The power of tests employing log-transformed volume in detecting abnormal trading
This paper shows that the simulation-based rejection percentages for detecting abnormal log-transformed volume reported in Ajinkya and Jain [AJ] (1989) are sensitive to the method of inducing abnormal volume. We present an alternative inducement method that possesses desirable distributional properties. Under this method, for example, with fifty firm portfolios and one-day event periods a 20% volume increase is detected just over 40% of the time, while AJ suggest a detection rate of over 90%. Rejection percentages for abnormal volume at earnings announcement dates also are more consistent with our alternative method than with AJ's method.