Causal forest estimation of heterogeneous and time-varying environmental policy effects

Empirically evaluating environmental policies requires grappling with impacts that exhibit not only cross-sectional heterogeneity, but also variation across time. Phased policy roll-outs offer opportunities for improvement across cohorts and policy effects can grow or decay, especially when natural processes are involved. Focusing on a subset of these factors can lead to erroneous inference, while considering them jointly magnifies specification challenges. To address these challenges, I extend and apply causal forests, a nonparametric method for estimating heterogeneous treatment effects, to simultaneously examine how effects vary across time. I first adapt causal forests to a panel setting with staggered policy introduction by incorporating dynamic selection assumptions and estimators. After illustrating the method's performance on simulated data, I use it to reanalyze how individual quota programs have affected fisheries catches around the world. Estimates reveal substantial heterogeneity and time dependencies and suggest that longer policy exposure may be less beneficial than previously thought. More generally, the approach has potential value for evaluating impacts of a range of environmental policies as well as environmental shocks.