panel data regression at higher level of aggregation

[charlie held]

Hi,
I have data in individual level but want to run a regression to test for the correlation between variables at a higher aggregation level (county). Both the dependent and independent (and controls) are at the county level. Hence all individuals within a county are identical. My first solution is to only keep one observation within each county to use in the regression but I've noticed in other studies that they keep the full data-set and instead run the regression with clustered errors at the county level. Is this a correct way to do it? and if so, what are the pors and cons? Thankful for any help on the issue!

[Carlo Lazzaro]

Charlie:
welcome to this forum.
You can -cluster()- SEs at -country- level.
Otherwise, assuming that individuals are nested within counties, you can consider -mixed-.

[charlie held]

Thanks Carlo fo the quick reply! Is this a more proper way to do than to just run a regression using only one observation per county?

[Carlo Lazzaro]

Yes, it is.