Areg absorb(year)

[Michaela Frici]

Hello everyone,
I am trying to replicate research by Chetty et al. on Salience and Taxation. In the original paper, they used the function -areg and absorbed time variable a(year):

areg $ylist $xlist, absorb(year)

I would like to ask whether the areg command in this form runs a random effects panel regression, or whether it runs a fixed effects regression? Are the coefficients and standard errors equivalent to either:
a. xtreg $ylist $xlist i.year, re
or b. xtreg $ylist $xlist i.year, fe

Thank you very much,
Best,

Michaela

[Andrew Musau]

I would like to ask whether the areg command in this form runs a random effects panel regression, or whether it runs a fixed effects regression? Are the coefficients and standard errors equivalent to either:
a. xtreg $ylist $xlist i.year, re
or b. xtreg $ylist $xlist i.year, fe

Neither. You are pooling your panels and have year fixed effects, so the equivalent is

regress $ylist $xlist i.year

or more unconventionally

xtset year
xtreg $ylist $xlist, fe

[Michaela Frici]

Thank you so much for the quick reply!
So model A. assumes that observations in each cross section do not necessarily refer to the same unit? So it is simple Pooled OLS with time fixed effects?
Model A. areg $ylist $xlist, absorb(year) And then if region fixed effects are added, then it would become a random effects model?
Model B. areg $ylist $xlist i.region, absorb(year)

[Andrew Musau]

So model A. assumes that observations in each cross section do not necessarily refer to the same unit? So it is simple Pooled OLS with time fixed effects?
Model A. areg $ylist $xlist, absorb(year

You are assuming that there is no cross-sectional heterogeneity, so yes, it's pooled OLS with year dummies - or a year fixed effects model.

And then if region fixed effects are added, then it would become a random effects model?
Model B. areg $ylist $xlist i.region, absorb(year)

No, it's still a fixed effects model. For a random effects model, you need to use the commands -xtreg- or mixed. You cannot estimate using areg.

[Michaela Frici]

Thank you so much!