Week 5

The Economic Consequences of Banks’ Derivatives Use in Good Times and Bad Times

Ken B. Cyree & Pinghsun Huang & James T. Lindley

Received: 23 November 2009 /Revised: 28 March 2011 /Accepted: 29 March 2011 / Published online: 13 April 2011 # Springer Science+Business Media, LLC 2011

Abstract In a sample of 335 commercial banks, we do not detect a systematic effect on bank values from derivatives use in either the high growth period of 2003–2005 or the low growth period of 2007–2009. These findings apply to all types of derivatives including credit default swaps. Our results suggest that banks take a more balanced approach and restrict their derivative activities to providing derivative services for customers and risk management. We also find that the market disciplined banks significantly for taking TARP funds, indicating that receiving TARP funds was a signal that the banks were financially distressed. Lastly, we cannot discern valuation effects resulting from derivatives use even in large and poorly capitalized banks that are more likely to take risk-shifting opportunities. Collectively, we find no compelling evidence supporting the widespread allegation that derivatives use increased banks’ speculating behaviors and significantly contributed to the loss of value during the subprime mortgage crisis.

Keywords Credit default swaps . Derivatives . Deposit insurance . Subprime crisis

Jel Classification G30 . G32 . G34

“Derivative contracts—including credit defaults swaps—serve a useful function in mitigating risk and making the capital markets more efficient. They did not cause the crisis, but they did introduce greater interconnectedness as well as embedded and

J Financ Serv Res (2012) 41:121–144 DOI 10.1007/s10693-011-0106-y

K. B. Cyree School of Business Administration, University of Mississippi, 227 Holman Hall North, Oxford 38677 MS, USA e-mail: [email protected]

P. Huang (*) Department of Accountancy and Graduate Institute of Finance and Banking, National Cheng Kung University, Tainan 701, Taiwan e-mail: [email protected]

J. T. Lindley Department of Economics and Finance, University of Southern Mississippi, 314-H Joseph Greene Hall, Hattiesburg 39406 MS, USA e-mail: [email protected]

hidden leverage into financial institutions’ balance sheets, and, in many instances, they magnified the effects of other risks.” –Henry Paulson, Financial Crisis Inquiry Commission Hearing, May 6, 2010

1 Introduction

The causes of the most recent banking crisis are likely intermingled, with each cause being a contributing factor to the crisis and recession that followed. Brunnermeier (2009) describes many of the causes ranging from the originate-to-distribute model and securitization to leveraged borrowers’ balance sheets. Beltratti and Stulz (2009) report that banks with shareholder-friendly boards fared worse during the crisis. Allen and Carletti (2010) focus on the lack of liquidity during the crisis and the fear of contagion. A common thread in most studies of the crisis is that moral hazard led bank managers to take excessive risks of one kind or another since they do not bear all of the downside risk. Banks took risks that in bad economic times resulted in at least a partial government bailout in the form of direct investment, a deposit insurance fund guarantee or payout, or other forms such as government subsidized loans.

The popular press has reported extensively on the causes of the crisis. For example, a recent article in Time (Saporito, 2009) reports that self-interest or greed did not compel financial institutions to adopt adequate risk controls. During the crisis, American International Group (AIG) nearly collapsed because it had underwritten systemic risk by improperly trading credit default swaps (CDSs). The AIG episode indicates that the full- blown financial panic was spread to not only the banks that had overinvested in mortgage- backed securities, but also those that had written subprime-linked CDSs without adequate tools for risk management. The banking crisis was further aggravated by the opaque interconnectedness of large financial institutions emanating from interest rates, foreign exchange, and other derivatives (Greenberger, 2010; Paletta and Patterson 2010).

While banks’ derivatives use allegedly was linked to the credit crunch, Stulz (2009) notes that a dearth of serious empirical studies on the economic consequences of derivatives use makes it unclear if financial derivatives destabilized the banking system, which in turn led to the financial meltdown. To gain a better understanding of this controversy, we investigate the role of derivatives in the banking industry and their effects on valuation before and during the crisis. Rather than focus on the short-run wealth effects, we investigate whether or not derivatives (such as credit default swaps) led to differential performance before and during the crisis. We focus on the valuation effects of derivatives use by commercial banks prior to the crisis to ascertain whether or not commercial banks used these instruments to create value by leveraging government guarantees. Since CDSs and other derivatives are often blamed for the crisis, we focus on the effects of derivatives use during the crisis to assess whether banks using these derivatives experienced differential market performance compared to banks that did not use derivatives.

As a whole, the net economic consequences of banks’ derivatives utilization rest on two distinct motives: risk taking and risk insulating. Merton (1977) contends that insurance premiums that are insensitive to bank risk provide an incentive for depository institutions to increase the bank’s risk by exercising the deposit insurance put option. The risk-shifting incentives have been reinforced by implicit bailouts and subsidized loans from government.1 On the other hand, Keeley (1990) and Diamond (1991) postulate that the

1 This view implicitly assumes that either default risk remains constant, or that stockholders believe the market will respond favorably to the higher risk. The latter seems implausible in competitive markets.

122 J Financ Serv Res (2012) 41:121–144

potential loss of a valuable charter induces banks to limit their risk taking. Similarly, Park (1997) argues that value-maximizing banks take a more balanced approach by selecting the asset portfolios and the capital ratios that will reduce the probability of bankruptcy and limit the risk of losing their charter. While derivatives can insure against market-wide shocks in bad times thus protecting the charter value, they enable banks to exploit lucrative trading opportunities in periods of economic growth.

The asymmetric effects of derivatives use on shareholder value between good times and bad times provides a powerful setting to examine the process by which banks manage their financial derivatives. If banks utilize derivatives for speculation and trading purposes, we would expect that derivatives will trade at a premium in good times and a discount in bad times. To the extent that derivatives act as a precautionary hedge against downside risk, the use of derivatives by depository institutions will be more valuable in bad times than in good times. Conversely, banks can construct a well-diversified portfolio of derivative securities to hedge against financial risks while concurrently providing dealing services to their customers. Consequently, we would not observe a systematic valuation effect of derivatives use or detect its valuation differentially across economic upturns and downturns.

To test these predications, we collect a sample of 335 commercial banks with 1,746 bank-years in the period 2003 through 2009, but omit 2006 as a transition year. We split our sample into the expanding markets of 2003–2005 and the declining markets during 2007– 2009 (which coincided with the subprime mortgage crisis). Consistent with prior studies (e.g., Guay and Kothari, 2003), we use the gross notional principal of derivatives holdings to capture a bank’s derivatives position. Additionally, we use three measures to evaluate long- run performance: buy-and-hold returns, buy-and-hold abnormal returns based on size and book-to-market matched portfolios, and Sharpe ratios.

Not surprisingly, our univariate results reveal that banks market valuations are significantly better in good times than in bad times regardless of market or accounting performance measures. On average, banks experienced a higher probability of growth, had higher net interest margins, and carried substantially lower past due loans in good times than in bad times. There is no discernible difference in derivatives use between economic good times and bad times except for credit default swaps. Banks’ usage of credit default swaps increases during bad times, albeit in a relatively small magnitude. This finding seems intuitive since a series of bond defaults can trigger the demand of, and supply for, credit derivatives. A close analysis of our sample shows that CDSs are concentrated in large banks such as JP Morgan Chase, Citigroup, and Bank of America with more than 10% of their assets invested in CDSs, thus the economic impact of CDSs could be larger for large banks vis-à-vis small banks.

In a battery of tests, our estimates of the effects of derivatives on market valuation are not statistically distinguishable from zero in either good times or bad times. Our inference is not limited to interest rate and foreign currency derivatives, it also extends to credit default swaps. The latter finding is particularly important because the financial meltdown enables us to identify a common risk factor among banks and isolate the part of stock price movements that were attributable to the credit crunch. This in turn allows us to assess more precisely the economic consequences of credit derivatives in response to the subprime mortgage crisis event.

We also include a variable that measures direct government subsidies in the form of funding from the Troubled Asset Relief Program (TARP). The TARP program was created to buy assets, primarily mortgages and mortgage-based securities, which had little or no liquidity. In practice, the TARP program provided direct assistance to banks through preferred or common stock purchases, and therefore obtaining TARP funds either directly or indirectly was a bailout for the bank. We find that TARP recipients underperformed non-

J Financ Serv Res (2012) 41:121–144 123

TARP recipients. This result is consistent with the premise that the market disciplined banks that had relied on government funds as a partial bailout, thus rendering risk taking more expensive for banks to pursue.

To the extent that the moral hazard problem is more severe for banks that are poorly capitalized or are too big to fail, the economic performance of derivatives use in bad times will be worse for big banks or poorly capitalized banks than for their counterparts that are smaller or better capitalized. We find that the use of derivatives in small and poorly capitalized banks do not measurably influence bank value, supporting our previous findings. Taken together, our results lend no credence to the speculation that derivatives’ use facilitated banks’ speculating behavior nor did it contribute significantly to the credit crunch. Rather, it appears that banks took a more balanced approach of mixing their derivative instruments between dealing and risk management activities.

In Section 2 below, we review prior research and develop testable hypothesis, and in Section 3 we describe the data and empirical models. In Section 4, we present the empirical results, and Section 5 contains the summary and concludes the paper.

2 Prior research and hypothesis development

There are numerous studies that investigate the wealth effect of corporate derivatives use. Allayannis and Weston (2001) document that the use of foreign currency derivatives adds nearly five percent to market value, on average, to industrial firms. Similarly, Carter, Rogers, and Simkins (2006) report that the premium generated from the use of jet fuel derivatives is as large as 10% for the US airline companies. In contrast, Guay and Kothari (2003) and Adam and Fernado (2006) show that derivatives yield, at best, provide moderate benefits for nonfinancial firms. Moreover, Jin and Jorion (2006) find that financial derivatives have no influence on the market value of oil and gas producers. Thus, the effects of derivative contracts on market valuation are mixed and still open to debate.

