Offsetable Derivative Exposures and Financial Stability. Jed J. Neilson Pennsylvania State University

Transcription

1 Offsetable Derivative Exposures and Financial Stability Jed J. Neilson Pennsylvania State University K. Philip Wang University of Florida Christopher D. Williams University of Michigan Biqin Xie Pennsylvania State University April 24, 2018 Abstract U.S. GAAP allows banks to offset reciprocal derivative assets and liabilities with the same counterparty and report only the net amount on the balance sheet, resulting in trillions of dollars of offsetable derivative assets and liabilities going unrecognized. These offsetable derivative exposures create the single largest quantitative difference between amounts recognized under U.S. GAAP and those under IFRS, contributing to a long-standing debate about whether offsetable derivatives are informative of banks risk. In this paper, we examine whether offsetable derivatives have implications for the stability of the financial system. We find that offsetable derivatives are positively associated with banks default risk and their exposure to systemic risk. Further analyses suggest that offsetable derivatives reflect banks exposure to market risk and counterparty credit risk. Finally, we find that adjusted regulatory capital ratios that fully incorporate offsetable derivatives have stronger explanatory power for banks default risk and systemic risk than the standard version of these ratios currently used by banking regulators, which largely ignore these offsetable derivatives. We acknowledge helpful comments from Orie Barron, Gauri Bhat (discussant), Jeremiah Green, Guojin Gong, Yihong Jiang, Henock Louis, Karl Muller, Hong Qu, and workshop participants at the 2017 BYU Accounting Research Symposium, University of Illinois at Chicago, Pennsylvania State University, and the 2018 FARS Midyear Meeting.

2 1. Introduction U.S. GAAP allows banks to offset reciprocal derivative assets and liabilities with the same counterparty and report only the net amount of derivatives on the balance sheet if the two firms have a legal setoff right under a master netting agreement. 1 These unrecognized derivative exposures, which amount to trillions of dollars and represent more than 90% of derivative assets and liabilities of U.S. banks, have led to significant debates about whether offsetable derivative exposures are informative of banks risk. Indeed, this is one of the reasons that IFRS requires stricter offsetting criteria than U.S. GAAP, which has created the single largest quantitative difference in the amounts presented in statements of financial position prepared in accordance with U.S. GAAP and in the amounts presented in those prepared in accordance with IFRS (FASB and IASB 2011, p. 1). Since banking regulators generally use recognized U.S. GAAP numbers in this case the net derivative exposures to calculate banks regulatory capital ratios and monitor risk in bank derivatives, offsetable derivative exposures are essentially ignored in evaluating bank risks (see, e.g., Office of the Comptroller of the Currency 2008). In this paper, we examine the implications of offsetable derivatives for financial stability. Specifically, we investigate whether offsetable derivatives are associated with a bank s default risk and its potential exposure to systemic risk. Both U.S. bank holding companies (hereafter, banks ) and bank regulators have strongly opposed efforts by the FASB to converge with IFRS to limit the extent to which U.S entities may 1 For example, suppose Bank A has $100 of derivatives with positive fair values (i.e., derivative assets) and $90 of derivatives with negative fair values (i.e., derivative liabilities) with Bank B, under a master netting agreement. U.S. GAAP allows Bank A to net the $90 of derivative liabilities against the $100 of derivative assets, thus recognizing only $10 of net derivative assets on the balance sheet. In this paper, we call the $90 of reciprocal derivative assets and liabilities that are allowed to be eliminated from the balance sheet offsetable derivative exposure or offsetable derivatives, using the terms interchangeably. Similarly, we call the $10 of net derivative assets reported on the balance sheet as net derivative exposures, net derivative assets, or net derivatives, using the terms interchangeably. The U.S. GAAP offsetting guidance is set out in Section 45 of ASC Balance Sheet Offsetting and Section 45 of ASC Derivatives and Hedging Overall. 1

3 offset their derivative assets and derivative liabilities. In a joint response to a January 2011 Exposure Draft on this topic, five regulatory agencies, including the Federal Reserve and the Federal Deposit Insurance Corporation (FDIC), stated that including offsetable derivatives on the balance sheet would significantly overstate the reported assets and liabilities of financial institutions and impair rather than improve financial reporting by providing less relevant information to financial statement users (FASB 2011a, p. 2). 2 Similarly, the four largest U.S. banks (i.e., JP Morgan, Citigroup, Bank of America, and Wells Fargo) strongly opposed the proposed changes to U.S. GAAP that would bring a substantial amount of offsetable derivatives onto the balance sheet (e.g., FASB 2011b, p. 2; FASB 2011c, p. 2). The FASB eventually decided to retain its existing (substantially more lenient) offsetting model and instead bridge the differences in offsetting criteria between U.S. GAAP and IFRS by adopting common requirements for footnote disclosure of derivatives. 3 The position of U.S. banks and regulators on this issue relies largely on the fact that counterparties in derivative markets often have a legal setoff right under a master netting agreement, which allows them to offset derivative assets against derivative liabilities with the same counterparty in the event of counterparty default or bankruptcy. In those events, the setoff right grants the derivative counterparties the privilege of immediately collecting their debt by setting off their derivative assets against their derivative liabilities to the defaulting counterparty, leaving 2 The Exposure Draft jointly issued by the FASB and the IASB in January 2011 is titled Proposed Accounting Standards Update Balance Sheet (Topic 210): Offsetting. 3 The FASB and the IASB jointly issued these common disclosure requirements in December 2011 (Accounting Standards Update No , Balance Sheet [Topic 210]: Disclosures about Offsetting Assets and Liabilities). Note that examining the relative desirability of balance-sheet recognition versus footnote disclosure of offsetable derivatives is beyond the scope of this paper. Instead, our focus is on the fundamental question that underlies the debate on this topic does the practice of netting derivative assets and liabilities obscure information about bank risk and threats to financial stability? 2

