The Value-Relevance of Derivative Disclosures by Commercial Banks ...

Review of Quantitative Finance and Accounting, 25: 413–427, 2005 c 2005 Springer Science + Business Media, Inc. Manufactured in The Netherlands.

The Value-Relevance of Derivative Disclosures by Commercial Banks: A Comprehensive Study of Information Content Under SFAS Nos. 119 and 133 LI WANG Department of Accounting, Kent State University, College of Business Administration, Kent, Ohio 44242 E-mail: [email protected] PERVAIZ ALAM∗ Department of Accounting, Kent State University, College of Business Administration, Kent, Ohio 44242, Tel.: 330-672-1121, Fax: 330-672-2548 E-mail: [email protected] STEPHEN MAKAR Department of Accounting, University of Wisconsin Oshkosh, College of Business Administration, Oshkosh, WI 54901-8677 E-mail: [email protected]

Abstract. This study examines the value-relevance of banks’ derivative disclosures under Statements of Financial Accounting Standards (SFAS) Nos. 119 and 133. Using the complete time-series of SFAS No. 119 disaggregated notional value disclosures and the most recently available SFAS No. 133 fair value data, this study investigates whether such expanded disclosures provide incremental information content beyond earnings and book value. Our results indicate that banks’ notional principal amount disclosures are value-relevant, and that this evidence of incremental information content is robust to the inclusion of recently available fair value data and alternative model specifications. Key words: derivatives, banks, information content JEL Classification: M41, G21

1.

Introduction

The ongoing dramatic growth in the use of derivatives, together with the accompanying derivative debacles, has motivated accounting regulators to develop and expand disclosure requirements. In the U.S., for example, accounting regulators (Financial Accounting Standards Board, FASB) began their derivatives project in 1986. The FASB subsequently sought to improve the usefulness of publicly available derivatives information with significant revisions to disclosure requirements in 1994 and 1998. This study investigates whether such expanded derivative disclosures provide incremental information content beyond earnings and book value. ∗ Corresponding

author.

414

WANG, ALAM AND MAKAR

Corresponding to the FASB’s derivatives project, the notional amount of derivatives increased more than twelve-fold, from $1.5 trillion in 1986 to $20.2 trillion in 1994 (Bank for International Settlements, 2004). Derivative debacles in 1994-1995 alone included the bankruptcy of Orange County in California, and substantial financial losses at major U.S. corporations such as Procter and Gamble, and Gibson Greetings. By the end of 2003, the notional amount of derivatives reached $234 trillion, and was accompanied by U.S. business scandals at Enron and Fannie Mae that centered on their improper derivatives reporting. The increasing use of derivatives suggests that risk management practices have changed over the years. In fact, previous research suggests that derivatives are an endogenous component of the firm’s risk management strategy (Nance, Smith and Smithson, 1993; Géczy, Minton and Schrand, 1997; Henstschel and Kothari, 2001). Before the U.S. disclosure requirements on derivatives were written, firms selectively reported their derivative transactions, including any associated gains or losses. The lack of consistent disclosure became a major issue when the substantial losses became apparent. In an attempt to address this lack of useful publicly available information, the FASB issued Statements of Financial Accounting Standards (SFAS) No. 119 in 1994 and SFAS No. 133 in 1998. However, prior research on the information content of derivative disclosures has been inconclusive (e.g., Khurana and Kim, 2003). Some previous studies were performed before the accounting standards were fully implemented (e.g., Riffe, 1997), while others use only two years of post-implementation data (e.g., Barth, Beaver and Landsman, 1996; Nelson, 1996; Venkatachalam, 1996). Seow and Tam (2002) contribute to this literature using large banks’ disclosure data for 1990–1996 in a returns regression framework, but do not find any evidence that the notional principal amounts of derivatives are value-relevant. These authors call for more research on the information content of derivative notional amounts in particular, as well as on the usefulness of subsequently available derivative disclosures. This study investigates the value-relevance1 of banks’ derivative disclosures for the entire SFAS No. 119 reporting period (i.e., 1994–2000) as well as for the most recent SFAS No. 133 reporting period (i.e., 2001–2002). Using data from the Chicago Federal Reserve Bank’s Call and Income Reports, we examine the disaggregated notional values of derivatives by risk category (i.e., interest rate and foreign exchange) and by intended use (i.e., trading and non-trading). Our investigation focuses on the banking industry because of the uniqueness of the industry and the size of derivative losses some banks have reported (see Springett, 1995; Kaplan, 1997). Recent data show that the five largest banks have entered into derivative contracts of nearly $78 trillion (Office of the Comptroller, 2004). In the banking industry, there is also a severe moral hazard problem due to both the heavy federal and state regulation, in general, and the federal insurance protection of bank deposits, in particular. The research design used in this study follows the work of Ohlson (1995), which is noted for its rigorous specification of the association between accounting information and stock prices (Richardson and Tinaikar, 2004). Our results indicate that the notional principal amounts of derivatives are economically significant and provide incremental information content beyond earnings and book value. Moreover, such results are robust to the inclusion of SFAS No. 133 fair value data as well as to alternative model specifications. The remainder of this paper is organized as follows. Section 2 describes the FASB’s derivatives project and the accompanying disclosure requirements. In addition, hypotheses

415

THE VALUE-RELEVANCE OF DERIVATIVE DISCLOSURES

pertaining to such reporting requirements are developed in relation to prior research. Section 3 details the empirical models, Section 4 discusses sample selection and data description, and Section 5 presents the empirical results. Finally, Section 6 provides a summary and suggestions for future research.

2.

