The abundance of risk metrics stems from the effort to measure the difference between
the expected and actual returns, under a hypothesis of normality. Under the assumption
of risk aversion, investors are likely to quantify risk using metrics which measure
returns lower than the expected average. These include the semi-variance of returns
smaller than the average, the risk of loss – a return under a chosen level,
usually 0%, and value-at-risk, for the greatest losses, with a probability of less
than 1-5% in a given period of time. The Basel II accord improves on the way risks
are measured, by allowing banks greater flexibility. There is an increase in the
complexity of measuring credit risks, the market risks measurement methods remain
the same, and the measurement of operational risk is introduced for the first time.
The most advanced (and widely-used) risk metrics are based on VaR. However, it must
be noted that VaR calculations are statistical, and therefore unlikely to forecast
extraordinary events. So the quality of a VaR calculation must be checked using back-testing,
and if the VaR value fails in a percentage of 1-5% of the cases, then the premises
of the model must be changed.