1.
Understanding Children’s Work Project Working Paper Series, November 2006 - Revised 2008
an issue paper
Child labour and Education For All:
L. Guarcello
F. C. Rosati
S. Lyon
64520
Child labour and Education For All:
an issue paper
L. Guarcello*
S. Lyon*
F.C. Rosati**
Working paper
November 2006
Revised June 2008
Understanding Children’s Work (UCW) Project
University of Rome “Tor Vergata�?
Faculty of Economics
Via Columbia 2 , 00133 Rome
Tel: +39 06.7259.5618
Fax: +39 06.2020.687
Email: info@ucw-project.org
As part of broader efforts towards durable solutions to child labor, the International
Labour Organization (ILO), the United Nations Children’s Fund (UNICEF), and the
World Bank initiated the interagency Understanding Children’s Work (UCW) project
in December 2000. The project is guided by the Oslo Agenda for Action, which laid
out the priorities for the international community in the fight against child labor.
Through a variety of data collection, research, and assessment activities, the UCW
project is broadly directed toward improving understanding of child labor, its causes
and effects, how it can be measured, and effective policies for addressing it. For
further information, see the project website at www.ucw-project.org.
This paper is part of the research carried out within UCW (Understanding Children's
Work), a joint ILO, World Bank and UNICEF project. The views expressed here are
those of the authors' and should not be attributed to the ILO, the World Bank,
UNICEF or any of these agencies’ member countries.
*
UCW-Project
**
UCW-Project and University of Rome “Tor Vergata�?
Child labour and Education For All:
an issue paper
Working Paper
November 2006
Revised June 2008
ABSTRACT
Education is a key element in the prevention of child labour; at the same time,
child labour is one of the main obstacles to Education for All (EFA).
Understanding the interplay between education and child labour is therefore
critical to achieving both EFA and child labour elimination goals. This paper
forms part of UCW broader efforts towards improving this understanding of
education-child labour links, providing a brief overview of relevant research
and key knowledge gaps. The study largely confirm the conventional wisdom
that child labour harms children's ability to enter and survive in the school
system, and makes it more difficult for children to derive educational benefit
from schooling once in the system. The evidence also suggested that these
negative effects are not limited to economic activity but also extend to
household chores, and that the intensity of work (in economic activity or
household chores) is particularly important in determining the impact of work
on schooling. As regards the link between education provision and child
labour, it pointed to the important role of inadequate schooling in keeping
children out of the classroom and into work. This evidence indicated that both
the school quality and school access can play an important role in household
decisions concerning whether children study or work.
Child labour and Education For All:
an issue paper
Working Paper
November 2006
Revised June 2008
CONTENTS
1. Introduction.............................................................................................................................. 1
2. Child labour as an obstacle to Education For All: how work affects
children's ability to attend and benefit from schooling ...................................................... 1
2.1 Child labour and school attendance: descriptive evidence ..................................... 2
2.2 Child labour and school attendance: causal links ..................................................... 6
2.3 Child labour and learning achievement ................................................................... 10
2.4 Child labour and schooling: student and teacher perceptions ............................. 14
3. Education provision as a factor in child labour: how inadequate
schooling can “push�? children into work........................................................................... 15
3.1 Impact of supply constraints..................................................................................... 16
3.2 Impact of school quality ............................................................................................ 18
3.3 Impact of special transitional education and flexible schooling
programme on child labour............................................................................................... 23
4. Conclusion .............................................................................................................................. 25
References...................................................................................................................................... 28
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
1 JUNE 2008
1. INTRODUCTION
1. The international community's efforts to achieve Education For All (EFA) and
the progressive elimination of child labour are inextricably linked. Education – and,
in particular, education of good quality up to the minimum age for entering into
employment – is a key element in the prevention of child labour. There is broad
consensus that the single most effective way to stem the flow of school age children
into work is to extend and improve access to school, so that families have the
opportunity to invest in their children’s education and the returns to such an
investment are greater than those associated with involving children in work.
Conversely, when the expected returns to education are low or education costs are
high, schooling is likely to be seen by households as a less attractive alternative to
work for their children.
2. At the same time, child labour is one of the main obstacles to EFA, as
involvement in child labour is generally at a cost to children’s ability to attend and
perform in school. According to UNESCO, there were 104 million children of
primary-school going age not enrolled in school at the turn of the millennium, the
majority of whom are working children. Child labour also adversely affects the
academic achievement of the considerable number of children who combine work
and school, often resulting in these children leaving school prematurely and entering
into work.
3. Understanding the interplay between education and child labour is therefore
critical to achieving both EFA and child labour elimination goals. This paper forms
part of UCW broader efforts towards improving this understanding of education-child
labour links, providing a brief overview of relevant research and key knowledge gaps.
4. The paper is structured as follows. The next section examines child labour as an
obstacle to achieving EFA, reviewing descriptive and econometric evidence of the
costs of child labour in terms of school entry, school survival and learning
achievement. Section 3 then looks at education provision as a factor in child labour,
reviewing empirical evidence of how school access and quality influence household
decisions on the allocation of children’s time between work and school. Section 3
also looks at information gaps that need to be filled in order to assess the potential of
transitional education and flexible schooling initiatives in supporting national efforts
towards EFA and child labour reduction. Section 4 concludes.
2. CHILD LABOUR AS AN OBSTACLE TO EDUCATION FOR ALL: HOW
WORK AFFECTS CHILDREN'S ABILITY TO ATTEND AND BENEFIT
FROM SCHOOLING
5. This section reviews evidence relating to the impact of work on school
attendance, learning achievement and school life. It highlights the constraint that
child labour poses to achieving Education For All. The section looks firstly at the
effects of child labour on children's ability to enter and survive in the school system,
and secondly at the effect of child labour on children's ability to derive educational
benefit from schooling once in the system. Obviously, the two issues are closely
related, but a distinction is useful for expositional purposes.
2 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
2.1 Child labour and school attendance: descriptive evidence
6. Working children are disadvantaged vis-à-vis their non-working counterparts in
terms of their ability to attend school in many of the countries where child labour is
common. As shown in Figure 1, in a sample of 60 developing countries from the
UCW Country Statistics,2 working children face an attendance disadvantage of at
least 10 percent in 30 countries, of at least 20 percent in 16 countries and of at least
30 percent in 10 countries. In seven countries, on the other hand, working children
actually have a slight attendance advantage and in five others the attendance rates of
working and non-working children are virtually equal.
7. The wide cross-country variation in terms of the relative success of working
children in attending school could reflect underlying differences in the nature or
intensity of work carried out by children as well as structural differences in the way
that education systems accommodate the exigencies of children’s work.3 To the
extent that the latter explanation holds, the large cross-country variation suggests
substantial scope for policy intervention aimed at bringing and retaining working
children in school.
Figure 1. School attendance disadvantage(a) of working children, 7-14 years age group, selected countries
1.3
1.2
school attendance disadvantage index
1.1
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Ecuador
Senegal
Chad
Guyana
Jamaica
Togo
Uzbekistan
Burundi
Domenican Rep.
Lesotho
Moldova
Panama
El Salvador
Timor-East
Malawi
Argentina
Brazil
Mali
Portugal
Vietnam
Cameroon
Albania
Angola
Comores
Congo
Ghana
Guatemala
Iraq
Kenya
Mexico
Nicaragua
Philippines
CAR
Romania
Rwuanda
Sierra Leone
Trinidad Tobago
Azerbaijan
Yemen
Belice
Bosnia
Chile
Colombia
Gambia
Guinea Bissau
Honduras
Paraguay
Swaziland
Tanzania
Zimbawe
Turkey
Bolivia
Mongolia
namibia
Peru
South Africa
Sri Lanka
Zambia
Cote d'Ivoire
Sudan North
Notes: (a) School attendance disadvantage index refers to the school attendance rate of economically-active children expressed as a ratio of
the school attendance rate of non-economically active children. The smaller is the index value, the higher is the disadvantage faced by
economically-active children compared to children not involved in economic activity.
Source: UCW calculations based on household survey datasets
8. High levels of child labour therefore translate into large numbers of out-of-school
children in many national contexts, which in turn means lower overall attendance
rates and slower progress towards achieving Education For All (EFA). This negative
correlation between child labour and overall school attendance is illustrated in Figure
2
UCW Country Statistics consist of a core set of child labour and schooling indicators for over
70 countries. They are based on nationally-representative household surveys conducted as
part of ILO/IPEC SIMPOC, UNICEF MICS, World Bank LSMS and national household survey
programmes. The Country Statistics can be found at the UCW website (ucw-project.org).
