Title: Penalized expectiles optimal portfolios
Authors: Gabriele Torri - University of Bergamo (Italy) [presenting]
Rosella Giacometti - Università di Bergamo (Italy)
Abstract: Expectiles are risk measures increasingly popular in recent years among academics and practitioners, thanks to their good theoretical properties: they are the only risk measure that is both coherent and elicitable. Moreover, they have an intuitive economic explanation and interesting algebraic connections can be established with Value at Risk (VaR) and Expected Shortfall (ES). Recent works explored their usage in portfolio optimization, showing how to build optimal risk-return portfolios using expectiles as risk measure. However, real-world application to portfolios with a large number of assets are limited by estimation error, that typically leads to bad out of sample performances. We propose a novel derivation of the linear programming formulation of the minimum EVaR portfolio, similar to the ones available in the literature, but computationally faster and characterized by a straightforward economic interpretation. We also introduce a ridge penalization to the portfolio weights in order to improve the finite sample performances, and we test the model on a variety of datasets.