SSRN Id2542898

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Is the Information Obtained from European Options on Equally

weighted Baskets enough to Determine the Prices of Exotic


Derivatives such as Worst of Options?
Jacinto Marabel Romo∗

First version: March 2014


This version: December 2014

Abstract

In recent years there has been a remarkable growth of multi-asset options. These options exhibit
sensitivity to the volatility of the underlying assets, as well as to their correlations. The call versus
call is a product commonly used to trade correlation within the inter-dealer broker markets. The
buyer of correlation buys a European call on the equally weighted basket option and sells a weighted
average of European calls on each asset. In this case, the following important question arises: Is the
information provided by equally weighted basket options enough to price other European multi-asset
exotic derivatives such as worst-of or outperformance options? This article investigates this issue
under a stochastic correlation framework. Importantly, this article shows that, when pricing multi-
asset exotic derivatives, matching the prices of European equally weighted basket options, quoted in
the market, does not guaranty the absence of model risk even in the case where the exotic payoff is
observed only at maturity.

Keywords: stochastic correlation, stochastic volatility, multifactor, worst of options, outperfor-


mance options, equally weighted basket options.

JEL: G1, G2, G12, G13.

∗ The content of this paper represents the author’s personal opinion and does not reflect the views of BBVA. I am grateful

to Francisco Gomez Casanova, Javier Rubén Madrid, Guillaume Guilmard, Miguel Canteli and Martino Grasselli for useful
discussion. I also thank seminar participants at the 21th Global Derivatives Trading and Risk Management Conference
2014 (Amsterdam, Holland). All errors are my own.
1 Introduction

In recent years there has been a remarkable growth of multi-asset options. These options exhibit sensitiv-

ity to the volatility of the underlying assets, as well as to their correlations. Therefore, banks and other

financial institutions have become increasingly exposed to correlation between the underlying assets. The

fact that volatilities are stochastic is widely recognized and the correlations between financial quantities

are notoriously unstable. In this sense, empirical evidence shows the existence of a positive link between

realized correlation and realized volatility (see, for instance, Solnik et al., 1996). In particular, in 2008

the crisis in the credit market led to an increase in correlations and volatilities that resulted in large losses

for several exotic trading desks. Moreover, empirical evidence, comparing the implied volatility skew of

an index against is components, tends to show the existence of a implied correlation skew. Nevertheless,

multi-asset options are usually priced using constant correlations.

The call versus call is a product commonly used to trade correlation within the inter-dealer broker

markets. The buyer of correlation buys a European call on the equally weighted basket option and sells

a weighted average of European calls on each asset. The weights are usually chosen so that the structure

exhibits zero vega. Within the class of multi-asset options, structured products with embedded worst-of

options are quite popular between final investors. Although these products are pretty popular, their

liquidity in the inter-bank and inter-dealer broker markets is quite low. Hence, usually the financial

institutions that issue this kind of structures have to rely on internal models to price these derivatives.

In the single-asset world, Carr and Madan (2001) show that it is possible to price any European

derivative with a given maturity, even those that are not quoted in the market, if we have available

prices of plain vanilla options on that underlying asset for that maturity. Hence, given the price of these

vanilla options, the valuation of single-asset European derivatives is model-free. In the multi-asset case,

the following important question arises: Is the information provided by equally weighted basket options

enough to price other European multi-asset exotic derivatives such as worst-of or outperformance options?

This article investigates this issue under a stochastic correlation framework. To this end I consider a

multi-asset model based on Wishart processes, which accounts for stochastic volatility and for stochastic

correlation between the underlying assets returns, between their instantaneous volatilities and between

the volatilities and the assets returns. The Wishart process was mathematically developed in Bru (1991)

and was introduced into finance by Gourieroux and Sufana (2004). In their model the Wishart process

describes the dynamics of the covariance matrix and it is assumed to be independent of the assets noises.

On the other hand, Da Fonseca et al. (2008) introduce a correlation structure between the single asset

1
noise and the volatility factors and da Fonseca et al. (2007), as well as Marabel Romo (2012) consider

the pricing of multi-asset structures under a Wishart specification.

2 The multi-asset Wishart volatility model


0
n o
Along the lines of Da Fonseca et al. (2007), let St := (S1t , S2t ) ∈ R2 t ≥ 0 be the price process of
0
the assets and let us denote by Yt = (ln S1t , ln S2t ) the log-return vector. For simplicity, let us assume

that the continuously compounded risk-free rate r and dividend yield qi , for i = 1, 2, are constant.

