Predicting Accruals based on Problems with Cash Flows

We construct empirical proxies for cash-flow timing and matching problems and provide evidence on the ability of this accrual generating factor to explain accruals. Researchers lament our meager understanding of the process that spawns accruals (Owens et al., 2013, Ball, 2013, McNichols, 2000). Our aim is to map the explanatory power boundaries demarcated by our current knowledge of the role of accruals—not to enter an empirical horse race. We produce three measures of cross-sectional variation in cash-flow problems: (1) negative serial correlation in cash flows (2) the predictive power of current cash flows for future cash flows, and (3) the relative explanatory power of two-year cash flows to one-year cash flows for returns. We find that the Jones-model r-squared increases significantly when cash-flow problems are incorporated and that these increases in explanatory power are larger than those produced by incorporating asymmetrically timely loss recognition into the Jones model (Ball and Shivakumar, 2005). We also show that the explanatory power of the Dechow and Dichev (2002) model increases significantly in subsamples exhibiting strong negative serial correlation in cash flows, suggesting that the explanatory power of this model is tied to its incorporation of cash-flow problems. Our results suggest a way to understand accruals at the firm-year level can be found via improved estimates of the transitory component in current operating cash flows. We thank participants in the Washington University in St. Louis accounting brown bag for comments.