Counterparty credit risk in credit default swap
market
☆Qin Jia
a *a
Faculty of Economics and Finance, University of Groningen, Groningen, 9741 JB, Netherlands Student number: 2350491 Supervisor: Jakob Bosma
Abstract
This study examines the counterparty credit risk involved in credit default swap (CDS) market. We investigate this issue with ordinary least squares using the average quotes contributed by Markit Group which collects the data from more than thirty CDS dealers. We find strong evidence that counterparty credit risk in CDS transactions is priced in most cases and the pricing effect indicated in empirical analysis is small on average. Moreover, pricing effect for each dealer differs in magnitudes and signs. Our results indicate that the small pricing effect results from influence of collateralization while the large pricing effect is due to the influence of some large complex financial institutions. Meanwhile, negative pricing effect in our analysis suggests a competent relation between dealers and references while positive effect indicates a contagion spillover among them.
Key words: Counterparty credit risk; Credit default swaps; Ordinary least squares; CDS spread; Competition and contagion effects
☆
1. Introduction
Counterparty credit risk is one of the major types of risks confronted by investors, firms and institutions in financial markets. The global financial crisis that erupted in 2007 has caused the U.S. financial markets to be increasingly volatile and then began to adversely influence other countries around the world. One of the main side effects of the crisis has been to reveal that the trading counterparties are not reliable (Blanchet-Scalliet, 2008). The fall of several major companies causes the concerns for counterparty credit risk in financial market to reach summit. On September 15 Lehman Brothers defaulted on its debt and swap obligations and filed for bankruptcy protection due to the fact that it suffered too large losses to survive under the adverse impact of the crisis. The next day the Federal Reserve declared 85 billion dollar loans to support the American International Group (AIG) since AIG had got caught in tight liquidity and suffered three consecutive quarters of net losses. Specifically, the losses of AIG in CDS businesses accumulated up to 25 billion dollars during the second quarter of 20081. Consequently investors of CDS derivatives began to pay more attention to the risks or moral hazards they may confront in transactions. This paper is mainly based on the work of Arora and Longstaff (2012) and discusses the counterparty credit risks that investors would face in credit default swap market. Like Arora and Longstaff we find that the counterparty credit risk is priced in CDS market; however, pricing effects in our analysis is comparatively larger and more volatile among different dealers. This paper also extends the paper of Jorion and Zhang (2007) and the results indicate that both contagion and competition effects are prevalent among the firms involved in our data set.
Generally there is no counterparty credit risk in transactions of exchange-traded derivatives since the role of exchange guarantees that the cash flows promised by the derivative contracts are regularly paid to the counterparties. Over-the-counter market is
1
often called off-exchange market since there is no administration or regulation to supervise the transactions from the exchanges. Counterparty credit risk in CDS transactions is thus specific to the over-the-counter markets and is a result of the fact that one counterparty of a CDS contract will default prior to the expiration of the contract and fail to make the full payments required by the contract (Pykhtin and Zhu, 2006). Generally speaking, in a standard contract, payments are made by the protection buyer semiannually or quarterly in arrears. If the reference firm defaults, there will be a final accrual payments covering the period from the previous payments to the default date and payments then stop (Hull and White, 2003). There exist two methods following to settle down the transactions: the protection sellers can buy the securities hold by the buyers at the notional amount, or the sellers pay the difference between the notional amount and the post-default market value of the bond by cash.
Transactions of CDS contracts became prevalent since 1998 when the International Swaps and Derivatives Association (ISDA) established the fundamental principles and terms for a standard CDS contract. During the first half year of 2008 the notional amount of outstanding CDSs reached as much as $58 trillions2. Later, many large corporations, Fannie Mae and Freddie Mac for example, fell into distress as a result of the deterioration of the financial market. Accordingly investors’ concerns about the Domino Effects of counterparty credit risks in CDS market reached unprecedented high. Distrust and suspicions on the CDS markets drove banks, investors and financial institutions to seek for more efficient and plausible regulatory principles or practices, e.g. creating a central clearing house for CDS transactions (Duffie and Zhu, 2009)3or trading the CDSs through an exchange. On the other hand governments start to interfere with the financial market and some bailout programs
2
The size of CDS trading on May 22, 2008 comes from the estimates by the Bank for International Settlements. See http://www.bis.org/press/p080522.htm
3
have been executed to rescue some too-big-to-fail4 companies. One of the main bailout programs is the Troubled Asset Relief Program (TARP)5 implemented by the U.S. government.
