WeierstraÿInstitut

für Angewandte Analysis und Sto hastik

im Fors hungsverbund Berlin e.V.

Preprint ISSN 0946  8633

E ient Computation of Option Pri e Sensitivities

Using Homogeneity and other Tri ks
Oliver Reiÿ1, Uwe Wystup2
submitted: May 24th 2000

1 Weierstrass-Institute for 2 Commerzbank

Applied Analysis and Sto hasti s Treasury and Finan ial Produ ts
Mohrenstraÿe 39 Neue Mainzer Straÿe 32-36
D - 10117 Berlin D - 60261 Frankfurt am Main
Germany Germany
E-Mail: reisswias-berlin.de E-Mail: wystupmathnan e.de
URL: http://www.wias-berlin.de/reiss URL: http://www.mathfinan e.de

Preprint No. 584

Berlin 2000

WIAS
2000 Mathemati s Subje t Classi ation. 91-08, 91B28.
Key words and phrases. Cal ulation of Greeks, Derivatives of option pri es, Homogeneity
properties of nan ial markets.
Oliver Reiÿ is partially aliated to Delft University, by support of NWO Netherlands.
Edited by
WeierstraÿInstitut für Angewandte Analysis und Sto hastik (WIAS)
Mohrenstraÿe 39
D  10117 Berlin
Germany

Fax: + 49 30 2044975
E-Mail (X.400): =de;a=d400-gw;p=WIAS-BERLIN;s=preprint
E-Mail (Internet): preprintwias-berlin.de
World Wide Web: http://www.wias-berlin.de/
 Abstra t
No front-o e software an survive without providing derivatives of op-
tions pri es with respe t to underlying market or model parameters, the so
alled Greeks. We present a list of ommon Greeks and exploit homogene-
ity properties of nan ial markets to derive relationships between Greeks out
of whi h many are model-independent. We apply the results to European
style options, rainbow options, as well as options pri ed in Heston's sto hasti
volatility model and avoid exorbitant and time- onsuming omputations of
derivatives whi h even strong symboli al ulators fail to produ e.

1 Introdu tion
The omputation of sensitivities of option pri es, the so- alled Greeks, is often
umbersome - both for the mathemati ian and for symboli al ulators. This paper
provides methods to avoid dierentiation as mu h as possible. Many Greeks are
related among ea h other. These relations are based on model-independent homo-
geneity of time and pri e level of a nan ial produ t on the one hand and model
dependent relations su h as the partial dierential equation the value fun tion must
satisfy and relations implied by the assumed distribution of the underlying. The
basi market model we use is the Bla k-S holes model with sto ks paying a ontin-
uous dividend yield and a riskless ash bond. This model supports the homogeneity
properties whi h are valid in general, but its stru ture is so simple, that we an
on entrate on the essential statements of this paper. We will also dis uss how to
extend our work to more general market models.

We list the ommonly used Greeks and their symbols. We do not laim this list
to be omplete, be ause one an always dene more derivatives of the option pri e
fun tion.

As spe ial ases we look at the Greeks of European options in the Bla k-S holes
model in one dimension. It turns out, that one only needs to know two Greeks in
order to al ulate all the other Greeks without dierentiating.

Another interesting example is a European derivative se urity depending on two

assets. For su h rainbow options the analysis of the risk due to hanging orrela-
tion of the two assets is very important. We will show how this risk is related to
simultaneous hanges of the two underlying se urities.

There are several appli ations of these homogeneity relations.

1
 1. It helps saving time in omputing derivatives.

2. It produ es a robust implementation ompared to Greeks via dieren e quo-

tients.

3. It allows to he k the quality and onsisten y of Greeks produ ed by nite-

dieren e-, tree- or Monte Carlo methods.

4. It admits a omputation of Greeks for Monte Carlo based values.

5. It shows relationships between Greeks whi h wouldn't be noti ed merely by

looking at dieren e quotients.

1.1 Notation

S sto k pri e or sto k pri e pro ess

B ash bond, usually with risk free interest rate r
r risk free interest rate
q dividend yield ( ontinuously paid)
 volatility of one sto k, or volatility matrix of several sto ks
 orrelation in the two-asset market model
t date of evaluation (today)
T date of maturity
 =T t time to maturity of an option
x sto k pri e at time t
f (
) payo fun tion
v (x; t; : : :) value of an option
k strike of an option
l level of an option
vx partial derivation of v with respe t to x (and analogous)

The standard normal distribution and density fun tions are dened by

n ( t) = p1 e
1 t2
2 (1)
Z x2
N ( x)

= n(t) dt (2)
1

1 x2 2xy + y 2!
n2 (x; y ; ) = p
2 1  2 exp 2(1 2 )
(3)

Zx Zy
N2(x; y; ) =
1 1
n2 (u; v ; ) du dv (4)

See http://www.MathFinan e.de/fronto e.html for a sour e ode to ompute N2.

2
1.2 The Greeks

Delta
vx
Gamma vxx
Theta  vt
Rho  vr in the one-sto k model
Rhor r vr in the two-sto k model
Rhoq q vq
Vega  v
Kappa  v orrelation sensitivity (two-sto k model)

Greeks, not so ommonly used:

 xv
, sometimes alled gearing
Leverage
v x sometimes
Vomma 0 v
Speed vxxx
Charm vxt
Color vxxt
Cross vx
Forward Delta
F vF
Driftless Delta
dl
eq
Dual Theta Dual  vT
Strike Delta
k vk
Strike Gamma
k vkk
Level Delta
l vl
Level Gamma
l vll
Beta 12 1  two-sto k model
2

2 Fundamental Properties
2.1 Homogeneity of Time

In most ases the pri e of the option is not a fun tion of both the urrent time t and
the maturity time T , but rather only a fun tion of the time to maturity  = T t
implying the relations

 = vt = v = vT = Dual : (5)

This relationship extends naturally to the situation of options depending on several

intermediate times su h as ompound or Bermuda options.

3
2.2 S ale-Invarian e of Time

We present the prin iple of the s ale-invarian e of time in this se tion, be ause this
prin iple holds in general. In a market model parameters may be quoted on an
annual basis. We illustrate this idea in a Bla k-S holes framework, in whi h the
volatility is su h a model parameter. The same idea an easily be applied to other
market models.

We may want to measure time in units other than years in whi h ase interest rates
and volatilities, whi h are normally quoted on an annual basis, must be hanged
a ording to the following rules for all a > 0.

 ! a
r ! ar
q ! aq
 ! pa (6)

The option's value must be invariant under this res aling, i.e.,

 p
v (x; ; r; q; ; : : :) = v (x; ; ar; aq; a; : : :) (7)
a
We dierentiate this equation with respe t to a and obtain for a=1
1
0 =   + r + qq +  ; (8)
2
a general relation between the Greeks theta, rho, rhoq and vega. Based on the
relation

v (x1 ; :::; xn ; ; r; q1 ; :::; qn ; 11 ; :::; nn ) =

 p
v (x1 ; :::; xn ; ; ar; aq1 ; :::; aqn ; a11 ; :::; ann )
p (9)
a
we obtain

Theorem 1 (s ale invarian e of time)

n
X 1 X n
0 =   + r + qi qi +   ;
2 i;j =1 ij ij
(10)
i=1

where ij denotes the dierentiation of v with respe t to ij .

2.3 S ale Invarian e of Pri es

The general idea is that value of se urities may be measured in a dierent unit, just
like values of European sto ks are now measured in Euro instead of in- urren ies.
Option ontra ts usually depend on strikes and barrier levels. Res aling an have

4
dierent ee ts on the value of the option. Essentially we may onsider the following
types of homogeneity lasses. Let v (x; k) be the value fun tion of an option, where
x is the spot (or a ve tor of spots) and k the strike or barrier or a ve tor of strikes
or barriers. Let a be a positive real number.

Denition 1 (homogeneity lasses) We all a value fun tion k-homogeneous of

degree n if for all a > 0
v (ax; ak) = an v (x; k): (11)

We all an options whose value fun tion is strike-homogeneous of degree 1 a strike-

dened option and similarly an option whose value fun tion is level-homogeneous of
degree 0 a level-dened option.

The value fun tion of a European all or put option with strike K is then K-
homogeneous of degree 1, a digital option whi h pays a xed amount if the sto k
pri e is higher than a level L is L-homogeneous of degree 0. The path-independent
barrier all option paying (S k)+ IfS>K g is (k; K )-homogeneous of degree 1. A
power all with ap paying min(C; ((S K )+ )2 ) has a homogeneity stru ture of
v (aS; aK; a2 C ) = a2 v (S; K; C ).
We show how su h a s ale invarian e an be used to determine some relations among
the Greeks. We explain this with two examples. In the rst example we analyze a
strike-dened option and in the se ond one we on entrate on a level dened option.
The generalization to options with some more parameters like the mentioned path-
independent barrier all or power- all an easily be done. For the barrier all one
an use the results from the multi-dimensional strike-dened option (26) and (27).

2.3.1 Strike-Delta and Strike-Gamma

For a strike-dened value fun tion we have for all a; b > 0
abv (x; k) = v (abx; abk): (12)

We dierentiate with respe t to a and get for a=1

bv (x; k) = bxvx (bx; bk) + bkvk (bx; bk): (13)

We now dierentiate with respe t to b get for b=1

v (x; k) = xvx + xvxx x + xvxk k + kvk + kvkx x + kvkk k (14)

= x
+ x2 + 2xkvxk + k
k + k2 k : (15)

If we evaluate equation (13) at b = 1 we get

v = x
+ k
k : (16)

5
We dierentiate this equation with respe t to k and obtain

k = xvkx +
k + k k ; (17)

kxvkx = k2 k : (18)

Together with equation (15) we on lude

x2 = k2 k : (19)

2.3.2 Level-Delta and Level-Gamma

For a level-dened value fun tion we have for all a; b > 0
v (x; l) = v (abx; abl): (20)

We dierentiate with respe t to a and get at a=1

0 = vx (bx; bl)bx + vl (bx; bl)bl: (21)

If we set b = 1 we get the relation

x +
l l = 0: (22)

Now we dierentiate equation (21) with respe t to b and get at b=1

0 = vxx x2 + 2vxl xl + vll l2 : (23)

One the other hand we an dierentiate the relation between delta and level-delta
with respe t to l and get

vxl x + l l +
l = 0: (24)

Together with equation (23) we on lude

x2 + x
= l2 l + l
l : (25)

In general we obtain

Theorem 2 (pri e homogeneity)

n
X m
X
v= xi
i + kj
kj (26)
i=1 j =1
n
X m
X
xi xj = ki kj k
ij ij (27)
i;j =1 i;j =1
for strike-dened options and
n
X m
X
0= xi
i + lj
lj (28)
i=1 j =1
n
X n
X m
X m
X
xi xj ij + xi
i = li lj lij + li
li (29)
i;j =1 i=1 i;j =1 i=1
for level-dened options.

6
3 European Options in the Bla k-S holes
Model
We start with relations among Greeks for European laims in the n-dimensional
Bla k-S holes model

dSi (t) = Si (t)[(r qi ) dt + i dWi (t)℄; i = 1; : : : ; n (30)

Cov(Wi(t); Wj (t)) = ij t; (31)

where r is the risk-free rate, qi the dividend rate of asset i or foreign interest rate of
ex hange rate i, i the volatility of asset i and (W1 ; : : : ; Wn ) a standard Brownian
motion (under the risk-neutral measure) with orrelation matrix . Let v denote
today's value of the payo f (S1 (T ); : : : ; Sn (T )) at maturity T . Then it is known
that v satises the Bla k-S holes partial dierential equation

n
X 1 X n
0 = v rv + xi (r qi )vxi + ( Æ  T )ij xi xj vxi xj : (32)
i=1 2 i;j =1

3.1 Relations among Greeks Based on the Log-Normal Dis-

tribution

The value fun tion v has a representation given by the n-fold integral
Z  px +  
v=e r f : : : ; Si (0)ei i i ; : : : g (~x; ) d~x; (33)

i = r qi 1 2 g (~x; ) n-variate
where
2 i and is the standard normal density with
orrelation matrix . Sin e we do not want to assume dierentiability of the payo
f, but we know that the transition density
p g is dierentiable, we dene a hange

 xi +i  , whi h leads to
the variables yi = Si (0)e i

Z !
ln Siy(0)
i i  d~y
v=e r f (: : : ; yi; : : :)g p
 
; Q p :
yi i 
(34)
i

3.1.1 Properties of the Normal Distribution

We olle t some properties of the multivariate normal density fun tion g. We sup-
pose that the ve tor X of n random variables with means zero and unit varian es
has a nonsingular normal multivariate distribution with probability density fun tion

1 T 
g (x1 ; : : : ; xn ; 11 ; : : : ; nn) = (2 )
1
2n jCj 1
2 exp
2
x Cx : (35)

Here C is the inverse of the ovarian e matrix of X , whi h is denoted by . Then

the following identity published in [3℄ an be proved easily by writing the density in
terms of its hara teristi fun tion.

7
Theorem 3 (Pla kett's Identity)
g 2g
= : (36)
ij xi xj

In the two-dimensional ase this reads as

n2 (x; y ; )  2 n2 (x; y ; )

= ; (37)
 xy
whi h an be extended readily to the orresponding umulative distribution fun tion,
i.e.,
 N2 (x; y ; )  2 N2 (x; y ; )
= = n2 (x; y ; ): (38)
 xy

3.1.2 Correlation Risk and Cross-Gamma

Using the abbreviation g

 2 g the ross-gamma and orrelation risk are
=
jk xj xk

2v 1 Z d~y
Sj (0)Sk (0)
= e r
Sj (0)Sk (0)j k 
f (: : : ; yi; : : :)gjk Q p ;
yii 
(39)

Z
v d~y
jk
= e r f (: : : ; yi; : : :)gjk Q p :
yi i 
(40)

Invoking Pla kett's identity (36) saying that gjk = gjk leads to

Theorem 4 ( ross-gamma- orrelation-risk relationship)

v 2v
= Sj (0)Sk (0)j k  : (41)
jk Sj (0)Sk (0)

3.1.3 Interest Rate Risk and Delta

A similar omputation yields

Theorem 5 (delta-rho relationship)

v v
= Sj (0) ; (42)
qj Sj (0)
0 1
v n
X v A
=  v Sj (0) : (43)
r j =1 Sj (0)

8
3.1.4 Volatility Risk and Gamma
The rst and se ond derivative of the density g satisfy

n
X
gj = g xi Cij ; (44)
i=1
Xn Xn
gjk = g xi Cij xi Cik gCkj : (45)
i=1 i=1

For the j -th vega we nd thus

Z n !
v X d~y
j
j
= e r f
g
xi Cij xj 1 Q p ;
yii 
(46)
i=1
yi

ln Si (0) (r i q + 21 i2 ) p
xj = p
 
= xj j  ; (47)
i

where we omit the arguments of f and g to simplify the notation. For the ross
gammas we derive

2v Z d~y
j k Sj (0)Sk (0)
Sj (0)Sk (0)
=e r f
g
Bjk Q p ;
yii 
(48)

Xn Xn
Bjk = xi Cij xi Cik Ckj
n
X p
xi Cij k  Æjk : (49)
i=1 i=1 i=1

We now multiply by jk , sum over k, remember that  is the inverse matrix of C
and obtain

n
X 2v Z d~y
jk j k Sj (0)Sk (0)
Sj (0)Sk (0)
=e r f
g
Dj Q p ;
yi i 
(50)
k=1
n n n

X
Dj = xi Cij xj 1
X
xi Cij xj +
X
xi Cij xj : (51)
i=1 i=1 i=1

In summary we obtain

Theorem 6 (gamma-vega relationship)

v Xn 2v
j = jk j k Sj (0)Sk (0) : (52)
j k=1 Sj (0)Sk (0)

In dimension one the gamma-vega and delta-rho relationships are also mentioned
in [4℄. Shaw shows there that v  S 2 (t)vS (t)S (t) satises the Bla k-S holes partial
dierential equation and is hen e identi ally zero for path-independent options. We
note that the gamma-vega relationship does not hold for barrier options, simply
be ause gamma and vega are not equal at the barrier.

