[lal mohan kumar]

Hi

1) -regress-: if you detect both heteroskedasticity and autocorrelation, you should go -vce(cluster);
2) -xtreg-: if you detect both heteroskedasticity and/or autocorrelation, you can go -robust- or -vce(cluster panelid).

I have one more query. Recently I have read that in a hand-out that one should always go for robust standard errors to be on the safer side. So can we directly go for robust standard errors, rather than using it on the basis of results obtained from the post estimation commands?

[Carlo Lazzaro]

Ial:
actually, some econometricians invoke clustered robust standard errors by default.

[lal mohan kumar]

actually, some econometricians invoke clustered robust standard errors by default.

Yes, since most economics papers report robust standard errors!.
Thanks once again Carlo

[Andrew Musau]

The reason behind Carlo Lazzaro's remark in #17 is that robust standard errors are also valid under the assumption of homoscedasticity and no serial correlation. Therefore, rather than test for the presence of serial correlation and heteroscedasticity (needed to assess the validity of conventional standard errors), one uses robust standard errors and foregoes testing. A simple simulation will show you that in the absence of heteroscedasticity and serial correlation, robust standard errors become conventional standard errors.

[Carlo Lazzaro]

Andrew Musau is obviously correct.
However, by automatically invoking clustered robust standard errors one may miss the chance to delve into interesting/challenging facets of her/his regression model, such as heteroskedasticity and serial correlation.
That said, when pressed by delivery deadline, going robust clustered by default makes even more sense.

[Andrew Musau]

Agreed Carlo. I teach a course in introductory econometrics and a significant number of lectures are dedicated to topics of heteroscedasticity and serial correlation. Robust standard errors fall under "remedies", among other approaches such as the not-so-favored weighted least squares (WLS) for the former.