xi in dynamic panel data

[japie999]

Hello,

I can use your help!

I estimate the following model, with GMM (dynamic panel data)
yit = c + xi + xit-1 + yit-1, like in Arrelano and Bond (1991).

Time period: 1998-2007, yit-1 because of autoregression.

xi leads to problems:

1. When I give xi the same value for each year the error ´near singular matrix´ occurs.
2. When I give xi only an value in one year (the other years ´NA´) the error ´insufficient number of observations´

Is there an other way of structurering my xi-data? Or is there a function that ´tell´ the model xi is stable during the period?

Thanks a lot!

[tchaithonov]

For 2, can you turn NAs to zeros instead? I had the same problem a while ago with my panel and the fix was:

Code: Select all

smpl @all series x = @nan(x,0)

Not sure it would help you much but give it a try and see if it works.

[EViews Gareth]

Putting in Zeros rather than NAs will not work.

Your first approach should work. If it isn't, you've probably created a singular matrix somewhere by mistake by including co-linear regressors.

[EViews Glenn]

There's a deeper issue here. The DPD transformations (first-difference, orthogonal deviations) will make any non-varying variable identically equal to zero. As in the standard fixed effects estimator, you cannot estimate the coefficients on these variables using this estimator.

[japie999]

Thanks for your reactions!!

So i'm in deep...
Nice.

Do you, or someone else, have a suggestion using what kind of model instead?

Any suggestion help!!!

Thanks, again!

[startz]

If you have fixed effects, drop the X variables that are fixed within a cross-section.