31.05.2014 Views

Processus de Lévy en Finance - Laboratoire de Probabilités et ...

Processus de Lévy en Finance - Laboratoire de Probabilités et ...

Processus de Lévy en Finance - Laboratoire de Probabilités et ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

3.5. NUMERICAL SOLUTION 115<br />

of X t un<strong>de</strong>r Q at the points u − i for real u:<br />

ψ Q (u − i) = − A M−1<br />

2 u(u − i) − (1 + iu) ∑<br />

Introducing uniform grids<br />

j=0<br />

M−1<br />

∑<br />

(e x j<br />

− 1)q j −<br />

j=0<br />

M−1<br />

∑<br />

q j +<br />

j=0<br />

e iux j<br />

e x j<br />

q j .<br />

x j = x 0 + jd and u k = u 0 − k∆,<br />

the last term becomes:<br />

M−1<br />

∑<br />

j=0<br />

e iu kx j<br />

e x j<br />

q j = e −ik∆x 0<br />

M−1<br />

∑<br />

j=0<br />

e −ikjd∆ e (iu 0+1)x j<br />

q j<br />

and comparing this to Equation (1.27), we see that if d∆ = 2π M , the last term of {ψQ (u k −i)} M−1<br />

k=0<br />

becomes a discr<strong>et</strong>e Fourier transform (cf. Equation (1.27)):<br />

ψ k := ψ Q (u k − i) = − A M−1<br />

2 u ∑<br />

k(u k − i) − (1 + iu k )<br />

j=0<br />

(e x j<br />

− 1)q j −<br />

M−1<br />

∑<br />

j=0<br />

q j<br />

+ e −ik∆x 0<br />

DFT k [e (iu 0+1)x j<br />

q j ]. (3.28)<br />

This expression can therefore be computed using the fast Fourier transform algorithm. Note<br />

that at this stage all computations are exact: there are no truncation or discr<strong>et</strong>ization errors.<br />

For a giv<strong>en</strong> maturity date T we now need to compute the Fourier transform ˜ζ T<br />

modified time value of options (see Equation (1.25)) at the points {u k } M−1<br />

k=0 :<br />

of the<br />

˜ζ T (u k ) = e iu krT eT ψ k<br />

− e − AT<br />

2 (u k(u k −i))<br />

iu k (1 + iu k )<br />

. (3.29)<br />

Option time values are th<strong>en</strong> approximated using Equation (1.28) on the same grid of log-strikes<br />

{x j } M−1<br />

j=0<br />

that was used to discr<strong>et</strong>ize the Lévy measure. 2 In the following equation and below<br />

we <strong>de</strong>note approximated quantities by putting a hat over the corresponding variables.<br />

ˆ˜z T (x j ) = ∆ M−1<br />

∑<br />

2π e−ix ju M−1<br />

k=0<br />

w k ˜ζT (u M−1−k )e −ix 0∆k e −2πijk/M<br />

= ∆ 2π e−ix ju M−1<br />

DFT j [w k ˜ζT (u M−1−k )e −ix 0∆k ], (3.30)<br />

2 Actually, the grid of log-strikes may be shifted but we do not do this here to simplify the formulas.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!