Working papers
- Extrapolation of Treatment Effect Estimates Across Contexts and Policies: An Application to Cash Transfer Experiments, March 2024
- Predicting the effects of a new policy often relies on existing evidence of the same policy in other contexts. Such cross-context extrapolation yields accurate predictions if local factors are orthogonal to the policy effects. When the policy effects are heterogeneous in local factors, however, one can alternatively make predictions using adjacent policies within the same context, thereby holding local factors constant. I examine this trade-off between extrapolation across contexts and policies, using cash transfer programs in Malawi and Morocco. By predicting the treatment effect of Moroccan CCTs on school enrollment rates with either Malawian CCTs or Moroccan labeled cash transfers (LCTs), I show that predictions based on the Moroccan LCTs are more accurate than the Malawian CCTs. Through the lens of my structural model, I find that perceived returns to schooling relative to outside options explain the prediction differences. My model also suggests that differences in the underlying mechanisms between the Malawian and Moroccan CCTs are reflected in the cross-contexts variation of the relative returns.
- How the Political Power of Teacher Unions Affects Education (joint with Eduardo Campillo Betancourt), June 2024
- Teacher unions play an important role in determining the quality of public education, especially when they have political power. However, the effects of teacher unions on education are theoretically ambiguous and empirical evidence is limited, particularly in developing countries. This paper studies how politically powerful teacher unions affect public education, focusing on the largest corporatist teacher union in Mexico and a performance-pay program regarded as an union’s patronage tool for rewarding teachers based on their electoral support. We show that the number of public secondary school teachers who got promoted in the program increased in the municipalities supporting the union-affiliated candidate during the 2006 presidential election, compared to less supportive municipalities, after that election. However, we also show that the increased promotion was not associated with improved learning outcomes. Combined with robustness checks, our results suggest that the implementation of the program was distorted when the union gained political influence through the alliance with the ruling party.