This preview shows page 1. Sign up to view the full content.
Unformatted text preview: University of Rochester William E. Simon Graduate School of Business Administration
The Bradley Policy Research Center Financial Research and Policy Working Paper No. FR 03-25 August 21, 2003 Conservatism in Accounting Part II: Evidence and Research Opportunities
Ross L. Watts Simon School of Business, University of Rochester This paper can be downloaded from the Social Science Research Network Electronic Paper Collection: http://ssrn.com/abstract=438662 Conservatism in Accounting Part II: Evidence and Research Opportunities Ross L. Watts William E. Simon Graduate School of Business Administration University of Rochester Rochester, NY 14627 August 21, 2003 This paper was written while I was visiting the Sloan School of Management at the Massachusetts Institute of Technology. Financial support from the Sloan School and the Bradley Policy Research Center, William E. Simon Graduate School of Business Administration is gratefully acknowledged. I am also grateful for the helpful comments of William Baber, Sudipta Basu, George Benston, Richard Frankel, Carla Hayn, Ludger Hentschel, S.P. Kothari, Thomas Lys, Stan Markov, Stewart Myers, Suresh Radhakrishnan, Charles Wasley, Greg Waymire, Joseph Weber, Joanna Wu, Peter Wysocki and Jerold Zimmerman. Conservatism in Accounting Part II: Evidence and Research Opportunities SYNOPSIS This paper is Part II in a two part series on conservatism in accounting. Part I examines alternative explanations for conservatism in accounting and their implications for accounting regulators (SEC and FASB). Part II summarizes the empirical evidence on the existence of conservatism, conservatism's increase over time and conservatism's alternative explanations. It also discusses opportunities for future research on conservatism. Conservatism is defined as the differential verifiability required for recognition of profits versus losses. In its extreme form the definition incorporates the traditional conservatism adage: "anticipate no profit, but anticipate all losses." Despite criticism from many quarters, the formal evidence suggests conservatism not only exists in modern day financial reporting, it also suggests conservatism has increased in the last 30 years. The empirical literature uses a variety of conservatism measures in time-series and cross-sectional tests of contracting, shareholder litigation, taxation, and accounting regulation explanations for conservatism. The tests' results suggest the importance of all four explanations. Two non-conservatism explanations (earnings management and the abandonment option) cannot individually or jointly explain the observed systematic understatement of net assets that is the hallmark of conservatism. Researchers should note that accounting's effects on managerial behavior play a central role in the evolution of both accounting and financial reporting. Assessing the relevance of an accounting method to financial statement users' decisions requires assessing managers' abilities to use that method to manipulate accounting numbers and commit fraud. The evidence on conservatism suggests asymmetric verifiability is critical to constraining manipulation and fraud. INTRODUCTION This paper's first objective is to summarize the formal empirical evidence on accounting conservatism whether it exists, has increased in recent times and is consistent with the alternative conservatism explanations examined in Part I. The second objective is to suggest potentially fruitful lines of research in conservatism. As in Part I, conservatism is defined as a stronger verifiability requirement for the recognition of gains than for the recognition of losses. The extreme form of this definition is the old adage: "anticipate no profit but anticipate all losses." Existing empirical research investigates the former, more general, definition of conservatism that implies merely a differential or asymmetric standard of verifiability for gains and losses. Empirical research uses a variety of measures to assess whether conservatism exists. Those measures flow from conservatism's definition and, on average, suggest conservatism exists and has increased in recent times. The measures play an important role in testing the alternative conservatism explanations: contracting, litigation, taxation and accounting regulation. Because the different explanations generate different hypotheses about time-series and cross-sectional variation in conservatism, variations in the conservatism measures provide an opportunity to discriminate among the explanations. MEASURING CONSERVATISM Researchers use three types of measures to assess conservatism: Net asset measures, Earnings and accrual measures, and Earnings/stock returns relation measures. All measures rely on the effect of conservatism's asymmetric recognition of gains and losses on reported accounting numbers, in particular net assets, earnings and accruals. In this section I Explain how asymmetric recognition affects net assets, earnings and accruals, Describe the measures of those effects, and Summarize the evidence those measures provide on conservatism. Net Asset Measures The market value of the assets and liabilities comprising net assets change every period but all these changes are not recorded in the accounts and the financial reports. Under conservatism increases in asset values (gains) that are not verifiable are not recorded while decreases of similar verifiability are recorded. The result is that net assets are understated are below market value. Researchers obtain estimates of this understatement using models of the valuation of the firm's shares and/or the ratio of the firm's book value of net assets to its equity value (book-to-market ratio). 2 Valuation Model Measures Feltham-Ohlson valuation models (Feltham and Ohlson, 1995 and 1996) are usually employed to estimate the extent of undervaluation of net assets. Those models include parameters that reflect the degree of understatement of operating assets. The models induce the understatement by assuming accounting depreciation exceeds economic depreciation. Conservatism parameter estimates are obtained from estimation of the valuation model and from time series estimation of the relation between accounting variables that are inputs to the valuation model. The valuation model estimate comes from cross-sectional regressions of firm market value on abnormal earnings, assets and investment. A valuation estimation example is Ahmed, Morton and Schaefer (2000) who regress firms' goodwill on abnormal earnings, lagged operating assets and contemporaneous investment in operating assets. Goodwill is equal to the market value of equity minus the book value of net assets. To the extent the book value of net assets is understated, goodwill is overstated. The coefficient of lagged operating assets should be positive if conservatism has understated the lagged assets. A time-series conservatism parameter estimate is obtained from the time series regression of abnormal earnings on lagged abnormal earnings and lagged book value of operating assets (see Myers, 1999). Again, the coefficient of lagged operating assets should be positive when conservatism exists. To understand these predictions, note that the more excess depreciation understates the operating assets, the greater is the coefficient applied to the lagged operating assets to explain either goodwill or abnormal earnings. 