Option pricing l Cutting edge

Smile dynamics II
In an article published in Risk in September 2004, Lorenzo Bergomi highlighted how traditional
stochastic volatility and jump/Lévy models impose structural constraints on the relationship
between the forward skew, the spot/volatility correlation and the term structure of the volatility
of volatility. Here, he proposes a model that enables them to be controlled separately and also
prices options on realised variance consistently. He presents pricing examples for a reverse
cliquet, a Napoleon, an accumulator and an option on variance
VTt 1+,Tdt2 – VTt 1,T2 . Let us buy (T2 – t)/(T2 – T1)er(T2 – t) VS of maturity T2 and sell

A
common feature of the recent breed of exotic options such as
Napoleons and reverse cliquets is that their price depends on as- (T1 – t)/(T2 – T1)er(T1 – t) VS of maturity T1. This is done at no cost; our prof-
sumptions made for the joint dynamics of the underlying and its im- it and loss at time t′ = t + dt is:
plied volatilities. These fall into three categories:
T2 − t  Vtt ′ (t ′ − t ) + Vt ′ 2 (T2 − t ′ )  r T −t − r T −t
h T
■ The dynamics of implied volatilities, and more specifically the term struc- P&L=  − VtT2  e ( 2 )e ( 2 ′ )
ture of the volatility of volatility. T2 − T1  T2 − t 
■ The forward skew.
T1 − t  Vtt ′ (t ′ − t ) + Vt ′ 1 (T1 − t )  r T −t − r T −t
h T
■ The spot/volatility correlation. −  − VtT1  e ( 1 )e ( 1 ′ )
In a previous article (Bergomi, 2004), we analysed popular stochastic T2 − T1  T1 − t 
volatility and jump/Lévy models and pointed out that, although these
models produce prices that include an estimation of the three effects list- ( ) (
= VtT′ 1 ,T2 − VtT1 ,T2 e − r (t ′− t ) = VtT+1dt,T2 − VtT1 ,T2 (1 − rdt ) )
ed above, they impose structural constraints on how these features of the This position generates a profit and loss linear in VTt 1+,Tdt2 – VTt 1,T2 at low-
joint dynamics of the spot and implied volatilities are related. Another of est order in dt, at zero initial cost. Thus the pricing drift of any forward FV
their drawbacks is that they are based on a specification of the spot VTt 1,T2 is zero.2
process and they fail to take into account the fact that variance swaps We now specify a dynamics for the VS curve. Let us introduce
(VSs) should be considered as hedge instruments too, and be endowed ξTt = VtT, T, the value of the variance for date T, observed at time t.
with their own dynamics. ■ A one-factor model. We are free to specify any dynamics on the ξT(t)
This article is a natural continuation of our first one: here, we propose that complies with the requirement that ξT(t) be drift-less. However, for
a model that aims at pricing both standard exotic options and general op- practical pricing purposes, we would like to drive the dynamics of all of
tions on variance in a consistent manner, and lets us independently set the ξT(t) with a small number of factors. Here, we show how this can be
requirements for: done by carefully choosing the volatility function of ξT(t).
■ The dynamics of VS volatilities. Let us assume ξT(t) is lognormally distributed and that its volatility is a
■ The level of short-term forward skew. function of T – t so that the model is translationally invariant through time:
■ The correlation between the underlying and short and long VS volatilities. d ξT = ω (T − t ) ξT dU t
The article is organised as follows. First, we set up a general framework
for the dynamics of forward variance swap variances (FVs). Then we spec- where Ut is a Brownian motion. Let us choose the form ω(τ) = ωe–k1τ.
ify a dynamics for the underlying that is consistent with that of variances. ξT(t) can be written as:
In the next section, we specify a particular choice for the dynamics of for-
ward variances and the underlying. We then focus on practical features of ξT ( t ) = ξT ( 0) e
(ωe − k1(T −t )
X t − ω2 e−2 k1(T −t ) E  X t2 
2
) (1)
the model, such as the term structure of the volatility of volatility and the
term structure of the skew. Then a section focuses on pricing examples: where Xt is an Ornstein-Ühlenbeck process:
we consider a reverse cliquet, a Napoleon, an accumulator and a call on
X t = ∫0 e − k1 (t − u ) dU u
t
variance. The concluding section summarises our work.
whose dynamics reads:
Modelling variances dX t = − k1 X t dt + dU t
A VS pays at maturity VtTh – VT, where V h is the annualised variance of the
t tT
spot, realised over the interval [t, T] and VTt is the implied VS variance, ob- X0 = 0
served at time t for maturity T. Because VSs are statically replicable by ξT(t) is drift-less by construction. Knowing Xt, we can generate Xt + δ
vanilla options, VTt depends only on the implied volatilities seen at time t through:
for maturity T.1 Because of the definition of VTt, the VS contract has zero X t + δ = e − k1δ X t + xδ
value at inception.
Now consider the FV V Tt 1,T2 defined as: where xδ is a centred Gaussian random variable such that E[x 2δ] = (1 –
e–2k1δ)/(2k1).
VtT1 ,T2 =
(T2 − t )VtT − (T1 − t )VtT
2 1
Starting from known values for Xt and E[X2t] at time t we can generate
T2 − T1 the FV curve ξT(t + δ) at time t + δ by using the following relationship:
where T1, T2 > t.
To write a pricing equation for an option on V Tt 1,T2 we first need to 1 As well as on how dividends are modelled and assumptions on interest rate volatility
know the cost of entering a trade whose payout at time t + dt is linear in 2 The drift-less nature of forward VS variances had been noticed before (see Dupire, 1996)

