Country-industry fixed effects

[Alex Boulders]

Dear Stata users,

I am trying to include year and country-industry dummies to my regression analysis in the following form:

egen country_industry=group( iso isic )
xtset country_industry year
reg growth var1 var2 var3 var4  i.year i.country_industry , cluster( country_industry)

However, Stata takes very long time to produce this regression output and I wonder if I am doing it wrong. Fyi, I have over 7000 unique values for country_industry. I would appreciate if you could have any suggestions to speed up the calculation process.

If not, should I use just year and industry dummies instead of country_industry dummies in this case?

Many Thanks

[Carlo Lazzaro]

Alex:
no wonder that you came across that kind of problem with more that 7000 unique values for i.country_industry-; I'm surprised that a -matsize- warning has not creeped up in the Result window.
I'm also not clear wit your going -regress- instead of -xtreg,fe, since you have -xtset- your data before.
If I were you, I would go -xtreg.fe- (I assume that you ruled out -re- specification on the grounds of -hausman- test outcome),,deleting -i.country_industry-:

xtreg growth var1 var2 var3 var4 i.year, fe cluster(country_industry)

[Alex Boulders]

Dear Carlo

Thanks much for your help. Indeed, Stata made a matsize warning and I set a higher matsize. -xtreg,fe- option works fine.

Best Wishes