Abstract:

This talk reviews recent advances in causal inference, focusing on uncovering treatment effect heterogeneity. Uncovering sources of effect heterogeneity is key for understanding the implications of the distribution of scarce resources, effectively assigning treatments to maximize average outcomes, and generalizing to populations beyond those under study. The talk focusses on propensity and covariate-based partitioning strategies and tree-based methods for assessing effect heterogeneity. To demonstrate methods for assessing effect heterogeneity, I draw on an empirical application of estimating the heterogeneous effects of completing a four-year college degree in the U.S. on reducing socioeconomic disadvantage over the career.