Bond Market Event Study Methods

Issues in the construction of corporate bond event studies using bond transaction data are explored. We show that the procedures used in studies to date have relatively low power because they fail to control for the substantial heteroskedasticity in bond returns due to differences in term-to-maturity, rating, and other characteristics. Standardization of bond returns by their estimated volatility leads to much more powerful tests. We further find that due to infrequent trading, use of transaction price observations over several days before and after an event, while giving more weight to returns calculated from transactions closest to the event, yields considerably more powerful tests than returns based solely on transactions the day before and day after the event (while also increasing coverage of small firms). We also explore how test power depends on which trades are included and how transaction prices are weighted in calculating daily prices, and on how individual bond returns are weighted in calculating firm returns. As an example of the power improvement our suggestions yield, the likelihood of detecting a fifteen basis point event impacting 300 firms using the t-test rises from approximately 24% for the best current state-of-the art procedure to 84% following our suggested procedures