MA in Mulitvariate GARCH,

[NTH_Le]

Hi,

I'm estimating Bivariate GARCH in Eviews and want to allow for MA() in the mean equation. I can add both AR() and MA() for the single equation GARCH case but for the multivariate GARCH case, only AR() is ok, MA() is said to be "not defined" when I add it as equation specific regressor in the system.

Can anybody help me? Do I need a code? Or how to allow for MA() in the conditional mean equation for bivariate GARCH models?

Thanks.

[trubador]

EViews does not allow MA, SMA and SAR terms in Multivariate GARCH estimation and yes you have to write your own programming code if you want to carry out this analysis. On the other hand, from the econometric point of view, I am not sure if it is necessary to complicate the mean-variance relationship in such estimations. In principle, you should try to keep the conditional mean equation as parsimonious as possible for the sake of computation. I believe GARCH-in-mean specification can be justified more easily than that of MA terms for practical significance.

[NTH_Le]

Thank you, Trubador.

Actually, I've attempt GARCH-in-mean, (bivariate) symmetric BEKK-GARCH and asymmetric BEKK-GARCH (TGARCH) but they all seem to be misspecified. I mean the portmanteau tests for the residuals are insignificant. Can you have any suggestions to correct the models? Thanks in advance.