While much attention has been paid to the role of derivatives in industrial firms, extant research has overlooked the economic consequences of derivatives use in the banking sector. However, banks’ use of derivatives has been in the spotlight since the outbreak of the catastrophic credit crunch. For instance, Gorton (2008) asserts that the financial panic was attributable to the opaque nature of derivatives. A recent Wall Street Journal article (Paletta and Patterson 2010) also cites Capital Hill’s concern that the lack of prudent government oversight may have led some banks to undertake derivatives predominately for potentially lucrative trading, thus fueling the economic turmoil. Equivalently, profit- maximizing behavior by banks may have created systemic risk. There is some evidence that derivative instruments are heavily used merely in a small number of large banks (Singh and Aitken, 2009; Fratianni and Marchionne, 2009). Taken together, moral hazard could be the driving force of some banks’ uncharacteristically large derivative holdings. However, it remains an empirical question of whether the reckless use of derivatives was rampant in the average bank. That is, did derivatives use in the average bank at least partially contribute to the disastrous financial turmoil?

It is important to note that banks might not utilize derivatives exactly in the same manner as non-banking firms. In contrast to industrial firms, banks have an institutional incentive to increase risks to exploit the moral hazard frictions embedded in government bailouts or deposit insurance (Merton, 1977; Calomiris and Mason, 2003). This view manifests itself in the housing bubble in that value-maximizing banks were prone to lend imprudently to subprime borrowers who had poor credit records. The recent bailouts of AIG, Bear Stearns, and

124 J Financ Serv Res (2012) 41:121–144

Goldman Sachs are examples of the “Bernanke Put,” which allegedly propelled financial institutions to originate and distribute mortgage-based securities to recklessly take advantage of the risk-shifting opportunities (Saporito, 2009). The incentive to increase risk intensifies when banking firms become financially distressed. Supporting this perspective, Kane (1989) and Cole (1993) find that perverse incentives created by deposit insurance resulted in a large number of the thrift and bank failures of the 1980s and early 1990s.2 Laeven and Levine (2009) also demonstrate that banks with more powerful owners are more likely to take risks, underscoring banks’ incentives to maximize shareholder value by increasing risk exposures.

Similar to non-banking firms, banks also have an incentive to protect themselves against a string of catastrophic events.3 As Kahane (1977) points out, a financial intermediary is more likely to minimize the variance of earnings than to exploit the insurance subsidy because low variance in earnings reduces the likelihood of being classified as a risky intermediary by regulators. Similarly, Buser, Chen, and Kane (1981) contend that a wealth- maximizing bank tends to operate in a low-risk fashion that closely mimics the FDIC’s regulatory norm to protect the bank’s charter value and other intangible capital. Moreover, banks have an incentive to lower risks and avoid herding so that they have sufficient liquidity to finance investment or buy distressed banks at low prices in a financial downturn (Acharya and Yorulmazer, 2008; Shleifer and Vishny, 2010). This risk-sharing notion, commonly referred to as the charter value paradigm in Saunders and Wilson (2001), is also consistent with Froot et al.’s (1993) assertion that hedging attenuates the underinvestment problem.

As a whole, given the conventional wisdom about the incentives created by deposit insurance and bailouts, banks could have a stronger incentive than non-banking firms to employ derivatives for potentially profitable trading. However, increasing leverage by means of derivatives can aggravate the risk of bankruptcy and the loss of the charter. Thus, banks are expected to trade off between the marginal benefit of increasing leverage with their derivatives operations and the marginal cost of losing their charter.

There is limited empirical evidence on the determinants of the use of derivatives by banks. Carter and Sinkey (1998) report that depository institutions facing more interest rate risk engage in more derivatives use, indicating that banks tend to reduce the likelihood of failure, presumably in the best interests of shareholders. Conversely, Whidbee and Wohar (1999) find that banks with larger insider holdings are less likely to hedge, suggesting that managers maximize shareholder value by not hedging. Moreover, Minton, Stulz, and Williamson (2009) show that a small number of banks use credit derivatives, and most of their gross positions are not for risk management but for dealer activities.

Prior empirical work on the valuation effect of banks’ derivatives holdings is sparse and tends to focus on the period prior to the subprime mortgage crisis. Sinkey and Carter (1999) find banks that held derivatives suffered negative abnormal returns around the

2 See Strunk and Case (1988) for a long list of factors that also contributed to the thrift failures. Key issues were the legal limits on interest paid on deposits by Savings and Loans and their undiversified portfolios of long term mortgages which had been mandated by law, in conjunction with rapidly rising interest rates caused by inflation. 3 Smith and Stulz (1985) argue that hedging reduces the probability of incurring bankruptcy costs and therefore benefits shareholders. The transaction costs of financial distress can stem from direct bankruptcy costs such as legal fees and accounting expenses in the bankruptcy or indirect bankruptcy costs like business disruption and loss of reputation. To the extent that corporate risk management reduces the likelihood of financial distress, it can add value. Froot, Scharfstein, and Stein (1993) also propose that costly access to external financing makes corporate hedging a value-enhancing strategy. When external financing is sufficiently expensive and internally generated cash flows are not sufficient to fund growth opportunities, the underinvestment problem arises. Moreover, Leland (1998) and Graham and Rogers (2002) show that firms’ use of derivatives can increase debt capacity and interest deductions by lowering the volatility of earnings.

J Financ Serv Res (2012) 41:121–144 125

announcement of large losses by Bankers’ Trust. The market reaction was negative for banks that had been derivatives dealers as well as for those banks that had used derivatives to hedge and speculate. In contrast, Brewer, Jackson, and Moser (Brewer et al. 2001) report that banks using interest rate derivatives in the period 1986–1994 did not experience significantly different ROA or ROE. Their results indicate that there was little effect from derivatives usage on accounting profits during this period. Overall, empirical evidence on this matter is largely inconclusive.

While these authors provide insights into the economic consequences of banks’ use of derivatives, they do not assess whether the market valuation of banks’ derivatives use varies widely in response to macroeconomic shocks. This research question is especially important because cyclical fluctuations in bank performance hinge largely on the intermediary’s appetite for risk (Rajan, 2006; Adrian and Shin, 2008; Shleifer and Vishny, 2010). Thus, tracing a potential link from the use of derivatives by banks to stock return performance around the time of the credit crunch provides important implications for the role of derivatives in the recent financial chaos. Moreover, the market reaction to the government’s injection of liquidity through the TARP in the immediate aftermath of the subprime crisis can shed light on the impact of external discipline on banks’ risk-shifting behavior.

To the extent that financial intermediaries exploit derivative contracts such as credit default swaps to take more leveraged bets when the economy is in an upswing (or perhaps a bubble), the economic consequences of financial derivatives will be procyclical. It will be even more so when measuring the reward to volatility trade-off, because economic uncertainty in bad times tends to magnify the fluctuation of share price (Adrian and Shin, 2008). In this circumstance, derivatives tend to move in tandem with the overall economy. In contrast, if banks use derivatives mainly to reduce the probability of “left-tail” outcomes, banks with derivatives should fare better in the time of crisis as compared to those without derivatives. Conversely, derivatives use could erode the value of bank shares during good times since the enormous costs in managing a multitude of instruments can outweigh the limited benefits from absorbing tail risks. Equivalently, if derivatives serve banks’ risk- insulating functions, their financial effects should be countercyclical. In the event of a credit crunch, such phenomena, if any, should be more pronounced with credit default swaps than with other derivatives due to a spike in defaults. On balance, stock performance is expected to be insensitive to derivative operations if banks hedge most of their exposures resulting from the dealing activities in derivative products.

3 Data and empirical models

3.1 A. Sample selection and data sources

The data for our study are from several different sources. We obtain data for all banks for the years 2003 through 2009 from the FR Y-9 reports for bank holding companies and the Call Reports for banks that are not holding companies. Both data sources provide information on the notional principal of derivatives that are classified as interest rate, foreign exchange, credit derivatives including credit default swaps, and other derivatives.

We obtain market values for banks, monthly stock prices, and price indices from the Center for Research in Securities Prices (CRSP). We have data from 2003 through 2009 for all publicly traded banks or bank holding companies. For the model using year-end accounting data, the sample contains almost 900 bank-year pooled observations for the pre- crash period of 2003–2005, and 850 for the period 2007–2009 during the crash.

126 J Financ Serv Res (2012) 41:121–144

3.2 B. Estimation methodology for multivariate analysis

We use multivariate analysis to examine the effect of derivatives use on bank value by modeling bank value as a function of bank characteristics. Our regression models become:

Yit ¼ g0 þ g1Xit þ g2Dit þ "it ð1Þ where Xit is a set of exogenous observable bank characteristics, Dit is a derivatives use variable, g ¼ g0; g1; g2f g is a vector of parameters to be estimated, and εit is an error term.

We define the dependent variable Yit in several ways. In particular, our performance measures include buy-and-hold returns (BHR), buy-and-hold abnormal returns (BHAR), and Sharpe ratios (SHARPE). They are calculated as follows:

BHRi;y ¼ Y12

t¼1 1 þ ri;t � �

� 1 ð2Þ

BHARi;y ¼ Y12

t¼1 1 þ ri;t � �

� Y12

t¼1 1 þ rMi;t

� � ð3Þ

SHARPEi;y ¼ ri;t � rf ;t SDi;t

» ffiffiffiffiffi 12

p ð4Þ

where ri,t denotes the raw return on stock i in month t,r M i;t is the size and book-to-market

matched portfolio return over month t (Fama and French, 1993), BHARi,y is the abnormal return for holding bank i stock over year y, SHARPEi,y represents bank i’s annualized Sharpe ratio, ri;tis the simple average of monthly returns for the one-year period, rf ;t is the average of one-month T-bill returns over the year, and SDi,t is the standard deviation of the monthly returns in year y.

