Searching for heterogeneous treatment effects using generalized random forest
One approach to estimate the treatment effect is the generalized random forest method of Athey et al. (2019). I apply this method on the dataset of Borcan et al. (2017) to investigate the possibility of additional heterogeneous effects. Borcan et al. (2017) estimated the effect of the anticorruption campaign on education in Romania. In the aftermath of an anticorruption campaign, mostly poor students were negatively impacted by the policy. In my research I take all available variables into account to access which variable has been mostly affected by the anticorruption campaign. I find that in addition to a significant heterogeneous effect for poor students there is also heterogeneous effects for the (expected) number of students taking the Baccalaureate exam, county’s trust score for justice and unemployment rate. Due to the additional heterogeneous findings, both policy makers and researchers should be cautious when using parametric estimation methods for policy inference as it might not capture the heterogeneous effects well enough.