Sie sind auf Seite 1von 22

Journal of Development Economics

Vol. 64 Ž2001. 81–102


www.elsevier.comrlocatereconbase

Credit constraints and the phenomenon


of child labor
Priya Ranjan )
Department of Economics, 3151 Social Science Plaza, UniÕersity of California, IrÕine, CA 92697 USA

Abstract

This paper develops an overlapping generations general equilibrium model where


inefficient child labor arises due to credit constraints. It derives a positive relationship
between inequality in the distribution of income and the incidence of child labor. Looking
at policy implications, it shows how trade sanctions against countries using child labor may
fail to reduce the incidence of child labor. It discusses some alternative policies to reduce
the incidence of child labor. q 2001 Elsevier Science B.V. All rights reserved.

JEL classification: J24; D60


Keywords: Child labor; Credit constraints; Income inequality; Trade sanctions

1. Introduction

The phenomenon of child labor is widespread in the developing world. The


International Labour Organization Ž1996. estimates that between 100 million and
200 million people under age 15 work in the world. Of these young workers, 95%
are in developing countries, with half of these in Asia. More importantly, about
110 million school age children worldwide Žaround 20% of the corresponding age
group. receive no primary education. Again a vast majority of these are concen-

)
Tel.: q1-949-824-1926; fax: q1-949-824-2182.
E-mail address: pranjan@uci.edu ŽP. Ranjan..

0304-3878r01r$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 4 - 3 8 7 8 Ž 0 0 . 0 0 1 2 5 - 5
82 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

trated in low income countries. In recent times, due to increased awareness of the
problem there has been a growing momentum in the west to impose international
minimum labor standards banning child labor. During the Uruguay round of
GATT negotiations, a good deal of pressure was put by some developed countries
to introduce environmental and social standards as a legitimate basis for trade
intervention, much to the chagrin of developing countries. However, in order for a
policy intervention to be effective in reducing the incidence of child labor, it
should be informed by the reasons for the existence of child labor. Therefore, it is
imperative that policy in this area should be based on a careful analysis of the
phenomenon of child labor.
Several empirical studies document the high returns on all levels of education
in developing countries.1 A natural question to ask then is why do we observe
inefficient child labor. Baland and Robinson Ž1999. show how inefficient child
labor could arise despite parental altruism due to either zero bequests arising from
poverty, or when capital markets are imperfect. Further, even if the non-negativity
constraint on bequests is relaxed by introducing reverse altruism, imperfect capital
markets still give rise to inefficient child labor.2 A similar result was obtained by
Ranjan Ž1999. in a simple two-period model where despite parental altruism, the
non-existence of market for loans against the future earnings of children gave rise
to inefficient child labor.
This paper studies the relationship between inequality in the distribution of
income and incidence of child labor in the presence of credit constraints and,
derives some policy implications. The credit constraints in the paper can be
understood in either of the following two ways as was made clear in Baland and
Robinson Ž1999.: either parents are unable to borrow with the promise by children
that the parents will be reimbursed for this in the future; or, children are unable to
borrow to reimburse their parents of the foregone earnings while going to school.
With either interpretation the net result is that the borrowing against the future
earnings of children is not possible. It is shown in a model where individuals have
differing abilities that in the presence of credit constraints, greater inequality Žin
the sense of second-order stochastic dominance. is associated with greater inci-
dence of child labor. For each level of ability there is a threshold level of parental
income such that households below that threshold send their children to work.

1
Psacharopoulos Ž1994. documents that the returns to all levels of education Žboth private and
social. are much higher than the returns to physical capital for all regions of the world. For the
countries where at least some children worked full time, the average rate of return on primary
education based on the data in Psacharopoulos Ž1994. was 26.7%.
2
Becker and Murphy Ž1988. and Nerlove et al. Ž1988. were the first to point out that a
non-negativity constraint on bequests could lead to underinvestment in the human capital of children.
Parsons and Goldin Ž1989. also showed that the sub-optimality of investment in education is closely
related to the availability of efficient capital markets.
P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102 83

Since the high marginal utility of current consumption gives rise to borrowing
constraint, a convex marginal utility from consumption implies a convex threshold
curve. Given this, redistributing income from the rich to the poor increases the
probability of a poor household rising above the threshold more than the increase
in the probability of a rich household falling below the threshold. Therefore, such
a transfer is likely to reduce the incidence of child labor. An implication of this
result is that policies that redistribute income from the rich to the poor can reduce
the incidence of child labor.
Next, the impact of trade sanctions on the incidence of child labor is discussed.
The incidence of child labor is high in poor countries having a comparative
advantage in the production of unskilled labor intensive goods. The impact of
trade sanctions on factor prices in such a country is to reduce the unskilled wage
and to increase the skilled wage through the standard Stolper–Samuelson effect.
This by itself would increase the returns to schooling and hence induce the
altruistic parents to send their children to school. However, a decline in the
unskilled wage reduces the income of the parents who are unskilled. In a world
with credit constraints, this effect would tend to increase the incidence of child
labor for children of the unskilled. It is shown that taking both these effects into
account, trade sanctions may fail to reduce the incidence of child labor. It is well
known that a trade sanction removes the gains that accrue from specialization; if
its impact on the incidence of child labor is ambiguous, the case for this as a
policy tool to reduce the incidence of child labor is a weak one.
Among related literature, in a recent interesting paper, Basu and Van Ž1998.
generate child labor in a general equilibrium setting by assuming that a household
would not send its children to work if its income from non-child labor sources
were sufficiently high Žthey call this assumption Luxury Axiom.. This generates
multiple equilibria in which a ban on child labor can move an economy from an
equilibrium with low wage where children work to another with high wage where
children do not work. Thus, child labor is observed in this model for low levels of
parental income due to the Luxury Axiom, while in our paper it arises due to
credit constraints.3
Concerning the issue of distribution, there is a literature that shows the
relationship between distribution of income and investments in human capital in
the presence of imperfect capital markets. In Loury Ž1981. the effect of income
inequality on investments in human capital is due to the assumption that returns to
human capital are decreasing at individual level and that human capital investment
is perfectly divisible. In Galor and Zeira Ž1993. the link between distribution of
income and investment in human capital arises due to the indivisibility in human
capital investment. However, greater inequality may increase investments in

