Detecting Trading Response Using Transaction-Based Research Designs.

This paper examines issues pertinent to using number of transactions in event-type studies of market trading response. It presents rejection frequencies for portfolios of 20, 50, and 100 observations corresponding to different abnormal trading measures and test approaches. Findings include: (1) A cross-sectional t-statistic approach performs almost as well as more elaborate test approaches and is robust to nonnormality in the number of transactions distribution; (2) a logarithmic transformation induces near-normality in the number of transactions distribution, and (3) for equal induced percentage increases in trading, rejection rates for number of transactions exceed those for percentage of outstanding shares trades (i.e., volume). The advantage of transaction over volume is relatively more important for small sample studies (e.g., n Copyright 1995 by Kluwer Academic Publishers