Explaining analysis with 'cluster' and 'robust' command

[Daniele Duci]

Dear reader,

I have a short and simple question. After doing an xtreg analysis, using a random effects model (', re' command) with the 'cluster' (for home country of a firm) and 'robust' command, I am unsure how to explain this briefly in my text. What I have now is the following:

"Additionally, I cluster the standard error by the home country of the parent firm and ran robust under the standard errors"

My question is if this is the correct way to mention that I used the 'robust' command when doing my analyses?

Thank you in advance and looking forward to your reply!

[Jesse Wursten]

The -robust- option is redundant if cluster(...) is specified. For more info, I recommend checking out this guide (I also have a presentation on the subject here, but I think the actual paper by Cameron&Miller will be more helpful).

[Joro Kolev]

Nowadays robust and clustered variances are prevalent, so instead of "Additionally, I cluster the standard error by the home country of the parent firm and ran robust under the standard errors" you should simply say:


"Additionally, I cluster the standard errors by the home country of the parent firm."

and by the way you should do this only if you have plenty of home countries of the parent firm. Nobody knows what is plenty, but everybody would agree that less than 20 is certainly not plenty, between 20 and 50 is debatable whether it is plenty, and probably above 50 can be safely called plenty.

I myself am rather brave person as asymptotics are concerned, so to me more than 30 degrees of freedom (say you have 50 observations and 20 regressors) is plenty.

[David Benson]

In addition to noting "I cluster the standard errors by the home country of the parent firm" in the text, in the actual table (it's description or text) you can say, "Robust standard errors, clustered by home country of the parent firm, are in brackets."

[Daniele Duci]

Dear all,

Thank you for your useful suggestions, much appreciated!