3 Stober (1996), Dechow, Hutton and Sloan (1999), Myers (1999) and Ahmed, et al. (2000) estimate positive conservatism parameters from their valuation regressions. All these studies find net assets to be understated. However, these studies estimate negative conservatism parameters from the time series abnormal earnings regressions, not positive parameters as predicted. The authors attribute this inconsistency to a misspecified relation between the accounting variables, an arbitrary specification not guided by any theory of conservatism or accruals other than the assumption that the accounting depreciation rate is too high. Such excessive depreciation is not the source of conservatism under the contracting explanation given in Part I of this paper, the source is the lack of appreciation of book value when there are gains. Also, the specification ignores known relations in the time-series of earnings, especially negative serial correlation in earnings and earnings changes generated by accrual estimation errors (Ball and Watts, 1972) or by conservatism, as discussed below. Book-to-Market Measures Beaver and Ryan (2000) measure conservatism using firms' book-to-market ratios based on the notion that, ceteris paribus, firms using conservative accounting report lower net assets and lower-book-to-market ratios. Using pooled time series and cross sectional data they regress book-to-market ratios on individual year and firm dummy variables and on individual firm stock returns for the current and previous five years. The estimated coefficient of an individual firm's dummy captures the persistent portion of the difference between the firm's book and market values of equity. The lower the coefficient, labeled the "bias component", the more book value of net assets is biased downward and the more conservative the firm's accounting. By construction, because the mean coefficient 4 is zero, the coefficient estimates relative conservatism and not aggregate conservatism. This book-to-market measure is used to proxy for the extent to which conservatism varies across firms. Earnings/Accrual Measures Conservatism implies that gains tend to be more persistent than losses. Unverifiable increases in asset values (gains) are not recognized at the time they occur but over future periods as the cash flows generating those increases are realized. For example, if an asset's value increases because it is expected to throw off more future cash flows, the gain is recognized over the future years as the increased cash flows are realized. This means that gains tend to be persistent. Since firms with positive earnings or earnings changes are likely to have had gains, positive earnings and earnings increases are also likely to be persistent. Losses of the same degree of verifiability as the unverifiable gains tend to be recognized as they occur rather than in the future as the cash flow decreases are realized there is a lump sum drop in earnings at the time of the loss rather than a flow of reduced earnings in the future. Firms whose earnings are negative or decrease are more likely to have recognized losses. Since, on average, the losses do not recur in future periods, negative earnings and earnings decreases are less likely to persist than positive earnings and earnings increases. Those negative earnings and earnings decreases are transitory. The persistence or transience of earnings and earnings changes provides measures of conservatism. 5 Conservatism's asymmetrical treatment of gains and losses produces an asymmetry in accruals. Losses tend to be fully accrued while gains do not. This causes accruals to tend to be negative and cumulated accruals to be understated. As a result negative periodic net accruals and negative cumulative accruals (cumulated over periods) are used as measures of conservatism. In addition, conservatism suggests losses, with their capitalization of future flows, generate more very large accruals than do gains. This predicts negatively skewed distributions of accruals and earnings and suggests estimates of the negative skewness of distributions of earnings, earnings changes and accruals are measures of conservatism. Earnings Measures Basu (1997) and Watts (1993, p. 11) predict negative earnings changes are more likely to reverse in the next period than positive earnings changes. Evidence that negative earnings changes are more likely to reverse than positive earnings changes already existed in the literature before Basu (1997), in particular in Brooks and Buckmaster (1976) and Elgers and Lo (1994). In confirming this result, Basu regresses earnings changes, deflated by beginning-of-period price, on lagged deflated earnings changes for samples of positive and negative earnings changes. The estimated lagged earnings coefficient for positive earnings changes is insignificantly different from zero, consistent with positive earnings changes being permanent and not reversing. In contrast, the estimated lagged earnings coefficient for negative earnings changes is significantly negative (-.69), but is not significantly different from minus one, the value expected when negative earnings changes are completely transitory. This result is consistent with write- 6 offs due to conservatism causing negative earnings changes. When those write-offs capture all expected future losses on the assets, they are completely transitory. Accrual Measures Givoly and Hayn (2000) note that conservatism reduces cumulative reported earnings over time. They suggest the sign and magnitude of accumulated accruals over time are measures of conservatism. For firms in a steady state with no growth and neutral accounting, earnings converge to cash flows and periodic accruals converge to zero. "A consistent predominance of negative accruals across firms over a long period is, ceteris paribus, an indication of conservatism, while the rate of accumulation of negative accruals is an indication of the shift in the degree of conservatism over time" (Givoly and Hayn, 2000, p. 292). Consistent with conservatism, Givoly and Hayn find that the distribution of return on assets,whether derived from time-series of individual firms or the cross-section of firm-years, is negatively skewed for most of the period they examine (1956-1999). They also find that over the period 1965-1998 accruals, other than depreciation, cumulate to a negative amount equal to16% of cumulative earnings over the same period. This accumulation occurs from 1982-1998 and is consistent with the timing of a large increase in conservatism observed in the time-series evidence on the earnings/stock return relation. Earnings/Stock Returns Relation Measures Stock market prices tend to reflect asset value changes at the time those changes occur whether those changes imply losses or gains in asset value stock returns tend to 7 be timely. Since conservatism predicts that accounting losses are recorded on a timely basis but gains are not, accounting losses are predicted to be more contemporaneous with stock returns than are accounting gains. Basu (1997) predicts that stock returns and earnings tend to reflect losses in the same period, but stock returns