Tests of Long-Term Abnormal Performance: Analysis of Power

In tests of long-term performance, researchers are faced with several research design choices. For instance, when estimating abnormal returns, what specific firm characteristics should be used as matching criteria to select control firms? What weights should be placed on each characteristic? Should an event firm be matched with one control firm or multiple control firms or with a reference portfolio? Should we use the calendar-time portfolio approach? We provide guidance to researchers on these questions by evaluating the power of the test using simulation analyses. We find that the quality of matching when selecting control firms has little impact on the power of the test. Among the alternative approaches studied, the Fama-French calendar-time portfolio approach obtains the highest power in random samples. Interestingly, we find that the higher power of this approach is attributable to the return aggregation method rather than the insample fit or the use of multiple risk factors in the model. In non-random samples, the FamaFrench approach obtains the highest power but is severely misspecified. The correction for misspecification involves matching on appropriate firm characteristics to estimate the empirical distribution of pseudo-alphas; thus, in general, the calendar-time portfolio approach in conjunction with matching produces well-specified tests with the highest power. Even so, for a reasonable sample size, the power of the best available methodology to detect economically significant abnormal returns is quite low.