Relation between Market Model Prediction Errors and Omitted Variables: A Methodological Note

Abstract

A number of accounting studies have examined market model prediction errors to evaluate the share price effects of various events, such as voluntary or mandatory accounting changes. A common approach is to examine the prediction errors around the event date of a portfolio of potentially affected firms. This approach has been extended recently through the use of cross-sectional regressions (e.g., Leftwich [1981], Holthausen [1981], Collins, Rozeff, and Dhaliwal [1981], Sepe [1982], and Chow [1983]) in which prediction errors or raw returns are regressed on firm-specific variables such as size.' The purpose of this paper is to examine the implications of some recent findings in finance for the various finance-based empirical accounting studies. In general, two types of omitted variables from the standard capital assets pricing model have been identified. First, several studies have reported that certain firm-specific variables such as size (Banz [1981] and Reinganum [1981]), earnings yield (Basu [1975; 1983]), and dividend-payout ratio (Litzenberger and Ramaswamy [1979]) may be omitted variables from the standard CAPM. In addition, other studies have found some special time-series characteristics of stock returns, such