This paper is based on Arora and Longstaff (2012) and will focus on the different pricing effects from 13 specific dealers respectively. Arora and Longstaff (2012), however, concentrate their research on the overall pricing effects of 14 dealers and discuss the influences of each dealer on the pricing effect in their supplementary section. They find that the pricing effect of counterparty credit risk is supported even they control the impact of different industries and dealers. Their empirical analysis also displays that different trading volumes of dealers would not make any difference to the pricing effect, which is supported by Tang and Yan (2011). Tang and Yan (2011) mainly investigate what determines the change of CDS spread and find that aggregate trading volume is unrelated to CDS spread changes. This paper employs quotes of CDS spreads for 229 companies from 8 industries and 33 countries. Quotes for the references and dealers in our data set come from Markit Group and are contributed by more than thirty CDS dealers. The sample period ranges from January 2, 2004 to January 10, 2011, which is assumed to be appropriate for our research since it includes the typical year 2008 when the financial market is most volatile and gluts considerable counterparty default risks.
4
The too big to fail theory states that some financial institutions are so important and interconnected that their demise might be disastrous to the whole economy; therefore they should become recipients of beneficial policies from government and central banks and supported by government when they face difficulty. (en.Wikipedia.org)
5
There are several main results derived from our empirical analysis. First and foremost, counterparty credit risks from the 13 specific dealers are priced in the spreads at which the credit protection is sold. Secondly, the pricing effect for the counterparty credit risks varies distinctly in magnitude among different dealers. Thirdly, the impact of dealers’ credit risk on the protection price sold is small on average. Fourthly, some dealers (Credit Suisse, BNP, RBS, Citigroup, Goldman Sachs and Lehman) have positive pricing effect while some (UBS, Deutsche Bank, HSBC, BOA and JP Morgan) are subject to negative effect. Finally, dealers have bigger power than references in modeling the pricing effect in CDS market.
Thestructureof this paper is as follows: section 2 is an introduction for a standard CDS contract and the counterparty credit risks involved in CDS transactions; section 3 is a detailed description of the data, including the information about the data sources, the involved reference firms and the 13 specific dealers; section 4 is about the methodology employed in our regression and some essential transformation of our basic framework; the next section is empirical analysis and some economic intuitions behind the preliminary results. The final section is about conclusions and appendixes.
2. The credit default swap contract and corresponding counterparty credit risk
This part briefly introduces the credit default swap contract and explains how the counterparty credit risks are produced and can be mitigated.
2.1 A standard credit default swap contract
credit event; generally it pays a cash amount equal to the notional amount of the contract minus the prevailing market value of obligations of the reference firms (3) the reference firms pay (receive) nothing to (from) either counterparty. The transaction of credit default swap is distinguished from other credit guarantees or insurance in that it requires a credit event instead of a loss to trigger compensation from the protection seller.
2.2 Credit default swap spread
The rate of payments made per year from the protection buyer to the protection seller is known as the credit default swap spread (Hull and Predescu, 2004). In this paper, CDS spreads for the 13 specific dealers embody the counterparty credit risks from them respectively while spreads for the references refer to the credit risks from the references respectively. Spreads for the references are the average prices at which the dealers are willing to sell the credit insurances. A higher spread for a dealer denotes a higher counterparty credit risk from that dealer. A higher spread for a reference implies a higher risk from the reference and that buyers have to pay a higher premium to the sellers to get credit protection. This paper investigates the pricing effect of counterparty credit risk in CDS market by examining the relation between the spreads for the references and the spreads for 13 specific dealers.