WWW.RISK.NET ● OCTOBER 2005 RISK 67

Cutting edge l Option pricing

where we use the same recursions as above for Xt, Yt, E[X 2t], E[Y 2t] and
X t + δ = e − k1δ X t + xδ
E[XtYt].
1 − e −2 k1δ While this set-up for the dynamics of the ξi is reminiscent of the Libor
E  X t2+ δ  = e −2 k1δ E  X t2  +
2k1 market models used in fixed income, there are as yet no market quotes
for prices of caps/floors and swaptions on forward variances, on which to
and expression (1). calibrate volatilities and correlations for the ξi.
Thus, by choosing an exponentially decaying form for ω(τ) the model ■ An N-factor model. We may generally write:
becomes Markovian: all ξT(t) are functions of just one Gaussian factor Xt. ωi2t
ξ i ( t ) = ξ i ( 0) e
ωi Z ti − 2
■ A two-factor model. To achieve greater flexibility in the range of term
structures of volatilities of volatilities that can be generated, we prefer to where ωi and ρ(Zi, Zj) are chosen at will. Later in this article, we will com-
work with two factors. We then write: pare pricing results obtained in the two-factor model with those obtained
(
d ξT = ωξT e − k1 (T − t ) dU t + θe − k2 (T − t ) dWt ) in an N-factor model for which ωi = ω, a constant, and the correlation struc-
ture of the Zi is:
where Wt is a Brownian motion. Its correlation with Ut is ρ. We can run
through the same derivation as above. ξT(t) now reads: ( )
ρ Z i , Z j = θρ0 + (1 − θ) β j − i (4)

 (
ξT (t ) = ξT (0) exp ω e − k1 (T − t ) X t + θe − k2 (T − t )Yt 

where θ, ρ0, β ∈ [0, 1].
It should be noted that, when pricing an option of maturity T, in con-
− ω2 e −2 k1 (T − t ) E  X t2  + θ 2e −2 k2 (T − t ) E Yt 2 
2
trast with the two-factor model, the number of factors driving the dynam-
 (2)

)
ics of variances in the N-factor model is proportional to T, thus the pricing
+ 2θe ( 1 2 )( ) E [ X tYt ]
− k + k T −t
time will grow like T 2.