9
4 The One-Dimensional Case
4.1 Results for European Claims in the Bla k-S holes Model

We list several relations for European options.

1
0 =   + r + qq +   s ale invarian e of time (53)
2
v = x
+ k
k pri e homogeneity and strikes (54)

x 2 = k 2 k pri e homogeneity and strikes (55)

x
= l
l pri e homogeneity and levels (56)

2 2
x + x
= l + l
l l pri e homogeneity and levels (57)

 =  (v x
) delta-rho relationship (58)

 + q =  v rates symmetry (59)

1
rv =  + (r q )x
+  2 x2 Bla k-S holes PDE (60)
2
1
qv =  + (q r)k
k +  2 k2 k dual Bla k-S holes (strike) (61)
2
1
rv =  + (q r +  )l
l +  2 l2 l dual Bla k-S holes (level)
2 (62)
2
q =  x
delta-rho relationship (63)

 =  k
k ombination of (63) and (54) (64)

 =  x2 gamma-vega relationship (65)

An interpretation of equation (65) an be found in [6℄. We would like to point out

that this relationship is based on a fa t on erning the normal distribution fun tion,
namely dening

n(t;  ) = p1 2e
t2
2 2 ; (66)
Z x2


N (x; ) = n(t;  ) dt; (67)
1
one an verify that

2 N (x;  ) =  N (x;  ):
xx (68)

There are surely more relations one an prove, but the next theorem will give a
deeper insight into the relations of the Greeks.

Theorem 7 If the pri e and two Greeks g1 ; g2 of a European option are given with
g1 2 G1 = f
;
k ;
l ; ; q g; (69)

g2 2 G2 = f ; ; ; ; g;
k l (70)

then all the other Greeks (2 G1 [ G2 ) an be al ulated. Furthermore, if  and

another Greek from G2 is given, it is also possible, to determine all other Greeks.

10
Proof. The relations (53) to (60) are independent of ea h other. The relations
(61) to (63) are on lusions. To get an overview over all these relations, we list the
appearan e of ea h Greek in all these relations. With X or O we denote, that the
marked Greek appears in the relation. The relations marked with X show, that
there is a relation between Greeks of G1 and G2 and the O shows, that this relation
on erns only the Greeks of one set.

Greeks 2 G1 Greeks 2 G2
equation v

k
l  q k l  
(53) X X X X
(54) O O O
(55) O O
(56) O O
(57) X X X X
(58) O O O
(59) O O O
(60) X X X X
(61) X X X X
(62) X X X X
(64) O O
(65) O O
(63) O O
Let us now assume the option pri e and one Greek from the set G1 are given. Then
a look at the table shows that all Greeks of the set G1 an be evaluated. If all
Greeks of the set G1 are known and additionally one Greek of the set G2 is given,
all other Greeks an be determined. One the other hand, only eight equations are
independent, so the knowledge of two Greeks is also the minimum knowledge one
needs to determine all ten Greeks. This is the proof of the rst statement.

If  and another Greek from G2 is given, then it is always possible to determine

one Greek of the set G1 and one applies the part of this theorem already proved. If
; k or
l is given, one an use one of the Bla k-S holes equations (60) to (62). If
vega  is given, one an use (65) to get .

We on lude this se tion with an example. In the spe ial ase of plain vanilla alls
and puts in a foreign ex hange market all relations for the Greeks presented above
are valid. These formulas are well known and an be found in [7℄.

4.2 A Path-Independent Barrier Call

4.2.1 Value
The payo of a path-independent down-and-out barrier all is given by

f (ST ; k; K ) = (S k)+
IfST >K g (71)

11
We assume k<K - otherwise it would be a plain vanilla all - and therefore the
payo an be written as (ST k)IfST >K g. We laim that k and K are strikes, be ause
this option has the s aling behavior f (aST ; ak; aK ) = af (ST ; k; K ). Intuitively one
would all K a level; but we dened a level by its s aling behavior in se tion 2.3.2,
whi h is not valid in this ase. Therefore the path-independent barrier all is an
example for a strike-dened option.

Using the abbreviation

ln( SK0 ) + (p
d =
r q )  21 2 ; (72)
2
the value of a path-independent down-and-out barrier all is given by

!
Z1 s (r q ) + 12  2  )2
v (S0 ; k; K ) = e r p k2 exp (ln( S0 )
s
2 2 
ds
K s 2 
= S0 e N (d+) ke r N (d ):
q
(73)

We now want to al ulate all Greeks of this option. We show that Theorem 7 an
be used to organize the al ulation of the Greeks.

