Download - HEDGE FUND - CDS

8/13/2019 HEDGE FUND - CDS

1/27

CREDIT DEFAULT SWAPTIONS

Alan L. Tucker, Ph.D. 1 Associate Professor of Finance

Lubin School of BusinessPace University

Jason Z. WeiAssociate Professor of FinanceRotman School of Management

University of Toronto

April 26, 2005

1Contact author: Alan L. Tucker, Department of Finance, Lubin School of Business, Pace University, 1Pace Plaza, New York, NY 10038, 212-618-6524 (voice), 212-346-1673 (fax), [email protected] .
mailto:[email protected]:[email protected]


2/27


Abstract

Credit derivatives were arguably invented by Bankers Trust (now part of Deutsche Bank)in 1991, with the product market not taking off until 1996 due to a period of tight creditspreads and generally favorable credit market conditions witnessed during the first half ofthe 1990s. Product education, advances in pricing, and more adverse credit events duringthe last decade have served to accelerate market growth. Indeed, the credit derivativesmarket is widely regarded as the fastest growing sector of the derivatives industry andnow exhibits over $5 trillion in average outstanding notional principal worldwide. Creditdefault swaps (CDSs) account for approximately 72.5% of the marketplace, with theremaining 27.5% spread mostly across credit spread swaps, total rate of return swaps, andcredit spread options. Options on credit default swaps known as CDS swaptions have

only recently become popular among end users. CDS swaptions come in two generalvarieties: Calls and puts written on CDSs, and cancelable CDSs. A cancelable CDScontains an embedded option to terminate an existing CDS (an embedded CDSswaption). This paper describes credit default swaptions, provides illustrations of theiruses, for example, in creating synthetic collateralized debt obligations, and presents andillustrates valuation models. The pricing models offered here are more accessible thanthose presented in the working papers of Sch nbucher (2000), Jamshidian (2002) and

Schmidt (2004).

Keywords: Credit default swaps, swaptions, option pricing

JEL Classifications: G13

2


3/27


1. Introduction

Credit derivatives have been traded since 1991 and credit default swaps (CDSs) account

for the vast majority of trading. Only recently have end users begun to take interest in

options on CDSs. Analogous to interest rate swaptions for the interest rate marketplace,

credit default swaptions represent a potentially important derivative product for credit

markets. Indeed, a CDS that is cancelable contains an embedded credit default swaption.

A cancelable long CDS position (where long means that the CDS trader is paying a

fixed swap rate and is thus the buyer of credit protection) is simply a package of a

straight (read non-cancelable) long CDS plus a put-style CDS swaption an option to

enter a CDS short and thus close the already outstanding long position. 2 A cancelable

short CDS represents a combination of a short position in a straight CDS plus a call-style

CDS swaption. To the extent that existing CDSs are cancelable and most are in

practice then ignoring the value of the embedded CDS swaption can lead to pricing

errors and thus arbitrage opportunities. 3 In fact, we opine that methods used to establish

initial swap rates on cancelable CDSs, as well as methods used to value seasoned CDSs

2 See, for example, Hull (2003, Chapter 27), for a discussion of straight credit default swaps.3 CDSs are commonly cancelable because they are written on a particular reference credit asset, forexample, a junk bond. To reverse-trade a CDS without an embedded option to cancel, the trader wouldhave to find another counterparty willing to execute a CDS on the particular reference credit asset. This is

plausible, but may not be realistic depending on the nature (read liquidity) of said asset. This contrastswith say, an interest rate swap, wherein a trader can readily reverse trade and close an outstanding position

because the underlying is a generic variable, for example, the s.a. $LIBOR.

3


4/27

that are cancelable, typically ignore the embedded swaption to terminate the position, and

thus these CDSs may be mis-priced. 4

The purpose of this paper is three-fold: to describe CDS swaptions; to illustrate some of

their applications; and, most importantly, to present valuation models which are

accessible to the reader. Sections 2, 3 and 4 address these goals, respectively. Section 5

offers a brief conclusion and suggestions for future research.

2. Product description We describe CDS swaptions through an example. Assume that all counter parties

(dealers and buy side) are AA-rated, either because they are already AA-rated or because

they have been credit enhanced to AA through collateral agreements, mid-market

agreements, netting agreements, and other well known credit enhancement techniques.

Assume that the CDS which underlies the swaption has a 3-year maturity, semi-annual

payment dates, and a swap rate (the strike rate on the swaption) of 150 basis points (bps).

