You are on page 1of 6

Geodesic strikes for composite, basket, Asian, and spread options

Peter Jckel
a
First version: 28th July 2012
This version: 28th July 2012

Abstract

In practice, it is not uncommon for X to be the market price of a standard investment asset such as an
equity share, denominated in its domestic currency
DOM, and for Y to be an FX rate that converts the
nal investment asset value to a target currency TAR.
As a consequence of the natural direction of the FX
rate, i.e., DOMTAR, meaning the value of one DOM
currency unit expressed in units of TAR, the valuation of a composite option may also incur a quanto
eect since the asset X and the FX rate DOMTAR
cannot be martingales in the same measure, and this
has to be taken into account. For the purpose of this
article, however, we shall assume that all involved
underlyings are martingales in the same measure. In
practice, this may mean that we have to determine
an eective quanto forward for the asset X in the target currency TAR by other means of approximation
prior to being able to commence with our eective
geodesic strike procedure. We shall return to this
point at the end of section 3.
Basket options are, conventionally, derivatives
with a payo of the form

We discuss simple methodologies for the selection of


most relevant or eective strikes for the assessment
of appropriate implied volatilities used for the valuation of composite, basket, Asian, and spread options
following the spirit of geodesic strikes [ABOBF02].

Introduction

In the context of vanilla, or near-vanilla, derivatives


trading, we often also encounter composite, basket,
Asian, and spread options. Whilst their respective
payo rules do in principle warrant treating them as
genuinely exotic options, it is often desirable to use
a simple approximation for the sake of tractability.
A popular approach is to use a relatively simple valuation methodology based on multivariate geometric Brownian motion, the well-known Black-ScholesMerton framework, and to nd a suitable implied
volatility (or term structure thereof) for each underlying selected by the concept of a most relevant, or
eective, strike for each observation date. For a plain
vanilla option, clearly, the volatility must be taken
from the implied volatility surface at the options expiry date and at the options strike. For composite,
basket, Asian, and spread options, the most suitable
strike for the looking up of implied volatility is not
necessarily as obvious. In this document, we suggest
a systematic procedure that addresses this question.

( [

K])+ .

(2.2)

This is a vanilla option on a linearly weighted average


of a number of underlyings, whence it is also referred
to as an arithmetic basket option. In contrast, whilst
rarely traded, there is also the geometric basket option
( [ i Xiwi (T ) K])+ .
(2.3)
Asian options are in the framework of multivariate
geometric Brownian motion merely a special case of
arithmetic basket options in that the xings that contribute to the average are from the same underlying,
but for dierent observation times:

Composite, Basket, Asian, and


spread options

A composite option is a contract of European style


which pays at a future payment date Tpay the payo
( [X (T ) Y (T ) K])+ ,

i wi Xi (T )

( [

i wi X

(Ti ) K])+ .

(2.4)

(2.1) Spread options usually simply pay according to

( [(X(T ) Y (T )) K])+
(2.5)
with = 1 for calls and puts, based on two underlyings X and Y observed on the expiry date T Tpay . for two underlyings X and Y .
By allowing the subscript i on an observation index

Deputy head of Quantitative Research, VTB Capital


Xi to indicate a specic underlying as well as xing
Key words and phrases. eective strikes, geodesic, compostime, and by additionally permitting weights to be
ite options, basket options, Asian options, spread options.
1

negative as well as positive, it is clear that all of bas- relating to the dispersion matrix A according to
ket, Asian, and spread options take on the form of
C =AA .
(3.5)
an arithmetic average option
(2.6) The approximation of eective geodesic strikes is now
to nd a set of logarithmic shift coecients , which
Equally, it is evident that both composite and geo- relates to the eective strike for underlying #i as
metric basket options appear as a geometric average

Ki = Xi ei ,
(3.6)
option
wi
( [ i Xi K])+ .
(2.7)
such that the multivariate probability density of
Subsequently, we will therefore concentrate on these under the joint normal law with
two generic cases: geometric and arithmetic average
options.

=C
(3.7)
( [

i wi Xi

K])+ .

is maximal at = subject to the constraint

Geodesic strikes

f (K ) = K .