2.3 What drives the counterparty credit risk?
There are three parties involved in a CDS contract: the buyer, the seller, and the reference. The logic behind the transaction is: the seller provides insurance for the buyer against a credit event from the reference. The counterparty involved in the transaction generally refers to the buyer and the seller: it is these two parties that sign a CDS contract. Counterparty credit risk refers to the risks originating from both parties and there are two possibilities: (1) the buyers default (2) the sellers default.
no loss, even though the buyers stop making the periodic payments. However, it is possible that buyers do not pay the required premiums regularly to the sellers but instead account this obligation as payables in their balance sheet. Under this circumstance the sellers might suffer a loss due to the fact that buyers fail to pay the premiums (payables).
the third party may seize and sell the collaterals. Therefore rehypothecation of collaterals would raise the counterparty credit risks from the dealers and drives the dealers’ CDS spreads higher.
2.4 How the counterparty credit risk can be mitigated?
Mitigation of counterparty credit risks becomes increasingly significant since the risks have already resulted in many investors to go through huge losses, especially during the 2007 global financial crisis.
The crisis erupted in 2007 has triggered an impetus to evolve from the less regulated over-the-counter derivative contracts (including CDSs) to the centralized counterparties (CCPs) rather than the bilateral clearing that has been widely used nowadays. CCP is an organization which interposes itself between the counterparties of CDS contracts and set two new contracts with each of the contracting parties. Applying CCPs is subject to both advantages and costs. Dufflie and Zhu (2010) argue that adding CCPs would reduce netting efficiency if we use separate CCPs for each derivative class and CCPs could also raise average counterparty exposures. Study conducted by Cecchetti and Gyntelberg (2009) find that CCPs would help manage counterparty credit risk by facilitating multilateral netting and improving transparency in financial market. But they also reveal that CCPs alone are not sufficient to ensure the resilience and efficiency of derivative markets when confronting large shocks. Manmohan Singh (2009) indicates in his paper that there need to be more research on the cost produced by offloading derivative contracts to CCPs. On the same time, some suggests to trade CDSs through exchange market instead of OTC market. On account of this view, Stulz (2009) discloses that exchange trading has both advantages and costs; over-the-counter CDS market works well during much of the first year of the credit crisis. Therefore transforming OTC CDSs to exchange-traded CDSs cannot guarantee that the credit risks will be mitigated.
provisions for potential losses. Through collateralization each party could control counterparty risk by requiring daily posting of collaterals which indicate marking-to-market changes in the value of CDS contract. In this way losses of each party would be controlled since posting collaterals could utmost maintain the market value of contracts. By the same token, CCPs require margins from its vast counterparties to cover their increased net liabilities by issuing margin calls when value of contracts decreases.
3. Data
There are two ways, as a common rule, to obtain the CDS spreads which are necessary for our research. The first way is to use the quotes of CDS spreads contributed by some major dealers. The second way is just to employ the actual CDS transaction prices executed in the trading. This paper uses the quotes instead of real transaction prices for several reasons:
Firstly, as CDS contracts are traded in over-the-counter market, most of the transactions are done directly between two counterparties and in forms that are custom-made to the investors. Therefore it may be difficult or even impossible to obtain the actual transaction price.
Secondly, the real transaction prices might be deviated and biased from their true values due to complicacies or heterogeneities of the trading. CDSs are actively traded in the over-the-counter market and lacks supervision from centralized exchanges. There is no guarantee for the liquidity, transparency or legitimacy of CDS trading. Therefore the actual transaction prices incorporate numerous unpredictable elements
and might fail to reflect the general perception towards CDS prices from the market. Quotes, on the other hand, are given by different dealers and represent equilibrium
in CDS market. We obtain from table 1 that all the 13 CDS dealers are actively making market in the sample period, which implies that the quotes contributed by each dealer should be competitive and executable.
14 major dealers respectively and there exist fourteen different prices for a reference firm in a specific date. This paper differs from Arora and Longstaff (2012) in that we use the average quotes which based on estimates from more than thirty dealers and thus there is only one quote for one reference in a specific date. This paper employ the average value since the individual prices for each reference from the dealers respectively are unavailable for us.