The strike rate assumes semi-annual compounding the same periodicity (or tenor) of the

CDS. The credit default swaps underlying reference credit asset is a BB-rated 10-year,

8%-coupon bond with $100 million par. The CDS swaption is a call, European-style,

with a maturity of 6 months. Thus the CDS swaption owner has the right, in 6 months, to

enter the underlying CDS long, that is, paying 150 bps.

Suppose that in 6 months, when the swaption matures, the bid-offer swap rates on newly-

minted 3-year credit default swaps (with semi-annual tenors) on the same reference

4 However, see the discussion below regarding pari passu assets.

4


5/27

credit asset (or pari passu asset) are 200 bps by 220 bps. 5 So, the underlying bond has

exhibited credit deterioration, for example, having been downgraded to a weak single B.

The call swaption is exercised, meaning that its owner can now long the same swap

paying just 150 bps. By engaging in a reversing trade, that is, entering a short CDS, the

swaption owner locks into an annuity of 50 bps (the bid of 200 bps less the strike rate of

150 bps) on $50 million for the next 6 semi-annual periods. This annuity is present

valued (monetized) at the interest rate swap mid-rate on a newly-minted 3-year

seminal annual (s.a.) $LIBOR swap since both counter parties are AA-rated.

If the swaption is a put and at expiry newly-minted CDS rates are 100 bps by 110 bps

(perhaps because the bond is now a weak single A), then the payoff to the CDS swaption

would be the present value (again, discounted at the 3-year interest rate swap mid-rate) of

six annuity payments of $50 million times 40 bps (the strike rate of 150 bps less the offer

of 110 bps). 6

CDS swaptions that are traded outright are likely to be European-style. However, a

cancelable CDS will contain either an American- or, more likely, Bermudian-style

5 Commonly, a CDS that is physically settled requires the long trader to deliver (read transfer ownership)to the short the reference credit asset, or an equivalent asset known as a pari passu asset. The short traderthen pays the long the face value of the reference credit asset (the notional on the CDS). A cash-settledCDS entails the short trader paying the long the difference between the face value and the post-defaultvalue of the reference credit asset, where said value is determined by a calculation agent. The agent

typically ascribes a value by taking the mean of the bid and offer prices quoted by dealers of the referencecredit asset. CDS dealers tend to prefer physical settlement in order to work the reference credit asset,that is, because they feel they can obtain better value than indicated by the calculation agent. Importantly,notice that the ability to trade pari passu assets, and other CDSs on pari passu assets, tends to mitigate thevalue of the embedded swaption to terminate an existing CDS. In other words, to trade pari passu assetsserves to give the CDS greater secondary market liquidity, a la an interest rate swap written on a genericreference rate such as seminal annual $LIBOR.6 In these illustrations we ignore the day count convention, that is, we assume that markets operatecontinually and time can be carved into perfect one-half year intervals. The usual day count conventionfor a CDS or CDS swaption is Actual/360.

5


6/27

swaption. For example, consider a long CDS giving the buyer of credit protection the

option to terminate the swap every six months. Assume that the CDSs underlying

reference credit asset is unique, illiquid, and has no pari passu substitutes. Then this

CDS represents a package of a straight CDS plus a potentially valuable Bermudian-style

put swaption the ability to short the CDS, at six-month intervals, thus closing the

original long position.

Besides plain-vanilla CDS swaptions whether they are American, Bermudian,

European, calls, puts, outright, or embedded in cancelable CDSs there can exist ofcourse a variety of more exotic CDS swaptions. For instance, there can be swaptions

written on binary and basket CDSs. There can be barrier CDS swaptions. And so on. It

should be interesting to witness changes in the market for CDS swaptions per se as the

market for credit derivatives in general continues to grow and re-invent.

2.1. Default-triggering exercise for outright European-style CDS call swaptions

Suppose that the reference credit asset (our 8%-coupon bond) has a default-triggering

event (such as a missed coupon date) prior to the 6-month maturity of the swaption. This

would have the effect of terminating the CDS (physical or cash settlement), and therefore

presents the problem of having a CDS swaption with no existing underlying. If the CDS

swaption is a put, then the issue is moot; the put swaption owner would not want to

exercise, because the credit spread on the defaulted bond would presumably explode to

something well above the original (150 bps) strike rate. However, the call swaption

owner would want to exercise. The call owner therefore needs a mechanism to capture

6


7/27

value. Hence, European-style CDS call swaptions contain legalese that permits for early

exercise in the event that the reference credit asset exhibits a default-triggering event

prior to the maturity of the CDS swaption. 7

In the event that the CDS call swaption is exercised early due to default of the reference

credit asset, then the swaption owner should be required to make a payment to the writer.

