rehdfe option estfe returns indicater _cons in table

[Rune Schmidt Qvist]

Hi

This may be too niche for you all to help, but I'll give it a shot. I'm using two user-written commands, available through SSC: reghdfe and esttab (the latter is a part of the estout package). As you may or may not know, reghdfe has an undocumented option, estfe, that posts a list of absorbed variables to `r(indicate)' for use with the esttab option indicate() (read more). Now, when using that with my model, I somehow end up with a row in my estimation table that says _cons and has a yes for my OLS estimation and a no for the rest. I've posted code and output below. As I work on a research PC that I can't copy from, please note that the output is something I've produced myself to mirror my layout issue without giving away information that I can't give away.

eststo _1: reghdfe outcomevar explvar, vce(cl id)
eststo _2: reghdfe outcomevar explvar, a(industry) vce(cl id)
eststo _3: reghdfe outcomevar explvar, a(industry year) vce(cl id)

estfe *, labels(industry "Industry FE" year "Year FE") 
esttab _1 _2 _3, nogap not noconstant indicate(`r(indicate)')

	(1)	(2)	(3)
explvar	1.03	0.88	0.85
Industry FE	No	Yes	Yes
Year FE	No	No	Yes
_cons	Yes	No	No
Observations	100,000	100,000	100,000

To be clear, I'm wondering why I get that _cons row. Removing the noconstant option from esttab doesn't change anything.

Best,
Rune

[Andrew Musau]

Should there not be a constant in OLS? I vote yes. The -nocons- option suppresses the coefficient, but you asked for a label for the constant by including the OLS regression in the estfe command. You should instead:

estfe _2 _3, labels(industry "Industry FE" year "Year FE")
esttab _1 _2 _3, nogap not noconstant indicate(`r(indicate)')

[Rune Schmidt Qvist]

Should there not be a constant in OLS? I vote yes. The -nocons- option suppresses the coefficient, but you asked for a label for the constant by including the OLS regression in the estfe command. You should instead:

estfe _2 _3, labels(industry "Industry FE" year "Year FE")
esttab _1 _2 _3, nogap not noconstant indicate(`r(indicate)')

Thank you, Andrew, that works! And I understand why it was happening now. Good spot!

Also, to be clear, I do want a constant in my OLS regression. I just dont want to report it.