MV-EGARCH with spillovers

[pitsikiou]

these are the data

[tahir1]

Dear Tom,
I am quite new to RATS and Running the MV-EGARCH with 3 markets By Koutmos G. (1995) Asymmetric volatility transmission in international stock markets, Journal of International Money and Finance.
I have the following questions:
Ist is that with the 3 markets the codes run and i have the results, however i get this one, does this mean that results are not correct and i need to run the model again?

MAXIMIZE - Estimation by BHHH
NO CONVERGENCE IN 500 ITERATIONS
LAST CRITERION WAS 0.0000000
Usable Observations 296
Skipped/Missing (from 297) 1
Function Value -2575.0414

2nd

I include political risk variable in the model as an exogenous variable to see the impact of political risk on the returns and the volatility of the selected markets, however the codes doesn't run.
I have attached the data, codes and results.

3rd which coefficient in the output is the asymmetric effect


I will be really thankful if you can guide me.
Best Regards

[TomDoan]

these are the data

Sorry to say, but once you've taken into account the serial correlation, there seem to be no GARCH properties. (Shouldn't those be logged -- it looks like you have have run them in levels, though the difference will be slight). For instance:

linreg logden
# constant logden{1 2}
@archtest %resids

gives

Lags Statistic Signif. Level
4 0.352 0.84226

[TomDoan]

Dear Tom,
I am quite new to RATS and Running the MV-EGARCH with 3 markets By Koutmos G. (1995) Asymmetric volatility transmission in international stock markets, Journal of International Money and Finance.
I have the following questions:
Ist is that with the 3 markets the codes run and i have the results, however i get this one, does this mean that results are not correct and i need to run the model again?

MAXIMIZE - Estimation by BHHH
NO CONVERGENCE IN 500 ITERATIONS
LAST CRITERION WAS 0.0000000
Usable Observations 296
Skipped/Missing (from 297) 1
Function Value -2575.0414

2nd

I include political risk variable in the model as an exogenous variable to see the impact of political risk on the returns and the volatility of the selected markets, however the codes doesn't run.
I have attached the data, codes and results.

3rd which coefficient in the output is the asymmetric effect


I will be really thankful if you can guide me.
Best Regards

I would suggest that you use the more modern program posted at

http://www.estima.com/forum/viewtopic.php?f=8&t=1940

as the base. You're working with a older and harder to customize program.

[pitsikiou]

i am sorry not to mention that..i transformed them into ldifference

[TomDoan]

i am sorry not to mention that..i transformed them into ldifference

I assumed as much, but that doesn't change the fact that

1. The monthly data are too coarse for GARCH effects
2. Eight years is probably not enough to really determine whether there is cointegration

[pitsikiou]

thank you very much
I ll take that into consideration and try to find more years.
Thank you for your advices

[tahir1]

Thanks Tom, I will use the new one.
I am using monthly data, what should be the minimum observation should be enough to run MV-EGARCH model.

Thanks

[TomDoan]

Thanks Tom, I will use the new one.
I am using monthly data, what should be the minimum observation should be enough to run MV-EGARCH model.

Thanks

It's not so much the number of observations, but that GARCH effects that might be present at daily or weekly frequencies often are largely dissipated by the time you get to monthly. As with the other poster, make sure that you even have usable GARCH effects at the univariate level before you commit yourself to coding up a multivariate.

[tahir1]

Hi tom,

I am using RATS 8.2, and trying to run the multivariate E-GARCH models propose by Koutmos. However i receive following error.
"## CP18. ADTEST is not the Name of a PROCEDURE. (Did you forget to SOURCE?) "
I will be really thankful for your help.

Further i want to include an exogenous variable in the model of Koutmos, like one variable in the mean equation and a different model in the variance equation.
how i can incorporate it in the Codes,

Thank you so much

Tahir

[TomDoan]

If you don't have ADTEST.SRC, download it from the web site.

To add variables to the mean, change the mean equation and everything else will work.

To add variables to the variance, you need to change the EGARCHSpillover function, and add a VECTOR for the coefficients for the shift variable.

Both of those are very simple with version 9 which adds the VARIANCES=KOUTMOS option to GARCH.

Note that this program is covered in considerable detail as part of the ARCH, GARCH and Volatility e-course.

[Lena]

Dear Tom,

I have a question concerning exogenous variables in the MV-EGARCH.
I'm able to incorporate exogenous variables in the mean equation but have some trouble in adding variables to the variance. I'm wondering whether I have to change the z(i) FRML's (as you said) or whether I should do something like

Code: Select all

do i= 1,n
    compute hlog=c(i)+g(i)*log(hhd(t-1)(i)+k(i)*ex
    do j=1,n
        compute hlog=hlog+a(i)(j)*z(j)(t-1)
   end do j
  compute hhd(t)(i) = exp(hlog)
end do i

with ex as the exogenous series and k(i) the vector of coefficients.
If exogenous variables should be included in the z(i)'s how do I interpret the A coefficients ?

I'd appreciate your help.
Lena

[TomDoan]

Thanks and sorry. You're correct. The adjustment is to EGARCHSpillover (as you did it) rather than to the Z FRML's. I changed the previous reply to reflect that.

As mentioned in the (now edited) reply, this is all very simple with version 9.