Let Θ denote the probability measure defined on a probability space (Λ, z, Θ) such that asset prices

expressed in terms of the money market account are martingales. We denote this probability measure

the risk-neutral measure. Let us assume the following dynamics for the return process Yt under Θ:

 
1 p
dYt = r1−q − diag (Xt ) dt + Xt dZt
2
0
where 1 is a 2 × 1 vector of ones, q = (q1 , q2 ) and the vector diag(Xt ) ∈ R2 has the i-th component

equal to Xiit
h 0 0
i p 0 0p
dXt = ΩΩ + M Xt + Xt M dt + Xt dBt Q + Q dBt Xt

where Ω, M, Q ∈ R2×2 and Bt is a matrix of standard Brownian motions in R2×2 . The previous equation

characterizes the evolution of the variance-covariance matrix Xt and represents the matrix analogue of

the square root mean-reverting process associated with the instantaneous variance process under the

Heston (1993) model. The matrix Q characterizes the randomness of the state variable X, whereas −M

represents the speed of mean reversion. In order to grant that Xt is positive semi-definite and the typical

mean-reverting feature of the volatility, the matrix M is negative semi-definite, while Ω satisfies

0 0
ΩΩ = βQ Q

with the real parameter β > 1. This condition ensures that X is positive semi-definite at every point in

time if X0 is1 . Z ∈ R2 is a standard Brownian motion of the form:

p
Z t = Bt ρ + 1 − ρ0 ρWt

0
where W ∈ R2 is another standard Brownian motion independent of B and ρ = (ρ1 , ρ2 ) is a fixed

correlation vector, with ρj ∈ [−1, 1], that determines the correlation between the returns and the state

variables. As said previously, the matrix Q characterizes the randomness of the state variable X. The
1 In the general case where we have n assets, this condition becomes β > n − 1.

2
mean reversion level X, associated with this variable, is given by the following expression:

0 0
ΩΩ + M X + XM = 0

2.1 Variance-covariance structure

The Wishart process allows us to capture stochastic volatility and stochastic correlation. In particular,

the conditional covariance matrix of the return process is:

1
Vt := V ar (dYt ) = Xt
dt

The correlation between the assets returns is given by:

X12t
ρ(t)S1 S2 := Corr (dY1t , dY2t ) = √
X11t X22t

On the other hand, correlation between each asset return and its own instantaneous variance is

constant and it can be expressed as:

ρ1 Q1i + ρ2 Q2i
ρSi Vii := Corr (dYit , dViit ) = p 2 i = 1, 2.
Q1i + Q22i

In this model the correlation between each variance process is stochastic and it is given by:

ρ(t)V11 V22 := Corr (dV11t , dV22t ) =


 
X11t Q212
+ Q222 + X22t Q211 + Q221 ρ(t)S S (Q11 Q12 + Q21 Q22 )
p + p 1 22
4 X11t X22t (Q211 + Q221 ) (Q212 + Q222 ) 2 (Q11 + Q221 ) (Q212 + Q222 )

Finally, the cross-asset-variance correlations are also stochastic and can be expressed as:

ρ(t)S1 V22 : = Corr (dY1t , dV22t ) = ρ(t)S1 S2 ρS2 V22

ρ (t)S2 V11 : = Corr (dY2t , dV11t ) = ρ(t)S1 S2 ρS1 V11

Note that this correlation increases when the correlation between assets returns increases.

3 The pricing problem

Under the specification corresponding to the multi-asset Wishart volatility model it is possible to obtain

semi-closed-form solutions for the price of standard-start and forward-start European options on each

underlying asset or multi-asset options, such as worst-of options and outperformance options using

generalized Fourier transforms along the lines of Lewis (2000), da Fonseca et al. (2007) and Marabel

Romo (2012). For instance, let us consider the payoff associated with a worst-of call on asset 1 and asset

3
2 with strike K and maturity t = T :

   +  +
S1T S2T
min , −K = emin(Y1T −Y10 ,Y2T −Y20 ) − eln K
S10 S20

where the strike is expressed in percentage terms. For simplicity, let us assume that Si0 = 1 (for i = 1, 2)

so that Yi0 = 0. Hence, the payoff is

 +
w (Y1T , Y2T ) = emin(Y1T ,Y2T ) − eln K

The Fourier transform corresponding to w (Y1T , Y2T ) is given by:

Z
w
b (z) := w
b (z1 , z2 ) = eihz,YT i w (Y1T , Y2T ) dYT
R2

0
where z = (z1 , z2 ) ∈ C2 and where h., .i represents the scalar product. The Laplace transform of the

assets returns is defined as:


h 0 i
Ψ (λ; Y0 , T ) := EΘ eλ YT λ ∈ R2

Along the lines of da Fonseca et al. (2007) it is possible to express the time t = 0 price corresponding to

the exotic European option on both underlying assets, EOP0 , as follows:

e−rT
Z
EOP0 = 2 Ψ (−iz; Y0 , T ) w
b (z) dz
(2π) χ

where χ ⊂ C2 is the admissible integration domain in the complex plane corresponding to the general-

ized Fourier transform associated with the payoff function w


b (z). Unfortunately, the Laplace transform

associated with the payoff of a European equally weighted basket option is not available in closed-form.

