time invariant variables in gravity model

[Anne Wanyonyi]

Dear all,

I have read various literature which says that time invariant variables tend to be absorbed once the bilateral pair fixed effects are incorporated in the gravity model.

In my case, all the time invariant variables (distance, language, common border and colonizer) have not been absorbed. I can see their coefficients after running the ppml regression (though they are very different from the coefficients when the bilateral pair fixed effects have not been included). Does this mean, there is something which is wrong, or one shouldn't include the time-invariant variables in the model from the onset in the gravity model once the pair-fixed effects have been added?

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions.

If you're seeing parameters estimated on variables when you do fixed effects, it means those variable vary within the panels defined by the fixed effects. Whether this is a problem would depend on precisely how you have set up your panels and whether those variables should vary within those panels.

[Anne Wanyonyi]

Thank you Phil for your response.

Would you please clarify this

Whether this is a problem would depend on precisely how you have set up your panels and whether those variables should vary within those panels.

I set up the panel by using this syntax:

egen pairID=group(importer exporter)
xtset pairID Year

Generally, distance, language and common border shouldn't vary within the panel. For each country, they take a value of either 1 or 0 for all the years considered.

[Joao Santos Silva]

See here.