I am using two-step system GMM to find impact of two independent business related variables on governance of an economy.
Dataset size: 8 Year data for 71 countries
I am not using any control variable in the main models but reporting 20 different results for different types of countries using the independent variables and their one year lags. For example, for the full dataset, I checked impact of x1, x2, and L1.x1, L1. x2. Then ran similar regression only for the developed countries and so on.
If I find fairly consistent results, and shed discussion on impact of different control variables only in additional analysis, would it suffice?
(Prior literature used 2-4 control variables but they used only OLS. Many of those papers were published in ABDC A/B ranked journals).
Thanks for your cooperation.
Sir,
I think this paper will be a help:
Roodman, D., 2009. How to do xtabond2: An introduction to difference and system GMM in Stata. The stata journal, 9(1), pp.86-136.
It explains both theory and application.
As far as I have found in literature, if hansen J and AR (2) is not significant and number of instruments are below number of groups all the instruments used will be valid and will not be an issue.
Good luck Sir.
In addition to Mohammad Omar Faruq, the J Hansen must not be too nigh, probably around 0.3 to 0.35.. Anything close to 1 is an indication of a misspecification and wrong instrument. Another important assumption of system hmm which is often ignored is the stationarity..
Mohammad Omar Faruq thanks a lot. I got the paper immediately after I posted! Started reading it already!
(The AR(1) AR(2) and Sargan test P values are fine when I run the models. Besides, System GMM resolves unobserved heterogeneity. So, using few controls maybe fine)
Hamid Muili Thanks a lot. As far as I know, Sargan test can be an alternative as well. But found some experts claiming superiority of Hansen J stat.
As far system Gmm is concerned, Hansen J is superior to sargan test and it will even give a better result if robust option specified.
"....It made the Windmeijer (2005) finite-sample correction to the reported standard errors in two-step estimation, without which those standard errors tend to be severely downward biased..." Roodman (2005) page 87.
Using the robust option at the end of the command will resolve the issue.
For example:
xtabond2 DepVar l.DepVar IndepVar year_dum, gmm(l.DepVar,collapse) iv(IndepVar Year_dum,equation(level)) twostep robust
Good luck dear Sir!
Got the following video tutorial series. Might be of help to others.
https://www.youtube.com/watch?v=JkBFrjYXEyM
Yes. Its a great video from Dr Adeleye.. I will also implore you to read the conference of Prof Kripfganz : Generalized method of moments estimation
of linear dynamic panel data models.
Mohammad Omar Faruq , Hamid Muili Thanks a lot for your informative cooperation.
Two-step system GMM estimates without control variables can be credible if only time dummies are used with two independent variables. However, it is important to note that the credibility of the estimates depends on the specific context of the analysis.
In summary, two-step system GMM estimates without control variables can be credible if only time dummies are used with two independent variables, but the credibility of the estimates depends on the specific context of the analysis.
- Badges
- Science topic
- Similar topics
- Correction