Evaluation of Value-at-Risk Models Using Historical Data

Recent studies have underscored the need for market participants to develop reliable methods of measuring risk. One increasingly popular technique is the use of "value-at-risk" models, which convey estimates of market risk for an entire portfolio in one number. The author explores how well these models actually perform by applying twelve value-at-risk approaches to 1,000 randomly chosen foreign exchange portfolios. Using nine criteria to evaluate model performance, he finds that the approaches generally capture the risk that they set out to assess and tend to produce risk estimates that are similar in average size. No approach, however, appears to be superior by every measure.