4 The impact of stochastic correlation

Let us assume that M is diagonal with M11 = M22 . In this case, we have the following evolution

corresponding to dX11t and dX22t respectively:

p
Θ
dX11t = κ1 (θ1 − X11t ) + ζ1 X11t dWX11 ,t

p
Θ
dX22t = κ2 (θ2 − X22t ) + ζ2 X22t dWX22 ,t

Θ Θ
where WX 11 ,t
and WX 22 ,t
are Wiener processes under Θ and where κ1 = −2M11 , κ2 = −2M11 , θ1 =

− 2Mβ11 Q211 + Q221 , θ2 = − 2Mβ11 Q222 + Q212 , ζ1 = 2 Q211 + Q221 and ζ2 = 2 Q222 + Q212
  p p

4
Hence, when M is diagonal, the evolution of each individual asset can be represented through a

Heston (1993) model and, therefore, we will have the same individual implied volatility surfaces under

the Wishart model and under the Heston (1993) model. Since both models account for the existence

of stochastic volatility, the difference in the prices of multi-asset options obtained under both models

will represent the stochastic correlation premium. Note that, in the presence of stochastic correlation

premium, both models can yield the same price for plain vanilla options but, at the same time, different

prices for multi-asset derivatives.

4.1 Case 1: High correlation and flat term structure of correlation


4.1.1 Equally weighted basket options calibration

Let us assume that r = 0.015 and qi = 0.03 for i = 1, 2 and let us consider the following risk-neutral

dynamics corresponding to the evolution of the assets prices and their instantaneous variances vit , for

i = 1, 2, under the probability measure Θ:


     √ 
dS1t
S1t (r − q1 ) v1t dWSΘ1 ,t
 dS2t     √ Θ


 S2t
 
= (r − q2 )   dt +  √v2t dWS2 ,t
 

 dv   κ1 (θ1 − v1t )   ζ v dW Θ 
1t  1 1t v1 ,t

    
dv2t κ2 (θ2 − v2t ) ζ2 v2t dWvΘ2, t

with v10 = X110 and v20 = X220 . Let us assume that market prices are generated by the Wishart

specification with stochastic covariances of table 1. Table 2 provides the instantaneous correlations

associated with this specification. In this case, all the initial correlation levels coincide with the long-

term values and all the correlations are relatively high in absolute terms. Hence, I denote this specification

high correlation and flat term structure (HCFTS) specification.

Table 1: Parameters specification in a bivariate Wishart volatility model for assets returns

Parameter Value Parameter Value


M11 -1.2500 X220 0.0609
M12 0.0000 Q11 0.3700
M21 0.0000 Q12 0.1900
M22 -1.2500 Q21 0.1900
X110 0.0727 Q22 0.3300
X120 0.0559 ρ1 -0.7100
X210 0.0559 ρ2 -0.6900
β 1.0500
Note. X0 represents the initial level
associated with the covariance matrix.

5
Panel A: Asset 1 Panel B: Asset 2

0.7
0.7
0.6
0.6

0.5 0.5
Implied Volatility

0.4

Implied Volatility
0.4
0.3

0.3
0.2

0.1 0.2

0
4 0.1
3.5
0
3
0
2.5 0.5 0
4
2
0.5
1.5 1 3
1 2 1
1.5
0.5 1 1.5
0 2 Strike
Maturity (years) 0
2 Strike
Maturity (years)

Figure 1: Implied volatility surfaces corresponding to the Heston model under the HCFTS specification. Strike
prices are expressed as a percentage of the asset price, whereas maturities are expressed in years. The surfaces are
calculated using the parameters of table 3 together with the following parameters values qi = 0.03, i = 1, 2, r =
0.015.

Table 2: Correlations associated with the specification of table 1 corresponding to the bivariate Wishart
volatility model.

Initial levels Long-term levels


ρ(0)V11 V22 0.8526 ρV11 V22 0.8526
ρ(0)S1 S2 0.8397 ρS1 S2 0.8397
ρ(0)S1 V22 -0.7996 ρS1 V22 -0.7996
ρ(0)S2 V11 -0.7951 ρS2 V11 -0.7951
Notes. The long-term levels are calculated using
X instead of X0 .

