Causal Forests for Heterogeneous Policy Effects in Federations

Wang et al. (2024) highlight DiD’s vulnerability to heterogeneous treatment effects in staggered rollouts. This research leverages causal forests—an extension of Kreif & DiazOrdaz’s (2019) ML toolkit—to map effect heterogeneity across subnational units (e.g., U.S. states). Unlike standard ATE estimates, causal forests identify which institutional features (e.g., gubernatorial power, judiciary independence) amplify or dampen policy impacts (e.g., Medicaid expansion). By combining Tortú et al.’s (2020) network interference approach with subgroup discovery, we resolve conflicts in federal policy studies (e.g., why some states benefit from environmental regulations while others resist). This directly challenges the "one-size-fits-all" assumption in DiD designs.

References:

1. Advances in Difference-in-differences Methods for Policy Evaluation Research. Guangyi Wang, Rita Hamad, Justin S. White (2024). Epidemiology.
2. Machine learning in policy evaluation: new tools for causal inference. N. Kreif, Karla DiazOrdaz (2019). Oxford Research Encyclopedia of Economics and Finance.
3. Modelling Network Interference with Multi-valued Treatments: the Causal Effect of Immigration Policy on Crime Rates. C. Tortú, L. Forastiere, I. Crimaldi, F. Mealli (2020).