The formal derivation of the procedure for the calculation geodesic strikes in [ABOBF02] involves concepts of projection onto an eective local volatility
representation of a basket process for large deviations, as well as Varadhans geodesic theorem [Var67],
and is rather technical. In this document, we shall
attempt to obtain a similar result with a somewhat
less rigorous, though tractable, argument, bearing in
mind that the sole purpose of the exercise is to arrive at a set of suitably chosen eective strikes for
the lookup of implied volatilities from the underlyings implied volatility smiles. These volatilities are
then to be used in whatever near-vanilla approximation that is chosen for the respective target product.
We emphasize that it is clear that this process cannot
possibly arrive at a sophisticated exotic product pricing framework. Instead, it merely is intended to give
a procedure that suces for the simplistic, but very
fast, valuation of some near-vanilla products whilst
preserving some sensible consistency conditions.
The starting point is that all of the involved
stochastic nancial observables X are governed by
a joint log-normal law, and that there is a critical
level K for a function f () of the nancial observables, identied by the fact that the payout is of the
form
( (f (X) K))+ .
(3.1)

In other words, we seek


= arg max () f (K ) = K

aij zj

() :=

e 2

C 1

(2)n |C|

(3.10)

and K being given elementwise in equation (3.6).


The log-bilinear form of (3.10) allows us to simplify (3.9) to
z = arg min z z f (K ) = K
z

(3.11)

with
= A z .

(3.12)

As we shall see below, it turns out that the eective strikes themselves depend on (implied) volatilities. This makes the task of eective geodesic strike
calculation ultimately an implicit problem, since we
need the eective strikes to be able to look up the
implied volatilities in the rst place. In practice, we
resolve this by the approximation that all volatilities that show up in the eective strike formul are
to be taken as at-the-forward implied volatilities. In
this context, we recall that we mentioned at the end
of the rst paragraph in section 2 that, when some
of the underlyings require translation into the target
valuation measure, an approximation for this translation of forward and implied volatility smile has to
be employed separately and prior to the invocation of
the eective geodesic strike procedure. When choosing the quanto-translation procedure, it is useful to
bear in mind that the subsequent geodesic strike selection is only an approximation for the sake of analytic tractability for a range of near-vanilla products,

(3.2)

with

X= X ,

(3.9)

with

As is well known, without loss of generality, we can


transform the vector of nancial variables X to a
vector of independent standard Gaussian variables z
according to

Xi = Xi e 2 cii +

(3.8)

(3.3)

the matrix A being the dispersion or factor loading


matrix, and the log-covariance matrix C whose elements are
cij = ln Xi , ln Xj
(3.4)
2

As for the eective geodesic strikes, equation (3.6)


gives us

and thus judge the required level of sophistication


for the quanto translation in line with the overall
level of the chain of approximations. We emphasize,
however, that whilst the geodesic strike procedure is
only an approximation, its purpose is to be consistent
and accurate in a certain asymptotic sense, namely
that of the local volatility projection for large deviations, i.e, for out-of-the-money options, and that
of the geodesic distance asymptotics of [ABOBF02]
and [Var67].

3.1

Ki = Xi ei
with
i :=

Xiwi .

(X) =

The constraint (3.8) becomes


Az

ln

XY

X (X +XY Y )
2
2
X +2X XY Y +Y

KY = Y e

ln

XY

Y (Y +XY X )
2
2
X +2X XY Y +Y

(3.25)

lim KX =

(3.14)

Y 0

(3.26)

which is consistent with the exact plain vanilla op(3.15) tion we arrive at in this limit for a composite option,
and we obtain of course the symmetric equivalent for
X 0. When correlation is zero, the log-moneyness
of the composite option translates to log-moneyness
(3.16) of the underlyings as a function of the log-volatilityratio
:= 2 ln(X /Y )
(3.27)
(3.17) according to the logistic functions

with

Xiwi

G :=

(3.24)

w C w

(3.13) It is worth reecting on equation (3.25) with respect


to a few benchmark cases. First, let us consider the
case when Y 0. In that case, we have

K = G ew

j cij wj

KX = X e

We consider the payo f () being given by


geometric
average

for geometric average options. For composite options


as dened in (2.1), we have wi 1 and

Geodesic strikes for geometric average options

(3.23)

which we write as
w A z = 0
with

:= ln(K/G) .