3
See Basu Ž1999. for a comprehensive survey of the literature on child labor.
84 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

human capital in their model where agents have identical ability. Chiu Ž1998.
extends the analysis of indivisible investment in human capital to allow for
heterogeneous ability and shows how a decrease in inequality can increase
aggregate human capital accumulation. However, none of these papers deal with
the subject of inequality and the incidence of child labor. Also, unlike Galor and
Zeira Ž1993. and Chiu Ž1998. where parents have Awarm glowB preferences, in
our model parents care directly about the utility of their children. Since in Basu
and Van Ž1998. child labor arises for low levels of parental income, their model
has distributional implications too. This has been explored in a recent work by
Swinnerton and Rogers Ž1999. who extend the Basu–Van model by introducing
heterogeneity in the ownership of non-labor income and show that if non-labor
income is distributed with sufficient equality then there is no child labor in
equilibrium. Rogers and Swinnerton Ž1999. extend the model further by allowing
for a more general utility function and find that the relationship between the
incidence of child labor and the distribution of non-labor income is ambiguous. In
both these papers the measure of inequality is the fraction of households receiving
dividends, while it is assumed that dividends are equally distributed between
dividend receiving households. In contrast, we look at inequality in the distribution
of total parental income and our measure of inequality is second-order stochastic
dominance, which corresponds to Lorenz dominance for distributions with same
mean.
Finally, there has been some work on the issue of using trade policy to enforce
labor standards, including standards regarding child labor Žsee Basu Ž1999. for
references.. For example, Rodrik Ž1995. justifies using trade policy if lower labor
standards are viewed as unfair and exploitative by a wide segment of society in the
importing country. Bhagwati Ž1995. opposes the use of trade sanctions on these
grounds because it is difficult to find universally condemned practices given the
diversity of labor practices and standards resulting from diverse cultural values,
economic conditions, analytical beliefs etc. He therefore cautions that labor
standards may become a front for protectionism. Unlike our paper, however, these
studies do not discuss the channels through which a trade sanction can impact the
incidence of child labor.4
The plan of rest of the paper is as follows. Section 2 presents the formal model.
Section 3 does the comparative statics with respect to initial distribution of income
and changes in factor prices. Section 4 derives some policy implications. Section 5
concludes.

4
On a related issue, Basu Ž1999. discusses the welfare implications of a ban on child labor in the
presence or absence of international capital mobility, and shows how a country imposing a unilateral
ban in the presence of capital mobility can become worse off. This calls for policy co-ordination among
countries.
P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102 85

2. The model

2.1. The production structure

The economy can produce two final goods, H and L, with two factors of
production: skilled labor denoted by S, and two types of unskilled labor—child
labor and adult unskilled labor—both denoted by U. Child labor and adult
unskilled labor are substitutes for each other in production. In efficiency units one
unit of child labor is equivalent to 0 - u - 1 units of adult unskilled labor. Both
final goods employ both factors of production with constant returns to scale
technology. H is the high-tech good that is more skill intensive than the low-tech
good L at all factor prices. L is also the numeraire good. The production functions
are given by

H s Ah F h Ž S h ,Uh . Ž 1.

L s Al F l Ž Sl ,Ul . Ž 2.
Since child labor and adult unskilled labor are substitutes, if adult unskilled wage
is wu , child wage is wc s u wu . The skilled wage is denote by ws . The factor
supplies are endogenously determined through the occupational choices of individ-
uals. A small open economy assumption is made, which takes the relative price of
goods as given. The relative price of the high-tech good is denoted by p. Given
the above production structure, relative goods price fixes the relative factor price,5
v s wsrwu .

2.2. Household behaÕior

Assume that each individual has one parent and one child. Further, assume that
each individual lives for two periods. In the first period they live with their
parents, and in the second period they live with their children. Parents have to
decide whether they want to send their children to work or school. Assume that
each individual i is born with a talent si g ws, s x, which is independent across
families at a point in time. Denote by F Ž s . the distribution function of talent for
each generation, and by f Ž s . the corresponding density function. The talent is
basically the amount of human capital that an individual can acquire upon going to

5
This is true only in a diversified equilibrium where both goods are produced. Throughout the paper
we assume a diversified equilibrium.
86 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

school. Therefore, if an individual with talent si goes to school, the income of the
individual in the second period is si ws . If the individual works as unskilled in the
second period of his life, the talent is of no use, and his income remains wu .
Assume for simplicity that the talent remains constant across generations within a
dynasty. Denote time by t. Further assume that in period t s 0 each adult is
endowed with income b 0 g wb,b x, which has a distribution function GŽ b 0 ..
As far as the preferences are concerned, the utility of a parent is assumed to
depend on family consumption during his tenure, and on the well-being that the
offspring experiences after taking up the family leadership. Using subscript t to
denote time, and denoting the utility of parents in period t by Vt , Vt is given by

Vt s U Ž Ct . q b Vtq1 Ž 3.
where Ct s C ght C lt1y g is an index of consumption, which depends on the actual
consumption of the two goods, where C ht is the consumption of high-tech good
and C lt is the consumption of low-tech good. Further, utility function U: Rq R ™
is bounded, continuously differentiable, strictly increasing and strictly concave,
with lim c ™ 0 U X Ž C . s `. 0 - b - 1 is the discount factor. Even though parents
care about their children, they discount their well being.
Given the above utility from consumption function, UŽ Ct ., if the expenditure
on goods in period t is Et , and the relative price of the high-tech good is fixed for
the small open economy at p, the indirect utility function is given by

IU Ž Et , p . s U Ž C V Et . Ž 4.
where C V is a constant which is a function of g , and p. To avoid cluttering
notations, normalize this constant to unity, and write Eq. Ž3. as

Vt s U Ž Et . q b Vtq1 Ž 5.

