You are on page 1of 9

Contents

1. Motivation
2. Basic notions
3. Lévy-Ito decomposition and path properties
● Levy-Khinchin representation and Levy-Ito decomposition
● Properties of sample paths
● Subordinators
● Levy processes and martingales
● …
4. Financial modelling with Levy processes
● Building Levy processes
5. Simulating Levy processes

6. Stochastic calculus with Levy processes


● Stochastic integrals, Ito formula
7. Measure transformations for Levy processes
8. Options pricing
● Risk-neutral modelling with exponential Levy processes

1
Paul Lévy (1886 – 1971)

2
4. Properties

 Definition: Characteristic function of process (Xt):


𝚽𝒕 𝒛 = 𝑬 𝒆𝒊𝒛𝑿𝒕 ∀ 𝒛 ∈ ℝ, 𝒕 ∈ ℝ+
● The characteristic function characterizes the distribution of
the process at any instant

 By stationarity and independence of increments, we see that


𝑡 ↦ Φ𝑡 (𝑧) is multiplicative :

𝐸 𝑒 𝑖𝑧𝑋𝑡+𝑠 = 𝐸 𝑒 𝑖𝑧(𝑋𝑡+𝑠−𝑋𝑠 ) 𝑒 𝑖𝑧𝑋𝑠 = 𝐸 𝑒 𝑖𝑧(𝑋𝑡+𝑠−𝑋𝑠) ] 𝐸[𝑒 𝑖𝑧𝑋𝑠


= 𝐸 𝑒 𝑖𝑧𝑋𝑡 ] 𝐸[𝑒 𝑖𝑧𝑋𝑠 = Φ𝑡 𝑧 Φ𝑠 (𝑧)

3
4. Properties

 Moreover, by the stochastic continuity, this function (as a


function of t) is continuous
● 𝑋𝑠 → 𝑋𝑡 in probability if 𝑠 → 𝑡 , which implies convergence in
distribution, and hence convergence of characteristic
functions: Φ𝑠 𝑧 → Φ𝑡 (𝑧) for any z fixed

 Multiplicative and continuous ⇒ it can be written as an


exponential function of t:
there exists 𝜓 = 𝜓(𝑧) such that Φ𝑡 𝑧 = 𝑒 𝜓 𝑧 𝑡

● 𝜓 is called the characteristic exponent of the process X

 Now, 𝜓 is also a continuous function of z, as Φ is continuous in z

4
4. Properties

 Proposition: Let X be a Levy process on ℝ. Then there exists a


continuous function 𝜓: ℝ → ℝ, called characteristic exponent of
X, such that
Φ𝑡 𝑧 = 𝐸 𝑒 𝑖𝑧𝑋𝑡 = 𝑒 𝑡𝜓(𝑧)
 The distribution of 𝑋𝑡 is hence entirely determined by the
characteristic exponent 𝜓
 Moreover, as Φ1 𝑧 = 𝑒 𝜓(𝑧) , we can write:
Φ𝑡 𝑧 = 𝑒 𝑡𝜓(𝑧) = 𝑒 𝑡 ln Φ1 (𝑧) ,
 This shows that the distribution of 𝑿𝒕 is entirely determined by
the distribution of 𝑿𝟏
 The Levy process is entirely determined by the distribution of 𝑋1
● This is the only “degree of freedom” for specifying a Levy process
5
4. Properties

 Proposition: If X is a compound Poisson process, then its


characteristic function is given by:

𝜆𝑡 ℝ 𝑒 𝑖𝑧𝑥 −1 𝑑𝐹(𝑥)
Φ𝑡 𝑧 = 𝐸 𝑒 𝑖𝑧𝑋𝑡 = 𝑒 ∀𝑧 ∈ ℝ

→ 𝜓 𝑧 =𝜆 𝑒 𝑖𝑧𝑥 − 1 𝑑𝐹(𝑥)

 Proof:…

6
4. Properties

 Definition: Levy measure of a compound Poisson process


𝑁𝑡

𝑋𝑡 = 𝑌𝑘 , 𝑌𝑘 𝑖. 𝑖. 𝑑 𝐹
𝑘=1
We introduce the new (positive) measure ν defined on ℝ, ℬ by:
𝜈 𝐴 =𝜆 𝑑𝐹 𝑥 = 𝜆ℙ 𝑌𝑘 ∈ 𝐴 ∀𝐴 ∈ ℬ
𝐴
Then the last result becomes:
+∞
Φ𝑡 𝑧 = exp[𝑡 𝑒 𝑖𝑧𝑥 − 1 𝜈(𝑑𝑥)]
−∞
Or:
+∞
𝜓 𝑧 = 𝑒 𝑖𝑧𝑥 − 1 𝜈(𝑑𝑥)
−∞

 Special case of Levy-Khinchin representation theorem


7
4. Properties

 The Levy measure 𝜈 is a deterministic measure


 It is not a probability measure (in general), but only a positive
measure as 𝜈 ℝ = 𝜆 ( ≠ 1 a priori)

 𝜆 = average number of jumps of the Poisson process by time


unit ⇒ 𝜈 𝐴 can be interpreted as :
𝜈 𝐴 = average number of jumps of size in A
observed by time unit

 In particular, 𝜈 ℝ = 𝜆 < ∞
 we speak about a “finite activity” Levy process

8
4. Properties

 Definition: Poisson random measure


 Definition: jump measure of a compound Poisson process
 Properties of the jump measure of a compound Poisson process
 Computation of the jump measure of a compound Poisson
process
 Representation of a compound Poisson process by a Poisson
integral

 Generalisation to any Levy process


 Levy-Ito decomposition

You might also like