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
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
426 14. Gestion <strong>de</strong> Portefeuilles multimoments <strong>et</strong> équilibre <strong>de</strong> marché<br />
entangled with the Mean-Variance Portfolio Mo<strong>de</strong>l. In<strong>de</strong>ed both of them fundamentally rely on the <strong>de</strong>scription<br />
of the probability distribution function (pdf) of ass<strong>et</strong> r<strong>et</strong>urns in terms of Gaussian functions. The<br />
Mean-Variance <strong>de</strong>scription is thus <strong>at</strong> the basis of Markovitz’s portfolio theory (Markovitz 1959) and of the<br />
CAPM (see for instance (Merton 1990)).<br />
Otherwise, the d<strong>et</strong>ermin<strong>at</strong>ion of the risks and r<strong>et</strong>urns associ<strong>at</strong>ed with a given portfolio constituted of N<br />
ass<strong>et</strong>s is compl<strong>et</strong>ely embed<strong>de</strong>d in the knowledge of their multivari<strong>at</strong>e distribution of r<strong>et</strong>urns. In<strong>de</strong>ed, the<br />
<strong>de</strong>pen<strong>de</strong>nce b<strong>et</strong>ween random variables is compl<strong>et</strong>ely <strong>de</strong>scribed by their joint distribution. This remark<br />
entails the two major problems of portfolio theory: 1) d<strong>et</strong>ermine the multivari<strong>at</strong>e distribution function of<br />
ass<strong>et</strong> r<strong>et</strong>urns; 2) <strong>de</strong>rive from it useful measures of portfolio risks and use them to analyze and optimize<br />
portfolios.<br />
The variance (or vol<strong>at</strong>ility) of portfolio r<strong>et</strong>urns provi<strong>de</strong>s the simplest way to quantify its fluctu<strong>at</strong>ions and<br />
is <strong>at</strong> the fund<strong>at</strong>ion of the (Markovitz 1959)’s portfolio selection theory. Non<strong>et</strong>heless, the variance of a<br />
portfolio offers only a limited quantific<strong>at</strong>ion of incurred risks (in terms of fluctu<strong>at</strong>ions), as the empirical<br />
distributions of r<strong>et</strong>urns have “f<strong>at</strong> tails” (Lux 1996, Gopikrishnan <strong>et</strong> al. 1998, among many others) and the<br />
<strong>de</strong>pen<strong>de</strong>nces b<strong>et</strong>ween ass<strong>et</strong>s are only imperfectly accounted for by the covariance m<strong>at</strong>rix (Litterman and<br />
Winkelmann 1998). It is thus essential to extend portfolio theory and the CAPM to tackle these empirical<br />
facts.<br />
The Value-<strong>at</strong>-Risk (Jorion 1997) and many other measures of risks (Artzner <strong>et</strong> al. 1997, Sorn<strong>et</strong>te 1998,<br />
Artzner <strong>et</strong> al. 1999, Bouchaud <strong>et</strong> al. 1998, Sorn<strong>et</strong>te <strong>et</strong> al. 2000b) have then been <strong>de</strong>veloped to account for the<br />
larger moves allowed by non-Gaussian distributions and non-linear correl<strong>at</strong>ions but they mainly allow for the<br />
assessment of down-si<strong>de</strong> risks. Here, we consi<strong>de</strong>r both-si<strong>de</strong> risk and <strong>de</strong>fine general measures of fluctu<strong>at</strong>ions.<br />
It is the first goal of this article. In<strong>de</strong>ed, consi<strong>de</strong>ring the minimum s<strong>et</strong> of properties a fluctu<strong>at</strong>ion measure<br />
must fulfil, we characterize these measures. In particular, we show th<strong>at</strong> any absolute central moments and<br />
some cumulants s<strong>at</strong>isfy these requirement as well as do any combin<strong>at</strong>ion of these quantities. Moreover,<br />
the weights involved in these combin<strong>at</strong>ions can be interpr<strong>et</strong>ed in terms of the portfolio manager’s aversion<br />
against large fluctu<strong>at</strong>ions.<br />
Once the <strong>de</strong>finition of the fluctu<strong>at</strong>ion measures have been s<strong>et</strong>, it is possible to classify the ass<strong>et</strong>s and portfolios<br />
using for instance a risk adjustment m<strong>et</strong>hod (Sharpe 1994, Dowd 2000) and to <strong>de</strong>velop a portfolio<br />
selection and optimiz<strong>at</strong>ion approach. It is the second goal of this article.<br />
Then a new mo<strong>de</strong>l of mark<strong>et</strong> equilibrium can be <strong>de</strong>rived, which generalizes the usual Capital Ass<strong>et</strong> Pricing<br />
Mo<strong>de</strong>l (CAPM). This is the third goal of our paper. This improvement is necessary since, although the use<br />
of the CAPM is still wi<strong>de</strong>ly spread, its empirical justific<strong>at</strong>ion has been found less and less convincing in the<br />
past years (Lim 1989, Harvey and Siddique 2000).<br />
The last goal of this article is to present an efficient param<strong>et</strong>ric m<strong>et</strong>hod allowing for the estim<strong>at</strong>ion of the<br />
centered moments and cumulants, based upon a maximum entropy principle. This param<strong>et</strong>eriz<strong>at</strong>ion of<br />
the problem is necessary in or<strong>de</strong>r to obtain accur<strong>at</strong>e estim<strong>at</strong>es of the high or<strong>de</strong>r moment-based quantities<br />
involved the portfolio optimiz<strong>at</strong>ion problem with our generalized measures of fluctu<strong>at</strong>ions.<br />
The paper is organized as follows.<br />
Section 2 presents a new s<strong>et</strong> of consistent measures of risks, in terms of the semi-invariants of pdf’s, such as<br />
the centered moments and the cumulants of the portfolio distribution of r<strong>et</strong>urns, for example.<br />
Section 3 <strong>de</strong>rives the generalized efficient frontiers, based on these novel measures of risks. Both cases with<br />
and without risk-free ass<strong>et</strong> are analyzed.<br />
Section 4 offers a generaliz<strong>at</strong>ion of the Sharpe r<strong>at</strong>io and thus provi<strong>de</strong>s new tools to classify ass<strong>et</strong>s with<br />
respect to their risk adjusted performance. In particular, we show th<strong>at</strong> this classific<strong>at</strong>ion may <strong>de</strong>pend on the<br />
2