Market Wages, Family Composition and the Time Allocation of
Children in Agricultural Households.‖ Journal of Development Studies 30(2): 33560.
Stinebrickner, Ralph and Todd R. Stinebrickner. 2003. ―Working During School and Academic
Performance.‖ Journal of Labor Economics 21(2): 47391.
Sultana, Ronald G. 2007. The Girls’ Education Initiative in Egypt. Report presented to the
National Council for Childhood and Motherhood (NCCM), Cairo, Egypt.
Tyler, John H. 2003. Using State Child Labor Laws to Identify the Effect of School-Year Work
on High School Achievement. Journal of Labor Economics 21(2): 381408.
Tzannatos, Zafiris. 2003. ―Child Labor and School Enrollment in Thailand in the 1990s.‖
Economics of Education Review 22(5): 52336.
Wahba, Jackline. 2006. ―The Influence of Market Wages and Parental History on Child Labour
and Schooling in Egypt.‖ Journal of Population Economics 19: 82352.
Zibani, Nadia. 2002. ―Gender Differentials in Children’s Work Activities: 1988–98,‖ in Ragui
Assaad, ed. The Egyptian Labor Market in an Era of Reform, pp. 178220. New York:
American University in Cairo Press.
71
APPENDIX
Household wealth proxy
To capture the effect of wealth on child labor and schooling we constructed a composite
variable based on the ownership by the household of a list of 23 durable goods and on a series of
housing characteristics, such as type of floor and ceiling and number of rooms. Following Deon
Filmer and Lant Pritchett (2001), we used factor analysis to obtain the weights that combine the
various indicators into a single composite wealth score. Since we are using the asset index as a
proxy for household wealth, it refers to the long-run economic status of households rather than
its current poverty or consumption status.
Because wealth in urban and rural areas take different forms, a single index was not
adequate to rank urban and rural households along a wealth continuum. We therefore opted to
construct separate wealth scores for urban and rural households. To allow for non-linearity in its
effect, we used the wealth score to construct a series of dummy variables indicating whether the
household is in the bottom, next to bottom, or top three quintiles of wealth distribution in urban
and rural areas respectively. We created the set of wealth quintile variables included in the
regression by interacting the urban and rural indices (in quintiles) with the corresponding
urban/rural dummy variables. This allows the wealth index to differ for urban and rural
households. The reference category describes residents of greater Cairo who are in the lowest
wealth quintile.
The analysis combines measures of wealth that can be broken down conceptually into
housing assets, durable goods, and financial assets, as follows:
72
(1) Housing quality variables: number of rooms; quality of material of walls, floor,
and roof; presence of a telephone; and access to the electricity network. (Each of these
indicators was ranked from worst to best on a two or three point scale).
(2) Ownership of durable goods including: fridge, freezer, dishwasher, color TV,
black and white TV, VCR, air conditioner, microwave, gas stove, kerosene stove, fan, water
heater, space heater, sewing machine, iron, radio, washing machine, camera, bicycle,
motorcycle or scooter, automobile, taxicab, and truck
(3) Access to interest or dividend income from financial assets and bank deposits.
Like Filmer and Pritchett (2001) we decided to only retain the first factor, which captures
the largest amount of information that is common to all the variables. Appendix Table 1 reports
the scoring coefficients for this factor. These estimates are obtained using the maximum-
likelihood estimation method without rotation, retaining only one factor. The results are very
robust to the retention of more factors and to estimation method. Theoretically, the asset index is
a standardized variable with zero mean and a standard deviation of 1. The estimation routines
never yield a standard deviation of exactly one, unless an exact solution to the factor model is
found. Our model yields standard deviations of 0.941 in urban areas and 0.937 in rural areas.
Since most variables are measures as 0 or 1, a change from zero to 1 would change the asset
score by the value of the variable’s coefficient divided by its standard deviation. So, for example,
owning a telephone increases the asset score by 0.23 in urban areas and by 0.25 in rural areas.