In order to minimize the L2 -norm of z subject to the


constraint (3.16), we solve the Lagrange multiplier
problem

z = arg min z z + w A z
z

X =

1 + e

Y =

and

1 + e

(3.28)

as shown in gure 1. The graph illustrates the sym-

(3.18)
*
X

From

*
Y

/2

z z+ w Az

=0
z=z

(3.19)

0
-4

we have
z =

A w
2

-3

-2

-1

Figure 1: Underlying log-moneyness for composite options with


zero correlation according to equation (3.28).

(3.20)

metry X () + Y () = 1 which is intuitively appealing. Also, when volatilities are equal, the log
= 2
.
(3.21) moneyness is equally distributed over both underw C w
lyings, which is again what one would intuitively expect from a sensible eective strike approximation
Upon resubstitution into (3.20), we arrive at
formula. In the limit of XY = 1, we obtain
A w

z =
.
(3.22)
X =
and
Y =

(3.29)
w C w

2
1+e
1 + e2

which we substitute into (3.16) yielding


k = K , and the contribution of the continuous part
will be largely centered near k and tail o rapidly
*
X
as k due to the rapid decay of out-of-the
*
money options value B(X, k , (k )), assuming that

Y
/2
(k ) rises only moderately such as would be consis
tent with nite second and higher moments. Hence,
while the selection of an eective implied volatility

at k = K for the underlying does not exactly re0


-4
-3
-2
-1
0
1
2
3
4
produce the smile dependence one can obtain from
Figure 2: Underlying log-moneyness for composite options with the continuous integration over all strikes to innity (3.33), we at least capture the fact that the domiperfect correlation according to equation (3.29).
nant contribution does indeed come from the implied

volatility near k = K , as intended.


which we display in gure 2. We see in gure 2 that
with diverging volatilities, i.e., as diverges from 0,
3.2 Geodesic strikes for arithmetic averthe underlyings log-moneyness is less rapidly shifted
age options
from an equal split at = 0 to just one of the underlyings, as indicated by the lesser slope near zero, Here, we have
when comparing with the case XY = 0 in gure 1.
f
(X) =
wi Xi .
(3.34)
This behaviour is also intuitively desirable.
arithmetic
average
i
Finally, we consider the case of perfect correlation
given by the option under consideration actually be- The constraint (3.8) is
ing a square option
g(z ) = K
(3.35)
X2 K + .
(3.30)
with
In comparison, by a standard argument of continuous

g(z) :=
wi Xi e j aij zj .
(3.36)
replication, we can value this option based on the
i
Taylor expansion with complete remainder for any
From
smooth function h(x) around k

h(x) = h(k) + h (k)(x k) +

z z + g(z) K

h (z)(x z) dz

=0

(3.37)

z=z

(3.31) we obtain
whence

zi +

h(x) 1{x>k} = h(k) 1{x>k} + h (k)(x k)+

ali wl Xl e

alj zj

= 0

(3.38)

Equations (3.35) and (3.38) can be solved for and


the elements of the vector z by the aid of a nonk
linear root nding algorithm such as [DGW81]. The

Choosing h(X) = X 2 K and k := K , and taking eective geodesic strikes are then
the expectation over X, we obtain for the call option

Ki = Xi ei
(3.39)
2
E X K +
(3.33)
with

= 2kB(X, k, (k)) + 2 B(X, k , (k )) dk

= A z .
(3.40)
k
+

h (z)(x z)+ dz .

(3.32)

wherein (k ) is the implied volatility of the asset for

strike k , and B(X, K, ) is the Black vanilla call op


tion function for forward X, strike K, and implied
volatility . Equation (3.33) has two parts. First,

there is a vanilla call option struck at k = K with


absolute weight 2k. Second, we have a continuum of
vanilla call options for all strikes above k with density 2. For signicantly out-of-the-money options, as
stipulated by the large deviations asymptotics at the
heart of the geodesic argument, the dominant part
will be the discretely weighted call option struck at

As for an initial guess for z in equations (3.35)


and (3.38), we can either use zero, or proceed to nd
an expansion as follows.
Dening

X :=

wi Xi

:= ln(K/X) , (3.41)

and

we rewrite (3.35) as

wi Xi e
i

aij zj

= X e .

