Statalist The Stata Listserver

[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: t-GARCH/ GARCH-models with conditional non-normality

From   "Roman Goldbach" <>
Subject   st: t-GARCH/ GARCH-models with conditional non-normality
Date   Sat, 10 Jun 2006 12:54:01 +0200

Dear All,

I am conducting a time series study on credit risks in EMU-bonds; the data show clear autocorrelation, unconditional non-normality (excessive skewnesss and excess kurtosis) and ARCH-processes. Hence I decided for a GARCH-type approach.

However, the generic GARCH(1/1) delivers non-normal residuals, i.e. conditional non-normality exists. This is said to be typical for financial time series. Acoording to Campell, Lo, MacKinlay (1997: 488-9) three possibilities exist:
- use the generic GARCH(1/1), but with robust standard errors - several studies report, that this leads to unbiased estimators and reliable p-values.
2) model another distribution, which accounts for the fat tails - most often a form of t-distribution; Bollerslev (1987) did that
3) use a semi-parametric approach, as in Engle and Gonzalez-Rivera (1991)

Is it possible to calculate a t-GarCH (or other GARCH based on the assumption of a t-type distribution) within STATA ? - I was not able to answer this question positively myself; However, the basic approach - as discussed in Bollerslev (1987) seems 'pretty simple' to me - "The model can be derived as a simple subordinate stochastic process by including an additive unobservable error term in the conditional variance equation."
I'm just a user, hence I am not aware, how I could do this myself in STATA.

1) Can STATA calculate a t-GARCH, or alike model?
2) If yes, how?
3) Is STATA able to run a semiparametric GARCH?
4) If yes, how?
5) Which of the three approaches (according to some experience) would you follow?

Thanks very much in advance.

* For searches and help try:

© Copyright 1996–2020 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index