The overall uncertainty inherent in financial statements has increased in recent decades, but the related reports and required level of audit assurance have changed very little. In our study, “Extreme Estimation Uncertainty in Fair Value Estimates: Implications for Audit Assurance” (Christensen et al. 2012a), we examine estimates reported by public companies and find that estimates based on management's subjective models and inputs contain estimation uncertainty that is many times greater than typical audit materiality. We do not question the value that audits provide to the marketplace or the ability of auditors to deploy up-to-date auditing techniques. Rather, we suggest that the convergence of relatively recent events is placing an increasingly difficult and perhaps, in some cases, unrealistic burden on auditors. We discuss potential changes to financial reporting and auditing standards that may improve the information provided to users, and also address the concerns raised in our study.
The prevalence of fair value and other estimates in financial statements, as well as their inherent estimation uncertainty, has increased dramatically in recent years and has been recognized as a significant issue by regulators.1 In our recently published study, “Extreme Estimation Uncertainty in Fair Value Estimates: Implications for Audit Assurance” (Christensen et al. 2012a), we provide examples from publicly available data that small, ostensibly reasonable changes to fair value inputs can result in changes to account values as large as 50 times audit materiality. Although complexity and estimation uncertainty have increased over time, the content of the audit report and the information conveyed on the face of the financial statements have changed relatively little. Our study questions whether auditing and financial reporting standards provide for effective conveyance of the uncertainty contained in financial statements. Further, we consider whether the expectations of standard setters, regulators, and users are placing an increasingly difficult, and likely unrealistic, burden on auditors.2
Our study does not question the value provided by audits or auditors' competence. Rather, we suggest that auditors are operating as well as can be expected given the current state of auditing and financial reporting standards. No amount of auditing can remove the underlying estimation uncertainty in reported values that are determined by management-derived estimation models that are hypersensitive to small changes in inputs. We consider how recent events have seemingly resulted in higher expectations and tighter constraints, thus placing a potentially unrealistic burden on auditors, essentially requiring them to provide a product that may be beyond their reach. We also suggest potential revisions to auditing and reporting standards that would improve the nature and quality of information provided to users, as well as revisions that would clarify the auditor's responsibilities with respect to significant estimates that contain extreme measurement uncertainty.
PUBLIC COMPANY EXAMPLES OF ESTIMATION UNCERTAINTY
In our study, we use estimates reported by Wells Fargo and General Motors (GM) to illustrate how changes in estimation model inputs impact fair value point estimates. In our illustrations, we compare estimation uncertainty in the point estimates, as reported by management, to audit materiality for the financial statements taken as a whole.3 The level of estimation uncertainty highlights potential challenges that auditors face in providing assurance on account balance estimates with uncertainty ranges that often are many times larger than materiality for the financial statements taken as a whole.
Wells Fargo's Mortgage-Backed Securities
We include in this paper Table 1 from our original study to analyze Wells Fargo's mortgage-backed securities from 2003 to 2008. In Panel A, the data in the first column are taken directly from Wells Fargo's sensitivity analysis showing the effect of a 200 bps (i.e., basis points) change in the interest rate input, whereas the other two columns report calculations from our study of (1) the increase in units of materiality implied by the potential effects reported in the first column, and (2) the bps changes that would induce changes in the values of Wells Fargo's mortgage-backed securities equal to the size of the audit materiality estimates shown in Table 1, Panel B. We show that changes to the interest rate input as small as 3.70 and 27.04 bps in 2008 and 2007, respectively, would yield material swings in this account's reported value—changes that directly impact income (in this instance, other comprehensive income).
Although Wells Fargo did not indicate that a change in interest rates of 200 bps was a reasonable range of possible values, it seemed to be within reason given recent changes in market rates.4 We are not suggesting that Wells Fargo used the “wrong” rate; rather, we are highlighting the fact that changes to a management-selected input (because it is relatively unobservable in the market and, thus, requires management judgment) as small as 3.70 and 27.04 bps materially affect the final estimate's value, and that such small changes are almost certain to be within a reasonable range. We also examined Wells Fargo's fair value estimates for mortgage servicing rights and found similar results. Thus, our study provides evidence that the values reported in the financial statements are sensitive to very small changes in just one valuation input, which, in turn, suggests that reasonable estimation ranges can be many times greater than quantitative materiality.