We use several control variables in our model that previous research has shown to be related to derivative use by banks (e.g., Sinkey and Carter, 2000). However, these same variables are correlated with the value of the bank. Consequently, Dit will be correlated with εit in Eq. (1). The OLS estimate of +2 will be biased due to a failure to control for those bank characteristics that result in banks taking derivatives positions. That is, are banks different because they use derivatives, or are banks different because they engage in economic activities whose risk leads to derivative use?4

Because of this endogeneity problem, we utilize Heckman’s (1979) two-stage procedure to control for the self-selection of banks that enter into derivatives contracts. We first estimate the following equation to explain the decision to use derivatives:

D »

it ¼ d

Y

it

þ mit Dit ¼ 1 if D»it > 0 Dit ¼ 0 if D

»

it < 0;

ð5Þ

where D » it is an unobserved latent variable, Πit is a set of bank characteristics that could

affect the decision to use derivatives, and μit is an error term. We use a probit regression

4 Stated differently, do banks have different production activities and thus use derivatives, or is the use of derivatives the only difference between the banks?

J Financ Serv Res (2012) 41:121–144 127

on the full sample of 1,746 bank-years to estimate Eq. (5). We then obtain d̂ Q

it, the fitted value of the probit regression index function, and calculate inverse Mills ratios

(i.e., 4ðd̂

Q it Þ

6ðd̂ Q

it Þ ), where Ω is the standard normal density and Φ is the normal cumulative

probability. We include the inverse Mills ratios to correct for potential self-selection bias in equation (1).

The vector Πit in Eq. (5) contains the control variables that previous research has found to be related to a bank’s decision to use derivatives (Carter and Sinkey, 1998; Whidbee and Wohar, 1999; Sinkey and Carter, 2000) and those that affect bank market valuations. We include the natural logarithm of total assets (LNASSETS) as a control variable for size since Lang and Stulz (1994) report that size can be related to accounting profitability. Large firms are also expected to use derivatives more extensively because of an economy of scale in risk management (Nance, Smith, and Smithson, 1993). The variable INSPCT is defined as the percentage of insider equity holdings of shares outstanding. INSPCT accounts for the variations in agency costs and managerial entrenchment identified by Gorton and Rosen (1995) and because concentrated ownership can have both beneficial and detrimental impacts on shareholders (Morck, Shleifer, and Vishny, 1988). A capital ratio variable (CAPRATIO) is included because there is an inverse relation between the capital-to-asset ratio and the probability of financial distress (Jensen and Meckling, 1976).

We also account for a variety of bank risks by including the variables GROWTH, GAP12, PD902TA, OBSL, NIM, EQISSUE, and LOAN2DEP. GROWTH is defined as the average growth rate in assets over the last three years, and we add it as a control variable based on the finding of Smith and Watts (1992) that firm value is related to investment growth. GAP12 is the 12-month maturity re-pricing gap divided by assets as a measure of interest rate risk. We define PD902TA as the percentage of loans in assets that are past due for more than 90 days as a measure of on-balance sheet credit risk. OBSL is off-balance sheet liabilities scaled by assets. We measure NIM by net interest income less net interest expense scaled by earning assets. EQISSUE is the amount of new equity issued, if any, normalized by assets, and LOAN2DEP is the loan-to-deposit ratio. ROA and ROE are included as variables in Table 1, and they are significantly lower during the crisis period. These accounting performance measures are highly correlated with the market performance regressors, so they are not used as variables in the remaining tables.

The bank-specific variables of interest in the vector Dit include indicator variables taking the value of one if the bank is involved in Interest Rate, Foreign Exchange, Other, and Credit Default Swap derivatives activities, respectively, and zero, otherwise. The dummy variables are INTDUM, FXDUM, OTHERDUM, and CDSDUM for Interest Rate, Foreign Exchange, Other, and CDS derivatives uses, respectively. While investigating the nature of derivatives activities in commercial banking, we alternatively replace the indicators with notional values scaled by assets size (INTSIZE, FXSIZE, OTHERSIZE, and CDSSIZE) to identify the types and scale of derivatives use by banks and the possible relation to market valuations before and during the financial crisis.5

As an initial test, we explore whether the marginal value of derivatives use largely depends on the state of the economy. Our intuition is that if banks hedge all their credit exposures as well as interest rate and foreign currency risks in which they might not have a comparative advantage to manage, derivatives use is more likely to be beneficial to

5 Another specification is to include interaction terms of the hedging dummy variable with other cross sectional variables. When we attempted the interactions, there was severe multicollinearity and these variables had to be dropped from the regressions.

128 J Financ Serv Res (2012) 41:121–144

shareholders in an economic bust than in an economic boom. Alternatively, if banks are inclined to exploit their derivative trading operations, the economic consequences of banks’ derivatives use should be worse in a down economy than a growing economy. If banks utilize derivatives in an attempt to maximize shareholder wealth and ensure safety-and- soundness practices simultaneously, we would not observe significant effects during different economic conditions.

It seems plausible that bank performance could be highly associated with the governmental injection of liquidity into the banking sector in the aftermath of the financial chaos (Fahlenbrach and Stulz, 2011). This sort of assistance appeared to be especially common during the credit crunch period. In particular, the Troubled Asset Relief Program, also known as TARP, was part of the US government’s rescue efforts to purchase impaired assets and equity from financial intermediaries in trouble. To address this possibility, we alternatively add to equation (1) an indicator variable that equals one if the bank is a TARP

Table 1 Means and t-tests for differences in means. In the table, BHR is buy-and-hold return, BHAR is buy- and-hold abnormal return using control firms matched on size and book-to-market ratio, and SHARPE represents annualized Sharpe ratio. ROA is return on assets, and ROE is return on equity. INTSIZE is the notional amount of interest rate derivatives divided by assets; FXSIZE is the notional amount of foreign exchange derivatives divided by assets; OTHERDUM is the notional amount of “other” derivatives divided by assets; CDSSIZE is the notional amount of credit default swaps divided by assets. INSPCT is insider holdings as a percentage of total shares outstanding. LNASSETS is the natural logarithm of total assets. CAPRATIO is the equity-to-assets ratio; GROWTH is the average growth rate in assets for the prior three years; GAP12 is 12-month repricing gap scaled by assets; NIM is net interest margin; PD902TA is loans more than 90 days past due scaled by assets; OBSL is the ratio of off-balance sheet liabilities to assets; LOAN2DEP is the loan-to-deposit ratio

Variable Pre-Crash (2003–2005) Mean (N=896)

Crash (2007–2009) Mean (N=850)

t-stat for difference in group means

BHR 0.1931 −0.2381 30.35b

BHAR −0.0489 −0.1630 7.77b

SHARPE 0.8836 −0.6773 30.54b

ROA 0.0110 0.0000 14.64b

ROE 0.1213 −0.1643 3.96b

INTSIZE 0.2932 0.3169 −0.19 FXSIZE 0.0425 0.0503 −0.45 OTHERSIZE 0.0070 0.0132 −1.43 CDSIZE 0.0001 0.0019 −2.38a

INSPCT 0.0590 0.0442 2.53a

LNASSETS 14.4912 14.9443 −6.06b

CAPRATIO 0.0912 0.0934 −1.09 GROWTH 0.1374 0.0857 6.39b

GAP12 0.1715 0.1421 4.78b

NIM 0.0383 0.0362 5.64b

PD902TA 0.0047 0.0199 −18.81b

OBSL 0.1927 0.1760 2.93b

LOAN2DEP 0.9932 1.0012 −0.18 EQISSUE 0.0026 0.0016 2.88b

a significant at the five-percent level b significant at the one-percent level

J Financ Serv Res (2012) 41:121–144 129

recipient in the year, and zero, otherwise. The market’s reaction to the granting of TARP funds is also expected to provide insights into the efficacy of market discipline in reshaping the bank’s profile of risk and return.

4 Empirical results

In this section, we summarize the univariate comparisons to shed light on the differences in banks’ risk and performance across the sample periods. In further analysis, we detail multivariate regression results before and during the financial crisis to account for the confounding effects.

4.1 A. Univariate results

Table 1 contains a comparison of means of selected variables for banks in the pre-crash period of 2003–2005 versus the crash period of 2007–2009. The year 2006 is omitted as a transition year. The five performance measures show significantly worse performance during the crash compared to the pre-crash period. BHR, BHAR, SHARPE, ROA, and ROE are all lower, and all the differences across the periods are significant. Lower performance is not surprising since the financial problems beginning in 2007 are some of the worst since the Great Depression. Strikingly, the buy-and-hold returns across the two periods show a difference of more than 43 percentage points. Other measures are not much better with ROE falling more than 28 percentage points, and ROA going to zero for the industry in the three-year period.

Insider holdings declined, but part of this decline could be due to the significant increase in asset size. Surprisingly, the average capital ratio did not decline during the crash, at least on average. Interest rate risk, as measured by GAP, net interest margins, and off-balance sheet liabilities fell. The proportion of past due loans increased by a factor of more than four, indicating the distress facing borrowers during the crisis. Although the proportions of derivatives holdings in all types increased, only CDSs showed a statistically significant increase. As illustrated in Table 2, the increase in CDSs was mostly attributable to a few banks. In fact, JP Morgan/Chase went from 2.42% in 2003 to more than 48% CDS notional value as a percentage of total assets in 2008.

