Is it possible to include dummy variables in panel data?

[Guest]

Dear Statalist community

I am trying to find the relationship between public spending (dependent variable) and the amount of immigration (independent variable). The format is in panel data.
Right now I am going to include dummy variable (adv = advanced economies is equal to 1 and vice versa) and interaction term (migrant_adv = migrant_pop*adv) into the regression. However the dummy is omitted because of collinearity as shown at the end of the post***. I understand that this might be because 'adv' does not vary through countries and time... However, I am quite confused as this means that I can-not include dummy variables into the equation at all? so my questions are.....

1.Is it possible to include dummy variables in panel data? and how?
2. If my focus is on the interaction term (controlling for the omitted variable bias that can happen because of group of advanced countries), can I just use the equation...

 xtreg govex_gdp migrant_pop migrant_adv  i.year, fe

? (without dummy in the regression)

__________________________________________________ ____________

***

 xtreg govex_gdp migrant_pop adv migrant_adv  i.year, fe

i.year _Iyear_1990-2010 (naturally coded; _Iyear_1990 omitted)
note: adv omitted because of collinearity

Fixed-effects (within) regression Number of obs = 163
Group variable: country Number of groups = 35

R-sq: within = 0.2446 Obs per group: min = 1
between = 0.0141 avg = 4.7
overall = 0.0014 max = 5

F(6,122) = 6.58
corr(u_i, Xb) = -0.7408 Prob > F = 0.0000

------------------------------------------------------------------------------
govex_gdp | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
migrant_pop | .6229395 .3631192 1.72 0.089 -.0958912 1.34177
adv | 0 (omitted)
migrant_adv | -.1574857 .3894843 -0.40 0.687 -.9285088 .6135374
_Iyear_1995 | .205332 .5277572 0.39 0.698 -.8394161 1.25008
_Iyear_2000 | -1.247214 .5390116 -2.31 0.022 -2.314241 -.1801866
_Iyear_2005 | -1.109176 .5801929 -1.91 0.058 -2.257726 .0393739
_Iyear_2010 | -.3361596 .6259008 -0.54 0.592 -1.575193 .9028735
_cons | 15.39561 .9841033 15.64 0.000 13.44748 17.34374
-------------+----------------------------------------------------------------
sigma_u | 4.3948045
sigma_e | 1.9360928
rho | .8374672 (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0: F(34, 122) = 10.77 Prob > F = 0.0000

Thank you
Rintarn

[Marcos Almeida]

The title of your question may lead to some misunderstanding.

If I understood right, your question only relates to the "within" - fe - effects model in panel data analysis.

Shall you use random-effects or between-effects, you may add time-invariant variables in the model.

The reason for that is reported in the Stata Manual.

[Carlo Lazzaro]

Guest:
as an aside to Marcos' helpful insighs, adv might have been omitted due to collinearity with the fixed effect.
I would also investigate (via -hausman-) whether -re- specification might be better for your data.

[Guest]

Dear Marcos and Carlo

Thank you so much! I think I was confused and panic for a bit but now I understand the whole point. Thank you so much!

Guest