4 only the net claims exposed to the risk of the defaulting counterparty (see, e.g., Summe 2010, p. 66; Charles 2009). While the mitigation of counterparty credit risk inherent in the contractual arrangement of offsetting is clear and intuitive, opponents of offsetting argue that bank risks are understated when only the net amount of derivative assets and liabilities is used to measure and monitor risk in bank derivatives. 4 For example, economists Admati and Hellwig (2013, p. 86) state that the practice of netting that is allowed under U.S. accounting rules for derivatives masks important risks. The IASB and some FDIC officials argue that netting under existing U.S. GAAP obscures important information about an entity s financial position, leverage, and exposure to market risk, operational risk, and counterparty credit risk (FASB and IASB 2011, BC33-34 in p. 52; Hoenig 2013). Thomas Hoenig, the Vice Chairman of the FDIC, believe that banks should be required to hold capital against their offsetable derivatives (Hoenig 2013). Opponents of netting generally support their position by citing one or more of the following concerns (discussed in more detail in Section 2.2). First, derivative assets and derivative liabilities that are offset under U.S. GAAP come from different derivative contracts with different terms and conditions and with cash flows sensitive to different market factors. While the net fair values provide a snapshot of how derivative assets and liabilities offset on the balance sheet date, they often represent less than 10% of the gross fair values, and thus may understate the potential for significant divergences in the fair values of different derivative contracts due to subsequent changes in underlying asset prices such as interest rates, foreign exchange prices, or commodity prices (Hoenig 2013). Second, by their very nature derivatives have high embedded leverage, and hence even nominal fluctuations in the underlying asset prices can lead to substantial changes in the fair value of derivative instruments, further 4 We use the two terms offsetting and netting interchangeably in this paper. 3

5 amplifying the market risk and counterparty credit risk of derivatives. Third, since derivative trades between banks form a complex network of cross exposures, it is possible that the size of a bank s offsetable derivatives reflects the bank s exposure to an interconnected network of counterparties in the financial system and hence may reflect the bank s potential exposure to systemic risk. In this paper we explore the implications of offsetable derivatives for financial stability by first examining whether offsetable derivatives are positively associated with a bank s default risk. We then explore the possibility that the magnitude of these offsetable derivatives also reflects a bank s potential exposure to systemic risk. We calculate offsetable derivatives as the gross (i.e., total) amount of derivative assets disclosed in banks FR Y-9C regulatory filings minus the net amount of derivative assets reported on the balance sheet, scaled by total assets. Our initial tests examine the relation between offsetable derivatives and default risk. We measure default risk using banks bond yield spreads, calculated as bond yields from secondary market trades less the interest rate on Treasury bonds of similar maturity (Barth et al. 2012; Blankespoor et al. 2013). We find a positive association between offsetable derivatives and bond yield spreads, suggesting that these offsetable exposures have implications for default risk. This result is robust to controlling for other potential sources of bank risk, namely, the quality of a bank s risk management function, and risk-taking incentives emanating from executive compensation structure. We also perform robustness tests using credit default swap (CDS) spreads as an alternative measure of default risk, and obtain results similar to those using bond yield spreads. Given our finding of a positive relation between offsetable derivatives and default risk, and given that the size of offsetable derivatives may reflect a bank s exposure to an interconnected network of counterparties, we next explore the relation between offsetable derivatives and a bank s 4

6 exposure to systemic risk. We use the SRISK measure developed by Acharya et al. (2012) and further refined by Brownlees and Engle (2016). SRISK captures the expected capital shortfall of a financial firm, conditional on a severe market decline. We find that offsetable derivative exposures are positively associated with SRISK, suggesting that these offsetable derivatives relate to a bank s exposure to systemic risk. To further understand how offsetable derivatives may contribute to a bank s default risk and exposure to systemic risk, we more closely investigate the role of market risk and counterparty credit risk. We capture the ex-post realization of a bank exposure to market risk using the volatility of the bank s trading revenue from derivatives and cash instruments. We capture the ex-post realization of a bank s exposure to derivative counterparty risk using the volatility of the impact on the bank s trading revenue resulting from changes in the creditworthiness of the bank s derivative counterparties. We find that offsetable derivatives are positively associated with both volatility measures. The results suggest that offsetable derivatives reflect a bank s exposure to market risk and counterparty credit risk, which supports our main findings that offsetable derivatives are positively associated with a bank s default risk and exposure to systemic risk. Finally, in an attempt to shed light on potential policy and regulatory implications of offsetable derivative exposures, we examine whether adjusted regulatory capital ratios that fully incorporate offsetable derivatives have stronger explanatory power for a bank s default risk and systemic risk than the standard version of these ratios currently used by regulators, which largely ignore offsetable derivatives. We find that these adjusted capital ratios do contain information about banks default risk and exposure to systemic risk incremental to the standard ratios currently used by the regulators. 5