4.2.2 Greeks
Delta. Sin e dierentiation annot be avoided entirely, we hoose the derivative
with respe t to k, whi h is obviously
vk = e r N (d ): (74)

Next we dierentiate the integral representation of v with respe t to K and

obtain
0 1
K k (ln( SK0 ) (r q ) + 12  2  )2 A
vK = e r p 2 exp 
K 2  2 2 
k K 1
=
K
p 2 e r n(d ): (75)

In Theorem 7 we had assumed only one strike. In our example we have two
strikes, and therefore we need two Greeks from the set G1 to determine all
other Greeks of this set. >From the pri e homogeneity we know that the
relation

v = S0 vS0 + kvk + KvK (76)

holds, when e we obtain for the spot delta

vS0 = e q N (d+) + KS k p 12 e r n(d ): (77)

0 

12
Rho. We use relations (58) and (63) and obtain

K k
vr =  ke r N (d )+ p e r n(d ); (78)
2
K k
vq =  S0 e q N (d+)  p e r n(d ): (79)
2

Gamma. We have al ulated all Greeks in G1 . To determine some other Greeks

without dierentiation we need at least one Greek of the set G2 . In the theorem
above we assumed, that the option will be des ribed by one strike, but the
option we analyze depends on two strikes. So we have to dierentiate tri e to
get all dual gammas.

vkk = 0 (80)
1 1
vkK = p
K 2
e r n(d ) (81)

k e r k K e r
vKK = p
K 2 2
n(d ) +
K 2 2
n(d )d (82)

The extension of (55) to the ase of one sto k and two strikes is the equation
(27) with n = 1 and m = 2. In our example this relation is given by

S02 = k2 kk + 2kK kK + K2 KK : (83)

>From this relation, whi h follows from the homogeneity of v, we obtain for
the spot gamma without dierentiation

ke r k K e r
vS0 S0 = p n ( d ) +
S02

 2  n(d )d :
S02  2 
(84)

Vega. >From (65) we get

v =
p ke r n(d ) (K k )e r 1
n(d )d : (85)


Theta. >From the s ale invarian e of time (53) we obtain

vt = v = rke r N (d ) + qS0e q N (d+)

K k 
(r q ) p 2 e r n(d ) p ke r n(d )
 2 
1 r n(d
+ (K k )e )d (86)
2

13
5 A European Claim in the Two-Dimensional
Bla k-S holes Model
5.1 Pri ing of a European Option

Rainbow options are nan ial instruments whi h depend on several risky assets.
Many of them are very sensitive to hanges of orrelation. We all kappa ( ) the 
derivative of the option value v with respe t to the orrelation .
The omputational eort to ompute the kappa is hard, even in a simple framework,
but in the Bla k-S holes model with two sto ks and one ash bond we an use the
ross-gamma- orrelation-risk relationship whi h an be used easily to nd kappa.

Let the sto k pri e pro esses S1 and S2 be des ribed by

S ( ) 1 2
ln 1 = (r q1  ) + 1 W1 ;
2 1
(87)
S1 (0)
S ( ) 1 2 q
ln 2 = (r q2 2 ) + 2 W1 + 2 1 2 W2 : (88)
S2 (0) 2
W 1 and W 2 are two independent Brownian motions under the risk neutral measure.
The probability density for the distribution of S1 ( ) is denoted by h1 (x) and is given
by the log-normal density
!
1 1 A2
h1 (x) = q exp ;
212 
(89)
212  x
!

x 1
A = ln r + q1  + 12 : (90)
S1 (0) 2
The equation for the se ond sto k pri e pro ess an be written as
! !
S ( ) 1 2  S ( ) 1 2
ln 2 = (r q2 2 ) + 2 ln 1 (r q1  )
S2 (0) 2 1 S1 (0) 2 1
q
+2 1 2 W2 : (91)

The onditional distribution of S2 ( ) given S1 ( ) is thus log-normal with density

!
1 B2
h2j1 (y jx) = q exp ;
222 (1 2 )
(92)
y 222 (1 2 )
" ! #

y 1 2 2 
B = ln r + q2  + 2  A : (93)
S2 (0) 2 1

The joint distribution of S1 ( ) and S2 ( ) is given by the produ t of h1 and h2

h(x; y ) = h1 (x)
h2j1 (y jx): (94)

14
A European option with maturity  and payo f (S1 ( ); S2 ( )) will be pri ed by

Z1 Z1
v = e r h(x; y )
f (x; y )dxdy: (95)
0 0
This integral has exa tly the stru ture of the integrals studied in se tion 3.1. Using
the results provided above, one an olle t several relationships for the Greeks in the
two-dimensional ase. Additional, the fundamental symmetry s ale invarian e of
time is valid too. Be ause we on entrate on European options, the two dimensional
Bla k-S holes-PDE also holds.

