model with fixed effect

[Sam Jung]

Hello,

I have a beginner question. I am trying to run a model with fixed effect. Trying some tests, I found that fe model with vce(robust) option would be good.

xtreg y x1 x2, fe vce(robust)

And I also want to check a treatment effect. So, I changed it to LSDV (dummy variable) using vce(robust) option.

areg y I.treatment|x1 I.treatment|x2, absorb(id) vce(robust)

My concern is that the robust standard errors are not the same. I would appreciate if you let me know how to make it.
Thank you.

[Andrew Musau]

-robust- in areg calculates White standard errors which are not appropriate for panel data. You need

areg y I.treatment|x1 I.treatment|x2, absorb(id) cluster(id)

to match the -robust- option in xtreg. This assumes your xtset command was

xtset id year

[Sam Jung]

Thank you for quick answer Andrew.

Yes, xtset is like that. And the robust standard errors are not exactly the same but become very close.

I compared the two like:

 
 xtreg y x1 x2, fe vce(robust)

and

 
 areg y x1 x2, absorb(id) cluster(id)

[Andrew Musau]

There is a degrees of freedom adjustment that areg applies as it absorbs the indicators. You can get the same standard errors by asking xtreg to do the same using the option -dfadj-.

xtreg y x1 x2, fe vce(robust) dfadj

[Sam Jung]

Thanks a lot. I will try it.