Bootstrap inference for conditional risk measures

Abstract

Risk measures play a key role in financial risk management and are enforced by current legislation to protect financial stability. In particular Value-at-Risk (VaR) and Expected Shortfall (ES) are used to assess the market risks associated with financial assets. These risk measures are frequently applied conditionally to account for the temporal dependence of financial data. To quantify the uncertainty induced by parameter estimation, practitioners often construct confidence intervals by resorting to resampling methods. This thesis provides a theoretical justification for intervals constructed for conditional risk measures. New resampling methods are proposed and validated to quantify the parameter uncertainty around the conditional VaR and ES estimates.