25.06.2013 Views

statistique, théorie et gestion de portefeuille - Docs at ISFA

statistique, théorie et gestion de portefeuille - Docs at ISFA

statistique, théorie et gestion de portefeuille - Docs at ISFA

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.

446 14. Gestion <strong>de</strong> Portefeuilles multimoments <strong>et</strong> équilibre <strong>de</strong> marché<br />

The first point to note is the difference b<strong>et</strong>ween currencies and stocks. For small as well as for large r<strong>et</strong>urns,<br />

the exponents c− and c+ for currencies (excepted Poland and Thailand) are all close to each other. Additional<br />

tests are required to establish wh<strong>et</strong>her their rel<strong>at</strong>ively small differences are st<strong>at</strong>istically significant. Similarly,<br />

the scale factors are also comparable. In contrast, many stocks exhibit a large asymm<strong>et</strong>ric behavior for large<br />

r<strong>et</strong>urns with c+ −c− 0.5 in about one-half of the investig<strong>at</strong>ed stocks. This means th<strong>at</strong> the tails of the large<br />

neg<strong>at</strong>ive r<strong>et</strong>urns (“crashes”) are often much f<strong>at</strong>ter than those of the large positive r<strong>et</strong>urns (“rallies”).<br />

The second important point is th<strong>at</strong>, for small r<strong>et</strong>urns, many stocks have an exponent 〈c+〉 ≈ 〈c−〉 2 and<br />

thus have a behavior not far from a pure Gaussian in the bulk of the distribution, while the average exponent<br />

for currencies is about 1.5 in the same “small r<strong>et</strong>urn” regime. Therefore, even for small r<strong>et</strong>urns, currencies<br />

exhibit a strong <strong>de</strong>parture from Gaussian behavior.<br />

In conclusion, this empirical study shows th<strong>at</strong> the modified Weibull param<strong>et</strong>eriz<strong>at</strong>ion, although not exact on<br />

the entire range of vari<strong>at</strong>ion of the r<strong>et</strong>urns X, remains consistent within each of the two regimes of small<br />

versus large r<strong>et</strong>urns, with a sharp transition b<strong>et</strong>ween them. It seems especially relevant in the tails of the<br />

r<strong>et</strong>urn distributions, on which we shall focus our <strong>at</strong>tention next.<br />

8 Cumulant expansion of the portfolio r<strong>et</strong>urn distribution<br />

8.1 link b<strong>et</strong>ween moments and cumulants<br />

Before <strong>de</strong>riving the main result of this section, we recall a standard rel<strong>at</strong>ion b<strong>et</strong>ween moments and cumulants<br />

th<strong>at</strong> we need below.<br />

The moments Mn of the distribution P are <strong>de</strong>fined by<br />

ˆP (k) =<br />

+∞<br />

n=0<br />

(ik) n<br />

where ˆ P is the characteristic function, i.e., the Fourier transform of P :<br />

Similarly, the cumulants Cn are given by<br />

Differenti<strong>at</strong>ing n times the equ<strong>at</strong>ion<br />

ln<br />

ˆP (k) =<br />

ˆP (k) = exp<br />

+∞<br />

n=0<br />

(ik) n<br />

+∞<br />

−∞<br />

n! Mn<br />

+∞<br />

n! Mn , (61)<br />

dS P (S)e ikS . (62)<br />

n=1<br />

<br />

=<br />

(ik) n<br />

n! Cn<br />

+∞<br />

n=1<br />

<br />

(ik) n<br />

. (63)<br />

n! Cn , (64)<br />

we obtain the following recurrence rel<strong>at</strong>ions b<strong>et</strong>ween the moments and the cumulants :<br />

Mn =<br />

n−1 <br />

<br />

n − 1<br />

MpCn−p ,<br />

p<br />

(65)<br />

p=0<br />

n−1 <br />

<br />

n − 1<br />

Cn = Mn −<br />

CpMn−p . (66)<br />

n − p<br />

p=1<br />

22

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

Saved successfully!

Ooh no, something went wrong!