Owning a refrigerator increases it by 0.38 and 0.37 respectively.
73
scoring
coefficients
Mean SD
scoring
coefficients
Mean SD
number of rooms in dwelling 0.056 3.584 1.151 0.033 4.048 1.778
quality of wall materials 0.020 0.803 0.398 0.037 0.594 0.491
quality of floor materials 0.064 0.94 0.237 0.084 0.576 0.494
quality of roof materials 0.065 0.892 0.31 0.073 0.506 0.5
own a phone 0.117 0.484 0.5 0.090 0.149 0.357
own a fridge 0.127 0.875 0.331 0.187 0.46 0.499
own a stand alone freezer 0.039 0.074 0.262 0.027 0.012 0.111
own a dishwasher 0.017 0.026 0.158 0.015 0.005 0.071
own a color TV 0.161 0.758 0.429 0.148 0.329 0.47
own a black and white TV -0.049 0.275 0.446 -0.031 0.521 0.5
own a VCR 0.066 0.209 0.407 0.046 0.03 0.17
own an air-conditioner 0.035 0.054 0.225 0.019 0.005 0.073
own a microwave 0.009 0.017 0.129 0.012 0.011 0.106
own a gas stove 0.089 0.852 0.355 0.122 0.557 0.497
own a kerosene stove -0.062 0.437 0.496 -0.049 0.724 0.447
own an electrical fan 0.091 0.773 0.419 0.105 0.496 0.5
own a water heater 0.153 0.538 0.499 0.105 0.102 0.303
own a space heater 0.044 0.099 0.299 0.032 0.022 0.147
own a sewing machine 0.038 0.229 0.42 0.037 0.097 0.295
own an iron 0.130 0.792 0.406 0.162 0.461 0.499
own a radio 0.065 0.832 0.374 0.059 0.689 0.463
own a washing machine 0.078 0.925 0.264 0.109 0.687 0.464
own a camera 0.055 0.162 0.369 0.048 0.047 0.212
own a bicycle 0.021 0.18 0.384 0.026 0.188 0.391
own a motorcycle or scooter 0.006 0.015 0.122 0.020 0.018 0.135
own an automobile 0.050 0.108 0.31 0.036 0.019 0.137
own a taxicab 0.006 0.01 0.101 0.007 0.008 0.089
own a truck 0.008 0.011 0.104 0.013 0.006 0.076
Asset Index 0 0.941 0 0.937
Appendix Table 1. Scoring Coefficients and Summary Statistics for Variables Entering the Computation
of the First Factor
urban households
rural households
Notes: All variables except for number of rooms take on a value of zero and one. Number of rooms ranges from 1
to 12 in urban areas and 1 to 20 in rural areas.
74
... Webbink, Smits, and de Jong (2012) show that excluding unpaid household and family business work, which they call "hidden" child labor, underestimates its prevalence. Assaad et al. (2010), working in Egypt, and Levison and Moe (1998), in Peru, emphasize the importance of including domestic activity in child labor, as it especially burdens girls and may reduce school attendance. Existing research suggests that domestic child labor is likely undercounted (Pocock et al., 2021), and excluding it may impact work-school tradeoff estimates, especially among girls. ...
... We measure only attendance, not educational outcomes, but other researchers have linked attending school to low test scores. Emerson et al. (2017) Where research shows that school attendance is lower among working than non-working children (Assaad et al., 2010;Beegle et al., 2009;Putnick and Bornstein, 2015;Ray and Lancaster, 2005), some evidence suggests only a partial tradeoff between school and work among child laborers, as the two activities are not perfect substitutes (Attanasio et al., 2010;Edmonds and Shrestha, 2014;Ravallion and Wodon, 2000). Evidence from compulsory schooling reform in Turkey points to higher substitutability in rural than urban areas (Dayioglu-Tayfur and Kirdar, 2020). ...