Derivative disclosures and prior research

Table 1 shows that derivatives held by commercial banks and trust companies at the end of September 2004 totaled $82 trillion, of which $78 trillion were held by the 5 largest banks. Table 1 also shows that most of the derivatives held by all banks and trust companies ($79.69 trillion) were held for trading purposes. As introduced in the previous section, the U.S. accounting regulators (FASB) began their study of derivatives in 1986. Because of the intricacies of derivative instruments, the FASB elected to address this problem in a series of Statements of Financial Accounting Standards: SFAS Nos. 105, 107, 119, and 133. Among the required disclosures of SFAS No. 105 (FASB, 1990), firms reported the face, contract or notional principal amount of financial instruments with off-balance-sheet risk. SFAS No. 107 (FASB, 1991) expanded such derivatives reporting to include the fair value amounts of all financial instruments (assets and liabilities) in notes to the financial statements, suggesting that FASB believed that the information was relevant to financial statement users. Barth (1994) demonstrates that the fair value amounts of banks’ financial investment securities are value-relevant, and that such information content exceeds that of historical costs. In an attempt to substantively improve the derivative disclosures required under both of these standards, the FASB issued SFAS No. 119 in 1994 (effective in December 1994). Among the expanded reporting mandated in this standard is the requirement that firms provide disaggregated notional value disclosures (e.g., asset versus liability positions). The fourth and most recent accounting standard, SFAS No. 133 (FASB, 1998), was effective for fiscal years beginning after June 2000. Subsequent to its issuance, the FASB’s

Table 1. Notional amounts of derivatives contracts held by commercial banks and trusts, September 30, 2004 (millions of dollars)

Rank

Bank name

Total assets

1 JPMorgan Chase Bank $661,772 2 Bank of America NA 740,695 3 Citibank NA 651,346 4 Wachovia Bank NA 380,236 5 HSBC Bank USA NA 118,454 Top 5 banks & trusts with derivatives $2,552,503 Other 662 banks & trusts with derivatives 4,202,500 Total amounts for 667 banks & trusts $6,755,003

Total derivatives

Total held for trading MTM

Total not traded MTM

$42,128,402 16,165,883 15,154,814 2,947,107 1,655,178 $78,051,384 4,215,941 $82,267,325

$42,028,862 15,586,689 14,927,224 2,597,611 1,639,177 $76,779,563 2,909,841 $79,689,404

$99,540 579,194 227,590 349,496 16,002 $1,271,822 1,306,100 $2,477,922

Notes: Credit derivatives are excluded from the sum of total derivatives. Data source: Office of the Comptroller of the Currency (2004). MTM = marked to market.

416


Derivatives Implementation Group has issued more than 180 guidelines to help firms understand and apply this standard’s complex reporting requirements. SFAS No. 133 hedge accounting, in particular, has been criticized as being “so idiosyncratic and . . . esoteric that auditing departments don’t have the expertise to implement this without bringing in specialist expertise” (Hawser, 2004, p. 45). For example, derivatives designated as one of three types of hedges (cash flow hedge, fair value hedge, net investment hedge) receive special accounting treatment by deferring gains or losses until the underlying transaction is complete. On the other hand, when derivatives are not designated as hedges, or if the hedge is ineffective, derivatives are marked-to-market on the balance sheet, with the corresponding gain or loss reported either on the income statement or as a separate component of equity. Indeed, SFAS No. 133 is “notorious for being the most complex of any of the FASB’s pronouncements” (Kawaller, 2004, p. 24), and thus the information content of the most recently available derivative disclosures are an open question. This study investigates the usefulness of notional and fair value derivative disclosures by commercial banks under SFAS Nos. 119 and 133. Sample firms are drawn from the banking industry in light of their large-scale use of derivatives and their comprehensive derivatives reporting. Given the existing gap in such prior research, we are interested primarily in the information content of derivative notional principal amounts. In contrast to prior research, we focus our analysis on the entire reporting period of SFAS No. 119, and take advantage of recently available SFAS No. 133 data to examine the sensitivity of notional value disclosures to fair value data. As stated above, SFAS No. 119 is noteworthy in that it requires firms to provide disaggregated notional value disclosures by asset versus liability position, by category of derivative (e.g., interest rate and foreign exchange), and by purpose for which the derivative is held (e.g., trading and non-trading). By incorporating such expanded disclosures, we are able to construct more powerful tests of incremental information content than was possible prior to SFAS No. 119. While both notional and fair value disclosures have limitations (e.g., Hentschel and Kothari, 2001), recent research has cautioned that banks’ fair value disclosures, in particular, may not be reliable (Nissim, 2003; Khurana and Kim, 2003). Indeed, Fannie Mae recently has been charged by the SEC with manipulating SFAS No. 133 fair value adjustments to smooth earnings fluctuations (Hagerty, 2004). As introduced in Section 1, prior research on the value-relevance of derivative disclosures is inconclusive. Riffe (1997) reports that notional value disclosures by banks for 1986–1989 are positively related to their market values. In contrast, Venkatachalam (1996) provides evidence that suggests notional derivative disclosures for 1993–1994 have negative valuation implications. Moreover, Venkatachalam reports that the value-relevance of banks’ notional value disclosures is incremental to fair value, and vice versa. Similarly, Eccher, Ramesh and Thiagarajan (1996) and Nelson (1996) control for notional principal amounts, but provide limited results pertaining to the information content of fair value disclosures for 1992–1993. Taking advantage of a longer time-series of derivatives data, Seow and Tam (2002) find that all derivative disclosures for 1990–1996, except notional principal amounts, are valuerelevant. As noted in Section 1, we respond to these authors’ call for additional research on the information content of notional derivative value disclosures.