3
Readers should also note that differences in data sources and survey instruments mean that
cross-country comparisons must be made with caution).
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
3 JUNE 2008
2, which plots rates of child economic activity and school attendance for boys and
girls for countries included in the UCW Country Statistics.
Figure 2. School attendance(a) and child labour, children aged 7-14 years, by sex
Notes: (a) School attendance rate refers to the number of 7-14 year-olds attending school expressed as a percentage of total
children in this age group.
Sources: UCW calculations based on household survey datasets, various countries
9. The preceding figures make clear that reducing child labour will be critical to
achieving EFA in many national contexts. But it is important to identify which work
categories or work settings are most detrimental to children’s school attendance in
order to guide policy towards EFA. Figure 3 looks at differences in school attendance
by general work category (i.e., economic or household chores) and by work setting
(i.e., family or non-family). The figure suggests that both distinctions are potentially
important.4 Household chores appear to pose a lesser barrier to school attendance than
economic activity, and family-based economic activity appears to interfere less with
schooling than similar work performed outside the family. This may be because
family work is more flexible to the exigencies of school, or because families have a
greater interest in safeguarding their children’s education.
10. But this evidence is only suggestive of possible differential impact of various
forms of works on attendance and therefore should be interpreted with caution. Some
of the children, for example, might be performing both economic and non economic
activities or both family and non family work. It could also be that household chores
and family-based economic activity are performed for fewer hours each week, leaving
more time for going to school (the issue of work intensity and school attendance is
looked at in the next section). More detailed evidence is required on the links between
work category/setting and school attendance in order to draw firmer conclusions.
Some of this evidence is presented later on, but more research is needed in this area.
4
The left hand graph plots the school attendance rate of children involved in economic activity versus that
of children involved in household chores, and right hand graph plots the school attendance rate of children
in family work versus that of children in non-family work. For each graph, observations lying along the 45
degree line indicate that the attendance rate of the two groups being plotted is the same. If the
observations lie above the 45 degree line, the attendance of the group plotted on the vertical axis is higher
than the attendance of the group on the horizontal axis, while if the observations lie below the 45 degree
line, the opposite holds true.
4 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
Figure 3. School attendance, work type (economic or household chores(a)), and work setting (family or non-family), children aged 7-14
years
Notes: (a) Children carrying out household chore for at least one hour during the reference week;
Sources: UCW calculation based on household survey datasets, various countries
11. Information on the school history of non-student working children is also
important in understanding the links between work and school attendance.
Particularly relevant in this context is the distinction among out-of-school working
children who are non school entrants (i.e., children never entering school), late
entrants (i.e., children not yet enrolled but who eventually will be) and those who are
early school leavers. The first group is undoubtedly worst off, denied the benefit of
formal education altogether, and therefore constitutes a particular policy priority. As
shown in Figure 4, in countries characterized by a relatively high prevalence of
children’s work in the age group 7-14 years, the ratio of children that enter school at
any age is lower. This is an indication that the higher rate of children’s work, the
higher the number of children that never enter school. For instance, in the case of
Ethiopia, 63% of children aged 10–14 have no formal schooling at all, and many
more from this age group enter school after the official starting age of six years.
12. Figure 5 and Figure 6 suggest that late entrants and early leavers also form
important components of the non-student working children population.5 Children’s
work is associated with a smaller proportion of children entering school at the official
entrance age (Figure 5) and with a higher proportion of children leaving the schooling
system prematurely (Figure 6) All three effects – non-entrance, delayed entrance and
early leaving – combine to reduce the total time working children spend in school
(Figure 7). These results underscore the fact that attention needs to be given to
analysing and addressing the role of children’s work at both ends of the primary
school cycle, i.e., to its role in preventing or delaying school entry and in children
leaving school prematurely.
5
Evidence suggests that the former is often contributing factor to the latter, i.e., late entrance
increases the chances that children will also leave the school system prematurely. See, for
example, UCW (2005b) Children’s work in Cambodia: a challenge for growth and poverty
reduction, June 2005
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
5 JUNE 2008
Figure 4. Gross school intake(a) and child labour, children aged 7-14 years, by sex
Notes: (a) Gross intake rate grade 1 refers to the number of new entrants in the first grade of primary education regardless of age, expressed as a
percentage of the population of the official primary school entrance age.
Sources: (1) UNESCO, EFA Global Monitoring Report 2005 (for gross intake rate); (2) UCW calculation based on household survey datasets, various
countries (for economically-active children)
Figure 5. Net school intake(a) and child labour, children aged 7-14 years, by sex
80
80
Fitted values Fitted values
60
60
40
40
20 20
0 0
20 40 60 80 100 20 40 60 80 100
net intake rate grade 1 (MALE) net intake rate grade 1 (FEMALE)
Notes: (a) Net intake rate grade 1 refers to the number of new entrants in the first grade of primary education of the official primary school entrance
age, expressed as a percentage of the population of the official primary school entrance age.
Sources: (1) UNESCO, EFA Global Monitoring Report 2005 (for gross intake rate); (2) UCW calculation based on household survey datasets, various
countries (for economically-active children)
Figure 6. School drop-out(a) and child labour, children aged 7-14 years, by sex
Notes: (a) Primary level drop-out rate refers to the percentage of pupils or students who drop out from a given grade or grades in a given school year.
It is the difference between 100% and the sum of the promotion and repetition rates.
Sources: (1) UNESCO, EFA Global Monitoring Report 2005 (for drop-out rate); (2) UCW calculation based on household survey datasets, various
countries (for economically-active children)
6 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
Figure 7. School life expectancy(a) and child labour, children aged 7-14 years, by sex
Notes: (a) School life expectancy (SLE) refers to the number of years a child of school entrance age is expected to spend at school or university,
including years spent on repetition.
Sources: (1) UNESCO, EFA Global Monitoring Report 2005 (for school life expectancy); (2) UCW calculation based on household survey datasets,
various countries (for economically-active children)
13. A fourth, frequently overlooked, group of out-of-school working children is
comprised of irregular school attendees (i.e., children formally enrolled school but
not attending for extended periods of time). Little is known about the size of this
group, owing to the fact that data on attendance regularity are rarely collected as part
of household surveys or government education statistics. But the often large
discrepancies between official school enrolment estimates (capturing children
formally enrolled) and attendance estimates (capturing children actually in class)
from household surveys suggest that this group of irregular attendees may be
considerable in many countries. Evidence from school-based surveys also suggests
that working children have more difficulty in attending class regularly in some
contexts (ILO/IPEC and UCW, 2005a). It stands to reason, therefore, that at least part
of the school attendance disadvantage of working children reported in Figure 1 is a
reflection of the fact that working children are forced to miss class more frequently
than their non-working counterparts.
2.2 Child labour and school attendance: causal links
14. In the previous section we presented descriptive evidence of the negative link
between school attendance and child labour. But for policy purposes it is important to
go beyond descriptive evidence to assess to what extent work involvement is a cause
of low school attendance (and of poor learning achievement, as discussed in next
section). While there has been considerable discussion of this issue in the literature
[see for example, Grootaert and Patrinos (1999), and Pushkar and Ray (2002)], there
have been very few analyses where the causal link between work involvement and
school attendance is definitively identified.
15. Establishing causality is complicated by the fact that child labour and school
attendance are usually the result of a joint decision on the part of the household, and
by the fact that this decision may be influenced by possibly unobserved factors such
as innate talent, family behavior and or family preferences. In fact, low returns to
education, the poor quality of schooling or high monetary and non-monetary costs of
schooling might make school attendance a less desirable than work for children.
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
7 JUNE 2008
Finally, one child's work might help to pay for his own or his siblings' schooling, so
again child labor might paradoxically increase school attendance (Manacorda, 2006).
This means that on the basis of cross-sectional data alone it is difficult to know, for
example, if it is low talent that induces a child not to go to school and hence start to
work, or if it is the preference or need to work that then induces a child to drop out of
school.
The use of panel data can help to address at least some of these issues and to get
firmer results in terms of causality. Panel data unfortunately remain scarce,
constituting an important obstacle to informed policy design. Where these data are
available, they underscore the importance of children’s as an obstacle to schooling.
Table 1. Determinants of school attendance in China (random effects logistic regression)(a)
School attendance Coef. Std. Err. z
Female -0.114 0.082 -1.39
School attendance 1989 0.788* 0.104* 7.61*
Household size -0.089** 0.035* -2.54*
Number of children -0.101 0.090 -1.12
Hours non-market wk -0.043* 0.011* -3.89*
Age 0.999** 0.141** 7.1**
Age squared -0.055** 0.006** -9.33*
Household head female 0.170 0.132 1.29
Water access 0.255** 0.098** 2.61*
Notes: * statistically significant at the 5% level.
