Estimating the Impacts of Child Labor on Schooling in Tanzania

Besufekad Alemu, University of Minnesota

Research that looks at the impacts of child labor on schooling proposes multiple ways in which the causality between schooling status (enrollment, interrupted schooling, and grade for age minus grade) and hours worked can be disentangled. While panel data methods are preferred to cross-sectional methods, there is still much difficulty in inferring causality even with panel data methods. This paper analyzes the problems of inferring causality associated with the different cross-sectional and panel estimation methods using household data from Tanzania. The results show that a child level fixed effects framework that utilizes an instrumental variable approach has the most promise. Nonetheless, this model is not without limitations, especially in the face of weak instruments.