Abstract An issue of fundamental importance to accountants concerns the qualities possessed by, or that should be possessed by, accounting information. The usual intuition is that any distortion of accounting numbers relative to the true underlying cash flows of the firm is generally undesirable from the standpoint of investors. Indeed, among the primary qualitative characteristics of accounting information championed by the FASB in their conceptual framework is reliability, which is defined as "the quality of information that gives assurance that it is reasonably free of error and bias and is a faithful representation" (emphasis added). Normative statements about the desirable qualities of accounting information are often problematic since such information typically serves multiple purposes. For example, Gjesdal (1981) identifies a decision-making rote and a stewardship role for accounting information and shows that the two needs may not be served equally well. The objective of this paper is to demonstrate that "distorted" accounting information may actually be preferred if the focus is on the stewardship value of accounting information. "Distorted" accounting information is characterized as information signals that are biased relative to the expected value of the firm or that measure the value of the firm with error (white noise). Bias and noise are meant to be representative of the many inadequacies usually attributed to accounting information. Current Generally Accepted Accounting Principles prevent or delay the recognition of certain assets and liabilities and their income statement counterparts, generating what may be thought of as bias. For example, the asymmetric recognition of certain gains and losses, accounting for R&D and advertising expenditures, and the absence of information about customers and suppliers in current financial statements imply that the consequences of certain current managerial activities are not reflected in accounting information. This is bias. Similarly, current Generally Accepted Accounting Standards require numerous estimates generating what may be thought of as noise. For example, subjective assessments such as estimated useful lives of assets, loss contingencies, impairment of asset values, and the allocation of purchase price to individual assets and liabilities in corporate acquisitions are likely to introduce noise into accounting information, even in the absence of bias. A crucial assumption underlying our analysis is that managers allocate effort across many managerial activities, all of which contribute to improving the value of the firm. For example, we can imagine an executive dividing his energies across activities like controlling costs, implementing total quality management initiatives, developing a well-trained workforce, and creating an operating environment where innovation can flourish. While activities such as these have cash flow implications for a firm, we argue that it is difficult, or impossible, to disaggregate the results of operations into individual performance measures that cleanly isolate the effects of the different activities. As a result, compensation contracts in our analysis are based on aggregate performance measures such as accounting income and share price that do not unambiguously distinguish between individual managerial activities. This does not preclude the use of detailed or disaggregated accounting information other than net income as a measure of managerial performance. Rather, it captures the idea that accounting systems cannot realistically measure the economic consequences of all managerial activities in detail. Some aggregation is inevitable, and this imposes a contracting constraint. In our model, limiting the available performance measures to accounting information and share price renders efficient managerial incentives unattainable and creates a potential contracting value for bias and noise. Given that aggregated accounting information and share price are the only available performance measures, we derive conditions under which biased accounting information can be effectively utilized to mitigate the limitations of aggregated measures by better balancing incentives across different managerial activities. In particular, we show that biased accounting information, even if it contains substantial noise, can be better than unbiased accounting information, even if it contains no noise, given that price is also available as a measure of managerial performance. As long as there is a need to provide different incentives for different managerial activities, there is a need for biased accounting information since bias enables the de facto observation of the individual components of output. In addition to examining the role of biased accounting information, our analysis demonstrates the benefits of noisy accounting information. This suggestion seems counterintuitive since noise imposes risk on the manager without generating any benefits. While the usual intuition holds in most settings, it fails to recognize that there may be an equilibrium relation between accounting information and an endogenously determined share price. A reduction in the level of noise in accounting information motivates investors to curtail private information acquisition, which dilutes the information content of price and thus lessens the usefulness of price as a measure of managerial performance. Therefore, the optimal level of noise in accounting information is a tradeoff between the usefulness of price relative to the usefulness of accounting information as measures of managerial performance. It is important to emphasize that, in this paper, the benefit of distorted accounting information (bias or noise) that accrues to shareholders is independent of any managerial motivations to manipulate the disclosure system that characterize the "earnings management" literature (see Schipper 1989). Moreover, by focusing exclusively on the stewardship value of information, we show that bias and noise are desirable because they create a "bigger pie" to be shared by all. However, we do not consider the impact of bias and noise on other aspects of shareholder welfare. Bias and noise in accounting disclosures will also impact investors' risk-sharing opportunities both through the direct effect of the distorted public disclosure and the indirect effect on private information acquisition (see, e.g., Diamond 1985). Any overall equilibrium, of course, would have to consider the interaction of all these forces. The exact nature of these tradeoffs remains an unresolved issue.
Bushman et al. (Fri,) studied this question.