Driscoll Kraay Standard Error (DKSE) vs Panel Corrected Standard Error (PCSE)

[Ashish Chouhan]

Dear statalist members,
I am trying to estimate a panel model where my T (year) is 36 and N (district) is 37. I have issues of heteroscedasticity, serial correlation and cross sectional dependence. I am confused whether PCSE or DKSE is more suitable in my case as both T and N are almost equal.
Any suggestions are welcome.
Thank you.

[Manh Hoang Ba]

Hi Ashish Chouhan,
PCSE is designed for large T, small N tables, with N(N+1)/2 parameters of the covariance matrix estimated, perhaps unsuitable for tables with equivalent N and T.
DKSE performs cross-average and then applies Newey's method. Therefore, the size of N is not an issue.
In this case, I prefer the DK method.

[Ashish Chouhan]

Thanks a lot.