Stratification

[filippo dare]

Hi everyone. I have my regressor of interest "pensionable age" which depends on age, gender, and contribution. So I think that I should create cells, since individuals that share tha same characteristcs will have the same "pensionable age". However the outcome variable is on individuals level. How can I run a regression with this setting? thanks for suggestions and ideas.

[George Ford]

Maybe?

reg pensionableage i.age i.gender contribution
reghdfe pensionableage i.gender contribution, absorb(age)

[Fei Wang]

I guess the OP means the "pensionable age" is an independent variable which may have a same value for different individuals, while the dependent variable is individual specific. Well, just regress the outcome on pensionable age at the individual level, and I don't think there is any problem -- that means you don't have to create cells.

[George Ford]

With Social Security pensionable age determined by the years born. It has nothing to do with the gender or contributions. Or, is pensionable age when the person takes SS rather then when they are legally allowed to? If the latter, then Fei is correct. If the former, you simply need a dummy for being born in certain years (computable from age).

" 1943 and 1954, your full retirement age was 66. If you were born in 1955, it is 66 and 2 months. For those born between 1956 and 1959, it gradually increases, and for those born in 1960 or later, it is 67." forcing that into a linear relationship may do little harm, but is wrong.