This paper employs data of quotes for CDS spreads provided by Markit Group from January 2, 2004 to January 10, 2011 and are based on the daily observations of CDS spreads in all trading days during the years. Markit collects the quotes, screens the quotes, removes some extremely deviated values and computes a daily composite spread only when more than two dealers contribute. We choose the quotes of spreads for CDSs with the maturity of five years6, which are practically more liquid and have been traded more frequently in CDS market. For each reference firm, we choose the CDS contract in the currency with the potentially highest liquidity7.
Among the thirty dealers which contribute to our quotes, this paper only investigates pricing effect of 13 specific dealers. Table 1 reports the summary statistics of the CDS spreads for these 13 dealers. As we can see, the average CDS quotes range from as low as 39.11 basis points of BNP to as high as 118.27 basis points of Morgan Stanley. Interestingly, standard deviations of the 13 dealers follow a similar pattern with the movements of spreads: BNP has the lowest deviation while Morgan Stanley has the highest. Note that we also include the quotes for Lehman Brothers despite it has already bankrupted. The reason is that during Jan 2, 2004 to Sep 15, 2008 Lehman was actively making markets in trading CDSs, and thus its quotes contain a great deal of information regarding the pricing effect of counterparty credit risks.
6
Most CDSs have maturities between 1 and 10 years; five years is the most typical maturity.
7
This data set contains CDS spreads for all the trading days from Jan 2, 2004 to Jan 10, 2011, during which the erupt of subprime mortgage crisis, the fall of Lehman and Bear Stearns, the distress of AIG, Fannie Mae and Freddie Mac are all included. This data set is thus quite informative for analyzing the pricing effect of counterparty credit risk.
Table 2 displays information about the firms underlying the CDS contracts. We obtain that there are 5 industries involved in our data. Among them bank sector has the most number of firms (153) with a percentage of 66.81%. Specifically, there is only one firm in financial service sector. Accordingly, bank sector has the largest CDS observations with a number of 187540 and financial service sector has the least (1786).
Table 1
Summary statistics for major dealers of CDS contracts
This table provides summary statistics for the CDS spreads (in basis point) of 13 specific CDS dealers. The spreads are based on the daily observations from Markit Group and are the average quotes which based on estimates of more than thirty dealers. N is the number of trading days which are available for the individual dealers from Jan 2, 2004 to Jan 10, 2011.
Dealer Mean Stand.dev. Min Median Max N
Barclays 59.21 60.22 5.46 13.97 261.94 1831 BNP 39.11 36.08 5.38 13.89 163.91 1831 Bank of American 74.98 74.71 4.87 48.20 383.54 1237 Citigroup 105.11 125.74 6.47 25.29 638.32 1831 Credit Suisse 68.70 54.20 8.95 66.46 261.43 1255 Deutsche Bank 54.16 45.92 8.92 21.63 190.00 1831 Goldman Sachs 88.94 84.69 17.23 39.00 579.29 1831 HSBC 48.43 39.98 4.97 51.67 171.59 1362 JP Morgan 58.31 43.38 10.87 44.96 227.33 1689 Lehman 68.22 84.00 17.20 30.69 675.63 1225 Morgan Stanley 118.27 140.16 16.55 41.12 1385.59 1831 Royal Bank of Scotland 70.32 72.72 3.97 14.61 315.50 1831
Table 2
The summary distribution of reference firms
This table provides information about the 229 firms underlying the CDS indexes in the data set. The panel in the left summarizes the information regarding the industry components of our data. The panel in the right exhibits observations for the spreads in each industry. N1 refers to the number of observations for reference firms in each industry. N2 refers to the number of observations for CDS spreads in each industry. Percentage1 and percentage 2 reveal the proportion of each sector in terms of the number of reference firms and spreads. The sample period is from Jan 2, 2004 to Jan 10, 2011.