reflect gains earlier than earnings. To provide estimates of his conservatism measure Basu regresses annual earnings on stock returns of the same year. He predicts a higher coefficient of stock returns and a higher R2 from this regression for a sample of firms with negative stock returns than for a sample of firms with positive returns. Using US data, Basu finds results consistent with his predictions. Using variations on this methodology, many other studies replicate the result, including Ball et al. (2000) and Holthausen and Watts (2001). Summary of the Evidence Overall, the evidence on the understatement of net assets, the behavior of earnings and accruals, and the earnings/stock return relation is strongly consistent with the existence of conservatism in US financial reporting. I believe that the one set of inconsistent results the negative estimated conservatism parameters in the abnormal earnings regressions is the weakest test. The positive sign predicted for the parameter does not incorporate the negative earnings dependence generated by conservatism, errors in accrual estimation, or the conservatism-generated asymmetry in the earnings distribution. EVIDENCE ON ALTERNATIVE CONSERVATISM EXPLANATIONS This section presents evidence on the extent to which conservatism measures vary across time, firms and countries in accord with the four conservatism explanations 8 discussed in Part I of this series. The section concludes with a discussion of the extent to which the evidence discriminates among the four explanations for conservatism. Time-series Evidence Part I argues that contracting likely explains the origins of both accounting and conservatism so that contracting's influence on accounting spans millennia. Thus one cannot identify a point in recent history when contracting's influence began. That same long history reduces the likelihood that recent contracting changes cause time-series variation in accounting conservatism. It is more likely that recent changes in the conservatism of GAAP cause changes in contracting (see later). This is not the case for litigation, however, as Watts (1993, p. 14) predicts that conservatism in US accounting varies with unexpected changes in legal liability. Part I notes that prior to 1966 litigation was extremely rare, so litigation alone does not explain conservatism in US published financial statements prior to that date. Thus, absent explanations other than litigation, one does not expect to see conservatism before 1966. Because other explanations do exist, in the US conservatism should increase after 1966. Further, Watts (1993) suggests that the post-1966 periods Kothari, et al. (1988) identify as different litigation regimes can be used to test whether conservatism varies with changes in liability exposure. Basu (1995, 1997) expects his conservatism measures to increase in the periods following liability increases and remain constant in periods following court decisions that restrict liability growth. Income taxes can explain the existence of conservatism in the US at least from 1909. Variations in the extent to which tax accounting is linked to financial reporting over time can be used to predict variations in the effect of income taxes on conservative 9 financial reporting. Based on the literature on the interaction between financial reporting and tax considerations, summarized in Shackelford and Shevlin (2001), one can predict that when the links between the two become closer, or taxes increase, financial reporting becomes more conservative. The Securities Acts that date from the 1930s are a potential source of a regulatory effect on conservatism, as the SEC's initial reaction was to be conservative for reasons explained in Part I (Zeff, 1972). Thus without other explanations, I expect conservatism to be first observed after the 1930s. Given the other explanations, I expect conservatism to increase after the 1930s. By the 1970s the SEC tended to be less conservative (Watts, 1977) and by 2000 we observe the FASB, presumably with the SEC's tacit approval, introducing unverifiable measures of assets. Hence, over the last 30 years if we can reliably measure the extent of conservatism in SEC rulings and FASB standards, we can predict variations in conservatism in accounting reports. Using the earnings/stock returns relation, Basu (1997) investigates conservatism in the US in four periods: 1963-66; 1967-75; 1976-83 and 1983-90. Kothari, et al. (1988) previously designated the periods as low, high, low, and high litigation growth periods respectively. Basu finds a significant increase in conservatism in the two high litigation growth periods and no increase in the low litigation growth periods, a result consistent with litigation generating conservatism. There is a significant level of conservatism in the last three periods but in the initial low litigation period (1963-66) the level of conservatism is insignificant, a result at odds with contracting and tax explanations and other evidence. 10 Holthausen and Watts (2001) argue that since contracting incentives for conservatism existed prior to the increase in the litigation in the late 1960's, US financial reporting should have been significantly conservative in the pre-1967 period also. Using the earnings/stock returns relation for a sample of large firms for the period 1927-1993, they find significant conservatism, not only in the last three litigation periods investigated by Basu, but also in the pre-litigation periods, 1927-41 and 1954-66. Conservatism is not significant in the period between 1942 and 1953, perhaps due to price controls during World War II and the Korean War.1 Holthausen and Watts (2001) find that the estimated coefficient for negative stock returns averages around .20 in the non-price control periods between 1927 and 1975. During 1976-82 the estimated coefficient increases to approximately .30, and by the 1983-93 period it is approximately .40. On the other hand, the estimated coefficient for positive stock returns averages around .10 in the non-price control periods until 1982 and drops to zero during1983-93. Both Basu (1997) and Ball, Kothari and Robin (1999) find a large increase in the negative returns coefficient and a drop in the positive returns coefficient in the periods 1983-1990 and 1985-95, respectively. This evidence leads me to conclude that by the end of the millennium US firms' accounting earnings are not timely at all in reflecting good news but are timely in reflecting bad news Because the significant increase in conservatism occurs under the FASB management of standards, it could be at least partially due to standard-setting. Accounting data were used under the World War II Office of Price Administration's "hold-the-line" rules to measure "out-of-pocket" expenses that could be used for a price increase. However, an asset write-off does not seem to qualify as an "out-of-pocket" expense. Further, shortages of fixed assets made decreases in asset values less likely (see Rockoff, 1995). 