38430230 Counter Party Credit Risk and CDS[1]

8/3/2019 38430230 Counter Party Credit Risk and CDS[1]

1/24

COUNTERPARTY CREDIT RISK AND

THE CREDIT DEFAULT SWAP MARKET

Navneet Arora

Priyank Gandhi

Francis A. Longstaff

Abstract. Counterparty credit risk has become one of the highest-profile risksfacing participants in the financial markets. Despite this, relatively little is knownabout how counterparty credit risk is actually priced. We examine this issue bystudying the CDS spreads quoted by a broad cross-section of dealers selling pro-

tection on the same underlying firm. This unique data set allows us to identifydirectly how the credit risk of the dealer affects the prices of these controversialcredit derivatives. We find that counterparty credit risk is significantly priced inthe CDS market. The magnitude of the effect, however, is relatively modest andis consistent with a market structure in which participants require collateralizationof swap liabilities by counterparties. We also find that the pricing of counterpartycredit risk became much more significant after the Lehman bankruptcy. Surpris-

ingly, counterparty credit risk is significantly related to the credit protection spreadsoffered by U.S. dealers, but not to those offered by non-U.S. dealers.

Current draft: July 2009.


2/24

1. INTRODUCTION

During the past several years, counterparty credit risk has emerged as one of themost-important factors driving financial markets and contributing to the globalcredit crisis. Concerns about counterparty credit risk were significantly heightenedin early 2008 by the collapse of Bear Stearns, but then skyrocketed later in theyear when Lehman Brothers declared Chapter 11 bankruptcy and defaulted onits debt and swap obligations.1 Fears of systemic defaults were so extreme in the

aftermath of the Lehman bankruptcy that Euro-denominated CDS contracts on theU.S. Treasury were quoted at spreads as high as 100 basis points.

Despite the significance of counterparty credit risk in the financial markets,however, there has been relatively little empirical research about how it affects theprices of contracts and derivatives in which counterparties may default. This isparticularly true for the $57.3 trillion notional credit default swap (CDS) marketin which defaultable counterparties sell credit protection (essentially insurance) to

other counterparties.2 The CDS markets have been the focus of much attentionrecently because it was AIGs massive losses on credit default swap positions thatled to the Treasurys $182.5 billion bailout of AIG. Furthermore, concerns aboutthe extent of counterparty credit risk in the CDS market underlie recent proposalsto create a central clearinghouse for CDS transactions.3

This paper uses a unique proprietary data set to examine how counterpartycredit risk affects the pricing of CDS contracts. Specifically, this data set includesthe simultaneous CDS quotes provided by 15 large CDS dealers for selling protectionon the same set of underlying reference firms. Thus, we can directly measure theextent to which a CDS dealers credit risk as a counterparty affects the prices atwhich the dealer can sell credit protection. A key aspect of the data set is that itincludes most of 2008, a period during which fears of counterparty defaults in theCDS market reached historical highs. Thus, this data set provides an ideal samplefor studying the effects of counterparty credit risk on prices in derivatives markets.

1Lehman Brothers filed for Chapter 11 bankruptcy on September 15, 2008. Dur-ing the same month, American International Group (AIG), Merrill Lynch, FannieMae, and Freddie Mac also failed or were placed under conservatorship by the U.S.government.

2The size of the CDS market as of June 30 2008 comes from estimates reported by


3/24

Four key results emerge from the empirical analysis. First, we find that there isa significant relation between the credit risk of the dealer and the prices at which thedealer can sell credit protection. As would be expected, the higher the dealers creditrisk, the lower is the price that the dealer can charge for selling credit protection.This confirms that prices in the CDS market respond rationally to the perceivedcounterparty risk of dealers selling credit protection.

Second, although there is a significant relation between dealer credit risk andthe cost of credit protection, we show that the effect on CDS spreads is relatively

small. In particular, an increase in the dealers credit spread of 400 basis pointsonly translates into a one-basis-point decline on average in the dealers spread forselling credit protection. This small effect is an order of magnitude smaller thanwhat would be expected if swap liabilities were uncollateralized. In contrast, thesize of the pricing effect is very consistent with the standard practice among dealersof having their counterparties fully collateralize swap liabilities.

Third, because the Lehman bankruptcy in September 2008 was such a major

counterparty credit event in the financial markets, we examine how the pricing ofcounterparty credit risk was affected by this event. Surprisingly, we find that thereis little evidence that dealer credit risk was priced in the CDS market prior to theLehman bankruptcy; the significant relation between dealer credit spreads and thecost of credit protection appears to be almost entirely due to the period after theLehman default. This result is consistent with market sources indicating that theLehman default highlighted the importance of a number of counterparty credit risksthat had previously been largely ignored, such as the risk of posted collateral not

being segregated or even being rehypothecated, leaving some counterparties withonly an unsecured general claim on the bankruptcy estate.

Fourth, given that there are significant differences between the legal protectionsprovided counterparties in the U.S. versus other jurisdictions, we examine whetherthe pricing of counterparty credit risk is different for U.S. dealers than for non-U.S. dealers. The results indicate that not only is counterparty credit risk pricedonly after the Lehman default, it is priced only in the cost of protection offered byU.S. CDS dealers. Thus, there appears to be a major difference in the credit riskpricing behavior of U.S. and non-U.S. financial institutions. This is particularlysurprising since U.S. legal protections for clients and counterparties of U.S. dealers(such as hedge funds using the dealer as a prime broker) are generally viewed asbeing stronger. Furthermore, many of the U.S. dealers in the sample also receivedTARP f di f th U S T hi h ld b i d idi


4/24

netting as largely successful in addressing counterparty credit risk concerns. Thus,proposals to create a central CDS exchange, may not actually be effective in reduc-ing counterparty credit risk further.