In summary, the univariate results show that banks performed poorly, but there is little evidence that the poor performance was sparked by derivatives use, at least on average and over a longer time period than events surrounding the beginning of the crisis.

B. Probit Selection Model Results

Table 3 presents the results from the probit model in the first stage of the regression models. Since the dependent variable is one if the bank uses derivatives, and zero, otherwise, a positive coefficient implies a larger likelihood of derivatives use. We find that larger banks, those with lower percentages of insider holdings, lower growth, larger amounts of off-balance-sheet liabilities, and lower NIM are more likely to use derivatives.

C. Multivariate Regression Results

We present our main test results across good times and bad times in Tables 4, 5, 6 and 7 to provide insights into how depository institutions have exploited derivatives. Panel A in Table 4 contains the results from the performance regressions for the pre-crash period using

130 J Financ Serv Res (2012) 41:121–144

indicator variables for derivatives use. In general, banks that use derivatives have no statistical differences from those that do not, with the exception of Other Derivatives use with BHAR as the performance measure at the conventional significance level. It appears that, at least on average, banks were not being rewarded or punished by market participants for using derivatives in this period.

Table 2 Percentage of Notional Value of Credit Default Swaps (CDSs) in Total Assets

Ticker Bank Name YEAR CDS/Assets Asset Size (000’s)

JPM JP Morgan Chase & Co. 2008 48.59% 2,175,052,000

C Citigroup 2008 22.29% 1,938,470,000

JPM JP Morgan Chase & Co. 2007 16.59% 1,562,147,000

JPM JP Morgan Chase & Co. 2009 16.52% 2,031,989,000

BAC Bank of America Corp. (DE) 2008 13.92% 1,822,068,028

BAC Bank of America Corp. (DE) 2009 13.35% 2,224,539,279

C Citigroup 2009 9.44% 1,856,646,000

C Citigroup 2007 7.67% 2,187,631,000

BAC Bank of America Corp. (DE) 2007 6.16% 1,720,688,423

JPM JP Morgan Chase & Co. 2005 3.52% 1,198,942,000

WFC Wells Fargo & Co. 2008 3.44% 1,309,639,000

JPM JP Morgan Chase & Co. 2003 2.42% 770,912,000

JPM JP Morgan Chase & Co. 2004 2.29% 1,157,248,000

WFC Wells Fargo & Co. 2009 1.95% 1,243,646,000

C Citigroup 2005 1.16% 1,494,037,000

BAC Bank of America Corp. (DE) 2005 0.79% 1,294,312,241

C Citigroup 2004 0.64% 1,484,101,000

BAC Bank of America Corp. (DE) 2004 0.41% 1,112,035,486

KEY KEYCORP NEW 2008 0.34% 105,231,004

BAC Bank of America Corp. (DE) 2003 0.34% 736,487,404

C Citigroup 2003 0.33% 1,264,032,000

STI SunTrust Banks Inc. 2008 0.26% 189,137,961

KEY KEYCORP NEW 2007 0.13% 99,567,393

PNC P N C BANK CORP 2008 0.13% 291,092,876

KEY KEYCORP NEW 2009 0.11% 93,381,546

PNC P N C BANK CORP 2007 0.11% 138,976,249

KEY KEYCORP NEW 2005 0.09% 92,844,997

KEY KEYCORP NEW 2004 0.08% 90,653,059

CMA COMERICA INC 2008 0.07% 67,912,580

NTRS NORTHERN TRUST CORP 2008 0.05% 82,053,626

CMA COMERICA INC 2009 0.05% 59,398,720

PNC P N C BANK CORP 2009 0.04% 269,921,958

STI SunTrust Banks Inc. 2005 0.04% 179,712,841

WFC Wells Fargo & Co. 2003 0.02% 387,798,000

STI SunTrust Banks Inc. 2007 0.02% 179,573,933

WFC Wells Fargo & Co. 2004 0.02% 427,849,000

PHBK PEOPLES HERITAGE FINL GROUP INC 2008 0.02% 122,745,454

J Financ Serv Res (2012) 41:121–144 131

The control variables in Table 4 indicate that banks with higher growth in assets have higher performance regardless of the measure. Banks with higher on-balance-sheet interest rate risk, as measured by 12-month GAP, have better market performance. Banks with higher net interest margin fare relatively well, as do banks with lower past due loans indicating that those banking firms that are successful at traditional bank lending activities are rewarded by the market. The other control variables are largely insignificant.

In Panel B of Table 4, we add a binary variable that equals one if the bank received TARP funds during the crisis, and zero otherwise, to test the sensitivity of our findings to a potential model misspecification. This additional control is important because it is a measure of whether or not our model has properly captured the known risks during the pre- crash period. If the TARP variable is significant in the pre-crash period, there might be uncaptured risk factors that are deemed important but have been left out from the set of our explanatory variables. As expected, the TARP variable is not significant in the pre-crash period, and little of the results have changed when adding the TARP variable. Thus, our results are not driven by bias attributable to omitted risk regressors.

To account for the possibility that the size of derivatives use could play a significant role in the market valuation, we alternatively use the notional amount of derivatives normalized by total assets as in Huang, Zhang, Deis, and Moffitt (2009). Similar to Table 4, Table 5 indicates that there is no statistical significance for our derivative measures in these performance regressions. That is, market participants did not significantly price derivatives use, at least on average, for banks during the pre-crash period. The lack of a reliable effect on performance by the size of derivatives use also holds in Panel B when the TARP variable is added. Consistent with Table 4, the TARP variable is insignificant. We again find faster growing banks and those with higher net interest margins, higher GAP, and lower past-due loans have higher performance.

In Panel A of Table 6, we focus on the performance regression results for the crash period (2007–2009) using indicator variables for derivatives use. While banks using interest

Variable Coefficient p-value

INTERCEPT −7.6227 <.0001 YR04 −0.0161 0.8915 YR05 0.0966 0.4095

YR06 0.1251 0.2950

YR07 0.1658 0.1724

YR08 0.1678 0.1801

YR09 0.1870 0.1570

LNASSETS 0.5432 <.0001

INSPCT −1.0478 <.0001 CAPRATIO 0.2229 0.7792

GROWTH −0.7001 0.0004 GAP12 −0.4271 0.1045 PD902TA −0.8962 0.6852 OBSL 1.6097 0.0003

NIM −16.1296 0.0002 LOAN2DEP 0.2228 0.2423

Likelihood Index

Table 3 Probit estimates for derivatives use. The dependent variable is one if the bank uses derivatives, and zero otherwise. YRXX is an indicator variable for year XX, and zero, otherwise. LNASSETS is the natural logarithm of total assets; INSPCT is insider holdings as a percentage of total shares outstanding; CAPRATIO is the equity-to-assets ratio; GROWTH is the average growth rate in assets for the prior three years; GAP12 is 12-month repricing gap scaled by assets; PD902TA is loans more than 90 days past due scaled by assets; OBSL is off-balance sheet liabilities to assets; NIM is net interest margin; LOAN2DEP is the loan- to-deposit ratio

N=1746

132 J Financ Serv Res (2012) 41:121–144

Table 4 Pre-crash performance regressions from 2003 through 2005 with indicator variables for derivatives type. The dependent variable uses several alternative measures of market performance: BHR is buy-and-hold return, BHAR is buy-and-hold abnormal return using control firms matched on size and book-to-market ratios, and SHARPE represents annualized Sharpe ratio. YR04 and YR05 are indicator variables equal to one in 2004 and 2005, respectively, and zero, otherwise. TARPBANK is an indicator variable equal to one if the bank received Troubled Asset Relief Program funds, and zero, otherwise. INTDUM is an indicator for using interest rate derivatives, and zero, otherwise; FXDUM is an indicator for foreign exchange derivatives use, and zero, otherwise; OTHERDUM is an indicator for using “other” derivatives; CDSDUM is an indicator for using credit default swaps. INSPCT is insider holdings as a percentage of total shares outstanding; LNASSETS is the natural logarithm of total assets; CAPRATIO is the equity-to-assets ratio; GROWTH is the average growth rate in assets for the prior three years; GAP12 is 12-month repricing gap scaled by assets; NIM is net interest margin; PD902TA is loans more than 90 days past due scaled by assets; OBSL is the ratio of off-balance sheet liabilities to assets; LOAN2DEP is the loan-to-deposit ratio. EQISSUE is the amount of new equity issued in a given year, divided by assets. LAMBDA is the inverse Mill’s Ratio from the first- stage Heckman correction Probit model

Variable BHAR BHR SHARPE

Estimate p-value Estimate p-value Estimate p-value

Panel A: Baseline Model

INTERCEPT −0.2014 0.0809 0.5778 <.0001 1.7870 0.0012 YR04 0.0921 <.0001 −0.2484 <.0001 −0.9005 <.0001 YR05 0.0385 0.0081 −0.3963 <.0001 −1.6431 <.0001 INTDUM 0.0637 0.1298 0.0791 0.1207 0.2274 0.2574

FXDUM −0.0103 0.6880 −0.0041 0.8944 0.0062 0.9594 OTHERDUM 0.0612 0.0463 0.0660 0.0761 0.2797 0.0565

CDSDUM −0.0362 0.3964 −0.0250 0.6289 −0.2145 0.2925 INSPCT 0.0163 0.7236 0.0156 0.7799 0.2326 0.2891

LNASSETS −0.0066 0.4713 −0.0287 0.0097 −0.0624 0.1526 CAPRATIO 0.0224 0.8853 0.0249 0.8948 0.8249 0.2656

GROWTH 0.2840 <.0001 0.3279 <.0001 0.9961 <.0001

GAP12 0.1652 0.0002 0.1459 0.0071 0.5918 0.0056

NIM 2.6012 0.0025 3.5061 0.0008 10.2393 0.0127

PD902TA −2.7072 0.0131 −3.3307 0.0118 −17.3143 0.0009 OBSL 0.0523 0.3982 0.0775 0.3017 0.4088 0.1667

LOAN2DEP 0.0070 0.1735 0.0051 0.4131 0.0359 0.1413

EQISSUE −0.5676 0.3919 −0.3537 0.6596 −0.9803 0.7567 LAMBDA −0.0358 0.1638 −0.0458 0.1414 −0.1427 0.2451 Adj. R-sq. 0.1444 0.3934 0.4056

