Fair Values, Comprehensive Income Reporting, and Bank Analysts’ Risk and Valuation Judgments

ABSTRACT

We investigate whether the measurement and reporting of comprehensive income in financial statements systematically affects commercial bank equity analysts’ investment risk assessments and valuation judgments. In an experiment in which 80 buy side analysts specializing in banking and financial institutions participated, we vary whether a bank is exposed to or hedged against interest rate risk across three different comprehensive income accounting regimes-piecemeal fair value accounting with comprehensive income reported in the statement of changes in equity, piecemeal fair value accounting with comprehensive income reported in a separate statement of performance, and full fair value accounting with comprehensive income reported in a separate statement of performance. When fair value changes are measured and reported on a piecemeal basis, we find no investment risk or valuation judgment differences across CI reporting regimes. Although we find that the analysts’ investment risk judgments are influenced by the banks’ risk management strategies, their valuation judgments are not. Only when fair value changes are measured completely and reported transparently (i.e., full fair value accounting) do analysts’ valuation judgments distinguish between banks with different levels of risk. The study contributes to three avenues of literature concerned with financial-reporting transparency, risk, and fair value accounting. First, our ex ante evidence informs accounting standard setters as they evaluate whether to move to full fair value accounting for all financial instruments, and assess the degree to which differences in comprehensive income measurement and reporting affect the information analysts obtain from financial statements. Second, the study adds to the experimental accounting literature that examines comprehensive income reporting format effects and valuation judgments. Third, we add to archival accounting research on the information in comprehensive income by testing market participants’ judgments under different comprehensive income reporting regimes, which cannot be directly investigated with archival data.

KEY WORDS: Fair value, comprehensive income, financial analysts, behavioral finance DATA AVAILABILITY: Contact the authors

Fair Values, Comprehensive Income Reporting, and Bank Analysts’ Risk and Valuation Judgments

I. INTRODUCTION

Statement of Financial Accounting Standards (SFAS) No. 130-Reporting Comprehensive Income (FASB 1997) requires firms to report comprehensive income in one of the primary financial statements. Although SFAS 130 did not change accounting measurement or recognition-it simply requires reporting all non-owner-related changes in equity in one place-the standard was controversial. Banks argued that piecemeal fair value recognition (i.e., under current rules, only fair value gains and losses on investment securities classified as available-for-sale are recognized) would result in biased measures of comprehensive income that misrepresent the underlying economics of banks that hedge their interest rate risk exposure by balancing their net financial asset and liability positions. Banks asserted that market participants might place undue focus on the distorted measure of comprehensive income, perceive excessive volatility in bank income, and misprice bank shares. One potential solution to this problem is more transparent and complete measurement and reporting of the fair value of financial instruments.

In this study we investigate whether differences in how commercial banks measure and report comprehensive income cause systematic differences in bank analysts’ investment risk assessments and valuation judgments. We conduct an experiment in which we vary whether a bank is exposed to or hedged against interest rate risk. In addition, we vary the bank’s comprehensive income accounting regime in three ways: (1) fair value gains and losses on available-for-sale investment securities are recognized in the statement of changes in equity, with fair value gains and losses on all other financial assets and liabilities not recognized in the financial statements but disclosed in the notes (consistent with banks’ prevailing piecemeal fair value reporting under SFAS 107 (FASB 1991) and SFAS 115 (FASB 1993))

of comprehensive income (analogous to the reporting standard currently being considered by the FASB and international standard setters).

We find that analysts assess significantly higher risk and lower share values for high-risk banks than for low-risk banks when fair value gains and losses on all financial assets and liabilities are recognized in a statement of comprehensive income. We find that under piecemeal recognition of fair value gains and losses, reporting format does not uniformly influence analysts’ risk and value judgments. Specifically, in the piecemeal fair-value conditions, we obtain evidence that bank analysts are cognizant of the banks’ risk management strategies and distinguish between them in terms of investment risk. However, in the absence of clear and complete reporting of the performance-relevant implications of those strategies, analysts’ valuation judgments do not make such a distinction. Taken together, these results suggest that although bank analysts are aware of the different interest rate risks assumed by high-risk and low-risk banks, clear performance reporting of the financial implications of these risks increases the likelihood that analysts will use the information in their investment-related judgments.

The study contributes to three areas of research related to comprehensive income, risk, and fair value accounting. First, the FASB is currently considering a proposal that would require all financial instruments be reported at fair value, with fair value gains and losses recognized in net income. We provide ex ante evidence that sheds light on the degree to which differences in financial performance measurement and reporting format affect the information analysts’ obtain from financial statements. Consistent with Ryan’s (1997) call for expanded fair value reporting, we demonstrate circumstances where analysts are more likely to incorporate fair value information into their judgments.

Second, we provide new evidence relevant to the FASB’s continuing deliberation of comprehensive income measurement and reporting (FASB 1997, par. 54). In particular, Hirst and Hopkins (1998) find that comprehensive income reporting formats influence the valuation judgments of analysts and conclude that there are benefits associated with transparent reporting of comprehensive income in a performance statement. Lipe (1998) notes several features of the Hirst and Hopkins (1998) design that limit the inferences that can be drawn from their study. In particular, she observes that the study relied on non-

specialist analysts who might not be familiar with the relevance of marketable securities gains and losses in a setting in which marketable securities are not a core component of operations. Furthermore, comprehensive income data were newly mandated and, therefore, possibly unfamiliar to analysts. Each of these concerns is addressed in the current study. Our findings are consistent with Hirst and Hopkins (1998) in that we find that clear and complete reporting improves analysts’ judgments.

In addition, this study also provides evidence on the ways in which economic context and disclosure regime interact to determine disclosure effectiveness. Contrary to the argument put forth by bankers in comment letters on the comprehensive income exposure draft, analysts do not penalize hedged banks when they asymmetrically highlight only the loss components of their hedged balance sheets under piecemeal fair-value measurement. As such, we sharpen some of the Hirst and Hopkins (1998) findings. Using a context where elements of other comprehensive income are part of core income, we find that bank analysts’ judgments are not differentially affected by where piecemeal CI is reported. However, analysts’ judgment performance improves when full fair value measurement is used. That is, analysts better incorporate into their judgments the interest rate risk of a bank when full fair value reporting is used.

Finally, we contribute to existing archival research evidence on the value relevance of fair value information. Specifically, we provide evidence on competing explanations for the puzzling findings in Barth, Landsman and Wahlen (1995), which suggest that incremental volatility in fair value net income relative to historical cost net income of banks is not priced as incremental risk. The Barth, et al., results could be explained if: a) banks hedge interest rate risk, or b) investors do not use fair value gains and losses disclosed in footnotes when assessing banks’ income volatility and risk

competing explanations in previous archival research studies, collect data on intermediate judgments, and provide ex ante evidence about a full-fair-value reporting regime that is not yet in place.

