Consistent Pricing of Equity and Volatility Derivatives

Ser-Huang Poon (Manchester Business School & RMI)

Joint work with He Xue Fei (RMI), Simon Acomb (MBS)

September 4, 2009

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 1 / 38

Who Trades Volatility?

Traders of Derivative Contracts

Hedging vega exposure
Hedging correlation exposure using dispersion trading
Asset Managers
Volatility has become a new asset class
Low correlation with other assets to produce portfolio diversi…cation
Proprietary Traders
Hedge funds and trading desks taking position in volatility
Proprietary traders of dispersion
As a hedge against credit due to the negative relationship between
credit and equity volatility
People with Structure Exposure to Volatility
Statistical arbitrage
Institutions where P&L is structurally exposed to volatility
(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 2 / 38
Equity and Volatility Derivatives

De…nition; any products or contracts that are sensitive to volatility

speci…cation

S&P 500, SPX and VIX products as examples

Demand for OTC contracts on other indices and individual stocks is

strong

Like other OTC contracts (e.g. interest rate derivatives), likely to

remain as OTC and grow into trillion if it is not already there

Many such contracts have existed for a long time.

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 3 / 38

Equity or Index Level Contracts

Index linked notes

SPX call and put options (CBOE), OTC warrants

Exotic

Digital Options
Knock-in, Knock-out, Barrier
Asian
CPPI (Constant proportion portfolio insurance)

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 4 / 38

Volatility Level Contracts

Volatility swaps

Volatility Index linked

Volatility futures
Volatility options

Exotic

Cliquet: Forward Starting Options

Napoleon

Use Variance and Volatility Interchangeably

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 5 / 38

Why a consistent pricing model is important?

A typical “Spot” model for equity and volatility

dS = µS (S, t ) dt + σS (S, t ) dWS

dV = µV (V , t ) dt + σV (V , t ) dWV
V = f (S, σV )
hdS, dV i = ρ σS σV dt

“Spot” vs. “Forward” as in the interest rate derivatives literature

Omit jumps in this entire study; if necessary use change (of business)
time

Assume all dynamics are under risk neutral measure

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 6 / 38

Volatility Consistent Pricing Model

dS = µS (S, t ) dt + σS (S, t ) dWS (1)

dV = µV (V , t ) dt + σV (V , t ) dWV (2)

Typically, (1) and (2) are used to price equity or index level contracts
eventhough volatility contracts (e.g. VIX options) suggests a
dynamics very di¤erent from (2).

Volatility level contracts are priced using (2) only without considering
the implications on (1).

Inconsistency is pervasive and yet volatility level products are hedged

or replicated using equity level products.

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 7 / 38

Example: S&P 500, SPX and VIX
They all share the common root

S&P 500 (Physical vs. Risk Neutral Measure)

#
SPX call and put options
#
Variance Swap and VIX
#
Volatility futures and options

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 8 / 38

Variance Swap

Forward contract on realised variance with payo¤ at maturity T :

VarSwapT = N σ2 Kvar
n rt2
σ2 = 252 ∑
t =1 n

where

σ2R is the realised variance (or annualised variance) of returns over the
life of the contract,

2 .
The strike price Kvar may be quoted as Kvol

N is the notional amount of the swap contract (typically $100,000 on

OTC)

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 9 / 38

Variance Swap
Demeter…, Derman, Kamal and Zou (1999)

Assume stock prices evolve without jump

dSt
= µ (t, ) dt + σ (t, ) dWt
St
= µt dt + σt dWt (3)

Then realised variance is the continuous integral

Z T
dSt 1
RV = = σ2t dt
St T 0
Z T
T 1
RV = σ2t dt.
2 2 0

From Ito’s lemma,

1 2
d ln St = µt σ dt + σt dWt . (4)
2 t
(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 10 / 38
Subtracting (4) from (3), we get

dSt 1 2
d ln St = σ
St 2 t
So Z T
2 dSt ST
RV = dt ln (5)
T 0 St S0
RT dS t 1
The …rst term of (5), 0 S t dt, is a long position in St share
continuously rebalanced.
The second term of (5), ln SST0 , is the payo¤ of a log contract.

