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.

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

B Generalized efficient frontier and two funds separ<strong>at</strong>ion theorem<br />

L<strong>et</strong> us consi<strong>de</strong>r a s<strong>et</strong> of N risky ass<strong>et</strong>s X1, · · · , XN and a risk-free ass<strong>et</strong> X0. The problem is to find the<br />

optimal alloc<strong>at</strong>ion of these ass<strong>et</strong>s in the following sense:<br />

⎧<br />

⎨<br />

⎩<br />

inf wi∈[0,1] ρα({wi})<br />

<br />

i≥0 wi = 1<br />

<br />

i≥0 wiµ(i) = µ ,<br />

In other words, we search for the portfolio P with minimum risk as measured by any risk measure ρα<br />

obeying axioms I-IV of section 2 for a given amount of expected r<strong>et</strong>urn µ and normalized weights wi.<br />

Short-sells are forbid<strong>de</strong>n except for the risk-free ass<strong>et</strong> which can be lent and borrowed <strong>at</strong> the same interest<br />

r<strong>at</strong>e µ0. Thus, the weights wi’s are assumed positive for all i ≥ 1.<br />

B.1 Case of in<strong>de</strong>pen<strong>de</strong>nt ass<strong>et</strong>s when the risk is measured by the cumulants<br />

To start with a simple example, l<strong>et</strong> us assume th<strong>at</strong> the risky ass<strong>et</strong>s are in<strong>de</strong>pen<strong>de</strong>nt and th<strong>at</strong> we choose to<br />

measure the risk with the cumulants of their distributions of r<strong>et</strong>urns. The case when the ass<strong>et</strong>s are <strong>de</strong>pen<strong>de</strong>nt<br />

and/or when the risk is measured by any ρα will be consi<strong>de</strong>red l<strong>at</strong>er. Since the ass<strong>et</strong>s are assumed<br />

in<strong>de</strong>pen<strong>de</strong>nt, the cumulant of or<strong>de</strong>r n of the pdf of r<strong>et</strong>urns of the portfolio is simply given by<br />

Cn =<br />

(90)<br />

N<br />

wi n Cn(i), (91)<br />

i=1<br />

where Cn(i) <strong>de</strong>notes the marginal nth or<strong>de</strong>r cumulant of the pdf of r<strong>et</strong>urns of the ass<strong>et</strong> i. In or<strong>de</strong>r to solve<br />

this problem, l<strong>et</strong> us introduce the Lagrangian<br />

<br />

N<br />

<br />

N<br />

<br />

L = Cn − λ1 wi µ(i) − µ − λ2 wi − 1 , (92)<br />

i=0<br />

where λ1 and λ2 are two Lagrange multipliers. Differenti<strong>at</strong>ing with respect to w0 yields<br />

i=0<br />

λ2 = µ0 λ1, (93)<br />

which by substitution in equ<strong>at</strong>ion (92) gives<br />

<br />

N<br />

<br />

L = Cn − λ1 wi (µ(i) − µ0) − (µ − µ0) . (94)<br />

i=1<br />

L<strong>et</strong> us now differenti<strong>at</strong>e L with respect to wi, i ≥ 1, we obtain<br />

so th<strong>at</strong><br />

n w ∗ i n−1 Cn(i) − λ1(µ(i) − µ0) = 0, (95)<br />

w ∗ 1<br />

i = λ1<br />

n−1<br />

Applying the normaliz<strong>at</strong>ion constraint yields<br />

1<br />

w0 + λ1<br />

n−1<br />

N<br />

i=1<br />

1<br />

µ(i) − µ0<br />

n−1<br />

. (96)<br />

n Cn(i)<br />

1<br />

µ(i) − µ0<br />

n−1<br />

n Cn(i)<br />

30<br />

= 1, (97)

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

Saved successfully!

Ooh no, something went wrong!