1 11 Cross-sectional Evidence Variation across Firms Beaver and Ryan (2000) predict that their net assets measure of conservatism (the bias component or BC) varies with three contemporaneous proxies for accounting conservatism: the ratio of accumulated depreciation to gross property, plant and equipment for firms that use accelerated depreciation; the ratio of the sum of research and development expense and advertising expense to sales; and the ratio of the LIFO reserve to total assets. Because BC becomes more negative as conservatism increases, the prediction is that BC varies negatively with those ratios. Ahmed, et al. (2001) assume conservatism evolves as an efficient contracting mechanism to mitigate dividend policy conflicts between shareholders and bondholders, as discussed in Part I. They predict that the more severe such conflicts, the more conservative the firm's accounting choices and the greater the firm's conservatism, the lower will be the cost of debt. In testing both those predictions Ahmed, et al. use Beaver and Ryan's bias component BC and a version of Givoly and Hayn's accumulated accruals measure. The first explicit prediction is: the larger the proxies for dividend conflicts, which are the ratio of dividends to assets, standard deviation of return on assets, and leverage, the more negative are BC and the cumulated accruals. The second explicit prediction is: the more conservatism reflected in the two conservatism measures, the lower will be the cost of debt implied by S&P debt ratings. Beaver and Ryan (2000) estimate the relation between their BC measure of conservatism and their accelerated depreciation, research and development and advertising expense, and LIFO reserve proxies for conservatism. They control for the 12 effects of growth, leverage and investment opportunity set on the bias in book-to-market ratios. Beaver and Ryan find significantly negative effects for accelerated depreciation and research and development and advertising expense, consistent with BC reflecting conservatism. The LIFO reserve proxy however has a significantly positive effect. They explain the LIFO result by the fact that large LIFO reserves in their 1981-1993 sample period occurred in firms operating in mature, unprofitable markets. In essence, the LIFO variable is a proxy for investment prospects. Using both the BC measure and a cumulative accruals measure, Ahmed, et al. (2001) find that, across firms, conservatism increases as conflicts over dividend payout grow. They also find that when the two measures report higher levels of conservatism, the cost of debt is lower. Variation across Countries A number of studies, starting with Ball, Kothari and Robin (2000), predict that Basu measures of conservatism vary across countries with different institutional arrangements.2 Ball, et al. predict that common law countries' use of published financial accounting statement numbers in contracts causes those countries' earnings numbers to be more conservative than those of code law countries. Information asymmetries among parties to code law country firms are resolved privately within the firm without the use of external contracts. Similarly, dividend distributions are determined privately within the firm. Ball, et al. also seek to jointly test hypotheses relating to the litigation and regulatory explanations of conservatism. They conclude that among the common law Other studies using the Basu measure to investigate cross-country variation in conservatism include Pope and Walker(1999) and Giner and Rees (2001) 2 13 countries they investigate, the UK provides the least incentive to be conservative under both explanations and so predict that conservative financial reporting is less evident there. The common law countries in the Ball, et al. (2000) sample are Australia, Canada, the UK and the USA; the code law countries are France, Germany and Japan. Ball, Robin and Wu (2002) hypothesize that four Asian countries Hong Kong, Malaysia, Singapore and Thailand that have some common-law heritage, including accounting standards, will exhibit less conservatism than common law countries. Those four countries have two code law attributes important to the relation between financial reporting and contracting: there are political influences on financial reporting; and information asymmetry is resolved through private channels of communication rather than through public disclosure. Family and other insider networks are more important than contracts based on audited financial statements. Using data from 1985-1995, Ball, et al. (2000) find that the earnings of common law country firms are much more conservative than those of code law country firms. Ball, et al. regress earnings on both positive and negative stock returns, thereby allowing for different coefficients on returns of different signs. Basu coefficient estimates for positive and negative stock returns constructed from that regression for common law countries are .02 and .33 respectively and for code law countries are .04 and .05 respectively. Note the very large differences between the negative return coefficients for the common law countries versus the code law countries. Note also that the adjusted R2 of .14 for the common law country regression and .05 for code law country regression suggests that accounting earnings are more closely associated with contemporaneous 14 stock returns, more "timely", in common law countries. This difference in timeliness is primarily due to the early recognition of losses in the countries that rely on accountingbased contracts the difference in timeliness of profits is much smaller. Ball, et al. (2000) find UK financial reporting is significantly less conservative than in Australia, Canada and the US and marginally more conservative than reporting in the code law countries. Given Ball, et al. `s arguments that the UK has less litigation and regulation than the other common law countries, and presumably the same contracting incentives, this evidence is consistent with one or both of the litigation and regulation explanations. Using the same common and code law countries as Ball, et al (2000), Ball, Robin and Wu (2002) find the four Asian countries to be more similar to code law countries than to common law countries in the conservatism of their firms' financial reporting during 1984-1996. If we accept Ball, Robin and Wu's claim that the important difference between the Asian countries and common law countries is their lack of reliance on public disclosure and contracts, these results support the contracting explanation for conservatism. Contractual Variations Based on the contracting explanation for conservatism, Watts (1977, p. 62) and Leftwich (1983, p. 35) predict that debt contracts adjust GAAP definitions of earnings and net assets to make net assets conservative when GAAP requires recognition of gains and assets that are not verifiable or to which the firm has no immediate legal claim. An example of an excluded non-verifiable asset is goodwill. Earnings from the equity method of accounting for unconsolidated subsidiaries is an example of earnings that are 15 excluded because the firm has no immediate legal claim on them. Without crossguarantees between the subsidiary and the parent, the parent's creditors and debt-holders may not be able to access those gains in bankruptcy. Leftwich's (1983) examination of an American Bar Association guide to writing debt covenants is the only empirical evidence on these propositions. Although his subjective assessment is consistent with conservatism, more formal studies will provide stronger evidence. Discrimination among