... Using longitudinal data from Indonesia, Sim et al. (2017) show that child labor negatively impacts math skills, and market work particularly hinders educational attainment. Nevertheless, extant literature suggests that children are as likely to trade schooling for domestic work as they are for market work (Assaad et al., 2010;Levison and Moe, 1998). Levison and Moe (1998) identify a gap in the literature: Due to data limitations, many studies focus only on whether children work or go to school. ...
Preprint
Full-text available
This paper examines school attendance among children in Nepal to evaluate the foregone human capital investment potentially associated with trading child labor for education. The researchers survey almost 2,900 household representatives and more than 3,400 children age 5 to 13. Response comparisons reveal that working children miss school more often than non-working children. Using a regression framework, this study confirms that working children are more likely to miss at least one day of school per week than non-working children. The children's self-reported data reveal that engagement in market work is associated with almost 1.5 missed school days per week for boys and girls. Girls engaged in domestic work attend 0.4 fewer school days per week. Using responses from household representatives, however, the researchers find a roughly 0.3-day reduction for children engaged in market work and no reduction associated with domestic work. Because the tradeoff diminishes using proxy responses, especially among girls, the researchers deploy a randomized survey experiment to investigate whether adults perceive schooling's benefits for boys and girls differently. The study finds that respondents spontaneously identify more benefits for boys than girls. The results warrant further research into which survey instruments accurately measure child labor and school attendance; furthermore, they highlight the importance of including domestic work in child labor research. JEL category: O10, J20
... This often starts at an early age. In Egyptian communities where basic public services are limited, young girls are asked to fetch water and fuel, dispose of garbage, do laundry, and care for younger siblings (Assaad, Levison & Zibani 2010). Although our study did not find that girls' participation in household chores impacted their time spent on school work, research in other contexts has suggested this (Kazeem 2013) and it will be important to monitor this. ...
... Rees (2017) in particular found that children from higher-income countries spend more time on leisure activities, such as watching television, listening to music, and using the computer than children from lowerincome countries. Other research has also shown that children in poorer living conditions may have less time available for leisure and study (Fonta et al. 2020) as they may be required to contribute more to family life, including participation in domestic chores, such as fetching water and disposing of waste (Assaad et al. 2010). ...
Article
Full-text available
Population-based information on how school-going children spend their daily lives outside of school from their own perspective in African contexts is limited. Using representative data from 2124 four- and six-graders who completed the 2018 International Survey of Children�s Wellbeing in Khomas region, Namibia, this article describes children�s out-of-school activities, the contextual factors that influence patterns of time use and how these patterns of time use affect subjective wellbeing. Results indicate that children in Namibia spend time on diverse activities related to school, family, and leisure. Furthermore, children�s participation in chores contributed positively to their life satisfaction and both play and spending time with family had a valuable impact on their wellbeing. By describing at a population level children�s out-of-school time use and satisfaction, this study contributes to understanding the richness of children�s lives in Khomas region and the valuable contributions they make to their families and communities.
... Child labour includes the violation of the human rights of minors and it is associated with various harmful activities (Edmonds and Pavcnik 2005). Several definitions of child labour have been presented to date, ranging from wage work for minors (Psacharopoulos 1997) and various forms of market work (Ray and Lancaster 2005) to domestic activities (Basu et al. 2010;Assaad et al. 2010). These findings are in line with Levison and Moe (1998) and Webbink et al. (2012), who also defined domestic activities as a "hidden" form of child labour. ...
... Secondly, it is difficult to estimate the hidden or non-paid work by minor employees, such as domestic activities, which are defined as a form of child labour (Levison and Moe 1998;Basu et al. 2010;Assaad et al. 2010;Webbink et al. 2012). Thirdly, only published scientific papers from different databases and reports from international organizations were selected and included. ...
