Hi,
Could someone please tell me what is the right way to account for time, country
and industry specific effects?
I am estimating a dynamic panel data model using the one step Robust GMM System
estimator (using xtabond2 command).
The data set is unbalanced with over 50,000 observations. It has over 6,000
firms, 37 countries and 7 industries. The sample period is 1987 ? 2002 with
minimum data coverage of 4 years and maximum of 16 years.
My approach to account for time, industry and country specific effects is to
include year dummies (yr1987-yr2002), industry dummies (IndPrimary-IndTrade) and
country dummies (count1-count37) among the regressors. So, overall I have 60
dummy variables.
Since I am dealing with three qualitative variables (year, industry & country),
so I do not suppress the constant term, and I drop one dummy variable for each
group of dummies. For year, I drop the first dummy: yr1987. For industry, I drop
the last dummy: IndTrade. For country, I drop the last dummy: count37.
The problem I am facing is that when I run the regression without any of the
dummies, the validity of the instruments is ACCEPTED by the Hansen J test of
overidentifying restrictions. But once I add the year, industry and country
dummies, the validity of the instruments get REJCETD by the Hansen J test.
Questions:
1- Could someone please tell me if adding dummies among the regressors is the
right way to account for time, industry and country specific effects ?.
if so, What is it that I am doing WRONG ?
2- If it is not the right way, please tell how to do it?.
Thank you and best regards,
Dahlia Anwar El- Hawary
Consultant
Financial Sector Operations and Policy Department
World Bank
Tel: 202 473 5238
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