This paper contributes to an extensive literature on the effect of counterpartycredit risk on derivatives valuation. Important research in this area includes Cooperand Mello (1991), Sorensen and Bollier (1994), Duffie and Huang (1996), Jarrowand Yu (2001), Hull and White (2001), Longstaff (2004, 2008), and many others.The paper most-closely related to our paper is Duffie and Zhu (2009) who study

whether the introduction of a central clearing counterparty into the CDS marketcould improve on existing credit mitigation mechanisms such as bilateral netting.They show that a central clearing counterparty might actually increase the amountof credit risk in the market. Thus, our empirical results support and complementthe theoretical analysis provided in Duffie and Zhu.

The remainder of this paper is organized as follows. Section 2 provides a briefintroduction to the CDS market. Section 3 discusses counterparty credit risk in the

context of the CDS markets. Section 4 describes the data. Section 5 examines theeffects of dealers credit risk on spreads in the CDS market. Section 6 summarizesthe results and presents concluding remarks.

2. THE CREDIT DEFAULT SWAP MARKET

In this section, we briefly review the basic features of a typical CDS contract. Wethen discuss the institutional structure of the CDS market.

2.1 CDS Contracts

A CDS contract is best thought of as a simple insurance contract on the eventthat a specific firm or entity defaults on its debt. As an example, imagine thatcounterparty A buys credit protection on Citigroup from counterparty B by payinga fixed spread of, say, 225 basis points per year for a term of five years. If Citigroupdoes not default during this period of time, then B does not make any payments toA. If there is a default by Citigroup, however, then B pays A the difference betweenthe par value of the bond and the post-default value (typically determined by asimple auction mechanism) of a specific Citigroup bond. In essence, the protectionb i bl t t th b d b k t th t ti b i th t f d f lt


5/24

2.2 The Structure of the CDS Market

Like interest rate swaps and other fixed income derivatives, CDS contracts aretraded in the over-the-counter market between large financial institutions. Duringthe past 10 years, CDS contracts have become one of the largest financial productsin the fixed income markets. As of June 30, 2008, the total notional amount ofCDS contracts outstanding was $57.325 trillion. Of this notional, $33.083 trillion iswith dealers, $13.683 trillion with banks, $0.398 trillion with insurance companies,$9.215 trillion with other financial institutions, and $0.944 trillion with nonfinancialcustomers.5

Early in the development of the CDS market, participants recognized the ad-vantages of having a standardized process for initiating, documenting, and closingout CDS contracts. The chartering of the International Swap and Derivatives As-sociation (ISDA) in 1985 led to the development of a common framework whichcould then be used by institutions as a uniform basis for their swap and derivativetransactions with each other. Currently, ISDA has 830 member institutions. These

institutions include virtually every participant in the swap and derivatives markets.As the central organization of the privately-negotiated derivatives industry, ISDAperforms many functions such as producing legal opinions on the enforceability ofnetting and collateral arrangements, advancing the understanding and treatment ofderivatives and risk management from public policy and regulator capital perspec-tives, and developing uniform standards and guidelines for the derivatives industry.6

3. COUNTERPARTY CREDIT RISK

In this section, we first review some of the sources of counterparty credit risk inthe CDS market. We then discuss ways in which the industry has attempted tomitigate the risk of losses stemming from the default of a counterparty to a CDScontract.

3.1 Sources of Counterparty Credit Risk

There are at least three ways in which a participant in the CDS market may sufferlosses when their counterparty enters into financial distress. First, consider thecase in which a market participant buys credit protection on a reference firm from


6/24

a protection seller. If the reference firm underlying the CDS contract defaults, theprotection buyer is then owed a payment from the counterparty. If the default wasunanticipated, however, then the protection seller could suddenly be faced with alarge loss. If the loss was severe enough, then the protection seller could potentiallybe driven into financial distress. Thus, the protection buyer might not receive thepromised protection payment.

Second, even if the reference firm underlying the CDS contract does not default,a participant in the CDS market could still experience a substantial loss in theevent that the counterparty to the contract entered financial distress. The reasonfor this is that while CDS contracts initially have value of zero when they areexecuted, their mark-to-market value may diverge significantly from zero over timeas credit spreads evolve. Specifically, consider the case where counterparty A hasan uncollateralized mark-to-market liability ofX to counterparty B. If counterpartyA were to enter bankruptcy, thereby canceling the CDS contract and making theliability immediately due and payable, then counterparty Bs only recourse wouldbe to attempt to collect its receivable of X from the bankruptcy estate. As such,

counterparty B would become a general unsecured creditor of counterparty A. Giventhat the debt and swap liabilities of Lehman Brothers were settled at only 8.625cents on the dollar, this could result in counterparty B suffering substantial lossesfrom the default of counterparty A.7

A third way in which a market participant could suffer losses through thebankruptcy of a counterparty is through the collateral channel. Specifically, considerthe case where counterparty A posts collateral with counterparty B, say because

counterparty B is counterparty As prime broker. Now imagine that the collateralis either not segregated from counterparty Bs general assets (as was very typicalprior to the Lehman default), or that counterparty B rehypothecates counterpartyAs collateral (also very common prior to the Lehman default). In this context,a rehypothecation of collateral is the situation in which counterparty B transferscounterparty As collateral to a third party (without transferring title to the col-lateral) in order to obtain a loan from the third party. Buhlman and Lane (2009)

argue that under certain circumstances, the rehypothecated securities become partof the bankruptcy estate. Thus, if counterparty B filed for bankruptcy after rehy-pothecating counterparty As collateral, or if counterparty As collateral was notlegally segregated, then counterparty A would become a general unsecured creditorof counterparty B for the amount of the collateral, again resulting in large poten-tial losses. An even more precarious situation would be when the rehypothecated


7/24

collateral itself was seized and sold by the third party in response to counterpartyBs default on the loan obtained using the rehypothecated securities as collateral.Observe that because of this collateral channel, counterparty A could suffer signif-icant credit losses from counterparty Bs bankruptcy, even if counterparty B doesnot actually have a mark-to-market liability to counterparty A stemming from theCDS contract.

