VaR may assist in creating a bubble
Mostly used within the trading environments at banks, Value-at-Risk (VaR) is a risk measure that some may say provides the maximum loss scenario given a level of confidence and amount of days to liquidation known as the ‘horizon period’. It actually gives you an answer of what’s the most you could possible lose within a reasonable bound:
There are three main approaches to calculating VaR: the Variance-Covariance approach, Historical Simulation and thirdly, through Monte Carlo simulation. My focus is on the widely used Historical VaR.
The other day (some month(s) ago) a fellow by the name of Julian Gough started tweeting about this market risk measure, describing it as “a guy sitting on the lip of a volcano that last exploded in 2008. And, each year, he says “This volcano hasn’t exploded for 5 years… 6 years… 7… It’s getting safer and safer!” The most beautiful description of Historical VaR I’ve come across. The volcano doesn’t get safer as time elapses, it’s the contrary, and we should take a hard look at the bubbles it creates.
Let’s start with what’s nice about Historical VaR:
The computations can be performed quickly depending on the size of the portfolio
It’s relatively easy to compute because the method is fairly easy to understand and easy to explain to senior management.
You don’t have to make distributional assumptions
Full correlation exists amongst all risk factors so they can all be stressed at the same time
The heavy downside:
The assumption that past performance dictates and reflects future results is problematic! If atypical events occur, VaR is distorted.
Because a limited amount of scenarios are used, the VaR is sensitive to a relatively small number of results.
The use of proxies when dealing with new instruments with no history.
So if nothing crazy happened yesterday, it’s more likely that today and tomorrow we’ll have good, sunny days. As the days go on, we are able to take on more risk because we’re of the idea that std dev (σ) is decreasing. With the assumption that volatility (or riskiness) is decreasing, we buy more and more, and so, prices are inflated.
I am inclined to say that it may be the best to not have VaR at all, it can be very misleading and very wrong. But, we do need a level of comfort that we are on the right track.
So, know the risks. The risk manager should not use VaR as their end all even if they backtest, use conditional VaR or implement a weighting scheme.