Source: UCW (2005a). Towards statistical standards for children’s non economic work: A discussion based
on household survey data. Guarcello L., Lyon S., Rosati F.C., and Valdivia C.
16. The effects of work on school attendance can also take a more indirect form.
Work can lead to late school entry, which, in turn, is often associated with early
school drop out and lack of completion of a course of study. Research in Cambodia
illustrates this, indicating that work tends to delay school entry (or prevent it
altogether), reducing the probability of completing primary school (UCW, 2005b).
This effect is strongest for economic activities and for boys in Cambodia. Performing
economic activity reduces the probability of entering school (as measured by the
probability of entering school by age 14) of boys by 25 percent, and the probability of
entering by the official school entry age by 17 percent. Non-economic activity also
has a strong influence on school entry, again particularly for boys. Involvement in
household chores makes it about 13 percent less likely that boys enter school by age
six, and also about 13 percent less likely that boys enter school at all.
8 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
Table 2. Effect of work on school entry, by outcome and sex(a)
Economic activity Non-economic activity
Boys Girls Boys Girls
School entry by age 14 -25.11* -8.95 -12.60* -4.70
School entry by age 6 -17.37* -8.90 -13.23* -5.60
Notes: * statistically significant at the 5% level. (a) Reported figures measure the percentage
change (expressed on a 0 to 100 scale) in the probability associated to each school entry outcome
as a result of working at each age
Source: UCW (2005b) Children’s work in Cambodia: a challenge for growth and poverty reduction.
17. As discussed above, data limitations prevent us from presenting an easy
replicable approach to identify the causal effects of work and of working hours on
education. It is nonetheless possible to make use of synthetic indicators (kernel
regression in the examples shown below) to offer a more direct and synthetic view of
the relationship between hours of work and schooling for monitoring and policy
design purposes. Instruments like these are suitable for describing the probabilistic
link between variables, but cannot be used to derive strict causal relationships. They
are basically reduced form estimates, and the relationship estimated is subject to
change if the underlying structure changes (for example, if the gender distribution of
employment changes). They must therefore be interpreted with care.
Figure 8.School non attendance versus hours spent performing household chores and
economic activity, selected countries (Kernel regression)
(a) Bolivia
0.50 Household chores
0.45
Economic activity
0.40
Prob. not attend school
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0.0 5.1 10.2 15.3 20.4 25.5 30.6 35.7 40.9 46.0
Weekly hours spent performing household chores and economic activity
(b) Cambodia
0.90
Household chores
0.80
Prob. not attend school
0.70
Economic activity
0.60
0.50
0.40
0.30
0.20
0.10
0.00
0.0 5.1 10.2 15.3 20.4 25.4 30.5 35.6 40.7 45.8 50.9 56.0
Weekly hours spent performing household chores and economic activity
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
9 JUNE 2008
(c) Mali
1 Economic activity
0.9
Prob. not attend school
Household chores
0.8
0.7
0.6
0.5
0.4
1 6 11 16 21 26 31 36 41 46 51 56 61 66
Weekly hours spent performing household chores and economic activity
(d) Senegal
1 Economic activity
0.9
Prob. not attend school
Household chores
0.8
0.7
0.6
0.5
0.4
1 6 11 16 21 26 31 36 41 46 51 56 61 66
Weekly hours spent performing household chores and economic activity
Source: UCW (2005). Towards statistical standards for children’s non economic work: A discussion based on
household survey data. Guarcello L., Lyon S., Rosati F.C., and Valdivia C.
18. Figure 8 presents the results of kernel regressions reflecting the relationship
between hours of work and the probability of attending school for four countries
(Bolivia, Mali, Cambodia and Senegal) (UCW, 2005a). The results provide further
evidence that work and education are competing activities, indicating clearly that the
probability of attending school declines with the increase of hours spent at work in
both economic activity and household chores. But Figure 8 also indicates that the
relationship between working hours and school attendance is very different across
countries. For example, in Cambodia there is a reduction in the probability of
attending school only if children work more than 30 hours a week, while in Senegal
the probability of attending school begin to decrease if the working load exceeds 15
hours per week. More research is needed to assess what generates such differences
and how they are related to school achievement (see also the following section).
19. The available evidence indicates that child work does negatively affect school
attendance and school survival, and that this negative effect is not limited to
economic activity, but extends also to household chores. The evidence also indicates
that the length of the working day, in economic and non economic activity, is an
essential dimension in assessing the detrimental effect of work on education. But
more research is needed to improve understanding of the determinants of the link
between child labour and school attendance. The relative importance and interplay of
work-related factors (e.g., sector, intensity, setting, work schedule, etc.) and school-
related factors (e.g., duration of the school day, flexibility of the school calendar,
school distance, etc.) remain poorly understood, constituting an obstacle to
identifying forms of work most disruptive of schooling as well as to designing
policies aimed at making schooling and (benign) work more compatible. Much of the
knowledge gap stems from the lack of adequate data, and specifically the lack of
panel and retrospective data. This data gap is beginning to close as new panel surveys
are underway or planned (e.g., Tanzania SIMPOC survey) and retrospective questions
are increasingly included in SIMPOC and other survey questionnaires. In absence of
10 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
appropriate data, information can be gathered by alternative techniques such as the
synthetic indicators presented above.
2.3 Child labour and learning achievement
20. The preceding sections presented evidence underscoring the role of child labour
as a constraint to school attendance. But child labour is not only an obstacle to getting
children into school but also to ensuring they that are able to learn effectively once
there. While the group of working students has been subject of relatively little
research, it stands to reason that children who are exhausted by the demands of work,
or whose work schedule leaves them little time for homework, are less likely to
derive educational benefit from their classroom time than their non-working
counterparts. Working students may also have their interest directed away from
academic pursuits, or be led to place less value on formal learning.
21. For all these reasons, school attendance alone is an incomplete indicator of the
educational impact of child labour. There is also a need to measure the effect of child
labour on actual classroom learning. Indeed, in terms of policy, it is learning
achievement rather than school attendance that is of most relevance: the public or
private return to investment in school is not realised if children fail to learn
effectively while in the classroom. And school attendance and achievement are of
course closely linked. A wide body of evidence indicates that children who perform
poorly in the classroom or are forced to repeat grades are much more likely to leave
the school system prematurely.
22. Grade repetition rates in the countries covered by the UCW Country Statistics
provide indirect evidence of a link between child labour and school performance.
Figure 1, which plots economic activity and primary level repetition rates, shows a
positive correlation between child labour and repetition for boys and girls alike. This
lends support to the conventional wisdom that working children are in a
disadvantaged position in the classroom leaving them more prone to repetition, to the
detriment of both the children concerned and to the internal efficiency of education
systems. But repetition is an imprecise indicator of school performance at best:
promotion criteria can differ widely across countries and indeed even across school
districts and schools within countries. In addition, causality might run in the opposite
direction: Manacorda (2008) using data form Uruguay also shows that school
repetition leads to school drop out, hence potentially increasing the incentives for
children to work.
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
11 JUNE 2008
Figure 9. Grade repetition(a) and child labour, children aged 7-14 years, by sex
% economically active (FEMALE)
% economically active (MALE)
Notes: (a) Primary repetition rate refers to the number of students enrolled in the same grade as in the previous year, expressed
as a percentage of all students enrolled in primary school.
Sources: (1) UNESCO, EFA Global Monitoring Report 2005 (for primary repetition rate); (2) UCW calculation based on
household survey datasets, various countries (for economically-active children)
23. Student test scores are for this reason a much better indicator for investigating
links between child labour and learning achievement. The First Comparative
International Study of Language, Mathematics and Associated Factors (FCIS) and the
Third International Mathematics and Science Study (TIMSS) are among the most
important of the very limited number of surveys containing information on student
test scores matched with student work status. Household survey instruments typically
used for analysing information on child labour, e.g., ILO/IPEC SIMPOC surveys,
World Bank LSMS surveys and UNICEF MICS surveys, are poorly suited for
collecting information on learning achievement, meaning that internationally
comparable data beyond FCIS and TIMSS are limited.