Said payment represents a type of premium accrual on a default insurance policy written

on the reference credit asset. For example, for our illustration, if at inception of the

swaption there exists a 6-month bond insurance policy that pays the difference betweenthe face value and the recovery value of the bond, and said policy has a cost of 10 bps of

face value, then, assuming that the reference credit asset defaults mid-way through the

life of the swaption (and ignoring day count conventions), the CDS call swaption owner

would be required to pay 5 bps of $100 million. 8

Note that the existence of a pari passu provision on the underlying CDS does not obviate

the need for the call swaption owner to exercise in the event of the default of the

reference credit asset. Finally, note that the matter addressed here does not affect

cancelable CDSs. A default-triggering event terminates the CDS and therefore the

embedded option to cancel the CDS.

7 This is, of course, different than implying that the CDS call swaption is American-style. An American-style (or Bermudian-style) call swaption could be exercised prematurely for reasons other than thetermination of the underlying CDS occasioned by the default of the CDSs reference credit asset.

7


8/27

3. Product application

To illustrate the use of CDS swaptions, let us consider three product applications: to

reduce a banks regulatory capital; to create a synthetic credit-linked note; and to create a

synthetic collateralized debt obligation.

3.1. Reducing bank regulatory capital

Suppose that a bank is carrying a large number of commercial loans which in turn is

stretching the banks regulatory capital. The bank cannot sell all of the loans becausemost are not assignable. The bank needs to reduce its regulatory capital requirements. It

can sell one loan and use the proceeds to purchase a call-style credit default swaption

where the underlying reference credit asset is a portfolio of the remaining loans (or a

highly correlated basket thereof). By purchasing this basket CDS call swaption, the bank

should obtain regulatory capital relief in a manner analogous to being long a basket CDS.

Of course, the principle advantage of buying the CDS call swaption (versus entering a

long position in a basket CDS) is the returns earned on the loans should their credit

quality improve; the principle disadvantage of the swaption vis--vis the basket CDS is

the cost of the former.

3.2. Creating a synthetic credit-linked note

Suppose that a hedge fund buys a 4-year floating-rate note issued by a highly rated bank

sponsor. The note pays s.a. $LIBOR plus 5 bps. The fund manager can enhance the

8 Presumably, the bond insurance price of 10 bps would be provided by a traditional bond insurancecompany such as CapMac/MBIA.

8


9/27

coupon to s.a. $LIBOR plus 45 bps if she agrees to bear the default risk associated with

an altogether different bond (in addition to the credit risk of the note she is buying). This

is a common credit-linked note (and a common way that dealers lay off their credit risk

from engaging in short CDS positions). Instead, the manager can effectively enhance the

coupon on the note by writing a put-style CDS swaption on the same/second bond. By

purchasing the floating-rate note and writing the put CDS swaption, the hedge fund

manager is long a synthetic credit-linked note.

3.3. Creating a synthetic collateralized debt obligation Suppose that an asset manager wants to create a synthetic collateralized debt obligation

(CDO), so he issues/sponsors a CDO (through a special purpose vehicle) with four debt

tranches and one equity tranche. The total issuance is for $200 million. Of this, $175

million represents the debt tranches and $25 million represents the equity tranche, which

the sponsor keeps. The $200 million is then invested in high quality agency securities.

The manager then shorts a CDS (as a credit protection seller) on 20 different high yield

bonds with an average notional principal of $10 million each. For writing these credit

default swaps the CDO will receive an average of 520 bps per year. The average yield on

the agency bonds held is 4.41%. Thus, with the pick up of 5.20%, the synthetic high

yield assets are yielding 9.61%.

Suppose further that the funding costs (the debt tranches of the CDO) have an average

yield of 5.63%. The manager wins if the losses from default are less than 398 bps per

9


10/27

annum. The losses will be determined by the number and sizes of the high yield bonds

that default and the recovery rates on those defaulted bonds.

The former represents the typical way to create a synthetic CDO. The CDO is said to be

synthetic because the yield enhancement (on the agency bonds) is occasioned by

shorting CDSs, rather than holding junk bonds. But instead of shorting CDSs, the

sponsor could write a call-style CDS swaption where the underlying CDS references the

same basket of high yield bonds. In other words, buying agency bonds and writing CDS

call swaptions presents an alternative means of creating a synthetic CDO.

4. Pricing CDS swaptions

In this section we first address the pricing of European-style CDS swaptions. A

discussion of how to obtain the two critical model inputs the forward CDS swap rate

and the forward volatility is presented. We then illustrate the valuation of Bermudian-

style CDS swaptions (with a parallel discussion of American-style CDS swaption

pricing). The reader should note that extant valuation models for CDS swaptions appear

in three working papers: Sch nbucher (2000), Jamshidian (2002) and Schmidt (2004).