Conservatism Explanations Evidence from reported estimates of the conservatism measures supports the existence of conservatism. The time-series, cross-sectional and contract variation evidence can help us discriminate between the alternative explanations for conservatism contracting, litigation, taxes and regulation so that we can assess where and when the various explanations influence practice. Overall, the time-series variation in the earnings/stock return relation is consistent with all four explanations for conservatism. The existence of conservatism prior to litigation and regulation is consistent with contracting and tax explanations, but is not predicted by the litigation and regulation explanations. Observing that conservatism continues until the present day is consistent with all four explanations. The litigation explanation predicts the increase in conservatism since 1967 and its variation with changes in the litigation environment. Finally, the overall increase in conservatism under the FASB's regulation of financial reporting is possibly consistent with the regulation explanation. We need finer tests of variations in regulatory conservatism similar to those 16 conducted for litigation periods to assess whether standard-setting generates conservative financial reporting. The cross-sectional tests using US firms provide some weak evidence consistent with the contracting explanation. Because the independent variables in the Beaver and Ryan paper are essentially proxies for conservatism they serve only to validate the BC conservatism measure. Ahmed, et al. find the greater the dividend conflicts, the more conservative the firm's reporting and, ceteris paribus, that more conservatism reduces interest costs. However, the variables used in the tests are jointly and not independently determined. Firms that borrow have fewer growth options and are more likely to pay dividends (Smith and Watts, 1992). Lower growth options can generate both more borrowing and less conservatism. That dependence can only be controlled by carefully specifying the relations between the variables. The cross-country evidence, particularly the Ball, Kothari and Robin evidence directly comparing firms in code-law and common-law countries, supports the contracting explanation. Implications of these results for the tax explanation are not as apparent. Accounting numbers influence tax liabilities in the code law countries and provide incentives for firms in those countries to use conservative accounting, but the deals among stakeholders before publishing the numbers may eliminate that effect. Ball, Kothari and Robin argue that the lower conservatism in the UK jointly supports the litigation and regulation explanations. But because the theory and empirical evidence supporting that prediction is weak, the Ball, Kothari and Robin evidential support for those explanations is also weak. 17 Leftwich's general survey of the debt contract guide provides weak support for the contracting hypothesis. More detailed firm-level contract studies are required to produce stronger evidence. In the aggregate there is more evidence consistent with the contracting and litigation explanations. Although the tax and regulation explanations are also consistent with the evidence, no studies directly address those explanations. NON-CONSERVATISM EXPLANATIONS Non-conservatism explanations can also account for some of the evidence summarized above. One such explanation sees the results being caused by managers exercising abandonment options (Hayn, 1995). Another explanation envisions the evidence resulting from managers managing earnings to maximize their own welfare at the expense of other parties to the firm (Hanna, 2002). Abandonment Option Hayn (1995) predicts and finds an asymmetric relation between earnings and stock returns similar to that predicted by Basu (1997) but estimates a regression of returns on earnings rather than a regression of earnings on returns. The estimated coefficient on earnings in Hayn's regression is expected to be larger for profits than for losses similar to the coefficient in the earnings/return regression being lower for positive stock returns. The reason is that, like conservatism, exercise of the abandonment option produces losses that are more transitory than profits. However, Hayn's explanation for losses' transience is slightly different Hayn reasons that losses are transitory because the firm's 18 managers will not continue to lose money. Instead, they will exercise the abandonment option by liquidating the operations that are generating losses. Losses on unprofitable operations will therefore cease but management will not cease profitable operations. Basu (1997) argues that the abandonment option explanation does not predict the asymmetric timeliness of earnings, the systematic differences in the time-series properties of earnings and cash flows or the time-series variation in the earnings/stock returns relation attributed to litigation. The first two of these conclusions appear to rely on an implicit assumption that the only loss associated with abandonment is the operating loss that induces the abandonment. This assumption is not generally correct. If the abandoned operations involve firm-specific assets that have no market value if those operations become unprofitable, the losses realized on abandonment will also incorporate these potentially substantial reductions in net asset value. This in turn can generate the asymmetric timeliness of earnings and the time-series properties attributed to conservatism. Basu (1997, p.32) also argues that the abandonment option explanation has different implications for the R2 for the earnings/return regression. In particular, he states that Hayn's abandonment explanation implies a higher R2 for the earnings/return regression for good news or positive return firms, while his explanation predicts the bad news or negative return firms will have the higher R2. The basis for the prediction from his explanation is that: in the bad news regressions contemporaneous returns and earnings both incorporate the losses recognized under conservatism, while 19 in the good news regressions contemporaneous returns incorporate all the gains generated by good news, but earnings incorporate the good news only to the extent it produces current earnings. As noted in the previous paragraph, exercise of the abandonment option in bad news periods causes sale or abandonment of assets an action that generates accounting realization of losses in firm-specific asset values in the bad news periods. This produces the circumstance in the first bullet. Good news does not typically generate realization of gains, so the second bullet's circumstances are likely to be the same for both explanations. Once we take into account the recognition of losses in firm-specific asset values, abandonment option explanation can also predict that the bad news firm regression has the higher R2. It is difficult to justify Basu's prediction that the two explanations imply different rankings of the R2of the good and bad news regressions. Consequently, it is not surprising that the evidence in the Hayn and Basu papers does not support Basu's prediction. The major problem with the abandonment option explanation as a general explanation for the