As in the one-factor case, if Xt, Yt, E[X 2t], E[Y 2t ] and E[XtYt] are known Specifying a joint dynamics for the spot
at time t, they can be generated at time t + δ through the following ■ A continuous setting. We could use the dynamics of instantaneous for-
relationships: ward variances specified in equation (2) and write the following lognor-
X t + δ = e − k1δ X t + xδ mal dynamics on the underlying:
Yt + δ = e − k2δYt + yδ dS = ( r − q ) Sdt + ξt (t )SdZ t
and: with correlations ρSX and ρSY between Z and, respectively U and W. This
1 − e −2 k1δ yields a stochastic volatility model whose differences with standard mod-
E  X t2+ δ  = e −2 k1δ E  X t2  + els are:
2k1
■ It has two factors.
1 − e −2 k2δ ■ It is calibrated by construction to the term structure of VS volatilities.
E Yt 2+ δ  = e −2 k2δ E Yt 2  +
2k2 In such a model, the level of forward skew is determined by ρSX, ρSY ,
ρ, ω, k1, k2 and θ with no way of controlling it separately, just like in stan-
1− e ( 1 2)
− k +k δ
E [ X t + δYt + δ ] = e ( 1 2 ) E [ X tYt ] + ρ
− k +k δ
dard stochastic volatility models.
k1 + k2 ■ A discrete setting. Here we achieve our objective of controlling the
where, in the right-hand terms, the second component is, respectively, the forward skew – or, in other words, the skewness of the spot process for
variance of xδ, the variance of yδ and the covariance of xδ and yδ. Starting time scale ∆ – by using the discrete tenor structure defined above and the
from time t = 0 we can easily generate a FV curve at any future time t by dynamics of forward variances given by expression (3).
simulating two Gaussian factors. We choose k1 > k2 and call Xt the short At time t = Ti the VS volatility σ
^ for maturity T + ∆ is known. It is
VS i
factor and Yt the long factor. given by:
■ A discrete structure. Instead of modelling the set of all instantaneous
VS = ξi (t = T )
σ i
forward variances, it may be useful to set up a tenor structure and model
the dynamics of forward variances for discrete time intervals, in a way that To be able to specify the spot process over the interval [Ti, Ti + ∆], we
is analogous to Libor market models. make two more assumptions:
In the fixed-income world, this is motivated by the fact that forward ■ that the spot process over the time interval [Ti, Ti + ∆] is homogeneous:
Libor rates are the actual underliers over which options are written. In our the distribution of STi + ∆/STi does not depend on STi. The reason for this
case, it is motivated by the fact that we want to control the skew for a requirement is that we want to decouple the short forward skew and the
given time scale. spot/volatility correlation. Imposing this condition makes the skew of ma-
Let us define a set of equally spaced dates Ti = t0 + i∆, starting from t0, turity ∆ independent on the spot level. Thus the prices of cliquets with pe-
today’s date. We will model the dynamics of FVs defined over intervals of riod ∆ will not depend on the level of spot/volatility correlation.
width ∆: define ξi(t) = V tt0 + i∆, t0 + (i + 1)∆ for t ≤ t0 + i∆. ξi(t) is the value at ■ that the at-the-money-forward (ATMF) skew (dσ ^ /d ln K)| for maturity
K F
time t of the FV for the interval [t0 + i∆, t0 + (i +1)∆]. Ti + ∆ be a deterministic function of σ ^ or σ
VS
^
ATMF. In this article, we im-
ξi(t) is a random process until t = t0 + i∆.When t reaches t0 + i∆, the VS pose that it is constant or proportional to σ ^
ATMF. Other specifications for
variance for time interval [t, t + ∆] is known and is equal to ξi(t = t0 + i∆). the dependence of the ATMF skew on σ ^ or σ
VS
^
ATMF are easily accommo-
We model the ξi in the same way as their continuous counterparts: dated in our framework.
There are many processes available for fulfilling these objectives – note
(
ξi (t ) = ξi (0) exp ω e

− k1 (Ti − t )
X t + θe
− k2 (Ti − t )
Yt 

that we also need to correlate the spot process with that of forward vari-
ances ξi for j > i. We could use a Lévy process, especially one that has an
− ω2 e 1 ( i ) E  X t2  + θ 2e 2 ( i ) E Yt 2 
2 −2 k T − t −2 k T − t expression in terms of a subordinated Brownian motion.3 Here we decide
 (3)
to use a constant elasticity of variance (CEV) form of local volatility. Over
+ 2θe ( 1 2 )( i ) E [ X tYt ]
− k + k T −t
 ) 3 For example, the variance gamma and normal inverse Gaussian processes

68 RISK OCTOBER 2005 ● WWW.RISK.NET

the interval [Ti, Ti + ∆] the dynamics of St reads: 1. Term structure of the volatilities of VS
 S 
1− β volatilities for a one-month interval
dS = ( rt − qt ) Sdt + σ 0   SdZ t (5)
 STi  140
_____ Two-factor
where σ0(σ ^ ), β(σ
VS
^ ) are functions of σ
VS
^ = √ξi (T ) calibrated so that the
VS i 120 N-factor

Volatility of VS volatility (%)

VS volatility of maturity Ti + ∆ is σ^ and the condition on the ATMF skew
VS
is fulfilled. rt and qt are, respectively, the interest rate and the repo, inclu- 100
sive of dividend yield. Note that instead of – or in addition to – control-
ling the skew we could have controlled the convexity of the smile near 80
the money. This would be relevant in the forex or fixed-income world. In
this article, we restrict our attention to the skew. 60
This completely specifies our model and the pricing algorithm. We can
40
write the corresponding pricing equation as:

( )
2
20
dP σ STi0 , ξ , S
i0
dP d 2P
+ ( rt − qt ) S + S2
dt dS 2 dS 2 0
0 12 24 36 48 60
( )
2 2
1 d P d P
+ ∑ ρij ωi ω j ξi ξ j i j + ∑ ρSi ωi σ STi0 , ξi0 , S S ξi = rP Maturity of VS volatility (months)
2 i , j >i0 dξ dξ i > i0 dSd ξi
where i0(t) is such that t ∈ [Ti0, Ti0 + ∆[, ωi is the volatility of the ξi and ρij
is their correlations. 2. Term structure of the volatility of VS
In the N-factor model, ωi = ω and ρij = ρ(Zi, Zj). In the two-factor model, volatilities for a one-year interval
the dynamics of the ξi is driven by the processes X and Y. The pricing
equation can then be written more economically as:
140
dP σ (, S ) 2 d 2 P
2 Two-factor
dP dP dP
+ ( rt − qt ) S − k1 X − k2Y + S 120 N-factor
dt dS dX dY 2 dS 2
Volatility of VS volatility (%)

