The Incremental Predictive Ability of Accrual Models With Respect to Future Cash Flows

Abstract

Prior studies on the incremental predictive ability of accrual models over cash flow models with respect to future cash flows have led to conflicting results. This paper presents an accrual-based cash flow prediction model based on a random walk in cash flows adjusted for the reversal of current payables and receivables. Results indicate that this simple accrual model predicts future cash flows (out-of-sample) better than models based on current cash flows alone. This paper also provides a more sophisticated accrual model by extending the model of the accrual process developed by Barth, Cram, and Nelson (2001) to include cash flow implications of growth in future sales. This more sophisticated accrual-based prediction model estimated via WLS (while pooling the prior three years of observations) predicts future cash flows better than both the simple accrual reversal model and the cash flow-based models, indicating that the accrual model contains information about future cash flow beyond the simple mechanical reversal of accruals. One explanation is that accruals may contain information regarding future sales. Consistent with this explanation, the paper finds that the accrual-based WLS model is superior to the cash flow-based model in capturing the effect of future sales on future cash flows. To determine whether the improved forecast accuracy is large enough to affect decision-making by financial statement users, the deciles of firms ranked on forecasted cash flow are compared to the deciles of firms ranked on actual future cash flow. The accrual-based model is superior to the cash flow-based model in placing firms into the correct deciles of actual future cash flow.

Keywords: cash flow prediction, out of sample prediction, accruals

JEL Classification: M41, M43, M44, G12