Bilateral Gamma Processes

in Finance
Stefan Tappe
Vienna Institute of Finance
stefan.tappe@vif.ac.at

based on joint work with:

Uwe Küchler (Humboldt University Berlin)

PRisMa 2008
One-Day Workshop on Portfolio Risk Management

September 29th, 2008, Vienna

Stefan Tappe 2008/09/29

Introduction

6000

5000

4000

3000

2000

1000

0 100 200 300 400 500 600 700

• Stochastic models for risky assets in financial markets.

Bilateral Gamma Processes in Finance 1

Stefan Tappe 2008/09/29

Exponential Lévy models

• Two financial assets:

(
St = S0eXt ,
Bt = ert.

• X is a Lévy process.

σ2
• Example: Black-Scholes model with Xt = σWt + (µ − 2 )t.

• Idea: Take X = X + − X −, where X +, X − are independent subordinators.

Bilateral Gamma Processes in Finance 2

Stefan Tappe 2008/09/29

Bilateral Gamma distributions

• Four parameters:
– Shape parameters: α+, α− > 0,
– Scale parameters: λ+, λ− > 0.

• Γ(α+, λ+; α−, λ−) is the distribution of X + − X −, where

– X + and X − are independent,
– X + ∼ Γ(α+, λ+),
– X − ∼ Γ(α−, λ−).

• Family of bilateral Gamma distributions.

Bilateral Gamma Processes in Finance 3

Stefan Tappe 2008/09/29

Bilateral Gamma processes

• Characteristic function:
 +
α+  −
α−
λ λ
φ(z) = , z ∈ R.
λ+ − iz λ− + iz

• Infinitely divisible with Lévy measure

+ −
 
α −λ + α −λ −
F (dx) = e x
1(0,∞)(x) + e |x|
1(−∞,0)(x) dx.
x |x|

• We call the associated Lévy process a bilateral Gamma process.

Bilateral Gamma Processes in Finance 4

Stefan Tappe 2008/09/29

Outline of the talk

• Bilateral Gamma:
1. Bilateral Gamma distributions.
2. Bilateral Gamma processes.

• Applications to Finance:
1. Option pricing in bilateral Gamma stock models.
2. Illustration of the theory: DAX 1996-1998.

Bilateral Gamma Processes in Finance 5

Stefan Tappe 2008/09/29

Related distributions

• Generalized tempered stable distributions (Cont and Tankov 2004):

+ −
 
α −λ+ x α −λ− |x|
F (dx) = e 1(0,∞)(x) + 1+β − e 1(−∞,0)(x) dx.
x1+β + |x|

• CGMY distributions (Carr, Geman, Madan and Yor 1999):

 
C −M x C
F (dx) = e 1(0,∞) (x) + e−G|x|
1(−∞,0)(x) dx.
x1+Y |x|1+Y

• Variance Gamma distributions (Madan, Carr and Chang 1998).

Bilateral Gamma Processes in Finance 6

Stefan Tappe 2008/09/29

Characterization of Variance Gamma distributions

• Variance Gamma distribution V G(µ, σ 2, ν):

 2
 ν1
σ ν 2
φ(z) = 1 − izµν + z , z ∈ R.
2

• Let µ := Γ(α+, λ+; α−, λ−). There is equivalence between:

1. µ is Variance Gamma;
2. α+ = α−;
3. X is a time-changed Brownian motion Xt = WYt ;
4. µ is a limit of GH-distributions.

Bilateral Gamma Processes in Finance 7

Stefan Tappe 2008/09/29

Representation of the density

• Density function for x > 0:

+ −
(λ+)α (λ−)α 1 +
+α− )−1 x +
−λ− )
f (x) = 1 + +α− )
· x 2 (α · e− 2 (λ
(λ+ + λ−) 2 (α Γ(α+)
· W 1 (α+−α−), 1 (α++α−−1)(x(λ+ + λ−)).
2 2

• Wλ,µ denotes the Whittaker function

z ∞ µ+λ− 12
z λe− 2
Z 
µ−λ− 12 −t t
Wλ,µ(z) = t e 1+ dt, z>0
Γ(µ − λ + 12 ) 0 z

for µ − λ > − 12 .

Bilateral Gamma Processes in Finance 8

Stefan Tappe 2008/09/29

Properties of the density

• Unimodality: f is strictly increasing/decreasing on (−∞, x0)/(x0, ∞).

• Smoothness: Let N ∈ N be such that N < α+ + α− ≤ N + 1. Then

f ∈ C N −1(R) \ C N (R).

• Semi-heavy tails: We have the asymptotic behaviour

+ +
(
C1xα −1e−λ x for x → ∞
f (x) ∼ α− −1 −λ− |x|
C2|x| e for x → −∞.

Bilateral Gamma Processes in Finance 9

Stefan Tappe 2008/09/29

Density shapes

2 alpha-

alpha+

0 1 2

Bilateral Gamma Processes in Finance 10

Stefan Tappe 2008/09/29

Bilateral Gamma processes

• X = X + − X −, in particular FV process and no Gaussian part.