5.2 Relations among the Greeks

We spe ialize the relationships among the Greeks found in n dimensions. Some
results are

0 = q1 + S1 (0)
1 ; (96)

0 = q2 + S2 (0)
2 ; (97)

1 1
0 = q1 q1 + q2 q2 + 1 1 + 2 2 + rr +  ; (98)
2 2
0 =  rv + (r q1 )S1 (0)
1 + (r q2 )S2 (0)
2
1 1
+ 12 S1 (0)2 11 + 1 2 S1 (0)S2 (0) 12 + 22 S2 (0)2 22 ; (99)
2 2
 = 1 2  S1 (0)S2 (0) 12 ; (100)

0 =  1 1 + 12  S1 (0)2 11 ; (101)

0 =  2 2 + 22  S2 (0)2 22 ; (102)

0 = 1 1 2 2 12  S1 (0)2 11 + 22  S2 (0)2 22 ; (103)

r =  (v S1 (0)
1 S2 (0)
2 ) ; (104)

0 =  v + q1 + q2 + r : (105)

Of ourse one an get more relations by ombining some relations above. The
relations we have hosen to present are either similar to the one-dimensional ase or
have another natural interpretation.

 (96) and (97). These relations are a justi ation for the rough way to deal
with dividends. One subtra ts the dividends from the a tual spot pri e and
pri es the option with this pri e and without dividends. This relation is not
ee ted by the two-dimensionality of the problem.

 (98). This is the two-dimensional version of the general invarian e under time
s aling.

 (99). This is the Bla k-S holes dierential equation. This relation must hold,
be ause we on entrated on European laims. It turns out, that the dynami

15
 of an option pri e is des ribed by the market model and that the pri e of the
option is dened as a boundary problem.

 (100). This is the ross-gamma- orrelation-risk relationship; it is remarkable,

that this relationship has su h a simple stru ture.

 (101) and (102). These are the gamma-vega relationships. Noti e that one an
determine  only by knowledge of some derivatives with respe t to parameters
whi h on ern only one sto k. Of ourse, there is no dieren e between the
rst and the se ond sto k. These relations are valid in the one-dimensional
ase with   0.
 (103) follows from (100).

 (104). This is the delta-rho relationship. The interest rate risk is well known
to be the negative produ t of duration and the amount of money invested.
The term in the parentheses is exa tly the amount of money one would have
to invest in the ash bond in order to delta-hedge the option.

 (105). This relation is the two-dimensional rates symmetry, an extension of

equation (59). It follows from (104), (96) and (97).

In the following we treat one example in full detail. Further examples su h as outside
barrier options and spread options are available in [7℄.

5.3 European Options on the Minimum/Maximum of Two

Assets

We onsider the payo

[ ( min(S1 (T ); S2 (T )) K )℄+ : (106)

This is a European put or all on the minimum (  = +1) or maximum ( = 1) of

the two assets S1 (T ) and S2 (T ) with strike K . As usual, the binary variable  takes
the value +1 for a all and 1 for a put. Its value fun tion has been published in [5℄
and an be written as

v (t; S1 (t); S2 (t); K; T; q1; q2 ; r; 1 ; 2 ; ; ;  ) (107)

h
=  S1 (t)e q1  N2 (d1 ; d3; 1 )
+S2 (t)e q2  N2 (d2 ; d4 ; 2)
!#
Ke r
1  p
+ N2 ( (d1 1  );  (d2
p
  ); ) ;
2 2

2 + 2
2 = 21 2 ;
1 2 (108)

16

2 1 ;
1 = (109)


1 2
2 = ; (110)


ln(S1 (t)=K ) + p(r q1 + 21 12 )
d1 = ; (111)
1 

ln(S2 (t)=K ) + (r q2 + 21 22 )
d2 = p
2 
; (112)

ln(S2 (t)=S1 (t)) +p(q1 q2 21  2 ) ;

d3 = (113)
 

ln(S1 (t)=S2 (t)) +p(q2 q1 21  2 ) :
d4 = (114)
 

5.3.1 Greeks
Delta. Spa e homogeneity implies that

v v v
v = S1 (t) + S2 (t) +K : (115)
S1 (t) S2 (t) K
Using this equation one only has to dierentiate twi e in order to get all deltas.
It turns out, that the value fun tion is given in the natural representation,
whi h is presented in the appendix, and one is allowed to read o the deltas:

v
= e q1  N2 (d1 ; d3 ; 1 ); (116)
S1 (t)
v
= e q2  N2 (d2 ; d4 ; 2 ); (117)
S2 (t)
 p 
v
e r
1 
N2((d1 p
K
=
2
+ 1 2
  );  (d 2
  ); ) :
(118)

Gamma. Computing the gammas is a tually the last situation where dierentiation
is needed. We use the identities
!

N (x; y; ) = n(x)N
x 2
p1 x2 ;
y
(119)

!
 x y
N (x; y; ) = n(y)N
y 2
p1 2 ; (120)

and obtain

17
 " !
2v e q1   d d
2 = p n(d1 )N  3p 1 12
 (S1 (t)) S1 (t)  1 2 1 
!#
 d1 d3 1
n(d )N  p ;
 3 2 1 2
(121)

" !
2v e q2   d4 d2 2
= p n(d )N  p
 (S2 (t))2 S2 (t)  2 2 1 1 2
!#
 d2 d4 2
n(d )N  p ;
 4 1 1 2
(122)

!
2v e q1  d1 d3 1
=
S1 (t)S2 (t) S2 (t) 
p n(d3)N   p1 2 : (123)
2
Kappa. The sensitivity with respe t to orrelation is dire tly related to the ross-
gamma

v 2v
= 1 2  S1 (t)S2 (t) : (124)
 S1 (t)S2 (t)

