Description: “Many commentators have suggested that firms need to do stress tests to examine their vulnerability to adverse situations that are not within the data set used to parameterize their risk models. We suggest the adop¬tion of a terminology to describe stress tests and also a meth¬odology that can be adopted by any risk model user to test and communicate a test of the stability of model results.”

Description: People naturally observe risk in the form of the range of experienced gains and losses which is usually represented in statistical terms by standard deviation. In most cases, there are not enough observed values to reliably discern the true amount of losses that are possible in extreme situations. Statistical techniques that have long been applied to insurance company risks are being used to extrapolate from observed risk to extreme tails of the distribution of gains and losses. These processes are essentially an extrapolation from the “known” risk of volatility near the mean to “unknown” risk of extreme losses. “Fat tails” is a term often applied to risks where the likely losses in rare situations significantly exceeds the predictions of a Gaussian (“thin tailed”) distribution. This paper will look at the relationship between the observed risk and the extreme risk in statistical insurance r isk models and discuss how we might create a discussion among model users about the appropriateness of that relationship for specific types of risks.