Fixed effects

[Tom Howard]

Dear Stata community,

I am currently writing my dissertation and having trouble with some specification in my stata regression. I have a dataset for 20 years observed for 29 cities and adopting a difference-in-difference method to provide an estimate for each year after the treatment.

My regression is as follows: GDPpc = treatment*year dummies (DiD estimates) + controls + city fixed effects + year fixed effects

I have used to command "xtset cityid year"

Where I am getting confused, is whether I need to include i.year in my stata regression, in order to control for the year fixed effects, having already included the "fe" option. Will the fe option control for the time-varying year fixed effects as well as the city fixed effects, or is it necessary to include i.year?

As an aside, when I include i.year, my results are insignificant, however when I don't include it, some of my results become significant.

I hope someone can help me out.

Thanks,
Tom Howard

[Andrew Musau]

You do not show your regression command, but assuming that you are using xtreg, fe, the -fe- option adds unit fixed effects, but not time fixed effects. You need "i.year" to add time fixed effects. But you may want to look at xtdidregress that will adjust for both unit and time effects.

help xtdidregress