GARCH-M with dummy variables

[ssccffv]

Hi,

I'm asked to perform a GARCH model of excess return of sp500 ranging from 2003M01 to 2013M12.

I got my ARMA model with 2 dummy variables. The independent variables are: dum01 dum02 AR(1) MA(1) MA(2) MA(6). dum01=1 from 2003 to 2007. dum02=1 from 2008 to 2009.

But when I tried to estimate a GARCH(1,1)-M model adding variance in the mean equation, all fitted value after2004M01 were NA. I re-estimated the model without dummy variables and everything seemed fine. Is there any difference between with and without dummy variables when performing a GARCH model ?

Besides, could anyone tell me whether the information table below the equations indicates the information of the mean equation or the variance equation in a ARCH/GARCH output window?

Thank you very much for any help.

[EViews Gareth]

We'd need to see the workfile.

The summary statistics are for the mean equation.

[ssccffv]

We'd need to see the workfile.

The summary statistics are for the mean equation.

Hi Gareth,

Here is my workfile. https://www.dropbox.com/s/u2p9qkuxulavf ... 4.wf1?dl=0 The problem is in the equation "variance_in_mean".

Thanks a lot.

[EViews Gareth]

What did you do that resulted in NA fitted values?

[ssccffv]

What did you do that resulted in NA fitted values?

I went to Quick-Estimate Equation, then selected ARCH method. Line in the mean equation window was "ex_rt c dum01 dum02 AR(1) MA(1) MA(2) MA(6)". ARCH-M was "variance". For the variance section, model was GARCH/TGARCH. Order: ARCH 1, GARCH 1. Then just clicked "OK" I didn't change anything else.

[EViews Gareth]

And where did you see the NAs?

[ssccffv]

When the equation came out, I checked the "Resids" and the figure was weird so I went to "Actual, Fitted, Residual Table" and found those NAs.

[EViews Gareth]

What is the build date of your copy of EViews? You can view it under Help->About EViews.

[ssccffv]

The built date of the copy in my computer is 30 Jun 2011. It's Eviews 7.2. I also tried Eviews 8 in a computer in the school computer lab and it gave the same NAs. But I can't get the built date of that copy right now.

[EViews Gareth]

You should update your copy to the latest version.

When I take the steps you take, I receive perfectly normal looking results.

[ssccffv]

OK, I'll try it again from the very start in a school's computer to see if I can get a normal result. Thank you very much, Gareth.

BTW, could you give me the workfile which you got a good result with? Thanks.