The remainder of the paper is organized as follows. Section two provides background and describes our hypotheses. We describe our experiment in section three. In section four we describe the results. Section five draws some conclusions and offers some implications.

II. BACKGROUND AND HYPOTHESES

In this section we link the study to the authoritative literature on accounting for comprehensive income, provide a description of current comprehensive income reporting practices among banks, relate our study to prior archival and experimental financial reporting research, and motivate and propose our hypotheses.

Comprehensive Income Reporting

Comprehensive income reflects reported net income adjusted for changes in other comprehensive income items, including unrealized gains and losses on available-for-sale investment securities, derivatives qualifying as cash-flow hedges, foreign currency translation adjustments, and minimum pension liability adjustments. SFAS 130 requires companies to report comprehensive income and its components in one of the primary financial statements (FASB 1997, par. 22). Although SFAS 130 did not change recognition or measurement of the components of comprehensive income, companies were almost uniformly opposed to the standard’s issuance.1

Most of the banks responding to the Comprehensive Income Exposure Draft argued that separately reporting comprehensive income and its components in a statement of performance would be misleading because not all financial instruments are recognized at fair values. In particular, under SFAS 115- Accounting for Certain Investments in Debt and Equity Securities (FASB 1993), companies are required

1 For example, (Hirst, Hopkins, and Yen 2001) report that during the public comment period for the Comprehensive Income Exposure Draft (ED) (FASB 1997), 76 percent of the letters received by the FASB were unilaterally opposed to the proposed standard. Commercial banks were particularly vocal opponents of the proposed standard, writing 82 percent of the letters received from firms in the financial services sector and 33 percent of all letters received by the FASB on this ED. Only 15 out of 278 (five percent) of the letters were unilaterally in favor of the CI-reporting rules proposed in the Exposure Draft, but letter writers who were partially in favor of the Exposure Draft typically supported the need for increased disclosure.

to recognize at fair value only investment securities that are classified as available-for-sale. In contrast, all other financial assets and liabilities are recognized at amortized historical cost (with fair values disclosed in the notes, according to SFAS 107-Disclosures about Fair Value of Financial Instruments (FASB 1991)).

Commercial banks claimed that reporting piecemeal-fair-value-based comprehensive income and its components in statement of performance would misrepresent the economic risk and performance of financia l institutions. For example, banks that hedge interest rate risk by matching the rates and maturities of their financial assets and liabilities would report comprehensive income that only includes unrealized gains and losses on available for sale marketable securities because the unrealized gains and losses on all other interest-rate hedged financial assets (e.g., loans) and financial liabilities (e.g., deposits) are currently excluded from both net and comprehensive income. Therefore, many banks suggested that highlighting only a piece of the interest-rate hedge would cause financial-statement users to perceive artificial volatility in hedged banks’ performance and to be misled about their interest rate risk management strategy.2

In response to dissatisfaction with the piecemeal recognition of fair values, the FASB is currently developing accounting standards that reflect full fair value accounting for all financial instruments (FASB 1999, par. 334). As part of this initiative, the FASB issued a Preliminary Views document that outlines many of the issues it is considering as it develops a proposal that could substantially expand the scope of fair value measurement and recognition of financial instruments (FASB 1999). Consistent with the recommendation of the Joint Working Group of Standard Setters’ (2000), the Board has indicated that its preference is to require recognition of fair value gains and losses in net income. Our study is designed to provide standard setters with ex ante evidence on the effects of such reporting.

2 Partly in response to these arguments, the final version of SFAS 130 allowed firms to report comprehensive income in any of the primary financial statements. Although the location of a disclosure may seem like a minor issue, two of the Board’s members dissented in the issuance of the standard on the grounds that allowing entities to disclose items of other CI in the statement of changes in equity “will do little to enhance their visibility and will diminish their perceived importance” (FASB 1997, p. 10). This assertion is consistent with the opinions of analysts in Brown (1997) who indicated that the information contained in the SCE is not very useful.

Descriptive Statistics of Banks’ Comprehensive Income Reporting Practices

To assess whether banks’ comprehensive income reporting behavior is consistent with the comments included in their comment letters to the comprehensive income Exposure Draft, we collected data from the 1999 financial statements of the 100 largest U.S. banks (based on total assets). Consistent with their objections to the exposure draft, the data reported in Panel A of Table 1 indicate that 80 of 100 largest U.S. banks report the effects of comprehensive income in the statement of changes in equity, rather than in a statement of performance. Further, all of the 25 largest banks reported comprehensive income in the statement of changes in equity.

INSERT TABLE 1 ABOUT HERE

We also gathered data on the presentation of comprehensive income and the components typically reported by commercial banks. The data in Panel A of Table 1 indicate that nearly all of the top 100 banks in 1999 (92 of 100) elected to report the individua l components of comprehensive income in the body of the financial statements. Only eight of the top 100 banks reported an aggregate comprehensive income number, or an aggregate comprehensive income adjustment to net income, with disclosure of the separate components of comprehensive income in the notes to the financial statements. As expected, available for sale securities were the most common component of other comprehensive income (99 of 100), with a minority of banks also reporting foreign currency translation adjustments (16 of 100).

To assess the magnitude of the effects of comprehensive income relative to net income for banks, we collected additional data on our sample from SNL DataSource (see Panels B and C of Table 1). For the years 1997 to 1999, Panel B indicates that comprehensive income typically differs substantially from net income. The median ratio of comprehensive income to net income among these banks ranges from a low of 70.6% in 1999 to a high of 114.7% in 1997. The data in Panel C suggest that fair value gains and losses on investment securities account for most of the difference between comprehensive income and net

income. For the median bank, fair value gains and losses amounted to -32.3% of net income in 1999, and 6.4% of net income in 1997. 3

Fair Value Income Volatility and Commercial Bank Stock Prices

Banks’ concerns about the potentially adverse interactive effects of comprehensive income reporting and piecemeal fair value recognition are moot if (1) fair value income is not more volatile than reported net income and (2) fair-value-income volatility is not incrementally relevant to investors or other stakeholders. Evidence reported in Barth, Landsman, and Wahlen (BLW, 1995) support the pre-SFAS 115 assertions made by bank managers that earnings including fair value gains and losses from investment securities are more volatile than reported earnings. For the period 1971-1990, BLW find that the standard deviation of fair value-based income is 26 percent greater than reported net income for the median bank in their sample .4

BLW also investigate whether the incremental volatility in fair value-based income is reflected in bank share prices .5 A link between fair value-based income volatility and share prices should occur if two assumptions hold: (1) fair value-based income provides incremental information about systematic economic risk beyond that in historical cost income, and/or (2) investors do not distinguish between recognized and disclosed amounts when estimating economic earnings (BLW, p. 583). BLW find that volatility in fair value-based income is not associated with a lower earnings multiple implied by securities prices and conclude that “the volatility in fair value net income incremental to that in historical cost net income is not priced as risk” (p. 588).

