Dear subscribers,
I would deeply appreciate if anybody can help me out.
I have cross-country panel data with 5-year average.
The significance of the key variable that I am interested in changes
depending on the estimation methods.
1) OLS : significant.
2) Fixed effect: insignificant.
3) system GMM: significant.
It could be the result of the endogeneity bias but the AR(2) test and
Hansen test (between 0.2-0.8 (p-value)) lies within the reasonable
range.
In addition, the lagged dependent variable also lies between the
coeffieicent of OLS and that of fixed effect.
Can I expect any of endogenous independent variable to have downward
bias as in the case of lagged dependent variables?
Or, do I have to have the consistent significant fixed-effect
coefficient? (Well, in cross-sectional analysis, in order to validate
the instrument variables, such problem should not arise as far as I
know)
My estimation can be backed up or is completely wrong? Could you
recommend any reference for this?
Thank you very much in advance
Best regards,
Yee Kyoung
--
Yee Kyoung Kim
Department of Economics
PhD candidate, Seoul National University, Seoul, Korea
(Phone) 016-595-6148
*
* 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/
