[Paper Review] Multi-Moments Method for Portfolio Management: Generalized Capital Asset Pricing Model in Homogeneous and Heterogeneous markets

Yannick Malevergne, Didier Sornette|ArXiv.org|Jul 19, 2002

Financial Risk and Volatility ModelingEconomics, Econometrics and Finance29 references3 citations

TL;DR

This paper introduces a generalized Capital Asset Pricing Model (CAPM) using multi-moment risk measures based on cumulants and semi-invariants to capture tail risks beyond variance. It derives exact formulas for portfolio moments and cumulants under a flexible Weibull-based parametric model, enabling analytical optimization that can simultaneously increase expected return and reduce large risks—'having your cake and eating it too'—in both homogeneous and heterogeneous markets.

ABSTRACT

We introduce a new set of consistent measures of risks, in terms of the semi-invariants of pdf's, such that the centered moments and the cumulants of the portfolio distribution of returns that put more emphasis on the tail the distributions. We derive generalized efficient frontiers, based on these novel measures of risks and present the generalized CAPM, both in the cases of homogeneous and heterogeneous markets. Then, using a family of modified Weibull distributions, encompassing both sub-exponentials and super-exponentials, to parameterize the marginal distributions of asset returns and their natural multivariate generalizations, we offer exact formulas for the moments and cumulants of the distribution of returns of a portfolio made of an arbitrary composition of these assets. Using combinatorial and hypergeometric functions, we are in particular able to extend previous results to the case where the exponents of the Weibull distributions are different from asset to asset and in the presence of dependence between assets. In this parameterization, we treat in details the problem of risk minimization using the cumulants as measures of risks for a portfolio made of two assets and compare the theoretical predictions with direct empirical data. Our extended formulas enable us to determine analytically the conditions under which it is possible to ``have your cake and eat it too'', i.e., to construct a portfolio with both larger return and smaller ``large risks''.

Motivation & Objective

• To develop a generalized CAPM that extends beyond the traditional mean-variance framework by incorporating higher-order risk measures.
• To address the limitations of Gaussian assumptions in portfolio theory by using cumulants and semi-invariants as risk measures that emphasize tail events.
• To provide an analytically tractable parametric model for portfolio return distributions using modified Weibull distributions with arbitrary shape parameters and dependence.
• To derive conditions under which portfolios can achieve both higher expected return and lower large risks, challenging the traditional risk-return trade-off.
• To enable accurate estimation of high-order moments via a maximum entropy parameterization for practical portfolio optimization.

Proposed method

• Uses cumulants and absolute central moments as consistent, risk-averse measures of portfolio fluctuations, generalizing variance-based risk.
• Employs a family of modified Weibull distributions to model marginal asset return distributions, allowing for sub- and super-exponential tails.
• Applies multivariate generalizations of Weibull distributions to model dependent asset returns with arbitrary shape parameters across assets.
• Derives exact analytical expressions for portfolio cumulants and moments using combinatorial identities and hypergeometric functions, even when exponents differ across assets.
• Uses Lagrange multipliers to solve for optimal portfolio weights that minimize higher-order cumulants, under constraints on portfolio weights summing to one.
• Applies the maximum entropy principle to parameterize the return distribution, enabling robust estimation of high-order moments for optimization.

Experimental results

Research questions

• RQ1 Can a portfolio be constructed such that both expected return increases and large risks decrease simultaneously, defying the traditional risk-return trade-off?
• RQ2 What are the analytical conditions under which higher-order cumulants (beyond variance) can be minimized while maximizing expected return?
• RQ3 How can the joint distribution of dependent asset returns be parameterized to allow exact computation of portfolio moments and cumulants?
• RQ4 What is the role of different tail behaviors (sub- and super-exponential) in shaping optimal portfolio allocations under multi-moment risk measures?
• RQ5 How does the generalized CAPM differ from the classical CAPM when higher-order moments are used as risk measures in both homogeneous and heterogeneous markets?

Key findings

• The paper derives exact analytical formulas for the cumulants of a portfolio return distribution using hypergeometric functions, even when the Weibull shape parameters differ across assets.
• For a two-asset portfolio, the optimal weight minimizing the nth cumulant is given by $ w^* = \frac{C_n(2)^{1/(n-1)}}{C_n(1)^{1/(n-1)} + C_n(2)^{1/(n-1)}} $, enabling explicit risk-return trade-off analysis.
• The expected return of the cumulant-minimizing portfolio, $ \mu_n^* $, can be expressed as a weighted average of individual asset returns, with weights proportional to the $ (n-1) $th root of their cumulants.
• The condition $ \mu_n^* < \mu_{n+k}^* $ holds when $ (\mu(1)-\mu(2)) \left( C_n(1)^{1/(n-1)} C_{n+k}(2)^{1/(n+k-1)} - C_n(2)^{1/(n-1)} C_{n+k}(1)^{1/(n+k-1)} \right) > 0 $, proving that higher moments can yield better risk-return combinations.
• In the general N-asset case, optimal weights are $ w_i = \frac{\prod_{j \neq i} C_n(j)^{1/(n-1)}}{\sum_j C_n(j)^{1/(n-1)}} $, showing a power-weighting rule based on cumulant magnitudes.
• The model demonstrates analytically that it is possible to 'have your cake and eat it too'—simultaneously achieving higher return and lower large risks—under specific conditions on the higher-order moments of asset returns.