Industry N1 Percentage1 N2 Percentage2 Bank 153 66.81% 187540 63.36% Financial Service 1 0.44% 1786 0.60% Financial Subsidiary 5 2.18% 6357 2.15% Insurance 39 17.03% 56073 18.95% Investment 4 1.75% 7214 2.44% Lease 5 2.18% 4139 1.40% Private Equity Investing 2 0.87% 2892 0.98% Real Estate Investment Trust 20 8.73% 29967 10.13%
229 1 295968 1
4. Method
4.1 Basic framework
Since CDS spreads in our data set are both time and cross section composite, we use panel regression8 for our analysis. In order to explore how the counterparty credit risk is reflected in the price of CDSs, we form a basic framework to investigate the relation between contract price and price movements of the references and dealers:
CDSi,t=
α
i+γCDSi,t-1+
13 1 j j βjCDS9j,t-1+ε
i,t (1) 8A panel incorporates information across both time and space; specifically, a panel keeps the same firms and measures their quantity over time.
9
With CDSi,t denotes the price of protection sold on firm i at time t;
α
iis a firm specific fixed variable; CDSi, t-1 denotes the spreads of firm i at time t-1; CDSj,t-1refers to the CDS spread of dealer j at the previous day of time t.
Framework (1) is based on the model of Arora and Longstaff (2012). We are going to analyze whether dealers’ credit risk is reflected in the prices at which dealers charge to provide credit protection. Our null hypothesis is that the dealers’ credit risk is not priced, which can be expressed in framework (1) as: the value of the slope coefficient
β
j is zero4.2 Fixed effect intuition
The main objective of our analysis is to obtain consistent estimates of the partial effects
β
j and γ. Apart from the variables presented in the right of framework (1),there are other factors which are heterogeneous but also cause changes in CDS spreads. We account for those factors by deriving fixed effects10
.Consistency of the estimates for coefficients
β
j and γ would be affected by those fixed effects. We modelfixed effect by allowing the constant term
α
i to vary across reference firms.Framework (1) replicates the model of Arora and Longstaff (2012), in which
α
i,t isexploited to model the firm specific fixed effects that would change across time. However, we get that constant
α
i in framework (1) will remain fixed across timedespite that it varies across references. This is why we have CDSi, t-1 to model the
variables which will change as time passes and are specific to the references.
10
4.3 First differencing, OLS and robust statistics
The most common method to conduct our analysis is to use the ordinary least squares (OLS). There is one prerequisite for applying OLS: the constant term
α
i can be observed for all references. However, we get from 4.2 that it is likelyα
i contains some fixed effects that is unobservable and correlates with the independent variables. We employ first differencing specification to eliminate the impact of fixed effects before applying OLS. With first differencing we can drop the impact of fixed effects out of our regression by adopting time-ranging differences of the variables in framework (1) to be our new variables, which is:△CDSi,t=γ△CDSi, t-1+
1 3 1 j j βj△CDSj,t-1+△ε
i,t (3) △CDSi,t=γ(CDSi, t-1-
CDSi, t-2)+
13 1 j j βj(CDSj,t-1-
CDSj,t-2)+ (ε
i, t-ε
i,t-1) (4)Framework (4) has eliminated constant term
α
i, which is likely to contain fixed effects. However, here comes one problem in framework (4): in its right composition,CDSi, t-1 is determined by
ε
i,t-1 byconstruction, which suggests that in framework (3)one of the independent variables correlates with the disturbance term.
E[△
ε
i,t-1ￜ△CDSi, t-1] ≠0There are several assumptions regarding classic linear regression models (OLS). One of the assumptions11 is about the disturbance terms:
E[△
ε
i,t-1ￜ△CDSi, t-1]= 0
This denotes there should be no relation between the disturbance terms and independent variables. Apparently framework (4) goes against this assumption and consequently, the slope coefficient βj would be biased:
E[βj]=βj+biasj,
11
Greece12(2005) states that if the time dimension of the panel grows large, for a fixed
number of reference firms, the bias would converge to zero in the limit as T →∞. The time series in our analysis is as large as 1831 days and the number of reference firms is only 230. Therefore we believe that the biases of our coefficientsβj and γ converge
towards zero:
limbiasj=0
We are now able to conduct our regression with OLS.