2.3. No credit market case

Given the timing structure of the model, allowing parents to save is equivalent
to allowing them to leave a bequest. Allowing for negative bequests will mean that
parents are able to borrow against the future earnings of the children with the
promise by children that they would repay the loan. The imperfect credit markets
preclude such borrowing possibilities as discussed in the introduction. Therefore,
in the no credit market case discussed below it is assumed that parents are not
allowed to leave a negative bequest or borrow against the future earnings of
children. For simplicity it is also assumed that there are no savings or positive
bequests. With the latter assumption the analysis below is valid for those house-
holds for whom the non-negativity constraint on bequests is binding.
P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102 87

The parent in period t has to decide whether to: Ž1. send the child to work in
which case the child earns the child wage in period t and the unskilled adult wage
in period t q 1; or Ž2. send the child to school in period t in which case the child
earns no wage in period t and the skilled wage in period t q 1. The parent takes
the action that maximizes his own utility, which takes into account the utility of
his child. Denote the income from non-child labor sources by bt . In the absence of
saving or borrowing possibilities, all income is consumed in the period in which it
is earned. Now the parent solves the following optimization problem in period
t s 0.

V Ž b 0 , s . s Max  U Ž b 0 q wc . q b V Ž wu , s . ,U Ž b 0 . q b V Ž s ws , s . 4 Ž 6.
Since saving and borrowing are not possible, if we start with an initial distribution
of parental income in period t s 0, from period t s 1 onwards parental income can
be either wu or s ws . Therefore, we solve Eq. Ž6. in two steps. First we will solve
for V Ž wu , s . and V Ž s ws , s ., which is the problem faced by agents in period
t s 1, and then step back and see which action maximizes the value function in
period t s 0.

2.3.1. Solution of the problem in period t s 1


V Ž wu , s . and V Ž s ws , s . can be written as follows.

V Ž wu , s . s Max  U Ž wu q wc . q b V Ž wu , s . ,U Ž wu . q b V Ž s ws , s . 4 Ž 7 .

V Ž s ws , s . s Max  U Ž s ws q wc . q b V Ž wu , s . ,U Ž s ws . q b V Ž s ws , s . 4
Ž 8.
From the inspection of Eqs. Ž7. and Ž8. it can be guessed that depending on the
level of talent, s , the agent will decide to send the child to school or work in each
case. Define su as the solution to the following equation
U Ž wu q wc . b
s U Ž wu . q U Ž s ws . Ž 9.
1yb 1yb

We make the following assumption about the parameters, which ensures that a
skilled parent never sends the child to work.6

Assumption 1. sws ) Ž wcrb . q wu .

The following Lemma is proved in Appendix A.

6
If in a first-best world people can lend and borrow freely at a common rate of interest r and if
b G Ž1rŽ1q r .., then assumption 1 ensures that there is no child labor in a first-best world.
88 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

Lemma 1. Under Assumption 1 the following is true of the decision in period


t s 1. (a) A skilled parent always sends the child to school; (b) an unskilled parent
sends the child to school if s G su , and to work otherwise.

Thus, if s G su , then both types of parents want to send the child to school and
the values of V Ž wu , s . and V Ž s ws , s . are

b U Ž s ws .
V Ž wu , s . s U Ž wu . q U Ž s ws . ; V Ž s ws , s . s Ž 10 .
1yb 1yb

If s - su , then a skilled parent sends the child to school, while an unskilled


parent sends the child to work and the values of V Ž wu , s . and V Ž s ws , s . are

U Ž wu q wc . U Ž s ws .
V Ž wu , s . s ; V Ž s ws , s . s Ž 11 .
1yb 1yb

Having solved the problem in period t s 1, we go back and solve the problem
in period t s 0 and find out the incidence of child labor in steady state.

2.3.2. Incidence of child labor in steady state


It follows from the discussion of the problem in period t s 1 that for house-
holds with talent s G su , every child goes to school from period t s 1 onwards
irrespective of whether the parent in period t s 1 is skilled or unskilled. This
implies that in steady state all households having talent s G su are going to send
the child to school irrespective of what happens in period t s 0. Therefore, from
the steady state’s point of view what is interesting in period t s 0 is the decision
of households with s - su , which is what we look at next.
In this case using Eq. Ž11. the parent’s optimization problem in period t s 0
reduces to

V Ž b0 , s .

b b
½
s Max U Ž b 0 q wc . q
1yb
U Ž wu q wc . ,U Ž b 0 . q
1yb
U Ž s ws . 5
It can be easily seen that for each level of talent there exists a threshold level of
parental income b ) Ž s . such that if b 0 ) b ) Ž s ., the parent sends the child to
school, otherwise the child is sent to work. The threshold b ) Ž s . is a solution to
the following equation.

b U Ž s ws . s Ž 1 y b . U Ž b 0 q wc . y U Ž b 0 . q b U Ž wu q wc . Ž 12 .
P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102 89

It is easy to show that if U Y Ž C . - 0 then b )X Ž s . - 0. Further, if U Z Ž C . ) 0, i.e.


the marginal utility is convex, then b )Y Ž s . ) 0. Alternatively, we can define a
threshold level of talent for each level of income as s˜ Ž b 0 .. Again U Y Ž C . - 0 and
U Z Ž C . ) 0 imply s˜ X Ž b 0 . - 0 and s˜ Y Ž b 0 . ) 0, respectively. For the case of a
logarithmic utility function the threshold b ) Ž s . given by
b
1y b
Ž wu q wc .
b) Ž s . s b b Ž 13 .
1y b 1y b
Ž s ws . y Ž wu q wc .
Note that in Eq. Ž13. b )X Ž s . - 0 and b )Y Ž s . ) 0.
The incidence of child labor among households with talent s - su is given by
su
LC s Hs GŽ b) Ž s . . f Ž s . ds Ž 14 .