3.2 Mitigating Counterparty Credit Risk

One of the most important ways in which the CDS market attempts to mitigate

counterparty credit risk is through the market infrastructure provided by ISDA. Inparticular, ISDA has developed specific legal frameworks for standardized masteragreements, credit support annexes, and auction, close-out, credit support, andnovation protocols. These ISDA frameworks are widely used by market participantsand serve to significantly reduce the potential losses arising from the default of acounterparty in a swap or derivative contract.8

Master agreements are encompassing contracts between two counterparties thatdetail all aspects of how swap and derivative contracts are to be executed, confirmed,documented, settled, etc. Once signed, all subsequent swaps and derivative trans-actions become part of the original master swap agreement, thereby eliminating theneed to have separate contracts for each transaction. An important advantage ofthis structure is that it allows all contracts between two counterparties to be nettedin the event of a default by one of the counterparties. This netting feature impliesthat when default occurs, the market value of all contracts between counterparties

A and B are aggregated into a net amount, leaving one of the two counterpartieswith a net liability to the other. Without this feature, counterparties might haveincentives to demand payment on contracts on which they have a receivable, butrepudiate contracts on which they have a liability to the defaulting counterparty.

Credit support annexes are standardized agreements between counterpartiesgoverning how credit risk mitigation mechanisms are to be structured. For example,a specific type of credit risk mitigation mechanism is the use of margin calls in which

counterparty A demands collateral from counterparty B to cover the amount ofcounterparty Bs net liability to counterparty A. The credit support annex specifiesdetails such as the nature and type of collateral to be provided, the minimumcollateral transfer amount, how the collateral amount is to be calculated, etc.

ISDA protocols specify exactly how changes to master swap agreements and


8/24

the increasing tendency among market participants to close out positions throughnovation rather than by offsetting positions motivated the development of the 2006

ISDA Novation Protocol II. Similarly, the creation of a standardized auction mech-anism for settling CDS contracts on defaulting firms motivated the creation of the2005-2009 ISDA auction protocols and the 2009 ISDA close-out amount protocol.

An important second way in which counterparty credit risk is minimized isthrough the use of collateralization. Recall that the value of a CDS contract candiverge significantly from zero as the credit risk of the reference firm underlying thecontract varies over time. As a result, each counterparty could have a significantmark-to-market liability to the other at some point during the life of the contract.In light of the potential credit risk, full collateralization of CDS liabilities has be-come the market standard. For example, the ISDA Margin Survey 2009 reportsthat 74 percent of CDS contracts executed during 2008 were subject to collateralagreements and that the estimated amount of collateral in use at the end of 2008was approximately $4.0 trillion. Typically, collateral is posted in the form of cashor government securities. Participants in the Margin Survey indicate that approx-

imately 80 percent of the ISDA credit support agreements are bilateral, implyingtwo-way transfers of collateral between counterparties. Of the 20 largest respon-dents to the survey (all large CDS dealers), 50 percent of their collateral agreementsare with hedge funds and institutional investors, 15 percent are with corporations,13 percent are with banks, and 21 percent are with others.

The data set used in this study represents the CDS spreads at which the largestWall Street dealers are willing to sell credit protection. Both discussions with CDS

traders and margin survey evidence indicate that the standard practice by thesedealers is to require full collateralization of swap liabilities by both counterpartiesto a CDS contract. In fact, the CDS traders we spoke with reported that thelarge Wall Street dealers they trade with typically require that their non-dealercounterparties overcollateralize their CDS liabilities slightly. This is consistent withthe ISDA Margin Survey 2009 that documents that the 20 largest firms accountedfor 93 percent of all collateral received, but only 89 percent of all collateral delivered,

suggesting that there was a net inflow of collateral to the largest CDS dealers.Furthermore, the degree of overcollateralization required can vary over time. As anexample, one reason for the liquidity problems at AIG that led to emergency loansby the Federal Reserve was that AIG would have been required to post additionalcollateral to CDS counterparties if AIGs credit rating had downgraded further.9

At fi t l th k t t d d f f ll ll t li ti t t


9/24

that there may be little risk of a loss from the default of a Wall Street credit protec-tion seller. This follows since the protection buyer holds collateral in the amount of

the protection sellers CDS liability. In actuality, however, the Wall Street practiceof requiring non-dealer protection buyers to slightly overcollateralize their liabilitiesactually creates a subtle counterparty credit risk. To illustrate this, imagine that aprotection buyer has a mark-to-market liability to the protection seller of $15 per$100 notional amount. Furthermore, imagine that the protection seller requires theprotection buyer to post $17 in collateral. Now consider what occurs if the protec-tion seller defaults. The bankruptcy estate of the protection seller uses $15 of the

protection buyers collateral to offset the $15 mark-to-market liability. Rather thanreturning the additional $2 of collateral, however, this additional capital becomespart of the bankruptcy estate. This implies that the protection buyer is now anunsecured creditor in the amount of the $2 excess collateral. Thus, in this situation,the protection buyer could suffer a significant loss even though the buyer actuallyowed the defaulting counterparty on the CDS contract.

