I thought it is important to perform an ARCH type test (NR2) (with chi square distribution) to determine whether conditional standard deviation or unconditional std dev is suitable for your data. It may be interesting to take into account Asia financial crisis (AFC) and the Global financial crisis (

GFC in your mean and variance equations. You need to justify whether your variables in your mean equation should be estimated as in I(1) or I(0) format. Also, Why GARCH (1,1)? why not just an ARCH?

Although you can apply regression model, but "Regress" isn't the word you should have used since you want to apply the GARCH model. But then, do you have the justification to apply the GARCH model?

I think that is the first thing you should have done - to ascertain if the fit model is linear or non linear.

You don't choose model arbitrarily. I think the pre analysis - unit root, correlation matrix, VIF, differencing etc should be your first step.

However, there are other very good estimation/econometric software eg. RATS, Matlab, eview. But i recommend RATS and STATA for econometric graduate students.

For the stata estimation, you can visit- www.stata.com to see previous examples and command/codes or use this attached file. I am very sure you will get your problem sorted. Lest i forget, which of the STATA version are you using?

If there is no success, let me know.

Good luck!

Hi. You don´t have to use GARCH models. What I see is a lot simpler.

You can simply run a linear regression between GDP growth and a set of covariates, including one measure of exchange rate volatility. For the latter you can use the standard deviation of the returns of daily transactions over a quarter. That will be your direct measure of ER volatility for a given quarter.

These are measures of "realized volatility".

We talk something about this in our paper: "Do it with a smile: forecasting volatility with currency options".Chek this out in Google.

All the best!!

I thought it is important to perform an ARCH type test (NR2) (with chi square distribution) to determine whether conditional standard deviation or unconditional std dev is suitable for your data. It may be interesting to take into account Asia financial crisis (AFC) and the Global financial crisis (

GFC in your mean and variance equations. You need to justify whether your variables in your mean equation should be estimated as in I(1) or I(0) format. Also, Why GARCH (1,1)? why not just an ARCH?

you can do it easily through Eviews and Microfit.

Chuck A Arize Thank you very much Sir!

The reason why I chose Garch 1,1 is because a lot of the outstanding literature seemed to use that model.

1) I have done the ARCH type test - my data presents arch effects

2) I have preceded to estimate my mean equation

3) Following this I ran a GARCH (1,1) model - in which the coefficients where significant

4) I then tested for Homoscedasticity & Auto correlation in the GARCH model and both of them failed the test.

I am using STATA - my inputs are attached.

Using the same data and methodology, in model papers, I should be getting no heteroscedacity & no correlation for the GARCH(1,1) model but i am not.

If anyone could provide insight - it would be much appreciated.