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 ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
30 INTRODUCTION<br />
Lévy mo<strong>de</strong>ls by Fourier transform. This section is the only one of the first chapter to contain<br />
new results: we propose original improvem<strong>en</strong>ts to this m<strong>et</strong>hod and obtain estimates for the<br />
truncation and discr<strong>et</strong>ization errors, that are not giv<strong>en</strong> in the original refer<strong>en</strong>ce [23].<br />
The second chapter is <strong>de</strong>dicated to the theor<strong>et</strong>ical treatm<strong>en</strong>t of the calibration problem.<br />
We start by discussing the least squares calibration m<strong>et</strong>hod, commonly used by aca<strong>de</strong>mics<br />
and practitioners. We show that in the context of non-param<strong>et</strong>ric calibration of expon<strong>en</strong>tial<br />
Lévy mo<strong>de</strong>ls, least squares calibration does not always allow to find a solution and ev<strong>en</strong> if it<br />
does, the solution is typically very s<strong>en</strong>sitive to small perturbations of input data. To solve the<br />
calibration problem in a stable manner, we first reformulate it as the problem of finding an<br />
expon<strong>en</strong>tial Lévy mo<strong>de</strong>l that has the smallest relative <strong>en</strong>tropy with respect to the prior among<br />
all solutions of the least squares calibration problem. This problem, called minimal <strong>en</strong>tropy<br />
least squares calibration, is still ill-posed, so we regularize it using the technique of minimal<br />
<strong>en</strong>tropy regularization, from the theory of ill-posed inverse problems.<br />
The third chapter discusses the numerical implem<strong>en</strong>tation of our calibration algorithm. To<br />
solve the calibration problem numerically, it is expressed in terms of the characteristic tripl<strong>et</strong>s<br />
of the prior and the solution and the Lévy measure of the prior is discr<strong>et</strong>ized on a uniform grid<br />
so that the calibration problem becomes finite-dim<strong>en</strong>sional.<br />
In the fourth chapter we first review the two available m<strong>et</strong>hods to mo<strong>de</strong>l <strong>de</strong>p<strong>en</strong><strong>de</strong>nce b<strong>et</strong>we<strong>en</strong><br />
compon<strong>en</strong>ts of Lévy processes. Un<strong>de</strong>rstanding the drawbacks of these m<strong>et</strong>hods allows to<br />
formulate the <strong>de</strong>sirable properties of a multidim<strong>en</strong>sional mo<strong>de</strong>lling approach. These properties<br />
lead us to <strong>de</strong>fining the notion of Lévy copula, first in the case of Lévy processes with only<br />
positive jumps in every compon<strong>en</strong>t and th<strong>en</strong> in the g<strong>en</strong>eral case. After proving a repres<strong>en</strong>tation<br />
theorem, which shows that Lévy copulas compl<strong>et</strong>ely characterize <strong>de</strong>p<strong>en</strong><strong>de</strong>nce structures of Lévy<br />
processes, we compute Lévy copulas that correspond to various special types of <strong>de</strong>p<strong>en</strong><strong>de</strong>nce.<br />
The fifth and the last chapter provi<strong>de</strong>s the tools necessary to apply Lévy copulas to finance.<br />
We first give m<strong>et</strong>hods to construct param<strong>et</strong>ric families of Lévy copulas and th<strong>en</strong> <strong>de</strong>velop an effici<strong>en</strong>t<br />
algorithm to simulate multidim<strong>en</strong>sional Lévy processes with <strong>de</strong>p<strong>en</strong><strong>de</strong>nce structures giv<strong>en</strong><br />
by Lévy copulas. The last section of this chapter contains an example of a multidim<strong>en</strong>sional<br />
expon<strong>en</strong>tial Lévy mo<strong>de</strong>l for option pricing, constructed using Lévy copulas.