Table 3 provides the Heston (1993) parameters corresponding to each asset under the HCFTS speci-

fication, whereas figure 1 displays the implied volatility surfaces associated with each asset. The figure

reveals the existence a of negative volatility skew, which is most pronounced for near-term options. This

is a common pattern of behavior that has been widely observed in the equity options market.

Table 3: Heston parameters associated with the bivariate Wishart volatility model.

Asset 1 Asset 2
κ1 2.5000 κ2 2.5000
θ1 0.0727 θ2 0.0609
ζ1 0.8319 ζ2 0.7616
ρS1 V11 -0.9468 ρS2 V22 -0.9522
v10 0.0727 v20 0.0609

On the other hand, panel A of figure 2 displays the implied volatility surface associated with equally

6
Panel A: Implied volatility Surface Panel B: At−the−money implied volatility term structure
0.23

0.225

0.22

0.5

0.45
0.215
0.4

Implied volatility
0.35
Implied Volatility

0.21
0.3

0.25

0.2 0.205
0.15
0.4
0.1
0.2
0.05 0.6
4
0.8
3 0.195
1
2
1.2
1
Strike 0.19
0 1.4 0.5 1 1.5 2 2.5 3 3.5 4
Maturity (years) Time (years)

Figure 2: Implied volatility surface corresponding to equally weighted basket options under the HCFTS Wishart
specification (panel A) and at-the-money implied volatility term structure (panel B).

weighted basket options under the HCFTS Wishart specification, while panel B provides the at-the-

money implied volatility term structure. Note that, since we have considered a parametric specification

for the Wishart model where the term structure of correlation is flat, the at-the-money volatility term

structure is also quite flat.

Let us assume that market prices are generated by the Wishart specification with stochastic co-

variances and let us assume that we observe prices of equally weighted basket options. The reason for

this choice is that basket options are, by far, the most liquid product to trade correlation within the

inter-dealer broker markets. We want to price multi-asset European exotic options with maturity within

three years such as worst-of and outperformance options. I focus on this maturity because it is the most

common maturity for structured products, as well as within the inter-dealer broker markets. In this

case, the following important question arises: is it enough to match the price associated with equally

weighted basket options? To investigate this question I calibrate the free parameters associated with

the multi-asset version of the Heston (1993) model with constant correlations in order to match the

prices of the equally weighted basket options generated by the Wishart HCFTS specification. I call this

multi-asset version of the Heston (1993) model the multi-Heston model.

We have the following free parameters to calibrate under the multi-Heston model: the inter-asset

correlation ρS1 S2 , the cross-asset-variance correlations ρSi Vjj for i 6= j and, finally, the inter-variance

7
correlation ρV11 V22 . To price multi-asset derivatives under both models I use Monte Carlo simulations

with daily time steps and 80.000 trials. The calibrated parameters are given by: ρS1 S2 = 0.8757,

ρS1 V22 =ρS2 V11 =−0.9000 and ρV11 V22 = 0.9000.

Table 4 shows the difference of implied volatilities associated with equally weighted basket options,

expressed in percentage terms, generated under the HCFTS specification and under the multi-Heston

model. A negative value means that the implied volatility is lower under the HCFTS specification than

under multi-Heston specification. We can see from the table that, in this case where the Wishart speci-

fication exhibits a flat term structure of correlation, the multi-Heston model is able to provide a pretty

good fit of the implied volatility corresponding to equally weighted basket options for all the maturities

and, especially, for options with maturity within three years. Note that, although the instantaneous

correlations are constant under the multi-Heston model, this specification is able to provide a similar

skew to the one associated with the Wishart model for basket options. This question is related to the fact

that the multi-Heston specification accounts for stochastic volatility and the existence of negative cross-

asset-variance correlations introduces a covariance skew that allows the model to replicate the basket

skew generated by the stochastic correlation framework provided by the Wishart model.

Table 4: Implied volatility differences (expressed in percentage) corresponding to equally weighted


basket options between the HCFTS Wishart specification and the multi-factor Heston model for different
maturities (expressed in years).

