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

Create successful ePaper yourself

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

398 13. Gestion <strong>de</strong> <strong>portefeuille</strong> sous contraintes <strong>de</strong> capital économique<br />

2.3 Portfolio wealth un<strong>de</strong>r the Gaussian copula hypothesis<br />

2.3.1 Deriv<strong>at</strong>ion of the multivari<strong>at</strong>e distribution with a Gaussian copula and modified Weibull margins<br />

An advantage of the class of modified Weibull distributions (7) is th<strong>at</strong> the transform<strong>at</strong>ion into a Gaussian,<br />

and thus the calcul<strong>at</strong>ion of the vector y introduced in <strong>de</strong>finition 1, is particularly simple. It takes the form<br />

yk = sgn(xk) √ <br />

|xk|<br />

2<br />

χk<br />

c k 2<br />

, (38)<br />

where yk is normally distributed . These variables Yi then allow us to obtain the covariance m<strong>at</strong>rix V of the<br />

Gaussian copula :<br />

<br />

|xi|<br />

Vij = 2 · E sgn(xixj)<br />

χi<br />

c i<br />

2 |xj|<br />

χj<br />

cj <br />

2<br />

, (39)<br />

which always exists and can be efficiently estim<strong>at</strong>ed. The multivari<strong>at</strong>e <strong>de</strong>nsity P (x) is thus given by:<br />

P (x1, · · · , xN) = cV (x1, x2, · · · , xN)<br />

=<br />

1<br />

2 N π N/2√ V<br />

N<br />

pi(xi) (40)<br />

i=1<br />

N ci|xi| c/2−1<br />

⎡<br />

exp ⎣− <br />

i=1<br />

χ c/2<br />

i<br />

i,j<br />

V −1<br />

ij<br />

Obviously, similar transforms hold, mut<strong>at</strong>is mutandis, for the asymm<strong>et</strong>ric case (8,9).<br />

⎤<br />

c/2 c/2 |xi| |xj|<br />

⎦ . (41)<br />

2.3.2 Asymptotic distribution of a sum of modified Weibull variables with the same exponent c > 1<br />

We now consi<strong>de</strong>r a portfolio ma<strong>de</strong> of <strong>de</strong>pen<strong>de</strong>nt ass<strong>et</strong>s with pdf given by equ<strong>at</strong>ion (41) or its asymm<strong>et</strong>ric<br />

generaliz<strong>at</strong>ion. For such distributions of ass<strong>et</strong> r<strong>et</strong>urns, we obtain the following result<br />

THEOREM 5 (TAIL EQUIVALENCE FOR A SUM OF DEPENDENT RANDOM VARIABLES)<br />

L<strong>et</strong> X1, X2, · · · , XN be N random variables with a <strong>de</strong>pen<strong>de</strong>nce structure <strong>de</strong>scribed by the Gaussian copula<br />

with correl<strong>at</strong>ion m<strong>at</strong>rix V and such th<strong>at</strong> each Xi ∼ W(c, χi). L<strong>et</strong> w1, w2, · · · , wN be N (positive) nonrandom<br />

real coefficients. Then, the variable<br />

SN = w1X1 + w2X2 + · · · + wNXN<br />

is equivalent in the upper and the lower tail to Z ∼ W(c, ˆχ) with<br />

ˆχ =<br />

<br />

where the σi’s are the unique (positive) solution of<br />

<br />

<br />

i<br />

i<br />

wiχiσi<br />

c−1<br />

c<br />

χi<br />

χj<br />

(42)<br />

, (43)<br />

V −1<br />

ik σi c/2 = wkχk σ 1−c/2<br />

k , ∀k . (44)<br />

10

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

Saved successfully!

Ooh no, something went wrong!