Changes in accrual properties and operating environment: Implications for cash flow predictability

Highlights

• We reconcile the mixed evidence in prior literature on the relative ability of earnings and cash flows in predicting future cash flows. We document that the conflicting evidence is attributable to differences in the manner in which cash flows are measured.

• Operating cash flows outperform earnings in predicting future operating cash flows every year over the last two decades, both in the US and internationally.

• Accruals and its components, including those capturing non-articulating events, exhibit incremental predictive ability over current operating cash flows. Thus, our study sheds new light on interpreting prior research by Hribar and Collins (2002) on accrual measurement.

• Earnings' ability to predict future operating cash flows is increasing over the period 1989–2015. However, this trend is attributable to the increasing predictive ability of operating cash flows rather than accruals.

• Shortening operating cycles, decreasing levels of non-cash working capital, and increasing intensity of intangibles over time explain the increasing trend in cash flow predictability.

Abstract

This paper reconciles conflicting evidence in prior literature on the relative ability of earnings and cash flows in predicting future cash flows. Further, we investigate the implications of temporal shifts in accrual properties and operating environment for cash flow predictability. Three key insights emerge. First, cash flows consistently outperform earnings in predicting future cash flows. Second, accruals and its components, including those capturing non-articulating events, have incremental (albeit small) predictive ability over cash flows. Third, earnings’ ability to predict future cash flows has increased over the period 1989–2015, due to changes in operating environment rather than accrual properties.

Introduction

The Financial Accounting Standards Board (FASB) asserts in its conceptual framework that a primary objective of financial reporting is to help existing and potential investors, lenders, and other stakeholders assess the amount, timing, and uncertainty of future expected cash flows (Financial Accounting Standards Board (FASB), 1978, Financial Accounting Standards Board (FASB), 2010). The conceptual framework also states that earnings provide a better basis than current cash flows for assessing a firm's future expected cash flows. Earnings' superior ability as a summary measure to predict future cash flows is attributable to the timing role of accrual accounting, i.e., the ability of accruals to smooth temporary timing differences in cash flows. Following these assertions, prior research has examined both (i) the relative ability of earnings as compared to operating cash flows to predict future operating cash flows, and (ii) the incremental ability of accruals beyond operating cash flows in predicting future operating cash flows.¹ While prior research shows that accruals contain incremental information for predicting future cash flows, the evidence on the relative ability of earnings and cash flows for predicting future cash flows is mixed.² The first objective of our paper, therefore, is to reconcile the contradictory evidence in the literature regarding the relative predictive ability of earnings and cash flows.

As our second objective, we investigate time trends in the predictive ability of earnings and its components because of temporal shifts in accounting and operating environments, each of which has different implications for cash flow predictability. First, temporal shifts in the landscape of accrual accounting may affect the predictive ability of accruals and earnings over time. In particular, Bushman et al. (2016) document that the negative contemporaneous correlation between accruals and cash flows has been declining over time and has largely disappeared in recent years, primarily due to the increased incidence of non-timing-related accruals. Diminishing correlation between accruals and cash flows may limit the predictive ability of models that use earnings as a summary measure because they constrain the coefficients on accruals and cash flows to be the same.

Second, Lev, Li, and Sougiannis (2010) report a positive trend in the prevalence of managerial estimates embedded in accruals due primarily to the move towards fair-value accounting (see also Lev and Gu, 2016). On the one hand, managerial estimates could improve the predictive ability of accruals and earnings by incorporating forward-looking information. On the other hand, managerial estimation could diminish the predictive ability due to measurement error and accrual manipulations.³

Third, the nature of firms' operating environments has also changed significantly over time (Fama and French, 2004, Srivastava, 2014). For example, firms increasingly exhibit lower profitability, lower matching between revenues and expenses, higher growth, and higher cash flow volatility. Further, firms are becoming more intangible intensive (Srivastava, 2014, Bushman et al., 2016). Growing investments in intangibles increase cash flows' ability to predict future cash flows, as these investments do not often generate accruals due to the immediate expensing of these items. Finally, from a theoretical perspective, both the relative predictive ability of earnings over cash flows and the incremental predictive ability of accruals are a function of the length of firms' operating cycles and the magnitude of their working capital (Dechow et al., 1998). Changes in the operating environment could also affect these characteristics and hence the trends in accruals and earnings’ predictive ability. Thus, both accounting and operating factors could differentially affect the temporal variation in cash flow predictability.

We conduct our empirical analysis using Compustat annual data spanning the period 1989–2015 and use pooled estimations, year-wise cross-sectional estimations, and firm-specific time-series estimations to assess the relative predictive ability of earnings vis-à-vis cash flows and the incremental predictive ability of earnings components.⁴ As a starting point, we follow the cash flow statement–based approach to measure cash flows. In terms of relative predictive ability, we find that current cash flows outperform earnings in predicting future cash flows over the entire sample period. Notably, in every year of the sample period 1989–2015, cash flows outperform earnings. On average, cash flows’ predictive ability is about 1.5 times that of earnings. Our finding that cash flows outperform earnings is consistent with one stream of prior literature (e.g., Bowen et al., 1986, Finger, 1994, Burgstahler et al., 1998), but inconsistent with the other (e.g., Greenberg et al., 1986, Dechow et al., 1998, Kim and Kross, 2005).

