Glossary
Fat tails
Extreme moves occur more often than a normal distribution allows. Risk figures that assume the normal understate them.
A distribution has fat tails when large moves, in either direction, are more probable than the normal distribution with the same standard deviation implies. Daily market returns are a standard example. It is measured by excess kurtosis, which is positive for fat-tailed data, and by comparing the worst observed days with what the normal predicts.
It matters because risk limits and loss estimates built on the normal are too low in the tail. Example: under the normal, a fall of 4 standard deviations or more has a chance of about 0.0032% on any day, roughly one day in 31,574. Real return series show such days much more often.