Fixed effect regression for country panel

[Aneethan Azad]

Hello Statalisters,

I have a country panel which involves around 31 countries with data on variables gathered from 1996 to 2017. I want to have gathered immigration data from the US on country of origin, so I am looking at reasons why people migrate from their respective countries to the United states. I have around 5 main independent variables that are factors in affecting migration and another independent variable which is a control which is population. I have also calculated the same 5 independent variables but relative to the US that I want to run in a separate regression. Here are the commands I ran:
encode country, gen(country1)
xtset country1 year
xtreg immigration var1 var 2 var3 var4 var5 var6 i.year, fe vce(robust)

Do you see any issues with what I have done, and would you say the fixed effects model I am running is most suitable?
What ways would you recommend I check how good my model is currently? I am a beginner in econometrics so I do not have much knowledge, apologies if the answers to the question I posed is obvious.

Thanks,
Aneethan

[Carlo Lazzaro]

Aneethan:
welcome to this forum.
It's virtually impossible (for me, at least), to reply more than vaguely to your query.
Usually, theory and previous researches in your scientific field can help in pointing you out to -fe- or -re-, provided that the comparison between these two specifications should be made on the sample at hand.
That said, as you imposed non-default standard errors (probably to accomodate for heteroskedasticity and/or autocorrelation of the idiosyncratic error), the abovementioned comparison via -hausman- is out of debate, as it allows default standard errors only), whereas it is feasible with the community-contributed command -xtoverid- (just type -search xtoverid- from within Stata to spot and install it).

[Aneethan Azad]

Hello Carlo,

Thank you very much for your reply and the welcome to the forum, my variables include unemployment rate, fdi per capita and gdp per capita. As these are not time invariant would it be best to run a random effects model rather than a fixed effects model. Apologies if this is a basic question I am very new to econometrics.

Thanks,
Yasir

[Carlo Lazzaro]

Yasir:
one downside of fixed effects model is the impossibility of estimating time-invariant coefficients of predictors.
If, as it seems from your post, you have time-varying predictors, fixed effects specification should work Ok. Obviously, the issue of comparing fixed vs random effects specification via -xtoverid- still holds.