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 11 / 38

Log contract
Neuberger (1994)

The log contract is a futures style contract whose settlement price is

equal to the logarithmic of the price of the asset at T or ln SST0 .
Take
b b 1 1 b
ln = ln b + 1
a a a b a
Z b Z b
1 1 b
= dx + b dx + 1
a x a x2 a
Z b
1 b
= (b x) 2
dx + 1
a x a
Z b Z a
1 1
= 1b >a (b x ) 2 dx 1a >b (x b) dx
a x b x2
b
+ 1
a

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 12 / 38

Log contract
For x 2 [0, ∞)
Z ∞ Z a
b b 1 1
ln = 1 (b x )+ dx (x b )+ dx .
a a a x2 0 x2
Now let
eT
b = S
a = S0
x = K
Then
! Z ∞
eT
S eT
S 1
ln = 1 ( ST x )+ dK
S0 S0 S0 K2
Z S0
1
(x ST ) + dK
0 K2
(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 13 / 38
Realised Variance
Substitute the de…nition of log contract
Z T
!
T dSt eT
S
RV = dt 1
2 0 St S0
Z ∞ Z S0
1+ 1
+ ( ST x) dK + (x ST ) + dK
S0 K2 0 K2

1
The …rst term, as mentioned before, is a futures contract on St .
e
The second term, SST0 1 is a short position of S10 amount of
forward contract struck at S0 = KATM .
The third term a long position in K12 put struck at K for a continum
of K from 0 to S0 .
The forth term is a long position in K12 call struck at K for a
continum of K from S0 to ∞.
Under risk neutral expectation, the …rst and second terms are zero.
(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 14 / 38
Fair Value Pricing Principle

At contract initialisation, the fair value of volatility (or variance) is

the delivery price that makes the swap has zero value at initialisation
such that
E0Q (VarSwapT ) = E0Q σ2R Kvar = 0

Kvar = E0Q σ2R

Z ∞ Z S0
2 1 1
= ( ST x )+ dK + (x ST ) + dK
T S0 K2 0 K2

This means variance swap can be replicated by trading a continuum

of options each strike weighted by 1 K 2 .

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 15 / 38

Volatility Swap
p
Since E (x 2 ) > E (x ), variance swap is more expensive than volatility
swap, and the di¤erence is due to convexity adjustment with respect to
the volatility of volatility.

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 16 / 38

VIX Futures and Forward Variance
A forward starting variance swap is just the di¤erence between two
variance swaps.
Z T2 Z T1 Z T2
σ2t dt = σ2t dt + σ2t dt
0 0 T1
Z T2 Z T2 Z T1
σ2t dt = σ2t dt σ2t dt
T1 0 0
E σ2T 1 ,T 2 = T2 E σ2T 2 T1 E σ2T 1

A VIX future is the expectation today, of the square root of the future
expectation of variance. Hence, a VIX future is bounded above by the
fair volatility of a forward starting variance swap.
Value of the future re‡ects the SPX volatility term structure; an
upward sloping vol surface will produce bigger value of VIX futures
through time.
Futures de-correlate with time. Near dated futures are more
correlated than far dated futures.
(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 17 / 38
VIX Term Structure
VIX Term Structure data calculated using SPX option prices on
Wednesday 29 October, 2008.

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 18 / 38

VIX Options

Multiplier of $100 a point (compared with $1000 for futures)

Range of strikes traded at 2.5pt intervals
Maturities of the next two VIX futures expiries + quarterly cycle
Expire into the same VIX calculation.
European style exercise, call and puts.
Facilitates range of strategies. What combination would be
appropriate for the following ?
Protection from rise in volatility
Believe that volatility is going to become more unstable
Volatility term structure is going to become less steep.

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 19 / 38

Pricing VIX Options with Black-Scholes (or Black)

Carr and Wu (2006) treat VIX like a commodity with known forward price.
(i.e. VIX futures price in this case) Then

c = e rT [FN (d1 ) KN (d2 )]

ln KF + 0.5ηT p
d1 = p , d2 = d1 η T
η T

However, Greeks become di¢ cult

Delta is sensitivity to VIX (a lot like SPX vega), whereas Vega is
sensitivity to Vol of Vol, η.
There is no delta to the S&P (but it surely exists!!)
There is no mechanism to pricing more exotic instruments – e.g. S&P
option with expiries with VIX reaches 40.
There is no insight into the combined dynamics of S&P and volatility

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 20 / 38

VIX Options "Volatility" Smile/Skew

The Black-Scholes (Black) cannot be the right model!!!

