Schooling, Job prospect and Child labour in a Developing Economy

Diganta Mukherjee
Economic Research unit, Indian Statistical Institute, Kolkata
and
Uday Bhanu Sinha
Department of Economics, Delhi School of Economics, Delhi

Abstract: Incidence of child labour is widespread in the developing countries. The two broad
reasons for this are poverty and low return to basic education in the developing world. The
theoretical literature is mainly focused on the question of poverty and usually view child labour
through the lens of credit market imperfection. We develop a model of child labour vs. education
in which uncertainty in job prospect, paternal attitude towards education and the existence of
informal sector are crucial dimensions. Our model predicts that a child go though part schooling
and then join the informal sector. We examine the effectiveness of standard policies like trade
sanctions compulsory schooling or financial incentives in reducing the incidence of child labour.

JEL Classification: J20, D60, 010

Key words: Child labour, Schooling, Unemployment, Informal sector.

Address for Correspondence: Diganta Mukherjee, Economic Research unit, Indian Statistical
Institute, 203 B.T. Road, Kolkata 700108, diganta@isical.ac.in
1. Introduction

Historically child labour existed in all countries. In today’s world the incidence of child labour is
widespread in the developing countries. The International Labour Organization (ILO, 1996)
estimates that between 100 million and 200 million people under age 15 work worldwide, 95% of
them being in developing countries. More importantly, about 110 million school age children
worldwide (around 20% of the corresponding age group) receive no primary education. Again a
vast majority of these are concentrated in low income countries. It is reasonable to presume that
the existence of child labour is intimately linked with poverty.

Empirical studies point to two broad reasons for why children work. First, because of abject
poverty, many households find it necessary to send their children to work, and thereby, stop them
from receiving education (see Bhalotra, 1999; Ray, 1999). Thus, poverty is the greatest single
force, which creates the flow of children into the workplace. Second, there is very strong
empirical evidence that the rate of return to basic, primary-level education as provided to poor
children in many developing countries is very low, not only because of high rates of time
discount but also due to the poor quality of education.1 In such situations, poor families sending
any child to school must be a testimony to the fact that those poor families gain some satisfaction
purely from educating their children.2 In addition, since decisions on child labour are inherently
intertemporal ones, credit markets play a significant role in influencing them. Even if returns to
education were significantly high, sending a child to school instead of to work entails a sacrifice
of current income in favour of future earnings. If the family has access to credit at ‘reasonable’
terms, then it may not have to forego present consumption significantly, and the decision to send
a child to school may not be a painful one. The working of credit markets should therefore be an
important element in the analysis of child labour.

There is a surge in theoretical literature in the recent past, which have dealt with different reasons
for the existence and persistence of child labour and usefulness of different policy measures.
Apart from the focus on poverty, credit market imperfection and bad quality of education in
schools, the literature has also focused on the low returns to schooling as a reason for the parents

1
Ray (2000) has found a significant negative relationship between the supply of child labour and quality of
schooling in Ghana.
2
A recent survey conducted in Indian villages found that economic motives are not the
only reasons why
poor families want their children to go to school (see The Probe team, 1999, Chaps. 2 and
3).

1
not to send their children to school. However, none of the theoretical studies has explicitly
incorporated the second characteristic feature of a developing country apart from poverty. This is
unemployment (or underemployment) and also the existence of the large informal sector
providing employment to semi-educated or a part of the fully educated labour force. We develop
a model for the existence of child labour even though the parents intrinsically value education by
introducing the problem of educated underemployment and the informal sector absorbing the pool
of labour force, which are partly educated. Given the facts, as outlined above, we develop an
intertemporal model of child labour versus child education in which the uncertainty in job
prospects after schooling, subjective parental attitudes towards education, and the existence of
informal sector all play a role.

Several explanations have been proposed for the use and prevalence of child labor in developing
countries. In an interesting paper Basu and Van (1998) generate child labour in a general
equilibrium setting in which parents dislike child labour and withdraw their children from the
labour force once the adult minimum wage reaches a critical level. This model establishes the
possibility of two stable equilibria, a low wage equilibrium characterized by child labour and a
high wage equilibrium in which children are all attending school. Genicot (1998) builds a theory
of child labor based on efficiency wages. In Baland and Robinson (2000) inefficiently high
degree of child labour arises due to the commitment problem between parents and the children in
the presence of credit constraints. That the incidence of child labour can be inefficiently high due
to credit constraints is also studied by Ranjan (2001). In a dynamic context, a fertility based
argument for the persistence of child labour in poorer societies can be found in Chakraborty and
Das (2005). Thus, the devloping countries may be stuck in a low wage child labour trap (Basu
(1999)). Basu (1999) provides a survey of possible other causes, including social norms, social
stigma etc.