5. Empirical analysis
5.1. Is the counterparty credit risk priced?
In our null hypothesis the counterparty credit risk is not priced in CDS market. However, result of our empirical analysis shows that all the 13 dealers’ counterparty credit risks are priced except for Barclays and Morgan Stanley. As we can see in Table 3, the pricing effects on average are very small and vary distinctly among dealers: (1) the average coefficient of 13 dealers is 0.01038, which meanson average spread of the dealers has to increase by 96 basis points to result in one basis point increase in the price of protection sold (2) 11 dealers have significant slope coefficients; Barclays and Morgan Stanley have insignificant estimates (3) among the 11 dealers which have significant coefficients, UBS, Deutsche Bank, HSBC, Bank of America, JP Morgan have negative coefficients while Credit Suisse, BNP, RBS, Citigroup, Goldman Sachs and Lehman Brothers have positive coefficients (4) we can also obtain that the slope coefficient for CDSi,t-1 variable is 0.139, which denotes the price of CDSs at time t-1
has to increase by 7 basis points in order to cause one basis point increase in the price of CDSs at time t. There exists some persistence in the CDS spreads as time passes.
12
We make a comparison between specificationⅠandⅡin table 3. SpecificationⅠ investigates how firm specific variables will contribute to the cost of credit protection and we get that lagged CDS spreads will have a significant and negative pricing effect (-0.1023). This estimate, however, is unreliable as there exist omitted-variable bias (OVB): if we incorrectly leave out one or more important independent variables, the model will consequently overestimate or underestimate effect of other variables. Lagged CDS spreads in our data are contributed by more than thirty major dealers and therefore in table 3 all the variables are correlated with each other. Consequently, if we eliminate the 13 dealer variables (specificationⅠ) those correlations would be reflected in residual term and this would cause the coefficients for the lagged spreads to be biased. Comparatively, specificationⅡdoes a better job than specificationⅠin modeling the pricing effect: R-square of specificationⅡ(0.0540) is six times larger than R-square of specification Ⅰ (0.0085). We believe that framework (3) (specificationⅡ) in our analysis is appropriate and informative for modeling the pricing effects in CDS market.
Specifically, we observe that coefficient of the constant term is very significant and large. This may seem weird since our estimation is based on equation (3) and has dropped the constant terms
α
i. We can ignore this problem since the constant in table3 is systematically set in OLS regression and only represents a position in regression and thus has no economic intuition.
Table 3
Results from the regression of CDS spreads on the CDS spreads of the 13 dealers.
SpecificationⅡinvestigates how the dealers’ one-day lag spreads could determine the references’ spreads in the next day.
Ⅰ: △CDSi,t = γ△CDSi, t-1 + △εi,t Ⅱ: △CDSi,t = γ△CDSi, t-1 +
13 1 j j βj△CDSj,t-1 + △εi,tRegression specification Ⅰand Ⅱ
Variable Coefficient Ⅰ CoefficientⅡ Lagged CDSi,t-1 -0.1023** 0.1393***
[0.0493] [0.0425]
[0.0091] Morgan Stanley 0.0268 [0.0171] Constant 0.2605*** 0.1977*** [0.0662] [0.0303] Number of obs 293549 77409 R-squared 0.0085 0.0540 Mean 0.01038
5.2. Why magnitudes of some effect small while others large?
Since the slope coefficients differ in sign we employ the absolute value to describe the magnitude of the effect. In later part we will discuss the positive and negative pricing effects.