24. Calculations by Gunnarsson et al (2006) based on the FCIS dataset show a strong
and consistent pattern across all the nine countries and the two achievement tests
included in the survey: third- and fourth-graders “almost never�? involved in paid
work outside the family6 outperformed children involved in this form of work “only
some of the time�?, who in turn outperformed children “often�? involved in this work
(Figure 10). The differences in performance by work status were very large. In math,
for example, children almost never working in the nine countries scored 13 percent
higher than children working some of the time, and 22 percent higher than children
working often. Differences in language test scores were similarly large. The authors
show that the strong negative relationship holds up even when possible child-, family-
and school-related confounding factors (i.e., involvement in preschool education,
parental education, home learning environment, class instruction time, classroom
learning environment, compulsory education legislation, etc.) are controlled for and
the possible endogeneity of work is taken into account (Table 3).
6
The authors explain that they do not include work in the home in their empirical analysis because the lack
of meaningful variation in home work meant that the pattern of test scores against home work intensity was
unlikely to be reliable.
12 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
Figure 10. Third- and fourth-grade test scores,(a) by involvement in paid work outside the family, selected
Latin America countries
18 often works(b) sometimes works(c) never works(d)
16
test score
14
12
10
8
language math language math language math language math language math language math language math language math language math language math
Argentina Bolivia Brazil Chile Colombia Dom. Honduras Paraguay Peru All
Rep countries
Notes: (a) Simple mean test score over all children in the child labour group in the county.(b) Child often works outside the home
when not in school. (c) Child sometimes works outside the home when not in school. (d) Child never works outside the home.
Source: Gunnarsson V., Orazem P.F. and Sanchez M.A., Child labour and school achievement in Latin America, 2006.
Table 3. Impact of paid work outside the home on school performance (least squares and
instrumental variables equations on test scores)
Child Labour Exogenous(a) Child Labour Endogenous(b)
Variable
Mathematics Language Mathematics Language
Work outside -1.184(0.051)* -1.087(0.036) * -7.603(1.248) * -3.980(0.484) *
Beta coefficient(c) [-0.159] [-0.204] [-0.408] [-0.295]
Child
Age 0.097(0.027) * 0.045(0.019) * 0.309(0.070) * 0.162(0.024) *
Boy 0.731(0.079) * -0.165(0.056) * 2.480(0.358) * 0.679(0.155) *
No preschool -0.256(0.093) * -0.181(0.066) * -0.376(0.088) * -0.079(0.040) *
Parents/Household
Parent education 0.327(0.036) * 0.280(0.026) * -0.107(0.106) 0.134(0.042) *
Books at home 0.735(0.061) 0.497(0.042) * 0.196(0.100) * 0.258(0.037) *
School
Spanish enrolment/100 -0.046(0.008) * 0.022(0.006) * -0.079(0.010) * 0.007(0.005)
Inadequate classroom
-0.329(0.046) * -0.357(0.031) * 0.073(0.096) -0.140(0.038) *
environment
Math/week (Spanish/week) 0.027(0.017) 0.022(0.006) * -0.073(0.016) * -0.049(0.012) *
Community
Urban 0.730(0.107) * 0.240(0.076) * 1.847(0.225) * 0.794(0.117) *
Rural -0.692(0.122) * -0.893(0.087) * 1.641(0.410) * 0.275(0.202)
Constant 13.778(0.446) * 10.657(0.248) * 14.400(0.453)v 8.045(0.391) *
R2 0.084 0.127 0.063 0.091
N 20699 20290 20699 20290
Notes: (a) Standard errors in parentheses. (b) Bootstrap standard errors in parentheses. (c) The beta coefficients indicates the
number of standard deviation the test score will change from a one standard deviation increase in child labour. Regressions also
include dummy variables controlling for missing values.
* statistically significant at 5% level.
Source: Gunnarsson V., Orazem P.F. and Sanchez M.A., Child labour and school achievement in Latin America, 2006.
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
13 JUNE 2008
25. Orazen and Gunnarsson (2004) report similar findings using data from 10 poorer
countries7 included in the TIMSS survey. Working more than one hour outside the
home lowered seventh- and eighth-grade math scores by at least 10 percent and
science scores by between 11 and 15 percent, again controlling for possible
confounding factors and the endogeneity of work. Outside jobs performed for less
than the one hour per day threshold, however, had only a very small effect on science
scores and no effect on math scores, suggesting that it may not be work per se but
rather the intensity of work that is most damaging to achievement. The study findings
also suggest that work setting is an important factor in how work affects achievement:
1-2 hours per day of home-based work had a much smaller negative impact, lowering
test scores by only 1-2 percent.
26. Orazen and Gunnarsson (2004) note that the adverse effects of child labour on
the seventh- and eighth-graders in the TIMSS sample were much smaller than for the
third- and fourth-graders in the FCIS sample, pointing to the possibility that work is
more harmful to human capital development at younger ages when the building
blocks for more advanced knowledge acquisition are established.
27. Other, country-specific, studies yield similar conclusions to those emerging from
the FCIS and TIMSS survey datasets. ILO/IPEC and UCW (2005) found that while
involvement in economic activity per se did not affect the school performance of
children in Turkey, the intensity of economic activity did significantly influence test
scores. Ten additional hours of work per week, for example, raised the probability of
scoring “poorly�? in mathematics by almost four percentage points. Heady (2000)
found that work involvement had a significant negative effect on reading and
mathematics learning in Ghana, even after controlling for innate ability as measured
by the Raven’s Test.
28. World Bank (2005), using test score data from a nationally representative survey
of primary schools in Cambodia, reported that work had a significant detrimental
effect on learning achievement, particularly among fourth-graders. Estimated models
for literacy and numeracy test scores (including children, parental, household and
schooling characteristics) indicated that working every day before going to school
reduced literacy and numeracy test scores of Cambodian fourth-graders both by about
nine percentage points (Table 5).
Table 4. Estimated impact of children’s work on learning achievement, Cambodia(a)
Grade 4 Grade 6
Literacy Numeracy Literacy Numeracy
No school effects(b) -13.6* -16.2* -8.1* -9.3*
With school effects -9.1* -8.5* -1.3 -1.1
Notes: (a) Reported figures measure the change in percentage points (on a 0 to 100 scale) in test scores resulting from working
everyday before going to school.
* statistically significant at 5% level.
Source: World Bank (2005), Cambodia: Quality Basic Education for All.
29. The research evidence reviewed above clearly confirms the conventional wisdom
that working students face unique learning difficulties in the classroom. But beyond
7
The countries included were: Colombia, Czech Republic, Hungary, Iran, Latvia, Lithuania, Romania,
Russia, Slovak Republic and Thailand.
14 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
this general conclusion, many questions concerning the nature of the relationship
between work involvement and learning achievement remain unanswered.
Knowledge gaps of particular relevance in terms of policy formulation include the
relationship between work intensity and school performance; the extent to which
certain types of children’s productive activity by their nature are more damaging to
school performance than others; the relative importance of work type and work
intensity in influencing learning achievement; the degree to which work is more
damaging to learning at younger ages; and direction of the causal relationship
between work and school performance (i.e., the extent to which a child is a poor
student because s/he works, or alternatively works because s/he is a poor student).
2.4 Child labour and schooling: Student and teacher perceptions
30. A series of five recent ILO-supported school-based surveys in Brazil, Kenya,
Lebanon, Sri Lanka and Turkey capture the perceptions of teachers and of students
themselves regarding how work affects various dimensions of the school experience
(ILO 2003, 2004a, 2004b, 2004c, and 2004d). While the interpretation of the survey
results is subject to a number of caveats,8 the information they provide on student and
teacher perceptions nonetheless adds another layer to the understanding of the
relationship between work and schooling.
31. Survey feedback from students indicated that those working often had greater
difficulties in attending class regularly (Brazil, Sri Lanka, Turkey), arriving at class
on time (Sri Lanka, Turkey) and completing homework (Brazil, Kenya, Turkey), and
that these difficulties generally increased with work intensity. Teachers also saw the
learning of children as being frequently compromised by their involvement in work,
citing differences between working and non-working children in areas such as class
participation, homework completion, extra learning in the home, afterschool study, in
addition to the areas listed by students. In Lebanon, where teachers were asked about
student’s psychological and physical health, they indicated that children working only
in economic activity experienced recurring illness and depression more commonly
than other groups of children.
32. Using pooled data from the Brazil, Turkey and Kenya school-based surveys,
UCW and ILO/IPEC (2005) show that time in economic activity significantly
affected the probability of children reporting missing classes and reporting feeling
tired in class, even when controlling for child and household characteristics. In both
cases, however, the magnitude of the effect was relatively small (Table 5).
8
The survey results should be interpreted with caution for two main reasons: (1) Sample design: schools
and children selected are not always representative at country level, so a selection bias might influence the
results; and (2) Characteristics of working children: Children observed in the surveys worked a rather
limited number of hours in most of the countries, and work tended to be concentrated in a few days a week.
