FASB’s efforts to expand the use of fair value accounting have rekindled the debate on the costs and benefits of the approach. Proponents argue that expanding fair value accounting rules will make financial reporting more relevant to users. In support of this view, the CFA Institute Center, which represents the views of investment professionals, argued in a 2010 comment letter to FASB that fair value is the appropriate basis for measuring all financial instruments (“Fair Value as the Measurement Basis for Financial Instruments,” September 2010, http://cfa.is/1Xw1LkH). Former SEC chairman Arthur Levitt also contended in an interview that fair value provides the kind of transparency essential to restore public confidence in U.S. markets (“U.S. Moves to Ease ‘Fair Value’ Accounting Rules,” The Globe and Mail, Apr. 3, 2009, http://bit.ly/1WsgEpZ). In contrast, opponents of fair value have raised concerns about the impaired reliability of reported financial information, arguing that fair value accounting often relies on subjectivity in the measurement of various financial statement items. In particular, when active markets do not exist for assets and liabilities, the business must estimate the fair value by applying present value techniques, which opens up the possibility of intentional or unintentional errors in accounting estimates. Recent research into fair value accounting and its effects on investor confidence sheds some light on this debate.

Statement of Financial Accounting Standards 157

Before FASB issued Statement of Financial Accounting Standards (SFAS) 157, Fair Value Measurements, in September 2006, the amount of fair-valued assets measured by management was not available to financial statement users. Under SFAS 157, exchange-listed entities are required to classify their fair-valued assets into three categories (level 1, level 2, and level 3) on the basis of their liquidity or input reliability. If an asset has a liquid market with a readily available price, it is classified as level 1. If an asset does not have a liquid market, but a similar type of asset has a relevant market, it is classified as level 2. If an asset does not have any relevant market base for valuation, its fair value is measured by applying a valuation model, and it is classified as level 3. Because level 3 inputs are unobservable, the issue of the trustworthiness of level three asset values has been quite controversial. The adoption of SFAS 157 provided an opportunity to study how stock market participants perceive and value these opaque assets.

Fair Value Assets in the Financial Industry

Most studies of fair value accounting focus on the financial industry (banks, insurance companies, and other financial institutions) because fair value assets are more prevalent and important there. According to data obtained from the Standard & Poor’s Compustat database, the proportion of fair value assets to total assets of exchange-listed firms grew from 18.8% in 2008 to 20.3% in 2013. This upward trend is led by firms in the financial industry, where proportions grew from 25.8% to 29.3%, versus 16.3% to 17.9% in nonfinancial firms.

Furthermore, level 2 and level 3 assets represent a higher percentage (18.6% and 2.9%, respectively) of total assets in the financial industry compared with the nonfinancial industry (4.8% and 1.2%, respectively). In addition, given the highly leveraged situation of financial institutions, the size of level 3 assets relative to their common equity is quite large. This ratio is 10.5% for financial companies, which implies that the gains and losses from fair value changes in level 3 assets (hereafter, “level 3 gains”) are likely to have a significant effect on net earnings. The level 3 gains are not available in the Standard & Poor’s Compustat database; therefore, the annual fair value changes in the level 3 assets have been used as a proxy for the level 3 gains to compute the average absolute value of the ratio of level 3 gains to net income. For financial businesses, the average absolute value of the ratio is 63.2%, which is surprisingly high; thus, the effect of level 3 gains on net income in the financial industry is nontrivial, and the concern about level 3 fair values appears justified.

Pricing of Level 3 Assets and Market Implications

Chang Joon Song, Wayne B. Thomas, and Han Yi pioneered research on the three-level hierarchy after the adoption of SFAS 157 (“Value Relevance of FAS 157 Fair Value Hierarchy Information and the Impact of Corporate Governance Mechanisms,” Accounting Review, vol. 85, no. 4, pp. 1375–1410, 2010). Using the financial statements of 431 banks from 2008, they examined how stock market participants priced level 1, 2, and 3 assets. Their analysis presents evidence that the stock market values each dollar of level 1, 2, and 3 assets at $0.98, $0.97, and $0.68, respectively. The drop in valuation of level 3 assets indicates that investors are concerned about the reliability of management’s estimates of their fair values. This concern must be justified, given the well-known cases of fraudulent earnings management (e.g., Enron or WorldCom). Businesses could easily overstate their level 3 assets and recognize the gains from fair value changes in those assets whenever necessary to paint up decent earnings numbers.

