Value at risk or VaR has a bad name. But much of the problem is how we think about and how we talk about VaR. Words do matter and using the wrong terms and phrases often takes us down the wrong road.
The Financial Times, in talking about changes in Morgan Stanley’s VaR model, says that under their VaR model “[Morgan Stanley] expects to lose no more than $63m in a single trading day, within a certain probability”. (“M Stanley shows the ‘flaky’ side of model”, 18 October 2012) This is a common characterization of VaR but the phrase “expect to lose no more” is horribly misleading. A far more instructive way to say the same thing would be “the bank expects to lose more than $63m in a single trading day only infrequently (roughly one day out of 20)”.
This simple change in phrasing is a big improvement and points us in the right direction. First, it emphasizes that Morgan Stanley does expect to lose more than $63m, just not very often. Second, it pushes all of us (regulators, investors, managers) to think about the consequences of losing more than $63m. How much more? How often? How well can Morgan Stanley withstand losses of $63m? The world is an uncertain place and we can never know the maximum loss a firm might suffer. We all need to think carefully about those times when losses are more than the VaR.
This simple change in emphasis – talking about VaR “not as a ‘worst case,’ but rather as a regularly occurring event with which we should be comfortable” (to use Bob Litterman’s words) – goes far towards reminding us all that the proper role of VaR and quantitative risk tools is to inform and educate us about the uncertainty and randomness inherent in our world. There is no certainty in our world, no certainty that a bank will lose no more than $63m. The future is random and contingent and we need to embrace this uncertainty rather than obscure it.
(See my prior post – Rights and Wrongs of VaR – for more detail about how to think about VaR.)