This scenario is far from hypothetical. In actuality, a number of firms ex-

perienced major losses on swap contracts in the wake of the Lehman bankruptcybecause of their net exposure (swap liability and offsetting collateral) to Lehman.10

4. THE DATA

The data for the study consist of two different data sets. The first is publicly avail-able and includes daily quotations from the Bloomberg system for CDS contractsdirectly on the set of dealers from whom we will, in turn, obtain CDS quotations fora broader sample. Specifically, we include daily midmarket five-year CDS quotes for15 major CDS dealers during our sample period from March 31, 2008 to January20, 2009. The midmarket values for the CDS contracts reported by Bloombergcan be viewed as reflecting the average value of CDS spreads for these dealers inthe market. We will use these quotations to measure the degree of counterparty

credit risk inherent in the quotations provided by these 15 dealers in selling creditprotection on a broader set of firms.

Table 1 reports summary statistics for the CDS spreads for these 15 dealers.As shown, the average CDS spread ranges from a low of 59.40 basis points forBNP Paribas to a high of 355 10 basis points for Morgan Stanley Note that CDS


10/24

data for Lehman Brothers and Merrill Lynch are included in the data set eventhough these firms either went bankrupt or merged during the sample period. The

reason for including these firms is that both were actively making markets in sellingcredit protection through much of the sample period. Thus, their quotes maybe particularly informative about the impact of perceived counterparty credit riskon CDS quotations. The number of observations for each dealer varies since theBloomberg system does not provide quotations for each dealer for each day duringthe sample period.

The second data set is a proprietary data set provided by a major fixed-incomeasset management firm with trading ties to many of the 15 CDS dealers listedin Table 1. Specifically, this data set consists of daily quotes provided by these15 CDS dealers for selling five-year credit protection on the 125 individual firmsin the widely-followed CDX index. These quotes are provided via the Bloombergsystem to the asset management firm from the various CDS dealers throughoutthe trading day. These quotes represent firm offers to sell credit protection to ageneric counterparty in the asset management industry (quotes provided to multiple

counterparties via the Bloomberg system). Since these types of counterparties tendto have little leverage, it is realistic to assume that the CDS dealer views the creditrisk of the counterparty as relatively negligible (hence the willingness to providequotes to generic counterparties).

For each day in the sample period and for each of the firms in the CDX index,we extract quotes from the data set in the following way. First, we select 11:30AM as the reference time. For each of the 15 CDS dealers, we then include the

quote with time stamp nearest to 11:30 AM, but within 15 minutes (from 11:15to 11:45 AM). For many of these dealers, of course, there may not be a quotewithin this 30-minute time period. We repeat this process for all days and firmsin the sample. To ensure that the regression results are reliable, we only includethe 103 firms in the sample for which there at least 25 observations both prior toand after the Lehman bankruptcy in September 2009. Furthermore, for these firms,we only include observation dates in the sample for which there are two or more

contemporaneous quotations. This algorithm results in a set of 12,280 observationvectors of synchronous quotations by multiple CDS dealers for selling protectionon a common underlying reference firm. Since there are 212 days in the sampleperiod, this implies that we have data for multiple CDS dealers for an average of57.92 firms each day.

T bl 2 t t ti ti f th d t A h th b f


11/24

basis points, and the median range is 3 basis points.

5. EMPIRICAL ANALYSIS

In this section, we begin by briefly describing the methodology used in the empiricalanalysis. We then test whether counterparty credit risk is reflected in the pricesof CDS contracts. Next, we examine whether there was a change in the pricing

of counterparty credit risk after the Lehman bankruptcy filing in September 2008.Finally, motivated by the important differences in the legal protections provided toclients of U.S. and non-U.S. dealers, we test whether the pricing of counterpartycredit risk is different for the two types of CDS dealers.

5.1 Methodology

For each reference firm and for each date t in the sample, we have simultaneous

quotations from multiple CDS dealers for selling five-year credit protection on thatfirm. Thus, we can test directly whether counterparty credit risk is priced bya straightforward regression of the price of protection quoted by a dealer for thereference firm on the price of protection for the dealer itself providing that quotation.

To avoid the potential complexities of a full panel estimation, we will adoptthe more straightforward approach of simply estimating the regression separatelyfor each of the reference firms in the data set and including a vector of dummy

variables controlling for date-related fixed effects. Specifically, we estimate thefollowing regression for each reference firm,

CDSi,t = Ft + Si,t1 + i,t, (1)

where CDSi,t denotes the CDS spread quoted by the i-th dealer, is a vector ofregression coefficients, Ft is a fixed-effects vector with value one for the t-th element

and zero for all other elements, and Si,t1 is the CDS spread for the i-th dealer asof the end of the previous day.11 Under the null hypothesis that counterparty creditrisk is not priced, the slope coefficient is zero. The t-statistics for reported inthe tables are based on the White (1980) heteroskedastic-consistent estimate of thecovariance matrix.12


12/24

As shown in Table 2, there are a total of 12,280 observation vectors in thesample. On average, each observation vector consists of 3.089 distinct quotations for

selling credit protection on the reference firm, giving a total of 37,934 observationscollectively. Thus, there are 368.29 observations on average for each of the 103reference firms in the sample. In reporting the results from the regressions, ourapproach will be to provide summary statistics for the distribution of the 103 slopecoefficients and their corresponding t-statistics.

5.2 Is Counterparty Credit Risk Priced?

Although a formal model of the relation between a dealers credit risk and the priceat which the dealer could sell credit protection could be developed, the underlyingeconomics of the transaction makes it clear that there should be a negative relationbetween the two. Specifically, as the credit risk of a protection seller increases, thevalue of the protection being sold is diminished and market participants would notbe willing to pay as much for it. Thus, if counterparty credit risk is priced in themarket, the slope coefficient in the regressions should be negative.