Basu (2000) emphasizes that the labour market failures can contribute to child labour when it is
accompanied by adult unemployment. Our focus is on another aspect of labour market failure.
Parents may withdraw their children from school before completion when the prospect of getting
a job after school does not look good. In other words, the problem of educated unemployment in
the formal sector may force the parents to withdraw their children from school and put them to
work in the informal sector. Working in the informal sectors prepare the children, through
training and consequent learning by doing, to become an adult worker in the informal sector. The

2
informal sector in the developing countries accounts for a significant portion of their total
national income and employment. It is this sector which escapes all the prohibiting laws on child
labour whereas the formal sector can rarely bypass the laws to employ child labour. Thus, we find
that a portion of the children drops out of the school before completion as the probability of
getting a formal sector job is not high enough. This happens even when the parents intrinsically
value education and also the informal wage is a strictly positive function of the time spent in
school.3 The level of income of the parents plays a crucial role in the decision for school
completion. We do not consider the possibility of credit in our model. However, the existence of
credit may not solve the problem of child labour as the job prospect may be so low that it is not
worth while to take that credit.4

Our model predicts that in the childhood, a child would be in school for some time and
then drop out from school and join the informal sector for work. Thus, a part schooling
and part child labour is an alternative to full schooling. Of course, some schooling helps
in increasing the wage rate the candidate faces in the labour market. There are several
empirical studies that support this view. Heady (2000) notes that Participation in school
increases from the age group 7 - 9 to the age group 10 - 14 and then decreases for aged 15 – 19.
A survey for Bangladesh on the slum population of Dhaka city was conducted from 1995 to 1997
as part of the Urban Livelihoods Study (a joint research project of Proshika, the London School
of Hygiene and Tropical Medicine, and the University of Bath). This revealed that parents did
attach considerable value to education but often found that school fees were too expensive or felt
that work experience was more useful to future employment prospects than education. To some

3
In an empirical context, Smith and Welch (1989) document a sharp positive relationship between years of
schooling (even for incomplete schooling, like 0-4 years or 5-7 years of schooling) and wage for Black and
White American Males during the period 1940 - 1980. The structure of the labour market for Black
Americans would be similar to the informal markets in countries like India at the present time. This
provides indirect evidence for wage premium for (partial) education in the developing countries at the
present time.
4
It should be noted here that the issue under discussion here is not the "child labour due to poverty" or a
situation of "distress child labour" which along with other related issues has been beautifully researched in
the above mentioned papers. Neither are we considering a situation where the employment of very young
children is an alarming problem; the younger the child, the more vulnerable he or she is to physical,
chemical and other hazards at the workplace and, of course, to the economic exploitation of his or her
labour (Rogers and Swinnerton, 2002, Dessey and Pallage, 2003, 2005 and Raju, 2005). Also the issue of
school quality is not considered in this paper.

3
extent, these two factors were combined, with parents saying that education would help their
child get a good job if they stayed at school for many years, but they could not afford that.
Grootaert and Patrinos (1998) report the patterns of child work and school attendance for Côte
d’Ivoire, Colombia, urban Bolivia and Philippines. In urban Bolivia, full-time school attendance
is over 90 per cent until the age of 13 years. After that, child work becomes significant, with
somewhat more children working full-time than combining work with school. All these studies
provide a foundation for our theoretical model.

In section 2 we explain when it would be economically rational to leave school early and thus
forego formal employment opportunities in the future. We then look at some standard policy
measures in the context of trade sanction or schooling reforms that are typically discussed in the
context of child labour in section 3. Finally section 4 concludes.

2. The Model

We consider a family consisting of one parent and one child. Parent is altruistic (cares for child's
future income) and takes all the decisions. In each family income of parent is y where y ∈ [0, y ] .

Parent has a preference for the education of her child, which is represented by a parameter λ and
λ ∈ [−1,1] . A positive value of λ means parent intrinsically values schooling, a negative value
implies dislike.

We consider the schooling choice of a child as a continuous variable s . A child can be in school
for a part of its childhood or full. The total time of the childhood is normalized to unity and the
extent of education for a child is part of the childhood spent in school. Thus, s ∈ [0,1] . We

assume the parent's valuation of this schooling is: V (λ , s ) , which is increasing concave for

λ > 0 and decreasing convex otherwise.

We now specify the markets for both child and adult labour. There are two sectors in the
economy: formal and informal sectors. In the formal sector only adult with full education can
work, but the children cannot get formal employment due to the legal restriction on child labour
and also may be due to the full school required for formal sector jobs. The wage rate in the formal
sector is fixed may be due to ‘efficiency’ wage reason. However, not every adult worker with full

4
education can find employment in the formal sector. This is due to the fact that the size of the
formal sector is not large enough to employ all the educated adult workers. Only a fraction of
them is employed and the remaining adult workers go back to informal sector for employment.
Thus, any adult worker with full education has only a probability f of finding a formal sector
employment. The informal sector is capable of employing the remaining adult workers and the
child workers whoever is willing. It offers a wage which is dependent on the level of education of
any worker whether adult or child. The wage rate of a child is a fraction of that of an adult worker
with the same level of education. Thus, we have
1. Formal sector : adult wage = w F . Probability of getting a job here for an adult is f ∈ (0,1) if

s = 1 and 0 other wise. Children cannot get formal employment.

2. Informal sector: wage is dependent on the level of education /skill. Thus, informal sector
wage for an adult worker with education s is w( s ) = (1 − s ) wU + sw T , where wU , wT are the
wages of an adult worker with education zero or full respectively. We assume that a child
with education s receives α fraction of the adult wage w(s ) where 0<α<1. Therefore, the

total earning of a child who receives education for s fraction of its childhood would be
(1 − s )αw( s ) .

Parent’s utility function is U (c1 , s, y 2 ; λ ) = u (c1 ) + V (λ , s ) + β u ( y 2 ) , where c1 is aggregate

current consumption and y 2 is the child's future earning. β ∈ (0,1) is the altruism parameter.

u (c1 ) is a concave utility function.