3 The ratio of comprehensive income to net income is not significantly different across banks that report CI in a performance statement or the statement of changes in equity in 1997 (p = .52) or 1999 (p = .34). In 1998, the ratio was lower (.99) for banks that disclosed CI in a performance statement than those that did not (1.10). This difference is significant (p = .05) but inconsistent with a desire to highlight better performance. Based on these data, we conclude that there is no reliable signaling underlying the choice of CI format.

4 BLW do not directly compare time-series volatility in comprehensive income and net income, per se, because their sample predated mandated fair value recognition of available-for-sale securities and reporting of comprehensive income and its components.

5 Earnings variability should be reflected in share prices via a risk premium (e.g., Collins and Kothari 1989

Two alternative explanations may account for the observed lack of incremental association between volatility in fair value net income and security prices. First, BLW’s sample was based on fair values of investment securities disclosed in regulatory filings, but preceded mandated financial statement recognition of the fair values of investment securities (SFAS 115) and comprehensive income and its components (SFAS 130). Therefore, investors may have overlooked the fair value information because it was not included in the general-purpose financial statements. Second, BLW’s analysis only includes fair value gains and losses for investment securities. To the extent banks hedge the risks underlying their marketable securities holdings and bank share prices reflect these unrecognized hedges, then BLW’s proxy for fair value income (i.e., a form of piecemeal fair value measurement) does not reflect the underlying volatility of bank’s “true” economic income. Our experiment is designed, in part, to investigate the assumptions underlying BLW’s analysis and to provide evidence on the two alternative explanations.

Comprehensive Income Reporting Format and Investors’ Judgments

Hirst and Hopkins (1998) report that comprehensive income reporting format systematically affects buy-side analysts’ stock price judgments. They find that explicitly reporting unrealized gains and losses on marketable securities in comprehensive income increases the likelihood that analysts adjust their valuation judgments for marketable-securities-related earnings management. In addition, they find that analysts adjust their valuation judgments for this type of earnings management only when comprehensive income is reported in a statement of performance, and do not completely eliminate the effects when it is reported in the statement of changes in equity. Hirst and Hopkins (1998) attribute their findings to the greater degree of transparency for comprehensive income reported in a statement of performance and conclude that such transparency enhances analysts’ judgments.

As noted by Lipe (1998), the research design in Hirst and Hopkins (1998) includes features that limit the study’s generalizability. In particular, the study was conducted before SFAS 130 was mandated, perhaps causing analysts to unnaturally focus on comprehensive income when it was clearly reported. In addition, the experiment involved an electronics manufacturer (i.e., marketable securities are not related

to core operations), so analysts may have discounted the possibility of marketable-securities-based earnings management in the conditions in which comprehensive income was not clearly reported (i.e., in the statement of changes in equity).

Maines and McDaniel (2000) address one of these limitations by testing whether reporting format affects nonprofessional investors’ use of comprehensive income information in their performance assessments of an insurance company. In their experiment, incremental volatility of comprehensive income was manipulated by varying the unrealized gains and losses on marketable securities, a core asset for insurers. Their results suggest that nonprofessional investors evaluated the volatility of comprehensive income across all reporting format conditions. However, reporting format influenced the relative weight of comprehensive income volatility in their performance assessments only when comprehensive income and its components were reported in a separate performance statement. Although Maines and McDaniel (2000) do not examine the judgments of more sophisticated market participants, they propose that comprehensive income reporting format will have little or no effect when industry specialists evaluate the performance of companies for which comprehensive income information is related to core assets (e.g., commercial banks’ fair value gains and losses on investment securities) because analysts understand the importance of such information for assessing the performance of financial institutions and will search for this information (Hunton and McEwen (1997).

Hypotheses

Our hypotheses focus on two aspects of financial-reporting transparency for unrealized fair value gains and losses on interest-rate sensitive net assets: format and measurement. We first propose hypotheses related to comprehensive income reporting format (i.e., performance statement versus the statement of changes in equity) in the current piecemeal measurement environment. This is the setting in which Lipe (1998) and Maines and McDaniel (2000) made their format predictions for financial analysts and the setting in which bankers made their claims in opposition to performance-statement reporting of comprehensive income. Second, we propose hypotheses related to comprehensive income measurement (i.e., full fair value versus piecemeal recognition of unrealized gains and losses) to provide evidence on

the extent to which analysts process the recognized and disclosed aspects of fair value information in today’s piecemeal reporting environment and ex ante evidence on the effects of full fair value accounting.

We make our predictions in a setting in which bank analysts provide investment risk assessments and stock price judgments for a commercial bank. The analysts possess equivalent sets of fair value information on which to base their judgments, but the format and measurement of comprehensive income are independently varied. To allow for directional predictions in this setting, we assume that interest rates rise just before year-end, thereby triggering fair value losses (gains) on fixed rate financial assets (liabilities), which may be recognized in comprehensive income, depending on how it is measured. A critical feature of this setting is that banks’ reported net income for the year is not affected by the year¬end rate change because unrealized fair value gains and losses are not included in net income.

We also vary whether the bank has elected to take on a high degree of interest-rate risk (i.e., exposed) or a low degree of interest rate risk (i.e., hedged). The high-risk bank’s fixed rate assets have longer maturities than its fixed rate liabilities (banks commonly lend or invest at longer maturities and borrow at shorter maturities.) Because assets and liabilities do not hedge each other, the bank is exposed to a higher degree of interest rate risk. In a rising rate environment, the high-risk bank will experience relatively large net fair value losses on its financial assets and liabilities (i.e., large losses on its fixed rate assets, but only modest fair value gains on its fixed rate liabilities).6 The low-risk bank has identical financial assets as the high-risk bank

6 We limit our investigation to the risk associated with an adverse change in interest rates (i.e., resulting in net financial losses). Although normative models define risk as variability across the domains of gains and losses, research in psychology and applied fields suggest that people primarily perceive risk as the potential for loss. For example, in medicine, risk is commonly perceived as the possibility of some adverse outcome (Kleinbaum et al., 1982). Even in financial contexts, professional analysts and bond traders perceive risk as being the potential for monetary loss (Heisler 1994, Olsen 1997). See Hodder, Koonce and McAnally (2001) for a general discussion of risk perception and evidence that risk perceptions are not symmetric across the gain and loss domains.

