Hi there,

I was wondering about adding control variables to a DiD model. I'm following instructions of an assignment so I can't change the specification of certain variables, come up with a more efficient way, etc.

I examine whether the start of widespread use of the birth control pill (in the 70s) reduced the education gradient in the number of children a woman delivers. Hereby is the treatment group women who are low educated and the control group high educated women. They do this because use of the pill by low educated women in that time period affected fertility greatly, while fertility of high educated women did not change that much by using the pill.

The "time" variable is birth cohorts (1921-1930, 1931-1940, ..., 1971-1980). The treatment group is low educated women. For simplicity, I will ask my questions just about a 2-group 2-period model.


[ATTACH=CONFIG]temp_20557_1606873400300_796[/ATTACH]


About control variables. I read about control variables for DiD models that you should not include the control variable if this is affected by treatment (which makes sense). However, I was wondering what that exactly meant by treatment variable.

For example: household income. Household income is partly determined by the schooling of the mother. Does this mean household income is automatically affected by treatment (time*group)? Or is it necessary to also be affected by the actual treatment itself, i.e. increase in birth control pill usage. My question basically is: should (e.g) household income be affected by education, by birth control pill usage, or by the interaction between these?


I hope that I worded it okay and that my question is clear. It's hard to explain myself with this.


Thank you in advance (: