Robust value-at-risk: an information-theoretic approach

Abstract

We present a robust value-at-risk model that takes into account the possibility of model misspecification. In place of a single prior distribution, we utilize multiple priors in the form of an 'uncertainty set' around the estimated expected returns and covariance matrix, constructed using the information-theoretic notion of Kullback-Leibler divergence. An extension to conditional value-at-risk is also specified.