From a normative perspective, we can all agree that child labor is reprehensible and should be banned, particularly in its worst forms. According to the International Labor Organization (ILO), child labor has declined worldwide in the last fifteen years, but the numbers are still alarming. In 2012, 168 million children were a part of child labor with more than half of these kids involved in hazardous work.
A number of studies provide theoretical support for such a normative concern (Dessy and Knowles 2007), but this is not a completely undisputed matter. In a seminal paper, Basu and Van (1998) develop a model that rationalizes parents’ decision to send their children to work. The model allows for multiple equilibria, with or without child labor. Because all equilibria are Pareto optimal, they cannot be ranked. Based on that, the authors argue that governments could put forth a child labor ban policy to shift the economy from an equilibrium with child labor to an alternative equilibrium without child labor. The argument is embedded into the Second Welfare Theorem (under certain assumptions, an alternative Pareto allocation can be reached with some redistributive policy).
Baland and Robinson (2000) develop a two-period model and demonstrate that there could exist an efficient level of child labor in the economy. They also show that a policy implementing a ban could have redistributive consequences by affecting adult wages in both the short and long run.
Even though ban policies on child labor have been advocated by the ILO, we know very little about the effectiveness of these policies in reducing child labor. To date, rigorous evidence of the impact of child labor bans is still scant and limited to the US and the findings point to the limited effect of these policies in reducing child labor (Moehling 1999; Margo and Finegan 1996; and Lleras-Muney 2002).
Two recent studies attempted to say something about the impact of such policies in developing countries (Edmonds and Shrestha, 2012; Bharadwaj et al., 2013). The findings are not too encouraging. In fact, Bharadwaj et al. found that the Indian ban policy of 1986 increased child labor, a result that is consistent with prior theoretical predictions (Basu 2005). These findings may raise concerns regarding the effectiveness of such policies when households that rely on child labor might face multiple constraints.
If child labor is largely a phenomenon of poverty, any attempt to ban it through an enforceable minimum employment age policy could potentially be either innocuous or counterproductive. In addition, if the policy is well enforced only in the formal sector, a ban policy could increase participation in the informal sector as was seemingly the case in India.
A change in law in Brazil gave us the opportunity to explore these issues using a sharper identification strategy than was the case with previous studies. In 1998, Brazil passed a constitutional amendment increasing the minimum employment age from 14 to 16. Because Brazil is an economy with a very large informal sector –of nine percent of children aged 14 working in 1999, only a 1.3 percentage point worked as a registered worker, i.e., formally – one could question the effectiveness of such a policy in reducing child labor.
In this short paper we provide some preliminary estimates for the short run effect of the ban on the labor force participation rate of boys and girls. We use a difference-in-differences approach that is actually akin to a parametric regression discontinuity design (RDD) as we restrict our sample to a narrow age interval using the exact date of birth of a particular individual. We estimate a linear probability model with three outcome variables: participation rate, participation rate in the formal sector, and participation rate in the informal sector.
Our estimates point to a reduction in child labor of four percentage points for boys. In relative terms the effect amounts from 36 to 80 percent depending on the comparison group. Interestingly, the effect is almost fully driven by a reduction in participation rates in the informal sector, which suggests either that the law was enforced to some extent in the informal sector and/or that employers fearing being fined decided to not employ children under the age of 16. We found no impact on girls and we believe that this might be explained by the fact that, in our sample, girls participate much less than boys in the labor force (four percent as opposed to 11 percent).
In an upcoming working paper (here is a preliminary version of the paper and here you find the slides of a recent presentation) we look at the long-run consequences of the ban. Using skin color as a proxy for socio-economic background, we split the sample between white and non-white males. Our RDD estimates suggest that white males seem to have benefited from the ban as they ended up with higher education options and wages.
For non-white males our estimates indicate that they are less likely to be employed and to have a formal job. We interpret these findings as an indication that the labor market experience might have higher returns for disadvantaged youth (non-whites) than alternative options, such as working in the informal sector or attending poor quality schools.