The parameters of the system are: (i) wages w F > wT > wU , (ii) α , β and f .

The individual specific parameters are: λ and y .

The choice variable is s .

Since there does not exist any credit market opportunities (neither borrowing nor lending is
possible), the individual household budget constraints are defined as below. The first period
consumption c1 is the sum of parent’s income and the child’s income. The second period income
of the household depends on the employment opportunity of the adult and the level of education.
Thus, we write,

5
c1 = y + (1 − s )αw( s ) = y + α (1 − s )[(1 − s ) wU + sw T ] and

 w F w.p. f 
 for s = 1
y 2 =  wT w.p. (1 - f ) .
(1 − s ) wU + sw T for s ∈ [0,1)


Assumption 1: fw F = wT .

Assumption 1 implies that the expected earning from the formal sector employment is the same
as the wage rate in the informal sector when an adult has complete schooling in her childhood.
Thus in the second period every fully educated adult worker would try their luck for the formal
sector employment, those who fail would receive the wage rate wT in the informal sector for sure.
Therefore, the expected utility for the household choosing full schooling for a child would be
EU (c1 ,1, y 2 ; λ ) = u ( y ) + V (λ ,1) + β [ fu ( w F ) + (1 − f )u ( wT )] (1)

This is fixed given λ and y .

On the other hand, the households choosing partial schooling for their children would receive the
utility (for s ∈ [0,1) ),

U (c1 , s, y; λ ) = u (c1 ) + β u ( y 2 ) + V (λ , s )
(2)
= u ( y + α (1 − s )[(1 − s ) wU + swT ]) + βu ((1 − s ) wU + swT )] + V (λ , s )

∂U ∂V
Let Z ( s; y, λ ) = = u ' (.)[α (1 − 2 s ) wT − 2α (1 − s ) wU ] + βu ' (.)(wT − wU ) + for any
∂s ∂s
s ∈ [0,1). Then the FOC for an optimum s , say s * , is given by
Z ( s*; y, λ ) = 0 . (3)

2.1 Individuals who like schooling

We first consider the case when the parents intrinsically value some education for their children.
This is the case for the households with λ > 0 .

Note that, at the optimal choice s * , the term c* = (1 − 2 s*)wT − 2(1 − s*)wU must be negative
for the FOC to be satisfied (since all other terms are positive).

6
Let us consider the corner points first. We suppress the arguments y and λ of Z (.) , when they
are kept fixed, for convenience. The following two assumptions are the sufficient conditions for
the existence of an interior choice of s.
∂V
At s = 0 , Z (0) = u ' ( y + αwU )α ( wT − 2 wU ) + βu ' ( wU )( wT − wU ) + . (4)
∂s s =0

∂V
Note that u ′(.) > 0 and wT > wU and also > 0 by assumption.
∂s s =0

Assumption 2: wT − wU > wU .

Thus, Assumption 2 implies that Z (0) > 0 , which in turn implies that the equilibrium schooling

choice s* > 0.
∂V
At s = 1 , Z (1) = u ' ( y )α (− wU ) + βu ' ( wU )( wT − wU ) + (5)
∂s s =1

Assumption 3: wT − wU < αwT

Then we have Z (1) < 0 , ⇒ s* < 1. Then, an interior solution to (3) is ensured due to continuity

of Z (s ) . It is also easy to see from (3) that (1 − 2 s*)wT − 2(1 − s*)wU < 0 for equilibrium.

F z
O z′
C

0 s =1 s
s=0 ŝ s* z'

z
Figure 1

The graph of Z (s ) for different values of s is given by the curve zz in fig. 1. s * is the point

where it crosses the s-axis. Here the parent is altruistic ( β > 0 ), he/she values education

7
intrinsically ( V (λ , s ) increases in s for λ > 0 ) and the formal sector employment opportunity

( w F w.p. f and wT w.p. 1 − f ) strictly dominates the best opportunity in the informal sector

( wT ). In spite of these features we could have a situation like the one depicted in fig. 1 for certain
parametric configurations. This is because we have allowed for skill premium in wage in the
informal market, which adds to the attraction of the extra income available as a child labour (after
some amount of schooling). Of course, if β is too high then such a choice would not be observed
but that is only natural. So, for the rest of the analysis we will assume that the parent is not too
altruistic in order to highlight the implications of this model.

To establish this point, we provide the following numerical illustration where one can have a
situation where (2) may be inverted U shaped in s . Now the maximum utility obtained from
partial schooling is greater than the utility obtained from full schooling, then we have a choice of
schooling 0 < s* < 1 .

Example 1: Consider wU = 1, wT = 2.2, w F = 11, f = 0.2, α = 0.6, β = 0.4, y = 1. Take

u (x) = log( x), V(λ , s) = s (with λ = 1). Then routine calculation shows that for s * = 0.8,
U (c1 ,0.8,1;1) > U (1,1,1;1) .

This illustrates that even for a parent with highest valuation for education, it may be rational to
choose s * < 1 if he/she is not too altruistic and the probability of getting a formal sector
employment is low.