417

This study contributes to the literature on the information content of derivatives reporting by using the complete time-series of disaggregated derivative disclosures under SFAS No. 119 and the rigorous specification of the Ohlson (1995) firm valuation framework. In particular, we examine the following hypotheses (in alternative form). Hypothesis 1. The notional amounts of trading derivatives are value-relevant, and provide additional explanatory power beyond earnings and book value. Hypothesis 2. The notional amounts of non-trading derivatives are value-relevant, and provide additional explanatory power beyond earnings and book value. These two formal hypotheses, which distinguish a bank’s intent of holding derivatives, are investigated for each of the two major risk categories faced by our sample firms: interest rates and foreign exchange. Beyond these four tests pertaining to our primary analyses, we examine the robustness of such results to the inclusion of SFAS No. 133 fair value disclosures (i.e., fair value gains/losses for trading derivatives or fair value asset/liability positions for non-trading derivatives). We also examine the sensitivity of our primary results to alternative model specifications.

3.

Empirical models

Bank managers are likely to hedge when there is uncertainty about the variability of their cash flows and the consequent effect on firm value. For example, Géczy, Minton and Schrand (1997) and Graham and Rogers (2002) report that firms use derivatives to reduce cash flow variability, which allows them to raise additional funds and invest in new growth opportunities. Increased investment, in turn, may bring increased firm value. Following Dechow, Hutton and Sloan (1999) and Barth et al. (1998), we use Ohlson’s (1995) valuation framework to develop our empirical models for assessing the impact of derivatives use on firm value. Ohlson’s (1995) valuation framework can be expressed as follows: Pit = β0 + β1 E it + β2 BVit + β3 Vit Stock price (Pit ) proxies for firm valuation, and is calculated for each firm i three months after fiscal year-end period t. Earnings (E it ) are income before extraordinary items for firm i in period t, while book value (BVit ) is calculated as of the end of period t. Vit captures the information about future abnormal earnings based on non-financial statement variables. In our primary analyses, we examine hypothesis 1 pertaining to the value-relevance of trading derivatives in total (TDER), as well as trading interest rate derivatives (TINT) and trading foreign exchange derivatives (TFX), using the following OLS regression models. Pit = α0 + α1 E it + α2 BVit + α3 SALEGROWit + α4 TDERit + εit Pit = α0 + α1 E it + α2 BVit + α3 SALEGROWit + α4 TINTit + α5 TFXit + εit

(1) (2)

418


As noted above, these models are based on the well-known Ohlson (1995) firm valuation framework, which indicates that summary accounting measures such as earnings and book value are useful in explaining variations in firm value.2 The sales growth variable (SALEGROWit ) is used in prior research to control for omitted variables (e.g., Skinner, 1996). Specifically, SALEGROW operationalizes “other information” specified in the Ohlson (1995) framework, to proxy for future growth potential. This variable is calculated for each firm i over the three years ending in period t.3 With regard to the primary variables of interest, TDER, TINT and TFX are the notional amounts of trading derivatives disclosed under SFAS No. 119. In contrast to the control variables in our models, such notional amounts are not recognized in the financial statements, but may reveal important information regarding the magnitude and the purpose of a bank’s derivatives holdings. Thus, we predict that TDER, TINT and TFX will provide additional explanatory power beyond book value and earnings (i.e., the coefficients of TDER, TINT and TFX will be significant after controlling for book value, earnings, and sales growth). As summarized in Section 2, Riffe (1997) finds that the notional amounts of derivatives are positively related to bank equity, while Venkatachalam (1996) documents that the notional amounts are negatively related to bank equity value. If notional amounts serve as a proxy for the expected future net benefits, then the coefficient is expected to be positive. However, prior studies (e.g., Venkatachalam, 1996; McAnally 1996; Riffe, 1997) suggest that if the stock market perceives the notional amounts as a proxy for risk in off-balance sheet instruments, then the coefficient is expected to be negative. Accordingly, we do not predict the signs of the TDER, TINT and TFX estimated coefficients . To test the sensitivity of hypothesis 1 to the inclusion of SFAS No. 133 fair value gains (GAINS) or fair value losses (LOSSES), we estimate the following two OLS models. Pit = α0 + α1 E it + α2 BV2it + α3 SALEGROWit + α4 TINTit + α5 TFXit + α6 GAINSit + εit Pit = α0 + α1 E it + α2 BV2it + α3 SALEGROWit + α4 TINTit + α5 TFXit +α6 LOSSESit + εit

(3) (4)

While the variables of interest (TINTit and TFXit ), as well as the earnings (Eit ) and SALEGROW variables, are the same as in equations (1) and (2), book value (BV2it ) is calculated before SFAS No. 133 revaluation gains or losses of trading derivatives for firm i in period t. Equations (3) and (4) isolate the effect of SFAS No. 133 fair value disclosures of trading derivatives on stock price, after controlling for the notional amount disclosures as well as earnings, book value, and sales growth. To the extent that SFAS No. 133 fair value disclosures provide incremental information beyond notional amounts and the three control variables, we expect that the coefficient of GAINS (LOSSES) will be positive (negative) and significant. Moving to hypothesis 2, primary tests pertaining to the value-relevance of non-trading derivatives in total (NTDER), as well as non-trading interest rate derivatives (NTINT) and non-trading foreign exchange derivatives (NTFX), employ the following OLS regression models. The sensitivity of hypothesis 2 to SFAS No. 133 disclosures is considered by incorporating fair value data for assets (NTA) and liabilities (NTL) when non-trading derivative


419

values are positive or negative, respectively. Pit = α0 + α1 E it + α2 BVit + α3 SALEGROWit + α4 NTDERit + εit (5) Pit = α0 + α1 E it + α2 BVit + α3 SALEGROWit + α4 NTINTit + α5 NTFXit + εit (6) Pit = α0 + α1 E it + α2 BV3it + α3 SALEGROWit + α4 NTINTit + α5 NTFXit + α6 NTAit + α7 NTLit + εit

(7)