Vega. We refer to (101) and (102) to get the following formulas for the vegas,

v v + 12  (S1 (t))2 vS1 (t)S1 (t)

= (125)
1 1" !
= S1 (t)e q1  p 1 n(d3 )N
d d
 1p 3 12
2 1 
!#
d d
+ n(d1 )N  3p 1 12 ; (126)
2 1 
v v + 2  (S2 (t))2 vS2 (t)S2 (t)
2
= (127)
2 2" !
p d d
= S2 (t)e q2 
 2 n(d4 )N  2p 4 22
1 1 
!#
d d
+ n(d2 )N  4p 2 22 : (128)
1 1 

Rho. Looking at (96), (97) and (104) the rhos are given by

v v
= S1 (t) ; (129)
q1 S1 (t)
v v
= S2 (t) ; (130)
q2 S2 (t)
v v
= K : (131)
r K

18
Theta. Among the various ways to ompute theta one may use the one based
on (98).

v 1 1 2 
= q1 vq1 + q2 vq2 + rvr + v1 + v2 : (132)
t  2 2

6 Generalization to Higher Dimensions

and other Market Models
6.1 Beyond Bla k-S holes

Up to now we illustrated our ideas in the Bla k-S holes model and in some parts
we used spe i properties of this model. Nevertheless there are some properties,
whi h are so fundamental, that they should hold in any realisti market model.
These fundamental properties are the homogeneity of time, the s ale invarian e of
time and the s ale invarian e of pri es. For every market model one uses, one should
he k, if the model fullls these properties.

An example for a market model with non-deterministi volatility is Heston's sto h-

asti volatility model [2℄.

In this more general framework one needs to larify the notion of vega. A hange
of volatility ould mean a hange of the entire underlying volatility pro ess. If the
pri ing formula depends on input parameters su h as initial volatility, volatility of
volatility, mean reversion of volatility, then one an onsider derivatives with respe t
to su h parameters. It turns out that our strategy to ompute Greeks an still be
applied su essfully in a sto hasti volatility model.

6.2 Heston's Sto hasti Volatility Model

 q 
dSt = St  dt + v (t)dWt(1) ; (133)
q
dvt = ( vt ) dt +  v (t)dWt(2) ; (134)
h i
Cov dWt(1) ; dWt(2) =  dt; (135)

(S; v; t) = v: (136)

The model for the varian e vt is the same as the one used by Cox, Ingersoll and
Ross for the short term interest rate, see [1℄. We think of  > 0 as the long term
varian e, of >0 as the rate of mean-reversion. The quantity (S; v; t) is alled
the market pri e of volatility risk.

Heston provides a losed-form solution for European vanilla options paying

[ (ST K )℄+ : (137)

19
As usual, the binary variable  takes the value +1 for a all and 1 for a put, K
the strike in units of the domesti urren y, q the risk free rate of asset S, r the
domesti risk free rate and T the expiration time in years.

6.2.1 Abbreviations


a = (138)

1
u1 = (139)
2

1
u2 = (140)
2
b1

=  +   (141)

b2

= + (142)
q


dj = ('i bj )2  2 (2uj 'i '2 ) (143)

bj 'i + dj
g = j (144)
bj 'i dj

T
 = t (145)
" #

bj 'i + dj 1 edj 
Dj (; ') =
2
(146)
1 gj edj 

(r
Cj (; ') = q )'i
( " #)
a 1 gj edj 
+ 2 (bj 'i + d) 2 ln (147)
 1 edj 
fj (x; v; t; ') =
eCj (;')+Dj (;')v+i'x (148)
" #

1 1 Z 1 e i'y fj (x; v; ; ')
Pj (x; v; ; y ) = +
2  0
< i'
d' (149)

1 Z 1 h i'y i
pj (x; v; ; y ) =
 0
< e fj (x; v; ; ') d' (150)

1 
P+ () = + P1 (ln St ; vt ; ; ln K ) (151)
2

1
P () =

+ P2 (ln St ; vt ; ; ln K ) (152)
2
This notation is motivated by the fa t that the numbers Pj are the umulative
distribution fun tions (in the variable y) of the log-spot pri e after time  starting
at x for some drift . The numbers pj are the respe tive densities.

6.2.2 Value
The value fun tion for European vanilla options is given by
h i
V = e q S r P
t P+ () Ke () (153)

20
The value fun tion takes the form of the Bla k-S holes formula for vanilla options.
The probabilities P () orrespond to N (d) in the onstant volatility ase.

6.2.3 Greeks
We use the homogeneity of pri es, to obtain the deltas. But we must show, that the
pri e is given in its natural representation. So we use the following strategy.

We assume, that equation (153) gives the natural pri e representation, whi h is
dened in appendix A. Under this assumption we an read o the deltas, and from
the deltas we derive the gammas. Using Theorem 8 we show that the assumption
of (153) giving the natural pri e representation was orre t.

Spot delta.

V q P

= = e + () (154)
St

Dual delta.

V = e

K = r P () (155)
K
Gamma. Under the ondition, that the deltas are orre t, we obtain for the gam-
mas by dierentiation:

Spot Gamma.