7 In summary, we provide evidence that offsetable derivative exposures convey information about banks default risk and exposure to systemic risk, and that the risk relevance of regulatory capital ratios may be improved by fully incorporating these offsetable derivatives into the ratio calculation. This paper makes contributions along two major dimensions. First, it offers potential policy implications to accounting standard setters and banking regulators. Our findings stand in stark contrast to the claims of the banking regulators and the large banks that offsetable derivatives are uninformative, and even misleading, about bank risk. Although the relative desirability of balancesheet recognition versus footnote disclosure of offsetable derivatives is beyond the scope of this paper, our evidence suggests that the net amount of derivative assets recognized on the balance sheet may not sufficiently capture the threat of banks derivative exposures to financial stability. This is particularly important given that regulators rely on recognized U.S. GAAP numbers in this case the net amount of derivative assets in calculating banks regulatory capital ratios and in monitoring risk in the banking industry s derivative activities. Indeed, Laux and Leuz (2009) suggest that the use of GAAP numbers by regulators is an interesting issue for future research it might be more appropriate to adjust banking regulation, rather than the accounting system, given the accounting numbers are used in many other contexts. In the case of derivatives, bank regulators might consider the need to adjust regulatory capital ratios to incorporate offsetable derivative exposures, as has long been called for by the FDIC Vice Chairman Thomas Hoenig (Hoenig 2013). 5 U.S. bank regulators might also consider the need to add offsetable derivatives as an important metric in their periodic monitoring and evaluation of bank risk in derivative activities. 5 Note that most member countries in the Basel Committee on Banking Supervision (i.e., the primary global standard setter for the prudential regulation of banks), including the U.S., follow the Basel Capital Accord of 1988 as amended in 1995 to allow significant netting of derivatives (which is consistent with U.S. GAAP) in the calculation of bank regulatory capital ratios, resulting in offsetable derivatives being largely ignored in the determination of capital 6

8 Second, this paper contributes to the literature that examines the effect of financial reporting on financial stability. While the existing literature on this topic focuses on examining the effects of fair value accounting and securitizations on financial stability (e.g., Vyas 2011; Xie 2016; Dou et al. 2014; Cheng et al. 2011), Acharya and Ryan (2016) point out that minimal such research exists on financial reporting for risk-concentrated financial instruments such as derivatives (p. 319). 6 Acharya and Ryan (2016) further suggest that, relative to fair value accounting, accounting for derivatives with concentrated risks and covered by master netting agreements has more direct and significant implications for stability due to its greater effects on banks regulatory capital ratios (p. 285). The potential effects of bank derivatives are particularly meaningful because the failures of financial institutions impose an externality on the rest of the economy (Giesecke et al. 2014) and the use of derivatives by financial institutions has been highlighted as a contributing factor to the financial crisis (Financial Crisis Inquiry Commission 2011, pp. xxiv-xxv and p. 343). To the best of our knowledge, this paper is the first to empirically examine the implications of offsetable derivatives for financial stability and the effects of these derivatives on the risk relevance of regulatory capital ratios. In what follows, Section 2 provides institutional background and motivation. Section 3 describes research design and data. Sections 4 presents empirical results. Section 5 contains additional analysis. Section 6 concludes. adequacy, regardless of whether a bank follows U.S. GAAP or IFRS (Basel Committee on Banking Supervision 1995, pp. 3-6). Thus, the findings in our paper are also relevant to non-u.s. banks that use IFRS. 6 The existing literature related to derivatives largely focuses on how new accounting standards affect the risk relevance of the notional amount of derivatives. For example, McAnally (1996) finds that the notional amounts of derivatives disclosed under SFAS 109 are associated with lower systematic risk and lower industry risk (i.e., lower correlation between an individual bank s stock return and the return of the banking industry). Ahmed et al. (2011) find that SFAS 133 increases the risk relevance of the net notional amount of bank derivatives to bond markets, in that the net notional amount of bank derivatives becomes more negatively associated with bond yield spreads after SFAS