Table 3 reports the summary statistics for the estimated coefficients and theirt-statistics. As shown, the results strongly support the hypothesis that counterpartycredit risk is priced in the CDS market. In particular, 70 percent of the estimated coefficients have negative values. A simple binomial test for the hypothesis thatpositive and negative values are equally likely rejects this hypothesis soundly; the

p-value for the binomial test is less than 0.0001.

The mean value of across regressions is 0.00138. Assuming independenceacross the reference firms, the t-statistic for the mean value of is

3.33. The

results from the distribution of t-statistics are similar to those for the coefficientsthemselves. The mean value of t is 0.8639, and the t-statistic for the mean is4.89. Thus, by any measure, there is strong evidence that counterparty credit riskis being priced in the cost of protection offered by the CDS dealers in the sample.

5.3 Why is the Effect so Small?

Although statistically very significant, the average size of the coefficients in Table3 is actually relatively small in economic terms. In particular, the mean value of0.00138 for the coefficient implies that the credit spread of a CDS dealer wouldhave to increase by nearly 725 basis points to result in a one-basis-point decline inthe price of credit protection. As shown in Table 1, credit protection on most of


13/24

CDS dealers in the sample never even reached 725 basis during the period understudy. These results are consistent with the results in Table 2 suggesting that the

cross-sectional variation in the dealers quotes for selling credit protection on aspecific reference firm is only on the order of several basis points.

A number of papers have explored the theoretical magnitude of counterpartycredit risk on the pricing of interest rate swaps. Important examples of this lit-erature include Cooper and Mello (1991), Sorensen and Bollier (1994), and Duffieand Huang (1996). Typically, these papers find that since the notional amount isnot exchanged in an interest rate swap, the effect of counterparty credit risk on aninterest rate swap is very small, often only a basis point or two.

Unlike an interest rate swap, however, a CDS contract could involve a verylarge payment by the protection seller to the protection buyer. For example, sellersof protection on Lehman Brothers were required to pay $91.375 per $100 notionalto settle their obligations to protection buyers. Thus, the results from the interestrate swap literature may not necessarily be directly applicable to the CDS market.

A few recent papers have focused on the theoretical impact of counterpartycredit risk on the pricing of CDS contracts. Important examples of these papersinclude Jarrow and Yu (2001), Hull and White (2001), Brigo and Pallavicini (2006),Kraft and Steffensen (2007), Segoviano and Singh (2008), and Blanchet-Scallietand Patras (2008). In general, estimates of the size of the effect of counterpartycredit risk in this literature tend to be orders of magnitude larger than those in theliterature for interest rate swaps. For example, estimates of the potential size of

the pricing effect range from 7 basis points in Kraft and Steffensen to more than20 basis points in Hull and White, depending on assumptions about the defaultcorrelations of the protection seller and the underlying reference firm. Thus, thisliterature tends to imply counterparty credit risk pricing effects many times largerthan those we find in the data.

It is crucial to recognize, however, that this literature focuses almost exclusivelyon the case in which CDS contract liabilities are not collateralized. As was discussed

earlier, the standard market practice during the sample period would be to requirefull collateralization by both counterparties to a CDS contract. This would beparticularly true for CDS contracts in which one counterparty was a large WallStreet CDS dealer.

In theory, full collateralization of CDS contract liabilities would appear to im-


14/24

of the excess collateral. As discussed earlier, however, Wall Street CDS dealers oftenrequire a small amount of overcollateralization from their counterparties (typically

on the order of several percent) thus creating the possibility of a slight credit loss(ironically, however, only when the counterparty owes the bankrupt firm money).Second, the Lehman bankruptcy also showed that there were a number of legalpitfalls which many market participants have not previously appreciated. These in-clude the risk of unsegregated margin accounts or the disposition of rehypothecatedcollateral.

In summary, the size of the counterparty pricing effect in the CDS marketappears much too small to be explained by models which abstract from the collat-eralization of CDS contracts. Rather, the small size of the pricing effect appearsmore consistent with the standard market practice of full collateralization, or evenovercollateralization, of CDS contract liabilities.

5.4 Did the Pricing of Counterparty Credit Risk Change?

The discussion above suggests that the Lehman bankruptcy event may have forced

market participants to reevaluate the risks inherent in even fully collateralizedcounterparty relationships. If so, then the pricing of counterparty credit after theLehman bankruptcy might differ from the pricing in the CDS market previous tothe bankruptcy.

To explore this possibility, we split the sample into the pre-Lehman and post-Lehman periods, and then estimate the regressions separately for each of the peri-ods. Table 4 reports summary statistics for the coefficients and their t-statistics.

Surprisingly, Table 4 shows that prior to the Lehman bankruptcy, there islittle or no evidence that counterparty credit risk was reflected in the pricing ofCDS contracts. In particular, only 51.5 percent of the coefficients are negativeduring the pre-Lehman period, and neither the mean value of nor the mean valueof t is statistically different from zero.

In contrast, the results indicate that counterparty credit risk is very strongly

priced in the post-Lehman period. The coefficient is negative for 70.9 percentof the regressions and a binomial test strongly rejects the hypothesis of equality ofproportions. The mean value of is 0.0025 and has a t-statistic of4.25. Thisimplies that a 400-basis-point increase in a dealers spread maps into a one-basis-point decline in the dealers price for selling credit protection. The mean value oft i 0 8839 d h t t ti ti f 5 37


15/24

there were numerous legal pitfalls inherent in even a collateralized counterpartyrelationship.