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 21 / 38

VIX Options Risk Neutral Density

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 22 / 38

Heston for the Vol Surface

This is the model underlying Bloomberg’s variance swap platform!!!

Heston Process
p
ds = rsdt + s v dwts (6)
p
dv = k (θ v ) dt + η v dwtv (7)
corr (wts ,wtv ) = ρ

The distribution of vt is non-centred chi-squared under Heston.

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 23 / 38

Realized variance under Heston

Realized variance J (t, T ):

ZT
1
J (t, T ) = vτ d τ (8)
T t
t

I (t, T ) is the time t expectation of J (t, T ),

ZT
1
I (t, T ) = Et [J (t, T )] = E (vτ ) d τ (9)
T t
t

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 24 / 38

 " !# !
1 e k (T t) 1 e k (T t)
I (t, T ) = θ 1 + vt (10)
k (T t) k (T t)
= A(t, T ) + B (t, T )vt (11)

0 1
Zt
1 T t
Vtswap = @K vτ d τ I (t, T )A Not e r (T t)
T T
0

At inception, V0swap =0

1 e kT 1 e kT
K = I (0, T ) = θ 1 + v0
kT kT

See Lipton and Pugachevsky (1998), Chatsangar and Poon (2009)

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 25 / 38

Variance Swaption

If T is the maturity date of the swaption and T 0 is the maturity of the

swap and T < T 0 ,

Vtswaption = Et max K I (T , T 0 ) , 0 Not e r (T t)

De…ne

e K A(T , T 0 )
K =
B (T , T 0 )
h i
Vtswaption = B (T , T´)Et max e
K vT , 0 Not e r (T t)

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 26 / 38

Variance Swaption under Heston: A "closed form" solution

Lipton and Pugachevsky (1998), Chatsangar and Poon (2009)

+ h i 4kθ
Et e
K vT = e
K E (vT ) F e;
2c K , 2cvt e k (T t)
η2
e; 4kθ
+E (vT ) f 2c K + 2, 2cvt e k (T t)
η2

+vt e k (T t)
f e ; 4kθ + 4, 2cvt e
2c K k (T t)
η2

Fair value of volatility swap based on Taylor’s series approximation

q q
Var [J (0, T )]
E J (0, T ) E [J (0, T )]
8 (E [J (0, T )])3/2

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 27 / 38

Heston "Implied Vol" as of March 5, 2008

Heston cannot be the right model!!!

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 28 / 38

Proposed Solution:
LMM for Forward Variance

fS
dS (t ) = σS (t ) S (t ) d W

t T
# #
0 T1 T2 ... Tk (t ) 1 Tk ( t ) ... Tn Tn + 1

Q Tk
β e
d Ωk (t ) = vk Ωk k (t ) d Z (t )+1
k

Q T i +1
e
d Ωi (t ) = vi Ωi i (t ) d Z
β
i

Q T n +1
en
d Ωn (t ) = vn Ωn n (t ) d Z
β

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 29 / 38

LMM for Forward Variance

fS
dS (t ) = σS (t ) S (t ) d W
Q T i +1
e
d Ωi (t ) = vi Ωi i (t ) d Z
β
i

dWfS is an equivalent measure of dWS after appropriate Girsanov

QT
transform (and similarly for dZi i +1 ).
Aggregate variance Ωi (t ) is a martingale under its own terminal
measure with forward starting swap at Ti as its numeraire.
Q T i +1
e
Each d Z fS with coe¢ cient ρ .
is correlated with d W
i i
Could write vi = vτ and τ = Ti t to make “Vol of Vol” time
homogenous.

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 30 / 38

Forward Variance under Terminal Measure

The stock price dS (t ) is a driftless SV process with

fS
dS (t ) = σS (t ) S (t ) d W

Under the terminal measure for equity (derivatives)

Z T
1
σS (t ) = σS ,T = Et σu du
T t t
δ k (t ) t 2
= Ωk (t ) + + Ω2i (t ) + + Ω2T δ (t )
T t δ

where T is the maturity of derivative written on S (t ).