However, we humbly opine that the modeling presented in these papers is unnecessarily,

overly complex, and beyond the comprehension of all but the most mathematically

inclined. We hope that the following presentation is more accessible to the reader.

10


11/27

4.1. European-style CDS swaption pricing

If it is assumed that the forward credit default swap mid-rate (the one that prevails for the

reference credit asset at the time of exercise of the European-style CDS swaption) is

lognormal, then European-style CDS swaptions can be priced using a straightforward

modification of Blacks (1976) model. 9 In short, the CDS swaption can be priced using a

model that is analogous to the pricing of interest rate swaptions.

Begin by defining the following variables:

R 0 = the relevant forward CDS swap rate, expressed with compounding of m periods per annum, at time 0. In our illustration (see Section 2), this would be the

CDS swap rate, expected to prevail in 6 months and for 3 years, for the reference

credit asset (here a BB-rated, 8%-coupon, 10-year bond);

R K = the strike rate on the CDS swaption, also expressed with compounding of m

periods per annum. In our illustration, this is 150 bps;

T = the maturity of the CDS swaption. In our illustration, this is 0.50 (6 months);

= the standard deviation of the change in the natural logarithm of R 0, i.e., the

forward vol;

n = the maturity of the underlying CDS. In our example, this is 3 years;

m = the periodicity (or tenor) of the underlying CDS. In our example, this is 2

(semi-annual payments);

9 Another possibility here is to assume that the credit spread follows a process that is analogous to theinterest rate process found in the LIBOR market model of Brace et al (1997), Jamshidian (1997), andMiltersen et al (1997). Here an analytic approximation for the pricing of European-style credit defaultswaptions may exist. This is a subject worthy of future research. Hull and White (2000) derived ananalytic approximation for the pricing of European-style interest rate swaptions where the swaps referenceinterest rate was described by the LIBOR market model.

11


12/27


13/27

e(-0.03 x 1.0) + e (-0.03 x 1.5) + e (-0.03 x 2.0) + e (-0.03 x 2.5) + e (-0.03 x 3.0) ] = 2.785295. d 1 =

[ln(0.015/0.015) + (0.12) 2(0.50)/2]/(0.12) 0.50] = 0.04246. d 2 = 0.04246 (0.12) 0.50 =

-0.04239. N(0.04246) = 0.51696, and N(-0.04239) = 0.48307. Finally, C E = $141,590.

The value of the corresponding put is the same, that is, P E = $141,590, since both are

struck at-the-money.

It is interesting to note that a long (short) position in a CDS call swaption combined with

a short (long) position in a corresponding CDS put swaption creates a synthetic long

(short) forward-starting CDS (starting at time T and with swap rate R K ). This in turnimplies that combinations of CDS swaptions can be used just like CDSs can be used

to infer default rates and recovery rates for their underlying reference credit assets. 11

4.2. On obtaining R 0 and

The critical inputs to CDS swaption valuation are of course the relevant forward CDS

swap rate R 0 and the forward vol . Regarding the former, one can readily compute the

forward CDS swap rate if there exists a CDS swap curve for reference credit (or pari

passu ) asset. The methodology is completely analogous to obtaining a forward interest

rate swap rate from an interest rate swap curve. In practice, a term structure of CDS

swap rates normally exists. For example, in January 2001, that is, prior to its problems,

Enrons rating was Baa1 (Moodys) and the bid-offer mid-rates on Enron 3-, 5-, 7-, and

10-year credit default swaps were 115 bps, 125 bps, 137 bps, 207 bps, respectively.

11 See Hull (2003) beginning at page 641.

13


14/27


15/27

Unfortunately, the need for a no-arbitrage term structure model of credit default swap

rates has the effect of making the pricing of American- and Bermudian-style swaptions

extremely complicated at best. To invoke an analogy, consider the popularly-traded

Bermudian-style interest rate swaption. It permits the swaption owner to exercise on the

net payment dates. Most professional traders of this product employ a one-factor no-

arbitrage interest rate term structure model for pricing. While some experts have argued

that such an approach is prudent [Andersen and Andreasan (2001)], others contend that it

leads to substantial pricing error [Longstaff, Santa-Clara and Schwartz (2001)]. The pricing of American- and European-style CDS swaptions is no less complicated, and

controversial.

