This dissertation comprises of three chapters on the economics of harmful child labor. The first chapter is theoretical. Using household survey data from the Philippines, the second and third chapters examine whether a model assumption and an explanation for a model prediction in the first chapter are supported empirically. In Chapter 1, I model the labor market and welfare effects of banning harmful child labor. The effects are examined in two informational cases: (1) the parent has perfect information on harmful child labor and (2) the parent does not. The effects under both cases are contrasted between when the parent is the welfare evaluator for the household and when the child is, given imperfect parental altruism. Under both informational cases, the ban generates re-equilibrating labor market adjustments that expand employment and reduce wages in the non-harmful child labor market, as well as reduce child labor force participation. Under the first case, the ban is welfarereducing. Under the second case, it is generally welfare-reducing; under special conditions, it can be welfare-improving. Under both informational cases and when the child is the welfare evaluator, a ban is generally welfare-ambiguous; under special conditions, it can be welfare-improving. In Chapter 2, I examine the existence and magnitude of positive compensating wages for harmful child labor. Among the various harmful child labor measures examined, I find consistent evidence of a large and significant earnings premium for physically-strenuous labor at both the conditional mean and median. The result at the conditional mean is largely driven by the large and increasing premia as one moves down the lower half of the conditional earnings distribution. In Chapter 3, proxying the asymmetry in preferences and power statuses between the parent and child by the contradictory response of the parent when the child reports a work injury or illness, I examine whether parent-child injury report mismatches have an impact on the probability of harmful child labor. I find consistent evidence that mismatches have a large and significant positive effect on the probability of harmful child labor.