I was wondering if, when running a regression on panel data, clustered standard errors are already correcting for heteroskedasticity. Actually, I have run such a regression and detected heteroskedasticity. Since I used the pooled OLS model I have to cluster the standard errors anyway. Hence, I was hoping that I can address both issues simultaneously. Is that right?
$\begingroup$
$\endgroup$
1
-
2$\begingroup$ I think so, yes, but you might want to provide more detail on how you're handling the clustering. $\endgroup$– Jeremy MilesMay 28, 2014 at 18:25
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
Answering you question: Cluster Robust is also Heteroskedastic Consistent.
I would recommend that you read the A Practitioner's Guide to Cluster-Robust Inference which is a nice piece from Colin Cameron on several aspects of clustered/heteroskedastic robust errors.
Page 20 onward should help you out.
-
$\begingroup$ However, what happens if I correct for heteroscedasticity by means of clustered standard errors, even though there is prove that the initial results are homoscedastic. Is that a severe problem? $\endgroup$– PhilMay 29, 2014 at 11:03
-
$\begingroup$ If the errors are homoscedastic, Heteroskedastic consistent errors are biased. Jusha Angrist and Jorn Pischke have a nice discussion around that topic in the book Mostly Harmless Econometrics (Chapter 8) $\endgroup$– mmgmJun 24, 2014 at 13:28