Return Prediction and Portfolio Selection: A Distributional Approach
International Review of Economics & Finance 27: 209-223, 2011
34 Pages Posted: 25 Nov 2011 Last revised: 27 Sep 2017
Date Written: July 1, 2011
The inquiries to return predictability are traditionally limited to the first two moments, mean and volatility. Analogously, literature on portfolio selection also stems from a moment-based analysis with up to the fourth moment being considered. This paper develops a distribution-based framework for both return prediction and portfolio selection. More specifically, a time-varying return distribution is modeled through quantile regression and copulas, using the quantile approach to extract information in marginal distributions and copulas to capture dependence structure. A nonlinear utility function is proposed for portfolio selection which utilizes the full underlying return distribution. An empirical application to US data highlights not only the predictability of the stock and bond return distributions, but also the additional information provided by the distributional approach which cannot be captured by the traditional moment-based methods.
Keywords: Return distribution, Quantile regression, Copula, Heuristic Optimization, Portfolio selection
JEL Classification: C16, C53, C61, G11
Suggested Citation: Suggested Citation