It should be noted that despite recognizing the fact that by sending his own child
to work in period t s 0 a parent makes sure that all his future descendants are sent
to work, a parent may still do so. The reason is the high marginal utility of present
consumption for low-income households and their inability to smooth consump-
tion by borrowing, which induces them to use child labor as a consumption-
smoothing device.
The steady state incidence of child labor is given by the expression in Eq. Ž14..
Depending on the value of su there are two possible cases. If su F s, then there
will be no child labor in steady state because from period t s 1 onwards every
parent will send the child to school. If su ) s, then the incidence of child labor in
steady state is given by the mass of households having talent less than su whose
ancestors sent the child to work in period t s 0, which is given by Eq. Ž14.. It is
possible for there to be a poverty trap if su ) s and no household can afford to
send the child to school in period t s 0, i.e. b ) Ž s . ) b. In this case everyone in
this economy will be unskilled in the long run. Assuming su - s eliminates the
poverty trap case. Without loss of generality, we assume the following for the
range of su in the comparative statics below.

Assumption 2. s- su - s .

3. Comparative statics

Examination of Eq. Ž14. makes it clear that the incidence of child labor in
steady state depends on the initial distribution of income and the determinants of
90 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

su and b ) Ž s .. The latter two depend on the factor prices. In this section we are
going to do two types of comparative statics. First, we look at the impact of
changes in the initial distribution of income GŽ b 0 . holding the factor prices or
wages constant. Next, we look at the impact of changes in factor prices. The latter
will help in analyzing the impact of trade sanctions on the incidence of child labor.

3.1. Inequality and incidence of child labor

We want to examine the relationship between inequality in the initial distribu-


tion of parental income and the steady state incidence of child labor. The measure
of inequality used is the second-order stochastic dominance. Distribution G 1Ž b . is
more unequal than distribution G 2 Ž b . in this sense if the following is true.
X
b
Hb G 1 Ž b . y G 2 Ž b . d b G 0;bX g b,b Ž 15 .

If Eq. Ž15. is satisfied then distribution G 2 Ž b . dominates distribution G 1Ž b . in the


second-order stochastic sense. It will be assumed that for both initial distributions
ability is independent and identically distributed for each level of income.
We impose the following restrictions on the utility function and the parameters
of the model.

Assumption 3. U Z Ž C . ) 0, bG b ) Ž su ., f X Ž s . G 0.

It is proved in Appendix B that under Assumption 3, which lists a set of


sufficient conditions, the following proposition holds.

Proposition 1. If the initial distribution of parental income in economy 2, G2 (b),


dominates the distribution of parental income in economy 1, G1(b), in the
second-order stochastic sense, then the incidence of child labor in economy 1 is at
least as high as in economy 2.

The restriction U Z Ž C . ) 0, which gives rise to a convex threshold curve


Ž b s . ) 0 and s˜ Y Ž b 0 . ) 0., is satisfied by a lot of commonly used utility
)Y Ž

functions. This restriction is needed to generate a precautionary motive for saving,


and is satisfied by all utility functions belonging to the class known as constant
relative risk aversion utility function. Fig. 1 shows the relationship between the
initial distribution of income and the steady state incidence of child labor. The
incidence of child labor is given by the area below the threshold curve b ) Ž s .
between b and b and to the left of the vertical line at s s su . The threshold curve
b ) Ž s . is truncated at su because households with s G su will have their
P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102 91

Fig. 1. Determination of incidence of child labor.

descendants going to school in steady state, irrespective of their initial level of


income. Fig. 1 shows two alternative positions of b relative to b ) Ž su .. It is easy to
see that if bG b ) Ž su ., then the threshold curve is convex in the entire interval
wb,b x. However, this is not so if b- b ) Ž su . because of the vertical portion at
s s su . The condition b ) Ž su . F b implies that even the poorest households with
talent s G su will be able to send their child to school in period t s 0. This
condition ensures that the threshold is convex over the entire interval wb,b x and
also allows us to use stochastic dominance criterion to prove the result in
Proposition 1.
To gain intuition for the result in Proposition 1 suppose G 2 Ž b . has been
obtained from G 1Ž b . by transferring a dollar from a rich person to a poor person
without knowing the ability of each person. This transfer increases the threshold
level of ability required for the rich person to send the child to school, and lowers
it for the poor person. Convexity of the threshold Žwhich arises from convex
marginal utility. ensures that the magnitude of the former is less than the
magnitude of the latter, i.e. the magnitude of the rise in threshold for the rich
person is smaller than the magnitude of the fall in threshold for the poor person.
The incidence of child labor depends on what happens to the probabilities of these
two people sending the child to work. f X Ž s . G 0 is a sufficient condition to ensure
that the increase in the probability of sending the child to work for the rich person
is lower than the decrease in the probability of sending the child to work for the
92 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

poor person. Therefore, the incidence of child labor is likely to be lower under
distribution G 2 Ž b . than under distribution G1Ž b ..
The argument in the previous paragraph does not apply if the economy is in a
poverty trap situation, i.e. all children are sent to work in steady-state. In this case
even an inequality enhancing redistribution can reduce the incidence of child
labor. The assumption that su - s rules out this case. As well, the argument in
the previous paragraph is not applicable if for incomes below a certain cut off the
probability of sending the child to work is the same irrespective of the level of
initial income. For example, for all b - b ) Ž su . the probability of sending the
child to work is given by F Ž su .. In this case an inequality increasing redistribution
of income between two individuals below this cut off can reduce the incidence of
child labor because the probability of sending the child to school increases for the
individual receiving the transfer, without decreasing the probability of sending the
child to school for the individual making the transfer. Again, this case has been
ruled out by the condition bG b ) Ž su . in Assumption 3.
We have not yet talked about the steady state distribution of income. From the
discussion of the solution in period t s 1, it can be inferred that for all unskilled
households the steady state level of parental income is simply wu . For skilled
households the steady state level of parental income is s ws , where s g ws, s x.
The fraction of households with steady state level of income wu is given by the
value of LC in Eq. Ž14.. While Proposition 1 shows the link between inequality in
the initial distribution of income and the incidence of child labor, Proposition 2
below, which is also proved to be true Žunder Assumption 3. in Appendix C,
shows the link between initial distribution of income and the steady state distribu-
tion of income.

Proposition 2. If the initial distribution of parental income in economy 2, G2 (b),


dominates the initial distribution of parental income in economy 1, G1(b), in the
second-order stochastic sense, then the steady state distribution of income in
economy 2 dominates the steady state distribution of income in economy 1 in the
first order stochastic sense (and hence in the second-order stochastic sense).