There are some intuitions behind the different pricing effects in our analysis. Some dealers are faced with small pricing effect due to the common practice of full collateralization in CDS market. Unlike interest rate swap, which only trades the interest rate, credit default swap exchanges the notional amount between sellers and buyers. Notional amount of CDSs are most in the $10-$20 million range. Therefore if one reference defaults, then generally the protection seller will have to pay a large amount of money to settle its obligation. The buyers then may worry that the seller will default on its obligation and fail to pay in full value. At the same time the sellers may worry that the buyers will default and fail to pay the premium regularly. Due to the bilateral nature of credit risk, both the sellers and buyers would require collaterals from the counterparty. Full collateralization refers to the practice of posing collaterals which worth the full value of its obligation. There is full collateralization if the counterparty credit risk is very large, which is always the case in transactions. The rule of collateralization ensures that if one party defaults, the other party can declare the possession of the collateral and thus to a large extent minimize its loss. Then the credit risk of one party will have a small effect on the other party. However, the practice of full collateralization is not always obeyed due to the complicacies originating from unequal positions emerged in transactions. We believe that dealers with small pricing effect have complied well with the practice of collateralization and get little privileges in their transactions.
counterparty. LCFIs may avoid their high quality collaterals to flow to other firms or even lock up collaterals in their hoarding of assets and consequently reduce the supply of collaterals (Singh and Aitken, 2009). BNP is the largest bank in France regarding its net revenues and the second largest bank in Europe referring its capital value (Baidubaike). The increasing credit risk of BNP indicates that its probability of default is also increasing, under which circumstance counterparties of BNP’s CDSs are more likely to suffer a loss in the case BNP fails to collateralize adequately. In this sense credit risks of BNP, HSBC, Credit Suisse and UBS will strongly influence the risks of their counterparties.
5.3. Why some effect positive while others negative?
The empirical analysis with our data presents two diverging results: there is a negative correlation between the protection price sold and the counterparty credit risks of UBS, Deutsche Bank, HSBC, Bank of America, JP Morgan; while there is a positive correlation for Credit Suisse, BNP, RBS, Citigroup, Goldman Sachs and Lehman Brothers. We will explore the intuitions behind the diverging results with reference to the credit correlations across industries.
There is competition effect between references and dealer UBS, Deutsche Bank, HSBC, Bank of America, JP Morgan. Negative correlations in this paper denote that the price of protection sold is higher if the spread of dealers is lower. This means credit risk of the references is lower when credit risk of the dealers is higher: the dealer and the reference are competing with each other. In general competition arises when firms are in competent relations and one will benefit from another’s deterioration in transactions. The competition effect can be explained by the supply and demand relations of one product in one particular industry: if one firm failed and there is a fixed demand for one product, the surviving firms in the same industry can capture the customers from the demised firms and gain bigger market power, which is always the case for the firms in the same sector and sell substitute goods of each other.
6. Conclusions
This paper explores the relation between the credit risk from the dealers and the price at which the dealers will sell the protection. The results from empirical analysis reveal that counterparty credit risk is priced in CDS market in most cases. Credit Suisse, BNP, RBS, Citigroup, Goldman Sachs and Lehman has positive pricing effect while UBS, Deutsche Bank, HSBC, BOA and JP Morgan are subject to negative pricing effect. Positive pricing effect suggests credit contagion across firms while negative pricing effect denotes competition among them. The magnitudes of the pricing effects vary among different dealers. Some dealers face small pricing effect due to the practice of collateralization in CDS market. Other dealers confront big pricing effect as they are large complex financial institutions (LCFIs) and could largely impact the market.
of 13 specific dealers despite that there are more than thirty dealers involved in our data set. Moreover, this paper does not provide a very detailed explanation for the diverging pricing effects from each of the 13 dealers.
There are some suggestions for future study regarding counterparty credit risks in CDS market. For example, our empirical study shows that moving of CDS spreads is partly due to changing credit risks from the dealers (as high as 19.54% from BNP) and partly the effects of the references’ lag spreads (10.23% ). What are the other factors that may contribute to the changes in CDS spreads? How those other factors correlate and differ with the dealers’ risk factor and references’ lag spreads factor? There are many such issues left to explore in our future study.