1  d 2P d 2P d 2P   d 2P d 2P 
+  + + 2ρ  + σ (, S ) S  ρSX + ρSY = rP 100
2  dX 2
dY 2
dXdY   dSdX dSdY 
80
where ρSX and ρSY are, respectively, the correlation between Brownian mo-
tions Ut and Zt and the correlation between Wt and Zt. σ(···, S) is a short- 60
hand notation for:

( ) )
40
(
σ (, S ) ≡ σ STi , ξi0 X Ti , YTi , S
0 0 0
20

0
Pricing 0 12 24 36 48 60
We now turn to using the model for pricing, focusing on the two-factor Maturity of VS volatility (months)
model. In what follows, we choose as time scale ∆ = 1 month. By con-
struction, the model is calibrated at time t0 to the FV curve for all maturi-
ties t0 + i∆. We specify, in this order: the five-year volatility. Figure 1 displays the term structure of the volatili-
■ values for k1, k2, ω, ρ, θ ties of VS volatilities generated by the two-factor model with a flat initial
■ a value for the ATMF skew VS term structure at 20% volatility using these parameter values. We graph:
■ values for ρSX and ρSY.
  V ∆t , ∆t + τ  
These steps are discussed in the next three sections. 1
StDev ln  ∆t 
■ Setting a dynamics for implied VS volatilities. Our aim is to price ∆t   V ∆t , ∆t + τ  
options whose price is a very non-linear function of volatility; as we roll  0 
towards the option’s maturity, the maturity of the volatilities we are sensi- for a range of values of τ from one month to five years. We have picked
tive to shrinks as well. We thus need to be able to control the term struc- ∆t = 1 month. The value of ω is chosen so that, over the interval of ∆t = 1
ture of the volatilities of volatilities, be they ATMF or VS volatilities. In our month, the volatility of the one-month VS volatility is 120%.
framework, it is more natural to work with VS volatilities. In our model, We also display the term structure generated by the N-factor model
the dynamics of VS volatilities is controlled by k1, k2, ω, ρ and θ. As there using the following parameters:
is currently no active market for options on forward ATM or VS volatility,
σ = 240%, θ = 40%, ρ0 = 5%, β = 10%
these parameters cannot be calibrated on market prices. Thus their values
have to be chosen so that the level and term structure of volatility of volatil- These values are chosen so that, for ∆t = 1 month, the term structure of
ity are reasonably conservative when compared with historically observed the two-factor model is matched. Now let us measure volatilities over a
volatilities of implied volatilities.4 time interval of one year, instead of one month (see figure 2).
Here we choose the following values: They are very different. Although both models would yield similar prices
for options on VS variances observed one month from now, they would
ω = 2.827, ρ = 0, θ = 30%, k1 = 6 ( 2 months ) , k2 = 0.25 ( 4 years ) (6)
4Dealers trading Napoleons and reverse cliquets typically accumulate a negative gamma
so that the volatility of volatility for a one-month horizon is about 120% position on volatility. Ideally, bid and offer term structures of volatility of volatility will be
for the one-month VS volatility, 45% for the one-year volatility and 25% for used for pricing

WWW.RISK.NET ● OCTOBER 2005 RISK 69

Cutting edge l Option pricing

3. β and σ0 as a function of ^
σ VS in the case of turity, for the case of a flat term structure of VS volatilities, at order one in
a constant 5% skew both ω and the skew (dσ ^ /d ln K)| at time scale ∆, which we denote
K F
Skew∆. Given the skewness ST of the distribution of ln (ST/FT), the ATMF
skew is given at first order in ST by (Backus et al, 1997):
25 60
β ST
σ0 SkewT = (7)
50 6 T
20

40 where FT is the forward for maturity T.

15 Consider a maturity T = N∆ and let us calculate the second and third

σ0 (%)
30
moments of ln (ST /FT) = ΣNi= 1ri where returns ri are defined as:
β

10 S   S(i −1) ∆ 
20 ri = ln  i∆  − ln  
 F∆   F(i −1) ∆ 
5
10
While returns are not independent, they are uncorrelated. Thus, assuming
that ∆ is small, so that the drift term in E[ri] is negligible with respect to
0 0 the random term:
0 10 20 30 40 50 60
σ^ VS (%) N 
3
M 3T =  ∑ ri  = ∑ ri3 + 3∑ ri r j2
 i =1  i j >i

price differently options on VS variances observed in one year. In the two- Let us work at the lowest order in ∆. To derive an expression of the
factor model, volatilities of volatility will tend to decrease as the time scale third moment at order one in ω and S∆, we can use the following ap-
over which they are measured increases, due to the mean-reverting nature proximations:
of the driving processes. In the N-factor model, by contrast, they increase:
this is due to the fact that forward variances are lognormal. The term struc-
r j2 = ∆ξ j T j ( )
Ti +∆
ture would be unchanged if forward variances were normal. ri = ξi (Ti ) ∫ dZ t
Ti
■ Setting the short forward skew. We calibrate the dependence of σ0
and β to σ ^ , so that the one-month ATMF skew has a constant value –
VS Let us denote as ξ the constant value of the VS volatilities at time zero.
say, 5%. We use the 95–105% skew: We get, at order one in ω:



