command vce(robust) cluster at which level

[Lisa Space]

Dear statalist community,

very basic question:

If I do not specifically define at which level stata should cluster the standard errors, at which level does it cluster then?
So I used the vce(robust) robust command when I worked with panel data and I have data at the individual level.

Thank you in advance

Best

Lisa

[Clyde Schechter]

If you specify -vce(cluster)- without identifying the variable defining the cluster level you will just get a syntax error.

If you specify -vce(robust)-, in most commands you will get a non-clustered robust (Huber-White) standard error. There are some commands, including -xtreg, fe- and -xtpoisson, fe-, where, when -vce(robust)- is specified Stata substitutes -vce(cluster panelvar)-, where panelvar is the grouping variable that was specified when you -xtset- the data.

[Lisa Space]

Clyde Schechter thank you for that quick answer.
So that means when I did xtset the data with the individual identifier that means when performing xtreg, fe , stata will cluster at the individual level?

[Clyde Schechter]

That is correct, if you specify the -vce(robust)- option. If you leave -vce()- unspecified, then you get ordinary standard errors, with no robustness, nor clustering.