These three OLS models parallel the hypothesis 1 tests, except that the variables of interest in hypothesis 2 are non-trading derivatives in total (NTDER), non-trading interest rate derivatives (NTINT) and foreign exchange derivatives (NTFX). Similar to robustness checks of hypothesis 1, the equation (7) book value (BV3it ) has been adjusted for nontrading derivative assets (NTA) or non-trading derivative liabilities (NTL). To the extent that SFAS No. 119 notional value disclosures of non-traded derivatives provide information regarding the magnitude and purpose of banks’ derivative holding, we expect the estimated coefficients on NTDER, NTINT and NTFX to be significant. Similar to our hypothesis 1 tests, the signs of such coefficients are indeterminate (e.g., Venkatachalam, 1996; Riffe, 1997). While our formal hypotheses pertain to notional value disclosures, we take advantage of recently available fair value disclosures under SFAS No. 133. If such fair value data provide additional explanatory power, the estimated coefficients on NTA (NTL) will be positive (negative) and significant. As a final test of robustness, the dependent variable in the above seven equations (stock price, Pit ) is replaced with firm value deflated by book value. In this way, we respond to prior studies that caution that tests of fair value information content may be sensitive to model specification (e.g., Simko, 1999; Mozes, 2002). Given the use of book value as a deflator, the independent variables in the models below are deflated accordingly. Thus, for example, equation (1) becomes: MVit /BVit = α0 1/BVit + α1 Eit /BVit + α2 + α3 SALEGROWit /BVit + α4 TDERit /BVit + εit

(8)

In this deflated model, market value (MVit ) is defined as stock price times the number of shares outstanding for firm i at the end of period t. The deflator (BVit ) is the book value of equity for firm i at the end of period t, where the estimated intercept (α 2 ) can be interpreted as the estimated coefficient on book value (Core, Guay and Buskirk, 2003). All other variables are defined as before. 4.

Sample selection and data description

Data related to derivatives in this study come from the Consolidated Reports of Condition and Income for Banks with Domestic and Foreign Offices (FFIEC 031), filed with the Federal Financial Institutions Examination Council by state and national banks. These reports are available from the website of the Chicago Federal Reserve Bank. Other financial

420


Table 2. Descriptive statistics (millions of dollars) Variable All tests TA MV BV E SALEGROW Hypothesis 1 variables TINT TFX TCOM GAINS LOSSES Hypothesis 2 variables NTINT NTFX NTCOM NTA NTL

Mean

Std. dev.

Minimum

Maximum

3,873 1,002 375 60 754

10,651 3,503 1,054 189 19,064

93 1 5 (326) (65)

115,149 40,835 9,101 1,766 521,800

1,801 735 0 6 6

18,616 8,409 1 118 113

0 0 0 0 0

345,464 137,193 20 2,647 2,673

3,112 301 1 8 1

0 0 0 0 0

58,724 6,264 20 237 45

417 35 0 0 0

TAit is the total assets of firm i at the end of year t; MVit is stock price multiplied by the number of common stock shares outstanding for firm i at end of year t; BVit is the book value of firm i at the end of year t; Eit is the income before extraordinary items of firm i for year t; SALEGROWit is sales growth for firm i over three years ending in year t; TINTit is the notional amount of trading interest derivatives of firm i for year t; TFXit is the notional amount of trading foreign exchange derivatives of firm i for year t; TCOMit is the notional amount of trading commodity derivatives of firm i for year t; Gains and losses are revaluation gains and losses of trading derivatives of firm i in year t; NTINTit is the notional amount of non-trading interest derivatives of firm i in year t; NTFXit is the notional amount of non-trading foreign exchange derivatives of firm i in year t; NTCOMit is the notional amount of non-trading commodity derivatives of firm i in year t; NTAit is the positive fair values of non-trading derivatives for firm i in year t (reported as assets); and NTLit is the negative fair values of non-trading derivatives for i in year t (reported as liabilities).

data (stock price, number of shares outstanding, book value, earnings, sales growth and total assets) are obtained from Standard & Poor’s Research Insight. The 1994–2002 sample period includes the entire SFAS No. 119 reporting period (i.e., 1994 through May 2000) and the most recent SFAS No. 133 reporting period (i.e., June 2000 through year-end 2002). The final sample consists of 161 banks and 992 firm-year observations. Table 2 presents the descriptive statistics of the sample data. The sample data include both state and national banks, of varying size. For example, total assets range from $93 million to $115 billion, with an average of $3.9 billion. The average notional amount for trading interest derivatives (TINT) is about $1.8 billion, which is more than twice the average ($0.7 billion) for foreign exchange derivatives (TFX). In contrast, very few banks use commodity derivatives (TCOM), indicated by the zero mean amount of this type of derivatives. Therefore, these derivatives are excluded from our subsequent analyses. The average revaluation gains (GAINS) and losses (LOSSES) are about the same, both around $6 million. The revaluation gains and losses are only a small fraction of the notional amount of the trading derivatives. Such modest levels of fair value gains or losses are consistent