Routine comparative static exercises show that if we now vary y then with an increase in y ,

u ' (.) in the first term of Z (s*) decreases, and hence the value of Z (s*) increases (as u ' (.) has
a negative coefficient, see (3)). Thus to restore the equality Z ( s*) = 0 , the value of s * must

rise. However, at the same time Z (0) decreases (again due to a decrease in u ' (.) but now it has a

positive coefficient, see (4)) and Z (1) increases (similar argument for (5)). Thus, overall the
curve zz becomes flatter (say becomes z'z') when y increases and it crosses the s-axis at

progressively higher values of s * .

8
 w T − 2 wU
Also note that at s = = sˆ (say, < ½), Z (s ) is independent of y , as the first term in
2( wT − wU )
(3), which involves y , vanishes. Thus ŝ acts as a fixed point for Z (s ) as we vary y . This

implies that as y increases Z (s ) becomes flatter, pivoting around the value Z ( ŝ ) at ŝ .

Similar routine calculations show that with an increase in λ , Z (s ) increases, zz curve moves to

the right and hence s * increases in λ .

We summarise the above discussions in the following proposition.

Proposition 1: Given assumptions 1 - 3, the FOC and comparative static exercises show
(a) From the FOC, s * can be < 1 even for λ > 0 individuals.
(b) s * increases in λ for all values of λ > 0 .
(c) For λ >0, s * increases in y .

(d) For λ >0, s * ≥ ŝ for all y .

Corollary 1: (a) At low y , s * = 0 possible if assumption 2 is violated. (b) At high y , s * = 1

is likely.

2.2 Individuals who do not like schooling

We now consider the case λ < 0. Note that, now we can have the coefficient of u ' (.) in the first

term of Z (s*) (condition (3)) to be positive as now the second and third terms are of opposite

∂V
signs ( < 0 ). So we consider three cases.
∂s

Consider the situation when for some y , at s * the first term in (3) is exactly 0 (i.e., at s * = ŝ ).

Also suppose Z ( sˆ, λ , y ) = 0 . Then Z (sˆ, λ , y ) = 0 for any other value of y as the second and

third terms are independent of y . Thus, the optimal choice ŝ in this case would be invariant with

y . Curve (ii) in fig. 2 depicts this case in ( s, y ) space. Alternatively, consider the lowest possible

income y = wU . If s * > ŝ in this case, then the first term is negative at s * and hence the

9
optimal choice s * with respect to change in y in this case will behave in the same way as for

λ >0 (see curve (i) in fig. 2). Similar argument shows that if s * < ŝ for y = wU , then s * will
be decreasing in y (curve (iii) in fig. 2).

s*
(i)

ŝ (ii)

(iii)

0 y

Figure 2

Also note that as λ increases s * increases for any of the above situations.
We summarize the above discussion in the following proposition.

Proposition 2: Given assumptions 1 - 3, the FOC and comparative static exercises show that for
≥
λ < 0, at y = wU if s * sˆ then s * is respectively increasing or constant or decreasing in y .
<
As λ increases s * increases for any of the above situations.

Note that for the optimal education choice ŝ acts as the change point for λ < 0 individuals. If for
any value of income y, the optimal choice s* > sˆ then for additional income of the parent
education is attractive as the increase in wage (even in the informal sector) becomes more
important than increasing the time in the child labour market. This happens when the education
choice is s * (> ŝ ) already high and hence the time spent as a child labour is already low and this
bolsters the schooling choice. Child’s current period income becomes less important for wealthy
parents. The case s* < ŝ is very interesting. Here exactly the reverse happens and s * and y are

negatively related. In the margin the parent becomes indifferent at s* = ŝ as the FOC becomes

10
independent of income of the parent and the optimal choice of schooling depends on the altruistic
parameter, wage premium and parents attitude towards education, and none of them depends on
the income of the parent. Hence the education choice becomes independent of income of the
parent.

As corollary to the above proposition, we have the following.

Corollary 2: For λ < 0, V (λ , s ) being decreasing and convex in s ,

(a) s * < 1 is possible even for high y .

(b) s * > 0 for high y ⇒ s* > 0 for all y .

(c) s * < 1 for high y ⇒ s* < 1 for all y .

When assumption 2 is violated then we have sˆ < 0 . Then, it is straightforward to conclude the
following.

Corollary 3: For λ < 0, s * may be 0 even for high y . If so, then s * = 0 for all y .

As mentioned before, in the first part of the model (Section 2.1) we have stacked our cards in
favour of education. We have also not allowed for a situation of distress child labour. In our one
period (discrete time) model of childhood a question of survival while child is attending school is
not relevant. Thus, it is all the more noteworthy that even such parent may rationally choose not
to allow her child to complete schooling. This is amply illustrated in proposition 1. For those who
do not value education ( λ < 0) this is not so unexpected, but even for them a reduction in s *
with an increase in y comes as somewhat of a surprise. Anecdotal evidence exists in favour of
such a situation: small businessmen (Bania) class versus urban salaried middle class.

3. Policy analysis

We now focus on the efficiency implications of different anti-child labour policies in terms of our
model. We divide this discussion into two parts. The first part talks about the international
regulations, mostly related to trade sanctions. The national policies, in terms of schooling subsidy
or formal market reforms, are taken up in the second part.