Some solace can be taken in the fact that end-user demand for CDS swaptions is heavily

concentrated in the European-style. And for cancelable CDSs, the value of the embedded

American- or Bermudian-style option to terminate is largely minimized, or completely

eliminated, if the underlying reference credit asset, or a pari passu asset, is liquid

(thereby permitting a trader in a CDS to close the position by executing a

reversing/opposite trade in a new CDS written on the same or pari passu asset).

Still, one is left with the issue of how to go about pricing American- and Bermudian-style

CDS swaptions when the necessity arises. In subsection 4.4 below, we impart the pricing

of these products in a simplified setting, namely where there is a one-factor credit spread

term structure model. Then, in subsection 4.5, we suggest that the pricing of American-

15


16/27

and Bermudian-style CDS swaptions is probably best tackled using a richer credit spread

term structure model in conjunction with a Monte Carlo simulation valuation approach.

4.4. A one-factor model approach

The material presented in this subsection is mainly for pedagogical purposes. It is

designed to give the reader a sense of the complexity of pricing American- and

Bermudian-style CDS swaptions. And, in particular, how their prices are dependent on

the evolution of the credit spread term structure as well as the volatility of credit spreads.

The one-factor model of credit spreads invoked here is itself rather simple. In continuous

time, the one factor would be the instantaneous credit spread. The model does not permit

mean-reversion in the credit spread. 13 It assumes a flat credit spread volatility term

structure; that is, all credit spreads whether short-dated or long-dated have the same

volatility. 14 And as a one-factor model, it does not permit the possibility of short-term

and long-term credit spreads moving in opposite directions contemporaneously (a credit

spread twist). Still, the model does permit the credit spread term structure to shift in

non-parallel ways, and it accommodates a level effect whereby the volatility of credit

spreads increases (decreases) with a rise (fall) in credit spreads which tends to be true.

We examine a discrete-time version of the model where a time increment is equal to 0.5

years. Thus the one factor is the six-month credit spread. The reference credit assets are

13 If each yield comprising the credit spread is itself mean reverting, then, by definition, the credit spreaditself will be mean reverting, but probably at a much slower rate. So the degree of mean reversion may benominal.14 In reality, it is probable that shorter-term credit spreads are more volatile than longer-term credit spreads.

16


17/27

a series of risky, high-yield bonds, for instance, emerging market bonds. Given the credit

quality of the bonds issuer, suppose that the bonds current (time 0) credit spreads (out to

two years) are 554 basis points (bps) for a 0.5-year maturity, 545 bps for a 1.0-year

maturity, 547 bps for a 1.5-year maturity, and 550 bps for a 2.0-year maturity. These

four credit spreads represent our relevant credit spread term structure. These rates are

expressed with semi-annual compounding and already have been purged of any

contaminating factors; for example, the effects on yields and thereby credit spreads of

any embedded options in the reference credit assets have been controlled via an

application of option-adjusted spread analysis.

In our model the change in the short-term/six-month/one-factor credit spread is given by

the multiplicative term

emh h (3)

Where h represents a time increment, represents the volatility of the credit spread (the

standard deviation of the percentage change in the natural logarithm of the credit spread),

and m represents a drift term (or mean). Equation (3) obviously implies a recombining

binomial framework wherein the one factor the short-term credit spread and therefore

the entire credit spread term structure, can move up or down after a discrete increment of

time (h). Here h = 0.5. We will assume that = 0.17. That is, the volatility of the credit

spread (of any maturity) is 17% per annum a high volatility accompanying a substantial

average credit spread (due to the level effect). The drift terms m 1, m2, m3 and m 4 are

non-stochastic but can change each period. These terms are parameterized by forcing the

model to fit the current credit spread term structure (a no-arbitrage approach). When

17


18/27

doing so, we also force the risk-neutral probabilities of the up jumps (and down jumps) in

the single factor (and therefore entire credit spread term structure) to be 50%.

Given our initial credit spread term structure (0.5-year, 554 bps; 1.0-year, 545 bps; 1.5-

year, 547 bps; 2.0-year, 550 bps), we have the following values for m: m1 = -0.0797, m2

= 0.0422, m3 = 0.0169, and m 4 = 0.0015. These values are obtained via no arbitrage

arguments. For example, m 1 is calculated from

( )( )

.12/0554.01

2/0545.0120554.0][5.0

25.017.05.05.017.05.0 11

++=+ + mm ee

We also have the resulting tree of credit spread term structures for our high-yield,

emerging market bonds (Figure 1 below). In this tree, the top number in each node

represents the prevailing (at time 0) or subsequently prevailing (depending on the jump in

the term structure) 0.5-year credit spread; the second number (if it exists) represents the

prevailing or subsequently prevailing 1.0-year credit spread; the third number (if it exists)

represents the prevailing or subsequently prevailing 1.5-year credit spread; and the fourth

number at time 0 is the current 2.0-year credit spread. Again, the probabilities of the

upward and downward movements in the term structure illustrated above are 50% each.

