U.S. and international accounting standards mandate recognition and/or disclosure of fair value information for an increasing number of items. Fair value estimates require judgment, introducing the possibility of biases in measurements, measurers, and/or models. In addition, unanticipated changes in market risk result in realized values differing from fair value estimates. Accompanying the shift to fair value accounting is the emergence of voluntary disclosures in audited financial statement footnotes that alert investors to management’s concerns about the reliability of mandated fair value information. We refer to such disclosures as reliability disavowals (hereafter, disavowals). In our paper, Are voluntary disclosures that disavow the reliability of mandated fair value information informative or opportunistic? forthcoming in the Journal of Accounting and Economics as published by Elsevier, we examine whether disavowals are informative; that is whether they are a truthful revelation by management that their fair value estimates are unreliable. We also consider that managerial opportunism may contribute to—or even solely motivate—the decision to disavow.
Two issues motivate our study. First, little is known about voluntary disclosures that communicate concerns about the reliability of fair value disclosures. Scholars and regulators acknowledge that transparency requires recognized fair value estimates be supplemented with disclosures about reliability. In the U.S., this discussion culminated in SFAS 157, “Fair Value Measurement”. We document the existence of supplemental voluntary disclosures prior to the passage of SFAS 157, and we examine their veracity. Second, reliability is a primary qualitative characteristic of accounting information (SFAC No. 2) and impacts the weight financial statement users place on accounting information (Cotter and Zimmer 2003). Because voluntary disavowals can affect the degree to which financial statement users incorporate disavowed information into their judgments and decisions, it is important to understand whether disavowals do, in fact, reflect legitimate reliability concerns. Such research is particularly timely given concerns made salient by the global financial crisis.
Although managers disavow a variety of fair value estimates, we focus on the disavowal of the fair value estimate of employee stock option compensation (SOC) mandated under SFAS 123. We focus on SOC disavowals because reliability concerns about SOC expense are significant. In addition, concerns about the reliability of SOC fair value estimates are similar to concerns about the reliability of Level 3 estimates in the SFAS 157 fair value hierarchy. Reliability concerns associated with SFAS 157 center upon the unobservable model inputs on which Level 3 estimates are based. Similarly, SOC disavowals emphasize the subjectivity involved in estimating unobservable model inputs.
We hand collect disavowal data from 10-K reports for the population of ExecuComp firms in the year they adopted SFAS 123 (fiscal 1996 for most firms) and in each fiscal year from 2001 through 2005. Approximately 6.5% of sample firms disavow in the adoption year, and an average of 13.6% of the firms disavow in each fiscal year from 2001 through 2005. At least 2.5% of the firms either initiate or discontinue a disavowal disclosure in each fiscal year. Clients of Ernst and Young (E&Y) are approximately four times more likely to disavow than are the clients of other national audit firms, most likely due to E&Y including a disavowal as a sample supplemental disclosure in its SFAS 123 implementation guidance (Ernst and Young 1995).
We provide evidence on three predictions that will hold if disavowals of SOC fair value estimates reflect reliability concerns. We focus our analyses on firms’ estimates of stock return volatility over the expected life of the options because this estimate is a particularly critical option pricing model input. In addition, most SOC disavowals explicitly reference the highly subjective nature of this particular estimate.
We conduct three tests. The first examines whether the disavowal decision is associated positively with ex ante proxies for the unreliability of the stock volatility estimate and/or management opportunism. We assume that volatility is more difficult to estimate, and thus more likely to be unreliable, when (1) the firm’s trading history is short, (2) the firm’s historical volatility is itself volatile, (3) the firm has no exchange-traded options, or (4) there is a large difference between historical volatility and independent estimates of future volatility. Consistent with prior research, we assume management has incentives to act opportunistically when (1) SOC expense is high, (2) abnormal CEO compensation is high, or (3) the proportion of the firm’s total options granted to its top-five executives is high.
The disavowal decision is associated positively with all four proxies for unreliable stock volatility estimates in the initial period and one proxy for unreliable volatility in the 2001–2005 period. In contrast, none of the three opportunism proxies are associated positively with the disavowal decision in the initial period, but one, abnormal compensation, is positively associated in 2001–2005.
Our second and third sets of tests explore properties of managers’ volatility forecasts. In the second set of tests we examine forecast bias. We assume that a firm’s various SOC disclosures reflect the same managerial motives. Accordingly, if disavowals are opportunistic, disavowal managers’ volatility estimates also reflect opportunism and thus are downward biased. Similarly, if disavowals are informative, disavowal managers’ volatility estimates are truthful and thus unbiased. We assess forecast bias by comparing managers’ volatility estimates to both ex ante benchmark volatility forecasts (i.e., forecasts based on historical volatility, implied volatility, or the combination of firm and industry historical volatility) and ex post (i.e., realized) volatility. In support of informative disavowals, disavowal managers’ volatility estimates are not downward biased relative to realized volatility. Nor are their mean volatility estimates downward biased relative to any of the three ex ante volatility benchmarks. However, disavowal managers’ median volatility estimates exhibit downward bias relative to two of the three ex ante benchmarks, consistent with some disavowals being opportunistic.
Our third set of tests investigates forecast difficulty. Greater forecast difficulty implies less reliable volatility estimates, suggesting greater forecast difficulty for disavowal firms if disavowals are informative. Our results indicate that disavowal firms have greater volatility of realized volatility and that their benchmark forecasts have larger forecast errors. These results are consistent with disavowal firms facing a more difficult forecasting environment. Thus, their disavowals are informative.
Overall, results from each of the three tests support the hypothesis that disavowals are informative. We also find some support for disavowals being opportunistic. The results from our first test indicate that support for the informativeness hypothesis is stronger in the SFAS 123 adoption year than in the 2001–2005 period, while support for the opportunism hypothesis is concentrated in the 2001–2005 period. The diffusion-of-innovations literature provides a possible explanation for this pattern of results. Early adopters of the disavowal (i.e., disavowal firms in the SFAS 123 adoption year) responded to changes in the disclosure environment (i.e., the SFAS 123 requirement to disclose SOC estimates) with disclosure innovation. In contrast, the factors that motivate innovation and early adoption of the disclosure only partially explain the behavior of subsequent adopters.
Our informativeness results are important because reliability concerns impact the weight that financial statement users place on accounting information, and disavowals can affect the judgments of financial statement users. In addition, prior commentary argues that transparency requires recognized fair value estimates be supplemented with disclosures about reliability. The emergence of voluntary disavowal disclosures is consistent with a perception by managers that there are benefits to such supplemental disclosures.
The full paper is available for download here.