How to estimates autocorrelation-consistent standard errors while clustering at a higher level

[Jinchi Dong]

Hello,

I have a panel data. The observations is the GDP data for provinces in different countries in different years

xtset province year

Due to the serial correlation, I try to use the "ivreghdfe" to get the autocorrelation-consistent standard error. The regression code is as follows:

ivreghdfe gdp x, cluster(province year) bw(2)

However, many researchers suggest to cluster at coarsest level. Therefore, I try to use the following code to cluster at country level:

ivreghdfe gdp x, cluster(country year) bw(2)

But I got the following error:

 Error: cluster kernel-robust requires clustering on tsset panel & time vars.
tsset panel var=gdlcode1; tsset time var=year; cluster vars=country1,year

Is there any way to estimates autocorrelation-consistent standard errors while clustering at a higher level?

Many thanks