Information Content of Annual Earnings Announcements Revisited

Abstract

Studies which use residual variance measures to investigate the content of earnings reports (e.g., Beaver [1968] and May [1971]) exclude firms with fiscal years ending on December 31 for two reasons. First, since December 31 fiscal year-end firms' annual earnings announcements are clustered together in the months of January, February, and March, some of the information effects of firms' earnings may be eliminated when the market return is removed (via a market model) from an individual firm's security returns. Moreover, the residuals of such firms may be correlated due to extramarket factors such as industry factors. Second, the inclusion of December 31 fiscal year-end firms in the sample may induce a bias, since they are, on average, relatively large and as such are generally associated with greater flows of additional from sources other than earnings reports. Patell [1976], as a reference point in his study of earnings forecasts, employed the residual variance measures to examine the content of 273 annual earnings announcements of a subsample of firms with December 31 fiscal year-end. His results were less dramatic (about 10 percent above-normal price reactions) and were based on firms which also voluntarily disclosed their earnings forecasts along with annual earnings. However, Patell did not examine the validity of the methodological concerns alluded to by Beaver and May regarding December 31 fiscal year-end firms. In the next two sections of this paper, I discuss the methodological