Panel Data in Stata

[Marti Wall]

Hello!
For my panel data in STATA I am using a set of 6 countries for a time period of 20 years. I am examining the effect of FDI (% of GDP) on economic growth (GDP pc), and including some control variables which are backed up by economic literature (e.g. years of schooling/ physical investment etc etc). I wanted to run the OLS, FE and RE models, then check which one is most applicable by using the Hausman Test. Nonetheless, what I am noticing is that there seems to be a non-linear relationship between the dependent variable and FDI (% of GDP). Would that mean that I should include a quadratic term of FDI too?(FDI^2). My aim was to have a simple double log-linear Cobb-Douglass production function, meaning that GDP pc is transformed to a natural log, as well as all other variables. Would this mean that when I now add the squared term into the regression it doesn't make much sense anymore?
I cannot seem to find anything on a bell shaped relationship in panel data analysis in STATA (especially in regard to running FE models), help would be appreciated!!

*(Moreover, as well as having an overall effect on the 6 countries combined, what is the best (and easiest) STATA command/method to check the effect of FDI (with the control variables) on economic growth for the individual 6 countries?? Is including country dummies a way to go?)*

Thank you so much in advance!
Marti

[Carlo Lazzaro]

Marti:
welcome to this forum.
Please read and act on the FAQ to post more effectively. Thanks.
That said:
1) since you have a T>N panel dataset, you should consider -xtgls- and -xtregar-;
2) exploring non-linearity is easy if you plug in both the linear and the squared terms of the predictors;
3) if you already -xtset- your dataset with -country- as your -panelid-, there's no need to include a categorical predictor too.

[Marti Wall]

Thank you so much Carlo for your fast reply!

Could you please expand further what would running the -xtgls- and -xtregar- commands effectively do, and what results would I ideally like to see in order to use these models instead of OLS/ FE/ RE? Could you please also suggest the best way (by commands in STATA) to check the effect of FDI on GDP pc for the countries individually.

Many thanks!

[Carlo Lazzaro]

Marti:
1) please get yourself familiar with Stata built in commands for T>N panel datasets (-xtgls- and -xtregar-). See Stata .pdf manual.
2) a categorical predictor -i.country- is probably what you're looking for.