Abstract: We investigate the impact of an increase in deportations on deportees’ home-country firms by leveraging Secure Communities (SC) as a natural experiment. We analyze changes in Mexican firm outcomes using the variation in deportee intensity over time and geographies caused by the staggered adoption of SC across US-counties and the strength of the migrant network. We uncover that SC increased firm creation and the probability of firm survival in Mexico. These results are driven by an increase in local market size. We also find that SC caused an increase in exports, suggesting long-term productivity increases. 

Description: We investigate how local business cycles in the United States are transmitted to Mexican communities via migrant networks. We also aim to present empirical evidence that, conditional on housing construction, the local economic cycle can change the dynamic network due to the high elasticity of migration in terms of wage differentials among undocumented migrants. As a result, economic geography models must account for network dynamics that change often in a short period of time.

Abstract: Recent economics studies show a keen interest in finding heterogeneous treatment effects using machine learning methods. Whether treatment effects depend on covariates, i.e., the conditional average treatment effect (CATE), is a key empirical question. A popularized approach consists of obtaining consistent estimation and inference on features of the CATE rather than the CATE itself, as it avoids relying on strong assumptions needed for accuracy of machine learning methods. We adopt this approach and focus on a particular problem faced by practitioners; the optimal group selection for heterogeneous effects. We suggest a method based on the unsupervised learning algorithm k-means and show an application in a recent microcredit RCT.