Detecting Abnormal Bid-Ask Spread: A Comparison of Event Study Methods.

This study examines empirical issues associated with the use of bid-ask spreads in event studies. The simulation results indicate that the distribution of average standardized abnormal spread shows little deviation from normality. Simulation results also indicate that the widely used percent spread metric results in test statistics with low power. In contrast, use of a standardized raw spread metric and a simple mean-adjusted expectation model results in well specified and reasonably powerful Patell and Brown-Warner type test statistics. As the abnormal spread series is characterized by high first order serial correlation, it is important to adjust for this serial correlation when using multi-day event windows. Copyright 2000 by Kluwer Academic Publishers