PPML for the gravity model

[May Cheiw]

Dear Colleagues,

I have estimated the gravity model using ppmlhdfe and
it gives estimates of coefficients but value of Z are 0 for all estimates.
I would like to seek for advice what alternative method I should try.

Thank you very much in advance.

[Joao Santos Silva]

Dear May Cheiw,

We do not know much about what you are doing, but my guess is that you either have a very small sample or you have some highly collinear regressors.

Best wishes,

Joao

[May Cheiw]

Thank you very much for your comment, Professor Silva. I appreciate very much. I will look into samples again.

[May Cheiw]

Dear Professor Silva

I am sorry to disturb you again but I would like to seek your advice.
Now, I ran ppmlhdfe to estimate the gravity model using the following two sets of samples-;

1) sample contains 137 countries
2) sample contains only 28 EU memer countries

Both samples, independent variables are distance and contig, time variant exporting and importing country dummies.
I estimate the gravity model for several sectors.

For sample (1), in some sectors, z-values of all coefficients are zero.
For sample (2), all coefficients are estimated with z-values.

I do not understand why this happens.
Sample size seems not to be a problem.
Regressors are only two (distance and contig) and thus highly collinear regressors may not be the case.

I very much appreciate for your suggestions.

[Joao Santos Silva]

Dear May Cheiw,

If you can send me by email the data file and the commands to estimate the 2 models, I'll try to look into it.

Best wishes,

Joao

[May Cheiw]

Dear　Professor Silva

Thank you very much for your reply. I will send dataset shortly

[Joao Santos Silva]

Dear May,

Thank you for sending the data and the code. For the benefit of other possible readers, the problem was solved by absorbing a large number of fixed effects using ppmlhdfe.

Best wishes,

Joao