What Usain Bolt can teach banks about financial risk
By A.P.
THE banking industry did a bad job in the run-up to the financial crisis of assessing “tail risks”, extreme events that represent the least likely of a range of probable outcomes. The Basel Committee on Banking Supervision, which is the international standard-setter for bank capital, has proposed changes in the internal risk models that financial institutions use. In particular, it wants banks to shift from a technique called “value-at-risk” (VaR) to one called “expected shortfall” (ES).
VaR is a way of measuring a firm’s risk of suffering really big losses over a certain period (a day, a week, a month) to a certain level of “confidence”. A daily VaR of $1m at 1% probability means that there is a 99% chance that you will not lose more than $1m on any one day. The problem is that if you have that one bad day in 100, the potential losses could go much higher than $1m. VaR doesn’t have much to say about what those losses might be. The expected-shortfall approach is meant to provide an answer to that question. Instead of asking, “What are the chances that things get so bad that we lose $1m?” it asks, “If things do get that bad, how much would we actually lose?”
To do this, it uses a statistical method called “extreme value theory”, which looks specifically at what happens in the tail of distributions. To take a more trivial example of where extreme-value theory has been used, a 2011 paper by two researchers at Tilburg University collected data on the personal bests of elite athletes between 1991 and 2008, in order to try and calculate the “ultimate world record” for 100m sprints—the absolute edge of human performance given the times, equipment and drugs-policies that then prevailed. For the 100m for men, the boffins put the ultimate world record at 9.51 seconds, compared with the record that then prevailed of 9.72, and a current world best of 9.58, set by Usain Bolt in 2009. That looks pretty good: the model came up with a number that was well inside the mark that then prevailed, and is still a hefty improvement on the current record. If extreme-value theory is meant to help banks think through the extremes, this is encouraging.
But like every model in history, expected shortfall cannot predict the future. In an earlier 2006 paper, researchers from the same university tried to calculate ultimate world records for a wider range of events, including the men’s marathon. The researchers reckoned back in 2006 that the best possible running of that distance would yield a time of two hours, four minutes and six seconds. Yet the world record today stands at two hours, three minutes and 23 seconds (Wilson Kipsang in 2013). To be fair to the researchers, they did not claim that their ultimate record could not be broken. But whether bankers will remember that reality can be worse than expected is a different question. Expected shortfall is an improvement on VaR; it is not a crystal ball.
