FE not dropping indicator variables

[Rachel Vos]

Goodmorning!

I was using a fixed effect regression after doing the Hausman test. I forgot to remove my 2 variables i.INDUSTRY and i.YEAR, however in the regression results, there are not omitted and even statistically significant. I was taught Stata would omit static variables that do not vary over time while using FE.

How possibly come there are not omitted?

[William Lisowski]

Clearly, i.YEAR varies over time. Perhaps you will find that i.INDUSTRY does also.

You don't tell us anything about your data, so I will assume that you have a variable named ID that identifies each panel. Then

bysort ID (INDUSTRY): generate change = INDUSTRY[_N]!=INDUSTRY[1]
sort ID YEAR
list ID YEAR INDUSTRY if change, sepby(ID)

will show you any panel where INDUSTRY changes over time.

[Carlo Lazzaro]

Rachel:
as an side to William's excellent advice, I would double-check -i.INDUSTRY- and see if the same panelid had been coded differently (and mistakenky so), because it is quite rare that firms change their industry as time goes by.