st: IV test for nondynamic panel data estimation

[Hewan Belay <hewan_belay@yahoo.com>]

Dear Statalist,

My question concerns instrument-validity tests in (a) panel data estimations which (b) don’t have a dynamic element. For *dynamic* panel data models, there is extensive discussion on how to implement Hansen and Sargan tests of the instruments’ strength and relevance, i.e. for commands xtabond, xtdpd, and xtdpdsys. I don’t see similar discussion for any of the other xt commands. There is some discussion in association with the –gmm- command, but this appears to be again related to dynamic models. The –hausman- test is only discussed (in the case of panel data estimations) in the context of the possible correlation of RHS variables with individual effects, e.g. application of hausman to see whether random effects estimation will yield consistent estimates or whether fixed effects estimation will have to be used. However, I am interested in tests of instruments, which are intended to address endogeneity in the sense of RHS variable(s) being
 correlated with the error term, not with the individual effects. This is provided in the stata manual for non-panel-data contexts (e.g. –ivregress-, e.g. through the postest command -estat overid-) but not for panel data contexts (e.g. –xtivreg-).

So in sum, my question is, where do I find the commands which can be used to test IVs’ strength and relevance in nondynamic panel data models.

Thanks,
Hewan


      

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/