Hello everyone,

I'm currently working on a dyadic dataset which I want to analyze by using panel data regression. I want to examine the lagged effects of domestic war (indepvar) on Net ODAs (depvar). I use a balanced dataset consisting of 191912 observations. Moreover I use lagged control variables such as GDP pc of the recipient, population of the recipient and Oil exports of the recipient.

1.) Now I try to cluster the standard errors on the donor-recipient country level but I am not quite sure how the command should look like.

My regression has the following form:
reg ODA war_t_1 gdp_pc_t_1 pop_t_1 Oil_t_1, cluster(donorID recipientID)

but Stata is telling me, that "too many variables are specified" and that the "option cluster() is incorrectly specified".

But if I type in the following command before:

gen country = donorID + recipientID

and after that I use the regression:

reg ODA war_t_1 gdp_pc_t_1 pop_t_1 Oil_t_1, cluster(country)

the regression runs without problems. Is this the correct command or how do I have cluster the standard errors on donor-recipient country level?


2.)Besides the above I have a question adding fixed effect interactions such as donor*year FE. If my regression looks like this:

reg ODA war_t_1 gdp_pc_t_1 pop_t_1 Oil_t_1 i.yearXdonorID, cluster(country)

I receive one omitted variable (the last one of all the fixed effect interactions). Stata tells me: "note: 1288.yearXdonorID omitted because of collinearity

How can I deal with this issue or is it because of the lagged variables?


Thanks in advance and kind regards.
Marco