Area specific trend

[Katie Zhang]

If I want to control area specific trend (not linear). Which of the following three pieces of codes should I use? Thanks!!!!!!!



xtset area year

1) xtreg Y X1 X2 i.year##i.area, fe r

2) xtreg Y X1 X2 c.year##i.area, fe r

3) xtreg Y X1 X2 i.year c.year##i.area, fe r

[Phil Bromiley]

You'll increase your chances of a useful answer by following the FAQ on asking questions. For example, we don't know if you have multiple observations per area-year or what.

First, when you use i.area you'd controlling for area but not assuming there is a trend per se. (2) forces the influence of year both as a main effect and an interaction to be linear in year. So, 1) is a more general control. However, if you only have one observation per area per year, it won't estimate (and i.area would be redundant to the panel variable). 3) seems like an odd combination. It allows varying year main effects but linear on the interaction with area. So, I would tend to 1) but only if you have multiple observations per area-year.