Empirical Evidence on the Time‐Series Properties of Operating Cash Flows

This study investigates the time‐series properties of operating cash flows per share and earnings per share for all manufacturing firms on the Compustat Quarterly Industrial tape for which sufficient data are available. Both individually‐identified and “premier” models are compared on the basis of their relative fit and forecasting accuracy. The empirical results suggest that for both accounting variables the individually‐identified models outperform the premier models, although this advantage is larger for earnings, and for forecast horizons beyond one quarter ahead. A major conclusion of the study is that the time‐series properties of cash flows are quite different than those of earnings. In particular, the cash flow series are considerably less predictable, as shown by their relatively high incidence of white‐noise series and relatively large forecast errors.