Strike/Maturity 0.75 1 1.50 2 2.50 3 3.50 4


65 0.11 0.08 0.05 0.00 0.03 0.04 0.02 0.04
70 0.08 0.05 0.02 -0.01 0.01 0.02 0.01 0.03
75 0.05 0.02 0.01 -0.01 0.00 0.01 0.00 0.03
80 0.00 -0.02 -0.01 -0.02 -0.01 0.01 0.00 0.02
85 -0.04 -0.04 -0.03 -0.03 -0.01 0.01 0.00 0.02
90 -0.08 -0.07 -0.04 -0.04 -0.02 0.00 -0.01 0.01
95 -0.12 -0.09 -0.05 -0.05 -0.03 0.00 -0.01 0.00
100 -0.14 -0.11 -0.06 -0.06 -0.04 -0.01 -0.01 0.00
105 -0.16 -0.12 -0.07 -0.06 -0.04 -0.02 -0.02 -0.01
110 -0.18 -0.13 -0.07 -0.07 -0.05 -0.02 -0.03 -0.01
115 -0.19 -0.13 -0.07 -0.07 -0.05 -0.03 -0.03 -0.02
120 -0.21 -0.13 -0.07 -0.08 -0.06 -0.03 -0.04 -0.02
125 0.15 -0.12 -0.07 -0.07 -0.06 -0.03 -0.04 -0.03

Pricing results Table 5 displays the pricing performance of both models in the valuation of worst-of

derivatives with maturity within three years, whereas table 6 provides the same information for out-

performance options. We can see, from table 5, that the worst-of forward associated with the HCFTS

specification is much lower than the one obtained under the multi-Heston model. The pricing discrepan-

cies are also relevant for at-the-money worst-of calls and for in-the-money worst-of options. Importantly,

in-the-money options are the ones that exhibit more sensitivity with respect to the worst-of forward. Re-

8
garding the outperformance option, the price discrepancies between models are also relevant. Note that,

since there is an inverse relationship between the worst-of forward and the at-the-money outperformance

option2 , the price discrepancies between models, in absolute terms, are identical for both derivatives.

Table 5: Pricing worst-of derivatives with maturity within three years under the HCFTS Wishart
specification and the multi-factor Heston model. Prices and strikes are expressed in percentage terms.

worst-of Forward 86.77 87.21


worst-of Call worst-of Put

Strike Wishart Multi-Heston Wishart Multi-Heston


50 37.744 38.192 2.597 2.615
60 30.076 30.519 4.489 4.503
70 23.157 23.587 7.129 7.131
80 17.094 17.499 10.626 10.602
90 11.970 12.340 15.063 15.003
100 7.841 8.161 20.494 20.385
110 4.711 4.951 26.923 26.734
120 2.519 2.677 34.292 34.021
130 1.150 1.237 42.482 42.140
140 0.421 0.457 51.313 50.920
150 0.113 0.120 60.565 60.143

Table 6: Pricing outperformance options with maturity within three years under the Wishart specifi-
cation and the multi-factor Heston model. Prices and strikes are expressed in percentage terms.

Outperformance
max(S1T − S2T − K, 0)
Strike (K) -30 -20 -10 0 10 20 30
Wishart 29.619 21.279 14.063 8.446 4.613 2.294 1.041
Multi-Heston 29.406 20.945 13.622 7.997 4.143 1.892 0.758

4.1.2 Calibrating the worst-of forward

Let us assume that we have available market prices associated with an at-the-money spread option, which

is equivalent to have available the worst-of forward. As before, we focus on derivatives with maturity

within three years. If we try to calibrate the worst-of forward, the at-the-money European call on the

equally weighted basket, as well as the European call on the equally weighted basket with strike 120%,

in this case, there is no solution within the set of symmetric positive semi-definite correlation matrices.

Therefore, it is not possible to match at the same time equally weighted basket options and spread options

(worst/best-of forward) for this example. This fact indicates the existence of a stochastic correlation

premium since there are some combinations of parameters associated with the Wishart specification that

generate prices for multi-asset European options which a multi-asset stochastic covariance model with

constant correlations is not able to replicate.


2 In particular, we have that max (S1T − S2T , 0) = S1T − min (S1T , S2T ).

9
Since it is not possible to perfectly match all the available market instruments, I will find the param-

eters corresponding to the multi-Heston model that provide the best possible fit to the worst-of forward,

the at-the-money European call on the equally weighted basket and European call on the equally weighted

basket with strike 120%. The estimated parameters are given by: ρS1 S2 = 0.9050, ρS1 V22 =ρS2 V11 =−0.8600

and ρV11 V22 = 0.7500.

Table 7 shows the difference of implied volatilities associated with equally weighted basket options,

expressed in percentage terms, generated under the HCFTS specification and under the multi-Heston

model. We can see from the table that, contrary to the previous case, now we are not able to match the

whole volatility skew associated with equally weighted options. In particular, for this parametric com-

bination, the multi-Heston model generates less skew than the Wishart specification. This is consistent

with a lower estimated value, in absolute terms, associated with the cross-asset-variance correlations.

