GMM Estimation

[Olufemi]

Hi everyone,
I am planning to estimate GMM in Eviews 6 and after going through the Wizard, I always have an error message of "Number of instruments greater than Number of observation"
I have Seven Hundred and forty one observations and 9 regressors. Please what do I do

[EViews Gareth]

The settings you used in the Wizard have created too many instruments. You need to think about what your instruments should be, and set the Wizard accordingly.

[Olufemi]

Dear Qms Gareth,
I tried to follow the wizard but all efforts have proved abortive. I am trying to estimate a GMM equation which contains some dummy variables and some time invariant variables as well as those that vary with time. The dependent variable is Log of quantity while the regressors include GDP, distance, Exchange rate as well as some dummies. The panel is an unbalanced panel. Please can you guide me on the steps to go through with the wizard.

Best Regards

[Olufemi]

Dear QMS Gareth,
In addition, what do Eviews mean by strictly exogenous varaible, does it mean an instrument that is excluded from the equation to be estimated? If yes, can I still estimate the equation without the strictly exogenous variables? and in the step 5 of 6 of the wizard, will C, I mean the intercept term be in the Transform Box and be treated as strictly exogenous or in the No transformation Box?
I am using the Eviews 6.

Best Regards

[EViews Gareth]

I think you need to decide what your model is econometrically before we can figure out how to put that econometric model into EViews.

[Olufemi]

Dear QMS Gareth,
My model is a gravity model and I am using an unbalanced and undated panel for bilateral trade between countries with 741 observations.
The data was collected for a period of 20 years between 1989 and 2008

The model is specified thus:

Log(Tradequantity)=Log(GDPIM)+Log(GDPEXP)+Log(Distance)+Log(Exchange Rate) + Log(Trade Openess) + Breakpoint dummy + Colonial dummy + Language dummy

Log(Tradequantity)= Logarithm of quantity traded between two countries
Log(GDPIM)= Logarithm of the Gross domestic product for the importing country
Log(GDPEXP)= Logarithm of the Gross domestic product for the exporting country
Log(Distance)= Logarithm of distance between the two trading countries
Log(Exchange rate)= logarithm of Exchange rate with respect to the US dollar at the time of trade
Log(Trade openess)= Logarithm of trade openess with respect to the importing country. where trade openess is calculated by (Import+export)/GDP
Break point dummy is zero between 1989 and 2002 and one otherwise
Colonial dummy equals one if the trading partners have colonial ties and zero otherwise
Language dummy equals one if trading partners speak the same language and zero otherwise

Please I will appreciate me if you could guide me all through the process of estimation. Would be willing to make changes to the data structure if I have not done that properly in Eviews.
Please if the data will have to be specified as dated in Eviews, How will I go about it because you could have an example of a country trading with more than one country in a particular year and thus the date does not uniquely specify the observation.

Best Regards

[EViews Gareth]

Nothing you have said so far indicates that you want to, or should be using a dynamic gmm model.

[Olufemi]

Hello QMS Gareth,
I think I don't understand your question but based on the model I just specified, can you guide me on how to estimate a GMM model either by using the wizard or Menu or by programming. I need your help urgently please.

Best Regards

[EViews Gareth]

I cannot tell you how to perform a dynamic GMM estimation, if you cannot tell me what dynamic GMM estimation you want to perform.

All you have done so far is list a dependent variable and a bunch of independent variables. There is a lot more to GMM estimation than that.

[nadja123]

Hi Olufemi,

you may try to go through my posts in reply to the questions of hannibal, under:

Dynamic Panel / GMM (EV 6): Update
by nadja123 on Mon Mar 15, 2010 3:23 pm

Read it until the end, including the typo alert.