Lecture 11

Volatility expansion
Bruno Dupire
Bloomberg LP
IV. Volatility Expansion
Bruno Dupire
3
Introduction
This talk aims at providing a better
understanding of:

How local volatilities contribute to the value of
an option
How P&L is impacted when volatility is
misspecified
Link between implied and local volatility
Smile dynamics
Vega/gamma hedging relationship
Bruno Dupire
4
In the following, we specify the dynamics
of the spot in absolute convention (as
opposed to proportional in Black-Scholes)
and assume no rates:

: local (instantaneous) volatility
(possibly stochastic)
Implied volatility will be denoted by

dS dW
t t t
=o
o

o
Framework & definitions
Bruno Dupire
5
P&L
S
t t +A
S
t
O
Break-even
points
o At
o At
Option Value
S
t
C
t
C
t t +A
S
Delta
hedge
P&L of a delta hedged option
Bruno Dupire
6
P&L Histogram P&L Histogram
Expected P&L = 0 Expected P&L > 0
Correct volatility Volatility higher than expected
S
t t +A
S
t
S
t t +A
S
t
Ito: When , spot dependency disappears At 0
1std 1std
1std 1std
P&L of a delta hedged option (2)
Bruno Dupire
7
P&L is a balance between gain from I and loss from O:
( )
P L dt
t t dt
&
, +
= +
|
\

|
.
|
o
2
0 0
2
I O
=> discrepancy if o different from expected:
( ) gain over dt =
1
2
2
0
2
0
o o I dt
- >
- <

`
)
o o
o o
0
0
0
:
:
Profit
Loss
Magnified by I
O I
0
0
2
0
2
=
o
From Black-
Scholes PDE:
Black-Scholes PDE
Bruno Dupire
8
Total P&L over a path
= Sum of P&L over all small time intervals
P L
dt
T
&
( )
=

}
1
2
2
0
2
0
0
o o I
Spot
Time
Gamma
No assumption is made
on volatility so far
P&L over a path
Bruno Dupire
9
The probability density may correspond to the density of
a NON risk-neutral process (with some drift) with volatility .
Terminal wealth on each path is:
Taking the expectation, we get:
| |
E X E S dS dt
T

o o wealth
T
= +

} }
( ) [ ( )| ] E I
0 0
2
0
2
0 0
1
2
( is the initial price of the option)
( ) wealth =
T
X dt
T
E I
0
2
0
2
0
0
1
2
+
}
( ) o o
( ) X E
0
General case
Bruno Dupire
10
In a complete model (like Black-Scholes), the drift does
not affect option prices but alternative hedging strategies
lead to different expectations
Example: mean reverting process
towards L with high volatility around L
We then want to choose K (close to L) T
and
0
(small) to take advantage of it.
L
Profile of a call (L,T) for
different vol assumptions
In summary: gamma is a volatility
collector and it can be shaped by:
a choice of strike and maturity,
a choice of
0
, our hedging volatility.
Non Risk-Neutral world
Bruno Dupire
11
X X E S dS dt
T
( ) ( ) [ ( )| ] E E I = +

} } 0 0
2
0
2
0 0
1
2
o o
X X E S dS dt ( ) ( ) ( ) [ | ] E E I = +
} } 0
2
0
2
0
1
2
o o
C C E S dS dt ( ) ( ) ( [ | ] ) E E I = +
} } 0
2
0
2
0
1
2
o o
From now on, will designate the risk neutral density
associated with .
In this case, E[wealth
T
] is also and we have:
Path dependent option & deterministic vol:
European option & stochastic vol:
dS dW
t t
=o
( )
X E
Average P&L
Bruno Dupire
12
Buy a European option at 20% implied vol
Realised historical vol is 25%
Have you made money ?

