Estimating Response Coefficients : Pooled Versus Firm-Specific Models

While accounting research has shown that earnings response coefficients vary across firms many short-window accounting studies estimate earnings response coefficients using a pooled cross-sectional regression model which implicitly assumes that coefficients are identical across firms. We study the implications of this implicit assumption by comparing the cross-sectional regression approach to a firm-specific coefficient approach. In the samples we study the means of the firm-specific earnings response coefficients are on average more than 10 times as large as the corresponding coefficients estimated from the pooled cross-sectional regression model. We conclude that future research on the relation between accounting variables and returns should seriously consider the use of a firm-specific coefficient methodology in place of a cross-sectional regression methodology