problème de colinéarité avec le ppmlhdfe

[Blaise Likamata]

I am working on the effect of a common currency on trade flows. When applying the ppmlhdfe estimator, I encounter a collinearity issue. Specifically, I run the following command:
ppmlhdfe tradeflow_comtrade_o union_monetaire_africaine ln_pbii ln_pbij ln_popi ln_popj ln_distij , absorb(exp_time_num imp_time_num pair_id ) vce(cluster pair_id )
Even when I follow the recommendations of Baldwin and Taglioni (2006) and estimate a simpler specification:
ppmlhdfe tradeflow_comtrade_o union_monetaire_africaine, absorb(exp_time_num imp_time_num pair_id ) vce(cluster pair_id )
I still face the same collinearity issue, and the variable of interest (union_monetaire_africaine) and others control variables are dropped.

What should I do to resolve this?

Kind regards,

[Joao Santos Silva]

Dear Blaise Likamata,

As far as I can see, your variables (including the variable of interest) are perfectly collinear with the fixed effects, and that is why they are dropped. With these fixed effects, you can only identify the effects of bilateral variables that vary over time, so you probably need to add years to your data.

Best wishes,

Joao

[Blaise Likamata]

Dear Joao Santos,
Thank you very much for your clarification.
I have already incorporated the time dimension into my dataset.
However, given the issue of perfect collinearity with fixed effects, wouldn't it be preferable to use the standard PPML estimator in order to avoid this problem?

Kind regards,
Blaise Likamata

[Joao Santos Silva]

Dear Blaise Likamata,

If you exclude the fixed effects, your results are likely to be biased because you do not account for the multilateral resistance and for the possibly endogeneity of the FTAs. So, I would stick to the original specification with a long panel.

Best wishes,

Joao

[Blaise Likamata]

Cher Joao Santos ,
Thank you very much for your clarification.
I understand the importance of including fixed effects to control for multilateral resistance and the potential endogeneity of FTAs
I am not sure whether this syntax accounts for fixed effects:
ppml tradeflow_comtrade_o union_monetaire_africaine ln_pbii ln_pbij ln_popi ln_popj ln_distij exp_time_num imp_time_num pair_id
My panel is built using the CEPII database as follows: African countries are exporters (48 countries), and there are 98 importers (including the 48 African countries). The period covered is from 1995 to 2021.
Best regards,
Blaise Likamata

[Blaise Likamata]

Dear Joao Santos Silva ,

After reading your papers and the one by Larch et al. (2018), I now understand better how to use fixed effects in my estimations.
Thank you for your helpful explanations.

Best regards,
Blaise Likamata