* Dynamic panel models
* Prepared for 2010 ICPSR Summer Program
* 7/8/10	Finished
* Revised RJFjr for TAMU.TSCS 7/7/11

clear
use http://www-personal.umich.edu/~franzese/garmit_esspanel1.dta, clear

*First a series of models illustrating temporal dynamics & Hurwicz issues...

reg fdi captax
*This one just silly (bivariate and) to ignore unit effects or dynamics
more

reg fdi l.fdi captax
*This one probably overestimates temporal dependence b/c ignores unit effects (or any model/control thereof)
more

xtreg fdi l.fdi captax, fe
*This one probably underestimates temporal dependence b/c Hurwicz (order 1/T)
more

*help xtabond
more
help xtabond
more
xtabond fdi captax
* So this one is supposed to redress that Hurwicz issue, but notice...
* We can use AB's MA(1) errors test, and Sargan's test for overidentifying restrictions...
more

estat sargan
more

* more DPD model options: help xtdpdsys...
more
help xtdpdsys
more

xtdpdsys fdi captax
* ours has conditions abond perform poorly, but notice help indicates large N, small T for abover also...
* We can use AB's MA(1) errors test, and Sargan's test for overidentifying restrictions...
more

estat sargan
more

*explore some options beyond the default; max lag-depth for IVs of 2 or 10
more
xtabond fdi captax, maxldep(2)

estat sargan
more

xtabond fdi captax, maxldep(10)

estat sargan