Article
Full-text available
The consequences of the recent pandemic have been disproportionately disruptive to several social groups, including children. As developing economies have been firefighting the recent pandemic, the welfare of minors could be affected and children’s economic exploitation and abuse could increase. Therefore, the present research aims to shed light on and to investigate the association between child labour in developing countries and pandemics, including the coronavirus, through conducting a systematic literature review on previous empirical studies. The present research concludes that previous studies on non-COVID-19 pandemics have mainly focused on the African economies, while studies on the recent pandemic have focused on Asian countries. In addition, differences were observed in relation to the methodological approaches and the characteristics of minor employees and the protection services in certain countries have proven to be insufficient. Suggestions for future research and policy implications are presented.
... Children's time allocation plays an important role in their well-being, education and future life opportunities. Previous research on time allocation patterns provides valuable insights into the complexity of children's worlds, social hierarchies, economic realities (Assaad, Levison, and Zibani 2010;DeGraff and Bilsborrow 2003;Koissy-Kpein 2012) and its relevance for policy formation (Ilahi 2001). This paper contributes to these debates examining the practices of children's time allocation in rural Sierra Leone. ...
Article
Children’s time allocation plays an important role in their well-being, education and future chances in life. This paper examines children’s time use in rural Sierra Leone in the context of dynamics of power. It analyses the ways in which children co-construct their patterns of time use and create their own temporal spaces within the boundaries of existing expectations and norms. Drawing on rich qualitative data we argue that children’s routines are shaped by an interplay of various dimensions and expressions of power that create and re-create gender, generation and other social orders in a particular society. This power interplay creates polyphonic temporal spaces where school, work and play time overlap each other.The analysis of this polyphony of time use has demonstrated that children’s time allocation goes beyond simple one dimensional task distribution and includes a complex network of activities that may take place simultaneously.
... There is a vast body of research on the determinants of educational investment in children and many studies have focused on the nexus between child labour and education (Basu, 1999). However, few studies have addressed the role of unpaid household duties (Assaad et al., 2010;Levison & Moe, 1998;Levison, Moe, & Knaul, 2001). In addition to the main analysis, we explore the impact on child participation in housework. ...
Article
Full-text available
This study examines the effect of household water accessibility on children’s educational attainment in villages situated in the remote hilly and mountainous areas of Nepal. Educational attainment was measured based on school attendance, grade repetition, and completion of primary and lower-secondary schooling. The estimation results show that a one-hour increase in the time spent on a water-fetching trip will decrease the probability of girls completing primary school by about 17 percentage points (in the age group of 14–16 years). Although boys’ completion rate is less affected, they are more likely to repeat a grade. Additional analyses indicate that these results are driven by the increased participation of older boys and younger girls in household duties.
... In many religious societies, gender role divisions limit or restrict women's access to education (Subrahmanian, 2005). In Pakistan, for example, boys' education is perceived as familial and social capital, while young girls in the family are perceived as their mothers' helpers and are mainly expected to take care of their younger siblings (Assaad et al., 2010;Khalid & Mujahid-Mukhtar, 2002). Research on patriarchal societies suggests that education is a key factor for women's integration into the labor market, as it expands their employment opportunities, particularly into more profitable and respected fields (Abu- Rabia-Queder, 2016;Steiber et al., 2016). ...
Article
Full-text available
This paper presents the narratives of Druze women in Israel, focusing on their strategies in dealing with their parents and the clergy for obtaining secondary and higher education. These reveal their use of different forms of agency in their struggle for education, pointing to the key role played by their mothers. I argue that these patterns are related to the intersectionality of at least two subordinating and oppressive mechanisms: the Israeli state’s negligence with respect to Druze society and the social and religious structures imposed on Druze women. Their struggle to seek education is an example of women’s resistance against patriarchal oppression from within.
