Abstract
Event studies have become pivotal in securities fraud litigation. The typical use-cases are single firm / single event (SFSE) applications. These differ from the standard event study methodology in that inference on abnormal returns can only be based on the time-series variance of abnormal returns over a calibration period. We analyze robust inference in the SFSE setting using Monte Carlo and resampling experiments. We find that White- or Newey-West-correction of standard errors generally cannot be applied in the SFSE setting, but even extreme heteroscedasticity does not bias OLS inference. The one, but in applications crucial, problem of inference arises when calibration and event period occur in different regimes. To identify such situations, we develop a specification test for the presence of a structural break in return volatility. Finally, we show that GARCH estimation using intraday data is a suitable and, in most instances, preferred approach to conduct SFSE analyses.