Comprehensive Income Format

Hirst and Hopkins (1998) and Maines and McDaniel (2000) investigate the effects of two comprehensive income reporting formats in the current piecemeal fair value reporting environment: (1) comprehensive income reported in the statement of changes in equity (denoted “PFV-SCE”-i.e., piecemeal fair value accounting in the statement of changes in equity) and (2) comprehensive income reported in a separate statement of performance (denoted “PFV -IS”-i.e., piecemeal fair value accounting in a separate comprehensive income statement). Hirst and Hopkins (1998) estimate that approximately 50 percent of analysts receiving PFV-SCE financial statements in their sample did not process the comprehensive income information. This appears to have caused fewer analysts to incorporate value¬relevant marketable-securities information into their stock price judgments in the PFV-SCE conditions, resulting in stock-price judgments that were systematically different from the highly transparent PFV -IS conditions.

Lipe (1998) and Maines and McDaniel (2000) predict that these two comprehensive-income-reporting formats should not differentially affect bank analysts’ judgments because fair value and interest-rate-risk information is related to banks’ core operations. Therefore, bank analysts should use directed search (c.f., Hunton and McEwen 1997) to locate the value-relevant fair-value information, regardless of its location in the financial statements, notes, or management’s discussion and analysis (MD&A). As such, analysts’ risk and stock-price judgments should distinguish between the high and low-risk banks, regardless of comprehensive income reporting format (i.e., PFV-SCE vs. PFV-IS). We propose the following hypothesesæillustrated in the left half of Panels A and B of Figure 1æto jointly test the comprehensive¬income-format predictions of Lipe (1998) and Maines and McDaniel (2000) in piecemeal fair-value reporting environment.

H1: Under piecemeal recognition of fair value gains and losses, bank analysts’ investment risk and stock valuation judgments will not be differentially influenced by whether comprehensive income is reported in the statement of changes in equity or a separate a statement of comprehensive income.

H2: Under piecemeal recognition of fair value gains and losses, bank analysts’ investment risk (stock valuation judgments) for the interest-rate-risk exposed bank will be higher (lower), regardless of

whether the bank reports comprehensive income in the statement of changes in equity or a separate a statement of comprehensive income.

INSERT FIGURE 1 ABOUT HERE

Comprehensive Income Measurement

Banks opposed the FASB’s Comprehensive Income Exposure Draft (FASB 1997), in part, because of the mixture of piecemeal fair value accounting and the proposed requirement that comprehensive income and its components be reported in a separate statement of performance. Bank managers asserted that a piecemeal approach to fair value recognition that is highlighted in a separate statement of comprehensive income would misrepresent the interest rate risk and income volatility of commercial banks and would mislead investors.

A weakness in the bank managers’ claims is that banks convey interest-rate risk information via “gap tables” included in MD&A and that SFAS 107 requires explicit disclosure of fair value information for financial assets and liabilities.7 Lipe (1998) and Maines and McDaniel (2000) suggest that because bank analysts are industry experts, they should be able to use all available information to assess interest rate risk and share value. Therefore, they would predict that bank analysts’ risk and value judgments under piecemeal fair value reporting should fully reflect all of the fair value information embodied in the notes and MD&A.

A potentially significant information-processing problem that could affect the predictions offered by Lipe (1998) and Maines and McDaniel (2000) is that gap tables and SFAS 107 disclosures provide interest-rate-risk-related information at single points in time and require financial statement users to effortfully simulate the possible effects of unrealized gains and losses on banks’ current and future operating performance. Because analysts do not have unlimited time and cognitive resources, they may place greater weight on items that are transparently associated with a bank’s performance via an income¬type measure and may not fully process the static information disclosed in MD&A and footnotes. Therefore, analysts may not incorporate all relevant fair value information when comprehensive income

7 Gap tables provide a static measure of interest-rate exposure by summarizing net differences between interest rate sensitive assets and liabilities maturing at different times.

includes only piecemeal recognition of fair value gains and losses. This position is consistent with the findings of Hirst, Jackson, and Koonce (2001) who find that supplemental disclosures revealing prior financial-statement-estimate accuracy levels are more useful when they are transparent and complete in their implications.

We propose that fair-value gains and losses are most transparent and are more likely to be considered by analysts when all financial assets and liabilities are recognized at fair value and the unrealized gains and losses are explicitly recognized in a separate statement of performance (denoted “FFV-IS”-i.e., full fair value accounting in a separate comprehensive income statement). However, the effects of this transparency are conditioned on whether the bank is hedged or exposed. In the case of a fully interest¬rate-risk-hedged bank, full fair value reporting should not systematically influence analysts’ investment risk or price judgments because there is little incremental performance information in the comprehensive income information. If analysts believe managers’ assertions that they are fully hedged (via the gap table and/or SFAS 107 disclosures), then reported operating income provides sufficient information to assess the bank’s economic performance

However, when the bank is exposed to interest rate risk because of a mismatch between the maturities of assets and liabilities, comprehensive income provides incremental information that can help analysts understand the effects of interest rate changes on the bank’s operating performance. Therefore, relative to analysts who receive comprehensive income measured under PFV-IS, analysts who receive FFV-IS financial statements should more clearly and completely see the financial-performance-relevant effects of the bank’s exposure to interest-rate risk. Relative to the PFV-IS for the exposed bank, analysts should provide higher investment risk and lower stock price judgments in the case of an exposed bank reporting FFV-IS. The resulting interactionæproposed in the following hypothesisæis depicted in the right half of Panels A and B of Figure 1.

H3: Compared to piecemeal recognition of fair value gains and losses in a separate statement of comprehensive income, full recognition of fair value gains and losses in a separate statement of comprehensive income will lead analysts to provide higher investment risk and lower stock price judgments when the bank is exposed to interest rate risk, but will not differ when the bank is fully hedged with respect to interest rate risk.

Taken together, these hypotheses suggest that the two dimensions of the transparency of the interest¬rate risk faced by banksæreporting format (i.e., IS versus SCE) and measurement (i.e., FFV versus PFV) of comprehensive incomeæwill influence bank analysts’ risk assessments and stock price judgments differently depending on the extent to which the bank is hedged or exposed.

III. EXPERIMENT

To investigate the effects of comprehensive income measurement and reporting format on equity analysts’ risk and stock price judgments, we designed and conducted a 3 X 2 between-subjects experiment. Participants were 80 buy-side equity-security analysts and portfolio managers.8 All participants were recruited individually from the 2000 Association for Investment Management and Research Membership Directory (AIMR 2000) on the basis of their self-reported industry specialization (banking) and job descriptions. After securing their agreement to participate, the materials were distributed via overnight mail.