Article
Full-text available
Abstract: This paper investigated the probability of employment of mothers and children (aged 10–18) in Iranian urban household and the factors which influence them with expansion of neoclassic model. Bprobit model and the Stata Software were used to that end. The sample size includes 6517 Iranian urban households (in 2014) who had children aged 10–18. The data was extracted from the Urban Household National Survey. 46. 4% of the sample are females and 54. 6% are males. 93. 97% of children are unemployed and 6. 03% are employed. The results showed that children do not necessarily drop out of school to work and be employed. An increase in the level of education of parents negatively influences the employment of the children aged 10–18, and the effect of father’s education is more decisive in this regard. Furthermore, the increasing of father’s education reduces the possibility of mother’s employment in this household. The substitution effect showed that mother’s and child’s employment are substitutes, as is father’s employment with mother’s and child’s employment. The wealth effect showed that an increase in household per capita income reduces the possibility of children’s employment and raises the possibility of mother’s employment. Furthermore, the size of the family reduces the possibility of mother’s employment and raises the possibility of children’s employment.
Article
Children in Sub-Saharan Africa spend a non-trivial amount of their time cooking and collecting fuel for domestic use. This is particularly the case in rural areas where access to efficient energy is low, and children's academic performance is poor. This paper argues that households' use of cooking gas could reduce the time spent doing domestic chores, increase learning time, and improve children's school performance. We investigate this proposition using the Ghana Living Standards Survey data. We employ different instrumental variable estimations techniques to deal with the possible endogeneity problem. The results show that cooking with gas marginally improves the learning outcome of children in rural Ghana. Our results imply that the adoption of cooking gas could enhance human capital development in developing countries.
Article
Full-text available
Women make significant contributions in social and economic development of any society but face discrimination, marginalization and multiple barriers to formal education which could further challenge their developmental roles. Many communities in Tanzania still hold outdated traditional practices which favour the male child in the provision of basic rights such as education. This paper discusses the experiences of house girls in Tanzania. In poverty stricken families, the girl-child is forced to take up jobs like that of the house-girl which often limit or curtail their ability to acquire formal education. This study was conducted in Northern Tanzania, in the rural areas of Mwanza region. A qualitative research approach which utilized a critical ethnographic research design was employed. Data was collected through in-depth individual interviews and observation methods. A total of ten house-girls participated in the interviews. In addition, data obtained from a review of literature, anecdotal evidence and informal conversations with various community members also contributed to this study. The results highlight the circular and cumulative causational relationship in which education is both a symptom, and simultaneously a cause of the problem. The study indicates that house girl employment typically begins at a very young age and it disrupts girls' education and blights educational outcomes. Likewise, the study indicates that regardless of how hard a girl works does not lead to an improvement in economic outcomes. Recommendations are made for the government and non-governmental leaders to devise strategies and policies that will promote and safeguard the rights of girls.
Chapter
In this chapter we examine the dilemmas that exist in the formulation of child labor policies. While public opinion is often sensitive to the needs of children in general, and those of working children in particular, recent years have seen a debate on the most appropriate strategies to address the problem. Despite the fact that children are perceived as a vulnerable population segment, the introduction of the child agency perspective has meant in practice a questioning of the traditional models of protection in social work with children. This has led to qualify the objectives of “eradication” of child labor and to establish limits in international labor standards, with the counterbalance of local needs. In any case, income redistribution policies within a framework of consensus among social actors have proven to be particularly effective. Therefore, despite the controversies regarding the degree of relevance and adjustment, there is sufficient evidence to defend investment in education and health for preventive purposes.
Article
Full-text available
This article considers the estimation of the causal effect of fertility on female-labor-force participation equations. My main concern is to examine two considerations, the endogeneity of fertility and the impact of controlling for unobserved heterogeneity and for predetermined existing children. Using PSID data, a switching binary panel-data model that accounts for selectivity bias as well as for other forms of time-invariant unobserved heterogeneity is estimated. Individual effects are allowed to be correlated with the explanatory variables, which can be predetermined as opposed to strictly exogenous. Family sex composition is used as an instrument for exogenous fertility movements. The results indicate that exogeneity assumptions of children variables induce a downward bias in absolute value in the estimated negative effect of fertility on participation, although the failure to account for unobserved heterogeneity overstates this effect. Moreover, stronger effects of fertility are found when existing c...