All these results have direct policy implications. For instance, they show that the effectiveness of the ban depends on a child’s gender and socio-economic background, and that the ban may have long-lasting and redistributive consequences by increasing the wage gap between the economically better off and worse off households.
A number of studies provide theoretical support for such a normative concern (Dessy and Knowles 2007), but this is not a completely undisputed matter. In a seminal paper, Basu and Van (1998) develop a model that rationalizes parents’ decision to send their children to work. The model allows for multiple equilibria, with or without child labor. Because all equilibria are Pareto optimal, they cannot be ranked. Based on that, the authors argue that governments could put forth a child labor ban policy to shift the economy from an equilibrium with child labor to an alternative equilibrium without child labor. The argument is embedded into the Second Welfare Theorem (under certain assumptions, an alternative Pareto allocation can be reached with some redistributive policy).
Baland and Robinson (2000) develop a two-period model and demonstrate that there could exist an efficient level of child labor in the economy. They also show that a policy implementing a ban could have redistributive consequences by affecting adult wages in both the short and long run.
Even though ban policies on child labor have been advocated by the ILO, we know very little about the effectiveness of these policies in reducing child labor. To date, rigorous evidence of the impact of child labor bans is still scant and limited to the US and the findings point to the limited effect of these policies in reducing child labor (Moehling 1999; Margo and Finegan 1996; and Lleras-Muney 2002).
Two recent studies attempted to say something about the impact of such policies in developing countries (Edmonds and Shrestha, 2012; Bharadwaj et al., 2013). The findings are not too encouraging. In fact, Bharadwaj et al. found that the Indian ban policy of 1986 increased child labor, a result that is consistent with prior theoretical predictions (Basu 2005). These findings may raise concerns regarding the effectiveness of such policies when households that rely on child labor might face multiple constraints.
If child labor is largely a phenomenon of poverty, any attempt to ban it through an enforceable minimum employment age policy could potentially be either innocuous or counterproductive. In addition, if the policy is well enforced only in the formal sector, a ban policy could increase participation in the informal sector as was seemingly the case in India.
A change in law in Brazil gave us the opportunity to explore these issues using a sharper identification strategy than was the case with previous studies. In 1998, Brazil passed a constitutional amendment increasing the minimum employment age from 14 to 16. Because Brazil is an economy with a very large informal sector –of nine percent of children aged 14 working in 1999, only a 1.3 percentage point worked as a registered worker, i.e., formally – one could question the effectiveness of such a policy in reducing child labor.
In this short paper we provide some preliminary estimates for the short run effect of the ban on the labor force participation rate of boys and girls. We use a difference-in-differences approach that is actually akin to a parametric regression discontinuity design (RDD) as we restrict our sample to a narrow age interval using the exact date of birth of a particular individual. We estimate a linear probability model with three outcome variables: participation rate, participation rate in the formal sector, and participation rate in the informal sector.
Our estimates point to a reduction in child labor of four percentage points for boys. In relative terms the effect amounts from 36 to 80 percent depending on the comparison group. Interestingly, the effect is almost fully driven by a reduction in participation rates in the informal sector, which suggests either that the law was enforced to some extent in the informal sector and/or that employers fearing being fined decided to not employ children under the age of 16. We found no impact on girls and we believe that this might be explained by the fact that, in our sample, girls participate much less than boys in the labor force (four percent as opposed to 11 percent).
In an upcoming working paper (here is a preliminary version of the paper and here you find the slides of a recent presentation) we look at the long-run consequences of the ban. Using skin color as a proxy for socio-economic background, we split the sample between white and non-white males. Our RDD estimates suggest that white males seem to have benefited from the ban as they ended up with higher education options and wages.
For non-white males our estimates indicate that they are less likely to be employed and to have a formal job. We interpret these findings as an indication that the labor market experience might have higher returns for disadvantaged youth (non-whites) than alternative options, such as working in the informal sector or attending poor quality schools.
All these results have direct policy implications. For instance, they show that the effectiveness of the ban depends on a child’s gender and socio-economic background, and that the ban may have long-lasting and redistributive consequences by increasing the wage gap between the economically better off and worse off households.
Join the Conversation