Appendix Ⅰ
The regression order for thirteen CDS dealers Dealer1 Credit Suisse
Dealer2 UBS
Dealer3 Deutsche Bank Dealer4 BNP
Dealer5 Barclays
Dealer6 Royal Bank of Scotland (RBS) Dealer7 HSBC
Dealer8 Bank of American (BOA) Dealer9 Citigroup
Appendix Ⅱ
Companies included in US TARP list (http://projects.propublica.org/bailout/list)
Last update: May 3, 2013
Name Total Disbursed AIG $67,835,000,000
American Express Company $3,388,890,000
Bank of American $45,000,000,000
Capital One Financial Corp $3,555,199,000
Citigroup.inc $45,000,000,000
Fannie Mae $116,149,000,000
Freddie Mac Federal Home Loan Mort $71,336,000,000
Goldman Saches Group $10,000,000,000
JPM organ Chase co $25,000,000,000
Lincoln National Corporation $950,000,000
Morgan Stanley $10,000,000,000
Sun Trust Bank inc $4,850,000,000
Wachovia Corp $238,889
Reference
Navneet Arora, Priyank Grandhi, Francis A. Longstaff, 2012. Counterparty Credit Risk and the Credit Default Swap Market. Journal of Financial Economics 103, 280-293
Philippe Jorion, Gaiyan Zhang, 2007. Good and Bad Credit Contagion: Evidence from Credit Default Swaps. Journal of Financial Economics 84 (2007) 860-883
Damiano Brigo, Kyriakos Chourdakis. Counterparty Risk for Credit Default Swaps: Impact of Spread Volatility and Default Correlation. International Journal of Theoretical and Applied Finance, October 3, 2008
Michele Fratianni, Francesco Marchionne, 2013. The Banking Bailout of the Subprime Crisis: was the bang worth the buck? Journal of International Financial Markets, Institutions and Money, Volume 23, Pages 240-264
Darrell Duffie and Haoxiang Zhu, 2011. Does a Central Clearing Counterparty Reduce Counterparty Risk? Forthcoming, Review of Asset Pricing Studies
Grinblatt, Mark, 1995. An Analytic Solution for Interest Rate Swap Spreads. Recent Work, Finance, Anderson Graduate School of Management, UC Los Angeles
John Hull and Alan White, 2000. Valuing Credit Default Swaps Ⅰ: no counterparty default risk.
John Hull and Alan White, 2000. Valuing Credit Default Swap Ⅱ: modeling default correlations. Jan Ericsson, Kris Jacobs, and Rodolfo Oviedo, 2009. The Determinants of Credit Default Swap Premia. Journal of Financial and Quantitative Analysis, Vol. 44, No. 1
Dragon Tang, Hong Yan, 2008. Liquidity and Credit Default Swap. Inquire Europe Symposium “Liquidity, Credit and Risk”
Caitlin Ann Greatrex, 2008. The Credit Default Swap Market’s Determinants. Discussion Paper No: 2008-05
John Wiley & Sons Ltd, 2010. Counterparty Credit Risk: The new challenge for global financial markets. ISBM: 978-0-470-68576-1
Jun Liu, Francis A. Longstaff, Ravit E. Mandell, 2002. The Market Price of Credit Risk: An empirical analysis of interest rate swap spreads. Working Paper 8990, http://www.nber.org/papers/w8990
Dragon Yongjun Tang, Hong Yan, 2011. What Moves CDS Spreads?
Pierre Collin-Dufresne, Robert S. Goldstein, and J. Spencer Martin, 2001. The Determinants of Credit Spread Changes. The Journal of Finance, vol. LVI, No. 6
Steven H. Zhu, Michael Pykhtin, 2006. Measuring Counterparty Credit Risk for Trading Products Under Basel Ⅱ. Basel II handbook, Michael Ong, ed., RISK Books
William H. Greene, Econometric Analysis, sixth edition, Chapter nine, p 182-190 Chris Brooks, Introductory Econometrics for Finance, second edition
Laura Kodres, Inci Otker-Robe, 2009. Collateral, Netting and Systemic Risk in the OTC Derivatives Market. Prepared by Manmohan Singh. IMF Working Paper.
Manmohan Singh, James Aitken, 2009. Counterparty risk, impact on collateral flows and role for central counterparties. IMF Working Paper No. 09/173.
