Time-varying beta model

[kat15]

Dear EViewers,

I am trying to run a time-varying beta regression model, which looks like this:

ri,t= α + bt rm,t+ et

bt= c0+ c1*D1 + c2*D2 + c3*D3

ht= π+ αet-1^2+ βht-1

The first equation specifies the general relationship, while the second equation specifies the beta as a dynamic process (which depends on three dummy variables, D1, D2 & D3).
I have time series data for: ri,t, rm,t, D1, D2, D3.
I need to obtain the coefficients: bt, c0, c1, c2 & c3.
The third equation specifies a GARCH(1,1) model. I was told that this system of equations should be estimated simultaneously using maximum likelihood.

I cannot figure out the proper way of estimating this model in EViews ... do I specify a System, or use a state space object, or ...? The most important for me is estimating at least the first two equations together.

Please help! I am completely lost on this!!!

[trubador]

Actually, you can estimate this specific form of model via EViews' equation estimation dialog box (Quick/Estimate Equation) with selecting ARCH as the estimation method. Or you can simply run the following line from the command window:

Code: Select all

ARCH() r1t=c(1)+(c(2)+c(3)*D1+c(4)*D2+c(5)*D3)*rmt

After the estimation, you can generate the beta series:

Code: Select all

series betat = c(2)+c(3)*D1+c(4)*D2+c(5)*D3