Table 7: Implied volatility differences (expressed in percentage) corresponding to equally weighted


basket options between the HCFTS Wishart specification and the multi-factor Heston model for different
maturities (expressed in years).

Strike/Maturity 0.75 1 1.50 2 2.50 3 3.50 4


65 0.47 0.42 0.37 0.32 0.33 0.34 0.32 0.33
70 0.38 0.34 0.30 0.27 0.29 0.30 0.29 0.30
75 0.29 0.27 0.24 0.23 0.25 0.26 0.25 0.27
80 0.21 0.19 0.19 0.19 0.21 0.23 0.22 0.24
85 0.12 0.12 0.14 0.14 0.17 0.20 0.19 0.21
90 0.03 0.04 0.08 0.10 0.13 0.17 0.16 0.18
95 -0.07 -0.03 0.03 0.06 0.10 0.14 0.14 0.15
100 -0.16 -0.11 -0.02 0.01 0.06 0.10 0.11 0.13
105 -0.25 -0.18 -0.08 -0.03 0.02 0.07 0.08 0.10
110 -0.33 -0.24 -0.13 -0.07 -0.02 0.03 0.05 0.08
115 -0.40 -0.30 -0.17 -0.12 -0.06 0.00 0.02 0.05
120 -0.38 -0.33 -0.21 -0.16 -0.10 -0.04 -0.01 0.03
125 -0.25 -0.31 -0.25 -0.20 -0.14 -0.06 -0.04 0.00

Pricing results Table 8 displays the pricing performance of both models in the valuation of worst-of

derivatives with maturity within three years and table 9 provides the same information for outperfor-

mance options. For this example, the price differences corresponding to the worst-of forward under

both models is much lower. But even in this case, the price discrepancies associated with worst-of op-

tions are relevant especially for at-the-money options. In the case of outperformance options, the main

discrepancies correspond to out-of-the-money options.

10
Table 8: Pricing worst-of derivatives with maturity within three years under the HCFTS Wishart
specification and the multi-factor Heston model. Prices and strikes are expressed in percentage terms.

worst-of Forward 86.77 86.70


worst-of Call worst-of Put

Strike Wishart Multi-Heston Wishart Multi-Heston


50 37.744 37.827 2.597 2.741
60 30.076 30.226 4.489 4.700
70 23.157 23.374 7.129 7.408
80 17.094 17.367 10.626 10.960
90 11.970 12.274 15.063 15.427
100 7.841 8.149 20.494 20.862
110 4.711 4.980 26.923 27.254
120 2.519 2.725 34.292 34.558
130 1.150 1.277 42.482 42.670
140 0.421 0.482 51.313 51.435
150 0.113 0.132 60.565 60.645

Table 9: Pricing outperformance options with maturity within three years under the Wishart specifi-
cation and the multi-factor Heston model. Prices and strikes are expressed in percentage terms.

Outperformance
max(S1T − S2T − K, 0)
Strike (K) -30 -20 -10 0 10 20 30
Wishart 29.619 21.279 14.063 8.446 4.613 2.294 1.041
Multi-Heston 29.595 21.285 14.094 8.510 4.533 2.154 0.902

4.2 Case 2: Low correlation and positive term structure of correlation

Let us consider the parametric specification of table 10 associated with the Wishart model. Table 11

shows that, in this case, all the initial correlation levels are lower, in absolute value, than their long-

term counterparts. In this sense, I denote the specification of table 10 low correlation and positive term

structure LCPTS specification.

Table 10: Parameters specification in a bivariate Wishart volatility model for assets returns

Parameter Value Parameter Value


M11 -0.4500 X220 0.0425
M12 0.0000 Q11 0.2900
M21 0.0000 Q12 0.1250
M22 -0.4500 Q21 0.0650
X110 0.0500 Q22 0.2500
X120 0.0150 ρ1 -0.6300
X210 0.0150 ρ2 -0.6200
β 1.1000
Note. X0 represents the initial level
associated with the covariance matrix.

11
Panel A: Asset 1 Panel B: Asset 2

0.6

0.5 0.6
Implied Volatility

0.4 0.5

Implied Volatility
0.4
0.3
0.3
0.2
0.2

0.1 0.1
4 0
4
3.5
3 0 3 0.5

2.5
0.5
2 2 1
1 1.5
1 1 1.5
1.5
0.5 Strike
Maturity (years) 0 2 Strike Maturity (years) 0 2

Figure 3: Implied volatility surfaces corresponding to the Heston model. Strike prices are expressed as a
percentage of the asset price, whereas maturities are expressed in years. The surfaces are calculated using the
parameters of table 12 together with the following parameters values qi = 0.03, i = 1, 2, r = 0.015.

