Error Distribution

[gastonpresente]

Hi,

im using a GARCH model (1,1). That with a Normal Error Distribution its gives an alpha + beta greater than 1. And with a Student (with 3 df) gives an alpha + beta < 1.
Why is that?

Thanks a lot

Gaston

[trubador]

ARCH/GARCH behavior manifests itself as excessive kurtosis in residuals of the model. This property is known as fat tail or heavy-tailed distribution and means that extreme values have higher probabilities to occur. Since student-t has heavier tails than normal distribution, it has the better ability of generating large values (outliers) and therefore might be a better representative of conditional variance in your data, all else being equal.