Panel B: Baseline Model with TARP funds

INTERCEPT −0.2163 0.0620 0.5653 <.0001 1.7260 0.0019 YR04 0.0917 <.0001 −0.2488 <.0001 −0.9023 <.0001 YR05 0.0380 0.0091 −0.3968 <.0001 −1.6454 <.0001 TARPBANK −0.0173 0.1796 −0.0146 0.3498 −0.0709 0.2488 INTDUM 0.0600 0.1545 0.0760 0.1370 0.2121 0.2917

FXDUM −0.0111 0.6645 −0.0048 0.8767 0.0029 0.9814 OTHERDUM 0.0600 0.0508 0.0650 0.0809 0.2747 0.0611

CDSDUM −0.0345 0.4190 −0.0236 0.6490 −0.2075 0.3086 INSPCT 0.0199 0.6662 0.0186 0.7388 0.2474 0.2604

LNASSETS −0.0046 0.6221 −0.0270 0.0163 −0.0542 0.2207

J Financ Serv Res (2012) 41:121–144 133

rate derivatives had higher performance for all three performance measures, their corresponding coefficients are only marginally significant. Similar to what we find in the pre-crash period, we document banks that use Other Derivatives have higher performance, and in this case all three coefficients in the performance regressions are significant at the five-percent level or better. The uses of credit default swaps and foreign currency derivatives, however, do not affect bank value. Our control variables in Table 6 indicate that faster growth is related to higher market performance, and past due loans to assets is also strongly significant and negative. Collectively, these results suggest that market participants, on average, were more concerned with bank lending portfolios, especially the credit quality of those portfolios, than with derivatives use.

Panel B of Table 6 contains the performance regressions with the addition of a variable for receiving TARP funds. The TARP variable is negative and significant, vindicating the market belief that banks receiving TARP funds were either riskier or would have larger write-downs in the future.6 This evidence also supports the premise that the market disciplined banks during the distressed period for taking a federal bailout. Although higher performance remains for users of Other Derivatives in this model, the Interest Rate derivative use variable is statistically significant at only the ten-percent level. Control variables are largely consistent with Panel A.

In Table 7, we test whether our prior inferences are robust to a continuous measure of derivatives use. In both panels of Table 7, we find no evidence that derivatives use of any type has any bearing on banks’ performance at the conventional statistical level. While it is likely that banks using derivatives had differential market performance in short event windows around the failure of Lehman Brothers, the conversion of investment banks to commercial banks, and other events during the crisis, our results continue to indicate that a systematic relation between derivatives and market performance is generally absent. The TARP variable is again negative and significant in Panel B, reflecting the lower risk-

6 Note that riskier banks as indicated by receiving TARP funds would have future cash-flows discounted at a higher rate, and therefore their prices would fall, creating negative returns.

Table 4 (continued)

Variable BHAR BHR SHARPE

Estimate p-value Estimate p-value Estimate p-value

CAPRATIO −0.0046 0.9768 0.0021 0.9911 0.7144 0.3388 GROWTH 0.2838 <.0001 0.3278 <.0001 0.9953 <.0001

GAP12 0.1648 0.0002 0.1456 0.0073 0.5902 0.0057

NIM 2.5002 0.0038 3.4208 0.0011 9.8246 0.0172

PD902TA −2.7374 0.0121 −3.3562 0.0112 −17.4383 0.0008 OBSL 0.0563 0.3637 0.0808 0.2818 0.4251 0.1508

LOAN2DEP 0.0063 0.2180 0.0045 0.4664 0.0333 0.1743

EQISSUE −0.5173 0.4357 −0.3113 0.6987 −0.7740 0.8070 LAMBDA −0.0333 0.1966 −0.0437 0.1620 −0.1323 0.2823 Adj. R-sq. 0.1452 0.3933 0.4058

N=896

134 J Financ Serv Res (2012) 41:121–144

Table 5 Pre-crash performance regressions from 2003 through 2005 with size variables for derivatives type. The dependent variable uses several alternative measures of market performance: BHR is buy-and-hold return, BHAR is buy-and-hold abnormal return using control firms matched on size and book-to-market ratios, and SHARPE represents annualized Sharpe ratio. YR04 and YR05 are indicator variables equal to one in 2004 and 2005, respectively, and zero, otherwise. TARPBANK is an indicator variable equal to one if the bank received Troubled Asset Relief Program funds, and zero, otherwise. INTSIZE is the notional amount of interest rate derivatives divided by assets; FXSIZE is the notional amount of foreign exchange derivatives divided by assets; OTHERDUM is the notional amount of “other” derivatives divided by assets; CDSSIZE is the notional amount of credit default swaps divided by assets. INSPCT is insider holdings as a percentage of total shares outstanding; LNASSETS is the natural logarithm of total assets; CAPRATIO is the equity-to- assets ratio; GROWTH is the average growth rate in assets for the prior three years; GAP12 is 12-month repricing gap scaled by assets; NIM is net interest margin; PD902TA is loans more than 90 days past due scaled by assets; OBSL is the ratio of off-balance sheet liabilities to assets; LOAN2DEP is the loan-to- deposit ratio. EQISSUE is the amount of new equity issued in a given year, divided by assets. LAMBDA is the inverse Mill’s Ratio from the first-stage Heckman correction Probit model

Variable BHAR BHR SHARPE

Estimate p-value Estimate p-value Estimate p-value

Panel A: Baseline Model

INTERCEPT −0.2760 0.0004 0.4572 <.0001 1.3506 0.0003 YR04 0.0928 <.0001 −0.2481 <.0001 −0.9026 <.0001 YR05 0.0425 0.0036 −0.3927 <.0001 −1.6374 <.0001 INTSIZE 0.0053 0.4516 0.0025 0.7642 −0.0043 0.8964 FXSIZE 0.0264 0.3019 0.0174 0.5755 −0.0295 0.8090 OTHERSIZE −0.0554 0.7996 0.0052 0.9842 0.8458 0.4176 CDSSIZE −6.7202 0.5348 −2.1590 0.8693 −19.2568 0.7096 INSPCT 0.0039 0.9300 −0.0015 0.9782 0.1934 0.3598 LNASSETS 0.0012 0.8011 −0.0168 0.0044 −0.0214 0.3539 CAPRATIO 0.0316 0.8388 0.0327 0.8620 0.8341 0.2603

GROWTH 0.2727 <.0001 0.3119 <.0001 0.9588 <.0001

GAP12 0.1770 0.0001 0.1560 0.0046 0.5854 0.0070

NIM 2.2669 0.0060 3.0204 0.0025 8.8448 0.0247

PD902TA −2.7446 0.0122 −3.3572 0.0114 −17.4176 0.0009 OBSL 0.0548 0.3573 0.0885 0.2204 0.4829 0.0898

LOAN2DEP 0.0083 0.1020 0.0066 0.2817 0.0399 0.1010

EQISSUE −0.5426 0.4139 −0.3265 0.6849 −0.9124 0.7735 LAMBDA 0.0026 0.7635 0.0014 0.8899 −0.0049 0.9034 Adj. R-sq. 0.1412 0.3909 0.4028

Panel B: Baseline Model with TARP funds.

INTERCEPT −0.2851 0.0003 0.4495 <.0001 1.3172 0.0004 YR04 0.0924 <.0001 −0.2485 <.0001 −0.9042 <.0001 YR05 0.0418 0.0041 −0.3933 <.0001 −1.6399 <.0001 TARPBANK −0.0203 0.1168 −0.0174 0.2667 −0.0750 0.2238 INTSIZE 0.0055 0.4318 0.0027 0.7465 −0.0035 0.9166 FXSIZE 0.0278 0.2778 0.0185 0.5505 −0.0246 0.8408 OTHERSIZE −0.0823 0.7070 −0.0178 0.9465 0.7463 0.4757 CDSSIZE −6.2016 0.5667 −1.7138 0.8961 −17.3346 0.7375 INSPCT 0.0094 0.8326 0.0032 0.9519 0.2137 0.3130

LNASSETS 0.0028 0.5732 −0.0154 0.0102 −0.0156 0.5073

J Financ Serv Res (2012) 41:121–144 135

adjusted performance of banks that took TARP funds. Consistent with prior results, growth and lower past due loans are related to higher market performance.

To summarize, we find limited empirical evidence that banks’ use of derivatives either in good times or bad times was relevant to market valuations. In particular, we detect no systematic value differentials stemming from banks’ derivative positions across economic upturns and downturns. As a consequence, there is no reliable evidence that banks have maintained derivatives positions solely for speculative purposes. Nor have depository institutions held derivatives instruments purely for risk management. It appears that banks have sought to utilize financial derivatives to minimize variation in profitability while acting as dealers.

D. Robustness Checks

As a robustness test, we further ascertain whether the marginal value of derivatives use in both good times and bad times differs across banks facing a varying level of moral hazard problems. If moral hazard is indeed severe, we would observe that banks with deficient capital are more likely to increase risk since they do not bear all of the downside risk. Equivalently, if this hypothesis holds, derivatives use would be more detrimental to shareholders in poorly capitalized banks vis-à-vis well capitalized banks in a state of economic depression. In order to directly investigate the effect of risk, we add a binary variable that equals one if the bank’s equity-to-assets ratio is below the sample average in the year, and zero, otherwise. The variables of particular interest are the interactions of this indicator variable with our proxies for all types of derivatives use.

