F-Stat becomes <10, after controls added

[Apple Tan]

This question has been posted here: http://stats.stackexchange.com/question ... rols-added

I am using Eviews 7.

Model A:

Y = c + log(X) + A

Using a GMM IV, instrument = A_lag. F-stat > 10

Model B:

Y = c + log(X) + A + B

Using a GMM IV, instrument = A_lag. F-stat < 10



So why is it that after B is included my F-stat < 10? This means my instrument becomes weak after B is included. What does this imply? Is A_lag and B correlated?

Thank you

[startz]

You seem to have more variables than instruments. Something doesn't make sense.

[Apple Tan]

Hi

A_lag is the instrument in the true sense of the word, that is the instrument for the endogenous variable.

Of course, all the right-hand side variables in the 2nd stage are included in the 1st stage as instruments too.

Thank you

[startz]

You might want to post your EViews workfile, including the two equations you are estimating.

[Apple Tan]

Hi

M data is confidential so I am only posting my estimation equation and results (with the coefficients value removed).
Hope you can help

Thank you

[Apple Tan]

Model 2 is here

[startz]

That is a surprising outcome. I do not see why you should see that large a change.

At the same time, doing GMM with 21 observations is pretty questionable. GMM is a large sample estimator.

[Apple Tan]

Thanks for the reply. Then should I use TSLS instead of GMM?

Thank you

[Apple Tan]

i tried with TSLS.

My F-stat stll become <10 in Model 2.

Can anyone explain why this is happening?

Thank you