Arturo Leccadito (University of Calabria)

will give a presentation on

Predictive identification robust confidence sets with application to tail risk measures

Abstract:

This work proposes a novel method for constructing confidence sets for predictions in parametric or semi-parametric models, that are valid for out-of-sample inference and are identification-robust. This procedure involves two steps: first, a simultaneous robust confidence set for the model’s parameters that are hard to identify is constructed through the inversion of an out-of-sample goodness of fit test; second, this set is projected to construct a simultaneous confidence set for predictions. We focus on financial risk measures, specifically on Value at Risk and Expected Shortfall, for which an illustrative example on GARCH returns is provided, along with Monte Carlo simulations for coverage levels. Finally, an application of the proposed method is applied to the Technology Select Sector SPDR Fund’s returns.