Table 8 summarizes the results of the regressions with the inclusion of a low capital binary variable. For the sake of brevity, we present only the results from the regressions in which we use the gross notional principal of derivatives holdings scaled by assets to capture a bank’s derivatives positions. We find that none of those interaction terms are statistically significant for the pre-distress period. We report qualitatively similar results for the period of the financial crash. In further support of our prior findings that derivatives use does not significantly impact bank valuation across good times and bad times, our results with a variable for low capital are also insignificant. An insignificant relation to value for low

Table 5 (continued)

Variable BHAR BHR SHARPE

Estimate p-value Estimate p-value Estimate p-value

CAPRATIO −0.0006 0.9970 0.0051 0.9786 0.7150 0.3386 GROWTH 0.2735 <.0001 0.3126 <.0001 0.9619 <.0001

GAP12 0.1773 <.0001 0.1562 0.0046 0.5867 0.0069

NIM 2.1775 0.0084 2.9437 0.0033 8.5137 0.0309

PD902TA −2.7759 0.0112 −3.3840 0.0108 −17.5333 0.0008 OBSL 0.0578 0.3320 0.0910 0.2076 0.4938 0.0828

LOAN2DEP 0.0076 0.1371 0.0060 0.3317 0.0372 0.1279

EQISSUE −0.4846 0.4659 −0.2766 0.7313 −0.6973 0.8262 LAMBDA 0.0030 0.7241 0.0018 0.8609 −0.0033 0.9355 Adj. R-sq. 0.1427 0.3911 0.4031

N=896

136 J Financ Serv Res (2012) 41:121–144

Table 6 Crash performance regressions from 2007 through 2009 with indicator variables for derivatives type. The dependent variable uses several alternative measures of market performance: BHR is buy-and-hold return, BHAR is buy-and-hold abnormal return using control firms matched on size and book-to-market ratios, and SHARPE represents annualized Sharpe ratio. YR08 and YR09 are indicator variables equal to one in 2008 and 2009, respectively, and zero, otherwise. TARPBANK is an indicator variable equal to one if the bank received Troubled Asset Relief Program funds, and zero, otherwise. INTDUM is an indicator for using interest rate derivatives, and zero, otherwise; FXDUM is an indicator for foreign exchange derivatives use, and zero, otherwise; OTHERDUM is an indicator for using “other” derivatives; CDSDUM is an indicator for using credit default swaps. INSPCT is insider holdings as a percentage of total shares outstanding; LNASSETS is the natural logarithm of total assets; CAPRATIO is the equity-to-assets ratio; GROWTH is the average growth rate in assets for the prior three years; GAP12 is 12-month repricing gap scaled by assets; NIM is net interest margin; PD902TA is loans more than 90 days past due scaled by assets; OBSL is the ratio of off-balance sheet liabilities to assets; LOAN2DEP is the loan-to-deposit ratio. EQISSUE is the amount of new equity issued in a given year, divided by assets. LAMBDA is the inverse Mill’s Ratio from the first- stage Heckman correction Probit model

Variable BHAR BHR SHARPE

Estimate p-value Estimate p-value Estimate p-value

Panel A: Baseline Model

INTERCEPT −0.2212 0.2571 −0.2150 0.2081 −2.0714 0.0003 YR08 0.2541 <.0001 0.0508 0.0362 0.6920 <.0001

YR09 −0.0308 0.3113 0.1968 <.0001 1.2811 <.0001 INTDUM 0.0838 0.0653 0.0664 0.0949 0.2569 0.0550

FXDUM −0.0813 0.1007 −0.0297 0.4922 −0.0975 0.5034 OTHERDUM 0.1077 0.0230 0.1135 0.0062 0.4235 0.0024

CDSDUM −0.0074 0.9095 −0.0333 0.5591 −0.2736 0.1547 INSPCT 0.0734 0.4716 0.0465 0.6027 0.1359 0.6511

LNASSETS 0.0038 0.7892 −0.0042 0.7335 0.0451 0.2800 CAPRATIO 0.1763 0.4846 0.2453 0.2665 0.6360 0.3921

GROWTH 0.1321 0.0444 0.1312 0.0225 0.0323 0.8674

GAP12 −0.0510 0.5933 −0.0222 0.7906 0.0737 0.7934 NIM 1.8624 0.2407 1.7368 0.2111 6.5023 0.1644

PD902TA −7.5647 <.0001 −7.3219 <.0001 −13.5123 <.0001 OBSL −0.1640 0.1909 −0.0876 0.4247 −0.5194 0.1597 LOAN2DEP −0.0356 0.0788 −0.0234 0.1873 −0.0609 0.3069 EQISSUE −1.2177 0.5009 −0.9389 0.5530 −3.1579 0.5534 LAMBDA −0.0625 0.0217 −0.0510 0.0325 −0.2004 0.0125 Adj. R-sq. 0.3363 0.2567 0.2331

Panel B: Baseline Model with TARP funds

INTERCEPT −0.2551 0.1917 −0.2527 0.1386 −2.1592 0.0002 YR08 0.2542 <.0001 0.0509 0.0351 0.6923 <.0001

YR09 −0.0300 0.3237 0.1978 <.0001 1.2833 <.0001 TARPBANK −0.0520 0.0288 −0.0580 0.0053 −0.1348 0.0546 INTDUM 0.0842 0.0633 0.0669 0.0911 0.2581 0.0535

FXDUM −0.0832 0.0923 −0.0318 0.4601 −0.1024 0.4814 OTHERDUM 0.1073 0.0232 0.1130 0.0062 0.4224 0.0025

CDSDUM −0.0040 0.9512 −0.0295 0.6037 −0.2647 0.1679 INSPCT 0.0945 0.3553 0.0699 0.4334 0.1905 0.5272

LNASSETS 0.0083 0.5595 0.0008 0.9457 0.0569 0.1770

J Financ Serv Res (2012) 41:121–144 137

capital banks indicates that banks did not engage in derivative securities to take advantage of risk-shifting opportunities supporting Cordella and Yeyati (2003).

To address the concern that the correlations among the capital interaction terms and other explanatory variables cloud our inferences, we perform several sensitivity checks. First, we split the full sample into two sub-samples based on the annual average of the equity-to-assets ratio and then repeat the same analysis. In unreported results, we find that only the coefficient on the interest rate derivative variable is statistically positive at the five percent significance level for the subgroup of financially fragile banks during the crash period. Overall, our findings lend no support to the claim that banks lacking sufficient capital are more likely to utilize derivatives for speculation, thus resulting in relatively worse performance in the economic recession.

Next, we also estimate the regressions on the subsample of banks that use credit default swaps to determine if larger proportions of CDSs use are related to performance. None of the performance regressions using all three measures of performance in both sample periods have significant derivative variable estimates. Thus, even for the largest CDSs users in our sample, there is no statistically different performance before or during the financial crisis.

Lastly, in tests not reported, we discover that derivatives of all types do not contribute to large banks’ valuations differently than small banks’ valuations. These findings hold for periods before or during the financial meltdown, giving no support to the moral hazard hypothesis. It appears that while big banks may be too big to fail, they did not suffer losses during the financial panic due to financial derivatives use.

5 Summary and conclusions

In this paper, we use shocks to the US banking system during the recent financial meltdown as a natural experiment for explicitly testing the effects of economic fluctuations on the financial consequences of banks’ derivatives use. The crisis originated with subprime mortgage delinquencies during the bursting of the housing

Table 6 (continued)

Variable BHAR BHR SHARPE

Estimate p-value Estimate p-value Estimate p-value

CAPRATIO 0.1501 0.5514 0.2160 0.3263 0.5680 0.4444

GROWTH 0.1286 0.0500 0.1272 0.0264 0.0230 0.9052

GAP12 −0.0563 0.5552 −0.0280 0.7363 0.0601 0.8306 NIM 1.6287 0.3049 1.4762 0.2869 5.8966 0.2076

PD902TA −7.6594 <.0001 −7.4275 <.0001 −13.7577 <.0001 OBSL −0.1676 0.1807 −0.0915 0.4024 −0.5286 0.1519 LOAN2DEP −0.0386 0.0569 −0.0267 0.1315 −0.0686 0.2501 EQISSUE −0.9690 0.5921 −0.6618 0.6752 −2.5137 0.6372 LAMBDA −0.0647 0.0174 −0.0534 0.0245 −0.2061 0.0102 Adj. R-sq. 0.3394 0.2627 0.2355

N=850

138 J Financ Serv Res (2012) 41:121–144

Table 7 Crash performance regressions from 2007 through 2009 with size variables for derivatives type. The dependent variable uses several alternative measures of market performance: BHR is buy-and-hold return, BHAR is buy-and-hold abnormal return using control firms matched on size and book-to-market ratios, and SHARPE represents annualized Sharpe ratio. YR08 and YR09 are indicator variables equal to one in 2008 and 2009, respectively, and zero, otherwise. TARPBANK is an indicator variable equal to one if the bank received Troubled Asset Relief Program funds, and zero, otherwise. INTSIZE is the notional amount of interest rate derivatives divided by assets; FXSIZE is the notional amount of foreign exchange derivatives divided by assets; OTHERDUM is the notional amount of “other” derivatives divided by assets; CDSSIZE is the notional amount of credit default swaps divided by assets. INSPCT is insider holdings as a percentage of total shares outstanding; LNASSETS is the natural logarithm of total assets; CAPRATIO is the equity-to- assets ratio; GROWTH is the average growth rate in assets for the prior three years; GAP12 is 12-month repricing gap scaled by assets; NIM is net interest margin; PD902TA is loans more than 90 days past due scaled by assets; OBSL is the ratio of off-balance sheet liabilities to assets; LOAN2DEP is the loan-to- deposit ratio. EQISSUE is the amount of new equity issued in a given year, divided by assets. LAMBDA is the inverse Mill’s Ratio from the first-stage Heckman correction Probit model