421


with prior research findings that derivatives use has a relatively small effect on the total risk profile of banks (e.g., Hentschel and Kothari, 2001). Banks also recognize gains or losses as a result of interest rate changes. However, some banks are better able to hedge interest rate risk by matching maturities of assets and liabilities and/or by matching re-pricing dates. The notional amounts for non-trading derivatives are considerably smaller than for trading derivatives. The average non-trading interest derivatives (NTINT) is $417 million, which is only about one-fourth of the trading interest derivatives (TINT). The average notional amount for non-trading foreign exchange derivatives (NTFX) is $35 million, only a fraction of the trading foreign exchange derivatives (TFX) and much smaller than the non-trading interest derivatives. Also, very few banks have non-trading commodity derivatives (NTCOM). Similarly, the fair values of the non-trading derivatives are very small, indicated by the statistics of NTA and NTL. The averages for NTA and NTL are zero, and the maximums are only $237 million and $45 million, respectively. These numbers are only small fractions of the notional amounts for the non-trading derivative notional amounts, let alone of total assets and market value. The considerable difference between the fair values of the non-trading derivatives and their notional amounts, as well as total assets and market value, is consistent with Guay and Kothari (2003), who report that the effect of derivative use is modest relative to firm-level measures. Table 3 reports the Pearson and Spearman correlations for the variables specified in the models. All variables are deflated by number of shares. The correlation between revaluation gains and losses (GAINS and LOSSES) is very high. The Pearson and Spearman correlations are 1.00 and 0.89, respectively. As detailed in Table 1, the top 5 banks hold about 95% of the derivatives market (in terms of notional values). Moreover, as these banks trade positions largely with each other, it is not surprising that derivative gains and losses are perfectly correlated by one measure and highly correlated by another measure.4 Accordingly, we

Table 3. Correlations PRICE BV PRICE BV 0.59 E 0.78 TINT 0.34 TFX 0.25 GAINS 0.13 LOSSES 0.14 NTINT 0.30 NTFX 0.23 NTA 0.10 NTL 0.01 SALEGROW (0.10)

0.62

E 0.64 0.56

0.66 0.05 0.28 (0.01) 0.17 0.08 0.13 0.10 0.14 0.13 0.27 0.06 0.17 0.09 0.10 0.02 0.03 (0.28) (0.23)

TINT TFX 0.35 0.16 0.19 0.51 0.19 0.19 0.68 0.38 0.25 0.12 0.08

0.16 (0.04) 0.05 0.61

GAINS LOSSES NTINT NTFX NTA 0.14 0.09 0.10 0.62 0.29

0.17 0.17 0.89 0.37 0.25 0.65 0.13 0.03 (0.01) (0.01) 0.09 0.13 (0.13)

0.14 0.08 0.09 0.64 0.30 1.00 0.26 0.13 0.06 0.09 (0.12)

0.32 0.21 0.20 0.40 0.02 0.11 0.11 0.32 0.37 0.20 (0.03)

0.33 0.17 0.18 0.32 0.16 0.33 0.32 0.31

0.08 0.11 0.09 0.04 (0.01) (0.01) (0.01) 0.08 (0.01)

NTL

SALEGROW

(0.01) 0.01 0.02 (0.00) (0.01) 0.02 0.02 0.00 (0.01) (0.01)

(0.04) (0.04) (0.04) (0.01) (0.00) (0.00) (0.00) (0.01) (0.00) (0.00) (0.00)

0.03 (0.02) 0.11 0.07 (0.08) (0.06)

Pearson correlations are reported above the diagonal, Spearman correlations are below the diagonal. All variables are per-share. PRICE is the closing price of firm i, three months after fiscal year-end period t. All remaining variables are as defined in Table 2.

422


include only the GAINS variable in the Section 5 reported results to avoid multicollinearity problems. The empirical results for the LOSSES variable are footnoted. The dependent and independent variables in such analyses have been windsorized at both the top and bottom 1% to mitigate the effect of extreme values. 5.

Results

Table 4 presents the OLS regression results pertaining to our primary tests of hypothesis 1, where models 1–3 correspond to the first three models developed in Section 3.5 OLS estimates of model 0 are provided for completeness, and are consistent with prior research indicating that all three Ohlson model variables (BV, E, and SALEGROW) are useful in estimating firm value. Similarly, the variables of interest (TDER, TINT, and TFX) are positive and statistically significant (at a.01 level). Consistent with expectations, the notional amounts of trading derivatives provide incremental information content beyond earnings, book value, and sales growth. Moreover, these results are robust to the inclusion of SFAS No. 133 fair value gains (GAINS).6 The magnitude of the estimated coefficients on foreign exchange derivatives disclosures, in particular (TFX), is notable compared to the estimated coefficients on either the total trading derivatives (TDER) or the interest rate derivatives (TINT). Moving to the non-trading derivatives results, Table 5 details the OLS regression estimates pertaining to tests of hypothesis 2 in models 5–7. The estimated coefficients for the Ohlson model variables (BV, E, and SALEGROW) are comparable to the Table 4 results, and indicate that all three variables are useful in estimating firm value. Likewise, the variables Table 4. OLS regression—primary tests of hypothesis one (Trading Derivatives) Priceit = a + b ∗ Independent Variableit + eit Independent Variables

MODEL 0 MODEL 1 MODEL 2 MODEL 3

Estimate P-value Estimate P-value Estimate P-value Estimate P-value

Intercept BV

E

SALEGROW

2.81984 0.00000 3.25320 0.00000 3.12079 0.00000 3.12304 0.00000

8.76632 0.00000 8.22527 0.00000 8.12397 0.00000 8.11421 0.00000

0.02238 0.00000 0.02173 0.00000 0.02079 0.00000 0.02061 0.00000

0.32210 0.00000 0.31181 0.00000 0.33761 0.00000 0.33847 0.00000

TDER

TINT

TFX

GAINS

ADJ RSQ 0.6394

0.06451 0.00000

0.6616 0.04227 0.00811 0.04618 0.00767

0.86258 0.6698 0.00786 0.87858 (0.68057) 0.6698 0.00707 0.46808

Number of observations = 992. Tests of hypothesis one, that the notional amounts of trading derivatives are value-relevant, use pooled time-series cross-sectional OLS estimates of models 0–3. Models 1–3 correspond to Section 3 equations (1)–(3), while model 0 is provided for completeness. As noted in relation to Table 3, model 4 results are reported in a text footnote. A statistically significant coefficient on any of the variables of interest (TDER, TINT, or TFX) is consistent with notional derivative amounts providing incremental information content beyond Ohlson model control variables (BV, E, and SALEGROW) and SFAS No. 133 fair value gains or losses (GAINS, LOSSES). See Tables 2 and 3 for variable definitions. All p-values are based on White (1980) heteroscedastcity corrected t-values and two-tailed tests.