11
3.1 International Legislation

3.1.1 Trade Sanction

Given the widespread existence of child labour in the developing countries, the international
organizations and economics researchers often prescribe trade sanctions against the countries
producing goods using child labour. This policy issue holds a significant place in the current era
of globalization, when the countries are more integrated in terms of trade and investment than
ever before. Most of the exported commodities from a developing country are either made in the
informal sector or uses intermediate inputs from the informal sector. Thus, in order to reduce the
extent of child labour in these countries, the importing countries may impose a trade sanction,
thereby reducing total value of imports and hence adversely affecting the wages in these informal
sectors. Theoretical literature on child labour, however, does not support trade sanction as an
effective mechanism to contain child labour in the developing countries. Ranjan (2001) argued
that trade sanctions may not always reduce incidence of child labour. Similarly the impact of
trade sanctions is also ambiguous in Jafarey and Lahiri (2002). In both papers the overall effect
largely depends on how the credit constraints are relaxed due to trade sanction. Trade sanction
may also increase the incidence of child labour (Gupta, 2002).

We will now discuss the implication of such a sanction in terms of our model. With trade
sanction, the demand for labour in the informal sector will go down and this will reduce wages in
this sector. As wT is pegged at fw F , in our model this will mean a reduction in wU . Now

differentiating Z (s*) with respect to wU we get

∂Z
= u ' (.)(−2α (1 − s*)) + u ' ' (.)αs * (1 − s*) − β u ' (.) + β ( wT − wU )u ' ' (.)(1 − s*) (6)
∂wU

As all the terms in (6) are negative, we have the following result.

Proposition 3: s * increases with a reduction in wU .5

An alternative impact of trade sanction could be that w , w and w are all reduced by a factor say θ
5 U T F

where 0 < θ < 1 . Again it is easy to see that the education choice for the children would go up in this case.

12
3.1.2 Certification

Another common international policy is for the importing country to demand a

certification from the exporting countries that no child labour has been used in
manufacturing the product. The consequence of such a policy would be to reduce child
labour employment in the exporting country, thereby reducing the child wage. In terms of
our model, we can study the implication of such a policy by considering a reduction in
α . Again differentiating Z (s*) with respect to α we get

∂Z
= u ' (.)[(1 − 2 s*)wT − 2(1 − s*)wU ] + α [(1 − 2 s*)wT
∂α
− 2(1 − s*)wU ]u ' ' (.)(1 − s*)[(1 − s*)wU + s * wT ]
= [(1 − 2 s*)wT − 2(1 − s*)wU ][u ' (.) + u ' ' (.)α (1 − s*){(1 − s*)wU + s * wT }].
(7)
The sign of this is in general ambiguous. To get more insight into it, we note that the
second [.] bracketed term in the RHS of the above expression is of the form
u ' ( y + τ ) + u ' ' ( y + τ )τ , where τ = α (1 − s*){(1 − s*)wU + s * wT } . Using Taylor series,
we may say that this is ≅ u ' ( y + 2τ ) > 0 (due to the usual assumption on the utility
∂Z
function). In that case, the net effect, or will be < 0. Thus we have the following.
∂α

Proposition 4: The net effect of a reduction in α on the FOC is certainly weaker than
the effect of a reduction in wU and possibly ambiguous. So, it is not certain (though
possible) that s * increases with a reduction in α .

3.2 National Legislation

3.2.1 Compulsory Schooling

Often the domestic government tries to ensure a minimum length of schooling by setting a lower
limit on the time of school leaving. In this case, in terms of our model, the range of choices for s

13
becomes restricted. Suppose the earliest time one can leave school is set at s > 0. Then we can
infer the following; the proof of which is immediate given our analysis in section 2.

Proposition 5: (a) If s < sˆ ,

(i) There will be no effect on individuals with λ > 0.

(ii) If for λ < 0 and y = wU , s * ≥ sˆ > s , then there will be no effect on individuals with λ <
0 also.
(iii) If for λ < 0 and y = wU , s * < ŝ , then choice of s , say s * * will be Max{s*, s} .

(b) If s > sˆ then all segments of the population are affected. Choice of s will be Max{s*, s} for

all. But this will imply a high rate of dropout at s (for everybody for whom s * < s ).

So average enrollment will improve for those λ < 0 individuals for whom s * decreases in
y (curve (iii) in fig. 2).

3.2.2 Cost Subsidy for Schooling

An alternative way of improving school attendance is to make schooling less costly. This
can be done in two ways. One in which, through reduction in fixed costs like tuition,
books etc., V (λ , s ) is increased for each λ and y . The effect will be asymmetric depending
on the initial choice of s* prior to the policy of cost subsidy for schooling.

Some individuals who were operating at s * < 1 will move to s * = 1 as incentive for choosing
s * = 1 improves. But if there is no change in the variable cost (disutility from continuing in
∂V
school), i.e. is unchanged, then those who remain at an interior solution will continue
∂s
choosing the same s * as without the subsidy. This will result in an education gap among those
who switched to s * = 1 and those who remain at their pre-subsidy choices. Thus this policy
choice would lead to a polarization of population in education choice.

14
The second method in which a cost subsidy may be effected is through a reduction in the variable
cost by providing mid-day meals etc. This will have the same effect on V (λ , s ) as an increase in

λ and the effects will be the same as discussed in Proposition 1(b).

Overall, in both cases, school attendance will improve.

