Balancing treatment and control group with - weight -

[Bjorn Moller Ochsner]

Hi

I have a cross sectional study (individuals from 26 countries) in an RD-design. The treatment is Trump's electoral victory, dependent variable is satisfaction with democracy.

On the individual level, the respondents are as if-randomised. However, some countries are under- or overrepresented in either the control or the treatment group (due to differences in the exact dates when Eurobarometer conducted interviews in the different countries). For instance, Germans constitute 9 % of the control group, but only 3 % of the treatment group. Therefore I want to apply weights to the data. Concretely, treatment-Germans would have to weight 3x as much as control-Germans. The difficulty for me lies in that this weight must be different for every country. I.e. I have one variable (country) where a weight must be applied to EACH value on the variable (e.g. country 1=France, 2=Germany, etc.). Is this possible? Can I do something after creating dummy variables for each nationality?

Thanks!

[Carlo Lazzaro]

Bjorn:
maybe the fix is obtaining the population for each country for the year your cross-sectional data refer to (I mean something like the following example):

. use http://www.stata-press.com/data/r15/census12
(1980 Census data by state)

. summarize median_age [weight=pop]
(analytic weights assumed)

    Variable |     Obs      Weight        Mean   Std. Dev.       Min        Max
-------------+-----------------------------------------------------------------
  median_age |      50   225907472    30.11047    1.66933       24.2       34.7