Bi-variate BEKK, contagion effect dummy

[jongrim]

Dear Tom,

I am estimating a bivariate BEKK-model which includes a dummy variable in the VCV of coefficients to check any shift contagion between two markets. More specifically, I would like to change a21 => (a21+a21d*D) and b21=>(b21+b21d*D), for example. Is there any way to put those addational restrictions into Rats program? Any comment will be appreciated!

Respectfully,
Jongrim

[TomDoan]

Dummies or other XREGRESSORS only enter into the variance intercept in GARCH(MV=BEKK). If you want the GARCH coefficients to shift, you'll need to estimate that using MAXIMIZE.

[e1983]

Hi. I was including an independent variable (not a dummy variable) in the variance equations of a BEKK model; Tom, do I understand properly that the coefficients on the independent variable is essentially only shifting the constant terms in each equation?

Thanks

[TomDoan]

Yes. It makes the "constant" in the BEKK (C+Dx(t))'(C+Dx(t)) where both C and D are lower triangular.

[e1983]

thanks.