On average, the study participants have 12 years of experience as financial analysts (73% are CFAs). They spend an average of 51% of their time on equity-security analysis and another 34% on portfolio management. On average, they manage an $876 million portfolio (median $100 million), which includes 52 companies (median 40). In addition to the companies in their portfolios, on average these analysts follow another 92 (median 40) companies. On average, their employers have $24 billion (median $4.3 billion) of assets under management.

8 We received responses from 89 individuals. Because of the criticisms of prior research designs, we decided (prior to data collection) that we would only include responses from individuals who actively performed equity security analysis. Nine individuals indicated that they spent zero percent of their time performing equity security analysis and were eliminated from the sample.

Procedure

We provided participants with background information about a bank, industry average price-earnings ratios (15x) and ranges (10-20x), a description of the interest rate environment, summary historical financial information, and a stylized press release (as disseminated by Bloomberg Financial Services) reporting the bank’s annual earnings.9 The press release also included the current year’s financial statements, a summary of significant accounting policies, and a summary of significant risks including relevant footnote and MD&A disclosures regarding liquidity risk, credit risk, interest rate risk and fair values. After reviewing these materials, participants were asked to assess the risk of the bank and to provide an estimate of the value of the bank’s common stock. Participants also were asked to provide a description of the manner in which they determined the stock value. Following these questions, analysts assessed various types of risks faced by the bank, responded to a series of questions about the financial information in the case, several manipulation check questions, and provided demographic information. Appendices A-D include detailed descriptions of the financial accounting information included in the materials.

Materials and Independent Variables

To create materials representative of a typical commercial bank, we first created a model of the prototypical bank’s financial statements based on the financial statements for the years 1997 to 1999 for 11 of the 100 largest U.S. banks (measured based on total assets) as of year end 1999. The largest sample bank was Bank One (5th largest overall) and the smallest was National City Bancorporation (100th largest overall). As a starting point, we created a composite balance sheet, income statement, and statement of owners’ equity using the average account balances over the sample of 33 bank-years.

We then simulated the prototypical bank’s financial statements over a six-year span, using a set of baseline assumptions about asset and liability growth rates, credit losses, dividend payouts, tax rates, and non-interest income and expense items. These assumptions were applied equally across all conditions.

9 To reduce noise in analysts’ judgments, we provided banking-industry price-earnings ratio data to indicate a reasonable range within which the company’s stock price might fall. The materials did not include any stock price data.

Analysts were presented with the bank’s financial statements for the last three years of the simulation together with summary footnote and MD&A disclosures as part of the press release described earlier.10

Interest Rate Risk

The first independent variable is the relative level of the bank’s interest rate risk. To trigger variation in interest rate risk, we varied the extent to which the bank matched maturities of interest rate sensitive assets (loans and investment securities) and liabilities (deposits, federal funds, and short-term and long¬term liabilities). In the low-risk (hedged) condition, the bank matched the maturities of interest rate sensitive assets and liabilities, lending and borrowing at fixed rates over 5 years. By matching maturities each year, the low-risk bank had relatively small interest rate risk “gaps” (i.e., a measure of exposure based on the net differences between interest rate sensitive assets and liabilities maturing at different times). Following an upward movement in rates, the low-risk bank experienced fair value losses on its interest earning assets, but roughly equivalent fair value gains on its interest bearing liabilities.

In the high-risk (exposed) condition, the bank did not match the maturities of interest rate sensitive assets and liabilities. The high-risk bank’s loans and investment securities earned fixed rates over 5 years (exactly equivalent to the low-risk bank), but it borrowed funds at fixed rates over one year maturities. By not matching maturities, the high-risk bank had relatively large interest rate risk “gaps” each year .11 Following an upward movement in rates, the high-risk bank experienced relatively large fair value losses on its interest earning assets, but only modest gains on its interest bearing liabilities.

In the early years of the simulation, we assumed that interest rates were steady, varying a few basis points up or down each year. This assumption enabled us to hold the effects of interest rate risk constant across both the high and low-risk banks during the first two years included in the instrument. However, we introduced an interest rate shock (i.e., a 50 basis point increase in the Fed Funds target rate) at the end of the final year of the simulation. This interest rate shock triggered fair value losses that totaled roughly –

10 To simulate the footnote and MD&A disclosures in the instrument, we followed the relevant footnote and MD&A disclosures of these 11 banks in 1999.

$19.2 million on the interest rate sensitive assets for both the high- and low-risk bank. However, the shock also triggered fair value gains of $19.0 million on interest sensitive liabilities for the low-risk bank, thereby offsetting almost all of the fair value losses. On the other hand, the high-risk bank experienced fair value gains of only $7.1 million, leaving a net fair value loss of nearly -$12.1 million (before tax effects). This loss is roughly 40% of net income before tax and, thus, a significant figure.

In sum, we varied the interest rate risk of the bank by manipulating whether the bank hedged its interest rate sensitive assets and liabilities.12 Because interest rates were held steady until the end of the current year, prior year results were identical across the banks. The year-end interest rate increase had no effect on reported net income of the banks, but generated considerable differences in the fair value gains and losses on each bank’s financial instruments.

Comprehensive Income Measurement and Reporting Format

We vary the reporting of the effects of interest rate risk on comprehensive income three ways. In the PFV-SCE condition, fair value gains and losses on investment securities are recognized in the statement of owners’ equity, with fair value gains and losses on all other financial assets and liabilities disclosed in footnotes (consistent with prevailing practice under SFAS 107 and SFAS 115). In the PFV-IS condition, fair value gains and losses on investment securities are recognized in a separate statement of comprehensive income that follows the income statement, with fair value gains and losses on all other financial assets and liabilities disclosed in footnotes (consistent with a recommendation in SFAS 130). This condition is identical to the PFV-SCE condition except that comprehensive income and its components are reported in a performance statement. Finally, in the FFV-IS condition, fair values of all financial assets and liabilities are recognized on the balance sheet and fair value gains and losses on all the financial instruments are recognized in a separate statement of comprehensive income that follows the income statement (consistent with the new rule being considered by the FASB). This condition reports

11 For the high-risk bank, we repriced (i.e., applied the currently prevailing interest rates to) roughly 20% of the interest rate sensitive assets and 100% of the interest sensitive liabilities each year. By contrast, for the low risk bank, we repriced 20% of both the interest rate sensitive assets and liabilities each year.

comprehensive income in a performance statement (as in the PFV-IS condition) and recognizes fair value gains and losses for all financial instruments (unlike either of the PFV conditions).