Table 11: Correlations associated with the specification of table 10 corresponding to the bivariate
Wishart volatility model.

Initial levels Long-term levels


ρ(0)V11 V22 0.6029 ρV11 V22 0.6997
ρ(0)S1 S2 0.3254 ρS1 S2 0.6320
ρ(0)S1 V22 -0.2721 ρS1 V22 -0.5285
ρ(0)S2 V11 -0.2681 ρS2 V11 -0.4742
Notes. The long-term levels are calculated using
X instead of X0 .

Table 12 displays the Heston (1993) parameters associated with the LCPTS specification and figure

3 provides the corresponding volatility surfaces. Finally, figure 4 shows the implied volatility surface

associated with European equally weighted basket options under the LCPTS Wishart specification as

well as the at-the-money implied volatility term structure. We can see from the figure that, in this case,

there is a positive term structure of volatility consistent with the positive terms structure of correlation.

Table 12: Heston parameters associated with the bivariate Wishart volatility model.

Asset 1 Asset 2
κ1 0.9000 0.9000 κ2
θ1 0.1080 0.0955 θ2
ζ1 0.5944 0.5590 ζ2
ρS1 V11 -0.7503 ρS2 V22 -0.8363
v10 0.0500 v20 0.0425

12
Panel A: Implied volatility Surface Panel B: At−the−money implied volatility term structure
0.22

0.21

0.5
0.2
0.45

0.4
Implied Volatility

0.35

Implied volatility
0.19
0.3

0.25

0.2 0.18

0.15

0.1
0.17
4
3.5 0.4
3
0.6
2.5
0.16
2 0.8
1.5 1
1
1.2
0.5
0.15
0 1.4 Strike 0.5 1 1.5 2 2.5 3 3.5 4
Maturity (years)
Time (years)

Figure 4: Implied volatility surface corresponding to equally weighted basket options under the LCPTS Wishart
specification (panel A) and at-the-money implied volatility term structure (panel B).

4.2.1 Equally weighted calls (70-100) smile calibration & worst-of forward calibration

In this case, we focus on derivatives with maturity within three years and we try to calibrate the worst-

of forward, the at-the-money European call on the equally weighted basket and the European call on

the equally weighted basket with strike 70%. The calibration results are given by: ρS1 S2 = 0.5447,

ρS1 V22 =ρS2 V11 =−0.7400 and ρV11 V22 = 0.7500. Table 13 shows the difference of implied volatilities

associated with equally weighted basket options, expressed in percentage terms, generated under the

LCPTS specification and under the multi-Heston model. For the analyzed maturity of three years we

are able to match the at-the-money implied volatility and the one corresponding to the 70% strike. But

for out-of-the-money calls we have discrepancies. In particular, the smile is less steep under the LCPTS

specification than under the calibrated parametric specification associated with the multi-Heston model.

Note that in this case, where we have introduced correlation term structure within the Wishart model, the

multi-Heston model with constant instantaneous correlations is not able to match the implied volatility

skew associated with maturities different from three years.

13
Table 13: Implied volatility differences (expressed in percentage) corresponding to equally weighted
basket options between the LCPTS Wishart specification and the multi-factor Heston model.

Strike/Maturity 0.75 1 1.50 2 2.50 3 3.50 4


65 -0.17 -0.19 -0.13 -0.04 0.01 0.04 0.05 0.10
70 -0.31 -0.29 -0.19 -0.08 -0.03 0.02 0.03 0.09
75 -0.43 -0.38 -0.23 -0.12 -0.06 0.00 0.01 0.08
80 -0.54 -0.44 -0.27 -0.15 -0.08 -0.02 0.00 0.07
85 -0.62 -0.50 -0.30 -0.17 -0.10 -0.03 0.00 0.07
90 -0.69 -0.54 -0.32 -0.18 -0.10 -0.03 0.01 0.07
95 -0.74 -0.56 -0.32 -0.18 -0.10 -0.02 0.02 0.08
100 -0.74 -0.55 -0.30 -0.16 -0.08 0.00 0.04 0.10
105 -0.66 -0.48 -0.24 -0.12 -0.05 0.03 0.06 0.11
110 -0.42 -0.30 -0.14 -0.05 0.00 0.06 0.09 0.14
115 0.03 0.00 0.03 0.05 0.07 0.12 0.14 0.17
120 0.46 0.37 0.25 0.19 0.17 0.19 0.19 0.20
125 0.66 0.62 0.49 0.35 0.29 0.27 0.25 0.25

Pricing results Table 14 displays the pricing performance of both models in the valuation of worst-of

derivatives with maturity within three years, and table 15 provides the same information for outperfor-

mance options. Regarding worst-of derivatives, the main discrepancies correspond to out-of-the-money

worst-of calls where the multi-Heston model misprices these options when compared with the LCPTS

Wishart specification. Note that, from table 13, the implied volatility associated with out-of-the-money

equally weighted basket calls is higher under the LCPTS that under the multi-Heston model with con-

stant correlations. In the case of outperformance options, as before, the main discrepancies correspond

to out-of-the-money options.

