Estimating an MGARCH model in EViews

[bbollen]

Hi ALL,

I am fairly new to Eviews but have managed to estimate a number of models so far. My issue is that I cannot work out how to programmatically estimate a bivariate MGARCH (BEKK) model. If I have 2 return series (say rS and rF) what commands do I use to estimate the MGARCH model. I understand it has something to do with the SYSTEM command but I get lost after that! After the model is estimated I would like to have a look the conditional variance & covariance forecasts.

Thanks,
Bernie

[lsym]

Hi,

You can actually select the returns series and open them as a system. Then you press estimate and select ARCH in the estimation method. The options are the VECH model, the CCC (Bollerslev, 1990) and the diagonal BEKK model. However, you can find a code to estimate the Bivariate GARCH (BEKK) model as well as the trivariate case, in the Examples folder, Sample Programs/logl. You can find these folders in the place where you installed the application, probably Program files. It is quite easy to modify these codes. The estimation is done with Maximum Likelihood.

I hope that helps.

[bbollen]

Thankyou for you assistance ,

I can now estimate an MGARCH model. What I wish to do is obtain the conditional variance & covariance of the model at the time of the final observation of the return series. To do this I use the following code.

' MGARCH hedging ratio
system sysMGARCH
sysMGARCH.append rf01=c(1)
sysMGARCH.append rs01=c(2)
smpl @all
sysMGARCH.arch @diagbekk c(fullrank) arch(1) garch(1)
sysMGARCH.makegarch
scalar mgarch_ff = @elem(garch_01 ,250)
scalar mgarch_fs = @elem(garch_01_02 , 250)
HR_V(!i,1) = mgarch_fs / mgarch_ff

I get an error at the line scalar "mgarch_ff = @elem(garch_01 , 250)" - Error unknown object cannot be converted to string.
Can any-one suggest what mistake I am making!

Also, can anyone tell me how I might forecast variances & covariances from this model.

Thanks,
Bernie