B1256
Title: Extreme expectile estimation for heavy-tailed time series
Authors: Simone Padoan - Bocconi University (Italy) [presenting]
Gilles Stupfler - University of Angers (France)
Abstract: Extreme quantiles estimation have been widely discussed in the literature, since that for instance the value at risk is an important ``measure'' for the risk quantification in many applied fields such as in finance and insurance. In finance, real-life time series often reveal a certain degree of dependence through time, therefore, in the last years, a big effort has been devoted in deriving inferential methods for extreme quantiles that accommodate for the temporal dependence. The tail expectile is a coherent risk measure, therefore it provides an appealing alternative to the value at risk. Recently, several estimation results for extreme expectiles are emerging, under the assumption that the data are independent. We investigate the behaviour (asymptotic properties and finite sample performances) of some expectiles estimators based on large observations of stationary time series, under mild assumptions on the serial dependence.