An implication of Propositions 1 and 2 is that the economies which start out


with greater equality have smaller incidence of child labor, greater per-capita
income, and a greater equality in the distribution of income in the steady state. We
will discuss the policy implications of this result in Section 4.

3.2. Changes in factor prices

First we look at the impact of changes in factor prices, holding the initial
distribution of income, GŽ b 0 ., constant, and subsequently we look at the impact of
changes in factor prices when the economy is already in a steady state equilibrium.
P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102 93

We use the relationship wc s u wu to substitute out child wage from Eqs. Ž9. and
Ž12. to get the following.
U Ž Ž 1 q u . wu . s Ž 1 y b . U Ž wu . q b U Ž s ws . Ž 16 .
b U Ž s ws . s Ž 1 y b . U Ž b 0 q u wu . y U Ž b 0 . q b U Ž Ž 1 q u . wu . Ž 17 .
Eq. Ž16., which determines su , implies that ŽEsurEws . - 0. As discussed earlier su
is the threshold level of ability beyond which even an unskilled parent would send
the child to school from period t s 1 onwards. An increase in the skilled wage
lowers this threshold level of ability by making the schooling option even more
attractive. The impact of an increase in the unskilled wage on su is less clear cut,
however. The sign of ŽEsurEwu . depends on the sign of Ž1 q u .U X ŽŽ1 q u . wu . y
Ž1 y b .U X Ž wu . as follows.
Esu
) 0 if Ž 1 q u . U X Ž Ž 1 q u . wu . ) Ž 1 y b . U X Ž wu .
Ewu
Esu
- 0 if Ž 1 q u . U X Ž Ž 1 q u . wu . - Ž 1 y b . U X Ž wu .
Ewu
While an increase in the unskilled wage increases the relative attractiveness of
sending the child to work because future generations are better off by remaining
unskilled compared to the situation before, an increase in the unskilled wage also
increases the parental income of the unskilled households. Since the primary
reason why an unskilled household sends the child to work is low income in the
presence of borrowing constraints, a rise in parental income would induce them to
send their child to school. Therefore, the net effect is ambiguous.
Eq. Ž17. implies that ŽEb ) rEws . - 0 and ŽEb ) rEwu . ) 0. ŽEb ) rEws . - 0 im-
plies that an increase in the skilled wage shifts the threshold curve depicted in Fig.
1 to the left, i.e. for each level of ability the threshold level of parental income is
less than before. This is because an increase in the skilled wage increases the
returns to schooling, which induces more parents to send their children to school
because they care about the well being of their children. The opposite happens for
an increase in the unskilled wage.
Therefore, ŽEb ) rEws . - 0 and ŽEsurEws . - 0 imply from Eq. Ž14. that by
holding the initial distribution of parental income GŽ b 0 . constant, a higher skilled
wage from period t s 0 onwards is associated with lower steady state incidence of
child labor. The impact of a change in the unskilled wage is ambiguous, however.
What happens if there are changes in factor prices after the economy has
reached a steady state equilibrium? Now the impact on the incidence of child labor
depends only on how su is affected. We have already seen that an increase in the
skilled wage reduces su ; therefore, it induces some unskilled parents to send their
child to school. Hence, this tends to decrease the steady state incidence of child
labor. However, since the impact of a rise in the unskilled wage on su is
94 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

ambiguous, the overall impact on the incidence of child labor is ambiguous as


well.

4. Policy implications

4.1. RedistributiÕe policy

An immediate implication of Proposition 1 is that greater equality in the initial


distribution of income is associated with lower incidence of child labor. One has to
be careful, however, in deriving redistributive implications from this result. Any
inequality reducing redistribution of parental income, which does not use informa-
tion on the ability of individuals reduces the incidence of child labor, and
according to Proposition 2 it also reduces the inequality in the steady-state
distribution of income. If a redistributive policy takes into account the ability of
individuals, then even some inequality enhancing redistributions can reduce the
incidence of child labor. For example, redistribution of income between two
parents with equal income who are both below the threshold level of income and
their children have identical ability can reduce the incidence of child labor.
Alternatively, introducing a lottery between two such individuals can also reduce
the incidence of child labor by allowing one of them to overcome the borrowing
constraint. Therefore, the result in Proposition 1 is not valid for redistributive
policies contingent on the ability. In reality, it is extremely difficult for a
policymaker to perfectly observe the ability, and so redistributive policies are
unlikely to be contingent on ability, and in these cases an inequality reducing
redistribution will reduce the incidence of child labor.
Things are different when the economy has already reached a steady-state. In
steady-state children of the unskilled are always sent to work, while the children of
the skilled are always sent to school. Now one has to make a distinction between a
one shot redistribution and a permanently redistributive policy. A one shot
redistribution of income can permanently reduce the incidence of child labor. It
can be shown that a small redistributive tax-subsidy scheme Žsmall proportional
tax and a lump sum subsidy. that leaves the future wages unaffected, will reduce
the incidence of child labor.7 The impact of a permanently redistributive policy is
ambiguous, however, as should be clear from the discussion of the comparative
statics with respect to factor prices. A permanently redistributive policy will
reduce the net of tax skilled wage and increase the net of tax unskilled wage. The
impact of this on the incidence of child labor is ambiguous.

