Event Study

An event study measures the causal effect of a discrete event by comparing the actual outcome in a post-event window to the outcome the pre-event model would have predicted. The difference is the abnormal return $AR_t = y_t - \hat y_t$. Significant abnormal returns concentrated around the event date are evidence that the event itself caused them.

When to use

Event studies are everywhere in finance (M&A announcements, regulatory actions, earnings surprises) and increasingly in environmental and policy economics. The PS_3 HNC analysis is the textbook environmental application: estimate a pre-1989 model of log_nox on weather + season + trend, predict it forward, then check whether actual post-policy pollution deviates from the prediction. Validity rests on the Estimation Period being long enough to estimate the baseline reliably and free of anticipation effects, and on the predictor set being rich enough to capture "business as usual."