Fair Value Accounting: Evidence from Investment Securities and the Market Valuation of Banks

This study investigates how disclosed fair value estimates of banks’ investment securities and securities gains and losses based on those estimates are reflected in share prices in comparison with historical costs. Fair value securities gains and losses are calculable because banks also disclose realized securities gains and losses. Thus, banks’ investment securities provide an opportunity to examine two measurement methods, historical cost and fair value, for both an asset and its related earnings component. Previous research does not provide strong evidence on value-relevance of asset fair value estimates. Errors in estimating the fair values is the primary explanation for this unexpected finding. Another explanation relates to cross-sectional differences in sample firms, e.g., industry membership. This study examines disclosed fair values of investment securities that can be considered more reliable than previously-studied fair value disclosures. Moreover, the sample firms here belong to one industry, banking. This study also investigates the Barth et al. (1990) suggestion that fair value securities gains and losses are value-relevant. By examining how share prices reflect historical costs and fair values, evidence is provided on the measures’ relevance and reliability. Because these are the FASB’s two principal criteria for choosing among accounting alternatives [Statement of Financial Accounting Concepts (SFAC) No. 2, FASB 1980], the evidence can inform the FASB’s deliberations on using fair value accounting for investment securities, to the extent the disclosed fair value estimates would be used to measure investment securities under fair value accounting. Share prices for a sample of banks are explained using investment securities historical costs and fair value estimates together with the book value of equity before investment securities. Similarly, bank stock returns are explained using securities gains and losses based on historical costs and on fair values together with earnings before securities gains and losses. The analyses provide evidence on the two methods’ incremental and relative explanatory power, and relative measurement error. The findings indicate that fair value estimates of investment securities provide significant explanatory power beyond that provided by historical costs. Strikingly, historical costs provide no significant explanatory power incremental to fair values. Using a measurement error model, investment securities’ fair values are found to have less measurement error than historical costs vis-a-vis the amount reflected in share prices. The findings for securities gains and losses are different. The significance of any incremental explanatory power for fair values beyond historical costs depends on the specification of the estimating equation. In some specifications, fair value securities gains and losses have no significant incremental explanatory power, but historical costs always provide explanatory power incremental to fair values. The findings based on a measurement error model indicate that fair value securities gains and losses also have more measurement error than historical costs. Thus, although fair value estimates of investment securities appear reliable and relevant to investors in valuing bank equity, fair value securities gains and losses do not. One interpretation for these findings is that although estimation error in the disclosed fair values is small enough that investment securities’ fair values appear value-relevant, when two annual fair value estimates are used to calculate securities gains and losses, the effect of the combined estimation errors renders securities gains and losses value-irrelevant. Another plausible interpretation is that securities gains and losses might be offset by unrecognized correlated gains and losses on other assets and liabilities. Why this affects securities gains and losses but not investment securities is an unresolved question. Evidence from supplemental analyses gives more credence to the first interpretation than to the second.