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 31 / 38

Forward Variance under Periodic Risk Neutral Measure

fS
dS (t ) = σS (t ) S (t ) d W

For periodic risk neutral measure with multiple maturities and

possibility of early exercise,

i t < i +1
σS (t ) = σS ,i
Z T i +1
1
= Et σu du
Ti + 1 Ti Ti

where σS ,i is the average (or step-wise constant) variance

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 32 / 38

A Model based on Forward Variance

Equity is a geometric process with stochastic vol

dS (t )
= (r q ) dt + σS (t ) dWS
S (t )

t T
# #
0 T1 T2 ... Tk (t ) 1 Tk ( t ) ... Tn Tn + 1

Calibration is based on aggregate variance

Z T i +1
1
ΩT , T
2
(t ) = Ω2i ( t ) = Et σ2u du
| i {zi +}1 Ti + 1 Ti Ti
[vol ref period]

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 33 / 38

Forward Variance: Parity Relationship

Z T i +1
1
Ω2i (t ) = Et σ2u du
Ti + 1 Ti T i
Z T i +1
1
Ω i ( 0 ) = E0
2
σ2u du
Ti + 1 Ti T i
= Fair price of a forward starting variance swap (12)

Z T1
1
Ω20 ( 0 ) = E0 σ2 du
T1 T0 T 0 u
= Fair price of a spot variance swap (13)

From (12) and (13)

Ω2t,T i ,T i +1 = Ω2t,t,T i +1 Ω2t,t,T i

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 34 / 38

Forward Variance vs. Instantaneous Volatility

fS
dS (t ) = σt S (t ) d W

1 Ω2t,T i ,T i +1 ! σ2t if Ti +1 ! Ti ! t

2 Average (or step-wise constant) variance

Z T i +1
1
σ2t,T i ,T i +1 = Ω2t,T i ,T i +1 = Et σ2u du
Ti + 1 Ti Ti

3 If t > i + 1, σ2t,T i ,T i +1 is the realised variance of the reference period.

4 If i < t < i + 1, σ2t,T i ,T i +1 is the sum of variance realised plus the
expectation of the remain variance.

(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 35 / 38

Dynamic of Forward Variance
Each of the forward volatility is a SABR model
β=0.5
d Ωt,T i ,T i +1 = vi Ωt,T i ,T i +1 dZi

Homogenise “Vol of Vol”

vi = vτ
τ = Ti t

Correlation with equity price

hdZi ,t , dWS ,t i = ρi ,S σS vi dt
Correlation between volatility

hdZi ,t , dZj ,t i = ρi ,j vτi vτj dt

Could reduce this to a factor structure or just model adjacent
variance rates.
(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 36 / 38
Calibration:

Assuming that there are only three time references, t, Ti and Ti +1 .

Volatility options and futures at di¤erent maturities will use used to

extract ρi ,j , vτi , vτj , Ωt,t,T i and Ωt,T i ,T i +1 .

Use Ωt,t,T i to price SPX options with maturity Ti under terminal

measure, with the variance swap expiring at Ti as the numeraire.

Use Ωt,t,T i and Ωt,T i ,T i +1 to construct Ωt,t,T i +1 and repeat the

previous step.

The calibration results will be checked against the …t of SPX call and
put options.

The calibrated models can be used to price exotics on equity and

volatility.
(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 37 / 38
References:

Carr Peter and Liuren Wu (2006) A tale of two indices, Journal of

Derivatives, 13, 3, 13-29.
Chatsangar Ratchada and Ser-Huang Poon (2009) Volatility
Derivatives: Volatility/Variance Swap and VIX Option, Working
Paper, Manchester Business School.
Demeter… Kresimir, Emanuel Derman, Michael Kamal and Joseph Zou
(1999) A guide to volatility and variance swap (previously more than
you ever want to know about Vol/Var swap), Journal of Derivatives,
6, 4, 9-32.
Lipton Alex and Dmitry Pugachevsky (1998) Pricing of
volatility-sensitive products in the Heston model framework, Working
Paper, Deutch Bank.
Neuberger Anthony (1994) The log contract, Journal of Portfolio
Management, 20, 2, 74-90.
(Ser-Huang Poon) Consistent Volatility Model September 4, 2009 38 / 38