اعتبار تعهدی، ثبات درآمد و قیمت سهام

This paper extends the work of Sloan (1996. The Accounting Review 71, 289) by linking accrual reliability to earnings persistence. We construct a model showing that less reliable accruals lead to lower earnings persistence. We then develop a comprehensive balance sheet categorization of accruals and rate each category according to the reliability of the underlying accruals. Empirical tests generally confirm that less reliable accruals lead to lower earnings persistence and that investors do not fully anticipate the lower earnings persistence, leading to significant security mispricing. These results suggest that there are significant costs associated with incorporating less reliable accrual information in financial statements.

This paper investigates the relation between accrual reliability and earnings persistence. The paper builds on the work of Sloan (1996), who shows that the accrual component of earnings is less persistent than the cash flow component of earnings and attributes this difference to the greater subjectivity of accruals. We draw a natural link between Sloan's notion of subjectivity and the well-known accounting concept of reliability. We formally model the implications of reliability for earnings persistence, with our model predicting that less reliable accruals result in lower earnings persistence. Our empirical tests employ a comprehensive categorization of accounting accruals in which each accrual category is rated according to its reliability. Consistent with the predictions of our model, the empirical tests generally confirm that less reliable accruals lead to lower earnings persistence. We also show that stock prices act as if investors do not anticipate the lower persistence of less reliable accruals, leading to significant security mispricing. Our paper makes three contributions to the existing literature. First, we directly link reliability to empirically observable properties of accounting numbers. Reliability, along with relevance, is considered to be one of the two primary qualities that make accounting information useful for decision-making.1 While a large body of research examines the value relevance of accounting numbers, there is relatively little research on reliability. One consequence of the emphasis on relevance has been calls for the recognition of more relevant and less reliable information in accounting numbers (e.g., Lev and Sougiannis, 1996). However, as recently articulated by Watts (2003), allowing less verifiable and hence less reliable estimates into accounting numbers can seriously compromise their usefulness. Our research highlights the crucial trade-off between relevance and reliability. Our analysis suggests that the recognition of less reliable accrual estimates introduces measurement error that reduces earnings’ persistence and leads to significant security mispricing. The second major contribution of our work is to provide a comprehensive definition and categorization of accruals. Following Healy (1985), a large body of research has employed a narrow definition of accruals that focuses on working capital accruals. Non-current accruals, such as capitalized software development costs and capitalized expenditures on plant and equipment, are omitted from Healy's definition. Ignoring such accruals results in noisy measures of both accruals and cash flows (because cash flows are typically computed as the difference between earnings and accruals). Our evidence indicates that many of the accruals that are omitted from Healy's definition are of low reliability, and we encourage subsequent research to incorporate such accruals. The third major contribution of our work is to corroborate and extend Sloan's findings regarding market efficiency with respect to information in accruals. Sloan shows that investors act as if they do not anticipate the lower persistence of the accrual component of earnings, resulting in significant security mispricing. We corroborate and extend Sloan's results in two ways. First, we develop a more comprehensive definition of accruals than Sloan and show that it is associated with even greater mispricing. Second, we show that both the persistence of earnings and the extent of the associated mispricing are directly related to the reliability of the underlying accruals. While this study examines the impact of accrual reliability on earnings persistence, we emphasize that there are other explanations for our results. In particular, Fairfield et al. (2003a) contend that Sloan's (1996) findings are a special case of a more general growth effect that is attributable to diminishing marginal returns to investment and/or conservative accounting. Another explanation is that extreme accruals may be caused by extreme changes in sales that have a transitory economic impact on earnings. Supporting our accrual reliability explanation, Xie (2001) and Richardson et al. (2004) find that the lower persistence of accruals persists even after controlling for sales growth. Nevertheless, it is possible that other explanations contribute to our results and additional research is required to discriminate between competing explanations. The remainder of the paper is organized as follows. Section 2 develops our research design. Section 3 describes our data, Section 4 presents our results and Section 5 concludes.

