Standard errors different for ivreghdfe and ivreg2 (even more so when using "small" option for ivreg2) about ivreghdfe HOT 3 CLOSED

Since I can't see the dataset it's a bit hard to tell, but I have two guesses:

1. ivreghdfe also drops singletons. To disable that, use the keepsingletons option and see how things change. I suspect this is the likely case b/c the more "finely-distributed" the FEs are, the more singletons there are.

2. If that doesn't work, maybe check lines 237-239 here: https://github.com/sergiocorreia/ivreghdfe/blob/master/src/ivreghdfe.ado#L235

In summary, ivreghdfe doesn't just activate the small option, but also the noconstant, as well as disables the nopartialsmall option.

from ivreghdfe.

Hi Sergio,

Looks like keepsingletons did the trick. I mistakenly thought that ivreg2 accounted for singleton observations (at least the non-iterative ones) in its standard errors. Thanks so much!

from ivreghdfe.

Great to hear that!

from ivreghdfe.