Not necessarily!
High vol with low gamma, low vol with high gamma
Quiz
Bruno Dupire
13
An important case is a European option with
deterministic vol:
C C dS dt ( ) ( ) ( ) E E I = +
} } 0
2
0
2
0
1
2
o o
The corrective term is a weighted average of the volatility
differences

This double integral can be approximated numerically
Expansion in volatility
Bruno Dupire
14
t
Delta = 100%
K
Delta = 0%
S
( ) ( ) C S K K t K t dt
T
( ) ( ) , , E = +
+
} 0
2
0
1
2
o
Extreme case:
o o
0 0
0 = = I
K
This is known as Tanakas formula
P&L: Stop Loss Start Gain
Bruno Dupire
15
13
15
17
19
21
23
I
m
p
l
i
e
d

v
o
l
Implied volatility
strike
maturity
11
15
19
23
27
31
L
o
c
a
l

v
o
l
Local volatility
spot
time
Aggregation
Differentiation
Local / Implied volatility relationship
Bruno Dupire
16
o
c
c
c
c
2
2
2
2 ( , ) S T
C
T
C
K
=
Stripping local vols from implied vols is the
inverse operation:
Involves differentiations
(Dupire 93)
Smile stripping: from implied to local
Bruno Dupire
17
Let us assume that local volatility is a
deterministic function of time only:
( )
dS t dW
t t
= o
In this model, we know how to combine local
vols to compute implied vol:
( )
( )

o
o
T
t dt
T
T
=
}
2
0
Question: can we get a formula with ?
( ) o S t ,
From local to implied: a simple case
Bruno Dupire
18
Implied Vol is a weighted average of Local Vols
(as a swap rate is a weighted average of FRA)
When = implied vol
o
0
solve by iterations depends on I
0
0
o

1
2
0
2
0
2
0
( ) o o =
}}
I dSdt
o
o

0
2
2
0
0
=
}}
}}
I
I
dSdt
dSdt

From local to implied volatility

Bruno Dupire
19
Weighting Scheme: proportional to
I
0

At the
money
case:

S
0
=100
K=100
Out of the
money
case:
S
0
=100
K=110
t
S
I
0

Weighting scheme
Bruno Dupire
20
Weighting scheme is roughly proportional to the
brownian bridge density
Brownian bridge density:
( )
| |
BB x t P S x S K
K T t T

,
, = = =
Weighting scheme (2)
Bruno Dupire
21

( ) ( ) o o o
2 2
=
}
S S dS
small
o T
large
o T
S
0
S
o(S)
S
0
S
K
o(S)
S
0
S
o(S)
S
0
S
K
o(S)
ATM (K=S
0
) OTM (K>S
0
)
( )
o

S
dt
dSdt
=
}
}}
I
I
0
0
Time homogeneous case
Bruno Dupire
22
and are
averages of the same local
vols with different weighting
schemes

o
K
1

o
K
2
=> New approach gives us a direct expression for
the smile from the knowledge of local volatilities

But can we say something about its dynamics?
S
0
K
2
K
1
t
Link with smile
Bruno Dupire
23
23.5
24
24.5
25
25.5
26
S
0
K
t
S
1
Weighting scheme imposes
some dynamics of the smile for
a move of the spot:
For a given strike K,

(we average lower volatilities)
S
K
| +

o
Smile today (Spot S
t
)
S
t
S
t+dt
Smile tomorrow (Spot S
t+dt
)
in sticky strike model
Smile tomorrow (Spot S
t+dt
)
if o
ATM
=constant
Smile tomorrow (Spot S
t+dt
)
in the smile model
&
Smile dynamics
Bruno Dupire
24
A sticky strike model ( ) is arbitrageable.
( )

o o
K K
t =
Let us consider two strikes K
1
< K
2

The model assumes constant vols o
1
> o
2
for example
o
1
dt
dt
2
o

o
1
dt
o
2
dt

By combining K
1
and K
2
options, we build a position with no gamma and
positive theta (sell 1 K
1
call, buy I
1
/I
2
K
2
calls)

Sticky strike model
Bruno Dupire
25
C C dSdt ( ) ( ) ( ) E E I = +
}} 0
2
0
2
0
1
2
o o
If & are constant o o
0
o o c
2
0
2
= +
}}
I + = + dSdt C C c o c o
0
2
0
2
0
2
1
) ( ) (

Vega =
2
2
co
c C
c
co
c
co
co
co
c
co
o
C C C
= =
2
2
2
2
Vega analysis
Bruno Dupire
26
Hedge in insensitive to realised
historical vol
If =0 everywhere, no sensitivity to
historical vol => no need to Vega hedge
Problem: impossible to cancel now for
the future
Need to roll option hedge
How to lock this future cost?
Answer: by vega hedging

