IGARCH ?
Posted: Tue Nov 12, 2013 9:21 pm
by nithineco
HI, I am getting results like alpha + beta in Garch model estimation more than one dat is integrated Garch. My 1Q is how should I get a results that is less than 1 2Q. how should I explain results this situation where beta+alpha is greater than 1.dependent variable exchange rate return and independent variable is intervention.
Re: IGARCH ?
Posted: Thu Apr 17, 2014 3:59 am
by BobJ
Alpha and beta have to sum to less than one for the GARCH model to be stationary. You can't really report results with alpha+beta>1, they don't mean much except that your model seems misspecified. An integrated GARCH is such alpha+beta sum to exactly one (not greater than one). Two suggestions: try restricting alpha+beta to equal 1 or try adding model lags of the squared returns and variances (instead of a GARCH (1,1) estimate a GARCH (2,2) and see what happens, continue experimenting with different lags of variances and squared returns like that).