Pricing CMS Spread Options

and Digital CMS Spread

Options with Smile
Mourad Berrahoui
Commerzbank ZRC

I Introduction We start by presenting the current approach used in different banks,

then we propose two different methods to take into account the smile.
This document deals with the smile of spread options in the Black The first method consists in changing the strike where the volatility of
framework. The price of spread options is sensitive to the entire smile of each underlying is taken and represents only a partial modeling of the
both underlyings. The classical approach uses the Black model without smile. The second method takes into account the full smile of each
smile. For each underlying, the corresponding at-the-money volatility is underlying and involves some numerical integration. These two meth-
taken. This approach ignores the effect of the smile and this is even more ods are used to show the errors generated by the old approach. In the
of a problem when we deal with digital options, as in this case there is a last section, we extend the two methods to CMS underlyings, we give
smile effect caused directly by the slope of the smile. some ideas how to generate an artificial smile and used the same
In general no closed formula exists for pricing a spread option when approach above.
the strike is different to zero. We don’t focus in this paper on the numer-
al method. A very detailed survey on the valuation of spread option is
given in Carmona and Durrleman [2003]. II Notations
Dempster and Hong [2001] propose to use the Fast Fourier Transform
(FFT) with stochastic volatility and interest rate environments The following notations are used throughout this document.
Alexander and Scourse [2003] propose to value spread options with a Let’s consider two assets F1 and F2 and an option of maturity T
bivariate normal mixture distribution. depending on those two assets. We assume that under the T forward
An interesting study has been done see Cherubini and Luciano [2002] probability, each Fi (i = 1, 2) follows a lognormal process according to
where a non Gaussian copula has been proposed to associate the margin- the stochastic differential equation:
al distribution. This copula is calibrated using historical data.
dFi (t)
The aim of this paper is to develop a simple approach, easy to imple- = µ(Fi (t), t)dt + σi (t)dW i (t) (1)
ment with exogenous input smile with some application on CMS product. Fi (t)

2 Wilmott magazine
 TECHNICAL ARTICLE 1

Correlation between the two assets is represented by the fact that the two price:
standard Brownian processes in equation (1) satisfy:
P = B(0, T) · EQT [Max(Q1 F1 (T) − Q2 F2 (T) − K; 0)] (7)
E[dW 1 · dW 2 ] = ρdt (2)
There is no closed-form formula but two different numerical methods
A spread option also called crack spreads, due to their use in the oil are available to calculate P : Monte Carlo and semi-analytical.
industry gives the holder the right to exchange F2 for F1 at expiry. The
III.2.2 Montecarlo approach
payoff is:
We simulate the two processes F1 and F2 . The price P will corresponds to
payoff = Max(Q1 F1 − Q2 F2 − K; 0) (3) the mean of (7) over the set of Monte Carlo paths.

where Q1 is the quantity of asset F1 , Q2 the quantity of asset F2 , and K the III.2.3 Semi-analytical approach
strike. Different approaches exist:
• Apply a conditioning technique to turn the two-dimensional inte-
III Current Approach without Smile gral into a single one ( Ravindran [1993] , Shimko [1994] )
P = B(0, T) · EQT {EQT [Max(Q1 F1 (T) − Q2 F2 (T) − K; 0)|F2 (T)]}
III.1 Spread option with zero strike
When K = 0, a closed-form formula exists (Margrabe, 1978). We assume • Fast Fourier transform (Carr and Madan [1999] , Dempster and Hong
that the drift in equation (1) is deterministic. The price P of this option is: [2001])

