Forecasting 10-Day-Ahead Portfolio Value-at-Risk and Expected Shortfall: Univariate vs. Multivariate Approaches

Work in Progress

Basel regulations require banks to forecast 10-day-ahead Value-at-Risk (VaR) and Expected Shortfall (ES) for their trading portfolios.

We compare a wide range of volatility models, including multivariate and copula-based approaches, evaluating predictive accuracy via the joint loss function of Fissler and Ziegel (2016) as well as formal backtesting procedures.

Simple univariate models applied directly to portfolio returns perform on par with, or better than, multivariate approaches modeled at the asset level.