Average working hours are about five per week in Turkey, less than two per hours during weekdays in
Lebanon; almost 80 percent of children work not more than 14 hours per week (including weekends) in
Kenya. Obviously, there is a problem of endogenous truncation in these cases. We cannot observe
children working long hours in school, as they might be out of school having dropped out or not enrolled.
So we might not observe those children for whom the working deeply conflicts with schooling.
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
15 JUNE 2008
Table 5. Determinants of attendance regularity, classroom fatigue and drop-out intentions (pooled data
for Brazil, Kenya and Turkey), marginal effects after probit estimation
Dependant variable
Explanatory variable attendance regularity(1) sleepiness(2)
dy/dx z dy/dx z
Child age 0.159 1.05 0.2989 1.17
Child age squared -0.005 -0.98 -0.0113 -1.18
Female child -0.014 -1.53 -0.0025 -0.12
Mothers education level 0.009 0.56 -0.0106 -0.39
Fathers education 0.020 1.72 -0.0341 -1.06
Weekly hours in economic
activity 0.001* 2.29* 0.0037* 5.10*
country_Brazil 0.178 3.9 0.4442* 10.14*
country_Kenya 0.056 3.16 -0.2249* -9.46*
Weekly hours in household
chores 0.000 0.25 0.0019 1.53
Notes: * Statistically significant at 5% level. (1) Dummy variable taking value of 1 if one or more classes missed and value of 0
otherwise. (2) Dummy variable taking value of 1 if student reported ever feeling sleepy in class and value of 0 otherwise.
Sources: UCW calculations based on data from Brazil school-based survey (Child Labour and Education School Survey, May 2004);
Kenya school-based survey (Child Work, School Attendance and Performance in Kenya, April 2004); and Ankara school-based survey
(Light Work, Academic Performance and School Attendance of Children in Turkey, Ankara, May 2004) as cited in ILO/IPEC and UCW
Project, Impact of Children’s Work on School Attendance and Performance: A Review of School Survey Evidence from Five Countries,
March 2005.
33. The perceptions of both students and teachers in the five countries suggested that
difficulties associated with work were largely limited to children performing
economic activity rather than those performing household chores. Indeed, in many
instances, children performing only household chores seemed to actually face fewer
learning difficulties than children not working at all. Multivariate analysis also
showed no significant relationship between time in household chores and the
likelihood of learning difficulties (Table 5). One possible explanation for these
findings is that children performing household chores are more responsible than their
non-working counterparts and therefore more likely to take their studies seriously.
Another is that the time use of children performing chores is supervised more closely
by the elders in the home, helping to ensure adequate time is allocated to study.
3. EDUCATION PROVISION AS A FACTOR IN CHILD LABOUR:
HOW INADEQUATE SCHOOLING CAN PUSH CHILDREN INTO
WORK
34. In examining the relationship between school non-enrolment and child labour, the
direction of causality is not always apparent. In some cases, children are "pushed"
into work by poor quality, irrelevant or inaccessible schools, while in other cases
children are "pulled" from school and into work by household poverty or other
economic motives. The policy implications of this distinction are clear: where "push"
factors prevail, supply-side policy measures targeting the school system hold
particular promise for reducing child labour; where "pull" factors are relevant,
16 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
demand-side policy measures targeting the household are also needed for an effective
response to child labour.
35. This section focuses on the former case, examining the push factors contributing
to child labour. It reviews empirical evidence relating to the specific links between
education access and quality, on one hand, and child labour, on the other. It also
reviews research gaps that need to be filled to assess the potential of transitional
education and flexible schooling in supporting national efforts towards EFA and child
reduction goals.
3.1 Impact of supply constraints
36. School access has long been recognised as an important element in determining
household choices concerning children’s time use. A wide range of results are
available showing that increased and eased access to school reduces children’s work
in both economic activities and household chores. The availability of a primary
school within the village/community and distance from school in particular have
significant effects on child work reduction. When households within the same village
are compared though, distance to school appears to affect school attendance but not
child labor (Kondylis and Manacorda, 2007, for Tanzania).
37. Even when school access constraints are limited to higher levels of schooling,
they can be part of the reason why children do not attend school at all or drop out of
the primary school. The most commonly used explanation for this finding is that
returns to education tend to be much higher for (lower) secondary than for primary.
Parents have hence an incentive to send their children to primary school rather than to
work if they know that their offspring will also have access to (lower) secondary
education, where the seed of the initial investment in human capital begin to bear
fruit.
Table 6. Effects of travel time to school on children’s activities(a)
Sex/ Work only School only Work and school No Activities
Country
residence Marginal effect z Marginal effect z Marginal effect z Marginal effect z
Yemen Male 0.0003* 8.4* -0.002* -12.6* 0.00 0.3 0.0017* 11.6*
Female 0.0005* 8.0* -0.003* -17* -0.00010* -3.8* 0.0029* 14.8*
Urban 0.0002* 4.1* -0.001* -5.2* 0.00010* 2.3* 0.0012* 4.5*
Rural 0.0007* 11.2* -0.003* -19.5* -0.00010* -3.5* 0.0024* 16.0*
Morocco Urban .0002 0.96 -.003* -2.42* -7.79E-08 -0.22 .003* 2.41*
Rural 0.001* 2.2* -0.002* -1.97* 0.00006 0.62 0.0002 0.26
Notes: * Statistically significant at 5% level.
Source: UCW calculations based on Yemen, NPS 1999; Morocco, LSMS 1998-99
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
17 JUNE 2008
Table 7. Effect of school availability in village/community on children’s activity(a)
Work only School only Work and school No Activities
Country Sex/ School type/level
residence Marginal Marginal Marginal Marginal
z z z z
effect effect effect effect
Yemen Male basic school* -0.0129* -6.4* 0.059* 8.2* -0.005* -2.2* -0.041* -6.3*
koranic school* -0.0016 -1.0 0.005 0.7 -0.002 -0.8 -0.002 -0.3
secondary school -0.0027* -2.0* 0.021* 3.6* 0.001 0.6 -0.019* -3.7*
Female basic school* -0.0058* -2.0* 0.042* 5* 0.001 1.5 -0.037* -4.4*
koranic school* -0.0162* -6.0* 0.054* 7.3* -0.002 -1.8 -0.036* -4.9*
secondary school -0.0170* -7.1* 0.082* 12.6* 0.001 0.8 -0.065* -10.1*
Urban basic school* -0.0020 -0.7 0.015 0.7 -0.001 -0.4 -0.012 -0.6
koranic school* -0.0003 -0.2 -0.018* -2.0* -0.002 -1.3 0.020* 2.3*
secondary school 0.0015 0.8 0.000 0.0 0.002 0.9 -0.003 -0.2
Rural basic school* -0.0258* -6.4* 0.073* 9.1* -0.001 -0.7 -0.046* -5.7*
koranic school* -0.0005 -0.1 0.006 0.7 0.001 0.4 -0.007 -0.7
secondary school -0.0091* -3.1* 0.036* 5.1* 0.001 0.5 -0.028 -4
Morocco Rural Primary school -0.067* -3.59* 0.123* 5.2* 0.003 1.1 -0.059* -2.8*
Cambodia Male lower secondary
school * -0.005* -2.0* 0.025* 2.6* -0.019* -2.0* -0.001 -0.8
Female lower secondary
school * -0.007* -2.1* 0.032* 2.3* -0.024 -1.7 -0.002 -1.3
Urban lower secondary
school * 0.001 0.3 -0.031 -1.5 0.030 1.5 -0.001 -0.3
Rural lower secondary
school * -0.007* -2.1* 0.029* 2.7* -0.021* -2.0* -0.001 -0.9
Notes: * Statistically significant at 5% level.
Source: UCW calculations based on Yemen, NPS 1999; Morocco, LSMS 1998-99; Cambodia, CSES 2003-2204, Cambodia EMIS 2003-2004
38. Table 6 and Table 7 report estimation results from recent UCW research (2003a,
2003b and 2006) and serve to illustrate the effects described above.9 The results
indicate that the availability of a school has well-defined impact on children’s work,
with some variation by sex and residence. The differences by sex appear to be
country specific, while school availability is not surprisingly especially relevant in
rural areas.
39. It should be noted, however, that in several cases increased school availability
seems to increase school attendance by reducing the number of “inactive�? children
(i.e., those neither in school nor working) more than by reducing the number of
working children. This seems to indicate that the decision to send children to work is
not easy to reverse by reducing only the cost of accessing education. The same
comments apply to effect of distance from school: reducing travel time to school does
reduce child work, but appears to generate an increase in school attendance mainly by
reducing the number of inactive children.