5.5 Is there a Difference between U.S. and Non-U.S. Dealers?

Continuing this theme, it is important to recognize that these types of legal pitfallsare not necessarily the same across all legal jurisdictions. For example, U.S. lawprovides much stronger protection to clients of prime brokers in regards to therehypothecation of collateral than does the U.K. In particular, a recent paper bySingh and Aitken (2009) explains that how the Securities Act of 1933, the Securities

Exchange Act of 1934, and the Securities Investor Protection Act of 1970 offerU.S. investors protections not available in the U.K. They argue that because ofthese statutory protections, customers of Lehman Brothers Inc. (U.S.) may betreated more advantageously that the customers of Lehman Brothers International(Europe). In light of the potential differences in legal structure and protection across

jurisdictions, we explore whether the pricing of counterparty credit risk depends onwhere the CDS dealer is headquartered.

Specifically, we estimate the following extended regression specification,

CDSi,t = Ft + US IUS Si,t1 + NUS INUS Si,t1 + i,t, (2)

where IUS is an indicator that takes value one if dealer i is headquartered in the

U.S. and zero otherwise, and vice versa for INUS for non-U.S. dealers. The dealersheadquartered in the U.S. are Bank of America, Citigroup, Goldman Sachs, JPMorgan, Lehman, Merrill Lynch, and Morgan Stanley. Thus, the sample is fairlyevenly split between U.S. and non-U.S. dealers. As in the previous section, weestimate the regression separately for the pre-Lehman and post-Lehman periods inthe sample.

Table 5 reports the summary statistics for the regression coefficients US and

NUS and their t-statistics. Surprisingly, the results show that counterparty creditrisk is significantly priced for the U.S. dealers in the post-Lehman period, but notfor the non-U.S. dealers. In particular, the mean value of US is 0.0022 with at-statistic of3.79. This implies that an increase of about 454 basis points in thecredit spread of a dealer maps into a one-basis-point decrease in price it offers forselling credit protection The mean value of t is 0 6959 with a t statistic of


16/24

These results are particularly remarkable given that there is some basis forthe view that legal protections in the U.S. for CDS counterparties may actually be

stronger than in other jurisdictions. Thus, these results definitely represent a puzzlethat bears further investigation in future research

6. CONCLUSION

We examine the extent to which the credit risk of a dealer offering to sell creditprotection is reflected in the prices at which the dealer can sell protection. We findstrong evidence that counterparty credit risk is priced in the market; the higherthe credit risk of a dealer, the lower is the price at which the dealer can sell creditprotection in the market. The magnitude of the effect, however, is relatively small.In particular, an increase in the credit spread of a dealer of 400 to 700 basis pointsmaps into a one-basis-point decline in the price of credit protection.

The price of counterparty credit risk appears to be too small to be explained bymodels that assume that CDS liabilities are unsecured. The pricing of counterpartycredit risk, however, seems consistent with the standard market practice of requiringfull collateralization, or even the overcollateralization of CDS liabilities. This viewappears supported by the evidence that counterparty credit risk was not priced inthe period prior to the Lehman default. Furthermore, finding that counterpartycredit risk is only priced after the Lehman default harmonizes with a number of

industry sources that the Lehman bankruptcy revealed potential weaknesses withexisting collateral protocols and/or legal protections. Thus, the pricing of counter-party credit risk in the CDS market is a relatively recent phenomenon deriving fromminor imperfections in a largely well-functioning system for mitigating counterpartyrisk exposure.

These results also have implications for current proposals about restructur-ing derivatives markets. For example, since market participants appear to price

counterparty credit risk as it were only a relatively minor concern, this suggeststhat attempts to mitigate counterparty credit risk through alternative approaches,such as the creation of a central clearinghouse for CDS contracts, may not be aseffective as might be anticipated. This implication parallels and complements theconclusions in the recent paper by Duffie and Zhu (2009).


17/24

REFERENCES

Blanchet-Scalliet, Christophette, and Frederic Patras, 2008, Counterparty Risk Valuationfor CDS, Working paper, University of Lyon.

Bliss, Robert, and George Kaufman, 2006, Derivatives and Systemic Risk: Netting, Collat-eral, and Closeout, Journal of Financial Stability 2, 55-70.

Brigo, Damiango, and Andrea Pallavicini, 2006, Counterparty Risk and Contingent CDSValuation under Correlation Between Interest Rates and Default, Working paper, Im-

perial College.

Buhlman, Robert, and Jason Lane, 2009, Counterpary Risk: Hard Lessons Learned, Prac-tical Compliance & Risk Management March-April, 35-42.

Cooper, Ian, and Antonio Mello, 1991, The Default Risk on Swaps, Journal of Finance 46,597-620.

Duffie, Darrell, and Ming Huang, 1996, Swap Rates and Credit Quality, Journal of Finance51, 921-949.

Duffie, Darrell and Haoziang Zhu, 2009, Does A Central Clearing Counterparty ReduceCounterparty Risk?, Working paper, Stanford University.

Hull, John, and Alan White, 2001, Valuing Credit Default Swaps II: Modeling DefaultCorrelations, Journal of Derivatives, Spring 8(3), 12-21.

ISDA Margin Survey 2009, International Swaps and Derivatives Association.

Jarrow, Robert, and Fan Yu, 2001, Counterparty Risk and the Pricing of Defaultable Secu-rities, Journal of Finance 56, 1765-1799.

Kraft, Holger, and Mogens Steffensen, 2007, Bankruptcy Counterparty Risk and Contagion,Review of Finance 11, 209-252.

Longstaff, Francis A., 2004, The Flight-to-Liquidity Premium in U.S. Treasury Bond Prices,Journal of Business 77, 511-526.

Longstaff, Francis A., 2008, The Subprime Credit Crisis and Contagion in Financial Mar-kets, Working paper, UCLA.

L t ff F i A S j Mith l d E i N i 2005 C t Yi ld S d D f lt


18/24

Segoviano, Miguel, and Manmohan Singh, 2008, Counterparty Risk in the Over-the-CounterDerivatives Market, Working paper 08/258, International Monetary Fund.