3.2.3 Cash Grant for School Attending Families

If a cash grant is made available to the family when the child attends school, i.e. the
income in the first period is augmented from y to y + δ , then the effect will be the same as
an increase in y as outlined in propositions 1 and 2. So school attendance will improve for all

whose s * > sˆ and deteriorate for all ( λ < 0 individuals) whose s * < ŝ initially. It is often
observed in actual situations that a cash grant is not effective in this respect. This may be due to
the fact that the target population had a high proportion of λ < 0 individuals for whom the
initial schooling choice was s * < ŝ .

3.2.4 Increasing Productivity of Education

So far, in the last few subsections, we have been discussing the implication of restrictive policies
on school attendance and child labour. We now discuss a different kind of policy that is
constructive in nature. In particular, we consider a government policy that makes education more
effective in that it improves the chance of formal sector employment. This can be achieved
through an improvement in the quality of schooling, by introducing training on job oriented skills
in school or better tutoring. The effect in terms of our model would be an increase in the
probability f and hence in an increase in wT (= fw F ) . Let us now explore the consequences of
such a change.

To start with, more individuals will now prefer the formal employment option and hence choose
s * = 1. To look at the effect on those who still choose to remain in the interior (choose s * < 1),
differentiate Z (s*) with respect to wT we get

15
 ∂Z
= u ' (.)α (1 − 2 s*) + α 2 (1 − s*)s * u ' ' (.)[(1 − 2s*)wT − 2(1 − s*)wU ]
∂w T

+ β u ' (.) + βs * ( wT − wU )u ' ' (.). (8)

The second and third terms on the RHS of (8) are positive and the fourth term is negative. Sign of
the first term depends on whether s * is > or < ½.. So if s * < (>) ½., the first term becomes
positive (negative) and the RHS of (8) is more likely to be positive (negative). We summarise this
in the following.

Proposition 6: For those individual who continues to choose an intermediate s* < 1 we have, (a)

If s * is initially low (< ½), then with an increase in wT , s * increases but if s * is initially high

(> ½), then we may see a reverse relationship between wT and s * .

(b) We will observe a clustering of s * choices around ½ as the people with low s * s increase
their schooling choice and the people with high s * s decrease their schooling choice towards ½.
This will also create an education gap.

4. Conclusion

This paper explains the existence of child labour when there is a problem of educated
unemployment after full schooling of the child. Thus, even the altruistic parents might withdraw
their children from school and put them to work in the informal sector. This happens even when
the parents intrinsically value education and also the informal wage is a strictly increasing
function of the time spent in school.

On the policy front, we have examined the effectiveness of the standard policies in terms of our
model. It is found that trade sanctions may be effective in reducing the incidence of child labour
but the certification may have an ambiguous effect. Compulsory schooling would ensure more
children to go to school in early ages, but there might be substantial drop out at the limit of the
compulsory schooling period. Cash grant may not have any desirable effect when the society has
very low level of education and parents do not like education. Increasing the productivity of
education may lead to clustering of education choice and as a result drop out after some schooling
may be significantly large.

16
Reference

1. Baland, Jean-Marie and Robinson , James A., 2000, Is Child labour efficient, Journal of
Political Economy, 108 (4), 663-679.

2. Basu, Kaushik & Van, Pham Hoang, 1998. "The Economics of Child Labor", American
Economic Review, American Economic Association, vol. 88(3), pages 412-27.

3. Basu, K. 1999. Child Labour: Cause, Consequence and Cure, With Remarks on International
Labour Standards, Journal of Economic Literature, 37 (3), 1083 - 1119.

4. Basu, Kaushik, 2000. "The Intriguing Relation between Adult Minimum Wage and Child
Labour," Economic Journal, Royal Economic Society, vol. 110(462), pages C50-61.

5. Chakraborty, Shankha and Das, Mausumi, 2005, Mortality, Fertility and Child labour,
Economics Letters, 86, 273-278.

6. Dessy, Sylvain & Pallage, Stéphane , 2003. "The Economics of Child Trafficking," Cahiers
de Recherche 03-23, CIRPEE.

7. Dessy, Sylvain & Pallage, Stéphane, 2005. "A Theory of the Worst Forms of Child Labour,"
Economic Journal, Royal Economic Society, vol. 115 (500), pages 68-87.

8. Carol Ann Rogers & Kenneth A. Swinnerton, 2002. "A Theory of Exploitative Child Labor"
U. S. Department of Labor, mimeo.

9. Genicot, Gerance, 1998, Malnutrition and child labour, Department of Economics, University
of California, Irvine.

10. Gupta, Manas Ranjan, 2002, Trade sanctions, adult unemployment and the supply of labour,
Development policy Review, 20 (3), 317-332.

11. Jafarey, Saquib and Lahiri, Sajal, 2002, Will trade sanctions reduce child labour? The rold of
credit markets, Journal of Development Economics, 68, 137-156.

17
12. Raju, D. (2005): Banning Harmful Child Labour: A Labour Market and Welfare Analysis,
mimeo, Department of Economics, Cornell University.

13. Ranjan, Priya, 2001, Credit constraints and the phenomenon of child labour, Journal of
Development Economics, 64, 81-102.

