Estimation of the expected shortfall given an extreme component under conditional extreme value model (Q2417999)

The authors consider the problem of estimating the Marginal Expected Shortfall (MES) for a class of bivariate distributions with \textit{asymptotic independence}. Let \(X,Y\) be nonnegative random variables (risk factors) with distribution functions \(F_{X},F_{Y}\), respectively. The MES is defined as \[ \theta \left(p \right) = \mathbb{E}\left[Y\mid X > F^{\gets}_{X}\left(1-p \right) \right], \] where \(F^{\gets}_{X}\) is the left-continuous inverse of \(F_{X}\) and \(p\) is a \textit{small} probability. In this paper the vector \(\left(X,Y \right) \) fulfills a Conditional Extreme Value assumption (Assumption 1). Assumption 1 implies that the random variable \(X\) is regularly varying. Note that there is no general relationship between the bivariate regular variation and Assumption 1. The authors construct an estimator of the marginal expected shortfall \(\theta \left(p \right)\) and establish its consistency and asymptotic normality. Next, simulation studies and real financial data analysis are presented. The paper is concluded with some open problems and comments.