New working paper with Jakob Vanschoonbeek: When agents are forward-looking, 'recentering' methods may remove the very treatment effects researchers want to estimate. We show this empirically in China's high-speed rail expansion and propose an alternative instrument. papers.ssrn.com/sol3/papers....
Borusyak & Hull (2023, BH) develop a recentering approach to identify causal treatment effects when treatments are randomly allocated but units differ in th