Singh, Manmohan, and James Aitken, 2009, Deleveraging after LehmanEvidence fromReduced Rehypothecation, Working paper 09/42, International Monetary Fund.

Sorensen, Eric and Thierry Bollier, 1994, Pricing Swap Default Risk, Financial AnalystsJournal 50 (May-June), 23-33.

White, Halbert, 1980, A Heteroskedasticity-Consistent Covariance Matrix Estimator and a

Direct Test for Heteroskedasticity, Econometrica 48, 817-838.


19/24

Table 1

Summary Statistics for CDS Contracts Referencing Dealers. This table provides summary statistics for the CDS spreads (in basis points)

for contracts referencing the dealers listed below. The spreads are based on daily observations obtained from the Bloomberg system.N

denotes thenumber of days on which Bloomberg quotes are available for the indicated dealer. The sample period is March 31, 2008 to January 20, 2009.

StandardDealer Mean Deviation Minimum Median Maximum N

Barclays 122.65 43.33 53.27 122.17 261.12 212

BNP Paribas 59.40 13.29 34.24 59.08 107.21 212Bank of America 121.60 35.77 61.97 119.75 206.85 209Citigroup 180.67 71.13 87.55 162.90 460.54 207Credit Suisse 111.66 37.20 57.59 101.40 194.22 212Deutsche Bank 96.88 29.70 51.92 90.11 172.00 212Goldman Sachs 230.58 110.62 79.83 232.69 545.14 177HSBC 75.41 21.94 41.84 67.59 128.30 212JP Morgan 110.86 27.96 62.54 107.68 196.34 209Lehman 291.79 89.01 154.04 285.12 641.91 84

Merrill Lynch 243.19 71.34 114.35 218.43 472.72 193Morgan Stanley 355.10 236.22 108.06 244.98 1360.00 187Royal Bank of Scotland 116.45 45.16 55.17 110.69 304.89 212Societe Generale 87.13 20.86 43.17 88.43 162.66 212UBS 139.09 56.81 55.45 126.24 320.80 212


20/24

Table 2

The Distribution of Dealer Quotes. This table provides summary statistics for the distribution of dealer quotes for CDS contracts referencing

the firms in the CDX index. The panel on the left summarizes the distribution in terms of the number of dealer quotes on a given day for a CDScontract referencing a specific firm. The panel on the right summarizes the distribution in terms of the range of dealer quotes (R measured in basispoints) on a given day for a CDS contract on a specific reference firm. Only days on which two or more simultaneous quotes are available for a specificfirm are included in the sample as an observation. The sample period is March 31, 2008 to January 20, 2009.

Number Range of of Quotes Observations Percentage Quotes Observations Percentage

2 4343 35.37 0 1069 8.703 4213 34.31 0 < R 1 1877 15.294 2450 19.95 1 < R 2 2175 17.715 950 7.73 2 < R 3 1834 14.936 233 1.90 3 < R 4 988 8.047 70 0.57 4 < R 5 1612 13.138 18 0.15 5 < R 10 1968 16.039 3 0.02 10 < R 20 615 5.01

20 < R 142 1.16

Total 12280 100.00 Total 12280 100.00


21/24

Table 3

Summary Statistics for the Distribution of Results from the Regression of CDS Quotations for Individual Firms on the CDS

Spreads of the Dealers providing the Quotations. This table reports the indicated summary statistics for the distribution of the coefficient and its White heteroskedasticity-consistent t-statistic from the following regression, where CDSi,t is the CDS quotation by dealer i for the underlyingfirm as of date t, is a vector of regression coefficients, Ft is the fixed-effect vector (a vector with t-th element one and zero for all other elements),and Si,t1 is the CDS spread for dealer i as of date t 1. The distribution of estimates is obtained by estimating the regression separately for eachof the 103 firms in the sample. The sample period for each regression is March 31, 2008 to January 20, 2009.

CDSi,t = Ft + Si,t1 + i,t

Range of the Range of the Coefficient Number Percentage t-Statistic for Number Percentage

0.0100 4 3.89 t 1.98 28 27.190.0100 < 0.0075 2 1.94 1.98 < t 1.67 3 2.910.0075 < 0.0050 9 8.74 1.67 < t 1.00 17 16.500.0050 < 0.0025 22 21.36 1.00 < t 0.50 14 13.590.0025 < 0.0000 35 33.98 0.50 < t 0.00 10 9.71

0.0000 < 0.0025 17 16.50 0.00 < t 0.50 8 7.770.0025 < 0.0050 8 7.77 0.50 < t 1.00 7 6.800.0050 < 0.0075 2 1.94 1.00 < t 1.67 7 6.800.0075 < 0.0100 1 0.97 1.67 < t 1.98 4 3.88

> 0.0100 3 2.91 t > 1.98 5 4.85

Fraction Negative 0.699 Fraction Negative 0.699Mean of 0.00138 Mean of t 0.8639

Std. Deviation 0.00419 Std. Deviation 1.7918t-Statistic for Mean 3.33 t-Statistic for Mean 4.89


22/24

Table 4

Summary Statistics for the Distribution of Results from the Subperiod Regressions of CDS Spreads for Individual Firms onthe CDS Spreads of the Dealers Providing the Quotations.

This table reports the indicated summary statistics for the distribution of thecoefficient and its White heteroskedasticity-consistent t-statistic from the following regression, where CDSi,t is the CDS quotation by dealer i forthe underlying firm as of date t, is a vector of regression coefficients, Ft is the fixed-effect vector (a vector with t-th element one and zeros for allother elements), and Si,t1 is the CDS spread for dealer i as of date t 1. The distribution of estimates is obtained by estimating the regressionseparately for each of the 103 firms in the sample. The sample period for the pre-Lehman sample is March 31, 2008 to September 14, 2008. Thesample period for the post-Lehman sample is September 15, 2008 to January 20, 2009.