40. While the evidence on the effects to school on child work and on the other
children activities is well established in general, more analysis is necessary in order to
understand the reasons for the cross country differences by sex in these effects and,
especially, the reasons for the differentiated effects on the various children’s
activities. This information is very important for policy design and for the selection of
the appropriate policy mix. In particular, given the amount of existing evidence on the
subject, it would be helpful to have it consolidated in a systematic overview aimed at
9
A more detailed review of the literature on school access and child labour is beyond the scope of this
paper.
18 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
understanding the difference in terms of impact by sex and by children’s activities. It
would be important to assess what are the conditions that affect the efficacy of
improved access, especially in relation to the characteristics of the country, of the
school system, etc.
41. Finally, even more important is to begin to move to the analysis in the direction
of comparing the relative efficacy of the different interventions (see also the section
on school quality) like access to school, school quality, income transfers etc., and of
assessing the factors affecting efficacy. As school enrolment increases, the children
(working or inactive) left out become increasingly more difficult to reach and the
identification of effective and cost efficient policy mix gains in importance.
3.2 Impact of school quality
42. The quality of education is currently at the centre of the education reform debate.
It also constitutes an important component of the EFA objectives, and an in-depth
analysis of the link between quality of education and student achievements is
contained in the 2005 EFA Report (UNESCO, 2005). This section discusses the role
of school quality in determining school attendance and involvement in child work.
Evidence of the effects of school quality on school attendance and, especially, on
child labour, is limited. The little existing evidence has been reviewed in the 2005
EFA Report and in a companion paper developed by UCW (2006). This section
therefore reports only on some recent results from UCW research and on remaining
research gaps.
43. Before proceeding, it is worth noting that the issue of education quality is of
particular policy relevance, as underlying it is the question of whether, in order to
promote school enrolment and reduce child labour, providing “quality�? education is
necessary in addition to providing access to education. It is obvious that better quality
education is preferable in general. It also clear that without adequate access, little
benefits can be derived from improving quality. However, in many countries, a
decision must be made on whether, at the margin, to use additional resources for
improving access or quality.
44. There is strong evidence that school quality affects expected returns to education,
thereby also influencing household decisions concerning investment in children’s
human capital. But there is much less defined view on what the constituents of school
quality are, and on how to measure them for practical policy purposes.
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
19 JUNE 2008
Figure 11. School quality indicators and their relationship to student learning
Source: U.S. Department of Education. National Center for Education Statistics. Monitoring School Quality:
An Indicators Report, NCES 2001–030 by Daniel P. Mayer, John E. Mullens, and Mary T. Moore. John Ralph,
Project Officer.Washington, DC: 2000.
45. But translating the complex relationships depicted in Figure 11 into measurable
indicators is not straightforward. Figure 12 illustrates how a set of commonly-used
indicators can be mapped back to this framework. As is easy to see, the proxy
indicators used in empirical studies only partially reflect the main elements of school
quality. In fact, the limited availability of satisfactory information on school quality is
one of the areas that future research should address. Keeping in mind the problem of
data availability, some evidence about the relationship between school quality and
child labour is looked at below. A full review of the literature in this area is beyond
the scope of this issues paper.
Figure 12. Summary of school quality indicators
46. What are the effects of school quality on child labour and school enrolment? A
look at the cross country data for the few available indicators provides a suggestive
but not very precise picture. Figure 13 shows that the pupil-teacher ratio is strongly
and positively correlated with the percentage of working children. As the number of
students per teacher increase, the percentage of working children in each country
20 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
rises. Not surprisingly, however, the variation is very large, as numerous other factors
are also at play in determining children’s work.
Figure 13. Pupil teacher ratio versus working children
Sources: (1) UNESCO 2005 EFA Report (for pupil teacher ratio); (2) UCW calculation based on
household survey, various countries (for working children)
47. The sex of the teacher also has an apparent influence on the level of child labour.
Figure 14 depicts a negative relationship between the percentage of female teachers
and the percentage of both male working children and female working children.
Again, there is a wide range of variation, particularly at low levels of child economic
activity. The link between the sex of the teacher and child labor might be explained at
least in part by research indicating that pupils taught by female teachers perform
better than pupils taught by male teachers (Postlethwaite T. N., 2004), thereby
helping to prevent them from dropping out of school and entering work
48. It is interesting to note that in both Figure 13 and Figure 14 the dispersion around
the regression line tends to decline as the percentage of working children for each
country decrease. This suggests that school quality seems to matter more at relatively
high levels of school attendance (and low levels of child work).
Figure 14. Presence of female teachers versus working children
Sources: (1) UNESCO 2005 EFA Report (for % female teachers); (2) UCW calculation based on household
survey, various countries (for working children)
These descriptive results are suggestive of a potential role of school quality in
addressing child labour. They are, however, far from identifying causal effects and
cannot be used for policy formulation with any confidence. In cross country panel
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
21 JUNE 2008
analysis, which also takes into consideration the role of other variables, the results are
not clearly defined10. In fact, the pupil-teacher ratio seems to be the only indicator,
among the few available, for which a causal link with child work can be
unambiguously established. While cross country evidence is useful, the lack of data
on several relevant determinants of child work and the limited number of
observations makes the use of micro data for a single country more robust.
Table 8. Impact of school quality on HH decisions regarding school and work, marginal effects
after bivariate probit, Yemen(a)
Economic activity Combining economic Neither in economic
School only
Sex and School quality only activity and school activity nor in school
residence Indicators Marginal Marginal Marginal Marginal
z z z z
effects effects effects effects
Male to female teacher ratio 0.0001* 2.6* -0.0008* -9.5* -0.0001* -3.8* 0.0008* 9.8*
Total
Classes to classroom ratio 0.0006 0.4 -0.0130* -2.4* -0.0014 -1.5 0.0138* 2.7*
Male to female teacher ratio 0.00001 0.8 -0.0004* -3.8* -0.0001* -2.4* 0.0005* 4.8*
Male
Classes to classroom ratio -0.0008 -0.6 -0.0148* -2.3* -0.0062* -3.2* 0.0218* 3.8*
Male to female teacher ratio 0.0001* 2.8* -0.0012* -10.0* -0.0001* -3.8* 0.0011* 9.4*
Female
Classes to classroom ratio 0.0051* 2.0* -0.0165* -2.1* 0.0004* 0.5 0.0110 1.4
Male to female teacher ratio 0.00001 1.2 -0.0009* -5.2* 0.00001 -0.9 0.0008* 5.3*
Urban
Classes to classroom ratio 0.0029* 2.2* 0.0019 0.2 0.0037* 2.6* -0.0085 -0.9
Male to female teacher ratio 0.0001* 2.1* -0.0008* -8.4* -0.0001* -4.3* 0.0008* 8.5*
Rural
Classes to classroom ratio -0.0016 -0.7 -0.0123* -2.1* -0.0033* -2.7* 0.0171* 3.0*
Notes: * Statistically significant at 5 % level.
Source: UCW calculations based on Yemen National Poverty Survey, 1999 and Yemen School-based survey, 2000
10
UCW (2006), Does school quality matter for out of school children?, UCW forthcoming
working paper.
22 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
Table 9. Impact of school quality on HH decisions regarding school and work, marginal effects
after bivariate probit, Cambodia
Economic activity Combining economic Neither in economic
School only
Sex and School quality only activity and school activity nor in school
residence Indicators Marginal Marginal Marginal Marginal
z z z z
effects effects effects effects
Pupil teacher ratio 0.0001* 2.1* -0.0007* -3.4* 0.0006* 2.9* 0.00001 0.4
% of primary schools with parent
Total 0.0060 1.0 -0.0140 -0.6 0.0056 0.3 0.00230 0.8
association
% of primary school with libraries -0.0161* -4.2* 0.0311* 2.2* -0.0085 -0.6 -0.00651* -3.5*
Pupil teacher ratio 0.0001* 2.0* -0.0008* -2.8* 0.0006* 2.3* 0.00002 0.8
% of primary schools with parent
Male 0.0035 0.5 -0.0091 -0.3 0.0043 0.1 0.00133 0.4
association
% of primary school with libraries -0.0055 -1.2 0.0216 1.1 -0.0143 -0.7 -0.00178 -0.8
Pupil teacher ratio 0.0001 0.9 -0.0006 -2.0 0.0005 1.8 -0.00001 -0.2
% of primary schools with parent
Female 0.0091 1.0 -0.0181 -0.6 0.0053 0.2 0.00366 0.7
association
% of primary school with libraries -0.0278* -4.5* 0.0412* 2.0* -0.0011 -0.1 -0.01230* -3.7*
Pupil teacher ratio 0.00001 -0.1 -0.0031* -5.0* 0.0033 5.4 -0.0002* -2.0*
% of primary schools with parent
Urban 0.0052 0.9 -0.0121 -0.3 0.0026 0.1 0.0043 0.9
association
% of primary school with libraries -0.0116* -2.5* 0.0910* 3.1* -0.0735* -2.6* -0.0059 -1.5
Pupil teacher ratio 0.0001* 2.0* -0.0004* -2.0* 0.0003 1.4 0.0000 1.0
% of primary schools with parent
Rural 0.0043 0.5 -0.0274 -1.1 0.0229 0.9 0.0002 0.1
association
% of primary school woth libraries -0.0178* -3.6* 0.0098 0.6 0.0153 1.0 -0.0073* -3.5*
Notes: * Statistically significant at 5% level.