14. Ranjan, Priya, 1999, An economic analysis of child labour, Economics Letters, 64, 99-105.

15. Ray, R., 2000. The determinants of child labour and child schooling in Ghana, mimeo,
University of Tasmania.

16. The Probe Team , 1999, Public report on basic education in India, New Delhi , Oxford
University Press.

17. ILO - International Programme on the Elimination of Child Labour (1996): Child Labour:
What is to be done, Document for discussion at the Informal Tripartite Meeting at the
Ministerial Level, International Labour Office, Geneva

18. Heady, Christopher (2000), “What is the Effect of Child Labour on Learning Achievement?
Evidence from Ghana”. Innocenti Working Paper No. 79. Florence: UNICEF Innocenti
Research Centre

19. Grootaert, C. and H.A. Patrinos, 1998, The policy analysis of child labor: a comparative
study, St. Martin’s Press.

20. J. P. Smith and F. R. Welch (1989): Black Economic Progress After Myrdal, Journal of
Economic Literature, 27, 519 - 64.

18
Childhood is a period of life that should be devoted not to work but to education and training; that
because of its nature or the conditions in which it takes place, the work done by children often
jeopardizes their chances of becoming productive adults who will be useful to society. In many
places child labour outside the family is on the increase. This reflects a shift in attitude towards
child labour itself. Initially seen mainly as a tool enabling the child to become socialized and
acquire skills that will be useful to him or her in the future, it is now increasingly often seen by
families as a means of supplementing their income.

Poverty is the greatest single force, which creates the flow of children into the workplace. It
forces many children to work full time for their own and their families' survival. Furthermore,
because of poverty, the acute need of many households to keep many family members working to
ensure income security makes it nearly impossible for them to invest in their children's education.
There is cause for concern first of all because many children are put to work at a very early age,
in particular in rural areas where they often begin to work as soon as they are 5 or 6 years old.
Although the great majority of economically active children belong to the 10 to 14 year age
group, the proportion of child workers under the age of 10 in the total is far from negligible; it
may be up to 20 per cent in some countries.

Smith and Welch (1989) document a sharp positive relationship between years of schooling
(even for incomplete schooling, like 0-4 years or 5-7 years of schooling) and wage for Black and
White American Males during the period 1940 - 1980. The structure of the labour market for Black
Americans would be similar to the informal markets in our country at the present time. This is sort
of indirect evidence for wage premium for (partial) education.

However, the issue under discussion here is not the "child labour due to poverty" or a situation of
"distress child labour" which along with other related issues has been beautifully researched in
Basu and Van (1998), Basu (1999) and Basu (2000). Neither are we considering a situation
where the employment of very young children is an alarming problem; the younger the child, the

19
more vulnerable he or she is to physical, chemical and other hazards at the workplace and, of
course, to the economic exploitation of his or her labour (Rogers and Swinnerton, 2002, Dessey
and Pallage, 2003, 2005 and Raju, 2005).

We rather think of a situation where, in addition to the family's inability to pay for school, many
children live in communities that do not have adequate school facilities, so they work. However,
even when schools are available, the comparatively high cost to poor families of investment in the
education of their children means that the return on such investment must also be high. In fact,
many schools serving the poor are of such abysmal quality, or chances of improved upward
mobility for graduates are so slim, that the expected return is not equal to the sacrifice made. The
literature is replete with testimony of families that would like to educate their children, but either
cannot do so, or do not feel that the inferior schools available to them merit the costs. While it is
true that many children drop out of school because they have to work, it is equally true that many
become so discouraged by school that they prefer to work. These problems mean that only 68 per
cent of the world's children complete primary education (until age 11).

There are several empirical studies that support this view. Heady (2000) notes that Participation
in school increases from the age group 7 - 9 to the age group 10 - 14 and then decreases for aged
15 - 19

Table 1: Ghana: Participation in School & Work (per cent)

Age in years: 7-9 10-14 15-19

Sex: Girls Boys Girls Boys Girls Boys
School attendance 62.4 78.0 74.4 81.4 39.9 55.1

Reading and Mathematics test results conducted among attending and non-attending students
show that attending students perform significantly better in some age groups but the difference is
insignificant in most cases. Also, this is not controlled for other factors.

Table 2: Pakistan: Participation in School & Work (per cent)

Age in years: 5-9 10-14 15-19
Sex: Girls Boys Girls Boys Girls Boys

20
 School attendance 31.0 53.3 30.6 72.9 11.5 41.5

Bangladesh
In contrast to the nationally representative sample surveys for Ghana and Pakistan, with a
resultant emphasis on rural child work, the quantitative survey for Bangladesh was focused on the
slum population of Dhaka city. It covered over 700 households and was conducted from 1995 to
1997 as part of the Urban Livelihoods Study, a joint research project of Proshika, the London
School of Hygiene and Tropical Medicine, and the University of Bath. The qualitative analysis
was based on detailed interviews, conducted with 14 of the households that had taken part in the
quantitative survey. The analysis uses the ILO definition of work and, in this urban setting, this
typically involved working for people outside the household.

Table 3 presents the pattern of school attendance and work from the quantitative survey. In the
younger age group, 8-11, school attendance for boys and girls is very similar at just over 60 per
cent with boys being more likely to combine school with work. Girls and boys are also similar in
the proportion that neither work nor go to school. However, boys are more likely to work than
girls. Turning to the older age group, 12-16 years of age, both boys and girls are much less likely
to attend school, but the reduction for girls is greater. This reduction in schooling is accompanied
by an increase in work, with boys again more likely to work. The number of girls doing neither is
the same as in the younger age group, but the proportion of idle boys is substantially lower.

