Dispersion of Beliefs, Stock Prices and the Earnings Surprise Measures - A Generalized Approach

Abstract

In this paper we address the issue of modelling the relation between the stock prices and accounting earnings in the presence of potential divergence of opinions regarding the earnings data generating process among the investors. In our model the market's earnings expectation is defined as the weighted average of both the time-series and analysts' forecasts, with the weights being estimated directly from stock returns. No assumptions are made on the functional form of the earnings surprise-stock returns relation, which makes our model flexible enough to incorporate a variety of models discussed in the previous literature. The model is estimated semiparametrically following Hardle et. al [Annals of Statistics, 1993]. Our key findings are as follows. First, we find that investors use both the time-series and analysts' forecasts to predict future earnings. Second, the proportion of investors using the time-series (analysts) forecasts is significantly higher (lower) for the stocks with low market capitalization. Third, we find that accounting for the dispersion of earnings forecasts leads to a substantial increase in the magnitude of the post-earnings announcement drift.