Variable BHAR BHR SHARPE

Estimate p-value Estimate p-value Estimate p-value

Panel A: Baseline Model

INTERCEPT −0.3022 0.0426 −0.3376 0.0096 −2.3837 <.0001 YR08 0.2547 <.0001 0.0523 0.0318 0.6945 <.0001

YR09 −0.0321 0.2898 0.1963 <.0001 1.2761 <.0001 INTSIZE 0.0051 0.5933 0.0046 0.5831 −0.0059 0.8320 FXSIZE −0.0175 0.6554 0.0103 0.7639 0.0463 0.6883 OTHERSIZE 0.1629 0.2909 0.1575 0.2426 0.7710 0.0902

CDSSIZE −1.1219 0.2331 −1.3118 0.1108 −2.9991 0.2796 INSPCT 0.0088 0.9293 0.0102 0.9058 0.0149 0.9592

LNASSETS 0.0136 0.1696 0.0070 0.4224 0.0767 0.0088

CAPRATIO 0.1762 0.4854 0.2702 0.2210 0.7980 0.2839

GROWTH 0.1210 0.0652 0.1258 0.0283 0.0001 0.9997

GAP12 −0.0525 0.5848 −0.0124 0.8827 0.0963 0.7338 NIM 1.7439 0.2697 1.7455 0.2062 6.7711 0.1462

PD902TA −7.5652 <.0001 −7.2761 <.0001 −13.2893 <.0001 OBSL −0.1914 0.1274 −0.0997 0.3636 −0.5659 0.1264 LOAN2DEP −0.0363 0.0717 −0.0226 0.1989 −0.0663 0.2634 EQISSUE −1.3931 0.4424 −1.1516 0.4675 −4.0195 0.4522 LAMBDA −0.0184 0.2183 −0.0147 0.2591 −0.0608 0.1671 Adj. R-sq. 0.3309 0.2521 0.2252

Panel B: Panel A: Baseline Model with TARP funds

INTERCEPT −0.3301 0.0271 −0.3695 0.0046 −2.4570 <.0001 YR08 0.2546 <.0001 0.0522 0.0315 0.6942 <.0001

YR09 −0.0315 0.2983 0.1970 <.0001 1.2778 <.0001 TARPBANK −0.0504 0.0354 −0.0577 0.0058 −0.1326 0.0606 INTSIZE 0.0057 0.5446 0.0053 0.5196 −0.0042 0.8815 FXSIZE −0.0136 0.7286 0.0147 0.6657 0.0565 0.6240 OTHERSIZE 0.1456 0.3449 0.1377 0.3057 0.7255 0.1107

CDSSIZE −1.1263 0.2303 −1.3168 0.1080 −3.0108 0.2770 INSPCT 0.0285 0.7737 0.0328 0.7044 0.0668 0.8193

LNASSETS 0.0177 0.0786 0.0117 0.1844 0.0876 0.0033

J Financ Serv Res (2012) 41:121–144 139

bubble. The ensuing collapses of Bear Sterns and Lehman Brothers are frequently considered to be manifestations of the risk management meltdown of the financial sector. This natural experiment enables us to examine the impact of adverse shocks on the value of derivative instruments use and provides insights into how financial derivatives have been used in the banking industry. The risk-sharing component of derivatives is likely to make bank performance countercyclical, whereas the risk-taking component of derivative use tends to make stock performance procyclical. To the extent that one component does not dominate the other, the wealth effect of bank derivatives will not respond significantly to unanticipated common economic shocks.

We find that the economic consequences of banks’ derivatives use did not vary considerably with cyclical fluctuations. In particular, the shock-induced credit crunch had no bearing on the stock performance of banks that used derivatives. The absence of a link between derivative use and the state of the economy remains for both financially fragile banks and large depository institutions. Accordingly, there is no evidence that the average bank utilized derivatives that inflicted damage on the banking system. Our evidence on the market’s unfavorable reaction to the granting of funds from the Troubled Asset Recovery Program suggests that expanding the balance sheet to take excessive risk in good times can lead to the instability of banks in bad times, thus yielding detrimental effects on shareholder value. Consistent with our prior finding, this result indicates that market participants discouraged banks from taking advantage of the risk-shifting opportunities provided by explicit government bailouts.

To summarize, we find that derivatives did not exacerbate the crisis even for banks that were most likely to exploit the moral hazard involved with government guarantees. On average, banks appeared not to compromise the principle of safety and soundness and utilized derivatives in a prudent fashion. While our results are consistent with Henry Paulson’s testimony that derivatives generally did not cause the financial turmoil, we can not rule out the possibility that some US banks exploited derivatives in a reckless manner. However, our cross-sectional evidence suggests that increased government oversight of derivatives will likely play a small role in preventing a repetition of the financial crisis.

Table 7 (continued)

Variable BHAR BHR SHARPE

Estimate p-value Estimate p-value Estimate p-value

CAPRATIO 0.1519 0.5472 0.2423 0.2709 0.7340 0.3241

GROWTH 0.1177 0.0724 0.1220 0.0327 −0.0087 0.9642 GAP12 −0.0564 0.5562 −0.0169 0.8399 0.0860 0.7611 NIM 1.5015 0.3422 1.4678 0.2871 6.1335 0.1885

PD902TA −7.6566 <.0001 −7.3807 <.0001 −13.5295 <.0001 OBSL −0.1984 0.1135 −0.1077 0.3246 −0.5843 0.1142 LOAN2DEP −0.0388 0.0539 −0.0255 0.1460 −0.0730 0.2184 EQISSUE −1.1523 0.5251 −0.8757 0.5798 −3.3860 0.5267 LAMBDA −0.0203 0.1736 −0.0169 0.1932 −0.0659 0.1345 Adj. R-sq. 0.3337 0.2581 0.2275

N=850

140 J Financ Serv Res (2012) 41:121–144

Table 8 Pre-crash and Crash period performance with a binary low capital variable and its interaction terms with derivatives positions. The dependent variable uses several alternative measures of market performance: BHR is buy-and-hold return, BHAR is buy-and-hold abnormal return using control firms matched on size and book-to-market ratios, and SHARPE represents annualized Sharpe ratio. LOWCAP is an indicator variable that equals one if the bank has capital below the annual average, and zero, otherwise. TARPBANK is a binary variable equals to one if the bank received Troubled Asset Relief Program funds, and zero, otherwise. INTSIZE is the notional amount of interest rate derivatives divided by assets. We define FXSIZE as the notional amount of foreign exchange derivatives divided by assets. OTHERDUM is the notional amount of “other” derivatives normalized by assets, and CDSSIZE is the notional amount of credit default swaps scaled by assets. INSPCT is insider holdings as a percentage of total shares outstanding. LNASSETS is the natural logarithm of total assets. CAPRATIO is the equity-to-assets ratio; GROWTH is the average growth rate in assets for the prior three years; GAP12 is 12-month repricing gap scaled by assets; NIM is net interest margin; PD902TA is loans more than 90 days past due scaled by assets; OBSL is the ratio of off- balance sheet liabilities to assets; LOAN2DEP is the loan-to-deposit ratio. EQISSUE is the amount of new equity issued in a given year, divided by assets. CAPCDS is the interaction term between LOWCAP and CDSSIZE; CAPINT is an interaction term between LOWCAP and INTSIZE; CAPFX is the interaction term between LOWCAP and FXSIZE; CAPOTH is the interaction term between LOWCAP and OTHSIZE. LAMBDA is the inverse Mill’s Ratio from the first-stage Heckman correction probit model

Variable BHAR BHR SHARPE

Estimate p-value Estimate p-value Estimate p-value

Panel A: Pre-Crash (2003–2005) N=896

INTERCEPT −0.3289 <.0001 0.1875 0.1251 0.2586 0.5972 LOWCAP 0.0590 <.0001 0.0590 0.0048 0.2669 0.0015

TARPBANK −0.0229 0.0809 −0.0079 0.6877 −0.0351 0.6557 INTSIZE 0.0495 0.1016 0.0581 0.1995 0.1936 0.2856

FXSIZE −0.2391 0.3487 −0.2641 0.4896 −1.4154 0.3554 OTHSIZE 0.0101 0.9663 0.0196 0.9562 1.0117 0.4798

CDSSIZE −15.2086 0.9729 −6.2676 0.9925 −363.2147 0.8925 INSPCT −0.0028 0.9506 0.0142 0.8356 0.2376 0.3852 LNASSETS 0.0033 0.5111 −0.0179 0.0175 −0.0257 0.3949 CAPRATIO 0.3070 0.0809 0.2621 0.3198 1.9500 0.0648

GROWTH 0.2853 <.0001 0.3050 <.0001 0.9421 0.0001

GAP12 0.1884 <.0001 0.2052 0.0030 0.8143 0.0032

NIM 2.5988 0.0025 3.4961 0.0066 9.8795 0.0550

PD902TA −3.0693 0.0056 0.5611 0.7346 −1.0984 0.8684 OBSL 0.0762 0.2086 −0.0795 0.3815 −0.1859 0.6095 LOAN2DEP 0.0047 0.3830 0.0063 0.4415 0.0403 0.2153

EQISSUE −0.0060 0.9929 0.3643 0.7186 2.5544 0.5282 CAPCDS 15.3248 0.9727 −4.1252 0.9951 299.1202 0.9114 CAPINT −0.0485 0.1231 −0.0454 0.3357 −0.1398 0.4588 CAPFX 0.2657 0.2997 0.3010 0.4329 1.4635 0.3412