Figure 1. Tree of credit spread term structures

Time 0 Time 0.5 Time 1.0 Time 1.5

7.864%6.915%

6.004% 6.968%6.089% 6.184%6.147%

5.54% 5.437%5.45% 5.479%5.47%

18


19/27

5.50% 4.721% 4.862%4.788%4.834% 4.275%

4.308%3.823%

Armed with the above tree of credit spread term structures, we are prepared to value a

Bermudian-style CDS swaption where exercise is permitted every six months. 15

Illustration. Consider a Bermudian-style CDS put swaption with two-year maturity and

strike rate 550 bps (with semi-annual compounding), where the underlying is anoriginally two-year CDS entailing $100 million face value of our reference credit asset

having 2-year maturity. In other words, the CDS swaption permits its owner every six

months for two years to opt to be a seller of credit protection and to receive 5.50% s.a.,

until year 2. As such, the swaption grants the right to short the CDS for zero, or, in other

words, enter a short position in the CDS (receiving 5.5% s.a. and paying an amount

contingent upon default of the 2-year emerging market bond) at no cost.

For simplicity, assume that the relevant interest rate curve (used for discounting all cash

flows) is flat at 3% s.a., and so every forward rate is also 3% s.a. and each expected six-

month discounting factor (d 0.5) is 1/[1 + (0.03/2)] = 0.9852. In addition, assume that the

volatility of each forward rate is zero. Obviously, this latter assumption is unrealistic. It

is not invoked to suggest, in any way, that credit spreads and credit risk-free (or nearly

15 To value a Bermudian-style CDS swaption with a different exercise periodicity (e.g., quarterly), then onewould need to change the value of h (e.g., to 0.25) and to address all that said change occasions. In order tovalue an American-style CDS swaption one of course would need to permit h to be much smaller, e.g., asingle trading day, in order to frequently test for early exercise and thus obtain an accurate price.

19


20/27

credit risk-free) interest rates are zero correlated, or that the volatility of credit spreads is

not possibly correlated with the level of credit risk-free (or nearly credit risk-free) interest

rates. Rather, this assumption is invoked purely for pedagogical reasons. As we will see

shortly, pricing CDS swaptions is tedious enough under this simplifying assumption.

Relaxing the assumption is discussed more in the following subsection (4.5). This

assumption does of course imply that the actual yields (as opposed to credit spreads per

se) on the issuers bonds again where said yields already have been purged of any

contaminating factors such as embedded options are 8.54% for 0.5-year, 8.45% for 1.0-

year, and so on.16

Given the binomial term structure environment depicted above, we have the following

values (in $millions) for a short position in the underlying CDS:

Figure 2. Tree of values for a short CDS ($Millions)

Time 0 Time 0.5 Time 1.0 Time 1.5

uuu-1.1645

uu-1.3867

u uud-0.9377 -0.3369

ud+0.0019 +0.0210

d udd

16 An important reminder is now imparted. Namely, when a typical counter party on a CDS sells credit protection, the reference credit asset is not credit enhanced to Treasury. Rather, it is only enhanced to thecredit quality of the seller of credit protection. Due to netting agreements, collateral agreements, mid-market agreements, and other standard credit protection provisions found in the over-the-counterderivatives industry, the credit spread on the reference credit asset should therefore reflect the spread

between said assets yield (again, adjusted for the influences of factors such as embedded options) and thecomparable-maturity interest rate swap rate not the Treasury rate. In our example, 3% s.a. flat thereforereflects the assumed shape of the $LIBOR swap curve.

20


21/27

+0.9727 +0.1595dd+1.4552

ddd+0.4193

How were the values of the short CDS computed in Figure 2? The intuition is

straightforward enough: Assume that at any node, the short CDS party can reverse trade

by entering a long CDS position, and therefore lock-in an annuity of future inflows (or

outflows) to be discounted at (here) 3% s.a. (the zero-volatility forward swap rate). The

annuity itself is given by the difference between the original CDS rate (here 5.50% s.a.)and the new rate, times one-half of $100 million. (Here we omit consideration of bid-

offer spreads.) The new rate is the appropriate swap rate on the new/long CDS. The

length of the annuity is obvious the remaining maturity of the original CDS or,

equivalently, the maturity of the new/long CDS.