Gamma hedging vs Vega hedging
Bruno Dupire
27
X X E S dS dt ( ) ( ) ( ) [ | ] E E I = +
} } 0
2
0
2
0
1
2
o o
For any option, in the deterministic vol case:
For a small shift in local variance around (S,t), we have:
( )
( ) ( )
X X E S
dX
d
E S t S S t
S t
( ) ( ) [ | ]
[ , | ] ,
( , )
E E I
I
= +
=
0 0
2
0
1
2
1
2
c
o

For a european option:
( )
( ) ( )
dC
d
S t S t
S t
o

( , )
, ,
2
0
1
2
= I
Superbuckets: local change in local vol
Bruno Dupire
28
Local change of implied volatility is obtained by combining
local changes in local volatility according a certain weighting
( ) ( )
( )
( )
dC
d
dC
d
d
d

o o
o
o
2 2
2
2
=
}
sensitivity in
local vol
weighting obtain
using stripping
formula
Thus:
cancel sensitivity to any move of implied vol
<=> cancel sensitivity to any move of local vol
<=> cancel all future gamma in expectation
Superbuckets: local change in implied vol
Bruno Dupire
29
This analysis shows that option prices are
based on how they capture local volatility

It reveals the link between local vol and
implied vol

It sheds some light on the equivalence
between full Vega hedge (superbuckets)
and average future gamma hedge
Conclusion
Bruno Dupire
30
Dual Equation

The stripping formula


can be expressed in terms of

When


solved by
KK
T
C K
C
2
2
2 = o
:

o
KX
K
o
o
o
o

+
=
0 T
( )
}
=
K
S
u u
du
X
0 ,

o
o
Bruno Dupire
31
Large Deviation Interpretation
The important quantity is

If then satisfies:

and


( )
}
K
S
u u
du
0 , o
( )dW x a dx =
( )
( )
}
=
x
x
u a
du
x y
0
dW dt dy + =
( )
( )
}
} }

= =
K
S
K
S
K
S
u u
du
S K
u u
du
u
du
0 ,
ln ln

0 ,

o
o
o o
o
( ) K y W K x
t t
= =
Bruno Dupire
32
Delta Hedging
We assume no interest rates, no dividends, and absolute
(as opposed to proportional) definition of volatility
Extend f(x) to f(x,v) as the Bachelier (normal BS) price of
f for start price x and variance v:

with f(x,0) = f(x)
Then,
We explore various delta hedging strategies
}

dy e y f
v
X f E v x f
v
x y
v x
2
) (
,
2
) (
2
1
)] ( [ ) , (
t
) , (
2
1
) , ( v x f v x f
xx v
=
Bruno Dupire
33
Calendar Time Delta Hedging
Delta hedging with constant vol: P&L depends on the
path of the volatility and on the path of the spot price.
Calendar time delta hedge: replication cost of


In particular, for sigma = 0, replication cost of


}
+
t
u xx
du dQV f T X f
0
2
, 0
2
0
) (
2
1
) . , ( o o
)) .( , (
2
t T X f
t
o
}
+
t
u xx
dQV f X f
0
, 0 0
2
1
) (
) (
t
X f
Bruno Dupire
34
Business Time Delta Hedging
Delta hedging according to the quadratic variation:
P&L that depends only on quadratic variation and
spot price


Hence, for


And the replicating cost of is
finances exactly the replication of f until
t x t xx t v t x t t
dX f dQV f dQV f dX f QV L X df = + =
, 0 , 0 , 0
2
1
) , (
,
, 0
L QV
T
s
t
t
u u x t t
dX QV L X f L X f QV L X f
}
+ =
0
, 0 0 , 0
) , ( ) , ( ) , (
) , (
, 0 t t
QV L X f ) , (
0
L X f
) , (
0
L X f L QV =
t
t
, 0
:
Bruno Dupire
35
Daily P&L Variation
Bruno Dupire
36
Tracking Error Comparison

