Glossary
Value at risk (VaR)
An estimate of the loss a portfolio should not exceed over a set period at a given confidence level, such as 95% over one day.
Value at risk (VaR) is a single number that summarizes how much a portfolio might lose in normal conditions. It has three parts: a time period, a confidence level and a loss amount. "One-day 95% VaR of $20,000" means that on 95% of days the loss is expected to be smaller than $20,000, and on 5% of days, about one trading day in twenty, larger.
Three ways to calculate it
- Parametric. Assume returns follow a bell curve and use the portfolio's standard deviation.
- Historical. Take the actual returns of the past, say, 500 days and find the loss exceeded on 5% of them.
- Monte Carlo. Simulate thousands of possible outcomes from a model.
What it is used for
Banks and funds use VaR to set trading limits, compare risk across desks holding different assets, and calculate regulatory capital. Its appeal is that it turns stocks, bonds, currencies and options into one comparable figure.
What it does not tell you
- How bad the bad days are. VaR marks the edge of the worst 5%, and is silent on what lies beyond. A 95% VaR of $20,000 is consistent with a worst case of $25,000 or $2 million. Expected shortfall (also called CVaR), the average loss on the days beyond VaR, addresses this and has replaced VaR in bank capital rules.
- Fat tails. Real markets produce extreme moves far more often than a bell curve allows. See black swan.
- Changing conditions. VaR based on a calm period understates risk just before turbulence. Correlations also rise in a crisis, removing diversification the model counted on.
- Liquidity. It assumes positions can be sold at market prices.
In 2007 and 2008, several banks reported losses many times their stated VaR on many days in a row. VaR is a measure of ordinary risk, and should be read alongside stress tests of extraordinary events.