• Infinitely many jumps in each compact interval.

• For 0 ≤ s < t we have:

Xt − Xs ∼ Γ(α+(t − s), λ+; α−(t − s), λ−).

• Efficient methods for simulating bilateral Gamma processes.

Bilateral Gamma Processes in Finance 11

Stefan Tappe 2008/09/29

Quick Review

• Bilateral Gamma distributions:

– Simple characteristic function.
– Densities: unimodal, semi-heavy tailed.

• Bilateral Gamma processes:

– FV sample paths with infinitely many jumps on every interval.
– Easy to simulate.

Bilateral Gamma Processes in Finance 12

Stefan Tappe 2008/09/29

Exponential bilateral Gamma models

• Two financial assets:

(
St = S0eXt ,
Bt = ert.

• X ∼ Γ(α+, λ+; α−, λ−) is a bilateral Gamma process.

• The market is:

– free of arbitrage,
– but not complete.

Bilateral Gamma Processes in Finance 13

Stefan Tappe 2008/09/29

Option Pricing

• Price of a European Call Option:

Π = EQ[(ST − K)+],

where Q ∼ P is a martingale measure.

• Fourier transformation: Under ”appropriate conditions”

−rT Z iν+∞  iz

e K K φT (−z)
Π=− dz,
2π iν−∞ S0 z(z − i)

where φT is the characteristic function of XT under Q.

Bilateral Gamma Processes in Finance 14

Stefan Tappe 2008/09/29

Requirements on the martingale measure

• There are several martingale measures Q ∼ P.

• Under Q, the characteristic function φT should be ”simple”.

• Recall that for a bilateral Gamma process:

 +
α+  −
α−
λ λ
φ(z) = , z ∈ R.
λ+ − iz λ− + iz

• Q should have an economic interpretation.

Bilateral Gamma Processes in Finance 15

Stefan Tappe 2008/09/29

Esscher transforms

• For θ ∈ (−λ−, λ+) we define Pθ ∼ P as

dPθ 
 := eθXt−Ψ(θ)t, t ≥ 0.
dP Ft

• The cumulant generating function Ψ of X is given by

+ −
   
λ λ
Ψ(z) = α+ ln + α− ln , z ∈ (−λ−, λ+).
λ+ − z λ− + z

Bilateral Gamma Processes in Finance 16

Stefan Tappe 2008/09/29

Esscher martingale measure

• There exists θ ∈ (−λ−, λ+) such that Pθ is a martingale measure iff

λ+ + λ− > 1.

• In this case, it is the unique solution of the equation

 +
α+  −
α−
λ −θ λ +θ
= er−q , θ ∈ (−λ−, λ+ − 1).
λ+ − θ − 1 λ− + θ + 1

• We have X ∼ Γ(α+, λ+ − θ; α−, λ− + θ).

Bilateral Gamma Processes in Finance 17

Stefan Tappe 2008/09/29

Relative entropy

• For each Q ∼ P define the relative entropy

 
dQ 

HFt (Q | P) = EQ ln  , t ≥ 0.
dP Ft

• Find a martingale measure Q∗ ∼ P such that

HFt (Q∗ | P) = min HFt (Q | P), t ≥ 0.

Q∈EMM

• Minimal entropy martingale measure (MEMM).

Bilateral Gamma Processes in Finance 18

Stefan Tappe 2008/09/29

Exponential transform

• Let X̃t := L(eXt−(r−q)t) be the exponential transform.

• X̃ is again a Lévy process.

• For θ ≤ 0 we define Pθ ∼ P as

dPθ 
 := eθX̃t−Ψ̃(θ)t, t ≥ 0.
dP Ft

• Ψ̃ denotes the cumulant generating function of X̃.

Bilateral Gamma Processes in Finance 19

Stefan Tappe 2008/09/29

Minimal entropy martingale measure

• If λ+ > 1, there exists θ ≤ 0 such that Pθ is a MEMM iff

+ −
   
+ λ − λ
α ln + α ln ≥ r − q.
λ+ − 1 λ− + 1

• In this case, it is the unique solution of the equation

−
∞ −λ+ x ∞
e−λ x
Z Z
+ e x x
θ(e −1) + −x θ(e−x −1)
α (e − 1)e dx + α (e − 1)e dx
0 x 0 x
= r − q, θ ≤ 0.

Bilateral Gamma Processes in Finance 20

Stefan Tappe 2008/09/29

Characteristic function

• X is a Lévy process under Pθ with

Z 
x
φ(z) = exp (eizx − 1)eθ(e −1)F (dx) , z ∈ R.
R

• The value of the minimal relative entropy is given by

HF1 (Pθ | P) = r − q
Z ∞ −λ+x Z ∞ −λ+x
+ e x
θ(e −1) − e θ(e−x −1)
−α (e − 1)dx − α (e − 1)dx.
0 x 0 x

Bilateral Gamma Processes in Finance 21

Stefan Tappe 2008/09/29

Martingale measures considered so far

• Esscher martingale measure:

– Pro: Easy to obtain, X remains bilateral Gamma under Pθ .
– Contra: No economic interpretation.

