L'approche dare pour une mesure de risque diversifiée

The objective of this paper is to provide a complete framework to aggregate different quantile and expectile models for obtaining more diversified Value-at-Risk and Expected Shortfall measures, by applying the diversification principle to model risk. Following Taylor [2008] and Gouriéroux and Jasiak [2008], we introduce a new class of models called Dynamic AutoRegressive Expectiles (dare). We first briefly present the main literature about VaR and es estimations, and we secondly explain the dare approach and how expectiles can be used to estimate quantile risk measures. We finally use the main validation tests to compare the dare approach to other traditional methods for computing extreme risk measures on the French stock market. Classification JEL : C14, C50, G11, G32.