PPMLHDFE Dummy Variable Error

[Rethabile Molapo]

Good day Community,

I am trying to run a PPMLhdfe/ poisson regression on the following set of data:

Country

Year

Export Value

Tariffs

ln_GDP

ln_Population

ln_distance

Preferential Trade Agreement

language

WTO Membership

ln_Home_GDP

Export Value is the dependent variable, Preferential Trade Agreement, language and WTO membership are dummies abd ln_distance is fixed.
At the moment I am able to run the regression as follows: ppmlhdfe ExportValue Tariffs ln_GDP ln_Population ln_distance PreferentialTradeAgreement WTOMembership ln_Home_GDP, absorb(Country Year)

I find that if I run this regression the dummy variables are omitted due to collinearity, can you'll please advise me on how I can fix this problem.

[Andrew Musau]

Since your estimated model includes country effects, if WTO membership or PTA status does not change for all countries over the sample period, these indicators will be collinear with the country effects. This is not a concern if they are merely control variables, as their effects (along with those of all other time-invariant variables) are captured by the country effects. However, if one of these variables is central to your research, you cannot simultaneously identify its effect while controlling for country effects.

[Rethabile Molapo]

Thanks Andrew Musau.

Another problem I have been finding is that distance is also omitted. Do you have any idea what could be causing this?

[George Ford]

I suspect the same problem--distance is constant within the cross section.

Maybe a Mundlak approach would get those coefficients for you.

[Joao Santos Silva]

Dear Rethabile Molapo,

As noted by others, the problem is that those variables are perfectly collinear with the fixed effects. I believe that this is because you either have a single importer or a single exporter, so one obvious solution is to use data for multiple importers and exporters. Even in that case, some variables will be omitted, but that does not matter, unless you are specifically interested in their coefficients.

Best wishes,

Joao