Hi there,

I am studying the effect of becoming a parent on the earnings of fathers versus mothers. I use an event study approach with year of child birth is t=0, and window t=-2 to t=5. earnings calculated as annual gross earnings.

For this estimation i use event time dummies, age dummies, and year dummies. I use the event time dummy from t=-1 as the reference category, so that the rest is calculated as an effect relative on the time right before birth. I use clustered standard errors at the individual level. i replace the event time -2 to +5 for +1 to +8, because you are not allowed to use negative values as dummy variables. hence, i leave 2.eventtime out.

You could regress men and women separately, but I choose to use 1 joint regression, as follows:

Code:
regress income ib2.eventtime i.age i.year ib2.eventtime#i.female i.age#i.female i.year#i.female i.female, robust cluster(id)
I want to know 2 things:

First, is the impact of child birth on earnings significant for men, (and s the impact of child birth on earnings significant for women?)

Do I only look at the coefficients of i.eventtime, or do I do for example a joint F test for all the coefficients included in the model for men (and after that for women). I want to know the impact of child birth so I assume I only look at eventtime right? And for women, do I look at i.eventtime + i.eventtime#i.female?

Second, I want to know if the effect of childbirth on earnings is significantly different for men and women. I read some things about the Chow break test. Instead of looking at 2 different time period, I look at men vs women. In that case, should I do a joint F test for ib2.eventtime#i.female i.age#i.female i.year#i.female i.female? It feels kind of contradicting since I want to know if there is a significant difference of the impact of child birth. My gut is telling me I should just do a joint F test for ib2.eventtime#i.female.

Thank you in advance