Tree-based Estimation of Heterogeneous Dynamic Policy Effects

• Environmental and Energy Economics Workshop

Date and Location

Wednesday, October 17, 2018, 3:30 PM - 5:00 PM
ARE Library Conference Room, 4101 Social Sciences and Humanities

Abstract

Understanding variation in policy eects in both cross-sectional and temporal dimensions can help improve policy targeting, set expectations regarding timing of eects, and assess whether policy implementation has improved over time. To that end, this paper extends recent work adapting regression trees and random forests for identification of heterogeneous treatment effects. Specifically, I adapt those methods to a panel setting, illustrating how dynamic selection assumptions and estimators can be combined with causal forest approaches to jointly investigate not only cross-sectional heterogeneity, but also dynamics of policy effects and changes in implementation effectiveness. To illustrate and evaluate the utility of the approach, I reanalyze how property rights impact the probability of natural resource collapse, finding substantial heterogeneity and nonlinear time dependencies not identified in prior work.