7
A sufficient condition for a redistributive tax-subsidy scheme to reduce the incidence of child labor
is that the average parental income in steady state is greater than sws .
P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102 95

4.2. Impact of trade sanctions

We discuss the impact of trade sanctions on the incidence of child labor when
the economy is in a steady state equilibrium. Since most countries having a high
incidence of child labor are exporters of unskilled labor intensive goods, we
discuss the impact of trade sanctions on the economy having a comparative
advantage in the unskilled labor intensive good. Now, a trade sanction for this
economy will lower the relative price of the unskilled labor intensive good. This
will translate into a lower unskilled wage and a higher skilled wage from the
familiar Stolper–Samuelson relationship between the product prices and factor
prices.
As discussed earlier, an increase in the skilled wage will reduce the incidence
of child labor. The impact of a decrease in the unskilled wage is ambiguous,
however. If ŽEsurEwu . ) 0, then a decrease in the unskilled wage induced by trade
sanction will reduce the incidence of child labor. Therefore, in this case trade
sanction causes an unambiguous decrease in the incidence of child labor. How-
ever, if ŽEsurEwu . - 0, then a trade sanction induced decrease in the unskilled
wage increases su . The net impact on su depends on the relative strengths of the
two opposite forces. Now if the impact of a lower unskilled wage is so strong as to
offset the effect of a higher skilled wage, then su will increase. For Cobb–Doug-
las production functions, the higher the skill intensity of the low-tech sector and
the lower the difference between the skill intensities in the two sectors the greater
the elasticity of unskilled wage with respect to the relative product price. In this
case a small increase in the relative price of the high tech good causes a small
increase in the skilled wage, but a large decrease in the unskilled wage. Therefore,
su may increase in this case. An increase in su , however, has no impact on the
incidence of child labor in a steady-state equilibrium. This happens because in
steady state all parents in households having ability greater than the initial level of
su are skilled, and hence unaffected by a rise in su . Therefore, a trade sanction
will fail to lower the incidence of child labor in this case.
Since a trade sanction may not lower the incidence of child labor, and we know
from trade theory that a policy that restricts trade leads to welfare losses by
preventing specialization, the case for trade sanctions to deal with the issue of
child labor is a weak one. Therefore, we discuss some alternative policies to deal
with the problem.

4.3. Other policies

A policy that has been frequently advocated is to impose a legislative ban on


child labor. If a ban can be enforced, it will be successful in eliminating child
labor. From the discussion of the steady state in Section 2 it is clear that the skilled
parents never send their children to work. Therefore, if compulsory schooling laws
can be enforced for one generation it will become self-sustaining, and in the new
96 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

steady state equilibrium there will be no child labor in the economy. However, it is
clear from the discussion in Section 2.3 that for poor households with b F b ) Ž s .,
given the high marginal utility of present consumption, the optimal decision is to
send the child to work. Since it is not in the interest of households to send the
children to school, a legislative ban is going to run into enforcement problems.
This is consistent with the findings of several studies about the ineffectiveness of
compulsory schooling laws that attempt to send children to school instead of
work.8
Since sending the child to work is an optimal decision for poor parents, a
legislative ban, by putting a constraint on their choices, reduces their welfare.
Therefore, there is no clear-cut argument in favor of legislative bans in our model
on purely welfare grounds. The general equilibrium effect of a ban on the wages
discussed in Basu and Van Ž1998. and Baland and Robinson Ž1999. is absent in
our model because wages are determined by the relative product price, which is
determined in the international market given our small open economy assumption.
Therefore, a decline in the supply of unskilled labor brought about by a ban on
child labor has no positive effect on the unskilled wage in our model.
Any policy that induces poor parents to send their children to school is to be
preferred over policies that coerce them into doing so. A policy in this spirit would
be to provide consumption support to the families of poor children to compensate
for the foregone earnings of the children attending school. This policy would
induce parents to send their children to school, and therefore is not going to run
into enforcement problems encountered in the case of a legislative ban. Bequele
and Myers Ž1995. report several case studies of policy initiatives with a similar
flavor. One example is from Mandsaur Slate Pencil Industry project commissioned
in 1988. The main thrust of the program is to provide education through special
schools imparting non-formal education and vocational training. Students are
given a stipend of Rs. 100 a month as an incentive to the parents to withdraw their
child from work and send him to school. A nutritional meal is also provided at the
school as an added incentive. All four schools under this program show high
enrollment.
In the policy discussion earlier, it was implicitly assumed that the amount of
schooling infrastructure was adequate and there was no direct cost of schooling.
Obviously, if the reason why people send their children to work is a lack of
schooling infrastructure, then the provision of adequate schooling infrastructure
should get primacy in policy making. In general, however, a mere provision of
schooling infrastructure will not be sufficient to eliminate the incidence of child

8
Krueger Ž1996. finds tremendous amount of non-compliance with compulsory schooling laws.
Compulsory schooling laws and minimum age requirements are complementary policies. A non-com-
pliance with one implies a non-compliance with the other. Also, see Jensen and Nielsen Ž1997. for
evidence on Zambia.
P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102 97

labor. Also, to the extent that there are direct costs of schooling in the form of
tuition, books etc. they will increase the incidence of child labor. It can be easily
seen that if we introduce a direct cost of schooling, d, in the model discussed in
Section 2, the threshold level of parental income below which parents send their
children to work will rise, and hence the incidence of child labor will rise. In that
case an effective policy will involve, among other things, eliminating the direct
cost of schooling by distributing free books and other materials.

5. Concluding remarks

Understanding the economics of child labor is important in devising effective


policy initiatives to address this problem. The paper presented an overlapping
generations general equilibrium model to understand the determinants of child
labor, and analyzed some common policy prescriptions in that framework. It
obtained some new results on the link between inequality in the distribution of
income and the incidence of child labor. It also showed why the case for trade
sanctions as a policy tool to reduce the incidence of child labor is a weak one. It
also discussed why a legislative ban on child labor is going to be difficult to
enforce.

Acknowledgements

I am grateful to Garance Genicot, Maitreesh Ghatak, Stergios Skaperdas, and


two anonymous referees for helpful comments. I would also like to thank the
seminar participants at the Indian Statistical Institute, Delhi, University of South-
ern California, and the Far Eastern Meeting of Econometric Society, Singapore,
for discussions and comments.

