Dear users,
I would like to find out how shocks affect household consumption. I have a balanced panel of about 1,500 households for two periods.
I try to implement a fixed effects model with household fixed effects. I suspect that outcomes of households are correlated in villages (richer/poorer neighbourhoods), especially since I am using rainfall shocks which I expect to be correlated in villages since extreme will affect the whole village and not just some households.
Therefore, I did:
xtset hhid (household id) wave (survey wave/year)
xi: i.wave*i.village
xtreg log_consumption x1 x2 (household characteristics) _I*, fe vce(cluster village)

Stata tells me that panels are not nested within clusters which is indeed the case since I identified about 90 households that moved between the two periods. Since my dataset is already very small, is there any other chance to cluster errors at village-level without excluding those 90 households?

Thank you in advance for your comments!
Best, Anne