V. Stochastic Volatility Models
Bruno Dupire
38
Stochastic volatility model Hull&White (87)
P
t t
P
t t
t
t
dZ dt d
dW rdt
S
dS
| o o
o
+ =
+ =
Incomplete model, depends on risk premium
Does not fit market smile
Strike price
I
m
p
l
i
e
d

v
o
l
a
t
i
l
i
t
y
Strike price
I
m
p
l
i
e
d

v
o
l
a
t
i
l
i
t
y

Z W ,
< 0

Z W ,
= 0
Hull & White
Bruno Dupire
39
Role of parameters
Correlation gives the short term skew
Mean reversion level determines the long term
value of volatility
Mean reversion strength
Determine the term structure of volatility
Dampens the skew for longer maturities
Volvol gives convexity to implied vol
Functional dependency on S has a similar effect
to correlation
Bruno Dupire
40
Heston Model
( )

= + =
+ =
dt dZ dW dZ v dt v v dv
dW v dt
S
dS
q

,
Solved by Fourier transform:
( ) ( ) ( ) t t t
t
, , , , , ,
ln
0 1 ,
v x P v x P e v x C
t T
K
FWD
x
x
T K
=
=





Bruno Dupire
46
2 ways to generate skew in a stochastic vol model







-Mostly equivalent: similar (St,ot ) patterns, similar
future
evolutions
-1) more flexible (and arbitrary!) than 2)
-For short horizons: stoch vol model local vol model
+ independent noise on vol.
Spot dependency
( ) ( )
( ) 0 , ) 2
0 , , , ) 1
=
= =
Z W
Z W t S f x
t t

o
0
S
S
T
o

S
T
o

0
S
Bruno Dupire
47
Convexity Bias
| |
( )

=
= = =
=
0 ,
? |
2
0
2 2
Z W
K S E dZ d
dW dS
t t t
t

o o o o
o
| |
2
0
2
only NO! o o =
t
E
t
o
likely to be high if
0 0
or S S S S
t t
<< >>
| | K S E
t t
= |
2
o
0
S
2
0
o
K
Bruno Dupire
48
Risk Neutral drift for instantaneous forward
variance
Markov Model:
fits initial smile with local vols


Impact on Models
( ) dW t S f
S
dS
t
o , =
( ) t S, o
( )
( )
] | [
,
,
2
2
S S E
t S
t S f
t t
=
=
o
o
Bruno Dupire
49
Skew case (r<0)





- ATM short term implied still follows the local vols


- Similar skews as local vol model for short horizons
- Common mistake when computing the smile for another
spot: just change S
0
forgetting the conditioning on o :
if S : S
0
S
+
where is the new o ?
Smile dynamics: Stoch Vol Model (1)
| | ( ) ( ) T K K S E
T
T ,
2 2
o o = =
Local vols

+
S
0
S

S Smile
0
Smile S
+
S Smile
K

o

Bruno Dupire
50
Smile dynamics: Stoch Vol Model (2)
Pure smile case (r=0)






ATM short term implied follows the local vols
Future skews quite flat, different from local vol
model
Again, do not forget conditioning of vol by S
Local vols

S Smile
0
Smile S
+
S Smile

S
0
S
+
S
K

o

Forward Skew
Bruno Dupire
52
Forward Skews
In the absence of jump :
model fits market
This constrains
a) the sensitivity of the ATM short term volatility wrt S;
b) the average level of the volatility conditioned to S
T
=K.

a) tells that the sensitivity and the hedge ratio of vanillas depend on the
calibration to the vanilla, not on local volatility/ stochastic volatility.
To change them, jumps are needed.
But b) does not say anything on the conditional forward skews.
) , ( ] [ ,
2 2
T K K S E T K
loc T T
o o = =
Bruno Dupire
53
Sensitivity of ATM volatility / S
S c
c
2
o
dS
S
t S
t S t t S S S S S E
loc
t loc t loc t t t t t t

c
c
~ + + = + =
+
) , (
) , ( ) , ( ] [
2
2 2 2 2
o
o o o o o o o
o o
2
ATM
o
At t, short term ATM implied volatility ~
t
.
As
t
is random, the sensitivity is defined only in average:
In average, follows .
Optimal hedge of vanilla under calibrated stochastic volatility corresponds to
perfect hedge ratio under LVM.
2
loc
o