Panel Data analysis. How to know the impact for each country.

[Elizabeth Real]

Good afternoon,

I am currently working with a small panel database that includes only 17 countries and 15 periods. First, for a general analysis of the impact that the independent variable has on human development in the Latin American region, I have proceeded as follows:

xtreg depvar indepvar, fe cluster(id)
estimates store fixed
xtreg depar indepar, re cluster(id)
estimates store random
xtoverid (intead of "Hausman test" to choose between FE or RE)

However, now my interest is to know or analyze the impact that my independent variable has BUT for each country of the sample, then my question is with respect to the command to be used on STATA, I´m using:

--xtreg depvar indepvar i.country1, vce(robust)

And I would like to know if this command is ok or there is another more effective way that can be suggested. I will really appreciate the help.

Thanks in advance.
Elizabeth.

[Phil Bromiley]

You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. You only need enough to show us your problem.

I don't think you want to xtset country and then put in country dummies in the model. Implicitly they're already there. You can predict the fixed effect for each panel - see xtreg postestimation.

But, "the impact" is a little tricky. While it might just be a change in the level (the fe approach), we often worry that the parameters on the x's vary. That is, we normally want to ask how much x influences y. If you want to compare this across countries, you may need to let the parameter on x vary across countries. This can be done in a reg with i.country and factor variable notation for the fixed effects approach or mixed for random effects.

[Elizabeth Real]

Thank you so much for your help !