Hello,
It's my first time here and hopefully not my last.
I am replicating a paper and I need to understand what is the best way to cluster at two levels. For those that want a bit more of information, it is this paper: http://papers.ssrn.com/sol3/papers.c...act_id=2448482 (start of page 13).
If you don't care about it, the paper we are replicating uses the ado file : "cluster 2" command.
However, the coefficients are not the same (similar sign, but not the same).
We also used the command "xtivreg2" which also allows for two level clustering, and the same thing happens: sign is in the same way of the paper but coefficients are not exact (coef off by more than 10%).
we also used reghdfe, and the same thing happens. The results are off by similar amount, (not exactly the same amount).
Here is the code line:
cluster2 rgl ni regcap lugl NI_UGL REG_UGL liquid sec vix unemp libor, fcluster(bank) tcluster(quarter)
We want to cluster bank and quarter. rgl is regulated capital, net income, interaction terms and some controls. the equation in theory is fine.
So, if anyone can she dsome light, it would be much appreciated.
It's my first time here and hopefully not my last.
I am replicating a paper and I need to understand what is the best way to cluster at two levels. For those that want a bit more of information, it is this paper: http://papers.ssrn.com/sol3/papers.c...act_id=2448482 (start of page 13).
If you don't care about it, the paper we are replicating uses the ado file : "cluster 2" command.
However, the coefficients are not the same (similar sign, but not the same).
We also used the command "xtivreg2" which also allows for two level clustering, and the same thing happens: sign is in the same way of the paper but coefficients are not exact (coef off by more than 10%).
we also used reghdfe, and the same thing happens. The results are off by similar amount, (not exactly the same amount).
Here is the code line:
cluster2 rgl ni regcap lugl NI_UGL REG_UGL liquid sec vix unemp libor, fcluster(bank) tcluster(quarter)
We want to cluster bank and quarter. rgl is regulated capital, net income, interaction terms and some controls. the equation in theory is fine.
So, if anyone can she dsome light, it would be much appreciated.
Comment