conservatism results is its inability to explain, by itself, the systematic understatement of net assets. With unbiased mark-to-market the asymmetry in the earnings/returns regression disappears and there is no net asset understatement. With neutral historical cost accounting with no conservatism, the abandonment option might harvest some unrealized losses and so produce an asymmetry in the recognition of unrealized gains and losses and an understatement of net assets. Note first, however, that a great many losses in asset value do not lead to abandonment to the extent the assets are general and not firm-specific, those losses increase the net present value of the 20 investment and make abandonment less likely. Second, there is also an expansion option to at least partially offset the abandonment option. Firms that have good news increase their investment and that additional investment causes net assets to be valued closer to their market value. A priori it seems unlikely that the abandonment option can explain the significant net asset understatement observed in the empirical literature. Earnings Management Hanna (2002) argues that Basu's findings result from active management of earnings by executives. Hanna points out that three of the five financial reporting problems identified by Arthur Levitt (1998) generate understatement of assets: big bath charges, creative acquisition accounting, and miscellaneous cookie jar "reserves." In each of those reporting problems management inappropriately writes down assets and/or increases liabilities in order to inflate earnings in later years. The assumed motivations are to increase managers' compensation and to mislead the stock market. At the time assets are written down or liabilities increased, their earnings effects are presumed not to affect compensation and/or stock price. For example, earnings before the manipulation might be lower than the lower bound for compensation so no bonus is lost. Alternatively, the manipulation could generate a nonrecurring charge that is excluded from the earnings measure used to determine compensation and/or from the earnings number analysts use in valuation. The "reserve" created by the manipulation is overstated and will be used to raise earnings in the following years in order to increase earnings-based compensation and/or stock price. 21 At first glance, the earnings management explanation seems to fit the conservatism literature results for the following reasons: Establishing the reserve understates net assets, Negative stock returns justify write-offs, potentially providing the asymmetric earnings/stock returns relation, and The initial losses prove to be transitory when followed by the persistently higher earnings generated by using the reserve. Even though earnings management surely occurs, that by itself cannot be the general explanation for the systematic long-term evidence discussed earlier. Earnings management cannot explain important parts of the conservatism evidence and is not a plausible general explanation of the financial reporting evidence consistent with conservatism over long periods of time. The earnings management flows from hypothesized compensation contract and/or stock market effects. Lets begin with the stock market effects. Stock Market Effects Management believes that manipulating earnings fools the stock market, increases firm value and raises their compensation. If management is wrong, in that the market is relatively efficient and on average sees through the manipulation, earnings management motivated write-downs cannot themselves generate the negative stock returns associated with those write-offs. And, earnings management suggests that many firms must have large negative returns every year, not caused by manipulation, to allow write-offs to generate the continuous average accumulation of negative accruals observed every year by Givoly and Hayn (2000). In assessing this possibility, remember that earnings 22 management suggests positive accruals in years when big-baths or write-offs are not taken. Overall, this scenario seems very unlikely to be consistent with the conservatism evidence. Suppose instead that the stock market does not see through the manipulation. Then the stock-price effect depends on whether management designates the manipulative write-off as a nonrecurring charge and. if so, whether analysts and the market ignore nonrecurring charges in valuing the firm. If the write-off is not designated as a nonrecurring charge, presumably analysts and the market treat it no differently than any other earnings component, and there again will be no asymmetry in the earnings/stock return relation. If the manipulative write-off is designated as a nonrecurring charge and analysts and the market ignore it for valuation purposes, in a Basu regression the coefficient on negative returns relating to write-offs and non-recurring charges is zero, not one, indicating no association between manipulative write-offs and returns. There will be an asymmetry, but in the wrong direction, suggesting that the explanation does not fit the evidence. Management Compensation Effects Since stock-price-based compensation implies stock price effects and earnings/stock price relations that are difficult to reconcile with the overall evidence, let's turn to the earnings-based compensation motivation for earnings management that utilizes write-offs. It seems that under earnings-based compensation management seeks to overstate cumulative earnings, and net assets, to increase their compensation and to take advantage of the non-linearity of the bonus formulas by transferring earnings 23 between years.3 A priori, I expect the more general income-increasing effect to dominate so that without conservatism and other control mechanisms being present, net assets will be overstated, not understated. Yet the evidence strongly suggests net assets are understated. Using earnings-based compensation to explain this suggests wholesale stock-market-wide manipulative transfers of earnings across years, an implication that seems quite implausible. Like the abandonment option explanation, the earnings management explanation is plausible and not mutually exclusive to the conservatism explanation, such that all three mechanisms may be at work. The important point, though, is that the abandonment and earnings management explanations are not individually or jointly consistent with the overall pattern of evidence on conservatism. For example, neither individually nor in combination can they plausibly explain the systematic understatement of net assets. Overall conservatism is a much better explanation for the pattern of evidence. RESEARCH OPPORTUNITIES Part I of the paper presents the elements of a theory of conservatism and Part II presents the evidence on the theory. There are a host of research opportunities in both areas. Theory I interpret theory broadly including both formal theory ("modeling") and informal theory. Formal theory can play an important role, perhaps by tightening the
Bonus formulas generally have an earnings level below which no bonus is earned and some have caps above which no incremental bonus is earned. These features can induce a non-linearity in the earnings/compensation relation and provide an incentive to shift earnings across years.
