Glossary
Kurtosis
A measure of how heavy the tails of a distribution are. The normal distribution has a kurtosis of 3.
Kurtosis is the average fourth power of deviations from the mean divided by the variance squared: mean((r - mean)^4) / variance^2. The normal distribution scores 3. Excess kurtosis subtracts 3, so the normal scores 0 and heavy-tailed data score above 0.
The fourth power gives a few large deviations a big weight, so kurtosis is unstable in small samples.
Example: ten returns of four +1, four -1, one +5 and one -5 (per cent) have a variance of 5.8, a fourth moment of 125.8 and a kurtosis of 125.8 / 5.8^2 = 3.74, an excess kurtosis of 0.74.