Respondent invariant variables in a panel regression

[Jin Woo]

Hi, I am new to STATA. I have panel data (based on survey results) with various respondent invariant independent variables like gender and age. I would like to investigate if these respondent invariant variables have an effect on my dependent variable of interest, car ownership, but am not sure which model to use. I read through various topics in the forum that deal with time-invariant variables. I saw suggestions to use xtreg, re or xtreg, be, since of course, the fixed effects model cannot be applied here.

May I know if I can apply xtreg, re or xtreg, be to respondent invariant independent variables?

Any help will be appreciated. Thank you.

[Carlo Lazzaro]

Jin:
you should compare -fe- vs -re- specification via -hausman-; if the test does not reach statistical significance, go -re-.

[Phil Bromiley]

You might also consider xthybrid (see related Stata Journal articles as well). While Carlo offers the generally accepted answer, if you really care about things that are invariant within panels, fe simply doesn't do the job. The Hausman can be seen as a test whether the between equals the within, but in some cases we might know they should differ and still want to estimate the between.