Can We Consistently Forecast a Firm's Earnings? Using Combination Forecast Methods to Predict the EPS of Dow Firms

Abstract

This paper forecasts earnings per share four - and eight - quarters ahead for 30 Dow firms using out-of-sample combination forecast methods. We show that many financial/economic variables, such as price-earnings ratio, dividend yield and Treasury bill rate, fail to predict out-of-sample EPS relative to a simple autoregressive model. In contrast, a combination forecast method that combines both firm-specific and macroeconomic variables leads to substantial improvements in predictive power relative to the autoregressive benchmark. For most Dow firms, principal component methods in particular lead to large improvements in out-of-sample mean squared forecast error. Our results highlight that reliably identifying a firm’s earnings is not based on a single variable, but on the wealth of information embodied in a host of economic and financial variables, and that combination forecast methods can consistently outperform an AR benchmark across most Dow firms.

Keywords: Combination Forecast Method, Earnings Per Share, Principal Components, Dow

JEL Classification: G10, G17, C53