Chapter 9 - Retrospective Assessment of Value at Risk

Abstract

This chapter offers a retrospective on value at risk (VaR) from two different but closely related points of view.. First, it addresses the question as to how good the VaR models are, or how prone the VaR models are to model risk. The chapter addresses this issue by focusing at some evidence of the extent of model risk in commercial VaR models. The term “model risk” encompasses the possibility of errors from a wide variety of sources. This includes the possibility of an incorrect model, but it can also include incorrect calibration, problems arising from the diverse ways in which models are implemented, data and programming problems, and even problems arising from traders and asset managers reacting to the ways in which VaR forecasts are used within an institution. The chapter also highlights some of the many sources of model risk in such models. It also discusses the way in which one can evaluate the forecasts of a VaR model. This raises the issue of back-testing. Back testing is the application of quantitative methods to determine whether the forecasts of a risk forecasting model are consistent with the assumptions on which the model is based, or to rank a group of such models against each other. The chapter discusses some of the principal approaches to back-testing..