Can Firm-Specific Models Predict Price Responses to Earnings News?

The primary question addressed in this study is whether firm-specific ERCs - i.e., slope coefficients obtained from time-series regressions of abnormal returns on earnings surprises - are helpful in predicting price responses to future earnings surprises. Fundamental analysis involves both predicting future earnings and understanding the implications of future earnings for firm value. We focus on out-of-sample predictive ability of ERCs in order to provide evidence on whether historical returns-earnings relations - as measured under various specifications - are useful in predicting future returns-earnings relations. We find that ERCs from pooled regressions provide more accurate predictions of price responses to future earnings surprises than ERCs from firm-specific regressions. In addition, out-of-sample predictions from firm-specific regressions estimated over actual firms are no more accurate than those from firm-specific regressions estimated over pseudo firms (which are created through random draws of returns-earnings data and therefore have no economic characteristics that extend beyond the period over which the coefficients are estimated). We also find that the accuracy of out-of-sample predictions from firm-specific regressions decreases for relatively young firms and for firms that have recently experienced a relatively large change in the book-to-market ratio