So, how then is the new swap rate determined? This is rather straightforward too (in our

environment). For example, at time 1.5 in the up- up- up-state (uuu), the new credit

spread is 7.864%. Since at this time mark there is just one more period remaining to the

original swap, the reversing trade would entail assuming a long position in a 0.5-year

CDS whose correct swap mid-rate clearly must be 7.864%. (It is correct in the sense of

ensuring that the new CDS has zero initial value, that is, is at-market, at this node. The

present value of the netted fixed and expected contingent (upon default of the reference

credit asset) payments is zero.) And so the short CDS is valued at (5.50% -

21


22/27

7.864%)($50MM)(0.9852) = -$1.1645MM. The same procedure gives us the values

presented in nodes uud, udd, and ddd of Figure 2.

Now consider an interior node like ud in Figure 2: Here we have a credit spread term

structure (from Figure 1) of 5.437% (0.5-year) and 5.479% (1.0-year). These rates imply

discount factors of d 0.5 = 0.9735 and d 1.0 = 0.9474. The par rate occasioned by these rates

- which is the same as the yield-to-maturity implied by these rates, which is also the

correct CDS swap rate - is: Par rate = 2(1 0.9474)/(0.9735 + 0.9474) = 0.05478. Thus

the value of the short CDS in node ud of Figure 2 is given by two payments of (5.50% -5.478%)($50MM), each discounted at 3% s.a. for a total of $0.021MM. The other values

(at nodes uu, dd, u, d, and at time-0) found in Figure 2 are calculated in an analogous

fashion. 17

Given the values of the short CDS position depicted in Figure 2, we can now compute the

four possible put swaption values at time 1.5. This is first time point necessary to value

the put swaption in our illustration (because the underlying CDS expires at the same time

of the swaption, and so the last time that any exercise would occur is at the 1.5-year time

mark):

Figure 3. Put swaption at time 1.5 ($Millions)

Short CDS Put SwaptionTime 1.5 Time 1.5

-1.1645 Max[ 1.1645,0] = 0

17 Note that the time-0 value of the short CDS in our illustration is not quite zero. It is $1,900. The at-market swap rate is slightly lower than 5.50% s.a. It is 5.499% s.a.

22


23/27

-0.3369 Max[ 0.3369,0] = 0

+0.1595 Max[+0.1595,0] = 0.1595

+0.4193 Max[+0.4193,0] = 0.4193

We next need to compute the three possible put swaption values at the 1.0-year time

mark, while checking for the prospect of early exercise (Figure 4). This entails

comparing the wait value (if any) to the early exercise value and entering in each node

the greater of the two values.

Figure 4. Put swaption at time 1.0 ($Millions)Short CDS Put SwaptionTime 1.0 Time 1.0 Time 1.5

0Exercise value < 0

-1.3867 Wait value = 0Swaption value = 0

0Exercise value = 0.0210

+0.0210 Wait value = [0.5(0 + 0.1595)](0.9852)Swaption value = 0.0786

0.1595Exercise value = 1.4552

+1.4552 Wait value = [0.5(0.1595 + 0.4193)](0.9852)Swaption value = 1.4552

0.4193

This process must be repeated at time 0.5 and at time 0 in order to obtain the value of the

Bermudian-style put swaption. As reflected in Figures 5 (time 0.5) and then 6 (time 0)

below, the value of the swaption is $498,200 in our illustration.

Figures 4 through 6 present intuitive results. For example, the put swaption value

increases, as does the incidence of early exercise, as the credit spread declines (the south-

23


24/27

eastern part of the trees). It is then that the swaption owner wants to become the seller of

credit protection at an attractive premium of 550 bps. Of course, the value of the

swaption would change (directly) with . And we could readily compute the swaptions

relevant risk metrics (Delta or DV01, Gamma, and Vega).

Figure 5. Put swaption at time 0.5 ($Millions)

Short CDS Put SwaptionTime 0.5 Time 0.5 Time 1.0

0

Exercise value < 0-0.9377 Wait value = [0.5(0 + 0.0786)](0.9852)Swaption value = 0.0387



1.4552

Figure 6. Put swaption at time 0 ($Millions)

Short CDS Put SwaptionTime 0.5 Time 0 Time 0.5



0.9727

4.5. Applying the Jarrow-Lando-Turnbull Model

A concern about the approach illustrated in subsection 4.4 is that the one-factor model

utilized is not sufficiently rich; for example, it assumes a flat credit spread volatility term

structure. To overcome this concern, we suggest the use of a multi-factor version of the

24


25/27

Markov model for the term structure of credit spreads as developed by Jarrow, Lando and

Turnbull (1997). 18 The Jarrow-Lando-Turnbull (JLT) model should be used to capture

the potential evolution of the credit spread as said model permits for a richer credit spread

environment.