105% ≅ − 1 d σ K

95% − σ
σ
10 d ln K
F
M 3T = ∑ ri3 + 3∑ ∆
i j >i
ξi (Ti ) ∫T i
T +∆
i
dZ t ξ j T j( )
instead of the local derivative (dσ ^ )/(d ln K). This defines the functions T +∆
K ⌠ i 1 + ω⌠ T j e − k1 (T j − u )dU
σ0(σ^ ) and β(σ
VS
^ ). This calibration is easily done numerically; we can also
VS   ⌡0 u
use analytical approximations.5 = ∑ ri3 + 3∑ ∆ ( )
ξi Ti  dZ t ξ j (0)
If needed, individual calibration of σ0(σ ^ ) and β(σ^ ) can be performed i j >i

+ θω∫ j e
T (
− k2 T j − u )dW 
VS VS  
u
for each interval [Ti, Ti + ∆]. Typically the same calibration will be used ⌡Ti 0

for all intervals except the first one, for which a specific calibration is per- = ∑ ri3
formed so as to match the short vanilla skew. Here we use the same cal- i
ibration for all intervals. Figure 3 shows functions σ0(σ ^ ) and β(σ ^ ) for Ti +∆
⌠ j  − k (T − u ) − k (T − u )
T
VS VS ⌠ 
the case of a constant 95–105% skew equal to 5%. + 3∑ ∆ωξ j (0) ξi (0) 
 dZ t   e 1 j dU u + θe 2 j dWu 
j >i ⌡Ti ⌡0  
The level of 95–105% skew can either be selected by the trader or cho-
sen so that the market prices of call spread cliquets of period ∆ (here one
+ ρωξ ∆ N ρSX ζ ( k1∆ , N ) + θρSY ζ ( k2 ∆ , N ) 
3
=∑ ri3 2 2 2
month) are matched. i
■ Setting correlations between the spot and short/long factors – the
term skew. ρSX and ρSY cannot be chosen independently, since X and Y where ζ(x, N) is defined by:
have correlation ρ. We use the following parametrisation:
 1 − e − x  ∑ τ =1 ( N − τ ) e ( )
N −1
− τ −1 x
ζ ( x, N ) =  
ρSY = ρSX ρ + χ 1 − ρ2SX 1 − ρ2  x  N2

with χ ∈ [–1, 1]. ρSX and ρSY control both the correlation between spot Since we have set the short skew to a value that is independent on
and short and long VS volatilities, and the term structure of the skew of the level of variance, expression (7) shows that the skewness of ri is
vanilla options. They can be chosen, calibrated to the market prices of call constant. Thus:
spread cliquets of a period larger than ∆, or calibrated to the vanilla skew
( )
3

ri3 = S∆ ∑ ∆ξi (Ti ) = N S∆ ( ∆ξ) 2

3
∑
2
for the maturity of the option considered. The dependence of the term
i i
skew on ρSX and ρSY is made explicit in the following section.
In the N-factor model, we need to specify correlations between the spot where S∆ is the skewness at time scale ∆. We then get:
process and all forward variances, in a manner that is consistent with cor-
(
M 3T = N S∆ (ξ∆ ) 2 + ρω ∆ (ξ∆ ) 2 N 2 ρSX ζ ( k1∆ , N ) + θρSY ζ ( k2 ∆ , N ) )
3 3
relations of variances, a non-trivial task that we leave outside the scope of
this article.
■ The term skew. To shed light on how our model generates skew, we
derive an approximate expression for the ATMF skew as a function of ma- 5 See, for example, Zhou (2003)

70 RISK OCTOBER 2005 ● WWW.RISK.NET

At order zero in S∆ and ω: 4. The 95–105% skew as a function
2 of maturity
 n 
M 2T =  ∑ ri  = N ξ∆
 i =1 
6
hence the following expression for ST = M T3/(M T2) 2:
3 Actual
Equation (8)
5
S
ST = ∆ + N ω ∆ ρSX ζ ( k1∆ , N ) + θρSY ζ ( k2 ∆ , N ) 

95–105% skew (%)

