| This study examines whether security analysts extract information about underlying economic profits from reported earnings to revise their expectations of future earnings. Analysts' forecast revisions after earnings announcements are modeled as a Bayesian learning process in which the reported earnings and the capital market's reaction to the reported earnings are two noisy signals that reflect underlying profits. The model predicts that analysts, in revising their forecasts, place a higher weight on the reported earnings relative to the market reaction when the earnings signal contains less noise or the market reaction is more highly influenced by noise trading during the earnings announcements. I use the variance of residuals from a time series regression of earnings, conditional on firm size, capital intensity, product type and degree of barrier-to-entry, as the proxy for the noise in earnings, and stock price reversals after earnings announcements as the proxy for the noise in the market reaction. In addition, the model predicts that if the noise in earnings is positively auto-correlated the analysts should use this property to forecast earnings and thus place a higher relative weight on earnings. Using a sample of 18,399 individual analysts forecasts from 1993 to 2001, the results are generally consistent with the model. Therefore, I conclude that security analysts appear to process accounting information rationally, and collectively act as effective information intermediaries to enhance the overall efficiency of the capital market. |
