Basel smashes VaR (May 2012)
Since early 1900 we have known that one or two big price movements empirically make the distribution highly non-gaussian (Ignored by most academics and blind followers). Based on this we know that no matter how good mathematical model you have you cannot accurately predict the tail probabilities. And these tails have BIG IMPACT. The solution is to focus on robustness in your hedge, reduced leverage etc. (there is much more to this, time will tell)
From the Basel (May 2012) report:
"Moving from value-at-risk to expected shortfall A number of weaknesses have been identified with using value-at-risk (VaR) for determining regulatory capital requirements, including its inability to capture “tail risk”. For this reason, the Committee has considered alternative risk metrics, in particular expected shortfall (ES). ES measures the riskiness of a position by considering both the size and the likelihood of losses above a certain confidence level. In other words, it is the expected value of those losses beyond a given confidence level. The Committee recognises that moving to ES could entail certain operational challenges; nonetheless it believes that these are outweighed by the benefits of replacing VaR with a measure that better captures tail risk. Accordingly, the Committee is proposing the use of ES for the internal models-based approach and also intends to determine risk weights for the standardised approach using an ES methodology."
"Shortcomings of the models-based approach: The metric used to capitalise trading book exposures was the 10-day value-at-risk (VaR) computed at the 99th percentile, one-tailed confidence interval. By construction, this is a measure aimed at capturing the risk of short-term fluctuations in market prices. While a 10-day VaR might be useful for day-to-day internal risk management purposes, it is questionable whether it meets the objectives of prudential regulation which seeks to ensure that banks have sufficient capital to survive low probability, or “tail”, events. Weaknesses identified with the 10-day VaR metric include: its inability to adequately capture credit risk; its inability to capture market liquidity risk; the provision of incentives for banks to take on tail risk; and, in some circumstances, the inadequate capture of basis risk. Perhaps more fundamentally, the models-based capital framework for market risk relied on a bank-specific perspective of risk, which might not be adequate from the perspective of the banking system as a whole. The pro-cyclicality of VaR-based capital charges based on recent historic data and the large number and size of backtesting exceptions observed during the crisis serve to highlight regulatory concerns with continued reliance on VaR."
See also
The Number That Killed Us: A Story of Modern Banking, Flawed Mathematics
I have not met that many naive VaR followers at the trading floor, but yes around in different corporations there are in particular plenty of management removed from the trading floor (taking the big decisions) , often with lack of knowledge about trading and market behaviour that on their desk wants to have simple risk numbers, giving them the whole picture, rather than "scary" worst case scenario think thank discussions with their senior traders. I guess they prefer VaR as a flawed risk benchmark rather than have to tell to the share holders they could go bust in this and that extreme scenario that happen every 20, 50, 100 year (?) (rather than every 2000 years as predicted by their naive VaR models). Or even better reduce the firms over all risk exposure, stop giving loans without any equity behind it etc. And yes there are plenty of academics naively thinking it is just to have the right mathematical model and they can calculate the exact probability for extreme events. Think again!
