I have an unbalanced panel data set for 670 villages, collected once
every two years.Although the data is supposed to be collected
every two years, the responder to the survey is the village headman.
The data is thus collected at the level of the village.
What this means is that my intra-panel correlation is quite high ~0.7.
I am estimating crop area equations for this dataset.
The probits and tobit results are not very stable because of the
high intra-panel correlation (the data is for 11 years, so its six waves
of data)
However if I estimate these equations at the district level, (i.e.
put -iis district rather than -iis village) these equations are
obviously very stable. (Intra-panel correlation is 0.13)
My concern is about the interpretation of these results at the district
level. Are the independent variables now computed as averages
at the district level and are the coefficients simply changes in the
district averages
and their impact on the district average dependent variable?
