Heteroscedasticity and autocorrelation problem

[Paul Olz]

Hello everyone,

Unfortunately, I am confused about the occurrence of heteroscedasticity and autocorrelation in my xtreg regression. Am I right, that one way of minimizing the effects is to add vce(robust) to the regression?

Are there any other approach to solve my problem besides adjusting the variables (what I already did)?

Thank you very much in advance!

[Maxence Morlet]

Try the community-contributed summclust command. That will take care of intra-cluster correlation & heteroscedasticity (i.e. make it ignorable, account for it).

This command will result in CV3 standard errors, see MacKinnon et al. (2022).

Do you have N > T?

What do you mean by "adjusting" your variables?

[Joro Kolev]

In -xtreg- adding the robust option computes cluster robust variance, where the id is taken as the cluster. It resolves the problem in the sense that your variance/standard errors would be correct in the face of arbitrary heteroskedasticity and arbitrary within id correlation.