Table 3: Work and School for Slum Children in Bangladesh

Age in years: 8-11 12-16
Sex: Girls Boys Girls Boys
Only work 13.4% 16.2% 55.3% 56.3%
Work and school 1.8% 6.6% 1.4% 6.4%
Only school 61.2% 54.8% 20.0% 24.7%
Neither 23.6% 22.4% 23.3% 12.6%

There are interesting comparisons with Ghana and Pakistan, although the fact that the age groups
do not correspond makes precise comparisons difficult. The school attendance is substantially
lower than in Ghana but not generally lower than in Pakistan. However, the difference between
boys’ and girls’ school attendance is much smaller than in Pakistan. The work participation rates

21
are not markedly different from either Ghana or Pakistan, but there is a major difference in the
ability of children to combine work and schooling: it is even less common to see schooling
combined with work than it is in Pakistan. A large part of this difference is probably due to the
fact that the children work outside the household.

Families were asked why their children were not attending school. The most common explanation
was that the children were too busy working. However, it is worth noting that other commonly
expressed reasons were “no suitable school facilities” (mostly meaning that available schools are
too expensive), “busy with housework” (almost entirely applied to girls) and “child does not want
to go” (mainly applied to boys). The effect of housework on girls is consistent with the higher
proportion of girls neither working nor attending school, and reflects the same strong views on
gender roles that are evident from the Pakistan data and, to a smaller extent, the Ghana data.
These reasons for not attending school were followed up in more detail in the qualitative analysis.
This revealed that parents did attach considerable value to education but often found that school
fees were too expensive or felt that work experience was more useful to future employment
prospects than education. To some extent, these two factors were combined, with parents saying
that education would help their child get a good job if they stayed at school for many years, but
they could not afford that. Thus, it was not a case of work reducing schooling, but poor schooling
encouraging work. Although the data show that the situation of children in urban slums in
Bangladesh is different from both Ghana and Pakistan, there is still a clear policy implication that
improvements in schooling and reductions in fees may be a better way of encouraging school
attendance than trying to curb child labour directly.

Grootaert and Patrinos (1998) report the patterns of child work and school attendance for Côte
d’Ivoire, Colombia, urban Bolivia and Philippines. Children in urban areas are more likely to
concentrate on schooling, and slightly more likely to combine work and schooling, with the
consequence that many fewer concentrate on work. In urban Bolivia, full-time school attendance
is over 90 per cent until the age of 13 years. After that, child work becomes significant, with
somewhat more children working full-time than combining work with school. Only about 3 per
cent neither work nor attend school, and this is more common for girls. There is little difference
between girls and boys in their labour force participation.

Our model thinks of part schooling and part child labour as an alternative to full schooling
because of the above mentioned reasons. Of course, some schooling helps in increasing the wage

22
rate the candidate faces in the labour market. In section 2 we explain when it would be
economically rational to leave school early and thus forego formal employment opportunities in
the future.

We then look at alternative policy measures that may help in reducing school dropout and study
their efficiency in terms of our model. Three fundamental types of action against child labour can
be provided only by the central government: (i) child labour legislation and appropriate
enforcement mechanisms; (ii) a national child labour policy that sets public priorities and reaches
out to engage all the important social actors; and (iii) a publicly funded system of basic education
that ensures quality schooling that is physically and economically accessible to children of even
the very poorest families. The last is the most important of all since, without it, whatever
initiatives against child labour are undertaken will achieve very limited success. Other levels of
government, especially municipal government, can play a crucial role in mobilizing and focusing
local human and material resources on specific child labour problems.

Improving child labour legislation and enforcement measures. Most countries have child labour
legislation that establishes a minimum age for admission to employment or work and regulates
working conditions for young persons. Another problem relates to the discrepancies which exist
in many countries between the minimum age required by the law to work and that at which it is
permitted to leave the school system. In several cases, the minimum age for admission to work is
lower than the school leaving age, giving children access to employment before they have
completed the minimum number of years of compulsory schooling.

Extending and improving schooling for the poor. The single most effective way to stem the flow
of school-age children into abusive forms of employment or work is to extend and improve
schooling so that it will attract and retain them. It has been widely demonstrated that families are
prepared to make major sacrifices for the education of their children when it is economically and
physically accessible and truly productive in terms of future employment prospects. For example,
IPEC has found that schools become more attractive to poor families when they include practical
training in skills that make the children more employable or employable at higher wage rates.
Getting working children back into school is often less difficult than expected. For example, in
areas where most children have usually worked full time instead of attending school, parental
demand for return to school has been stimulated after the first parents were enticed to break with
the tradition of child labour. In cases where the income from child work is, in fact, indispensable

23
to the family, it has been possible to convince parents to lighten the children's workload so that
school attendance could be accommodated while the child still worked.

The types of economic incentives usually found include the payment of cash grants to children or
their families, the provision of free school lunches, other in-kind payments for school (e.g.
stationary or clothing) and the waiver of school fees. They may also comprise income-generating
schemes for poor families in communities with a high concentration of working children and
apprenticeship or other school-work programmes for children that provide education or training
with income as an alternative to child labour (ILO - IPEC, 1996). This analysis is done in section
3. Finally section 4 concludes our discussion.

24