3 24 logic in the conservatism explanation. Kwon, Newman and Suh (2001) make a start by introducing limited liability in a formal model to explain conservatism. Others could incorporate features that more closely resemble arguments in the literature. Theory of both kinds should concentrate on the real issues in financial reporting. In researching financial reporting's role in capital markets we too often ignore real problems in favor of problems we think we can solve. Much of the financial reporting literature falls into the latter category. Part I explains how the manager's production of financial reports and that role generates agency costs that can differentially affect the value of alternative accounting methods for financial reporting. Yet the reporting literature tends to ignore these agency costs. If information is free and there are no agency costs there is no role for accountants or financial reports. Accounting and reporting exist because of such costs. Assuming the costs away in order to research and gain reporting inferences from the association between stock prices and accounting methods is like assuming the costs of mining are zero in order to study the viability of alternative coal-mining methods. Financial reporting theory that incorporates the costs of managerial response to alternative accounting methods is essential. There are many potentially important theory topics in studying reporting costs and conservatism: One of the more important is the extent to which verification standards differ for gains and losses, A related topic is: why do courts punish overstatement more than understatement of net assets? 25 Is the explanation for the courts' behavior to the contracting arguments for conservatism and/or to potential explanations for similar behavior by politicians and regulators? Theory to explain the behavior of courts and regulators in bodies such as the SEC could be invaluable in further empirical testing of conservatism. Theory can also play an important role in developing the abandonment option and earnings management explanations. Hopefully those explanations can be fleshed out to the point that empirical studies can test when and where the explanations play a role in accounting accruals. Empirical Studies Because the survey of the empirical literature raises many potential research opportunities, I cannot possibly address all of them here. Instead I give a few examples to illustrate the nature of studies that could be conducted. A point made in my literature survey is that there is a lack of time-series studies of changes in taxes and regulation similar to Basu's study of variation in litigation costs. Performing such studies requires identifying significant changes in taxes or regulation that are expected to change the degree of conservatism. Identifying similar changes in contracting is also difficult, but I look for empirical work that bypasses the effects of changes in contracting on conservatism and focuses on the effect of GAAP changes on conservatism and how contracting responds to those changes. Changes in debt contracts provide fertile grounds for research on conservatism, particularly in view of the debt contracting innovations emerging in the 1990s. One of those innovations is the more frequent use of "frozen" GAAP GAAP existing at the 26 time of the contract to calculate the contracts' accounting numbers (Beatty, Ramesh and Weber, 2002). I expect that contracting changes of this kind relate to both the frequency of changes in GAAP and to changes in the conservatism of GAAP and that the contracts attempt to reduce the effect of non-conservative GAAP changes. There are a few contracting changes that appear to contradict the contracting explanation. Although most debt contracts still exclude intangible assets from assets, some agreements do not. For example, Weil (2002) describes the problems caused by including intangible assets in AOL's debt contract's definition of net assets. My guess is that those inclusions are mistakes that I predict will disappear as the error becomes evident from the experiences of AOL and other companies. If the inclusions of intangibles do not disappear, one could investigate the circumstance in which they are used. Along the same lines as my prediction that AOL made a mistake in its debtcontracting, it will be interesting to see if a significant increase in frauds emerges from the FASB's apparent move towards firm valuation as reflected in SFAS No. 142. Also, because auditors do not have a comparative advantage at valuing divisions of firms, what steps are they taking to reduce potential litigation costs from that activity? Casual evidence suggests some use of external experts. If experts are used, empirical studies of any cross-sectional variation in that usage could provide insights. Cross-sectional predictions of variation in conservatism are difficult to test because of the inherent endogeneity problem mentioned in my earlier discussion of Ahmed, et al. (2001). That said, a cross-sectional prediction related to both conservatism and the abandonment option explanations is worth investigating. Losses accompanying 27 exercise of abandonment options are likely to be larger for investments that involve more firm-specific assets. Losses on investments involving more general assets that have more alternative uses are likely constrained by the values of those assets' values in the alternative uses. Hence, I expect to observe greater conservatism in the financial reporting of firms with firm-specific assets. CONCLUSIONS Overall, existing evidence that suggests accounting is conservative is most consistent with contracting and litigation explanations, but some of the evidence is also consistent with tax and regulatory explanations. Further, there is reason to believe that these four explanations are not independent, that conservatism is driven by a concern with overpayment by contracting parties, courts and government. The evidence does not rule out earnings management or abandonment option effects, but instead suggests conservatism's effects are more pervasive. Because of the importance of reporting costs that drive the conservatism explanations, accounting researchers should incorporate those costs into their predicted effects of alternative accounting and reporting methods. Conservatism provides many opportunities to research the effects of those costs and provide a basis for those predictions. 