Title: Tail risks in vast portfolio selection: A comparison of penalized quantile versus expectile models
Authors: Rosella Giacometti - Università di Bergamo (Italy) [presenting]
Gabriele Torri - University of Bergamo (Italy)
Sandra Paterlini - University of Trento (Italy)
Abstract: Estimating in an accurate way, and optimally controlling tail risk, is of utmost importance for building portfolios with desirable properties for investors, especially in presence of a large set of assets. In recent years, the financial literature witnessed an increase of interest towards expectiles as an alternative to more common risk measures such as Value at Risk (VaR) and Expected Shortfall (ES). Such a family of measures has good theoretical properties (it is the only risk measure that is both coherent and elicitable), and has a relevant financial interpretation (it can be thought as the amount of money that should be added to a position in order to have a sufficiently high gain-loss ratio). We combine expectile and quantile approaches with regularization to build optimal portfolio models, with the aim of providing parsimonious and robust portfolios with better out-of-sample performances. Simulation and real-world analysis allow us to critically discuss pros and cons of the proposed methods when compared to state-of-art benchmarks.