423

THE VALUE-RELEVANCE OF DERIVATIVE DISCLOSURES Table 5. OLS regression—primary tests of hypothesis two (Non-trading Derivatives) Priceit = a + b ∗ Independent Variableit + eit Independent variables

MODEL 5 Estimate P-value MODEL 6 Estimate P-value MODEL 7 Estimate P-value

Intercept BV

E

SALEGROW NTDER NTINT NTFX

3.1785 0.0000 3.4021 0.0000 3.5679 0.0000

8.4845 0.0000 8.2454 0.0000 8.2709 0.0000

0.0218 0.0000 0.0202 0.0000 0.0200 0.0000

0.3060 0.0000 0.3085 0.0000 0.2892 0.0000

0.0634 0.00004

NTA

NTL

ADJ RSQ 0.6466

0.03570 0.07035 0.03540 0.07535

3.36650 0.6573 0.00001 3.37350 2.37710 260.1000 0.6561 0.00001 0.62844 0.16602

Number of observations = 992. Tests of hypothesis two, that the notional amounts of non-trading derivatives are value-relevant, use pooled timeseries cross-sectional OLS estimates of models 5–7 which correspond to Section 3 equations (5)–(7). A statistically significant coefficient on any of the variables of interest (NTDER, NTINT, or NTFX) is consistent with notional derivative amounts providing incremental information content beyond Ohlson model control variables (BV, E, and SALEGROW) and SFAS No. 133 fair value derivative assets (NTA) or fair value derivative liabilities (NTL). See Tables 2 and 3 for variable definitions. All p-values are based on White (1980) heteroscedastcity corrected t-values and two-tailed tests.

of interest (NTDER, NTINT, and NTFX) are positive and statistically significant (NTDER and NTFX are significant at a.05 level, while NTINT is significant at a.10 level). The results are consistent with hypothesis 2. The model 5 results, for example, indicate that the notional value of total non-traded derivatives (NTDER) provides incremental information content beyond book value, earnings, and sales growth. Similarly, the notional values of both non-traded interest rate derivatives (NTINT) and non-traded foreign exchange derivatives (NTFX) in model 6 are value-relevant. Finally, the model 7 estimation results indicate that these results are robust to the inclusion of SFAS No. 133 fair value derivative assets (NTA) or liabilities (NTL). The absence of value-relevance evidence for the asset (NTA) and liability (NTL) variables in model 7 may be due, in part, to their economic insignificance, as discussed in relation to our Table 2 descriptive statistics. More generally, the statistical insignificance of the SFAS No. 133 variables in both Tables 4 and 5 is consistent with recent research, which cautions that such fair value data may not be reliable for banks (e.g., Nissim, 2003). Similarly, recent studies of SFAS No. 133 disclosures note that compliance has been mixed (Jones and Wei, 2004), and conclude that the standard’s “desired level of financial transparency on the use of derivative financial instruments is not being adequately achieved” (Bhamornsiri and Schroeder, 2004, p. 680). In summary, our primary results indicate that the notional principal amounts of both trading derivatives (hypothesis 1) and non-trading derivatives (hypothesis 2) are useful in explaining variations in bank firm values. These primary tests take advantage of the complete time-series of disaggregated data under SFAS No. 119 and the rigorous specification of the Ohlson model framework, for a sample of 166 state and national banks of varying size. Moreover, evidence is provided that such support for hypotheses 1 and 2 is not sensitive to the inclusion of SFAS No. 133 fair value data. Our findings are consistent with prior studies such as Riffe (1997) and Venkatachlam (1996), both of whom conclude that the

424


notional amounts of derivatives contain value-relevant information for their sample of 242 banks and 99 banks, respectively. In contrast, Seow and Tam (2002) do not find notional amounts significant in the return models for their sample of 35 NYSE traded banks. We conjecture that such inconsistency could be due to their relatively small sample of top banks, the specification of the reporting variable (square root of notional amounts) or the model specification. As an additional robustness check, the dependent variable in the above primary analyses (stock price) is replaced with market value deflated by book value. Tables 6 and 7 detail the

Table 6. OLS regression—additional tests of hypothesis one (Trading Derivatives) MVit = a + b ∗ Independent Variableit + eit Independent variables

MODEL 1a Estimate P-value MODEL 2a Estimate P-value MODEL 3a Estimate P-value

Intercept C

E

SALEGROW TDER

0.93414 0.00000 0.97213 0.00000 0.97513 0.00000

6.80836 0.00000 6.64460 0.00000 6.62578 0.00000

0.00245 0.00000 0.00232 0.00000 0.00228 0.00000

(3.04454) 0.00341 (3.38888) 0.00102 (3.40272) 0.00094

TINT

TFX

GAINS

0.12614 0.00000

ADJ RSQ 0.424

0.06310 0.02586 0.06812 0.01760

0.62588 0.4445 0.00002 0.66499 (1.86786) 0.4452 0.00002 0.23559

Number of observations = 992. Additional tests of hypothesis one use pooled time-series cross-sectional OLS estimates of models 1a–3a, which parallel models 1–3 in Table 4. The dependent variable in Table 4 primary analysis is replaced with firm value deflated by book value in Table 6. Market value (MV) is price multiplied by the number of shares outstanding for firm i at end of year t, and is deflated by book value of equity of firm i at the end of year t (BV); and the intercept (C) is 1/BV. All other variables are as defined in Table 2 except that they are deflated by BV in this table. All p-values are based on White (1980) heteroscedastcity corrected t-values and two-tailed tests. Table 7. OLS regression—additional tests of hypothesis two (Non-trading derivatives) MVit = a + b ∗ Independent Variableit + eit Independent Variables