CDSi,t = Ft + Si,t1 + i,t

Pre-Lehman Post-Lehman Pre-Lehman Post-Lehman

Range of the Range of theCoefficients Number Percentage Number Percentage t-Statistics Number Percentage Number Percentage

0.0100 8 7.77 5 4.85 t 1.98 23 22.33 26 25.250.0100 < 0.0075 2 1.94 9 8.74 1.98 < t 1.67 1 0.97 6 5.83

0.0075 < 0.0050 8 7.77 10 9.71 1.67 < t 1.00 9 8.74 17 16.500.0050 < 0.0025 13 12.62 21 20.39 1.00 < t 0.50 10 9.71 16 15.530.0025 < 0.0000 22 21.35 28 27.19 0.50 < t 0.00 10 9.71 8 7.77

0.0000 < 0.0025 21 20.39 20 19.42 0.00 < t 0.50 12 11.65 12 11.650.0025 < 0.0050 18 17.48 5 4.85 0.50 < t 1.00 11 10.68 4 3.880.0050 < 0.0075 4 3.88 1 0.97 1.00 < t 1.67 13 12.62 9 8.740.0075 < 0.0100 6 5.83 1 0.97 1.67 < t 1.98 2 1.94 0 0.00

> 0.0100 1 0.97 3 2.91 t > 1.98 12 11.65 5 4.85

Fraction Negative 0.515 0.709 Fraction Negative 0.515 0.709Mean of 0.0007 0.0025 Mean of t 0.2570 0.8839

Std. Deviation 0.0057 0.0058 Std. Deviation 2.0396 1.6720t-Statistic for Mean 1.34 4.25 t-Statistic for Mean 1.28 5.37


23/24

Table 5

Summary Statistics for the Distribution of Results from the US and Non-US Subperiod Regressions of CDS Spreads for IndividualFirms on the CDS Spreads of the Dealers Providing the Quotations. This table reports the indicated summary statistics for the distributionof the coefficients US and NUS and their White heteroskedasticity-consistent t-statistics from the following regression, where CDSi,t is the CDSquotation by dealer i for the underlying firm as of date t, is a vector of regression coefficients, Ft is the fixed-effect vector (a vector with t-th elementone and zeros for all other elements), Si,t1 is the CDS spread for dealer i as of date t 1, and IUS and INUS are indicator variables that take value1 if the dealer is headquartered in the U.S. and not headquartered in the U.S., respectively, and zero otherwise. The distribution of US and NUSestimates is obtained by estimating the regression separately for each of the 103 firms in the sample. The sample period for the pre-Lehman sampleis March 31, 2008 to September 14, 2008. The sample period for the post-Lehman sample is September 15, 2008 to January 20, 2009.

CDSi,t = Ft + US IUS Si,t1 + NUS INUS Si,t1 + i,t



US 0.0100 10 9.71 8 7.77 tUS 1.98 23 22.33 21 20.390.0100 < US 0.0075 5 4.85 4 3.88 1.98 < tUS 1.67 2 1.94 9 8.740.0075 < US 0.0050 7 6.80 10 9.71 1.67 < tUS 1.00 8 7.77 19 18.450.0050 < US 0.0025 16 15.53 17 16.50 1.00 < tUS 0.50 8 7.77 13 12.620.0025 < US 0.0000 14 13.59 33 32.05 0.50 < tUS 0.00 11 10.68 10 9.71

0.0000 < US 0.0025 18 17.48 19 18.45 0.00 < tUS 0.50 8 7.77 12 11.650.0025 < US 0.0050 15 14.57 5 4.85 0.50 < tUS 1.00 12 11.65 5 4.850.0050 < US 0.0075 9 8.74 3 2.91 1.00 < tUS 1.67 13 12.62 10 9.710.0075 < US 0.0100 5 4.85 1 0.97 1.67 < tUS 1.98 4 3.88 0 0.00

US > 0.0100 4 3.88 3 2.91 tUS > 1.98 14 13.59 4 3.88

Fraction Negative 0.505 0.699 Fraction Negative 0.505 0.699Mean of US 0.0010 0.0022 Mean of tUS 0.2595 0.6959



24/24

Table 5 Continued



NUS 0.0100 22 21.36 23 22.33 tNUS 1.98 17 16.51 4 3.880.0100 < NUS 0.0075 7 6.80 3 2.91 1.98 < tNUS 1.67 7 6.79 5 4.850.0075 < NUS 0.0050 4 3.88 3 2.91 1.67 < tNUS 1.00 7 6.79 11 10.680.0050 < NUS 0.0025 8 7.77 5 4.86 1.00 < tNUS 0.50 7 6.79 10 9.710.0025 < NUS 0.0000 11 10.68 13 12.62 0.50 < tNUS 0.00 14 13.60 17 16.51

0.0000 < NUS 0.0025 6 5.82 17 16.51 0.00 < tNUS 0.50 7 6.79 26 25.250.0025 < NUS 0.0050 6 5.82 11 10.68 0.50 < tNUS 1.00 14 13.60 12 11.650.0050 < NUS 0.0075 10 9.71 8 7.77 1.00 < tNUS 1.67 12 11.65 10 9.710.0075 < NUS 0.0100 10 9.71 3 2.91 1.67 < tNUS 1.98 6 5.83 5 4.85

NUS > 0.0100 19 18.45 17 16.50 tNUS > 1.98 12 11.65 3 2.91

Fraction Negative 0.505 0.456 Fraction Negative 0.505 0.456Mean of NUS 0.0011 0.0012 Mean of tNUS 0.0633 0.0193


Documents

38430230 Counter Party Credit Risk and CDS[1]