Hello everyone. I have an unbalanced panel for 105 countries with 15 years time period. have run the following regression using xtdpdsys: xtdpdsys d.depvar d.indepvar1 d.indepvar2 popusharworld finp, lags(1) pre(wtodum) endog(reer tot mehi) artests(2). Though I got desirable results as per my theoretical proposition, the number of instruments (458) far exceeds number of groups (62).
I have then run the same regression in xtabond2 as follows: xtabond2 d.depvar ld.depvar d.indepvar1 d.indepvar2 d.reer popusharworld finp mehi, gmm(ld.depvar d.reer mehi, collapse) iv(wtodum) robust artest(2). I got number of instruments much less than number of groups now but my results have changed or rather worsened drastically.
Please help me with the commands and suggest how can I retain my good results.