N 4

Using equation (7) again, we finally get the expression of SkewN∆ at 3

order one in Skew∆ and ω:
2
Skew∆ ω
SkewN ∆ = + ρSX ζ ( k1∆ , N ) + θρSY ζ ( k2 ∆ , N )  (8)
N 2
1
This expression is instructive as it makes apparent how much of the
skew of maturity T is contributed on the one hand by the intrinsic skew- 0
0 12 24 36 48 60
ness of the spot process at time scale ∆ and on the other hand by the
Maturity (months)
spot/volatility correlation.
When ω = 0, the skew decays as 1/T, as expected for a process of in-
dependent increments. The fact that volatility is stochastic and correlated
with the spot alters this behaviour. Inspection of the definition of function 5. The two contributions to the 95–105% skew
ζ in equation (8) shows that for N∆ >> 1/k1, 1/k2, ζ(x, N) ∝ 1/N, so that in equation (8)
SkewN∆ ∝ 1/N, again what we would expect.
Equation (8) also shows how ρSX and ρSY can naturally be used to con-
trol the term structure of the skew. 5
Spot/volatility correlation
Contribution to 95–105% skew (%)

Figure 4 shows how the approximate skew in equation (8) compares

Intrinsic
with the actual skew. We have chosen the following values: ∆ =1 month,
4
the one-month 95%/105% skew is 5%, ω = 2.827, ρ = 0, θ = 30%, k1 = 6,
and k2 = 0.25. The spot/volatility correlations are: ρSX = –70%, ρSY = –35.7%
(χ = –50%). Even though ω and Skew∆ are both large, the agreement is 3
very satisfactory.
The two contributions to SkewN∆ in equation (8) are illustrated in figure
5. ‘Intrinsic’ denotes the first piece and ‘spot/vol correlation’ denotes the sec- 2
ond piece in equation (8). While the contribution of Skew∆ to SkewN∆ is mo-
notonically decreasing, the contribution of the spot/volatility correlation is 1
not, as it starts from zero at short time scales. Depending on the relative mag-
nitude of both terms, the term structure of the skew can be non-monotonic.
We have derived expression (8) for the case of a flat VS term structure 0
but the general case poses no particular difficulty. 0 12 24 36 48 60
Maturity (months)
Pricing examples
Here, we use our model to price a reverse cliquet, a Napoleon, an accu-
mulator and a call on realised variance, and analyse the relative contribu- A. Model prices
tion of forward skew, volatility-of-volatility and spot/volatility correlation
effects to prices. We use zero interest rates and dividend yield. Model Reverse cliquet Napoleon Accumulator
For the sake of comparing prices, we need to specify how we calibrate Black-Scholes 0.25% 2.10% 1.90%
model parameters. While it is natural to calibrate to the vanilla smile when With forward skew 0.56% 2.13% 4.32%
pricing options that can be reasonably hedged with a static position in With volatility of volatility 2.92% 4.71% 1.90%
vanilla options, it is more natural to calibrate to call spread cliquets and Full 3.81% 4.45% 5.06%
ATM cliquets when pricing Napoleons and reverse cliquets, which have a
high sensitivity to forward volatility and skew.
These products are also very sensitive to volatility of volatility. They are B. Model prices
usually designed so that their price at inception is small but increases sig-
nificantly if implied volatility decreases.6 As there is as yet no active mar- Model Reverse cliquet Napoleon Accumulator
ket for options on variance, we use the volatility-of-volatility parameter Full – correlations halved 3.10% 4.01% 5.04%
values listed in (6). Full – proportional skew 3.05% 4.30% 4.15%
Unless forward skew is turned off, the constant 95–105% one-month
skew is calibrated so that the price of a three-year 95–105% one-month
call spread cliquet has a constant value, equal to its price when volatility the implied volatility of the three-year one-month ATM cliquet is 20%.
of volatility is turned off and the one-month 95–105% skew is 5%, which The values for ρSX and ρSY are ρSX = –70%, ρSY = –35.7% (χ = –50%).
is equal to 191.6%.
In all cases the level of the flat VS volatility has been calibrated so that 6 See figure 1 in Bergomi (2004)

WWW.RISK.NET ● OCTOBER 2005 RISK 71

Cutting edge l Option pricing

6. Implied volatility of a call option on realised volatility (Bergomi, 2004).

variance as a function of maturity in the two- To understand why forward skew increases the price, consider first that,
factor model in the four cases listed above, E[ΣNi= 1 r i–] is constant, by calibration on the
ATM cliquet. Next consider the last period of the reverse cliquet. The final
payout is a function of the final return; it is a call spread whose low and
200 high strikes are, respectively, –C + Σi < N r i– – if it is negative – and zero.
95–105% one-month skew = 5%
180 When forward skew is turned on, the implied volatility of the ATM strike
No forward skew
is unchanged, by calibration, while the implied volatility for lower strikes
160
increases, making the call spread more expensive. The same argument
140 holds for returns prior to the last one.
120 ■ Napoleon. The maturity is still three years and the option pays at the
end of each year a coupon given by:
%