General Motors Pension Accounting
In 2009, the SEC accused GM of manipulating two key inputs used in accounting for their pension plan (SEC versus General Motors 2009). Specifically, GM was accused of using aggressively high discount rates and expected returns on plan assets. By using a discount rate that was aggressively high, but still within an ostensibly reasonable range, GM was able to reduce its pension liability and accompanying pension expense by multiples of audit materiality. Additionally, by using an aggressively high expected return on plan assets, GM was able to inflate its pretax earnings (by reducing pension expense) by multiples of materiality.5 Given the reasonable range of possible values yielded by changes in just two of many pension-related estimation inputs, this example further illustrates the hypersensitivity of reported values to small changes in highly uncertain inputs.
The Wells Fargo and GM cases highlighted in our study demonstrate the extreme estimation uncertainty in some significant accounting estimates. Hypersensitivity to small changes in unobservable inputs, the large number of such inputs, the large number of estimates in the financial statements of complex entities that involve such inputs, and the level of management discretion involved in accounting estimates all add to the burden placed on auditors in providing assurance surrounding these estimates.
CONVERGENCE OF RECENT EVENTS
In our paper, we provide a comprehensive examination of recent changes in the auditing and reporting environment for public company audits that place an increasingly difficult burden on auditors, who are required to provide a high level of positive assurance that financial statements—including those containing items subject to enormous inherent estimation uncertainty, such as those described above—are fairly stated in all material respects. Here, we provide a brief overview of some of these changes and their impact on the audit profession.
Complexity in financial reporting, including the use of subjective fair value models, has increased dramatically over the past two decades. Despite this increase in the level of financial statement complexity, auditors are required to provide reasonable assurance, defined in auditing standards as “high” assurance, that the financial statements taken as a whole are fairly stated in all material respects.6 Collectively, practitioners and practice guides (e.g., American Institute of Certified Public Accountants [AICPA] Audit Guide: Audit Sampling ) have determined that “high assurance” is consistent with at least 90 to 95 percent confidence that the auditing procedures applied to individual accounts or classes of transaction would detect a material misstatement or material weakness in controls.7
Further, auditing standards and regulators have increasingly focused on quantitative materiality, particularly with respect to the implied precision around net income. Quantitative materiality, commonly computed as 5 percent of pretax income, is a relatively tight margin, especially when highly uncertain estimates are considered. Standards on auditing estimates indicate that it is desirable for the estimation uncertainty range to be less than performance materiality for the auditor to conclude whether the estimate is fairly stated (AICPA 2010, AU-C 540.A100). In other words, the audit opinion could be interpreted as providing at least 90 to 95 percent confidence that the reported values are fairly stated to within plus or minus 5 percent of pretax income. In the absence of highly uncertain estimates within the financial statements, this goal may be achievable; however, we suggest that in the current environment, it often is not.
When financial statements contain extreme estimation uncertainty, as in the Wells Fargo and GM cases discussed above, auditing standards indicate that the estimate may be recognized by the entity, and that the auditor is then required to make “additional evaluations” of the company's process. We note that these additional evaluations appear to provide evidence that is more consistent with negative than with positive assurance (AICPA 2010, AU-C 540). However, there is no relaxation to the requirement that the auditor provide high assurance that the financial statements containing the reported values are fairly stated. Another significant challenge faced by auditors is that with extreme estimation uncertainty, very small inappropriate changes (intentional or accidental) by management in an input used to develop an accounting estimate could result in a material misstatement, as was alleged in the GM case. The difficulty for auditors is that such small changes arguably fit within a seemingly reasonable range of uncertainty associated with a particular input.
We believe that auditors are doing their best within the requirements imposed by standard setters and regulators, but we suggest that it is time for those who set and regulate standards to consider ways to more clearly convey where extreme estimation uncertainty exists within financial statements, and to reconsider auditors' ability to provide positive, high-level audit assurance on these estimates.
Although we acknowledge that there is no single path forward, our study offers several recommendations for changes regarding the format of financial statements, audit reports, and the nature of audit assurance. With regard to the financial statements themselves, we recommend that standard setters consider ways to convey the accounts whose values are subject to extreme measurement uncertainty.8 This could involve a number of strategies, from simply flagging or highlighting accounts with measurement ranges greater than materiality, to disclosing a reasonable or practical range of uncertainty associated with estimates at some level of precision either numerically or graphically, in a form such as a box plot. Some academic research has analyzed how alternative reporting formats for uncertain estimates may impact users; for additional information, see studies such as Clor-Proell et al. (2010), Christensen et al. (2012b), and Glover et al. (2005).