9 2. Institutional Background and Motivation Our paper is motivated by the long-standing debate on whether offsetable derivative exposures contain information about bank risk. This debate is important, not only because the magnitude of derivatives that are unrecognized on the balance sheet due to offsetting is substantial as generally more than 90% of banks derivative assets and liabilities are offset (see Figure 2), but also because the use of derivatives by financial institutions has been found to have contributed to the financial crisis (Financial Crisis Inquiry Commission 2011, pp. xxiv-xxv and p. 343; CBS 2008; Stout 2014). 2.1 Why Offsetable Derivatives May Not Contain Information About Bank Risk Arguments that offsetable derivatives do not carry information about bank risk generally rely on the fact that counterparties in derivative markets often have a legal right, referred to as the setoff right, which allows them to offset derivative assets against derivative liabilities with the same counterparty in case of counterparty default or bankruptcy. 7 Such a legal setoff right in derivatives is typically obtained by entering a master netting agreement, a legally binding contract that allows the aggregation of transactions to cancel out reciprocal derivative payables and reach a single net payable or receivable (International Swaps and Derivatives Association 2012; Gregory 2010). 8 For example, suppose Bank A and Bank B are derivative counterparties in multiple derivative contracts under a master netting agreement, and Bank A has $100 of derivative assets 7 This right of setoff applies only to reciprocal assets and liabilities arising from derivatives transactions and repurchase agreements, not to other types of assets and liabilities. For example, suppose a bank extends a mortgage loan (which is an asset to the bank) to an individual from whom the bank also takes deposits (which is a liability to the bank). The bank is not allowed to offset the mortgage loan against the deposits with the individual, since the legal right of setoff does not apply to reciprocal assets and liabilities outside of derivatives and repurchase agreements. For banks, the amount of offsetable assets and liabilities from derivative transactions is typically much larger than those from repurchase agreements (the latter is item bhcka288 in banks FR Y-9C regulatory filings). Offsetting criteria under U.S. GAAP are set out in Section 45 of ASC Balance Sheet Offsetting and Section 45 of ASC Derivatives and Hedging Overall. 8 The most widely used master netting agreement in over-the-counter (OTC) derivative markets is International Swaps and Derivatives Association (ISDA) Master Netting Agreement developed by the global trade association ISDA. 8

10 and $90 of derivative liabilities with Bank B. The presence of the setoff right means that Bank A can net out $100 of assets against $90 of liabilities to reach a single net asset of $10 with Bank B in case Bank B defaults or goes bankrupt. The setoff right in derivative markets essentially exempts derivatives from the automatic stay. The automatic stay prevents a firm s regular creditors from collecting their debts once the firm declares bankruptcy, which means that a regular creditor with both assets and liabilities with a bankrupt firm must immediately pay off all of what it owes to the bankrupt firm, but has to wait (possibly for months or years) to recover (typically a fraction of) the amount owed by the bankrupt firm through bankruptcy proceeding, leaving the regular creditor s whole assets with the bankrupt firm exposed to the risk of the bankrupt firm. In contrast, the setoff right grants derivative counterparties the privilege of immediately collecting their debt by setting off their derivative assets against their derivative liabilities to the bankrupt firm, leaving only the remaining net claims after offsetting exposed to the risk of the bankrupt firm (see, e.g., Summe 2010, p. 66; Charles 2009). In the example above, if Bank B declares bankruptcy, the setoff right grants Bank A the right to immediately collect its claims of $90 from Bank B through netting, leaving only net assets of $10 (as opposed to $100) exposed to the risk of the bankrupt Bank B. 9 Thus, the netting granted by the legal setoff right in case of counterparty default or bankruptcy appears to make offsetable derivatives (the $90 of derivative assets and liabilities that are offset in the example above) irrelevant to bank risk, which largely forms the basis of the arguments made by opponents to a 2011 FASB proposal to limit offsetting. In a January 2011 exposure draft jointly issued with the IASB, titled Proposed Accounting Standards Update Balance Sheet (Topic 210): Offsetting, the FASB proposed tightening its 9 In contrast, if Bank A was a regular creditor, it would have to first pay $90 to Bank B to settle its $90 of liabilities with Bank B, and then wait along with other claimants to collect its $100 of assets from Bank B. 9

11 offsetting standards to converge more closely with the substantially stricter offsetting criteria of IFRS (FASB and IASB, 2011). 10 In their comment letters regarding the proposal, the large U.S. banks unvaryingly opposed this convergence, citing the right of setoff as the principal reason why offsetable derivatives do not contain information about bank risk. For example, Bank of America stated that we do not agree that the gross positions provide insight into liquidity or solvency of the entity. Instead, we would argue that the net position more clearly represents the counterparty credit exposure, the future cash flow requirements, and the liquidity and solvency of the reporting entity. (FASB 2011c, p. 2). 11 Similarly, the American Bankers Association stated in its comment letter that the gross amount has no practical relation whatsoever to any risk, whether credit, liquidity or other market risk. In fact, liquidity risks will often be distorted by gross presentation. gross presentation will also provide no clearer information on market risk than net presentation (FASB 2011d, p. 3). The International Swaps and Derivatives Association (ISDA), the trade organization that designs the most widely used master netting agreements used in derivative markets, stated in its comment letter that net presentation, in accordance with U.S. GAAP, provides the most faithful representation of an entity s financial position, solvency, and its exposure to credit and liquidity risk (FASB 2011e, p. 2). Perhaps surprisingly, the U.S. banking regulators also sided with the banks, stating in a joint comment letter that companies should be allowed to offset derivative (and repurchase agreement) assets and derivative liabilities as long as 10 Specifically, U.S. GAAP allows entities to offset derivative assets and liabilities with the same counterpart and report only the net amount on the balance sheet as long as a master netting agreement provides the legal setoff right to settle the contracts on a net basis in the event of counterparty default or bankruptcy. In contrast, IFRS requires offsetting and reporting on a net basis when, and only when, there is an unconditional right to set-off that is enforceable at all times and under all circumstances and the intention to settle on a net basis (see, e.g., International Swaps and Derivatives Association 2012). 11 In the comment letters cited in this paper, the gross positions (amount) refers to the gross (i.e., total) fair value of derivative assets and liabilities before offsetting, which equals the sum of the net positions (amount) and the offsetable positions (amount); gross (net) presentation refers to reporting the gross (net) amount of derivative assets and liabilities on the balance sheet. 10

