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

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

116 CHAPTER 3. NUMERICAL IMPLEMENTATION<br />

where w k are the weights corresponding to the chos<strong>en</strong> integration rule. For a giv<strong>en</strong> strike K<br />

such that x 0 ≤ log K ≤ x M−1 , the option price can now be computed by linear interpolation<br />

(other interpolation m<strong>et</strong>hods can also be used):<br />

Ĉ Q √<br />

(T, K) = C A<br />

BS (T, K) + log K − x n K ˆ˜z T (x nK +1) + x n K +1 − log K<br />

ˆ˜z T (x nK ), (3.31)<br />

x nK +1 − x nK x nK +1 − x nK<br />

√<br />

where n K := sup{n : x n ≤ log K} and C A<br />

BS<br />

(T, K) <strong>de</strong>notes the Black Scholes price of a call<br />

option with time to maturity T , strike K and volatility √ A.<br />

Error control The above m<strong>et</strong>hod of evaluating option prices contains three types of numerical<br />

errors: truncation and discr<strong>et</strong>ization errors appear wh<strong>en</strong> the integral is replaced by a finite sum<br />

using Equation (1.28), and interpolation error appears in Equation (3.31). Supposing that the<br />

grid in Fourier space is c<strong>en</strong>tered, that is, u 0 = L/2 and ∆ = L/(M − 1) for some constant L,<br />

Section 1.4 allows to obtain the following bounds for the first two types of error.<br />

Since we have supposed that A > 0, Equation (1.30) provi<strong>de</strong>s a uniform (with respect to Q)<br />

bound for the truncation error:<br />

|ε T | ≤<br />

16e−T<br />

AL2<br />

πT AL 3 .<br />

Proposition 1.13 does not give uniform bounds for the discr<strong>et</strong>ization (sampling) error, however,<br />

since the calibrated Lévy measures are usually similar to the Lévy measures of the prior<br />

process (see Section 3.6), the or<strong>de</strong>r of magnitu<strong>de</strong> of the discr<strong>et</strong>ization error can be estimated<br />

by computing the bounds of Proposition 1.13 for the prior process. The exact form of the error<br />

<strong>de</strong>p<strong>en</strong>ds on the integration rule used; for example for Simpson’s rule one has<br />

( L 4 )<br />

log L<br />

|ε D | = O<br />

M 4 .<br />

A uniform bound for the interpolation error <strong>de</strong>p<strong>en</strong>ds on the interpolation m<strong>et</strong>hod that one<br />

is using. For linear interpolation using Equation (3.31) one easily obtains<br />

|ε I | ≤ d2<br />

8<br />

max ˜z′′ T ,<br />

and the second <strong>de</strong>rivative of the time value is uniformly boun<strong>de</strong>d un<strong>de</strong>r the hypothesis A > 0

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

Saved successfully!

Ooh no, something went wrong!