clustered standard errors are different from different Stata command

[River Huang]

Dear All, I was asked this question. When one runs the following do file (along with dta file) replicate_bbb.do and macro_workfile.dta, we can obtain (need to ssc install reghdfe and reg2hdfe)

The estimates on _intra, as expected, are the same for all results (commands). However, the clustered standard errors are not. xtreg in (1) and reghdfe in (5) provide different clustered standard errors from reg (2), areg (3) and reg2hdfe in (4). I "conjecture" that is is because of degree of freedom adjustments but find no support for it. I appreciate any comments on this issue. Thanks.

[Carlo Lazzaro]

River:
from the -areg- entry, Stata .pdf manual:

The number of groups specified in absorb() are included in the degrees of freedom used in the finite-sample adjustment of the cluster–robust VCE estimator.

[River Huang]

Dear Carlo, Many thanks for this information.