Appendix A. Proof of Lemma 1

In solving Eqs. Ž7. and Ž8. we have to examine the following four possible
cases.
Case I: When an unskilled parent wants to send the child to work, while a
skilled parent wants to send the child to school.
From Eqs. Ž7. and Ž8. in the text we get the following:
V Ž wu , s . s U Ž wu q wc . q b V Ž wu , s . Ž 18 .
V Ž s ws , s . s U Ž s ws . q b V Ž s ws , s . Ž 19 .
Eqs. Ž18. and Ž19. imply the following values for the value functions.
U Ž wu q wc . U Ž s ws .
V Ž wu , s . s ; V Ž s ws , s . s Ž 20 .
1yb 1yb
98 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

Now we ask the question for what parametric configuration do we get Case I as an
equilibrium? In order for an unskilled parent to send the child to work the value
derived from sending the child to work should exceed the value derived from
sending the child to school. From Eqs. Ž7. and Ž20. it is clear that for an unskilled
parent to send the child to work, the following condition must be satisfied
U Ž wu q wc . b
) U Ž wu . q U Ž s ws . Ž 21 .
1yb 1yb
Note that Eq. Ž21. is satisfied if s - su , where su is the value of s for which the
two sides of Eq. Ž21. are equal. When s s su , an unskilled parent is indifferent
between sending the child to school or work. As a matter of convention, we will
assume that whenever parents are indifferent they send the child to school.
Similarly, from Eqs. Ž8. and Ž20. we can see that in order for a skilled parent to
send the child to school the following must be true
U Ž s ws . b
G U Ž s ws q wc . q U Ž wu q wc . Ž 22 .
1yb 1yb
The concavity of the utility function UŽ C . implies that a sufficient condition for
Eq. Ž22. to be true for all s G s is sws ) Ž wcrb . q wu . This is the Assumption 1
in the text. Therefore, Case I is satisfied for the level of talent s g ws, su ..
Case II: When both skilled and unskilled parents want to send their child to
school.
This implies the following:
V Ž wu , s . s U Ž wu . q b V Ž s ws , s . Ž 23 .
V Ž s ws , s . s U Ž s ws . q b V Ž s ws , s . Ž 24 .
Eqs. Ž23. and Ž24. imply the following.
b U Ž s ws .
V Ž wu , s . s U Ž wu . q U Ž s ws . ; V Ž s ws , s . s Ž 25 .
1yb 1yb
Since skilled parents are sending their child to school as in Case I, it can be easily
shown that in order for an unskilled parent to send the child to school, the
condition required is just the opposite of Eq. Ž21., and hence is satisfied when
s G su .
Using Eqs. Ž8. and Ž25. we see that in order for a skilled parent to send the
child to school the following must be true.
U Ž s ws . b
G U Ž s ws q wc . q b U Ž wu . q U Ž s ws . Ž 26 .
1yb 1yb
Again given Assumption 1, Eq. Ž26. is satisfied whenever s G s. Therefore, Case
II obtains for s G su .
P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102 99

Case III: When an unskilled parent wants to send the child to school, while a
skilled parent wants to send the child to work.
In this case we get the following:
V Ž wu , s . s U Ž wu . q b V Ž s ws , s . Ž 27 .
V Ž s ws , s . s U Ž s ws q wc . q b V Ž wu , s . Ž 28 .
Eqs. Ž27. and Ž28. imply the following
1
V Ž wu , s . s U Ž wu . q b U Ž s ws q wc . Ž 29 .
1yb 2
1
V Ž s ws , s . s U Ž s ws q wc . q b U Ž wu . Ž 30 .
1yb 2
Using ŽEqs. Ž7., Ž8., Ž29. and Ž30. we can write the conditions required for this
case to obtain as follows.
U Ž wu . q b U Ž s ws q wc . G Ž 1 q b . U Ž wu q wc . Ž 31 .
U Ž s ws q wc . q b U Ž wu . ) Ž 1 q b . U Ž s ws . Ž 32 .
It can be easily seen that Assumption 1 implies that Eq. Ž32. can never be satisfied
for any s G s. Therefore, this case never obtains.
Case IV: When both skilled and unskilled parents want to send their child to
work.
From the discussion of Case I, it is obvious that whenever an unskilled parent
wants to send the child to work a skilled parent wants to send the child to school
as long as Assumption 1 is satisfied. Therefore, Case IV never obtains.
From the discussion above it is clear that under Assumption 1 only Cases I and
II are possible and the result from those two cases is summarized in Lemma 1 in
the text.

Appendix B. Proof of Proposition 1

We prove this proposition assuming interior solution for su , i.e. su g Žs, s .. If


su ) s , then su in the proof should be replaced with s everywhere. Everything
else remains same. Write the difference in the incidence of child labor under two
distributions using Eq. Ž14. as follows.
su
LC1 y LC 2 s Hs G1Ž b ) Ž s . . y G 2 Ž b ) Ž s . . f Ž s . d s Ž 33.

Next, let b ) Ž s . s b. This implies s s b ) Ž b .


forming the change of integral we get
´ ds s b )y 1X Ž
b .d b. Further, per-

Hbb Ž ss.
. X X
LC1 y LC 2 s y G 1 Ž b . y G 2 Ž b . b )y 1 Ž b . f Ž b )y1 Ž b . . d b Ž 34 .
)
u
100 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

Eq. Ž34. can be written as

LC1 y LC 2
)Ž b
Hbb Ž ss.b
. )y 1X X
sy
)
u
Ž b . f Ž b )y1 Ž b . . d ½H b ) Ž su .
G1Ž y . y G 2 Ž y . d y d b 5
Ž 35 .
Integrating Eq. Ž35. by parts treating b )y 1X Ž b . f Ž b )y 1 Ž b .. as the first term and
d Hbb) Ž s u .w G 1Ž y . y G 2 Ž y .xd y4 as the second term we get

LC1 y LC 2
b ) Žs .
X b
sy b )y 1
Ž b. f Ž b )y1
Ž b. . ½H b ) Ž su .
G1Ž y . y G 2 Ž y . d y 5 b)Žs u.

)Ž b
Hbb Ž ss. Hb
. Y
q
)
u
ž )
Ž su .
G1Ž y . y G 2 Ž y . d y b )y 1 Ž b . f Ž b )y1 Ž b . .

X 2
qf X Ž b )y1 Ž b . .Ž b )y1 Ž b . .
/ db Ž 36 .

Further simplifying Eq. Ž36., we obtain


X b Žs .
LC1 y LC 2 s y b )y 1 Ž b ) Ž s . . f Ž s . Hb G1Ž y . y G 2 Ž y . d y
)
Ž su .