Article
Full-text available
This article draws upon a time allocation survey of a four‐year panel of households from six villages in semi‐arid tropic rural India. The inter‐relationships among adult male. female and child market wage rates, and time allocated to domestic activities, leisure and schooling by boys and girls are analysed. The results based on the pooled time‐series and cross‐section individual observations suggest that adult and child wage rates play an important role in the time allocation of both boys and girls.
Article
Child labor’s effect on academic achievement is estimated using unique data on third and fourth graders in nine Latin-American countries. Cross-country variation in truancy regulations provides an exogenous shift in the ages of children normally in these grades, providing exogenous variation in the opportunity cost of children’s time. Least squares estimates suggest that child labor lowers test scores, but those estimates are biased toward zero. Corrected estimates are still negative and statistically significant. Children working 1 standard deviation above the mean have average scores that are 16 percent lower on mathematics examinations and 11 percent lower on language examinations, consistent with the estimates of the adverse impact of child labor on returns to schooling.
Article
To investigate the differential school enrollment and expenditure behavior of farm and nonfarm populations revealed in a number of empirical studies, an intertemporal model of farm-family schooling decisions is formulated in which the roles of education in agricultural production are explicitly considered and in which the productivity of children can be augmented by increasing either the quantity or quality of schooling. Results obtained from fitting the empirical specification of the theoretical framework to U.S. state-level data show that the participation of farm children in agricultural production affects significantly the demand for the quality and quantity of schooling in rural-farm areas. In particular, increasing factor inputs other than child labor reduces the enrollment rate of farm teenagers but increases educational expenditures such that the overall quality of farm children rises.
Article
The identity of the people engaged in a transaction is a major determinant of the institutional mode of transaction. The family is the locale of transactions in which identity dominates; however, identity is also important in the market. The opportunity to reduce transaction costs by making specific investments in exchanges between identified parties affects the organization of social activity, the division of labor, and, in particular, the interaction between specialization by identity and by other dimensions of transactions. Economic and social development can be understood in terms of changes in the modes of transacting. Within this broad framework one can also analyze the transactions in which families have an advantage over other institutions, the conditions that make families of various types more or less efficient than alternative modes of transaction, and the implications of family membership for transactions with others.
Article
An international debate over student employment turns on the question of whether work generally helps or harms children's development. This article focuses on two indicators of child development that are goals in all education systems: math and science achievement. After reviewing the major theoretical perspectives on school achievement and employment, we propose a general framework for analyzing their relationship. We then present the results of our cross-national study. From the U.S., we use cross-sectional and longitudinal NELS data. In the U.S. and in 22 other nations, we use cross-sectional TIMSS data to examine the effects of after-school work during the eighth grade. Our findings from each investigation are consistent: For boys, and to a lesser extent for girls, there are negative effects on math and science achievement that are associated with adolescent employment, even after controlling for family background and, in the NELS, after controlling for prior achievement.
Article
Using the 1985-86 Peru Living Standard Survey (PLSS) data on hours spent in paid work, unpaid work, household chores, and school, we analyze the determinants of hours spent in chores and in school for adolescent girls in Peru. Generalized Tobit methods are employed to estimate hours in school and hours of chores separately. This paper addresses two major gaps in the literature. First, the focus of most studies has been on education alone or on paid labor market work as the primary deterrent to schooling. Since in many contexts boys are more likely to be engaged in labor market work than girls, much of the work that girls do has been ignored. Second, owing to a lack of data, the majority of studies have focused on whether children engage in activities or not, as opposed to how much time they spend on those activities. In contrast, this paper uses detailed information on hours spent in paid work, unpaid work, household chores, and school, and focuses on girls' time allocated to household chores and school. The results indicate that girls in better living conditions with more educated mothers spend fewer hours doing chores and more hours in school. Girls with more preschool-age siblings do more chores and spend less time in school, but this effect on chores can be offset by the presence of other girls and women with whom they presumably share household tasks. Surprisingly, mother's presence reduces school hours substantially, although it increases the likelihood of enrollment.