EViews.com

Like the many who have attempted to estimate their panel data using Dynamic Panel Data Wizard on first difference GMM, I run into the problem of "the number of instruments is greater than the number of observations". I worked with a panel data file of 1,456 observations (19601 20114x7) but I still get into this problem. Is it because my N<T (ie cross section data is smaller time series observations)? In the series that you have in abond_pan.wf1, you have 1031 observations (1976 1984x140), which makes the N>T.

It depends on which options you choose when doing DPD. But if you choose to have instruments that increase with T, then you generally need N>T.

Dear Gareth

That is true, but if we I want to use a difference option as in Arellano bond estimation, I run into this problem of the number of instruments is greater than the number of observations.

dear
i am also encountered with same problem.. kindly help me.... ur help will be very valuable for me. tell me what is the meaning of this message that "no. of instruments are greater than the number of observation" i am a new user don't know much about the mechanics of GMM kindly tell me in simple way.... i shall be extremely thankful to you

One option is to increase the number of cross section data or reduce the time series data. Run it and see how it goes. Another is not to use the difference option in the GMM specification box. It works well if you don't use the difference option.

ok thanks will try it and then let u knw

Dear Tinfah,

I have slightly the same problem. I want to test the following regression with cross section fixed effects:
s = s sr rat
All the independent variables have 5 lags

When I enter the following instrument in the wizard:
@dyn(_2y,-5)
I still get the error message :
order condition violated - insuffencient instruments

What kind of instrument should I enter?

Regards,
Amana

s stand for yield
rat stands for rating
and sr stands for stock exchange index (instrument variable)

After completing the wizard
I get the following list for instruments:

@dyn(_2y,-2) c rat(-1) rat(-2) rat(-3) rat(-4) rat(-5) sr(-1) sr(-2) sr(-3) sr(-4) sr(-5)

Can anybody show me how to apply GMM producure with Newey and West (1987) correction for autocorrelation for one lag? Thank you.