Country-Specific time trends in a fixed effects model

[Jane Wilson]

Dear all,

I am running a two-way fixed effects regression:

xtreg rate fertility unemp youth union i.year,fe

I also want to include country-specific time trends. I believe the correct way to specify this is i.country##c.year. My question is, will including these time trends, such as below

xtreg rate fertility unemp youth union i.year i.country##c.year,fe

remove the country fixed effects from the model?

[Clyde Schechter]

My question is, will including these time trends, such as below

Code:
xtreg rate fertility unemp youth union i.year i.country##c.year,fe
remove the country fixed effects from the model?

No. Actually, so blatantly no that I fear I have misunderstood your question, because I can't grasp why you might think that. What am I missing?

[Jane Wilson]

Hi Clyde,

Just because when I run the code on stata the output omits reporting the country effects due to collinearity.

[Clyde Schechter]

Yes, but that is completely expected. The use of the -fe- option to create a fixed-effects linear regression model is mathematically the equivalent of including i.country in the regression as a (set of) variable(s). When you also explicitly mention i.country in the list of regression variables, you have, in effect, listed it twice. So the explicitly mentioned one is omitted, but the -fe- generated one is still there. You just don't see it because Stata never reports results for those. If you find that unsettling, then just go back to

regress rate fertility unemp youth union i.year i.country##c.year

and Stata will show you the country level effects. This analysis is entirely equivalent to the one you ran.