28 REFERENCES Ahmed, A.S., B. Billings, M.S. Harris and R.M. Morton. 2001. Accounting conservatism and cost of debt: An empirical test of efficient contracting. Working paper, Syracuse University. Ahmed, A.S., R.M. Morton and T.F. Schaefer. 2000. Accounting conservatism and the valuation of accounting numbers: Evidence on the Feltham-Ohlson (1996) model. Journal of Accounting, Auditing & Finance 15 (Summer): 271-292. Ball, R., and R.L. Watts, 1972. Some time series properties of accounting income. Journal of Finance 27 (June): 663-682. Ball, R., S.P. Kothari and A. Robin. 2000. The effect of international institutional factors on properties of accounting earnings. Journal of Accounting & Economics 29 (February): 1-51 Ball, R., A. Robin and J.S. Wu. 2002. Incentives versus standards: Properties of accounting income in four Asian countries. Working paper, University of Chicago. Basu, S. 1995. Conservatism and the asymmetric timeliness of earnings. Ph.D. thesis, University of Rochester. Basu, S. 1997. The conservatism principle and the asymmetric timeliness of earnings. Journal of Accounting & Economics 24 (December): 3-37. Beatty, A., K. Ramesh and J. Weber. 2002. The importance of accounting changes in debt contracts: The cost of flexibility in covenant calculations. Journal of Accounting & Economics 33 (June): 205-227. 29 Beaver, W.H., and S.G. Ryan. 2000. Biases and lags in book value and their effects on the ability of the book-to-market ratio to predict book return on equity. Journal of Accounting Research 38 (Spring): 127-148. Brooks, L., and D. Buckmaster. 1976. Further evidence of the time series properties of accounting income. Journal of Finance 31(5): 1359-1373 Dechow, P., A. Hutton, and R. Sloan. 1999. An empirical assessment of residual income valuation model. Journal of Accounting & Economics 26 (January): 1-34. Elgers, P.T., and M.H. Lo, 1994. Reduction in analysts' annual earnings forecast errors using information in prior earnings and security returns. Journal of Accounting Research 32 (Autumn): 290-303. Feltham, G. and J.A. Ohlson. 1995. Valuation and clean surplus accounting for operating and financial activities. Contemporary Accounting Research 11 (Spring): 689731. Feltham, G. and J.A. Ohlson. 1996. Uncertainty resolution and the theory of depreciation measurement. Journal of Accounting Research 34 (Autumn): 209-234. Giner, B. and W.P. Rees. 2001. On the asymmetric recognition of good and bad news in France, Germany and the United Kingdom. Journal of Business Finance & Accounting 28 (November/December): 1285-1331. Givoly, D., and C. Hayn. 2000. The changing time-series properties of earnings, cash flows and accruals: Has financial accounting become more conservative? Journal of Accounting & Economics 29 (June): 287-320. Hanna, J.D. 2002. Conservatism and accounting measurement. Working paper, University of Chicago. 30 Hayn, C. 1995. The information content of losses. Journal of Accounting & Economics 20 (September): 125-153. Holthausen, R.W., and R.L. Watts. 2001. The relevance of value-relevance literature for financial accounting standard setting. Journal of Accounting & Economics 31 (September): 3-75. Kothari, S.P., T. Lys, C.W. Smith and R.L. Watts. 1988. Auditor liability and information disclosure. Journal of Accounting, Auditing and Finance 3 (Fall): 307-339. Kwon, Y.K., D.P. Newman and Y.S. Suh. 2001. The demand for accounting conservatism for management control. Review of Accounting Studies 6 (March): 29-51. Leftwich, R.W. 1983. Accounting information in private markets: Evidence from private lending agreements. The Accounting Review 58 (January): 23-42. Levitt, A. 1998. Text of speech given on September 28, 1998 at New York University Center for Law and Business. Myers, J. 1999. Implementing residual income valuation with linear information dynamics. The Accounting Review 74 (January): 1-28. Pope, P.F., and M. Walker. 1999. International differences in the timeliness, conservatism, and classification of earnings. Journal of Accounting Research 37 (Supplement): 53-87. Rockoff, H. 1995. Price and production controls in World War II. Working paper, Rutgers University. Shackelford, D.A., and T. Shevlin. 2001. Empirical tax research in accounting. Journal of Accounting & Economics 31 (September): 321-387. 31 Smith, C.W., and R.L. Watts. 1992. The investment opportunity set and corporate financing, dividend, and compensation policies. Journal of Financial Economics 32 (December): 263-292. Stober, T. 1996. Do prices behave as if accounting book values are conservative? Crosssectional tests of the Feltham-Ohlson (1995) valuation paper. Working paper, University of Notre Dame. Watts, R.L. 1977. Corporate financial statements, a product of the market and political processes. Australian Journal of Management 2 (April): 53-75. Watts, R.L. 1993. A proposal for research on conservatism. Working paper, University of Rochester (presented at American Accounting Association national meeting, San Francisco, CA). Weil, J. 2002. Another write-off may loom for AOL. The Wall Street Journal, August 23, p. C1. Zeff, S.A. 1972. Forging Accounting Principles in Five Countries: A History and Analysis of Trends 1971. Arthur Andersen Lecture Series. Champaign, IL: Stipes Publishing Company. 32 ...
View Full Document
This note was uploaded on 03/10/2008 for the course FINC 732 taught by Professor Noe during the Spring '08 term at Tulane.
- Spring '08