Reghdfe F test on fixed effects function disabled

[Naira Andreasian]

Hello!
A friend of mine and I are now working on a seminar paper, in which we need to replicate the econometric analysis of a published article. We have chosen an article by Bertrand&Schoar(2003) in which they investigate the affect of executives on company policies. They run regressions with multiple FE(time, firm, and various manager FE).

Since there are multiple FE, reghdfe package suits best for this analysis. But we need to include the F statistic for the fixed effects only, and the function e(F_absorb) is currently disabled.

I've searched the forum, and found out, that if you write 'old' after the reghdfe regression code, the historic version of the package will run, and then the function e(F_absorb) can be used. But this did not work for me.

So my question is: is there a way we can calculate the F statistic for our FE using reghdfe package? We have tried searching the Internet for an answer, but couldn't find one.

Thank you all in advance.

[Andrew Musau]

reghdfe is from SSC. For my version, this statistic is available if using the -old- option

webuse grunfeld, clear
qui reghdfe invest mvalue kstock, a(company year) old
di e(F_absorb)

Res.:

. di e(F_absorb)
17.403146

So, check your version.

. which reghdfe
C:\Users\709554\ado\plus\r\reghdfe.ado
*! version 5.7.3 13nov2019

Noting that this is just a test that all absorbed variables are jointly equal to 0, you can use testparm if you include these variables as dummies, e.g., below using LSDV

webuse grunfeld, clear
qui regress invest mvalue kstock i.company i.year
testparm i.company i.year

Res.:

      F( 28,   169) =   17.40
            Prob > F =    0.0000

Alternatively, ensuring that the estimation sample stays the same, you can run a likelihood ratio test because the model without absorbed variables (fixed effects) is nested in the one with.