INSERT FIGURE 2 ABOUT HERE

In all conditions, the financial information available to analysts is equivalent. Analysts’ seeking to adjust reported net income to reflect fair value gains and losses on any or all interest rate sensitive financial assets and liabilities can do so. Figure 2 demonstrates the equivalence of reported net income figures across all risk, measurement, and format conditions. Figure 2 also demonstrates how comprehensive income and full fair value income vary across risk, measurement, and format conditions.13

IV. RESULTS

Manipulation and Other Checks

After eliciting our dependent measures, we asked participants to recall the amount of the change in the Federal Funds Target Rate included in our materials. Ninety-five percent of the analysts correctly indicated that the rate increased by 50 basis points.14 This suggests that analysts correctly noticed the increase in interest rates across all conditions.

To ensure that results were not driven by differences in the perceived reliability of the fair value data-in particular that they might be viewed as more reliable in the FFV conditions where they were recognized versus the PFV conditions where they were only disclosed-we analyzed analysts’ perception of these data. A 2 x 3 ANOVA revealed no significant differences across conditions (all p’s > .360).

We asked several questions to determine whether analysts understood that the banks differed in their risk management strategies. First, we asked participants to indicate the extent to which the bank was exposed to interest rate risk at December 31, 20X3. Using a 15-point scale (endpoints labeled 1: “interest

12 Along with a description of the banks’ hedging strategies (which varied across high and low risk conditions), we described both liquidity risk and credit risk factors of the bank. Those were held constant.

13 We chose not to collect data for the FFV-SCE cells after carefully considering the benefits and costs of gathering the additional data. The theory underlying our PFV format predictions also predicts a lack of format effects across FFV contexts. However, we wanted to conduct our format tests in the same PFV measurement regime as Hirst and Hopkins (1998) and Maines and McDaniel (2000). Because these ex ante expectations are identical between the PFV and FFV contexts, we decided that the study would obtain very little incremental benefit from recruiting 33% additional bank-specialist analysts.

14 Two analysts indicated that interest rates decreased by 50 basis points and two analysts indicated that rates increased by 100 basis points. One analyst did not answer this manipulation check question.

rate risk is completely hedged” and 15: “interest rate risk is completely exposed”), analysts indicated that the high-risk bank (11.97) was more exposed to interest rate risk than the low-risk bank (7.57) (t = 7.118, p = .000). To ensure that analysts believed that the banks’ interest rate exposures would not change in the future, we also asked for their perceptions of the banks’ exposures two to three years hence. Analysts used a 15-point scale (endpoints labeled 1: “interest rate risk will be completely hedged” and 15: “interest rate risk will be completely exposed”). Participants responded that the high-risk bank (10.28) would be more exposed to interest rate risk than the low-risk bank (7.72) (t = 3.208, p = .002). Taken together, these findings suggest that analysts recognized the difference in interest rate risk across the low- and high¬risk bank and that analysts did not expect it to reverse in the near future.

We also asked participants to assess the banks’ market risk (defined as the possibility that changes in future market rates or prices will make positions less valuable). A 15-point scale (endpoints labeled 1: “much lower than the average bank” and 15: “much higher than the average bank” with a mid point of 8: “equal to the average bank”) was used to gather the responses. Analysts considered the high-risk bank (11.13) to be relatively riskier than the low-risk bank (8.29) (t = 6.33, p = .000). This held true across all format conditions (all p’s < .01).

Analysts were asked to recall where comprehensive income was disclosed. Eighty-six percent of the analysts correctly recalled the location. Interestingly, only 33% recalled correctly in the low-risk PFV¬SCE condition. This is the condition where it is least likely that a bank analyst would seek the CI data: they are not in a performance statement and, by virtue of the bank being hedged, there is little to gain by analyzing them. In the other five conditions, recall accuracy ranged from 83% to 100%.

Overall, we conclude that our manipulations worked as planned.

Hypothesis Tests

Consistent with our hypothesis development in Section 2, we first describe results for our format effect hypotheses and then describe results for our measurement hypotheses. All inferences are based on planned comparisons.

Reporting Format Hypotheses

The reporting format hypotheses (H1 and H2) focus on the two piecemeal fair value measurement conditions, PFV-IS and PFV-SCE. These two treatments measure comprehensive income identically (and incompletely) but present it in different financial statements. H1 predicts that when CI is reported on a piecemeal basis, bank analysts’ risk and valuation judgments will not be influenced by whether CI is reported in the SCE or as a performance measure in the IS. H2 predicts that bank analysts’ risk and valuation judgments will distinguish between high and low-risk banks, regardless of how PFV CI is disclosed.

We analyze two different dependent variables to capture analysts’ risk perceptions. First, we examine analysts’ assessments of the investment risk of the bank relative to that of an average bank of equivalent size. We use a 15-point scale (endpoints labeled 1: much lower than the average bank and 15: much higher than the average bank). Second, we investigate analysts’ assessments of the multiple with which they would capitalize the bank’s earnings (i.e., a PE multiple based on the bank’s trailing net income, which was identical acros s conditions).

The descriptive statistics for analysts’ investment risk and PE-ratio assessments are reported in Panel A of Tables 2 and 3, respectively. As predicted, analysts’ investment risk assessments across high and low risk conditions for PFV-IS (8.25) are not different from their investment risk judgments across the high and low risk conditions for PFV-SCE (8.36) (t = 0.17, p = .862). See Panel B of Table 2. Examination of format effects in each risk condition reveals a consistent pattern: there are no format effects across either the high or the low risk PFV cells (both p’s > .310).

INSERT TABLES 2 AND 3 ABOUT HERE

As predicted, analysts’ average PE ratio assessments across high and low risk conditions for PFV-IS (13.15) are not different from their PE ratio judgments across the high and low risk conditions for PFV¬SCE (12.81) (t = 0.57, p = .571). See Panel B of Table 3. Examination of format effects in each risk condition reveals a consistent pattern: again, there are no format effects across either the high or the low risk PFV cells (both p’s > .459).

Turning to the valuation judgments, Panel A of Table 4 reports descriptive statistics for analysts’ stock price judgments. Panel B of Table 4 reports that the average stock price in the PFV-IS conditions ($13.49) was not significantly different from that in the PFV-SCE conditions ($13.00) (t = 0.74, p = .462). As with the investment risk and PE ratio findings, and as predicted, there were no format effects for the valuation data across either the high or low risk PFV cells (both p’s > .580). Taken together, the investment risk, PE ratio, and stock price findings do not reject H1.

