Earnings Announcements and the Convergence (or Divergence) of Beliefs.

Abstract Research on analysts' earnings forecasts has produced two major results. First, security analysts provide more accurate forecasts than do time-series models (Brown et at 1987) and, second, analysts' forecasts become more accurate and less dispersed as the forecast horizon decreases (Brown et al. 1985). This study examines the effect of an annual earnings announcement on the dispersion of analysts' one-year-ahead forecasts. it seems logical that forecasts should be less dispersed after the release of a value-relevant publicly observable signal. However, we find the opposite; i.e., that forecasts become more dispersed than would be expected in the absence of an earnings announcement. Security analysts' forecasts have often served as a proxy for the unobservable market expectation of earnings. Similarly, the dispersion of analysts' forecasts may proxy for the diversity of investor beliefs about future earnings. A number of studies have suggested that diversity of beliefs is important in security pricing as well as being a determinant of trading volume. Insight into the effects of earnings announcements on the heterogeneity of investors' beliefs will improve our understanding of how information gets impounded into prices and how information alters investors' portfolio decisions. The Bayesian belief revision model developed in this study suggests that the surprise content of the signal and the diversity of the perceived precision of the signal are important factors in determining whether the in- formation event will cause a convergence or divergence of forecasts. Holt- hausen and Verrecchia (1990) reach similar conclusions when they examine the effect of information on consensus. In their model, a decrease in consensus (increase in diversity of beliefs) is possible if there is disagreement about the effect of the signal on the value of the firm. Bamber (1987) argues that the surprise content of the announcement and disagreement about the interpretation of the signal are related; i.e., "more surprising or informative announcements are likely to spawn a wide variety of interpretations…." Empirical results are consistent with the insights provided by our model. There is a greater divergence of forecasts when the earnings announcement contains a bigger surprise, where surprise is defined as the difference between reported earnings and analysts' predictions of those earnings. An alternative explanation for the empirical results is nonsynchronous updating of forecasts by analysts. By partitioning the sample on the length of time between the announcement date and the next IBES report, we are better able to identify which IBES report (the first or second after the announcement) contains updated forecasts. This partitioning provides a more accurate measure of changes in the dispersion of forecasts due to an earnings announcement.