CAPOTH −0.1363 0.8367 −0.1982 0.8415 −1.2673 0.7497 LAMBDA 0.0017 0.8463 −0.0022 0.8632 −0.0208 0.6893 Adj. R-sq. 0.1190 0.0443 0.0318

Panel B: During the Financial Crash (2007–2009) N=850

INTERCEPT −0.1683 0.3219 −0.3355 0.0182 −2.2747 <.0001 LOWCAP −0.0227 0.4101 −0.0386 0.0937 −0.2184 0.0094 TARPBANK −0.0533 0.0393 −0.0564 0.0091 −0.1249 0.1121 INTSIZE 0.0151 0.7852 0.0264 0.5682 −0.0008 0.9964

J Financ Serv Res (2012) 41:121–144 141

References

Acharya V, Yorulmazer T (2008) Cash-in-the-market pricing and optimal resolution of bank failures. Rev Econ Stud 21:2705–2742

Adam TR, Fernando CS (2006) Hedging, speculation, and shareholder value. J Financ Econ 81:283–309 Adrian T, Shin H (2008) Financial intermediary leverage and value at risk. Federal Reserve Bank Staff

Reports no. 338. Allayannis G, Weston JP (2001) The Use of foreign currency derivatives and firm market value. Rev Financ

Stud 14:243–276 Allen F, Carletti E (2010) The role of liquidity in financial Crises. Working Paper presented at the 2008

Jackson Hole Symposium on Maintaining Stability in a Changing Financial System. Beltratti A, Stulz R (2009) Why did some banks perform better during the credit crisis? A cross-country

study of the impact of governance and regulation. Working Paper, Ohio State University. Brewer E, Jackson WE, Moser JT (2001) The value of using interest rate derivatives to manage risk at U.S.

banking organizations. Federal Reserve Bank of Chicago Economic Perspectives 49–66. Brunnermeier MK (2009) Deciphering the liquidity and credit crunch 2007–2008. J Econ Perspect 23:77–

100 Buser S, Chen A, Kane E (1981) Federal deposit insurance, regulatory policy, and optimal bank capital. J

Finance 35:51–60 Calomiris C, Mason J (2003) Consequences of US bank distress during the depression. Am Econ Rev

93:937–947 Carter D, Rogers D, Simkins B (2006) Does hedging affect firm value: evidence from the U.S. airline

industry? Financ Manage 35:53–86

Table 8 (continued)

Variable BHAR BHR SHARPE

Estimate p-value Estimate p-value Estimate p-value

FXSIZE −0.0593 0.6973 −0.0625 0.6234 −0.2328 0.6156 OTHSIZE 0.5276 0.6796 1.3001 0.2233 4.3519 0.2629

CDSSIZE −3.4406 0.6748 −7.9384 0.2467 −8.8040 0.7241 INSPCT 0.0436 0.6834 0.0171 0.8482 −0.0532 0.8701 LNASSETS 0.0144 0.1901 0.0180 0.0503 0.1379 <.0001

CAPRATIO 0.1542 0.5988 0.0922 0.7066 0.0446 0.9601

GROWTH 0.1984 0.0048 0.0861 0.1425 −0.1456 0.4957 GAP12 −0.0442 0.6698 −0.0234 0.7867 0.0715 0.8204 NIM 0.3515 0.8382 1.7122 0.2340 7.0423 0.1789

PD902TA −7.7463 <.0001 −6.0848 <.0001 −4.9620 0.0036 OBSL −0.2035 0.1250 −0.2823 0.0109 −1.7479 <.0001 LOAN2DEP −0.0246 0.2566 −0.0389 0.0316 −0.1575 0.0169 EQISSUE −3.4612 0.0762 −0.0446 0.9782 −0.5608 0.9247 CAPCDS 3.3328 0.6869 6.7183 0.3308 6.9507 0.7823

CAPINT −0.0142 0.8018 −0.0175 0.7112 0.0054 0.9750 CAPFX 0.0475 0.7648 0.0816 0.5381 0.3342 0.4888

CAPOTH −0.4474 0.7287 −1.2274 0.2549 −3.9482 0.3145 LAMBDA −0.0206 0.2014 −0.0156 0.2450 −0.0562 0.2512 Adj. R-sq. 0.2242 0.2078 0.0436

142 J Financ Serv Res (2012) 41:121–144

Carter DA, Sinkey JF (1998) The use of interest rate derivatives by end-users: the case of large community banks. J Financ Serv Res 14:17–34

Cole RA (1993) When are thrift institutions closed? An agency-theoretic model. J Financ Serv Res 7:283– 307

Cordella T, Yeyati EL (2003) Bank bailouts: moral hazard vs. value effect. J Financ Intermed 12:300–330 Diamond D (1991) Monitoring and reputation: the choice between bank loans and directly placed debt. J

Polit Econ 99:689–721 Fama E, French K (1993) Common risk factors in the returns on stocks and bonds. J Financ Econ 33:3–56 Fahlenbrach R, Stulz R (2011) Bank CEO incentives and the credit crisis. J Financ Econ 99:11–26 Fratianni MU, Marchionne F (2009) The role of banks in the subprime crisis. Indiana University Working

Paper. Froot K, Scharfstein D, Stein J (1993) Risk management: coordinating investment and financing policies. J

Finance 48:1629–1658 Gorton G (2008) The subprime panic. Eur Financ Manag 15:10–46 Gorton G., Rosen R (1995) Banks and derivatives. Working Paper, National Bureau of Economic Research,

Cambridge, MA. Graham J, Rogers D (2002) Do firms hedge in response to tax incentives? J Finance 57:815–840 Greenberger M (2010). The role of derivatives in the financial crisis–credit default swaps and the economic

meltdown. Testimony to the Financial Crisis Inquiry Commission Hearing in Washington D.C. Guay W, Kothari SP (2003) How much do firms hedge with derivatives? J Financ Econ 70:423–461 Heckman JJ (1979) Sample selection bias as a specification error. Econometrica 47:153–161 Huang P, Zhang Y, Deis D, Moffitt J (2009) Do artificial income smoothing and real income smoothing

contribute to firm value equivalently? J Bank Finance 33:224–233 Jensen M, Meckling WH (1976) Theory of the Firm: managerial behavior, agency costs and ownership

structure. J Financ Econ 3:305–360 Jin Y, Jorion P (2006) Firm value and hedging: evidence from U.S. oil and gas producers. J Finance 61:893–

919 Kahane Y (1977) Capital adequacy and the regulation of financial intermediaries. J Bank Finance 1:207–218 Kane EJ (1989) Changing incentives facing financial-services regulators. J Financ Serv Res 2:265–274 Keeley M (1990) Deposit insurance, risk, and market power in banking. Am Econ Rev 80:1183–1200 Laeven L, Levine R (2009) Bank governance, regulation, and risk taking. J Financ Econ 93(2):259–275 Lang L, Stulz R (1994) Tobin’s Q, corporate diversification, and firm performance. J Polit Econ 102:1248–

1280 Leland H (1998) Agency costs, risk management, and capital structure. J Finance 53:1213–1243 Merton RC (1977) An analytic derivation of the cost of deposit insurance and loan guarantees. J Bank

Finance 1:3–11 Minton BA, Stulz R, Williamson R (2009) How much do banks use credit derivatives to hedge loans? J

Financ Serv Res 35:1–31 Morck R, Shleifer A, Vishny R (1988) Management ownership and market valuation: an empirical analysis. J

Financ Econ 20:293–316 Nance D, Smith C, Smithson C (1993) On the determinants of corporate hedging. J Finance 48:267–284 Paletta D, Patterson S (2010) Deal near on Derivatives. Wall Street Journal, April, 26 Park S (1997) Risk-taking behavior of banks under regulation. J Bank Finance 21:491–507 Paulson HM (2010) Testimony to the Financial Crisis Inquiry Commission, Dirksen Senate Office Building,

Room 538, Washington DC, May 6, 2010. Rajan R (2006) Has financial development made the world riskier? Eur Financ Manag

12:499–533 Saporito B. (2009) How AIG became too big to fall. Time, March 30. Saunders A, Wilson B (2001) An analysis of bank charter value and its risk-constraining incentives. J Financ

Serv Res 19:185–195 Shleifer A, Vishny R (2010) Unstable banking. J Financ Econ 97:306–318 Singh M, Aitken J (2009) Counterparty risk, impact on collateral flows, and the role for central

counterparties. International Monetary Fund Working Paper 09/173. Sinkey JF, Carter DA (1999) The reaction of bank stock prices to news of derivative losses by corporate

clients. J Bank Finance 23:1725–1943 Sinkey JF, Carter DA (2000) Evidence on the financial characteristics of banks that do and do not use

derivatives. Q Rev Econ Finance 40:431–449 Smith CW, Stulz RM (1985) The determinants of firms’ hedging policies. J Financ Quant Anal 20:391–

405

J Financ Serv Res (2012) 41:121–144 143

Smith CW, Watts RL (1992) The investment opportunity set and corporate financing, dividend, and compensation policies. J Financ Econ 32:263–292

Stulz R (2009) Credit default swaps and the credit crisis. Working Paper, Ohio State University. Strunk N, Case F (1988) Where deregulation went wrong: a look at the causes behind savings and loan

failures in the 1980s. United States League of Savings Institutions, Chicago Whidbee DA, Wohar M (1999) Derivative activities and managerial incentives in the banking industry. J Cor

Fince 5:251–276

144 J Financ Serv Res (2012) 41:121–144

Copyright of Journal of Financial Services Research is the property of Springer Science & Business Media B.V.

and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright

holder's express written permission. However, users may print, download, or email articles for individual use.