MODEL 5a Estimate P-value MODEL 6a Estimate P-value MODEL 7a Estimate P-value

Intercept C

E

SALEGROW NTDER NTINT NTFX NTA

0.9492 0.0000 0.9948 0.0000 1.0008 0.0000

7.0451 0.0000 6.6113 0.0000 6.6107 0.0000

0.0025 0.0000 0.0022 0.0000 0.0022 0.0000

(3.6728) 0.0005 (3.6224) 0.0004 (3.6841) 0.0003

0.1207 0.0014

NTL

ADJ RSQ 0.3957

0.0474 0.2418 0.0515 0.2107

5.0751 0.4338 0.0000 5.0358 (8.8354) (345.5000) 0.4335 0.0000 0.2486 0.0417

Number of observations = 992. Additional tests of hypothesis two use pooled time-series cross-sectional OLS estimates of models 5a–7a, which parallel models 5–7 in Table 5. The dependent variable in Table 5 primary analysis is replaced with firm value deflated by book value in Table 7. Market value (MV) is price multiplied by the number of shares outstanding for firm i at end of year t, and is deflated by book value of equity of firm i at the end of year t (BV); and intercept (C) is 1/BV. All other variables are as defined in Table 2, except that they are deflated by BV in this table. All p-values are based on White (1980) heteroscedastcity corrected t-values and two-tailed tests.


425

OLS regression results pertaining to these additional tests of hypotheses 1 and 2, respectively. The results are similar to the primary evidence presented in Tables 4 and 5. Thus, we conclude that the primary results are robust to alternative model specifications. 6.

Summary and suggestions for future research

This study investigates the value-relevance of banks’ expanded derivative disclosures under SFAS Nos. 119 and 133. In light of the existing literature gap on the information content of derivatives reporting (Seow and Tam, 2002), we focus our analysis on notional principal amounts. By incorporating a complete time-series of disaggregated SFAS No. 119 notional value data, we are able to construct more powerful tests than were possible prior to this standard. Using the rigorous specification of the Ohlson (1995) firm valuation framework, our results suggest that notional principal amounts of derivatives are economically significant and provide incremental information content beyond earnings and book value. Moreover, such results are robust to the inclusion of recently available SFAS No. 133 fair value data as well as to alternative model specifications. There are a number of limitations to this study. First, only two years of SFAS No. 133 fair value data were available, which might have reduced the power of the tests. Second, the study is limited to commercial banks. While our primary objective is to contribute to the empirical evidence of notional value information content for banks, the statistically insignificant results for our sample’s SFAS No. 133 fair value data are consistent with recent concerns regarding the complexity of this recent standard (e.g., Bhamornsiri and Schroeder, 2004) and, in particular, the reliability of banks’ fair value disclosures (e.g., Nissim, 2003). Future research can expand the scope of this study beyond banks, and take advantage of additional SFAS No. 133 data as it becomes available. Finally, the primary tests rely on the assumptions that underlie the modified Ohlson (1995) model. Future research can subject such assumptions to theoretical examination. Notes 1. Value-relevance, measured in terms of stock prices, refers to increases in firm value resulting from added disclosure by firms. 2. Dechow, Hutton and Sloan (1999) find that short-term forecasts of earnings contain value-relevant information, in empirical tests of Ohlson’s (1995) valuation model. However, because forecasts of earnings are a potentially biased measure (see, e.g., Abarbanell and Bernard, 1992; McNichols and O’Brien, 1997; Das, Levine and Sivaramakrishnan, 1998), we do not include earnings forecasts in our valuation models. 3. SALEGROW is calculated as ((SALE/SALE[−3])−1)∗ 100. Per Research Insight’s definition of item A12 for banks, SALE “. . . . . . 11. . . includes total current operating revenue and net pretax profit or loss on securities sold or redeemed.” 4. We thank an anonymous reviewer for drawing our attention to this trading behavior and its implications for analyses of SFAS No. 133 fair value gains and losses. 5. As noted in relation to the Table 3 descriptive statistics, model 4 results for the LOSSES variable are footnoted. With regard to model assessment issues, diagnostics used to assess OLS error-term assumptions include the Durbin-Watson test for autocorrelation, the White (1980) test for heteroscadasticity, and variance inflation factors to determine the linear independence among the explanatory variables. In total, the model diagnostics suggest that the OLS assumptions have been met.

426


6. The OLS estimated coefficients for the variables of interest (TINT and TFX) remain positive and statistically significant (at a.01 level), while the LOSSES variable is not significant ( p-value of 0.2333), after the White (1980) adjustment for heteroscadasticity.