)Ž b
Hbb Ž ss. Hb
.
q
)
u
ž )
Ž su .
G1Ž y . y G 2 Ž y . d y

Y X 2
= b )y1 Ž b . f Ž b )y1 Ž b . . q f X Ž b )y1 Ž b . .Ž b )y1 Ž b . .
/ db

Ž 37 .
Next we show that under Assumption 3, which we recall is U Z Ž C . ) 0, b ) Ž su .
F b, and f X Ž s . G 0, the expression in Eq. Ž37. is non-negative. Looking at the
first term on the right hand side in Eq. Ž37., we know that b )y 1X - 0. Further,
b ) Ž su . F b implies that we can use the definition of second-order stochastic
)
dominance to get Hbb) ŽŽs u . s. w G 1Ž y . y G 2 Ž y .xd y G 0. Therefore, the first term is
non-negative.
As discussed in the text U Z ) 0 implies b )Y Ž s . ) 0, which in turn implies
)y 1Y Ž .
b b ) 0. Given this, a sufficient condition for b )y 1Y Ž b . f Ž b )y 1 Ž b .. q
X
f Žb )y 1 Ž b ..Ž b )y 1 Ž b .. 2 ) 0 is f X Ž s . G 0. Again, given b ) Ž su . F b, the definition
of second-order stochastic dominance implies Hbb) Ž s u .w G1Ž y . y G 2 Ž y .xd y G 0.
P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102 101

Therefore, the second term is non-negative as well. Thus, we get LC1 y LC 2 G


0.Q. E. D

Appendix C. Proof of Proposition 2

Let us denote the steady state distributions in the two economies by G 1Ž z . and
G 2Ž z .. We know from the discussion in the text that the support of the steady state
distribution is w wu , s ws x. We want to show that G1Ž z . G G 2Ž z .,; z g w wu , s ws x.
The mass of people for whom z s wu is simply the fraction of children going to
work. Therefore, from Proposition 1 above it is obvious that G 1Ž wu . G G 2Ž wu ..
Also, the fraction of population earning greater than su ws is independent of the
initial distribution of income, and is given by 1 y F Ž su .. Therefore, G 1Ž su ws .
s G 2Ž su ws .. Similarly, for any z G su ws , G1Ž z . s G 2Ž z .. So, we need to show
G 1Ž z . G G 2Ž z . for z F su ws . For z g w wu , su ws x we have

G1 Ž z . y G 2 Ž z .
z
su ws
s Hs G1Ž b ) Ž s . . f Ž s . d s q Hs 1 y G1Ž b ) Ž s . . f Ž s . d s

z
su ws
y Hs G2 Ž b ) Ž s . . f Ž s . d s y Hs 1 y G 2 Ž b ) Ž s . . f Ž s . d s Ž 38 .

The above can be simplified as


su
G1 Ž z . y G 2 Ž z . s Hz G1Ž b ) Ž s . . y G 2 Ž b ) Ž s . . f Ž s . d s Ž 39 .
ws

Eq. Ž39. above is similar to Eq. Ž33. in the proof of Proposition 1. The only
difference is that s has been replaced by zrws . Therefore, b ) Ž s. from Eq. Ž34. l

onwards is replaced by b ) Ž zrws . and rest of the proof remains the same, and
hence we get G 1Ž z . y G 2Ž z . G 0.Q. E. D.

References

Baland, J.M., Robinson, J., 1999. Is child labor inefficient. forthcoming. Journal of Political Economy.
Basu, K., 1999. Child labor: causes, consequence, and cure, with remarks on international labor
standards. Journal of Economic Literature 37, 108–1119.
Basu, K., Van, P.H., 1998. The economics of child labor. American Economic Review 88, 412–427.
Becker, G.S., Murphy, K., 1988. The family and the state. Journal of Law and Economics 31, 1–18.
Bequele, A., Myers, W.H., 1995. First Things First in Child Labor: Eliminating Work Detrimental to
Children. UNICEF, ILO, Geneva.
102 P. Ranjanr Journal of DeÕelopment Economics 64 (2001) 81–102

Bhagwati, J., 1995. Trade liberalization and fair trade demands: addressing the environmental and labor
standard issues. World Economy 18, 745–759.
Chiu, W.H., 1998. Income inequality, human capital accumulation and economic performance.
Economic Journal 108, 44–59.
Galor, O., Zeira, J., 1993. Income distribution and macroeconomics. Review of Economic Studies 60,
35–52.
International Labour Organization, 1996. Child Labour Surveys: Results of Methodological Experi-
ments in Four Countries, 1992–93. International Programme on the Elimination of Child Labour
ŽIPEC.. ILO, Geneva.
Jensen, P., Nielsen, H.S., 1997. Child labour or school attendance? evidence from Zambia. Journal of
Population Economics 10, 407–424.
Krueger, A., 1996. Observations on international labor standards and trade. In: Bruno, M., Pleskovic,
B. ŽEds.., Annual Bank Conference on Development Economics, The World Bank, Washington,
D.C.
Loury, G.C., 1981. Intergenerational transfers and the distribution of earnings. Econometrica 49,
843–867.
Nerlove, M., Razin, A., Sadka, E., 1988. A bequest constrained economy: welfare analysis. Journal of
Public Economics 37, 203–220.
Parsons, D., Goldin, C., 1989. Parental altruism and self-interest: child labor among late nineteenth-
century american families. Economic Inquiry 27, 637–659.
Psacharopoulos, G., 1994. Returns to investment in education: a global update. World Development 22,
1325–1343.
Ranjan, P., 1999. An economic analysis of child labor. Economics Letters 64, 99–105.
Rodrik, D., 1995. Labor Standards and International Trade: Moving Beyond the Rhetoric. Overseas
Development Council.
Rogers, C.A., Swinnerton, K., 1999. Inequality, Productivity, and Child Labor. Georgetown University
Working Paper.
Swinnerton, K., Rogers, C.A., 1999. The economics of child labor: comment. American Economic
Review 89, 1382–1385.

Das könnte Ihnen auch gefallen