Autoregressive Modeling of Earnings‐Investment Causality

The purpose of this paper is to empirically test the relationships between corporate earnings and investment. In particular, the study investigates whether knowledge of past investments improves the prediction of future earnings beyond predictions that are based on past earnings alone. Similarly, it investigates whether knowledge of past earnings improves the prediction of future investments beyond knowledge of past investments alone. This is the empirical definition of Granger causality. The empirical results show that the bivariate past series of earnings and investments is superior to the univariate series in predicting future investments but not in predicting future earnings. 1987 The American Finance Association