With regard to the audit report, our study recommends that audit standard setters and regulators consider changes in the type of assurance provided for accounts with extreme estimation uncertainty and the form and content of the audit report. For example, for highly uncertain estimates, the auditor would provide positive assurance with respect to the rigor and soundness of the entity's estimation model, valuation processes, and related controls, and provide negative assurance with respect to the fairness of the reported point estimate. Such an approach is more honestly representative of the type of assurance that auditors are now able to provide for point estimates with extreme estimation uncertainty. Alternatively, standard setters and regulators could consider different levels of assurance (e.g., high, moderate, low) for accounts, depending on the verifiability and precision of the reported point estimate, similar to the “grades” proposed by the PCAOB's Standing Advisory Group (SAG) (PCAOB 2010, SAG April 7–8 meeting minutes). Finally, with respect to potential changes to the form and content of the audit report, we recommend that the report include a list of accounts with extreme measurement uncertainty (i.e., uncertainty that constitutes multiples of materiality).
Over the last few decades, business transactions and financial reporting standards have become more complex and estimates more uncertain, yet the face of the financial statements continues to contain point estimates without conveying the level of uncertainty. Additionally, the nature of audit assurance apparently provided to users on estimates containing extreme measurement uncertainty has changed very little, and the format of the auditor's report has essentially remained the same. Using publicly available data, we demonstrate that some estimates reported in public company financial statements contain extreme measurement uncertainty, where very small changes to unobservable inputs can change the value of the estimate and, consequently, net income by many multiples of materiality.
We question whether transaction and reporting complexity has outstripped the auditor's ability to provide a high level of positive assurance that financial statements are fairly stated in all material respects when they include accounts that contain extreme estimation uncertainty. We are not condemning fair value estimates, the audit of fair value, or the auditor's ability to apply valuation techniques. Rather, it is an acknowledgement that no amount of auditing can remove the extreme uncertainty inherent in reported values derived from management's valuation models based on unobservable inputs subject to estimation uncertainty. We suggest that financial reporting and auditing standards should be changed to produce more transparency and clearly convey the level of uncertainty in estimates on the face of the financial statements and in the audit report. Further, we also suggest that standard setters consider the nature of assurance that auditors should be required to provide for highly uncertain estimates. By implementing such changes, we believe that auditors could provide a more honest and faithful representation of the assurance that they actually are able to provide, and that users of financial statements would be better informed about the uncertainty inherent in the financial statements.
Regulators have recognized the importance of addressing this topic, as evidenced by a roundtable hosted by the Securities and Exchange Commission (SEC) in November 2011 (SEC 2011). We note that our study was cited as a source document in the roundtable's briefing document.
We calculated materiality by multiplying pretax income by 5 percent, a benchmark frequently used in practice.
For example, the prime interest rate went from 4 percent in 2003 to 8.25 percent in 2006, then back to 3.25 percent in 2008. The Federal Reserve in St. Louis publishes historical prime interest rate data. We note that historical volatility in an input, such as the interest rate, does not necessarily translate to estimation uncertainty as of a specific measurement date. The focus of this study is estimation uncertainty as of the measurement date, and we highlight situations in which management's uncertain inputs are selected from a reasonable range that is wide enough to yield materially different valuations depending on the input value selected.
Specifically, the SEC argues that the 2003 discount rate was overstated by 75 bps, resulting in an overstatement to pretax earnings of $510 million (SEC v. General Motors 2009). Our study shows that this overstatement was equal to 3.4 times audit materiality. Our study also presents analyses showing that reasonable changes to key inputs resulted in material swings in the value of GM's accumulated post-retirement benefit obligation (APBO) and related expense account, Other Post-Employment Benefits (OPEB).
In 2004, the Public Company Accounting Oversight Board (PCAOB) was the first to clarify the technical definition of “reasonable assurance” as “high” assurance (PCAOB 2004, AS 2.17, Interim AU 230.10). The International Auditing and Assurance Standards Board (IAASB) and the Accounting Standards Board (ASB) followed the PCAOB, and now similarly define reasonable assurance as a high, but not absolute, level of assurance (International Federation of Accountants [IFAC] 2009, Clarified AU 200.5, ISA 200.5).
Many of the insights shared in this section come from one of the authors' first-hand knowledge gained from working in and with national and global audit methodology and policy groups of Big 4 and other international accounting firms.
Standard setters and regulators argue that uncertainty already is addressed in the footnotes to the financial statements. However, we do not believe that enhanced disclosure will be adequate, as research indicates that disclosures do not provide the same information content as the numbers on the actual financial statements. We believe that additional disclosure will not remove the impression of certainty that currently is being conveyed in the point estimates reported in the financial statements.