References Abarbanell, J. and V. Bernard, “Tests of Analysts’ Overreaction/Underreaction to Earnings Information as an Explanation for Anomalous Stock Price Behavior.” Journal of Finance 47, 1181–1207, (1992). Bank for International Settlements, Statistical Annex (Washington, D.C.) BIS, (2004). Barth, M., “Fair Value Accounting: Evidence from Investment Securities and the Market Valuation of Banks.” The Accounting Review 69(1), 1–25, (1994). Barth, M., W. Beaver and W. Landsman, “Value-Relevance of Banks’ Fair Value Disclosures Under SFAS No. 107.” The Accounting Review 513–537, (1996). Barth, M., G. Clement, G. Foster and R. Kasznick, “Brand Values and Capital Market Valuation.” Review of Accounting Studies 3, 41–68, (1998). Bhamornsiri, S. and R. Schroeder, “The Disclosure of Information on Derivatives Under SFAS No. 133: Evidence from the Dow 30.” Managerial Auditing Journal 19, 669–680 (2004). Core, J., W. Guay and A. Buskirk, “Market Valuation in the New Economy: An Investigation of What has Changed.” Journal of Accounting and Economics 34, 43–67 (2003). Das, S., C. Levine and K. Sivaramakrishnan, “Earnings Predictability and Bias in Analysts’ Earnings Forecasts.” The Accounting Review 73, 277–294, (1998). Dechow, P. M., A. P. Hutton and R. G. Sloan, “An Empirical Assessment of the Residual Income Valuation.” Journal of Accounting and Economics 26, 1–34, (1999). Eccher, A., K. Ramesh and S. Thiagarajan, “Fair Value Disclosures of Bank Holding Companies.” Journal of Accounting and Economics 22, 79–117, (1996). Financial Accounting Standards Board, “Statement of Financial Accounting Standards No. 105, Disclosure of Off-balance Sheet Risk and Financial Instruments with Concentrations of Credit Risk.” Norwalk, Connecticut, FASB, 1990. Financial Accounting Standards Board, “Statement of Financial Accounting Standards No. 107, Disclosures about Fair Value of Financial Instruments.” Norwalk, Connecticut, FASB, 1991. Financial Accounting Standards Board, “Statement of Financial Accounting Standards No. 119, Disclosure about Derivative Financial Instruments and Fair Value of Financial Instruments.” Norwalk, Connecticut, FASB, 1994. Financial Accounting Standards Board, “Statement of Financial Accounting Standards No. 133, Accounting for Derivative Instruments and Hedging Activities.” Norwalk, Connecticut, FASB, 1998. Géczy, C., B. A. Minton and C. Schrand, “Why Firms Use Currency Derivatives.” Journal of Finance 52, 1323– 1354, (1997). Graham, J. R. and D. A. Rogers, “Do Firms Hedge in Response to Tax Incentives?” Journal of Finance 57, 815–839, (2002). Guay, W. and S. Kothari, “How Much do Firms Hedge With Derivatives?” Journal of Financial Economics 70, 423–461, (2003). Hagerty, J., “Fannie Warns of $9 Billion Loss if it Loses Ruling on Derivatives.” Wall Street Journal November 16, A3, (2004). Hawser, A., “Brought to Book.” Global Finance 18, 44–45, (2004). Hentschel, L. and S. Kothari, “Are Corporations Reducing or Taking Risks with Derivatives?” Journal of Financial and Quantitative Analysis 36, 93–118, (2001). Jones, S. and L. Wei, “Financial Accounting Standards Board rule 133 is the Rodney Dangerfield of bookkeeping regulation: It don’t Get no Respect.” Wall Street Journal December 3, C3, (2004). Kaplan, G. C., “Computer Model blamed for $83 million loss.” Wall Street Journal March 17, 25, (1997). Kawaller, I., “What Analysts Need to Know About Accounting for Derivatives.” Financial Analysts Journal 60, 24–29, (2004). Khurana, I. and M. Kim, “Relative Value-Relevance of Historical Costs vs. Fair Value: Evidence from Bank Holding Companies.” Journal of Accounting and Public Policy 22, 19–42, (2003).


427

McAnally, M. L., “Banks, Risk and FAS 105 Disclosures.” Journal of Accounting and Economics 11, 453–490, (1996). McNichols, M. and P. O’Brien, “Self-Selection and Analyst Coverage.” Journal of Accounting Research 35, 167–199, (1997). Mozes, H., “The Value-Relevance of Financial Institutions’ Fair Value Disclosures: A Study in the Difficulty of Linking Unrealized Gains and Losses to Equity Values.” ABACUS 38, 1–15, (2002). Nance, D., C. Smith and C. Smithson, “On the Determinants of Corporate Hedging.” Journal of Finance 48, 267–284, (1993). Nelson, K., “Fair Value Accounting for Commercial Banks: An Empirical Analysis of SFAS 107.” The Accounting Review 71, 161–182, (1996). Nissim, D., “Reliability of Banks’ Fair Value Disclosures for Loans.” Review of Quantitative Finance and Accounting 20, 355–384, (2003). Ohlson, J. A., “Earnings, Book Values and Dividends in Equity Valuation.” Contemporary Accounting Research 11 (2), 661–687, (1995). Office of the Comptroller of the Currency, OCC Bank Derivatives Report: Third Quarter. Washington, D.C., OCC, 2004. Richardson, G. and S. Tinaikar, “Accounting Based Valuation Models: What have We Learned?” Accounting and Finance 44, 223–255, (2004). Riffe, S., “The Valuation of Off-Balance Sheet Financial Instrument Disclosures in the Banking Industry,” In B. Schachter (ed.) Derivatives, Regulation and Banking: Advances in Finance, Investment and Banking Series. Amsterdam: North-Holland, pp. 93–121, 1997. Seow, G. and K. Tam, “The Usefulness of Derivative-Related Accounting Disclosures.” Review of Quantitative Finance and Accounting 18, 273–291, (2002). Simko, P., “Financial Instrument Fair Values and Non-Financial Firms.” Journal of Accounting, Auditing & Finance 14, 247–274, (1999). Skinner, D., “Are Disclosures About Bank Derivatives and Employee Stock Options ‘Value-Relevant’?” Journal of Accounting and Economics 22, 393–405, (1996). Springett, P., “The Barings Rescue: Apologies, but the Pressures have Become too Much to Bear.” The Guardian (London), March 7,(1995). Venkatachalam, M., “Value-Relevance of Banks’ Derivatives Disclosures.” Journal of Accounting and Economics 22, 327–355, (1996). White, H., “A Heteroscedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroscedasticity.” Econometrica 48, 817–838, (1980).