To make the JLT model operational in the sense of permitting the possibility of early

exercise, we suggest using the Monte Carlo method of either Longstaff and Schwartz

(2001) or Andersen (2000). These methods are ideally suited because they efficiently

accommodate American- and Bermudian-style options that depend on two or morestochastic variables. (In our context, the LJT model would include a second factor in

order to accommodate non-constant credit spread volatility.) The Monte Carlo method of

Longstaff and Schwartz or Andersen should be used in conjunction with a corrective

procedure (to correct for a sub-optimal exercise boundary) found in Andersen and

Broadie (2001).

5. Conclusion

The explosion in the trading of credit default swaps has recently occasioned a growing

interest in credit default swaptions. This paper described CDS swaptions, offered some

illustrations of their application to achieve a variety of financial goals, and presented and

discussed valuation models. For those deeply interested in CDS swaptions, we humbly

suggest the following two avenues for future research: implying default probabilities and

recovery rates for the underlying reference credit assets from the market prices of CDS

18 A similar credit spread term structure model is presented in Kijima (1998).

25


26/27

swaptions; and valuing more complex CDS swaptions such as swaptions written on or

embedded in binary and basket credit default swaps.

References

Andersen, L., 2000, A Simple Approach to the Pricing of Bermudian Swaptions in theMultifactor LIBOR Market Model, Journal of Computational Finance 3, 1-32.

Andersen, L., and J. Andreasen, 2001, Factor Dependence of Bermudan Swaptions: Factof Fiction, Journal of Financial Economics 62, 3-37.

Andersen, L., and M. Broadie, 2004, A Primal-Dual Simulation Algorithm for Pricing

Multi-Dimensional American Options, Management Science , 50(9), 1222-1234.Black, F., 1976, The Pricing of Commodity Contracts, Journal of Financial Economics 3, 167-179.

Brace, A., D. Gatarek, and M. Musiela, 1997, The Market Model of Interest RateDynamics, Mathematical Finance 7, 127-155.

Bollerslev, T., 1986, Generalized Autoregressive Conditional Heteroscedasticity, Journal of Econometrics 31, 307-327.

Cumby, R., S. Figlewski, and J. Hasbrook, 1993, Forecasting Volatilities andCorrelations with EGARCH Models, Journal of Derivatives 1, 51-63.

Engle, R., 1982, Autoregressive Conditional Heteroscedasticity with Estimates of theVariance of U.K. Inflation, Econometrica 50, 987-1008.

Hull, J., 2003, Options, Futures, and Other Derivatives, 5 th edition, Prentice Hall,Upper Saddle River, New Jersey.

Hull, J., and A. White, 2000, Forward Rate Volatilities, Swap Rate Volatilities, and theImplementation of the LIBOR Market Model, Journal of Fixed Income 10, 46-62.

Jamshidian, F., 1997, LIBOR and Swap Market Models and Measures, Finance andStochastics 1, 293-330.

Jamshidian, F., 2002, Valuation of Credit Default Swap and Swaptions, WorkingPaper, NIB Capital Bank.

26


27/27

Jarrow, R., D. Lando, and S. Turnbull, 1997, A Markov Model for the Term Structure ofCredit Spreads, Review of Financial Studies 10, 481-523.

Kijima, M., 1998, A Markov Chain Model for Valuing Credit Derivatives, Journal of Derivatives 6, 97-108.

Longstaff, F., P. Santa-Clara, and E. Schwartz, 2001, Throwing Away a Billion Dollars:The Cost of Suboptimal Exercise Strategies in the Swaption Market, Journal of

Financial Economics 62, 39-66.

Longstaff, F., and E. Schwartz, 2001, Valuing American Options by Simulation: ASimple Least Squares Approach, Review of Financial Studies 14, 113-147.

Miltersen, K., K. Sandmann, and D. Sondermann, 1997, Closed Form Solutions forTerm Structure Derivatives with LogNormal Interest Rate, Journal of Finance 52, 409-430.

Nelson, D., 1990, Conditional Heteroscadasticity and Asset Returns: A New Approach, Econometrica 59, 347-370.

Schmidt, T., 2004, Credit Risk Modeling with Gaussian Random Fields, WorkingPaper, Department of Mathematics, University of Leipzig.

Sch nbucher, P.J., 2000, A LIBOR Market Model with Default Risk, Working Paper,

Bonn University.