12 they are covered by legally enforceable master netting agreements (FASB 2011a, p. 2). This position by U.S. regulators may result from regulatory forbearance (Costello et al. 2016), and is consistent with U.S. regulators long-standing practice of using net derivatives (based on the more lenient offsetting criteria of U.S. GAAP) as the primary metric to monitor and evaluate credit risk in banks derivatives activities Why Offsetable Derivatives May Contain Information About Bank Risk While the legal setoff right may mitigate counterparty credit risk in the event of counterparty default or bankruptcy, opponents of derivative netting argue that offsetable derivative exposures may still contain information about bank risk, for the following reasons. First, derivative assets and derivative liabilities that are offset under U.S. GAAP come from different derivative contracts with different terms and with fair values sensitive to different market factors (Hoenig 2013). For example, the fair value of interest rate swaps may be offset with that of corn futures if these contracts are with the same counterparty under a master netting agreement. While the net fair values of these contracts provide a snapshot of how derivative assets and liabilities offset at a particular point in time, they often represent less than 10% of the gross fair values, and thus may understate the potential for significant divergences in the fair values of different derivative contracts due to subsequent changes in underlying asset prices. The IASB has raised concerns that netting may obscure information about an entity s exposure to such market risk, stating that zero gross exposure is different from zero net exposure, because the latter may have significant 12 Every quarter since 1995Q4, the U.S. banking regulator Office of the Comptroller of the Currency (OCC) has been issuing Quarterly Report on Bank Derivatives Activities. A standard sentence in these quarterly reports is that net current credit exposure is the primary metric used by the OCC to evaluate credit risk in bank derivatives activities (see, e.g., the 2008Q4 report at Net current credit exposure is the net amount of derivative assets after offsetting under netting agreements. A complete list of these quarterly reports can be found at 11

13 counterparty, operational or other risks. For example, a bank that has a large amount of derivatives contracts outstanding, but without any significant net exposure, could still make very large losses if prices change significantly or important counterparties fail and netting arrangements do not work (FASB and IASB 2011, BC34 in p. 52). Second, opponents of derivative netting have also raised concerns that banks exposure to market risk may be exacerbated by the leverage that is inherent in derivative contracts, which they say is obscured by netting. For example, the IASB has stated that net presentation reduces users ability to understand the implied economic leverage position of an entity. Leverage is of concern to users because of two effects: (a) it creates and increases the risk of default and (b) it increases the potential for rapid deleveraging (FASB and IASB 2011, BC33 in p. 52). Since generally more than 90% of derivative assets are offset under U.S. GAAP (see Figure 2), offsetable derivatives may better reflect the high economic leverage embedded in derivative contracts than net derivatives. Finally, in addition to understating a bank s market risk exposure and implied economic leverage, derivative netting may also obscure a bank s exposure to an interconnected network of counterparties, and thus understate its exposure to counterparty and systemic risk. Since most derivative trades are over-the-counter, the opaque nature of OTC derivative markets increases banks uncertainty about cross exposures banks only understand their own exposures, but they are increasingly uncertain about cross exposures of banks that are farther away from themselves in the network (Acharya 2013). The model in Caballero and Simsek (2013) shows that when a surprise liquidity shock hits parts of the network, cross exposures between banks dramatically amplify banks perceived counterparty risk as banks become concerned that they might be indirectly hit. Likewise, the model in Jarrow and Yu (2001) shows that counterparty risk that arise 12

14 from interfirm linkage can cause a single default to lead to a cascade of defaults among interdependent firms. Finally, the model in Elliott et al. (2014) predicts that increased financial interdependencies among counterparties can lead to cascades of failures among interdependent organizations. 2.3 The Effect of Derivative Offsetting on Regulatory Capital Ratios of U.S. Banks Like most other member countries of the Basel Committee on Banking Supervision, the U.S. follows the Basel Capital Accord of 1988 as amended in 1995 to allow significant netting of derivatives (which is consistent with U.S. GAAP) in the calculation of bank regulatory capital ratios, resulting in offsetable derivatives being largely ignored in the determination of banks capital adequacy (Hoenig 2013; Basel Committee on Banking Supervision 1995, pp. 3-6). Under the Basel Capital Accord as followed by U.S. banking regulators, banks are required to hold capital against the credit equivalents of derivative contracts. The credit equivalents is calculated as the sum of two components current exposure and potential future exposure, and the two components recognize the effects of netting in two important ways. First, regulators calculate current exposure as the net (rather than the gross) amount of derivative assets. Second, potential future exposure, which is estimated as the notional amount of derivative contracts multiplied by a (small) percentage number called a conversion factor that varies by the type and the maturity of derivative contracts, is adjusted downward using the ratio of net derivative assets to gross derivative assets. 13 Because offsetable derivatives are typically much larger than net derivatives (on average, the latter is only 5.9%-7.4% of the former since see Figure 3), the standard 13 We compare the Basel Capital Accord of 1988 as amended in 1995 (Basel Committee on Banking Supervision 1995, pp. 3-6) and the Federal Reserve's risk-based capital guidelines as well as FR Y-9C reporting instructions related to the determination of the amount of capital banks should hold against derivative contracts, and find the former to be very similar to the latter. See sec pdf. See also pp of FR Y-9C reporting instruction for 1997 Q2 available at 13

