Backtesting VaR and ES under the magnifying glass

Backtesting provides the means of determining the accuracy of risk forecasts and the corresponding risk model. Given that the actual return generating process is unknown, the evaluation methods rely on various assumptions in order to quantify the models inefficiencies and proceed with the model evaluation. These method specific assumptions, in conjunction with the regulatory policies can introduce distortions in the evaluation process, which affect the reliability of the evaluation results. To investigate such effects from a practitioner's perspective, this paper reviews the major Value at Risk and Expected Shortfall forecast evaluation methods and evaluates their performance under a common simulation and financial application framework. Our findings suggest that focusing on specific individual hypothesis tests provides a more reliable alternative than the corresponding conditional coverage ones. In addition, selecting a two-year out-of-sample period provides a significantly better power to relevance ratio than the more relevant but powerless regulatory one-year specification.