INSERT TABLES 2 AND 3 ABOUT HERE

In H2, we predict that, under piecemeal recognition of fair value gains and losses, bank analysts will judge the investment risk of the high-risk bank higher than that of the low-risk bank and assign lower PE ratios and share value estimates to the high-risk bank, regardless of the PFV CI format used. As reported in Panel B of Table 2, analysts’ investment risk judgments were significantly higher in the high-risk PFV conditions (9.31) than the low-risk conditions (7.28) (t = 3.29, p = .001). This result held across both the PFV-IS and the PFV-SCE conditions (both p’s > .066).

Panel B of Table 3 reports that analysts’ PE ratios were lower in the high-risk PFV conditions (12.68) than the low-risk conditions (13.32) but not significantly so (t = 1.01, p = .158). This was true in both the PFV-IS and PFV-SCE conditions (both p’s > .133). The stock price data in Table 4 mirror the PE ratio data. The high-risk bank is valued less ($12.86) than the low-risk bank ($13.69), but the difference is not significant (t = 1.26, p = .106). This, too, held across PFV-IS and PFV-SCE conditions (both p’s > .165).

Thus, as a whole in the PFV conditions, analysts perceived significantly different levels of investment risk across the two banks. However, when comprehensive income was reported on a piecemeal basis, bank analysts’ PE ratio estimates and valuation judgments did not distinguish between the two banks. As such H2 receives mixed support and the arguments in favor of more complete measurement and reporting of fair value and gains and losses increase their appeal. We turn now to the tests of the FFV data to shed light on this issue.

Meas u r em en t Hyp o th es es

The measurement hypothesis (H3) focuses on the full fair value (FFV-IS) cells and the piecemeal fair value cells where comprehensive income is reported in a performance statement (PFV -IS). Thus, across both measurement conditions, we hold constant the format of comprehensive income reporting (i.e., in a performance statement). What varies is the banks’ hedging strategy (high or low risk) and the completeness of fair value reporting (FFV-IS or PFV-IS).

H3 predicts that there will be a significant risk by measurement interaction for the investment risk, PE ratio, and stock price judgments in the FFV-IS and PFV-IS cells. In particular, the interaction will be driven by higher investment risk judgments and lower PE ratios and stock prices in the high-risk FFV-IS condition compared to the high-risk PFV-IS condition. The high-risk FFV-IS condition CI disclosure more clearly and completely reveals the extent of the risk taken by the bank by choosing not to hedge its market risk. This allows bank analysts to more-easily gauge the financial impact of the bank’s chosen risk management strategy and to incorporate that assessment into their stock prices. No differences are expected across the low-risk FFV-IS and PFV-IS conditions as the more-transparent and complete CI disclosure in the FFV-IS conditions does nothing more than confirm that the bank is hedged.15

Panel B of Table 2 indicates that the interaction for the investment risk variable is in the predicted direction and is significant (t = 1.94, p = .028). The difference in investment risk across the high and low risk FFV-IS conditions is significant (10.85 versus 7.09

15 The predictions of H3 assume that FFV-IS disclosure of CI for the high-risk bank reveals that the bank is riskier than perceived in the PFV conditions. Evidence that this was the case in our materials comes from analysts’ evaluation of the volatility of CI across the three high-risk conditions. Using a 15-point scale (labeled 1: low volatility and 15: high volatility), the perceived CI volatility of the high-risk bank was identical in the two PFV conditions (PFV-IS = 9.13, PFV-SCE: 9.22

Panel B of Table 3 indicates that the interaction for the analysts’ PE ratio assessments is in the predicted direction and is significant (t = 1.53, p = .065). The difference in PE ratios across the high and low risk FFV-IS conditions is significant (11.89 versus 13.89

Finally, Panel B of Table 4 indicates that the interaction for the analysts’ judgments also is in the predicted direction and is significant (t = 1.54, p = .069). The difference in price across the high and low risk FFV-IS conditions is significant ($11.25 versus $14.10

V. CONCLUSIONS

This study proposes hypotheses that predict differences in commercial bank analysts’ assessments of a bank’s investment risk and stock price as a function of the transparency of fair value gain and loss information. We test our predictions in an experiment in which we vary the accounting for the effects of interest rate risk on comprehensive income in three ways, across both low- and high-risk banks. The three comprehensive income accounting regimes we examine are: (1) piecemeal recognition where fair value gains and losses on investment securities are recognized in the statement of changes in equity and fair value gains and losses on all other financial assets and liabilities are disclosed in the notes (consistent with prevailing practice under SFAS 107 and SFAS 115)

We found that bank analysts do distinguish between banks in terms of their interest-rate risk. That is, we provide evidence that bank analysts attended to and understood the risk information. We also found that analysts’ investment risk and valuation judgments were not differentially influenced across piecemeal reporting formats. However, significant valuation differences did not emerge across high and low-risk banks under these piecemeal regimes. We attribute this to the difficulty even experienced bank analysts face in determining the financial impact of known risk differences. The two piecemeal fair value reporting regimes do little to help, as they are both incomplete in their measurement of fair value changes. The full fair value regime, on the other hand, allows bank analysts to more readily observe and process company¬specific performance-relevant effects of the hedging strategies. Consequently, under this complete measurement and transparent disclosure regime, significant differences in valuation judgments emerge.

This study contributes new evidence to three related but distinct lines of literature concerned with fair values, comprehensive income, and risk. First, the design and results of this study contribute to the accounting research literature that investigates the influence of financial reporting formats on valuation judgments (e.g., Hirst and Hopkins 1998, Maines and McDaniel 2000). We extend the existing boundaries of this research by examining risk and value judgments of specialists in the commercial bank industry, a context in which comprehensive income and interest rate risk are important elements of core operations. We address the limitations of Hirst and Hopkins (1998) that were raised by Lipe (1998) and Maines and McDaniel (2000) and their prediction that analysts who specialize in financial service firms will not be influenced by comprehensive income format choices. We find that in a context where comprehensive income is measured on a piecemeal basis, bank analysts’ judgments are not influenced by comprehensive income report format. Furthermore, concerns raised by bankers during the comprehensive income Exposure Draft public -comment period appear to be overstated. Piecemeal reporting of fair value gains and losses does not mislead bank analysts evaluating a fully hedged bank. 16

16 Recall that Maines and McDaniel (2000) find that in a PFV setting, individual investors did not fully use the CI information available to them. Just as our evidence suggests professional analysts are better served with FFV data, we believe that individual investors would be best served by the complete and transparent reporting of all fair value changes.