To reconcile the contrasting results documented in prior studies, we identify several possible explanations and explore each in sequence: (i) measurement approach (measures based on the cash flow statement versus the balance sheet), (ii) sample period selection and composition (large firms, small firms, constant sample), (iii) variable definitions (earnings, cash flows, accruals), and (iv) estimation methods (e.g., cross-sectional vs. firm-specific estimation). Of these, the main driver of the contradictory findings in prior literature is the measurement approach adopted for estimating cash flows. Hribar and Collins (2002) assert that cash flows imputed from the balance sheet are erroneous because of errors introduced by non-articulation events. To explore this explanation, we compare the results from both the balance sheet– and cash flow statement–based approaches. Our results are strikingly different when we use balance sheet–based cash flows; we find that earnings display greater predictive ability than cash flows – exactly the opposite of our finding above. This evidence suggests that the contrasting conclusions in the prior literature are an artifact of the measurement approach used. That is, the balance sheet method induces measurement error in cash flows that, in turn, drives earnings’ superior predictive ability. Motivated by this finding, we organize the past literature based on the measurement approach used. The pattern that emerges is that papers using the cash flow statement (balance sheet) approach show that cash flows (earnings) are the better predictor.

Next, we explore the incremental predictive ability of accruals and cash flows. We find that accruals contribute 2 percent incremental explanatory power over cash flows, whereas cash flows provide 37 percent incremental explanatory power over accruals. Disaggregating accruals into its major constituent components following Barth et al. (2001) increases the incremental predictive ability of accruals to about 5 percent. Alternatively, when we decompose accruals based on managerial estimates embedded in accruals (Lev et al., 2010), we find that the incremental predictive ability of accrual components, on average, is about 4 percent. While accruals estimated from the cash flow statement contain incremental information, they miss important economic transactions, particularly mergers, acquisitions or divestiture activities that have implications for future cash flows (Larson et al., 2018). However, these activities are incorporated into the accruals estimated from the balance sheet. Hence, when we replace accruals based on the cash flow statement approach with accruals based on the balance sheet approach, the incremental predictive ability of aggregate accruals increases from 2 percent to 3 percent. Together, the results suggest that the components of accruals, including non-articulating accruals, have incremental predictive ability.

To accomplish our second objective, we conduct time-trend analysis and find that earnings’ ability to predict future cash flows is increasing over time: the explanatory power (Adj. R²) increases from 14 percent in 1989 to 37 percent in 2015. On average, the explanatory power of earnings increases by 0.96 percent per year, and this increasing trend is significant at the 1 percent level. Further investigation reveals that the increasing trend in explanatory power is attributable to the cash flows component, not accruals. Moreover, we observe no time trends in the predictive ability of disaggregated accruals. Together, the evidence suggests that the increase in the predictive ability of earnings cannot be attributed to changing accounting factors.

What explains the increasing trend in the predictive ability of the cash flows component? Following the theoretical and empirical literature, we consider several economic factors: operating cycles, magnitude of non-cash working capital, and intangible intensity (Dechow et al., 1998, Bushman et al., 2016). During our sample period, we find that operating cycle decreases on average by 0.47 days per year, non-cash working capital decreases on average by 0.42 percent of assets per year, and intangible intensity increases on average by 0.10 percent of total expenses per year. Multivariate regression analyses suggest that these trends are partly responsible for the increasing predictive ability of cash flows (and hence earnings), with declining operating cycle being the most dominant driver. If declining operating cycles contribute to the increasing predictive ability of cash flows, we conjecture that the relative predictive ability of cash flows over earnings should diminish with shorter forecast horizons (e.g., quarterly and semi-annual frequency). Consistent with this conjecture, we find that the predictive ability of earnings is on average better than that of cash flows at shorter reporting frequencies.

Finally, we explore whether our main findings in this study are peculiar to the US or generalizable to international settings. Similar to the evidence based on the US sample, results using the international sample suggest that cash flows outperform earnings in predicting future cash flows. Also, we find that the predictive ability of both earnings and cash flows is increasing over time for the international sample, and that the increase in earnings’ predictive ability is almost entirely attributable to the cash flows component. Together, these findings reinforce the conclusion that economic forces rather than reporting standards are responsible for the observed trends.

Our study has implications for accounting educators, academics, practitioners, and policy setters. First, the importance of cash flows, relative to earnings, in predicting future cash flows is relevant for practitioners and educators who often promote and use earnings as the summary metric for cash flow prediction purposes. Accounting standard setters such as the FASB and the IASB would find our evidence relevant as it challenges an important tenet of financial reporting, i.e., accounting earnings, as a summary metric, provides a better basis for predicting future cash flows. Second, the evidence in this study enhances our understanding of when and to what extent accruals, either alone or when combined with cash flows to form a summary earnings measure, help with predicting future cash flows. Third, despite the changes in reporting and operating environments over the years, accruals and its components display stable predictive ability. Finally, our finding that the non-articulating accruals component inherent in balance sheet–based accruals provides incremental predictive ability challenges the view that non-articulating accruals represent measurement error, and thereby adds a nuanced interpretation to Hribar and Collins (2002) recommendation. That is, while our results support the use of the cash flow statement approach for measuring cash flows, we suggest that measuring accruals using a balance sheet approach is more appropriate for applications such as predicting future cash flows.