A more conservative valuation approach is warranted for level 3 estimates to protect investors and financial analysts from the consequences of errors.

Using financial data from 467 financial institutions, Edward J. Riedl and George Serafeim examined the effect of level 3 assets on a company’s cost of equity capital (“Information Risk and Fair Values: An Examination of Equity Betas,” Journal of Accounting Research, vol. 49, no. 4, pp. 1083–1122, 2011). They hypothesized that, given management’s discretion to estimate the value of level 3 assets and the incentives to overstate earnings, market participants might suspect management of overestimating future cash flows to value those assets. Thus, market participants might be likely to discount such valuation, which is ultimately reflected in stock prices. Riedl and Serafeim found evidence supporting this notion; specifically, that companies with higher exposure to level 3 assets have a higher cost of equity capital.

Michael Magnan, Andrea Menini, and Antonio Parbonetti examined the association between the amount of a company’s fair-valued assets and the properties of earnings forecasts made by financial analysts working at brokerage houses (“Fair Value Accounting: Information or Confusion for Financial Markets?” Review of Accounting Studies, vol. 20, no. 1, pp. 559–591, 2015). Specifically, they investigated analysts’ earnings forecast errors and dispersions for companies with a large proportion of fair-valued assets to total assets and found that both the errors and dispersions were higher for those firms. In short, a high proportion of fair-valued assets in a financial statement creates an “information bottleneck” that prevents analysts from obtaining the information they need to make reliable earnings forecasts. Therefore, fair value accounting does not necessarily lead to a better information environment.

In a related study using financial data from 120 European banks, Emanuel Bagna, Giuseppe Di Martino, and Davide Rossi investigated the stock market discount related to holding level 3 assets in the European markets (“An Anatomy of the Level 3 Fair Value Hierarchy Discount,” working paper, 2014, https://ideas.repec.org/p/pav/demw-pp/demwp0065.html). They suggested three reasons for such a discount: 1) the lack of disclosure, specifically regarding how management makes level 3 fair value estimates; 2) the possible use of level 3 valuations for “earnings management,” as suggested by Song, Thomas, and Yi above; and 3) the lack of liquidity. This liquidity concern relates to the absence of an active market for level 3 assets; as a result, those assets remain illiquid and cannot be readily converted into cash. Therefore, companies with a large amount of level 3 assets are riskier than others, resulting in a higher price discount in the stock market.

The Role of Fair Value Disclosure

Some business voluntarily disclose the processes and controls they use to make fair value estimates of level 3 assets. Song Gun Chung, Beng Wee Goh, Jeffrey Ng, and Kevin Ow Yong studied the determinants of such voluntary disclosure (or reliability disclosure) and its effect on the stock market (“Voluntary Fair Value Disclosures Beyond FAS 157’s Three-level Estimates,” working paper, 2014, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1335848). Out of 692 financial companies, approximately 20% provided such reliability disclosure. Some stated that they employed independent institutions to measure the fair values, whereas others asserted that correctly estimating the fair value of assets and liabilities is management’s ultimate responsibility.

Disclosures on assumptions and models adopted during estimation, which are currently unavailable, could help mitigate reliability concern for level 2 estimates.

The study first examined why some companies provided reliability disclosure, finding that those with a large amount of level 3 assets are more likely to provide such disclosure. Given prior studies’ findings regarding the heavy stock market discounts for companies with a large proportion of level 3 assets, this suggests an attempt to acknowledge and alleviate investors’ concerns. Second, the study investigated how the provision of reliability disclosure affects stock prices and found that the price discount for holding level 3 assets diminishes in entities that make a reliability disclosure. This suggests that management’s assurance about the integrity of the process and controls in place for making level 3 fair value estimates is effective in mitigating investors’ concerns about the reliability of those estimates.

Additionally, Bagna, Martino, and Rossi found evidence that the stock price discount related to level 3 assets is concentrated in companies that do not provide required disclosures about the fair value estimation process under International Financial Reporting Standards (IFRS) 7, Financial Instruments: Disclosures. Therefore, they contended that strong enforcement of fair value disclosure could mitigate the effect of level 3 assets on stock prices.

External Information Environment and Corporate Governance

Riedl and Serafeim also investigated whether the external information environment surrounding a business can decrease the measurement error problem associated with level 3 assets. Larger companies have a greater following among analysts and are covered more intensely by the media. In such cases, management is likely to be more careful when evaluating level 3 assets. Riedl and Serafeim found evidence in support of this premise, noting that the measurement error problem in valuing level 3 assets is less prominent in large companies.

