Treatment group is omitted because of collinearity

[Maria Chr]

I am using a difference-in-difference strategy with fixed effects -xtreg...,fe robust-. I am using as my treatment group mothers whose youngest child is younger than one.

I am investigating a few control groups. Firstly, mothers of older children and secondly childless mothers.

I define the variable treat as a dummy which takes the value of 1 for the treatment group and the value of 0 for my control group. I define post as my post reform period and I also control for some time-invariant factors. My dependent variable is wages.

When using xtreg with the first control group, nothing gets omitted, but when I use the second one then my treatment variable gets omitted because of collinearity.

I have tried to see if the number of childless mothers changes over time and it does so I don't understand why it gets omitted in the second case. I also have tried using no controls and I still get the same situation.

Any help is greatly appreciated

[Phil Bromiley]

You will increase your chances of a helpful answer by following the FAQ on asking questions – provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

I don't understand how you can control for some time invariant factors if you're using fixed effects. The fixed effects should control for all time invariant factors.

It is not clear what you're doing with the other conditions when you set up these control groups – what happens to childless mothers when you use mothers of older children as your control group and vice versa?

If your treatment is mothers whose younger child is less than one, and you have a dummy as well that differentiates between mothers who have children less than one and other mothers who are childless, then they would be the same as your treatment variable.