Population-weighted estimate with panel data

[Tommaso Felici]

Hi all,

I have a problem in estimation with panel data. I would like to study the correlation between log GDP pro capite (as indipendent variables) and log Water Footprint pro capite (as dependent variables) in a Panel Dataset with T=44 and N= 119. I'am using as estimator the CCEMG (common correlated effects mean group) and the AMG (augmented mean group). I want to weight my data for the population,so that if a country has got, for example, the 0,1% of the global population in a certain year, his observation in that year will count for 0,1%. I tried with the expansion of the dataset, where, for example, if in 1970 china has got 10 times the population of the country with less population of the dataset, in that year china has 10 lines, but the outcome was a database with 2 milions of lines and the CCEMG and AMG don't work with such a database.
Can you suggest me how to estimate the coefficients of the model weigthing for population or a different way to weight the observations before the estimation?

[Matteo Pinna Pintor]

Hello Tommaso,

many Stata commands support weighting obervations - including for things such as frequency that might apply to your case. See -help weights- for an introduction to the topic, including various types of weights and why they differ. You should have a look if the command you use supports them. If you post your code, chances are higher that users will be able to help.

[Tommaso Felici]

Hello Tommaso,

many Stata commands support weighting obervations - including for things such as frequency that might apply to your case. See -help weights- for an introduction to the topic, including various types of weights and why they differ. You should have a look if the command you use supports them. If you post your code, chances are higher that users will be able to help.

Hi Matteo,

thanks for your answer. If you mean the codes of the estimation commands, they are:
- xtmg wfppclog gdppclog gdp2pclog gdp3pclog, cce -> for the CCEMG
- xtmg wfppclog gdppclog gdp2pclog gdp3pclog, aug -> for the AMG

from what I understood reading the various helps of weights I can't use them directly into the command (for example the command doesn't work with "aweight"). The problem is that because of the cross-sectional dependence I can't change the estimator, so I was looking for some other way to weight the dataset before or during the estimations. Just ask me if you need some other informations (sorry for the incompleteness of the informations but I have just recently started to use stata and this is the first time that I'm asking something on the forum).