15 version of regulatory capital ratios used by the U.S. banking regulators can be significantly higher than capital ratios that fully incorporate offsetable derivatives. For example, JP Morgan s tier 1 risk-based capital ratio of as of the end of 2009Q1 would decrease to 6.51 if offsetable derivatives were fully included in the ratio calculation (See Figure 4). Some officials from the FDIC have raised concerns that banks are not required to hold sufficient capital against the potential risks inherent in their derivative activities (Hoenig 2013; Norton 2014). To highlight these concerns, Thomas Hoenig (Vice Chairman of the FDIC) has been publishing on the FDIC website a Global Capital Index report for Global Systemically Important Banks, twice a year since 2012Q2. His reports highlight that the leverage ratio based on tangible capital (with higher ratios corresponding to better-capitalized banks) of the largest U.S. banks would have been markedly lower if the ratio incorporated derivative assets recognized in accordance with IFRS rather than U.S. GAAP. 14 These reports prompted some commentators to question the Federal Reserve s conclusion that the largest U.S. banks passed the annual stress tests, as the stress tests (which follow U.S. GAAP) allow banks to remove trillions of dollars of derivative assets and liabilities from their balance sheet, thus ignoring a key element of banks vulnerability See, e.g., columns (6) and (8) and footnote 5 of Mr. Hoenig s report for 2017Q2 at His other Global Capital Index biannual reports since 2012Q2 are available at 15 The New York Times Editorial Board cited Mr. Hoenig s adjusted capital ratios when they concluded that the largest U.S. banks that passed the stress tests would have been considered unsafe and unsound had their capital ratios been calculated in accordance with the much stricter offsetting criteria under IFRS (New York Times 2015). Additionally, some practitioners believe that the IFRS approach better captures banks derivative risk. For example, Fred Cannon, the head of research at investment bank KBW which specializes in working with financial firms, believe that the gross amount that the European balance sheets use is a better measurement of derivative risk. Do things the U.S. way and you are disguising things (Fortune 2013). A news article in Bloomberg, titled U.S. Banks Bigger Than GDP as Accounting Rift Masks Risk, has also highlighted the risks in offsetable derivatives (Bloomberg 2013). 14

16 3. Research Design and Data 3.1 Research Design Testing the Relation Between Offsetable Derivative Exposures and Default Risk If offsetable derivatives reflect a bank s exposure to market and counterparty credit risk (as discussed in sections 1 and 2.2), we would expect this to be manifest in a bank s risk of default. Because banks are required to disclose the amount of their gross derivatives in their FR Y-9C regulatory filings, the information on gross derivatives (and hence offsetable derivatives) is available to investors, thus allowing us to use a market-based measure of default risk the yield spreads of banks senior unsecured bonds (Blankespoor et al. 2013). We estimate the following model Log_Yield_Spreadi,j,t = α0 + α1 Offsetable_Exposurei,t-1 + α2 Net_Derivativesi,t-1 + αk Bank-level Control n i,t-1 + αm Bond-level Control n i,t-1 + Year-month Fixed Effects + ɛi,,t. (1) in which the dependent variable, Log_Yield_Spreadi,j,t, is the natural logarithm of the average yield spread of bond j of bank i in month t less the interest rate on U.S. Treasury bonds of similar maturity. Offsetable derivative exposures, Offsetable_Exposure, is calculated as the gross amount of derivative assets less the net amount of derivative assets recognized on the balance sheet, scaled by total assets. A positive α1 would suggest that banks with higher offsetable derivatives exhibit higher default risk. Equation (1) controls for the net amount of derivative assets recognized on the balance sheet scaled by total assets (Net_Derivatives). Equation (1) includes six other bank-level control variables often included in empirical models to explain banks credit risk: (1) return on assets (ROA); (2) loan quality measured as the sum of non-performing loans and net loan charge-offs scaled by total assets (NPL_NCO_Share); (3) asset-liability maturity gap measured as the scaled difference between assets and liabilities 15

17 maturing or repricing within one year (Maturity_Gap) following Flannery and Sorescu (1996); (4) capital adequacy (Capital_Ratio_Mkt) following Schwert (2018); (5) liquidity of the balance sheet (Liquid_Asset_Share) following Barth et al. (2012); (6) reliance on deposit funding (Deposit_Liab) following Ahmed et al. (2011). Equation (1) also controls for the natural logarithm of total assets (Size). Bond-level control variables include the natural logarithm of bond amount (Log_Bond_Amount) and the remaining maturity of the bonds (Years_to_Maturity). We control for year-month fixed effects, and cluster the standard errors by bank and quarter to account for repeated observations of the same bank over time Testing the Relation Between Offsetable Derivative Exposures and Systemic Risk As discussed in Section 2.2, in addition to default risk, offsetable derivatives may contain information about a bank s exposure to systemic risk. We explore this question using SRISK, a measure of a financial firm s exposure to systemic risk developed by Acharya et al. (2012) and further refined by Brownlees and Engle (2016). SRISK captures the ex-ante expected capital shortfall of a financial firm conditional on a crisis, and therefore captures a financial firm s exposure to systemic risk. 16 Specifically, SRISKi,t measures how much capital financial firm i would need at time t during a crisis to maintain a given capital-to-asset ratio (because of regulation and/or prudential management). SRISKi,t is empirically estimated using balance sheet and market information: 16 Banks are subject to prudential capital regulation. For example, they must maintain tier 1 risk-based capital ratio of at least 6% to be considered well capitalized. The SRISK measure captures the distance between the expected capital ratio of a bank during a severe market decline and a prudential capital ratio maintained by well managed large banks in normal time. For more details on SRISK, see Brownlees and Engle (2016) and Acharya et al. (2012). 16