Song, Thomas, and Yi further evaluated the effect of corporate governance on the pricing of level 3 fair value assets. If the share price discount from holding level 3 assets truly reflects investors’ concerns about the reliability of level 3 fair values, the discount might be less in a company with a more rigid monitoring system over the financial statement. Song, Thomas, and Yi constructed a measure of firms’ corporate governance systems using 1) audit committee financial expertise, 2) the frequency of audit committee meetings, and 3) the size of the audit engagement office. They found that the stock price discount resulting from holding level 3 assets is mitigated for companies with higher audit committee effectiveness and auditor engagement effort.

However, lowering the stock price discount by engaging high quality auditors appears to have some costs. Michael Ettredge, Yang Xu, and Han Yi studied the association between the proportion of level 3 assets and the amount of audit fees (“Fair Value Measurements and Audit Fees: Evidence from the Banking Industry,” Auditing: A Journal of Practice & Theory, vol. 33, no. 3, pp. 33–58, 2014). They found that audit fees are higher for companies with a greater proportion of fair value–based assets, and that these higher fees are mainly driven by the presence of a large amount of level 3 assets. This implies that fair value accounting creates a more difficult task for auditors, especially when companies have a large proportion of level 3 assets.


An analysis of the existing research leads to three primary conclusions relevant for investors, auditors, and regulators. First, level 3 assets, whose fair values are subjectively determined by management, hurt companies’ market values in the form of larger share price discounts. These discounts seem to be driven by investors’ skepticism about the reliability of management’s estimates. Anecdotal evidence further supports such skepticism. Radian Group Inc. reported a net profit of $195 million in the first quarter of 2008, primarily driven by gains from level 3 liabilities. Without the fair value gains, Radian would have reported a loss of about $215 million. Both the findings of recent accounting research and the example of Radian suggest that a more conservative valuation approach is warranted for level 3 estimates to protect investors and financial analysts from the consequences of errors.

Second, prior studies highlight the importance of disclosure for level 3 fair value estimates. More complete disclosure of level 3 fair value estimation procedures can alleviate investors’ concerns, mitigate the discount associated with level 3 estimates, and help the capital market assess the economic value of level 3 estimates more accurately. Furthermore, greater disclosure about the level 2 fair value estimation process will make financial statements even more informative. Companies possess considerable amounts of level 2 assets, and despite the relatively less subjective estimation process, the measurement of level 2 fair values still relies on managerial discretion. For example, in estimating level 2 fair values, companies can use market inputs such as yield curve or empirical correlation, but the fair value still depends on which model the firm selects. In support of this point, Song, Thomas, and Yi, as well as Riedl and Serafeim, found evidence that level 2 assets contribute less to value than level 1 assets. Disclosures on assumptions and models adopted during estimation, which are currently unavailable, could help mitigate reliability concern for level 2 estimates and reduce the associated stock market discount.

Finally, past research (e.g., Song, Thomas, and Yi; Ettredge, Xu, and Yi) suggests that auditors play a crucial role in assuring the reliability of subjective fair value estimates. Outside stakeholders have very limited information about the managerial projection of fair value estimates, and regulators and investors expect auditors to mitigate the degree of uncertainty associated with “unobservable inputs” for level 2 and 3 fair value estimates. Recent reports from the PCAOB, however, indicate that auditing fair value estimates is one of the most challenging audit areas. For example, according to a PCAOB inspection report in 2012, about 25% of all identified audit deficiencies are subject to fair value measurement problems. The report strongly recommends that auditors enhance scrutiny of these subjective fair value measurements.

To better assess the economic value of subjective fair value measurements, accounting professionals must apply more professional skepticism.

Overall, the expansion of fair value accounting necessitates a new system of U.S. financial reporting standards, including expanded regulatory guidelines for fair value disclosure and estimates. To better assess the economic value of subjective fair value measurements, accounting professionals must apply more professional skepticism, and regulators must increase the quality and quantity of discourse of fair value measurement.

Sung Gon Chung, PhD is an assistant professor of accounting, at the Mike Ilitch School of Business, Wayne State University, Detroit, Mich.
Cheol Lee, PhD is an associate professor of accounting, at the Mike Ilitch School of Business, Wayne State University, Detroit, Mich.
Santanu Mitra, PhD is an associate professor of accounting, at the Mike Ilitch School of Business, Wayne State University, Detroit, Mich.