The calendar-time approach as applied in the preceding sections assumes constant factor loadings over the estimation period. Jaffe (1974) and Mandelker (1974) propose to measure abnormal returns on a firm-specific basis and then form monthly calendartime portfolios using abnormal returns. Mitchell and Stafford (2000) compare both calendar-time methods and conclude that there is no difference in inferences about abnormal returns, but the Jaffe and Mandelker variant generally leads to more conservative point estimates. Firm- and month-specific abnormal return are hard, if not impossible, to estimate using an asset pricing model. A potential solution is to lower the returns frequency down to daily data to get the necessary degrees of freedom, but 22 daily returns (on average) introduce too much noise to reliably estimate abnormal returns. Mitchell and Stafford therefore propose that one measures firm-specific abnormal returns over a certain post-event period using monthly data. As a consequence, the firm-specific loadings do not change from month to month (are not month-specific), but by revising the portfolio composition every month, the implicit loadings to estimate the abnormal return for this portfolio vary.
KeywordsMarket Capitalization Abnormal Return Information Risk Deviation Variable Asset Price Model
Unable to display preview. Download preview PDF.