Market Risk [17]

Go to: Summary | Previous | Next   
Bullet points include: If GARCH innovations are Gaussian (not fat-tailed) then process is still generally unconditionally fat-tailed. (Excess) kurtosis of time series of returns generated from a GARCH model > 0. Conditionally, if shocks are Gaussian then (excess) kurtosis should be about zero and large outliers unlikely. To have conditional fat tails (or skew) requires fat-tailed (skew) innovations. N.B. possible confusion over definition of kurtosis: does it include/exclude -3?

Contents | Prev | Next | ERM Lecture Series

Desktop view | Switch to Mobile