Time and Country Fixed effects- impact on predictors like GDP

[Neha Bhardwaj Up]

Hello Stata list,

My Panel data looks as follows:
Explained Varibale- EXPORTS
Explanatory Variables- GDP, Capital intensity, Pollution intensity, Labour, Unemployment Rate, Productivity, etc.

If I include BOTH country and time fixed effects in my panel data model by doing i.YEAR and i.COUNTRY, what does it do to variables like GDP (which remain same for a particular country for a year). Does it treat it as a 'time invariant' predictor and exclude them from the analysis?

Apologies in advance for this very naive question however I am hoping for some explanations.

Thank you!

[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.

If you have panel data, then move to a panel estimator like xtreg (you'll still need to include dummies for year). The issue is not whether something is constant for a year (an observation I assume) - it is whether it is constant over all the observations for a panel, and GDP certainly should not be. Before asking, you should have at least run this to see what you get.

However, I would be worried about explaining exports with GDP when GDP includes exports in its calculation. By arithmetic, there is an association between exports and gdp and you don't estimate regressions to find arithmetic relations. Unless you're lagging things on the rhs, many of your variables can be influenced by exports (pollution, labor, unemployment...) which gives you a very serious endogeneity problem.

[Neha Bhardwaj Up]

Thank you Phil. I am lagging the variables on RHS to take care of the endogeneity. Your answer is very helpful. Thank you again.