Backtesting

Abstract

This chapter surveys methods for backtesting risk models using the ex ante riskmeasure forecasts from the model and the ex post realized portfolio profit or loss. The risk measure forecast can take the form of a VaR, an Expected Shortfall, or a distribution forecast. The backtesting methods surveyed in this chapter can be seen as a final diagnostic check on the aggregate risk model carried out by the risk management team that constructed the risk model, or they can be used by external model-evaluators such as bank supervisors. Common for the approaches suggested is that they only require information on the daily ex ante risk model forecast and the daily ex post corresponding profit and loss. In particular, knowledge about the assumptions behind the risk model and its construction is not required.

Keywords: Value-at-Risk, expected shortfall, distribution, forecasting, model evaluation, testing, historical simulation

JEL Classification: G20