Testing b coefficients Panel models

[Gaetano Vecchione]

I'm running 9 different Panel model to test diaspora effects of migration on trade for 9 different countries. I would be able to compare parameter estimates obtained for one country with those of another country, the question is whether the hypothesis that both sets of coefficient estimates are the same is acceptable on the data or should be rejected. How can I test this for every pair of countries? Thank you

[Carlo Lazzaro]

I was really impressed by this cartoon that was aired on the Italian television many years ago for the first time.
Recently, an Italian movie with the same title was also available for Jeeg Robot's fans.
Unfortunately (and despite his superpowers) Jeeg Robot does not copy with the requirements of using real given and family name on this forum, for good reasons well explained at https://www.statalist.org/forums/help#realnames.

[Gaetano Vecchione]

Thank you Carlo for your funny answer. Because of the rush, I did not read requirements to participate in this forum. I provide immediately. Gaetano Vecchione

[Carlo Lazzaro]

Welcome on board, Gaetano!

[Carlo Lazzaro]

Gaetano:
you should provide more details about your data.
You do not say which kind of panel data you're dealing with; an example/excerpt of your data via -dataex- can surely help.
Interaction with -i.country- is often a viable approach to investigate the difference you're seemingly interested in.
That said, it's weird that you have to test 9 regression models on the same topic to spot "the best one" (whatever that may mean); it's usually the case that researchers are inspired by what others did in the past when presented with the same research goal (put differently: skim through the literature of your research field).

[Gaetano Vecchione]

Carlo, we are using an hybrid panel model that is, a random-effect (RE) model allowing for time-invariant variables (Allison, 2009; Schunck, 2013).

We estimate 9 separate models for each country approximately in this form: Y = X + Z. Where Y is trade from Italian region to foreign country at time 2000-2017, X controls at time 2000-2017 and Z our variable of interest: a proxy of migration flows in subperiod between 1876 and 1924 (from Italian region to single foreign country).

I would be able to compare parameter estimates obtained for one country with those of another country, the question is whether the hypothesis that both sets of coefficient estimates are the same is acceptable on the data. How can I test this for every pair of countries?

Thank you!
Gaetano

[Carlo Lazzaro]

Gaetano:
I think that the robustness of the hypothesis you're interested in testing should come from the literature.
As far as testing in Stata is concernes, you may want to take a look at: -help test-; -help testparm- and -help lincom-.

[Gaetano Vecchione]

Carlo,
thank you again for your answer.

Best,
Gaetano

PS the countries selected are those indicated by the literature.

[Carlo Lazzaro]

Gaetano:
I meant that the plausibiity of the assumption of the equality of coefficients between countries should come from literature.