Long-Term Earnings Forecast Models for Nonseasonal Firms

Abstract

This paper examines the long-term predictive ability of earnings forecast models for a sample of 167 firms whose quarterly earnings numbers exhibit nonseasonal behavior. Empirical evidence is provided showing that: 1) a large number of firms (n=167, i.e., 28.2% of the sample) exhibit nonseasonal patterns in their quarterly earnings series; 2) the use of quarterly ARIMA forecast models does not result in enhanced annual earnings predictions versus annual ARIMA models for nonseasonal firms; 3) the size effect documented by Bathke et al. (1989) for short-term quarterly earnings forecasts also pertains to long-term, annual earnings forecasts: and 4) larger firms’ earnings series display enhanced levels of earnings persistence versus those of smaller firms.