Our comprehensive examination of accrual reliability, earnings persistence and future stock returns generates several new insights. First, we document significant costs from recognizing less reliable information in financial statements. Less reliable accruals lead to lower earnings persistence and investors do not appear to fully anticipate this lower persistence, leading to significant security mispricing. Second, we provide a comprehensive definition and categorization of accruals that incorporates many accruals that have been ignored by previous research. Our results indicate that some of the accrual categories that have been ignored by previous research have particularly low reliability. Third, we show that the magnitude of the security mispricing related to accruals is significantly greater than originally documented by Sloan (1996). For example, a simple decile sort on a combination of the least reliable accrual categories produces annual hedge portfolio returns of 18%. Our analysis has several implications for existing research. First, it highlights the crucial trade-off between relevance and reliability in accrual accounting. We show that less reliable accruals introduce costs in the form of lower earnings persistence and the associated mispricing. Our findings provide a natural counterpoint to research calling for new categories of relevant but relatively unreliable information to be allowed under GAAP (e.g., Lev and Sougiannis, 1996). In particular, our results support Watts’ (2003) contention that recent moves by the FASB to introduce even less verifiable information into GAAP may prove extremely costly.20 We reiterate Watts’ conclusion that the case for allowing less reliable categories of accruals must recognize the problems that reliability evolved to address. We note in this respect that our analysis treats measurement error in accruals as an independent random variable. To the extent that strategic managerial reaction and/or accounting conventions such as conservatism cause the measurement error to behave in a non-random fashion, our analysis may underestimate the effects and costs. The second key implication of our analysis is that accruals-based research should consider broader definitions of accruals than the definition originally proposed by Healy (1985). Healy's definition has become synonymous with accruals in the accounting literature. Yet we show that his definition excludes several economically significant categories of accruals with low reliability. Non-current operating asset accruals (e.g., capitalization of operating costs as long-term assets) represent a good example. Such accruals were at the heart of the well-known accounting debacle at WorldCom. We recommend that future research employ broader measures of accruals in order to provide more powerful tests. Our analysis also highlights the differing degrees of reliability associated with different categories of accruals. In this respect our analysis is still quite crude, and further refinements should lead to additional insights into accrual reliability and earnings persistence. Third, our findings corroborate claims concerning the importance of distinguishing between earnings and ‘free cash flow’ in evaluating firm performance.21 The standard textbook definition of free cash flow adjusts earnings by adding back depreciation and amortization and subtracting changes in working capital and capital expenditures. As such, these adjustments are substantially similar to the operating accrual components that we develop in this paper. Financing accruals are typically not backed out in arriving at free cash flow. Thus, the accruals that we find to be the least reliable correspond closely to the adjustments that are made to earnings in arriving at free cash flow. Free cash flow represents a combination of actual cash flows plus the relatively reliable financing accruals. Finally, our analysis is subject to numerous limitations. First, we do not develop and test potential alternative economic explanations for the differential persistence coefficients on the cash flow and accrual components of earnings. In particular, the accrual component of earnings is associated with sales growth, and it is possible that economic factors cause earnings to be less persistent in growth firms. Research by Xie (2001) and by Richardson et al. (2004) finds that the lower persistence of accruals remains even after controlling for sales growth. Nevertheless, additional research to evaluate the relative importance of competing explanations would be useful. Second, we assume that measurement error in accruals is an independent random variable. In reality, measurement error is likely to be influenced by strategic managerial manipulation of earnings and by accounting conventions such as conservatism. Modeling and endogenizing the measurement error should lead to additional insights. Finally, our analysis assumes that any measurement error in cash flows is of second-order importance. Recent research suggests that the cash flows are also subject to managerial manipulation (see Roychowdhury, 2003; Graham et al., 2005). The simultaneous modeling of measurement error in cash flows and accruals should provide an improved understanding of earnings persistence.