Pooled Dynamic GMM

[ernestdautovic]

Hi evryone,

I'm trying to prgram an Arellano-Bond Dynamic GMM in a pooled panel data set

what i wrote is

sample 1992 2008
pool eq1
pool eq 1 "the list of cross sections"
eq1.gmm (cx=fd, gmm=perwhite, cov=stackedwhite, iter=seq) d(y?) d(x1?) d(x2?) d(x3?) d(y?(-1)) @ @dyn(y?) @dyn(x1?) @dyn(x2?) @dyn(x3?)

is there something wrong with this?

because when i run the programme I get the error message "GMM is not the view for eq1"

Please help.

Best

[EViews Gareth]

We don't support GMM estimation for Pools. You'll have to use a panel instead.

[ernestdautovic]

Ok, thank you for your swift reaction!

you mean panel gmm dynamic estimate? and how this should be written?

more importantly is this panel dynamic GMM estimate comparable with my previous estimates that are pooled panel data but not dynamic? If not, I would opt for instrumental variable estimator.

sorry for annoying and sorry my ignorance too.

Best

[EViews Gareth]

When you switch from a Pool to a Panel you should get identical results, so yes you can compare your panel results to earlier pool result (of course you might like to re-do any previous pool estimations in a panel, just to be sure).

You should read this post for more information on changing from a Pool to a Panel.

[ernestdautovic]

Ok thanks for you reply.

However, now there is another issue, namely which kind of options I should specify in the estimation?

I have a panel from 1992 until 2010 (forecast included) of 20 countries.

The estimate is obviously a first difference estimate and the point is which kind of cross section, period, gmm, cov and iter options to use?

Some examples

Code: Select all

equation EQ6.GMM(CX=FD,GMM=PERWIGHT,COV=STACKEDWHITE,ITER=ONEC) d(IRL_T) d(B_E2) d(DEBT_E2) d(GDP_E2) d(IRS_T) d(GDPD_E2) d(IRL_T(-1)) @ @DYN(IRL_T,-2,-6) @DYN(B_E1,-2,-3) @DYN(DEBT_E1,-2,-3) equation EQ7.GMM(CX=F,P=F,GMM=PERWIGHT,COV=STACKEDWHITE, ITER=ONEC) d(IRL_T) d(B_E1) d(DEBT_E1) d(GDP_E1) d(IRS_T) d(GDPD_E1) d(IRL_T(-1)) @ @DYN(IRL_T,-2) @DYN(B_E1,-2,-4) @DYN(DEBT_E1,-2,-4) @DYN(GDP_E1,-2,-4) @DYN(IRS_T,-2,-4) @DYN(GDPD_E1,-2,-4)

[EViews Gareth]

I cannot possibly tell you which options you should be using - that is entirely an econometrics question.

[ernestdautovic]

hi,
I figured out a possible solution for my options, to be honest I didn't follow any particular econometric reasoning but just went through almost all possible logical options.

After some templates it turned out that an option specification such as: (CX=FD,PER=F,GMM=PERWIGHT,COV=STACKEDWHITE) has highly insignificant results mainly due to the COV=STACKEDWHITE option. Indeed, I first tried with the COV=STACKEDWHITE because that was my main option for the covariance matrix in my previous Least Squared pooled static estimates.

Then I tied a different covariance matrix option, namely with no option at all (ordinary), and with COV=PERWHITE. It turned out that with ordinary and perwhite covariance methods my estimates are highly significant!

Now I don't really know why this happens?