Table 14: Pricing multi-asset options under the Wishart specification and the multi-factor Heston
model. Prices and strikes are expressed in percentage terms.

worst-of Forward 79.88 79.87


worst-of Call worst-of Put

Strike Wishart Multi-Heston Wishart Multi-Heston


50 32.359 32.402 3.794 3.842
60 25.177 25.218 6.173 6.218
70 18.795 18.820 9.351 9.380
80 13.319 13.311 13.435 13.431
90 8.833 8.771 18.508 18.451
100 5.379 5.254 24.614 24.494
110 2.936 2.761 31.732 31.561
120 1.399 1.212 39.754 39.572
130 0.568 0.416 48.483 48.336
140 0.194 0.102 57.669 57.582
150 0.058 0.017 67.093 67.057

14
Table 15: Pricing multi-asset options under the Wishart specification and the multi-factor Heston
model

Outperformance
max(S1T − S2T − K, 0)
Strike (K) -30 -20 -10 0 10 20 30
Wishart 33.477 26.396 20.214 15.052 10.941 7.789 5.446
Multi-Heston 33.542 26.461 20.261 15.057 10.868 7.644 5.247

The previous examples suggest that, conversely to what happens with single-assets options without

path dependency, equally weighted basket options do not include all the information.

5 Summary and conclusions

We have considered two multi-asset models that account for stochastic covariances. Under the Wishart

specification the instantaneous correlations are also stochastic whereas under the multi-Heston framework

instantaneous correlations are constant. The article has investigated if the consideration of stochastic

correlation has any impact on the valuation of exotic multi-asset derivatives when we calibrate both

models to market prices of European multi-asset derivatives such as European equally weighted basket

options and/or outperformance options. The results show that, even for European exotic derivatives

such as European worst-of options, price differences may be relevant for some parameters combinations.

This fact is related to the existence of model risk, which is the risk that arises from the utilization of

an inadequate model. Importantly, this article shows that, when pricing multi-asset exotic derivatives,

matching the prices of European equally weighted basket options, quoted in the market, does not guaranty

the absence of model risk even in the case where the exotic payoff is observed only at maturity. Note

that, as it is well known, the price discrepancies would have been much greater if we had considered a

local volatility framework with constant instantaneous correlation.

Note that in the examples of this article I have investigated the particular case where the matrix

M is diagonal, so that the Wishart model and the multi-Heston specification share the same individual

implied volatility surfaces. Moreover, I have considered a two-assets world. But, when we relax the

previous assumptions price discrepancies between models may be greater. Regarding the calibration of

both models for n assets (n > 2) we also would have to consider the equally weighted European options

and/or the worst-of forward for each pair of underlying assets within the whole basket.

15
References

[1] Bru, M. F., 1991. Wishart processes, Journal of Theoretical Probability, 4, 725-743.

[2] Carr, P., Madan D., 2001. Optimal Positioning in Derivative Securities. Quantitative Finance, 1,

19-37.

[3] Da Fonseca, J., Grasselli, M., Tebaldi, C., 2007. Option pricing when correlations are stochastic: an

analytical framework. Review of Derivatives Research 10, 151-180.

[4] Da Fonseca, J., Grasselli, M., Tebaldi, C., 2008. A multifactor volatility Heston model, Quantitative

Finance, 8, 591-604.

[5] Gourieroux, C., Sufana, R. Derivative pricing with multivariate stochastic volatility: application to

credit risk, Les Cahiers du CREF, 2004, Working Paper No. 04-09.

[6] Heston, S. L., 1993. A closed-form solution for options with stochastic volatility with applications to

bond and currency options. Review of Financial Studies 6, 327-343.

[7] Lewis, A.L., 2000. Option Valuation Under Stochastic Volatility, with Mathematica Code. Finance

Press, Newport Beach, California.

[8] Marabel Romo, J. 2012. Worst-of options and correlation skew under a stochastic correlation frame-

work. International Journal of Theoretical and Applied Finance, 15,1-32.

[9] Solnik, B., Boucrelle, C., Le Fur, Y., 1996. International market correlation and volatility. Financial

Analysts Journal, 52,17-34.

16

You might also like