A0250
Title: Automatic threshold selection for extreme value regression models
Authors: Julien Hambuckers - University of Liege (Belgium) [presenting]
Marie Kratz - ESSEC Business School, CREAR (France)
Antoine Usseglio-Carleve - Avignon Université (France)
Abstract: In finance, extreme value regression (EVR) has become a standard tool in the econometrician's toolbox to estimate and characterize risk measures in changing economic conditions. However, in this regression context, the threshold choice is a non-trivial task since it should also depend on the covariates and can have important consequences on the final estimates. We investigate this under-discussed issue and propose an efficient and robust solution to automatically estimate these thresholds with the help of the distributional regression machinery. We illustrate its properties through several simulation studies. The method is later applied to estimate hedge funds' tail risks, accounting for their heterogeneous investment strategies and the time-varying characteristics of financial markets.
