Good Morning.

I am trying to run a simple descriptive regression on a year-district (1964-2002, ~400 districts) panel:

reg Y Democracy##i.President i.Year i.District

Democracy = 0 if Year is in range 1970-1992,
Democracy = 1 if Year is in range 1964-1969 or 1993-2002

President = President A if Year in range 1964 - 1978
President = President B if Year in range 1979 - 2002

Both regimes go through autocratic and democratic periods, hence the interaction above. Democracy and President vars do not vary across districts (i.e. sample is one country).

Once I add Year FEs the interaction term (Democracy == 1 )*(President== President B) drops out, i.e. Democracy#i.President

It's not obvious to me where the collinearity occurs. Have also tried using areg and xtreg with the same result. Any diagnosis/solutions would be much appreciated.