P = Q1 F1 B(0, T)eµ 1 N(d1 ) − Q2 F2 B(0, T)eµ 2 N(d2 ) (4) IV New Approach with Smile
where IV.1 A simple way to take into account a partial smile
ln(Q1 F1 /Q2 F2 ) + (µ1 − µ2 + σ 2 /2)T The problem with the formulas presented in the last section in the pres-
d1 = √ ence of smile is what volatility to use for each index. In general, we use
σ T
√ the volatility at the money for each underlying.
d2 = d1 − σ T In some special cases it is possible to determine a strike at which to
 T
take the volatility of each underlying rather than the money. Let’s
µi = µi (t)dt; i = 1, 2
0 assume that the asset F2 is less volatile. So the spread option become sim-

ply a mono-underlying option and the volatility to use for F1 correspond
σ = σ12 + σ22 − 2ρσ1 σ2
 to the strike F2 (0) + K . On the other hand, if we suppose that the asset F1
 is less volatile, then the volatility to use for F2 corresponds to the strike
1 T
σi = σi (t)dt; i = 1, 2 F1 (0) − K .
T 0
On the basis of this reasoning, we propose to use in general:
and B(0, T) is the price of a zero coupon of maturity T . Vol(F1 ) = Vol(Strike = ATM(F2 ) + K)

III.2 Spread option with non-zero strike Vol(F2 ) = Vol(Strike = ATM(F1 ) − K)

We will show later how accurate this approximation is in comparison to
III.2.1 Theoretical price of a spread option
the habit to use at the money volatility in case of a deeply in/out money
To calculate the spread option price in the case where K = 0, it is neces- option.
sary to write equations (1) and (2) differently to use only independent Just to give an example, imagine that we try to price a spread USD CMS
Brownian motions W̃1 and W̃2 , as follows: 20Y and USD Libor 3M at 06/17/2003 (Libor 3M = 1.02%, Swap20Y = 4.299%)
dF1    with strike equal to 3.279% (4.299%–1.02%). When the option is at money
= µ1 dt + σ1 ρdW̃t1 + 1 − ρ 2 dW̃t2 (5) (as it is the case at the beginning of the trade), there is no difference
F1
dF2 between the two methods. However, when the spread moves, the option
= µ2 dt + σ2 dW t1 (6) becomes deeply out or in the money and the more convex the smile, the
F2
greater the difference between the two methods. Even if the option was
The price P is the discounted expectation of payoff (3) under the T for- dealt at zero strike, because the smile for the indexes LIBOR3M and
^
ward probability QT where T is the maturity of the option we want to CMS20Y is quite different, the two methods still lead to different prices.