Nonetheless, we provide evidence that comprehensive income reporting may be deficient under piecemeal fair value recognition. Although bank analysts do not seem to penalize hedged banks when they report piecemeal comprehensive income losses in a statement of performance, we provide evidence that current measurement of comprehensive income may be inadequate in its ability to disclose the risks associated with banks that choose not to hedge their interest rate risk. When comprehensive income was measured on a full fair value basis and reported in a statement of performance, bank analysts were able to assess the financial impact of a risky bank’s interest rate risk management strategy and make corresponding adjustments to the bank’s valuation.17

Second, the results of this study add to archival accounting research on the information in fair values and comprehensive income by addressing a question that is difficult to test using archival data. We vary the degree of transparency associated with the interest rate risk of a bank, holding all else constant, and examine whether market participants’ risk assessments and valuation judgments are influenced by alternative comprehensive income measurements and formats. Thus, we extend research by Barth, Landsman, and Wahlen (1995) who find that volatility in fair value income is not associated with a reduced price-earnings multiple for a sample of banks. Our results suggest that two factors may be at play. First, as Barth, Landsman, and Wahlen (1995) note, their results may be due to banks having hedged asset and liability positions. Our results for the low-risk bank demonstrate that fair value gains and losses on available for sale securities are not incrementally value relevant in circumstances where they are offset by other gains and losses. An open question remains as to whether a reanalysis of Barth, Landsman, and Wahlen (1995), with controls for differences among banks’ hedging strategies, would reveal the value relevance of fair value gains and losses. We find that for an unhedged bank, unless the fair value gains and losses are completely measured and transparently disclosed in a statement of performance (a disclosure regime that does not currently exist), experienced bank analysts do not appear to fully

17 In addition, more transparent information (i.e., information presented in a ready to use format) may be more likely to be used because of features inherent in analysts’ information environment. Consider, for example, that many analysts screen investment opportunities using data sources that omit footnote data. Thus, fair value information incorporated directly into the financial statements may help screen in (out) favorable (unfavorable) investment opportunities.

incorporate these items into their stock price judgments. They recognize the differential risk, but they do not incorporate the impact of that risk on performance.

Finally, we contribute ex ante evidence to standard setters that should be useful in their deliberations over whether to require all financial instruments to be reported at fair value, and whether fair value gains and losses should be recognized in a statement of performance or in the statement of owners’ equity. Consistent with the findings of Hirst and Hopkins (1998), we show that clear and complete disclosure of value-relevant financial information in a statement of performance does improve professional analysts’ judgments over and above the case where data are measured and reported on a piecemeal basis with other data scattered throughout the financial statements. We maintain that financial accountants and standard setters can, and should, play a role in determining not only how to measure items, but also how they should be disclosed.

REFERENCES

Barth, M., W. Landsman, and J. Wahlen. 1995. Fair Value Accounting: Effects on Banks’ Earnings Volatility, Regulatory Capital, and Value of Contractual Cash Flows. Journal of Banking and Finance 19 (June): 577-605.

Brown, P. R. “Financial Data and Decision Making by Sell-Side Analysts.” The Journal of Financial Statement Analysis. (Spring 1997): 43-48.

Collins, D. W. and S. P. Kothari. 1989. An Analysis of Intertemporal and Cross-sectional Determinants of Earnings Response Coefficients. Journal of Accounting and Economics 11 (2/3): 143-181.

Coopers & Lybrand L.L.P. 1997. Monthly Financial Reporting Release: Accounting and Auditing Developments. (July) New York, NY.

Easton, P. D. and M. E. Zmijewski. 1989. Cross-Sectional Variation in the Stock Market Response to Accounting Earnings Announcements. Journal of Accounting & Economics. 11(2,3): 117-141.

Financial Accounting Standards Board. 1991. Statement of Financial Accounting Standards No. 107, Disclosures about Fair Value of Financial Instruments. Norwalk, CT: FASB.

1993. Statement of Financial Accounting Standards No. 115, Accounting for Certain Investments in Debt and Equity Securities. Norwalk, CT: FASB.

1996. Exposure Draft: Reporting Comprehensive Income. Norwalk, CT: FASB.

1997. Statement of Financial Accounting Standards No. 130: Reporting Comprehensive Income. Norwalk, CT: FASB.

1999. Preliminary Views on Major Issues Related to Reporting Financial Instruments and Certain Related Assets and Liabilities at Fair Value. Norwalk, CT: FASB.

Heisler, J. 1994. Loss aversion in a futures market: An empirical test. Review of Futures Markets 13: 793-822.

Hirst, D. E. and P. E. Hopkins. 1998. Comprehensive Income Reporting and Analysts’ Valuation Judgments. Journal of Accounting Research. 36 (Supplement): 47-75.

, , and A. Yen. 2001. Anatomy of a New Accounting Standard: An Analysis of SFAS 130 Comment Letters. Working paper: University of Texas at Austin.

, K. E. Jackson, and L. Koonce. 2001. Improving Financial Reports by Revealing the Prior Accuracy of Estimates. Working paper: University of Texas at Austin.

Hodder, L., L. Koonce, and M.L. McAnally. 2001. SEC market risk disclosures: Implications for judgment and decision making. Accounting Horizons 15 (March): 49-70.

Hopkins, P. E., R. W. Houston, and M. F. Peters. 2000. Purchase, Pooling, and Equity Analysts’ Valuation Judgments. Accounting Review. 75 (July): 257-281.

Hunton, J.E., and R.A. McEwen. 1997. An assessment of the relation between analysts’ earnings forecast accuracy, motivational incentives, and cognitive information search strategy. The Accounting Review 72 (October): 497-515.

Joint Working Group Of Standard Setters. 2000. Recommendations on accounting for Financial Instruments and Similar Items. Norwalk, CT: Financial Accounting Standards Board.

Kleinbaum, D., L. Kupper, and H. Morgenstern. 1982. Epidemiologic Research: Principles and Quantitative Methods. Belmont, CA: Lifetime Learning Publications.

Lipe, M. G. 1998. Discussion of Comprehensive Income Reporting and Analysts’ Valuation Judgments. Journal of Accounting Research. 36 (Supplement): 77-83.

Maines, L. A. and L. S. McDaniel. 2000. Effects of Comprehensive-Income Characteristics on Nonprofessional Investors’ Judgments: The Role of Financial-Statement Presentation Format. Accounting Review. 75 (April): 179-207.

Olsen, R. 1997. Investment risk: The experts’ perspective. Financial Analysts Journal 53 (March/April): 62-66.

Ryan, S. (1997). “A Survey of Research Relating Accounting Numbers to Systematic Equity Risk, with

Implications for Risk Disclosure Policy and Future Research.” Accounting Horizons (June): 82-95.