Accrual Anomaly: A Cointegration Approach

Abstract

In this study, I provide a refinement to the accrual anomaly with a new measure of earnings reversal for a given level of accruals. This new measure is the predicted earnings reversal based on the firm specific cointegration relation between earnings and cash flows implied by the accounting identity that earnings is the sum of the cash flows and the accruals. More specifically, it is the amount of earnings reversal for a given level of accruals based on the long-run equilibrium relation between earnings and cash flows for each firm. As this earnings reversal measure captures the firm specific time-series characteristics of earnings, cash flows, and accruals, I predict that it will yield a better cross-sectional ranking than the level of accruals for the accrual strategy which in turn will yield higher size-adjusted returns. I provide evidence that investing in a hedge portfolio that takes short (long) positions in firms with negative (positive) forecasted earnings reversals implied by the accruals yields annualized size-adjusted returns of 21.53 percent using the in sample parameter estimates alone and 8.80 percent using the out of sample parameter estimates in an additive way to the level of accruals between 1964 and 2008. My findings, which are robust to various earnings measures, sub-periods, sub-samples, and model specifications suggests that the performance of the accrual strategy can be enhanced by considering the firm level time-series characteristics of earnings, cash flows, and accruals.