Wilmott magazine 3
 IV.2 How to take into account the entire smile Gaussian copula assumption
The formula given for the price of a spread option in the previous sec- Prob(F1 (T) > x1 , F2 (T) ≤ x2 |Full smile)
tions cannot be extended to calculate a price with smile. For this, we (12)
= Prob(F1 (T) > x̃1 , F2 (T) ≤ x̃2 |σ1 = 1 (T, x1 )); σ2 = 2 (T, x2 )))
need a more general expression for the price which does not assume that
F1 and F2 follow lognormal distributions. The following formula is true with x̃1 and x̃2 such that
independently of the distribution of the underlying:
 +∞ Prob(F1 (T) > x̃1 |σ1 = 1 (T, x1 )) = Prob(F1 (T) > x1 |Full smile) (13)
C = B(0, T) Prob(Q1 F1 (T) > x + K, Q2 F2 (T) ≤ x) dx (9) Prob(F2 (T) > x̃2 |σ2 = 2 (T, x2 )) = Prob(F2 (T) > x2 |Full smile) (14)
0
1 (T, x1 ) denotes the implied volatility of F1 (T) at strike x1 and 2 (T, x2 ) the
where Prob(. . .) is the bivariate cumulative distribution with correlation
equal to ρ . implied volatility of F2 (T) at strike x2 .
In order to prove (9), we need the following proposition. This assumption means that we are using a Gaussian Copula to repre-
sent the joint distribution of the random variables F1 (T) and F2 (T).
Proposition The following algorithm, which relies on the Gaussian copula
The spread option payoff is a sum of product of digital options: assumption, can then be used to calculate the price of a spread option
 +∞ with smile as in (11).
Max(Q1 F1 (T) − Q2 F2 (T) − K; 0) = 1{Q1 F1 (T)>x+K } ·1{Q2 F2 (T)≤x} dx (10)
0 Algorithm
Proof — Calculate Prob(F1 (T) > x1 |Full smile), i = 1, 2 , from the price of a call
We have just to change the boundary of the integral in (10) by spread.
x < Q1 F1 (T) − K — Solve equations (13) and (14) for x̃1 and x̃2 .
. — Estimate ρ from historical data for F1 (t) and F2 (t).
x ≥ Q2 F2 (T)
— Calculate the joint distribution (normal bivariate) of F1 (T) and F2 (T)
(9) is then obtained by taking the expectation of (10). using (12).
The integral in (9) can be calculated numerically using simple meth-
ods: Trapezoidal rule, Simpson’s rule. . ., or high-order methods: Gauss, IV.3 Extension to CMS spread options
Gauss-Kronrod. IV.3.1 Introduction
All those methods involve approximating (9) in the discrete form:
If we want to use the model we have proposed above, we need the smile
 surface for each underlying. This smile is more or less known in the mar-
P = B(0, T) · wi Prob(Q1 F1 (T) > xi + K, Q2 F2 (T) ≤ xi ) (11)
i
ket when the underlying is the short rate (Libor 1M,. . .,12M). But, when
the underlying is CMS, the smile is unknown. One idea is to use the swap-
where wi is a series of quadrature weights.
tion smile with the swap maturity equal to the tenor of this CMS.
We are now faced with the problem of calculating the probability in
Unfortunately this strategy is not arbitrage free -in theory- particularly
(11) in presence of smile. The probability that one asset is above a fixed
when the CMS cap/floor and swap are liquid. In the last part of this sec-
strike can be retrieved easily from prices of call options. Here we need to
tion, we propose an idea to build this smile using the prices of CMS
calculate a bivariate probability. By no arbitrage, we can found see
caps/floors and swaps. To introduce this idea, first we present the issues
Cherubini and Luciano [2002] a lower and upper limit
involved in pricing CMS products, with a specific section about the tim-
P1 − MIN(P1, P2) ≤ Prob(F1 (T) > x + K, F2 (T) ≤ x) ing adjustment necessary for CMS products with fixings in advance.
≤ P1 − MAX(P1 + P2 − 1, 0) Then we expose a simple approach, widely used in banks, to price CMS
swaps and caps/floors using the whole smile of swaptions. This approach
with is based on a simple idea of replication, which can be used for any com-
plex European payoff.
P1 = Prob(F1 (T) > x + K)
P2 = Prob(F1 (T) > x) IV.3.2 Issues in pricing CMS products
Theses limits represent the financial application of the minimal and Let us denote SRt the swap rate at time t. Its value at time t is:
maximal copulas of the Frechet-Hoeffding inequality. B(t, T0 ) − B(t, TN )
Copulas helps us to calculate the bivariate probability knowing the SRt =

N
marginal distribution for each underlying (call spread price), and for that B(t, Ti )τi
the following assumption is needed: i=1

4 Wilmott magazine
 TECHNICAL ARTICLE 1

The swap is starts a time T0 and its payments occur at times Ti (i = 1, We choose the strikes Kj to be equally spaced, using a discretization step
. . . , N) with T = T0 < T1 < . . . < TN . . So we have:
B(t, Ti ) is the price at time t of the bond which pay 1 unit at time Ti .
Kj = K + j; j = 1, . . . , M
T i −T i− 1
τi = if SRt is expressed in basis Act/365.
365 In our experience,  = 5 to 10 basis points is a good choice and M can be
SRt is then a martingale (i.e. a driftless process) under the numeraire chosen so that K is about 15%, but it really depends to what limit of strike
SMT defined as: the trader wants to hedge its CMS products.
N The calculus of the weight wi is straightforward.
SMT = B(T, Ti )τi
i=1 IV.3.4 Timing adjustment for CMS products with fixings in advance
Prices of FRAs and caplets are given by: We have seen that the replication technique is based on swaptions with
cash settlement, so it can only be used to price CMS products in which
FRA(t) = B(t, T) · EQT [SRT |Ft ] the swap rate is observed and paid at the same time. When we deal with
Caplet(t) = B(t, T) · EQT [Max(SRT − K; 0)|Ft ] CMS product with fixings in advance, e.g. CMS vanilla caps/floors/swaps,
the price has to be adjusted.
QT denotes the T forward measure. Under this measure, SRt is not driftless
If the swap rate is observed at time T and paid at T + δ , the forward
and it is difficult to calculate its drift.
swap rate SR0 has to be corrected by a timing adjustment (see Hull):
The price of a physical swaption is given by:

N SR0 δR0 ρσ σR T
−
Swaption(t) = ESM T [Max(SRT − K, 0)|Ft ] · B(t, Ti )τi 1 + R0 δ
i=1
with R0 is the value at time zero of the forward rate between T and
where ESM T denotes the expectation with respect to numeraire SMT . T + δ, σR is the volatility of this forward rate, σ is the at-the-money volatil-
We can apply the Black formula in this case, because SRt is driftless. ity of the forward swap rate and ρ is the correlation between the forward
From this short analysis, we can see that if we can express the payoff swap rate and the forward rate.
of FRAs/caps in terms of the payoff of the swaption, then pricing becomes
simple. It is the idea of the replication, which we develop now. Example
Note that in order to price a cash swaption, which is a more common Let’s take the example given by Hull (see ref.).
product than physical swaptions, one has to use instead of the physical SR0 = 5%
swap measure, the cash swap measure where the numeraire is:
R0 = 5%

N
1
SCashM t = σ = 15%
(1 + SRT )i
i=0 σR = 20%
IV.3.3 Replication of simple products on CMS δ = 0.5
In this section we develop the idea of replicating the payoff of a CMS ρ = 0.7
swap or cap as a linear combination of swaptions with different strikes. The forward rate has to be adjusted by —0.0000256 T.
In addition to the mathematical argument of easy derivation given in the
last section, another motivation for doing this is that the only simple and IV.3.5 Building CMS smile by arbitrage
liquid way to hedge a product on CMS is using swaptions. The process SRT can be written under the T-forward measure as follows:
We want to write a linear payoff (swap/cap/floor) of the form:

1
Max(SRT − K; 0) SRT = EQT [SRT ] exp − σ 2 T + σ WT (16)
2
in terms of a non-linear payoff (swaption with cash settlement) of the form: The application of the replication technique for FRAs gives the expecta-

N tion value EQT [SRT ] of SRT under the T -forward measure as:
1
Max(SRT − K; 0) · FRA
(1 + SRT )i EQT [SRT ] =
i=1
B(0, T)
So the idea is the find a set of weights wj and strikes Kj such that:
The price of the caplet/floorlet with strike K using the expression (16) of
 
N
1 the process SRT is simply given by Black’s formula:
Max(SRT − K; 0) = wj Max(SRT − Kj ; 0) · (15)
^
(1 + SRT )i  
j i=1 Caplet = Black EQ T [SRT ], σ (K), T, K

Wilmott magazine 5
 The unknown variable in this formula is the volatility σ (K). At the same V.2 Short rate spread option
time this price can be obtained using the replication technique
described above. Hence we can imply the volatility σ (K) by: We consider a spread option Libor6M–Libor3M. First, we consider strike
zero and volatility flat. We compare the Margrabe closed-form formula
σ (K) = Black −1 (Caplet) (without smile), the Monte Carlo approach (partial smile), and the full
We can apply this technique for every strike K and thus we build the CMS smile method.
smile.
Libor 6M = 0.99 %
We admit that it can be time consuming. At the first approximation we can take
Libor 3M = 0.95 %
swaption smile.
From the table below, we check that our model give the same results
V Tests as Margrabe’s formula in the case where the volatility is flat, for different
maturities.
V.1 Introduction
We first show the difference in price for short rate spread options
(Libor6M–Libor3M), for given market data: yield curve and smile, with
the three methods: TABLE 1: STRIKE = 0 , VOLATILITY IS
FLAT AT 20%, CORRELATION = 0.7
— The approach with taking at the money volatility for each index.
Margrabe MC method Full smile
— The same approach but with taking as strike for one index, the
formula (10 000 path)
money for the second index plus/minus the strike of the spread
1Y 9 9 9
option.
2Y 32 32 32
— Pricing with full smile as described in this document.
5Y 153 153 153
Then we do similar tests on CMS products. 7Y 275 275 275
In all our tests, we use the following features: 10Y 484 483 483
15Y 854 852 853
— Payment frequency : 6M
20Y 1181 1177 1176
— Day count : ACT/360
— Yield curve:

ATM swap rate The small difference can be due to the numeric integration method
1Y 1.14% used in our implementation.
2Y 1.55% Now, we consider the same option but with smiled volatility. We
5Y 2.63% notice that the difference become significant when the maturity
7Y 3.12% increases.
10Y 3.61%
15Y 4.12%
20Y 4.36% TABLE 2: STRIKE = 0,
VOLATILITY WITH SMILE,
—Volatility surface:
CORRELATION = 0.7
3% 4% 5% 6% 7% 8% 9% Margrabe Full smile
1Y 26.50 21.90 23.30 25.50 26.70 27.80 29.10 1Y 12 12
2Y 25.80 20.30 19.10 21.30 22.70 23.80 25.20 2Y 42 41
5Y 23.90 19.50 15.50 15.00 15.60 16.40 17.60 5Y 176 170
7Y 22.70 18.70 14.80 13.30 13.40 14.00 14.70 7Y 278 275
10Y 21.50 17.90 14.10 12.10 12.10 12.60 13.10 10Y 436 436
15Y 20.20 17.00 13.50 11.20 10.90 11.30 11.60 15Y 665 690
20Y 19.30 16.40 13.10 10.60 10.30 10.70 11.00 20Y 860 901

6 Wilmott magazine
 TECHNICAL ARTICLE 1

TABLE 3: STRIKE = 0.20% TABLE 5: SPREAD OPTION ON CMS 20Y

VOLATILITY WITH SMILE, AND CMS 2Y WITH STRIKE = 2.50% AND
CORRELATION = 0.7 CORRELATION = 0.7
Margrabe Our approach Vol at the money Partial smile Full smile
(in basis point) 1Y 47 46 47
1Y 4 4 2Y 79 80 84
2Y 24 23 5Y 139 155 166
5Y 131 126 7Y 170 194 210
7Y 217 216 10Y 211 250 269
10Y 347 356 15Y 273 334 351
15Y 552 583 20Y 339 415 428
20Y 725 773
V.5 Impact of Smile in Digital CMS
V.3 Building CMS smile surface We approximate a Digital option as a call spread with a strike shift equal
We compare the swaption volatility smile (for 10Y fixed swap maturity) with to 10 basis points. We compare the same three models again.
the CMS 10Y, after building the CMS smile as described in this document.

TABLE 6: CALL DIGITAL OPTION ON CMS20Y

TABLE 4: SMILE CMS 10 Y (VOLCMS10Y– AND CMS2Y WITH STRIKE = 1.50% AND
VOLSWAPTION NX10Y) CORRELATION = 0.7
3% 4% 5% 6% 7% 8% 9%
Vol at the money Partial smile Full smile
1Y -0.2 0.1 0.3 0.3 0.4 0.5 0.6
1Y 98 98 98
2Y -0.4 -0.1 0.3 0.5 0.6 0.7 0.8
2Y 179 180 176
5Y -0.8 -0.5 -0.2 0.1 0.4 0.5 0.6
5Y 320 309 308
7Y -0.9 -0.7 -0.4 -0.1 0.1 0.2 0.3
7Y 378 361 363
10Y -0.9 -0.8 -0.5 -0.2 0.1 0.2 0.2
10Y 441 417 421
20Y -0.9 -0.9 -0.7 -0.4 -0.1 -0.1 0.1
15Y 508 478 484
20Y 559 528 527
In general the CMS smile is less than the swaption smile with the
same swap maturity.
The graphs below show the differences between the prices with the
V.4 Impact of smile in CMS spread option different models as presented in the tables.
We consider the spread option on CMS 20Y and CMS 2Y
with strike equal 2.5% (in but not far from the money). Digital spread option
The first column gives the price of the spread option Price (in b p) Difference partial/without smile and full smile methods
priced with the volatility at the money for each index and (strike = 1.5%)
33
the second column with partial smile. The third column
28
shows the price with full smile. In the first two columns, the
23
price is calculated with Monte Carlo.
18
It is clear from table 5 that the model with partial smile Without smile method
13
is closer to the full smile model than the classical approach Partil smile method
8
without smile.
3
Those differences depends on
−2 1Y 2Y 5Y 7Y 10Y 15Y 20Y
— the convexity of the smile −7
— how far is the strike of the spread option from the −12
money. Maturity
^

