xtreg fe, drop estimated constant in stored results?

[Ben Knox]

Hi,

I have panel data (firmid, timeid) and want a table in latex (using esttab) with two model specifications as described below:

i) "eststo: xtreg dep ind, fe"
ii) "collapse dep ind, by(firmid)" followed by "eststo: reg dep ind".

I want the table in latex to include the constant in (ii) but not in (i). Running "esttab , nocon" looses the constant in both.

Is there a way to drop the estimated constant for the first specification so I can simply run "esttab"?

Thanks!

[Arthur Morris]

Posting a working example of your code will improve the quality of answers, and -- since your questions concern user generated commands -- it is helpful to point out where you got them.

It appears that the problem is that when you use esttab, nocon option esttab drops everything called _cons.

The nocons option in reghdfe (from Correia http://scorreia.com/software/reghdfe/) makes this quite simple (The nocons option is not allowed with xtreg, fe.):

webuse nlswork, clear
xtset idcode

*Non working example xtreg doesn't allow nocons:
eststo clear
eststo: xtreg ln_w age ttl_exp tenure not_smsa south, fe // adding nocons here will give an error
eststo: reg ln_w age ttl_exp tenure not_smsa south
esttab , nocon
esttab , drop(_cons)
* reghdfe allows nocons
eststo clear
eststo: reghdfe ln_wage age ttl_exp tenure not_smsa south, absorb(idcode) nocons 
eststo: reg ln_w age ttl_exp tenure not_smsa south
esttab

An alternative approach may be to use the esttab, rename() option with the esttab, drop() to change the name of the _cons produced by xtreg, fe and then drop that name. Maybe another commenter has experience with this approach.

Related note: The coefficient _cons reported by xtreg, fe is not the "constant" in the same sense that the coefficient _cons reported by regress is. What xtreg reports is the average of all of the individual constants (the individual fixed effects). See https://www.stata.com/support/faqs/s...effects-model/

[Ben Knox]

Hi Arthur ~ a belated thanks for your reply. Both suggestions work great. The reghdfe package is actually perfect for my requirements.

Ben