18 SRISKi,t = Et-1(Capital Shortfalli Crisis) = Et-1(k (Debti + Equityi ) Equityi Crisis) (2) =k Debti,t-1 (1 k) (1 LRMESi,t ) Equity, where k is a prudential level of book-equity capital to asset ratio 17. LRMES is the long-run marginal expected shortfall at time t for bank i, defined as expected equity loss conditional on a crisis. SRISKi,t is a function of a bank s size, leverage, and market risk. Since SRISKi,t is the dollar amount of expected capital shortfall, we follow Berger et al. (2016) to normalize it by scaling it with market capitalization, and call this scaled variable Normalized SRISK (NSRISK). NSRISK is the proportional expected capital shortfall conditional on crisis. We use NSRISK as the dependent variable in the following model to investigate whether offsetable derivative exposures are positively associated with a bank s exposure to systemic risk: NSRISKi,t = β0 + β1 Offsetable_Exposure i,t-1 + β2 Net_Derivativesi,t-1 + βk Bank-level Control n i,t-1 + (3) + Year-month Fixed Effectsi + ɛi,t. The dependent variable, NSRISKi,t, captures bank i s systemic risk in month t. The main independent variable of interest is Offsetable_Exposurei,t-1. The coefficient β1 on Offsetable_Exposurei,t-1 captures the effect of offsetable derivatives on a bank s systemic risk conditional on a crisis. A positive β1 would suggest that banks with greater offsetable derivatives exhibit greater exposure to systemic risk conditional on a crisis. Equation (3) includes the same set of bank-level control variables as those in Equation (1). 3.2 Data and Sample To construct the bond sample for estimating Equation (1), we obtain all unsecured senior bonds issued by U.S. banks from the Mergent Fixed Investment Securities Database (FISD). We obtain bond characteristics (e.g., bond types, maturity date) from FISD. For these same bonds, we 17 K is set at 8% in the SRISK data we obtained from the V-Lab as described in Section 3.2 below. 17

19 obtain bond yields of secondary market trades from the Trade Reporting and Compliance Engine (TRACE) database. Since TRACE coverage begins in July 2002, our bond sample begins in Our bond sample ends in 2014, as the Enhanced TRACE bond database we use have an 18-month lag in data availability (Dick-Nielsen 2014). To ensure that bond yield spreads are solely driven by the issuer s creditworthiness (instead of reflecting special features of the bonds), we follow the literature to exclude bonds that are callable, puttable, convertible, exchangeable, redeemable, asset-backed or with other credit enhancements, or with variable coupon rates or with zero coupon (e.g., Goplan et al. 2014; Campbell and Taksler 2003). To enhance the homogeneity of the sample, we eliminate subordinated bonds and thereby include only senior bonds in the sample. Finally, we exclude bonds with remaining maturity of less than one year (Acharya et al. 2016). To clean the TRACE data on bond trades, we follow the literature to delete trades that were canceled, incorrectly recorded, or subsequently reversed in TRACE (Dick-Nielsen 2009; Dick-Nielsen et al. 2012; Jankowitsch et al. 2014; Becker and Ivashina 2015). If there are multiple trades for a bond in a day, we use the bond yield of the last trade of the day. If multiple trades occur in the last minute of a day, we take the average of the yields of these trades. We calculate bond yield spread as the bond yield minus maturity-matched Treasury rate, and take the natural logarithm of the average bond yield spread of a bond during a month as our dependent variable in Equation (1). 18 In addition to bond yield, we require that all bank-quarter observations have the required financial variables for the regression analysis, as described above in Section We obtain banks quarterly financial data from their regulatory FR Y-9C filings downloaded from the Federal Reserve Bank of Chicago website. The final bond sample for the default risk analysis includes 18 The Treasury rate for 1, 2, 3, 5, 7, 10, 20, and 30 year bonds is obtained from the U.S. Department of the Treasure website. Similar to Collin-Dufresne et al. (2001) and Blankespoor et al. (2013), we use linear interpolation to estimate the omitted maturities. 18