Wilmott magazine 7
 TECHNICAL ARTICLE 1

Digital spread option could be a worthwhile alternative. For a digital option, in this
Price (in b p) Difference partial/without smile and full smile methods case one needs to consider:
(strike = 2.9%)
10 dC ∂C ∂C dσ
= + ∗
0 dK K =K0 ∂K K =K0 ∂σ K =K0 dK K =K0
1Y 2Y 5Y 7Y 10Y 15Y 20Y
−10
with C(K, σ (K)) is the call option price and σ (K) is a paramet-
−20 ric volatility function
−30 (Example: SABR model).
−40 Without smile method Another method which we propose is to price spread
Partial smile method
−50 options taking the volatility at a different strike than the
−60 money of each underlying the same time, as follows:
−70 Vol(F1 ) = Vol(Strike = ATM(F2 ) + K)
−80 Vol(F2 ) = Vol(Strike = ATM(F1 ) − K)
−90
Maturity This method is only a partial smile model but we show that it
is close to the first, full smile, method.
A separate section of this document is dedicated to deal-
Digital spread option
ing with CMS underlyings and building the CMS smile.
Price (in b p) Difference partial/without smile and full smile methods
(strike = 3.5%)
10
VII REFERENCES
8
6 ■ C. Alexander and A. Scourse [2003] “Bivariate Normal Mixture
4 Spread Option Valuation” ISMA Centre Discussion Papers in Finance
2 2003–15.
Without smile method ■ R. Carmona and V Durrleman [2003] “Pricing And Hedging Spread
0 Partial smile method
1Y 2Y 5Y 7Y 10Y 15Y 20Y
−2 Options in A Log-Normal Model” Working paper.
−4 ■ R. Carmona and V. Durrleman [2003] “Pricing And Hedging Spread
−6 Options” SIAM Review, 45:4 627–687.
−8 ■ P. Carr and.D.B Madan [1999] “Option Valuation Using the Fast
−10 Fourier Transform”, Journal of Computational Finance, 2, 4, 61–73.
Maturity ■ U. Cherubini and E. Luciano [2002] “Multivariate Option Pricing With
Copulas” Working paper.
■ S. Coutant, V. Durrleman, G. Rapuch and T. Roncalli [2001]
These graphs show that taking the volatility at the right strike (par-
“Copulas, multivariate risk-neutral distributions and implied dependence
tial smile) gives closer prices to the full smile method especially when functions” Working paper.
the option is deeply in or at the money. ■ M. Dempster and G. Hong “Pricing Spread Options with the Fast
For digital option at the money, the differences between the two Fourier Transform” Risk 2001 Europe.
models however are significant. ■ John C. Hull “Options, Futures & Other Derivatives”.
■ G. Rapuch and T. Roncalli [2001] “Some remarks on two-asset
VI Conclusions options pricing and stochastic dependence of asset prices” Working
paper.
In this document we have exposed two new methods to take into account ■ K. Ravindran [1993] “Low Fat spread” Risk Magazine 6:10, 66–7.
the smile for spread options and in particular digital spread options. ■ J. Rosenberg [2003] “Non parametric Pricing of Multivariate
The most advanced of those two methods is a numerical integration Contingent Claims” Research and Market Analysis Group, Federal
method based on a copula assumption, which uses the entire smile of Reserve Bank of New York.
each underlying. ■ D. Shimko [1994] “Option on Futures Spreads: Hedging, Speculation
If the smile is not smooth enough, this method can lead to instabili- and Valuation” The journal of Futures Markets 14:2,182–213.
■ Paul Wilmott, “Derivatives The Theory and Practice of Financial
ties. This is why, when this situation occurs, a parameterization of the
Engineering”.
smile and then using a closed-form formula for Pr ob(Fi,t > xi ; smile(Fi,t ))
W

8 Wilmott magazine