UC Berkeley

This dissertation attempts to answer three critical questions that have remained largely misunderstood in the literature of child labor. The first question is whether child labor can help child laborers gain more human capital, including both formal education and health status. The second question focuses on the mechanisms through which child labor impacts human capital. It asks how a positive causal impact from child labor to human capital can possibly take place. The third question discusses policy implications. Given the gain in human capital of child laborers due to child labor, what are the unintended consequences of current policies and what can we do to effectively combat child labor and at the same time help child laborers acquire more human capital? Because these three questions are intrinsically related I find it more productive to present them in form of one major study rather than in three separate papers.

To provide empirical evidence to the first question of whether child labor can help child laborers gain more human capital, I exploit a quasi-controlled experiment that took place between 2004-2009 in a poor rural area in Vietnam. Most children in this area were so poor that they dropped out of school prematurely. In order to help these children sustain their education, a non-governmental organization (NGO) decided to provide a cow to each poor household with school-aged children so that the children could spend time tending the cows, earn some income to pay for their schooling. Practically this intervention provided the children a means to convert their time into income.

Due to limited resources, the NGO could provide cows to only a subset of the poor children, effectively creating a controlled experiment in which some of the children had work (the treatment group) and the others did not (the control group). Since the children were not randomized into the treatment and control groups, the main concern was the selection bias. An examination of the bias shows that the children were selected into the treatment group on the basis of most urgent needs - which means those determined to be more likely to drop out of school in absence of the treatment were selected to receive the cows. The data collected verified that at baseline those in the treatment had indeed acquired less education, had higher dropout ratio, were poorer, had less land, lived further away from school, and their parents had lower levels of education. All of these socio-economic indicators suggest that the selected children would have been more likely to drop out of school if the status-quo had continued. Since the selection bias (being more likely to drop out) works against the treatment effect (acquiring more education), estimates of the impact are likely to be the lower bounds of the true effect and should be valid. I find that the poor children who worked gained a significant average of 0.59 years of education over a period of 5 years compared to those who did not have any work opportunity.

While the finding of a positive causal relationship between child labor and education is striking, this outcome per se is not very useful in terms of proposing new policy interventions because it does not explain how child labor results in more human capital. Imagine even if we have the luxury of running a perfect randomized controlled trial and the experiment shows that child labor leads to an increase in human capital, there remains a "black box". We still cannot explain how the positive impact takes place. Clearly unless we can explain what happen in the black box - unless we can explain with economic theories how child labor can positively affect human capital, we cannot construct informed policy interventions.

This immediately leads to the second question: what are the mechanisms through which child labor can result in more human capital? To answer this question, I construct a theoretical model which examines how a household would choose optimal levels of human capital under the treatment (where children can work) and under the control (where children are not allowed to work). This framework shows that child labor affects child laborers' human capital through a positive income effect and a negative time cost. On one hand, child work brings home more income to acquire more education and consumption (a positive impact on health status). On the other hand, child labor takes away time, a necessary input for schooling. The most important finding is that while child labor always generates a positive income effect, its opportunity cost of time in terms of the education forgone can be zero, leading to a positive net effect.

To see this, consider a household's choices of child labor and human capital as in a controlled experiment. Note that under the control when child labor is not allowed, the optimal level of schooling can be zero. For example, a hungry family that can afford only one meal per day would choose zero schooling, a costly expenditure in poor countries, in order to spend all income on food. In this case, the time cost of child labor in terms of forgone schooling is zero because in the absence of work children stay home anyway. When these hungry children can work, there are only two possibilities. First, they might choose to work full-time and spend all additional income on food. This case would lead to an increase in health status of the child laborers with no change in their education. Second, they might choose to work part-time and go to school part-time, using their additional income to pay for more food and more schooling. This case would lead to an increase in both their health status and education. The model shows that the income effect of child labor can dominate the time cost (because it can be zero), resulting in a positive net effect on human capital. The critical point that separates this research from the literature is that I use the amount of school time that would be chosen in the absence of work as the benchmark, not school time endowment, to measure the time cost of child labor.

New answers to question 1 and 2 immediately bring up question 3: what are the unintended consequences of current policies and what can we do to effectively combat child labor and promote human capital? I find that current interventions such as trade sanctions, consumer boycotts, legal penalties or an outright ban against child labor, which would diminish or eliminate child work opportunity, would unambiguously reduce the human capital of the poorest child laborers. Note that child labor restricting policies are grounded on the belief that a loss in household welfare due to the loss of child labor income can be offset by an increase in child schooling due to reduced child labor. However, this study shows that by restricting child labor, these policies would reduce not only household welfare but also the schooling and physiological capital of the poorest children. Such instruments will have unintended consequences on children's human capital, the very point that they advocate for.

Instead I find policies that make schooling more affordable such as such as reducing school fees, providing free meals and textbooks, providing cash transfer conditional on schooling, improving teacher/student ratios, improving curriculum, would simultaneously increase education and reduce child labor. In addition, I suggest new market-based policies that can enhance child laborers' education and health status without consuming additional public resources more than the status-quo. For example, encouraging the private sector to provide work to children with unemployed non-school time conditional on their school attendance would maximize education gain by capturing the income effect while excluding the substitution effect.

My research adds to the literature in a number of ways. This is the first study to provide empirical evidence that child labor can lead to an increase in the human capital of child laborers. Moreover, this research is also the first to provide a theoretical framework that explains the mechanism at work - that is child laborers can gain more human capital from working because child labor always generates a positive income effect while its opportunity cost of time in terms of forgone education can be zero. Most importantly, my work suggests a need for a major overhaul of current policies that are adversely affecting hundreds of million poor children around the globe. Reducing child labor by enforcing interventions that restrict child work opportunities will have the exact unintended consequences of reducing the human capital of the poorest child laborers. The best way to get the more than 100 million hungry children worldwide out of work is to subsidize schooling (i.e. even pay them to go to school). Such a subsidized education in poorest countries, however, more often than not is practically out of the question. In this situation, the hungry children need, not less, but more work opportunities to buy more food, and at times also buy more education.