100
 
80 max  0, C + min ri 
 i 
60
40 where ri are the 12 monthly returns observed each year. Here we use
C = 8%.
20
Again, we notice that volatility of volatility accounts for most of the
0 price. Forward skew seems to have no sizeable impact, though this is not
0 250 500 generic; its magnitude and sign depend on the coupon size. While the pay-
Maturity (days)
out is still a call spread as a function of the final return, both strikes lie
below the money. Also, in contrast to the case of the reverse cliquet,
E[min
i ri] is not constant in the four cases considered.
7. Implied volatility of a call option on realised ■ Accumulator. The maturity is again three years with one final payout,
variance as a function of maturity in the two- given as a function of the 36 monthly returns ri by:
factor and N-factor models
 N 
( )
max  0, ∑ max min ( ri , cap ) , floor 
 i =1 
120
Two-factor model where floor = –1% and cap = 1% – a standard product.
N-factor model
100 The largest contribution comes from forward skew. Notice that switch-
ing on the volatility of volatility in the case when there is no forward
80 skew has no material impact on the price while it does when forward
skew is switched on. To understand this, observe that, in Black-Scholes,
when both strikes are priced with the same volatility, a 99–101% one-
%

60
month call spread has negligible vega. However, when the call spread
is priced with a downward sloping skew, it acquires positive convexity
40
with respect to volatility shifts.
■ Effect of spot/volatility correlation – decoupling of the short
20 forward skew. In standard stochastic volatility models, changing the
spot/volatility correlation changes the forward skew and thus the price
0 of cliquets. In this model, because of the specification chosen for the
0 250 500 spot dynamics in equation (5), changing the spot/volatility correlation
Maturity (days) does not change the value of one-month cliquets. It only alters the
term skew.
Prices quoted above have been calculated using ρSX = –70% and ρSY =
The corresponding term skew is that of figure 4. –35.7%. Figure 4 shows that, with these values the three-year 95–105%
In addition to the Black-Scholes price, we calculate three other prices skew is 1.25%.
95% – σ 105% ≠ 0, ω Let us now halve the spot/volatility correlation: ρSX = –35% and ρSY
by switching on either the one-month forward skew (σ ^ ^
^
= 0) or the volatility of volatility (σ 95% – σ
^
105% = 0, ω ≠ 0) or both (full). = –18% (χ = –19.2%). The three-year 95–105% skew is now 0.75% – al-
These prices are listed in table A. We give the definition of each product most halved. The implied volatility of the three-year cliquet of one-month
and comment on pricing results in the following paragraphs. ATM calls remains 20% and the price of a 95–105% one-month call spread
■ Reverse cliquet. Here we consider a globally floored locally capped cliquet is unchanged, at 191.6%. The new prices appear on the first line
cliquet, which pays once at maturity: of table B. The difference with prices on the fourth line of table A mea-
sures the impact of the term skew, all else – in particular cliquet prices
 N  – being kept constant. The fact that prices decrease when the spot/volatil-
max  0, C + ∑ ri− 
 i =1  ity correlation is less negative is in line with the shape of the Black-Sc-
holes vega as a function of the spot value.7
The maturity is three years, returns ri are observed on a monthly basis ■ Making other assumptions on the short skew. Here, we want to
(N = 36), r –i = min(ri, 0) and the value of the coupon is C = 50%. highlight how a different model for the short skew alters prices, using the
Notice that corrections to the Black-Scholes price are by no means three payout examples studied above. We now calibrate functions σ0(σ ^ )
VS
small, the contribution of volatility of volatility being the largest. The and β(σ ^ ) so that, instead of being constant, the 95–105% skew for ma-
VS
fact that volatility of volatility makes the reverse cliquet more expensive
is expected: this option, as well as the Napoleon, is in essence a put on 7 See figure 1 in Bergomi (2004)

72 RISK OCTOBER 2005 ● WWW.RISK.NET

 l Cutting edge
turity ∆ is proportional to the ATMF volatility for maturity ∆.
We have calibrated the proportionality coefficient so that the three- REFERENCES
year cliquet of one-month 95–105% call spreads has the same value as
Backus D, S Foresi, K Li and L Wu, 1997
before. The flat VS volatility is still chosen so that the implied volatility Accounting for biases in Black-Scholes
of the three-year cliquet of a one-month ATM call is 20%. Prices are list- Unpublished
ed on the second line of table B. Bergomi L, 2004
The accumulator is now sizeably cheaper. One can check that, in Black- Smile dynamics
Scholes, the value of a symmetrical call spread as a function of ATM volatil- Risk September, pages 117–123

ity, when the skew is kept proportional to the ATM volatility, is almost a Dupire B, 1996
A unified theory of volatility
linear function of volatility – in contrast with the case when volatilities are Unpublished
shifted in parallel fashion, where it is a convex function of volatility – thus
Leland H, 1985
suggesting why volatility of volatility has much less impact than in the con- Option replication with transaction costs
stant skew case. Journal of Finance 40(5), pages 1,283–1,301
■ Option on realised variance. Here we consider a call option that pays Zhou F, 2003
at maturity: Black smirks

