Past is Prologue: Inference from the Cross Section of Returns Around an Event
Abstract
Confounding events can cause false positives when testing the relationship between short-term returns around a quasi-experimental event and firm characteristics. We show that this risk is severe in practice: return-characteristic relationships are often statistically significant at the 1% level on over 30% of all trading days. Benchmarking a relationship against the distribution of the same relationship on pre-event days is effective at addressing the problem. We introduce a novel GLS variation of this approach that achieves large gains in statistical power relative to OLS and provide Stata and Python modules that implement both procedures.
Cite
@article{cohn2026past,
title = {Past is Prologue: Inference from the Cross Section of Returns Around an Event},
author = {Jonathan Cohn and Travis Johnson and Zack Liu and Malcolm Wardlaw},
journal = {Journal of Financial Economics},
year = {2026},
doi = {10.1016/j.jfineco.2026.104278}
}