Source: UCW calculations based on Cambodia CSES 2003-04 and Cambodia EMIS 2003-04
49. Table 8 and Table 9 present the results of a recent UCW study on school quality
and child labour based on microdata from Yemen and Cambodia.11 In Yemen, both
the male to female teacher ratio and the classes to classroom ratio appear significant
in determining the time use patterns of children. Both quality indicators appear to be
relevant in determining school attendance in particular. In Cambodia, among the
several indicators of school quality which were available, only two appear to be
significant: the pupil to teacher ratio and the percentage of primary schools with a
library. In sum, the evidence from Yemen and Cambodia indicates that quality of
education does indeed matter for child labour and school attendance. However, the
effects of school quality appears to be relatively “small�?, with large improvement in
school quality potentially leading to only moderate reductions in child work and
increases in school attendance.
50. Several qualifications and further research and analysis is necessary before we
can go beyond the general statement that school quality matters for child labour. First,
as mentioned above, more and better information is needed on school quality
indicators. A systematic effort at the international level should possibly be developed
to design and collect such indicators. More analysis is also needed comparing the
effects of school access with those of school quality to be able to formulate
11
UCW (2006), Does school quality matter for out of school children?, UCW forthcoming
working paper.
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
23 JUNE 2008
recommendations on the appropriate policy mix between the two. Finally, if the
results presented above are taken at face value, there is a need for further
investigation into the effects of school quality in retaining children in school and in
avoiding early drop out. (Preliminary fragmentary evidence seems in fact to indicate
that school quality is more effective in retaining children to school rather in attracting
them to it for the first time).
3.3 Impact of special transitional education and flexible schooling
programme on child labour
51. The previous sections highlighted the important role of school access and school
quality in determining school attendance and children’s involvement in work. Special
transitional education (TE)12 and flexible schooling (FS)13 programmes constitute a
third important supply-side element influencing child labour and school attendance
outcomes in many national contexts. Such programs can take numerous forms, with
some serving as a bridge to entry or re-entry into the formal education system and
others serving as sources of remedial support or special needs education within the
formal system. Still others are designed to make the schooling system for
accommodating of children’s work exigencies and schedules. They are all based on
the premise that child labourers need special support in order to ensure that, once in
school, they remain there and are able to learn effectively.
52. Information on transitional education and flexible schooling programmes
unfortunately remains very limited. Little is known about the difference they are
making in reducing the exclusion from education of working children, about which
and how many child labourers are being reached, and with what impact. This limits
the lessons that current TE and FS efforts offer in terms of which policy approaches
are most effective or are best candidates for broad-scale replication. This section
briefly examines some of the research priorities and information gaps that need to be
filled in order to assess the potential of transitional education and flexible schooling
in supporting national efforts towards EFA and child labour reduction.
12
Transitional education programs are aimed at smoothing the transition of child labourers and
other vulnerable children into the formal school system. They are based on the premise that
child labourers are often difficult to insert directly (back) into the formal education system
because of their age, different life experiences and lack of familiarity with the school
environment. International programming experience points to two main policy options for
easing the transition of child labourers back into the formal school system (a) remedial
education, providing returning children and child labourers with special remedial support within
the regular classroom context; and (b) "bridging" education, involving intensive compensatory
or "catch-up" courses designed to raise academic proficiency, offered in either non-formal
community schools or in school facilities prior to, during or after regular classes.
13
Flexible schooling programs are targeted specifically to working children, and are designed
to make school more accommodating of the exigencies of work. These programs are not
therefore aimed primarily at reducing child work per se, but rather at increasing school
attendance and reducing drop-out among child labourers. Flexible schooling programs are
designed to balance the learning and earning needs of families and children by facilitating fluid
work/study schedules. They encompass formal, non-formal and work-based learning
arrangements, and, ideally, help children who need or want to work to move back and forth
between systems considered to be equally valid, rather than one the "poor cousin" of the other.
International programming experience points to three main policy options for helping children to
combine work and school more easily: (a) flexible delivery modes, designed to make schooling
more accommodative of children’s work schedules; (b) adaptive curricula, designed to make
course contents more relevant to the lives of working children; and (c) substitute non-formal
education, designed to impart basic literacy, numeracy and life skills at times not in conflict with
work.
24 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
53. A systematic mapping of the wide variety of policy and programmes experiences
in both transitional education and flexible schooling is needed as a first step towards
the identification of good practices. These programmes have taken a wide variety of
forms, either because of trying to address different needs or because of using different
approaches to address the same need. There is now a substantial body of programme
experience that could be used to compile a set of good practices and/or guidelines for
action. Such a mapping would need to bring together information on a wide variety of
qualitative and quantitative variables.14 The mapping of TE and FS programmes
would also need to aim at providing an assessment of the relative dimension of the
programmes, in order to obtain a picture not only of the instruments used, but also of
the distribution of resources invested. It would useful to compare the amount of
resources invested in TE and FS programmes with those utilised in other strategies to
cope with the needs of working children.
54. While piloting should ideally be short-term and catalytic, testing models which
can then be mainstreamed into national policies and replicated on a broader scale, this
mainstreaming and replication does not appear to be occurring in the case of many TE
and FS pilot programmes for working children. Why are these programs typically of
limited coverage? Answering this question will be critical to assessing the potential of
these programmes as vehicles for addressing the education rights of out-of-school
working children. The following areas of research seem of particular relevance in this
context: identification of the approaches suitable for scaling up, also looking at
international experience on the few large scale programs; the challenges of scaling up
(bottlenecks, institutional constraints, political constraints, the need for community
mobilization, the need for systematic evaluation of pilot experience to guide scaling
up, etc); links between non formal education, vocational training and labor market
outcomes; and how to address the issue of the links between the formal and non
formal education systems when the latter is of large scale.
55. Evaluations of TE and FS initiatives are relatively scarce, and more attention is
needed to piloting methodologically-sound evaluation studies. There are two main
directions that the researcher could follow: a) look for existing data that, through
matching with programme information, would allow reliable estimate; b) try to
address the issue at the source by designing and implementing the necessary data
collection jointly with the implementation of a programme. While care is necessary in
designing such data collection, the costs of the research are not necessarily large.
Treatment and control groups can be limited in size, especially if the program is also
of limited scope (e.g. limited coverage area, small target group, etc.), but still convey
very useful quantitative information on the impact of the program. Evaluation criteria
should include the following: programme sustainability, with special attention to the
issue of integration with the main education system or through other institutional
channels; programme replicability, i.e., the extent to which the approach followed is
dependent on local factors, thereby limiting its applicability to other contexts;
learning outcomes, i.e., student achievement tests including changes, positive or
negative, in the outcomes of other, non-beneficiary students; and school survival, i.e.,
the extent to which TE and FS programmes succeed in retaining child laborers in the
education system.
14
Including, for example, geographical distribution; classification of programme by type of
provider (e.g., community/faith-based, private, public or mixed); pedagogical approach;
geographic coverage; beneficiary population; number of teachers/instructors; per unit costs;
services provided (e.g., accelerated "catch-up" learning, specific skills training, basic literacy
and numeracy, etc.); physical facilities and instructional materials; management structure; and
stakeholder involvement.