= 
x = e
=
q
p (ln St ; vt ; ; ln K )
St 1
(156)
St x St

Dual Gamma.
K

K = 
K y = e
=
r
p (ln St ; vt ; ; ln K )
K 1
(157)
K y K

Proof of the natural representation assumption >From Theorem 8 we know,

that our initial guess for the deltas is orre t, if the relation

St2 = K2 K (158)

holds. In fa t, this equation is given by

q p r p
St e 1 (ln St ; vt ; ; ln K ) = Ke 2 (ln St ; vt ; ; ln K ); (159)

and this statement is true. So our al ulation for the deltas and gammas has
been nished.

21
Rho. Rho is onne ted to delta via equations (64) and (63).

V r  P
= Ke (); (160)
r
V q  P
= St e + (): (161)
q
Theta. Theta an be omputed using the partial dierential equation for the Heston
vanilla option

1 1
Vt + (r q )SVS + vVvv + vS 2 VSS + vSVvS qV
2 2
+[( v ) ℄Vv = 0; (162)

where the derivatives with respe t to initial varian e v must be evaluated

numeri ally.

7 Summary
We have learned how to employ homogeneity-based methods to ompute analyti al
formulas of Greeks for analyti ally known value fun tions of options in a one-and
higher-dimensional market. Restri ting the view to the Bla k-S holes model there
are numerous further relations between various Greeks whi h are of fundamental
interest. The method helps saving omputation time for the mathemati ian who has
to dierentiate ompli ated formulas as well as for the omputer, be ause analyti al
results for Greeks are usually faster to evaluate than nite dieren es involving at
least twi e the omputation of the option's value. Knowing how the Greeks are
related among ea h other an speed up nite-dieren e-, tree-, or Monte Carlo-
based omputation of Greeks or lead at least to a quality he k. Many of the
results are valid beyond the Bla k-S holes model. Most remarkably some relations
of the Greeks are based on properties of the normal distribution refreshing the a tive
interplay between mathemati s and nan ial markets.

A The Natural Pri e Representation for Homoge-

neous Options
We analyze the following problem. Let v (x; k) be the value of an option and v (x; k)
is homogeneous of degree 1. After Evaluating the integral to determine the option-
pri e, one obtains the following formula:

v (x; k) = xf (x; k) + kg (x; k) (163)

One the other hand, we know from the homogeneity of v:

v (x; k) = xvx (x; k) + kvk (x; k) (164)

22
So the question is: Can we on lude f (x; k) = vx (x; k) and g (x; k) = vy (x; k)? The
answer is: No, not in general.
Be ause of v f (x; k),g (x; k),vx (x; k)
being homogeneous of degree 1 we know, that
and vk (x; k ) are homogeneous of degree 0. Therefore we know that f (x; k ) has the
x
representation f ( ) and so on. Introdu ing the notation u =
x we nd from (163)
k k
and (164) that

uf (u) + g (u) = uvx (u) + vy (u) (165)

We dene h(u) = vx (u) f (u). The answer to the question above would be yes, if
and only if h(u) = 0 for all u. One an easily show, that the fun tion h(u) has the
following properties:

lim h(u) = 0
u!0
(166)

lim h(u) = 0
u!1
(167)

h(u) = uf 0 (u) + g 0 (u) (168)

So we ome to the following denition:

Denition 2 (Natural Representation of Homogeneous Fun tions)

Let v (x; k) be a homogeneous fun tion of degree 1. Then there is a unique represen-
tation
   
x x
v (x; k) = xf + kg with (169)
k k
0 0
0 = uf (u) + g (u) (170)

We all this the natural representation.

Of ourse, the denition of the natural representation an be extended to higher
dimensions. The question of this se tion an now be answered more exa tly. One
an read o the deltas if and only if the pri e formula is given in its natural repre-
sentation. This statement also holds in higher dimensions. For the two dimensional
ase, we summarize:

Theorem 8 Let v (x; k) be a homogeneous fun tion of degree 1. The representation

v (x; k) = xf (x; k) + kg (x; k) (171)

is the natural representation if and only if

x2 x f (x; k) = k2 k g (x; k) (172)

Theorem 9 Let v (x; k) = xf (x; k) + kg (x; k) be the natural representation of a

homogeneous fun tion of degree 1. Then the following equations hold:
x v (x; k) = f (x; k) (173)

k v (x; k) = g (x; k) (174)

23
Referen es
[1℄ COX, J.C., INGERSOLL, J.E. and ROSS, S.A. (1985). A Theory of the Term
Stru ture of Interest Rates. E onometri a 53, 385-407.
[2℄ HESTON, S. (1993). A Closed-Form Solution for Options with Sto hasti
Volatility with Appli ations to Bond and Curren y Options. The Review of
Finan ial Studies, Vol. 6, No. 2.
[3℄ PLACKETT, R. L. (1954). A Redu tion Formula for Normal Multivariate In-
tegrals. Biometrika. 41, pp. 351-360.
[4℄ SHAW, W. (1998). Modelling Finan ial Derivatives with Mathemati a. Cam-
bridge University Press.

[5℄ STULZ, R. (1982). Options on the Minimum or Maximum of Two Assets. Jour-
nal of Finan ial E onomi s. 10, pp. 161-185.
[6℄ TALEB, N. (1996). Dynami Hedging. Wiley, New York.
[7℄ WYSTUP, U. (1999). Vanilla Options. Formula Catalogue of
http://www.MathFinan e.de.

24