20 14,448 monthly bond yield observations across 525 senior bonds of 48 banks from 2002 through Our SRISK sample for testing Equation (3) regarding the relation between offsetable derivatives and banks systemic risk includes all large U.S. banks with market capitalization greater than $5 billion as of the end of June There are 35 such banks. We obtain daily SRISK data from New York University s V-Lab, the institute that publishes daily SRISK of domestic and global systemically important financial institutions (the director of V-Lab is Robert Engle, one of the researchers who developed the SRISK measure). We take the average of daily NSRISK during a month as our dependent variable in Equation (3). The final SRISK sample includes 3,697 monthly SRISK observations for 35 publicly traded large banks during the period. 3.3 Descriptive Statistics Figure 1 depicts the aggregate amount of offsetable derivative exposures in comparison to the aggregate amount of net derivative assets and the aggregate total assets of all U.S. bank holding companies (BHCs) in each year from 2002 to The aggregate offsetable derivatives experience a large increase in 2008 and range between $4.4 trillion and $6.9 trillion since then, which is a substantial amount given that the aggregate on-balance sheet total assets of all U.S. BHCs is $14-18 trillion. 19 The aggregate amount of net derivative assets is below $488 billion in all years. Figure 2 shows the ratio of the aggregate amount of net derivative assets to the aggregate amount of gross derivative assets for all BHCs; the ratio experiences a monotonic decrease from 2005 to 2012, and is below 10% in all years during the period except for In Figure 3, the ratio of the aggregate amount of net derivative assets to the aggregate amount of 19 The substantial increase in the aggregate amount of offsetable derivatives of all BHCs from 2007Q4 to 2008Q4 is partly because some large BHCs acquired some non-bhc financial institutions with large amounts of offsetable derivatives during Such acquisitions in 2008 include Bank of America s acquisition of Merrill Lynch, JP Morgan merge of Bear Stearns, and Wells Fargo s purchase of Wachovia. 19

21 offsetable derivatives for all BHCs exhibits a similar monotonic decrease from 2005 to 2013, and ranges between 5.9% and 7.4% since Table 1, Panel A reports descriptive statistics of the variables used in the default risk analysis. The mean (median) Log_Bond_Spread is 0.03 (0.02), corresponding to a mean (median) bond yield spread of 1.42 (1.02) percentage points. The mean (median) Offsetable_Exposure is 0.33 (0.09), which is many times larger than the net amount of derivative assets Net_Derivative whose mean (median) is 0.02 (0.02). Untabulated results show that Pearson correlation between Offsetable_Exposure and Log_Bond_Spread is (p < 0.001). Table 1, Panel B reports summary statistics of the variables used in the systemic risk analysis. The mean (median) NSRISK is 0.15 (-0.06), suggesting a right skewed distribution. The mean (median) Offsetable_Exposure for this sample is 0.10 (0.001). The mean (median) Net_Derivative is 0.01 (0.01). Untabulated results show that Pearson correlation between Offsetable_Exposure and NSRISK is (p < 0.001). 4. Empirical Results 4.1 The Relation Between Offsetable Derivative Exposures and Default Risk Table 2 reports the estimation of Equation (1) regarding the relation between offsetable derivatives and default risk. In column 1, where Net_Derivatives is not included in the regression, the coefficient on Offsetable_Exposure is significantly positive (0.575, p = 0.002). In column 2, where Offsetable_Exposure is not included in the regression, the coefficient on Net_Derivatives is also significantly positive (5.602, p = 0.060). In column 3 where both Offsetable_Exposure and Net_Derivatives are included in the regression, the coefficient on Offsetable_Exposure remains significantly positive (0.537, p = 0.003), while the coefficient on Net_Derivatives becomes statistically insignificant (p = 0.437). These results suggest that Offsetable_Exposure is positively 20

22 related to a bank s default risk, and its effect on banks default risk subsumes that of Net_Derivatives. Regarding the control variables, both profitability (ROA) and capital adequacy (Capital_Ratio_Mkt) are negatively associated with bond yield spreads, while poor loan quality (NPL_NCO_Share) is positively associated with bond yield spreads. Greater reliance on deposit financing (Deposit_Liab) is associated with lower bond yield spreads. Large banks (Size) have lower yield spreads, possibly reflecting investor expectation of government assistance to large banks (Acharya et al. 2016). Years to maturity is positively associated with bond yield spreads. 20 Table 2, column 4 reports the estimation of Equation (1) after adding bank fixed effects. Including bank fixed effects barely affects the magnitude of the coefficient on Offsetable_Exposure (0.492), although the statistical significance decreases slightly (p = 0.026). The slightly weaker results with firm fixed effects reflect the fact that most of the variation in Offsetable_Exposure arises in the cross-section rather than in the time series (probably because banks business models are relatively persistent), making firm fixed effects a poor match for our empirical setting (Roberts and Whited 2013; Coles et al. 2006; Zhou 2001). Overall, the results in Table 2 suggest that bond investors perceive that offsetable derivatives contain information about a bank s default risk that is incremental to the information contained in the net amount of derivative assets reported on the balance sheet. 4.2 Ruling out Alternative Explanations Related to Default Risk It is possible that the magnitude of a bank s offsetable derivatives somehow reflects aspects of bank risk that are unrelated to its derivative exposures. In particular, we consider the two 20 Amiram et al. (2018) find that the coupon choice is used as a signaling device to mitigate information asymmetry between lenders and borrowers about the credit quality of the borrower. In untabulated tests, we find that our results are robust to controlling for bond coupon rates. 21