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
25 JUNE 2008
56. Many non-formal TE initiatives have been criticized for creating a second,
inferior, education track for working children, and not acting as bridges to (re)entry
into the formal system. While stand-alone non-formal education programs may be
appropriate for older, long-term drop-outs, there seems to be some consensus that the
overarching emphasis of transitional education should be equipping children to enter
and succeed in regular schooling. A critical review of the work and experiences that
have led to this consensus and, eventually, a critical reappraisal of its main conclusion
would be of interest. This could possibly lead to an assessment of the role of
nonformal vis-à-vis the formal education system and to a clear identification of the
relative roles of the two systems. It would be of interest to identify the situations in
which experience and theory shows that the best interest of the children and youth is
achieved without mainstreaming non-formal efforts in the formal education system
(e.g. older children, children that have suffered severe physical or psychological
health damages, children that need reintegration also from traumatic experiences like
child soldiers).
4. CONCLUSION
57. The preceding sections provided a brief overview of research evidence
concerning the interplay between education and child labour. It also identified areas
where further research is needed to help guide policy towards the related goals of
EFA and child labour elimination.
58. Evidence reviewed of the impact of work on school attendance and performance
underscored the constraint that child labour poses to achieving Education For All.
This evidence largely confirmed the conventional wisdom that child labour harms
children's ability to enter and survive in the school system, and makes it more
difficult for children to derive educational benefit from schooling once in the system.
The evidence also suggested that these negative effects are not limited to economic
activity but also extend to household chores, and that the intensity of work (in
economic activity or household chores) is a particularly important in determining the
impact of work on schooling.
59. But beyond these general conclusions, many questions concerning the nature of
the relationship between work involvement and education remain unanswered in the
research literature. We need first of all more knowledge about the effect of work on
school entry and survival. There is a specific need to open the “black box�? of child
work, and look more closely at the effect of different forms of work on enrolling and
staying in school. For example, a lot can be potentially learned by looking at the
factors underlying the large cross-country variation in terms of the ability of child
labourers to combine school and work, and in particular by looking at the extent to
which these differences are institutional or policy related. More research is also
needed on learning achievement, and on how both school and work conditions affect
the ability of working student to perform in the classroom.
60. Research questions of particular relevance for identifying forms of work most
disruptive of schooling as well as for designing policies aimed at making schooling
and (benign) work more compatible include the following:
x work setting and schooling: the degree to which work performed within a family
setting is less disruptive to schooling than work performed outside the family
environment;
26 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
x work intensity and schooling: the degree to which schooling is only compromised
by work performed beyond a particular daily or weekly hours threshold (i.e.,
whether it is work per se or only work performing intensively that is detrimental
to schooling);
x work type and schooling: the extent to which certain types of children’s
productive activity by their nature are more damaging to school attendance and
performance than others;
x interplay among work characteristics: the relative importance of different work
characteristics (setting, intensity, type, etc.) in influencing schooling attendance
and performance, and the interplay among work characteristics;
x child age, work and schooling: the degree to which work is more damaging to
learning at younger ages;
x innate ability, work and schooling: the extent to which a child is a poor student
because s/he works, or alternatively works because s/he is a poor student; and
x cross-country variation in terms of how work effects schooling: reasons for the
large differences across countries in terms of the ability of working children to
attend and perform in school.
61. Evidence reviewed in the preceding sections concerning the link between
education provision and child labour pointed to the important role of inadequate
schooling in keeping children out of the classroom and into work. This evidence
indicated that both the school quality and school access can play an important role in
household decisions concerning whether children study or work. But again,
considerable further research is necessary before it is possible to go beyond the
general statement that school access and school quality matter for child labour. Better
information is needed regarding how access and quality (and their interaction) affect
household decisions in order to identify the best mix between quality and access
policy measures. It is also necessary to assess whether the main effect of school
quality is in improved retention or higher rates of entry. The analysis of the effects of
school quality requires better data reflecting the main elements of school quality.
62. Areas of further research of particular relevance to identifying supply-side
policies for reducing child labour and raising school attendance include the following:
x factors affecting the efficacy of improved schooling access: reasons for the large
cross-country variations in terms of how improved school access affects school
attendance and child labour;
x measuring school quality: developing proxy indicators reflecting the main
elements of school quality, and using these indicators to provide a more complete
picture of links between school quality and child labour;
x impact of access and quality interventions: assessing the relative efficacy of
access and quality interventions in order to formulate recommendations on the
appropriate policy mix between access and quality;
x school quality and retention: the effects of school quality in retaining children in
school and in avoiding drop out, in view of preliminary fragmentary evidence
suggesting that school quality might be more relevant in terms of retaining
children in school rather in attracting them to it for the first time; and
x relative importance of “push�? and “pull�? factors: the degree to which children
are "pushed" into work by poor quality, irrelevant or inaccessible schools, or,
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
27 JUNE 2008
alternatively, children are "pulled" from school and into work by household
poverty or other economic motives.
63. The paper dealt with special transitional education (TE) and flexible schooling
(FS) programmes as other important supply-side elements influencing child labour
and school attendance outcomes in many national contexts. Information on
transitional education and flexible schooling programmes unfortunately remains
scarce, limiting the lessons that current TE and FS efforts offer in terms of which
policy approaches are most effective or are best candidates for broad-scale
replication. Further research is needed, inter alia, about the difference these
programmes are making in reducing the exclusion from education of working
children, about which and how many child labourers are being reached, and about
why these programmes have for the most part been unable to expand to scale. The
role of non-formal education strategies generally in supporting national efforts
towards EFA and child labour reduction as needs to be assessed.
28 CHILD LABOUR AND EDUCATION FOR ALL: AN ISSUE PAPER
REFERENCES
Cigno, A. and F.C. Rosati (2005): “The economics of child labour�?, Oxford University
Press
Guarcello, L., Lyon S. and F.C. Rosati (2006): “Does School Quality matter for Working
Children?�? UCW forthcoming working paper.
Gunnarsson V. Orazem P. and Sanchez M. (2006): “Child labour and school
achievement in Latin America�?. World Bank Economic Review , Vol. 20, 1, pp. 31-54,
January.
Heady C. (2000): “What is the effect of child labour on learning achievement? Evidence
from Ghana.�? Innocenti Working Papers, no. 79.
Kondylis F. and M. Manacorda (2007), 'School Proximity and Child Labor: Evidence
from Rural Tanzania', mimeo, London School of Economics, 2007.
ILO/IPEC (2003): “Child Work, School Attendance and Performance: Case Study, Colombo.�?
Gunawardena et al, September 2003
ILO/IPEC (2004a): “Brazil Child Labour and Education School Survey�?, unpublished.
ILO/IPEC (2004b): “Child Work, School Attendance and Performance in Kenya�?,
unpublished
ILO/IPEC (2004c): “Impact of Child Work on School Attendance and Performance in
Lebanon.�? Consultation and Research Institute, unpublished.
ILO/IPEC (2004d): “Light Work, Academic Performance and School Attendance of
Children in Turkey�?, unpublished, Ankara, May 2004.
ILO/IPEC and UCW Project (2005): “Impact of Children’s Work on School
Attendance and Performance: A Review of School Survey Evidence from Five
Countries.�?
Manacorda, M. (2006), "Child Labor and the Labor Supply of Other Household
Members: Evidence from 1920 America", American Economic Review, December 2006,
96(5) 1788-1800.
Manacorda, M. (2008), The Cost of Grade Retention', mimeo, London School of
Economics, 2008.
Orazem P. and Gunnarsson V. (2004): “Child Labour, School Attendance and
Performance: A Review.�? Iowa State University, Department of Economics Working Papers
Series # 04001.
UCW WORKING PAPER SERIES, NOVEMBER 2006 – REVISED
29 JUNE 2008
Postlethwaite T. N. (2004): “What do International Assessment Studies tell us about the
Quality of School Systems? Background paper for EFA Global Monitoring Report
2005.�?
UCW (2003a): “Understanding Children’s work in Yemen.�? Inter-agency report to the
Government of Yemen, Sana’a, March 2003.
UCW (2003b): “Understanding Children’s work in Morocco.�? Inter-agency report to the
Government of Morocco, Rabat, March 2003.
UCW (2005a): “Towards statistical standards for children’s non economic work: A
discussion based on household survey data.�?
UCW (2005b): “Children’s work in Cambodia: a challenge for growth and poverty
reduction.�? Inter-agency report to the Government of Cambodia, Phnom Penh, June 2005.
UNESCO (2005): “EFA Global Monitoring Report – The quality imperative.�?
UNESCO, Institute for Statistics, 2005
U.S. Department of Education (2000): “Monitoring School Quality: An Indicators
Report.�? National Center for Education Statistics. Daniel P. Mayer, John E. Mullens, and
Mary T. Moore. Washington, DC, 2000.
World Bank (2005): “ Cambodia Education Quality For All.�? World Bank, Washington
DC, January 2005