Earnings forecasts and stock prices: An examination of the distribution of analyst's forecasts

Scope and method of study. This study develops and tests a model that posits the existence of skewed earnings forecast distributions. The first purpose is to determine the existence of skewness of earnings forecast distributions. The skewness of the distributions is examined for significant information with respect to unexpected earnings. Next, an examination of determinants of skewness of earnings forecast distributions is conducted. Finally, this study examines the securities market's use of information contained in the skewness of the earnings forecast distributions to price the underlying stock. This study uses the Shapiro-Wilks W statistic, Johnson ;Findings and conclusions. This study develops a theoretical model to explain why skewed earnings forecast distributions exist. Empirical tests of the theoretical model in this dissertation provide evidence consistent with the model. First of all, 6,505 earnings forecast distributions were significantly nonnormal representing 39.4 percent of the 16,529 tested distributions. There are 3,136 significantly skewed distributions. This study finds that the skewness of the earnings forecast distributions is significantly positively correlated to unexpected earnings. Skewness is negatively correlated with measures of institutional ownership, consistent with skewness increasing as access to private information is limited to fewer earnings forecasters. A test of biased earnings forecasts is shown to exist and the bias is significantly less for skewed distributions than normal earnings forecast distributions. Finally, this study finds evidence that the market values the information contained in the skewness of earnings forecast distributions to price the underlying stock. Significant cumulative average abnormal returns of.53 percent and.36 percent for positively and negatively skewed distributions, respectively, do exist over the three day period following the publication of skewed distributions. However, a cross sectional regression of the earnings forecast distribution characteristics on the abnormal returns does not provide evidence of a significant relationship.