)
Risk January, pages 87–91
max σ 2h − σ
1
K ,0
2

K
2σ 

where volatility σ ^ is the strike and σ2 is the annualised variance measured

K h volatilities of forward variances in the N-factor model are larger than in
using daily log returns. We assume there are 250 daily observations in a the two-factor model.
year, equally spaced. The variance of the distribution of σ2h has two sources: Finally, in addition to the effects discussed above, prices of calls on
■ The dynamics of VS variances. variance have to be adjusted to take into account bid/offer spreads on the
■ The fact that observations are discrete. In the case where VS variances VS hedge. These can be approximately included by shifting the level of
are static, it is the only contribution and it is determined by the distribu- volatility of volatility (Leland, 1985).
tion of spot returns, in particular its kurtosis, which depends on assump-
tions made for the short-maturity smile – in our context, the value of β. In Conclusion
the general case, it affects short-maturity options most noticeably. We have proposed a new model that, in contrast to popular stochastic
Prices are expressed as implied volatilities calculated with the Black- volatility and jump/Lévy models, gives us the flexibility to independently
Scholes formula with zero rate and repo. The underlying is the VS vari- control:
ance for the maturity of the option, whose initial value is given by the VS ■ The term structure of the volatility of volatility.
term structure observed at the pricing date. ■ The short-term skew.
In our model, daily returns are generated by the stochastic local volatil- ■ The correlation of spot and volatilities.
ity function form in equation (5). Their conditional kurtosis is a function This model consistently prices general exotic options and options on
of β, a parameter we use to control the short-term skew. The prices of op- variance or volatility. We achieve this by choosing a time scale ∆, specify-
tions on variance will thus depend on assumptions we make for the skew ing how the forward skew for maturity ∆ depends on the level of volatil-
at time scale ∆. Figure 6 shows implied volatilities of call options on vari- ity, and modelling the dynamics of VS variances. The model can be
ance, using a flat term structure of VS volatilities at 20%, the same corre- simultaneously calibrated on the short vanilla skew, the long vanilla skew,
lations as in the examples above (ρSX = –70%, ρSY = –35.7%), for the two an ATM cliquet and a call spread cliquet of period ∆, while letting us spec-
cases σ^
95% – σ105% = 5% and σ 95% – σ105% = 0%.
^ ^ ^ ify freely the dynamics of forward variances as well as how the short for-
Figure 6 illustrates how assumptions for the forward skew affect the ward skew depends on volatility.
distribution of returns significantly and thus the price of options on vari- By directly controlling the short forward skew we are able to adjust
ance, mostly for short options. The shortest maturity in the graph corre- the amount of skewness of the distribution of ln(ST) generated on the
sponds to options with a maturity of one month (20 days). Since we have one hand by the intrinsic skewness of the process at short time scales,
taken ∆ =1 month, the distribution of σ2h does not depend on the dynam- and the spot/volatility correlation on the other hand. Handling this issue
ics of variances ξi – it only depends on β. appropriately is, in our view, an essential task in the design of models
Note that, in our model, VS volatilities for maturities shorter than ∆ are that accurately capture the three effects mentioned above.
not frozen. Instead of being driven by equation (3), their dynamics is set Even though the choice of the time scale ∆ is natural for many pay-
by the value of β. outs – for example, Napoleons and reverse cliquets – it is more arbitrary
■ Using the N-factor model. It is instructive to compare prices of op- for other options, for example, options on variance. It would be useful
tions on realised variance generated by the two-factor and N-factor mod- to have scaling relationships relating parameter sets for different values
els. As figures 1 and 2 illustrate, even though the dynamics of VS of ∆ so that some model features remain unchanged, for example, the
volatilities in both models are similar for the short term, they become skew of vanillas or the implied forward skew of cliquets. This is left for
different for longer horizons. future work.
Here we price the same option on variance considered above using the As of today there are no market prices for caps/floors/swaptions on
N-factor model of forward variances. Parameter values for the dynamics forward variances. Choosing a value for the parameters governing the
of forward variances are the same as those used in figures 1 and 2. We dynamics of forward variances is thus a trading decision. It is the hope
95% – σ105% = 0). Also, to make prices com-
have taken no forward skew (σ ^ ^ of the author that the liquidity of options on volatility and variance in-
parable with those obtained in the two-factor model, we have taken zero creases so that we will soon be able to trade the smile of the volatility
correlation between spot and forward variances. Implied volatilities for of volatility! ■
both models are shown in figure 7.
Because ∆ = 1 month, for the shortest maturity considered – 20 days Lorenzo Bergomi is head of quantitative research in the equity
– the implied volatilities for both models coincide. For longer maturi- derivatives department at Société Générale. He wishes to thank
ties, the fact that implied volatilities are higher in the N-factor model is members of his team for helpful suggestions. Email:
in agreement with figure 2, which shows that, for longer horizons, the [email protected]

WWW.RISK.NET ● OCTOBER 2005 RISK 73