• Minimal entropy martingale measure:

– Pro: Easy to obtain, economic interpretation.
– Contra: Characteristic function of X under Pθ not in closed form.

Bilateral Gamma Processes in Finance 22

Stefan Tappe 2008/09/29

Bilateral Esscher transforms

• Recall that X = X + − X −.
+
,θ − )
• For θ+ ∈ (−∞, λ+) and θ− ∈ (−∞, λ−) we define P(θ ∼ P as

+ −
dP(θ ,θ )  θ + Xt+ −Ψ+ (θ + )t θ − Xt− −Ψ− (θ − )t
 := e ·e , t ≥ 0.
dP Ft

• Ψ+, Ψ− denote the cumulant generating functions of X +, X −.

Bilateral Gamma Processes in Finance 23

Stefan Tappe 2008/09/29

Bilateral Esscher martingale measures

• Define Φ : (−∞, λ+ − 1) → (−∞, λ−) as

+
 +
 α− −1
λ −θ α
Φ(θ) := λ− − e−(r−q) − 1 .
λ+ − θ − 1

• Then P(θ,Φ(θ)) is a martingale measure for each θ ∈ (−∞, λ+ − 1).

• We have X ∼ Γ(α+, λ+ − θ; α−, λ− − Φ(θ)) under P(θ,Φ(θ)). Thus:

−rT iν+∞  iz  +

α+T  −
α−T
λ −θ λ − Φ(θ)
Z
e K K dz
Π=− .
2π iν−∞ S0 λ+ − θ + iz λ− − Φ(θ) − iz z(z − i)

Bilateral Gamma Processes in Finance 24

Stefan Tappe 2008/09/29

Minimizing the relative entropy

• Find θ ∈ (∞, λ+ − 1) such that

HF1 (P(θ,Φ(θ)) | P) = min HF1 (P(ϑ,Φ(ϑ)) | P) ≥ min HF1 (Q | P).

ϑ∈(−∞,λ+ −1) Q∈EMM

• The relative entropy is given by:

+ +
  
λ λ
HF1 (P(θ,Φ(θ)) | P) = α+ − 1 − ln
λ+ − θ λ+ − θ
− −
  
− λ λ +
+α − 1 − ln , θ ∈ (∞, λ − 1).
λ− − Φ(θ) λ− − Φ(θ)

Bilateral Gamma Processes in Finance 25

Stefan Tappe 2008/09/29

Example: DAX 1996-1998

6000

5000

4000

3000

2000

1000

0 100 200 300 400 500 600 700

• We assume St = S0eXt , where X ∼ Γ(α+, λ+; α−, λ−).

Bilateral Gamma Processes in Finance 26

Stefan Tappe 2008/09/29

Estimates

• Maximum Likelihood Estimate:

(α+, λ+; α−, λ−) = (1.55, 133.96; 0.94, 88.92).

• With θ = −5.30 we have

HF1 (P(θ,Φ(θ)) | P) = min HF1 (P(ϑ,Φ(ϑ)) | P) = 0.0029411

ϑ∈(−∞,λ+ −1)

and X ∼ Γ(1.55, 139.26; 0.94, 83.65) under P(θ,Φ(θ)).

• Note that minQ∈EMM HF1 (Q | P) = 0.0029409.

Bilateral Gamma Processes in Finance 27

Stefan Tappe 2008/09/29

Black Scholes and Bilateral Gamma Prices

1200

1000

800

600

400

200

0
4000 4200 4400 4600 4800 5000 5200 5400 5600 5800 6000

Bilateral Gamma Processes in Finance 28

Stefan Tappe 2008/09/29

Implied volatility surface

4460
4440
4420
3*10^5*sigma

4400
4380 4000
4360
4400
4340
4320 4800

1000 5200
1200 K
1400 5600
1600
10*T 1800
6000
2000

Bilateral Gamma Processes in Finance 29

Stefan Tappe 2008/09/29

Relation to the Normal distribution

• The Central Limit Theorem yields:

+ − + −
 
α α α α
Γ(α+, λ+; α−, λ−) ≈ N +
− −
, + 2
+ − 2
.
λ λ (λ ) (λ )

• Berry-Esseen gives us:

c
sup |Fn(x) − Φn(x)| ≤ √ ,
x∈R n

where α+ → nα+ and α− → nα− for some n ∈ N.

Bilateral Gamma Processes in Finance 30

Stefan Tappe 2008/09/29

Conclusion

• Stock models based on bilateral Gamma processes.

• Option pricing using historical data.

• Current Research: Model calibration to option price data.

Bilateral Gamma Processes in Finance 31