Life after VaR

Abstract

Value at risk (VaR) is now universally used to measure and manage exposure to financial market risk. Yet its shortcomings are also well known, one of the most serious being that it is susceptible to being “gamed.” Since VaR focuses only on the loss at a single probability quantile, like 1%, it presents the opportunity for a tricky operator to inflate his returns by taking on unmeasured extreme risk exposure in the tail beyond the VaR cutoff, as the authors illustrate in the first part of this article. Use of the related concept of “conditional tail expectations” or “tail VaR” can mitigate this problem. But a second difficulty that is not addressed by tail VaR is that by focusing on a single horizon, the VaR methodology ignores how loss probabilities may evolve over the short run. For this, the authors offer “iterated conditional tail expectation,” a new dynamic risk measure that is based on repeated monitoring of tail VaR over the forecasting horizon.