(3.42)

whence

We now seek an expansion given by


= 0 + (1) + 2 (2) +
z = 0 + z (1) + 2 z (2) +

1
2

z (2) =

(3.43)
(3.44)

1+2

A
C

(3.45) z =

A A
C C
1
3
2
C
( C )

A
C

(3.56)
Substituting (3.44) into (3.42) and expanding in
with 1 denoting the identity matrix. This nally
around 0 gives
gives us the second order expansion
(1)

i 1 +
i

1
2

(2)

aij zj + 2 zj
(1) 2

aij zj

+ 2

= 1 + + 1 2 + O 3
2

1
2

(3.57)

1+2

A A
C C
1
3
2
C
( C )

A
+ O 3 .
C

where we have used the normalized eective weights As an example, we show in gure 3 the rst and second order expansion solutions for a monthly observed

wi Xi
i := .
(3.46) Asian option of one year maturity in comparison with
X
a numerical solution for an arbitrarily chosen term
2 , equa- structure of arbitrage-free implied volatility.
Matching coecients up to order O
tion (3.45) gives

35%
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

T
i i

= 1

A z (1) = 1

30%

(3.47)
(3.48)

-0.5
25%

-1
20%

Az

(2)

1
2

1
2

(1)

A Az

(1)

(T )

, (3.49)

15%
-1.5

with the matrix dened to be diagonal and its elements being equal to those of . Expanding (3.38)
in yields
(1)

+ 2 zi

(1)

NL2SOL

5%

implied volatility [right axis]

(3.50)
(2)

wl Xl ali 1 +

= O

10%
Lognormal expansion, second order
-2

-2.5

(2)

zi

Lognormal expansion, first order

Figure 3: An example for the eective strike adjustments (3.40)


of rst and second order as given in (3.57) in comparison with
a numerical solution by the aid of NL2SOL of equations (3.35)
and (3.38) for a monthly observed Asian option of one year
maturity with an arbitrarily chosen term structure of arbitragefree implied volatility. All forwards were equal to 1, and the
strike was at 0.3.

(1)
alj zj

from which we derive

z (1) = (1) X A

0%

(3.51)

When any of the weights are negative, it is possi

ble for X to be zero. As a consequence, ln(K/X) is

z (2) = (1) X A A z (1) (2) X A


undened, and none of the above equations can be
(3.52)
evaluated. In that case, it may be better to use an
by matching coecients of up to second order. expansion in

Combining (3.48) and (3.51) results in


:= K X ,

(3.58)
1
1
(1) =
X C

which gives us, to rst order

(3.53)

z =
C
X

and hence
z (1) =

A
.
C

i.e.,
(3.54)
zi =

Further, combining (3.49) and (3.52) produces

K X
vX

aij wj Xj

(3.59)

wi Xi cij wj Xj .

(3.60)

with

1
1 3 C C
1
(2) =


2
2 2
X C
( C )

vX :=

(3.55)

i,j

Acknowledgement
The author is grateful to Charles-Henri Roubinet,
Head of Quantitative Research at VTB Capital, for
pointing out the results in [ABOBF02], and for authorizing the release of the ndings presented here
(originally from mid 2011, and rst completed with
the second order expansion in March 2012) into the
public domain.

References
[ABOBF02] M. Avellaneda, D. Boyer-Olson, J. Busca, and P. Friz.
Reconstructing volatility. Risk, pages 9195, October 2002.
www.math.nyu.edu/faculty/avellane/
Avellaneda.pdf.
[DGW81]

J. E. Dennis, D. M. Gay, and R. E. Welsch. NL2SOL


an adaptive nonlinear least-squares algorithm. ACM
Transactions on Mathematical Software, 7:348368,
1981. www.netlib.org/toms/573.

[Var67]

S. Varadhan. Diusion processes in a small time interval. Communications on Pure and Applied Mathematics, 20(4):659685, November 1967.