[Joao Santos Silva]

Dear Az Us,

I am not sure what you mean when you say that you want to estimate the model for one country, is it for trade between regions?

The Poisson model with RE is not really attractive, just use PPML with FE.

Best wishes,

Joao

[Az Us]

Dear Prof. Joao Santos Silva,

Thank you for the reply.

I mean I want to estimate a gravity model for 1 country and its 30 trade partners.
In particular, I am studying Kyrgyzstan bilateral trade and the consequences of its entry to the Eurasian Economic Union. Basically, my approach is to compare log_dist's coefficients estimated for the same specification but within different time periods. from 1995-2014 and 1995-2018. Changes in this parameter will catch the effect of the implementation EEU, among other shocks

I am also interested in analyzing GDP coefficient since Kyrgyzstan is an entrepot trade hub for the goods from China to EEU. So my intuition is to if the country's export affected by GDP in conjunction with other trade factors.

Kind regards,

Azamat

[Joao Santos Silva]

Dear Az Us,

It looks as if your sample is very particular. This my be fine if that is really the population you care about.

Best wishes,

Joao

[Az Us]

Dear Prof. Joao Santos Silva ,

my result after executing following code:


ppml ln_exp_usd ln_gdp_p_current_o_usd ln_gdp_p_current_d_usd contig comlang_off comlang_ethno comcol ln_distcap, cluster(ln_distcap)

my question is why distance is not significant in this case? the same happens when I regress import against same variables.

best,

Azamat

[Joao Santos Silva]

Dear Az Us,

I do cannot really comment on this because I do not know the details of what you are doing, but notice that you have only 30 clusters, so you have a very small sample.

Best wishes,

Joao

[Az Us]

Dear Prof. Joao Santos Silva ,

I am regressing export of country A against its GDP, parner GDP, distance between them and some other dummies like contingency, common language.

actually, in terms of logic should I use in the left hand side logged value? or it must in levels? according to "Log gravity" the flow must be in levels.

best,

Azamat

[Joao Santos Silva]

Exports must be in levels, not logs.

[Az Us]

Dear Prof. Joao Santos Silva ,

Thank you for your answer.

I got another question. i am trying to add to my model exporter and importer effect. It seems impossible to do, since i am estimating this model only for bilateral trade of one country with several others.

best,

Azamat

[Joao Santos Silva]

In that case you can only include one set of FE; these are pair or partner FEs.

Joao

[Az Us]

Dear Prof. Joao Santos Silva ,

I added parner FEs into my equation. ppml ln_exp_usd contig comlang_off comlang_ethno comcol ln_dist, cluster(ln_dist)
it drastically decreased coefficient of ln_dist. IS there a solid explanation for that? how can interpret coefficient of each partner?

best,

Azamat

[Joao Santos Silva]

Dear Az Us

The coefficient on distance is not identified on that case so it is meaningless. Perhaps you should use ppmlhdfe whose results are easier to interpret?

Best wishes,

Joao

[Az Us]

Thank you Prof Joao Santos Silva ,

I will try. One more question about the other specification - ppml lgd_dlgd_o exp contig comlang_off comlang_ethno comcol ln_dist, cluster(ln_dist) . Are the results of this model consistent? Can i use them for predicting future exports of the country? I would like to remind you, that this model is for 1 exporter and 30 importers.

best,
Azamat

[Joao Santos Silva]

Dear Az Us,

The gravity equation is not generally used for prediction over time, it essentially explains cross-sectional trade flows.

Best wishes,

Joao

[Az Us]

Dear Prof Joao Santos Silva ,

thank you for answer. sorry for bothering you. but i still have questions. i am estimating traditional gravity equation ppml ex lngdp_o lngdp_d comcol contig comm_lang rta, clusted id for one reporter and several partners .

in this case can include remoteness of the partner? does it make sense in this case since i am considering not all countries?

best,

Azamat

[Joao Santos Silva]

That is for you to decide; I am not familiar enough with the problem to be able to comment.

Best wishes,

Joao