Title: Realized estimators of tail risk measures
Authors: Giuseppe Storti - University of Salerno (Italy) [presenting]
Ostap Okhrin - Technical University Dresden (Germany)
Abstract: A novel estimation method for the quantiles and tail expectation of the distribution of financial returns that exploits information on realized higher order moments built from intra-daily data is proposed. Building on recent results on the joint elicitability of VaR and ES, our approach can be sees as a data driven generalization of standard asymptotic expansions such as the Cornish-Fisher one. The proposed procedure can be used to generate realized proxies of conditional VaR and ES as well as it can be extended to generate predictions of the future values of these risk measures. The empirical performance of the proposed methods will be assessed via Monte Carlo simulations and applications to real stock market data.