Panel data, effect of childbirth on wages

[Else Spierings]

Dear all,

I am researching the effect of child birth on the wages of women and men. I would like to research the effect of the first born and differentiate the effect in the first year of birth, second year after birth etc. Furthermore I would like to correct for any further births (second born, third born etc). How can I do this? I have now managed to create a variable which indicates whether that year a baby is born or not.

Please find an example of my data:

sex id year age educnew kidplus grwage
Male 17301 1994 29 0 0 .1280499
Male 17301 1995 30 4 0 .1107853
Male 17301 1996 31 4 1 .1489137
Male 17301 1997 32 4 0 .0253252
Male 17301 1998 33 4 0 -.0175293
Male 17301 1999 34 4 1 .1754386
Male 17301 2002 37 3 0 .2313433
Male 17301 2004 39 3 0 -.5030303
Male 17301 2005 40 3 0 -.0510732
Male 17301 2006 41 3 0 .079525
Male 17301 2007 42 3 0 -.0833333
Male 17301 2008 43 3 0 .0909091
Male 17301 2010 45 3 0 .0119048
Female 17302 1995 25 3 0 .2637853
Female 17302 1996 26 3 1 -.1128027
Female 17302 1997 27 3 0 -.0127456
Female 17302 1998 28 3 0 .1335726
Female 17302 2000 30 3 0 -.1666667

kidplus=1 if a child was born that year.
grwage is the growth in salary that year
id is the unique identifier for each participant.
educnew is the level of education of the participant.
Participants who did not give birth to children are also included.

Please note that the period is 1993-2018 but there are quite some missings, not everyone filled out the survey every year.

So far I have run regressions using:
xtset id year
xtreg grwage age kidplus educnew if sex==1, fe
xtreg grwage age kidplus educnew if sex==2, fe

This obviously does not provide me with the information I would like. I have been researching the command: sort by id (year);, but can't figure out how to implement it in the right way for this regression.

Any thoughts and help are welcome and much appreciated . Thank you in advance.
Kind regards,
Elisabeth

[Else Spierings]

Already managed to solve my question!

[Carlo Lazzaro]

Else/Elisabeth:
it would be interesting to know how you solved that issue, because other on this forum could benefit from your solution.
That said, why in your code you did not include -i.sex- as a predictor instead of running two separate regressions?

[Else Spierings]

Hi Carlo,
Haha I seem to have mixed up my name . Else is an abbreviation of Elisabeth!

I left out the sex because I ran a fixed effects analysis and that would cause the sex to be dropped out because of collinearity.

I solved it by creating dummies for each year after child birth:
gen kid1 = 1 if kidsdiff>0 & kidsdiff !=.
bysort id (year): gen kid2 = 1 if kid1[_n-1]==1
bysort id (year): gen kid3 = 1 if kid2[_n-1]==1
bysort id (year): gen kid4 = 1 if kid3[_n-1]==1
bysort id (year): gen kid5 = 1 if kid4[_n-1]==1
bysort id (year): gen kid6 = 1 if kid5[_n-1]==1
bysort id (year): gen kid7 = 1 if kid6[_n-1]==1
bysort id (year): gen kid8 = 1 if kid7[_n-1]==1
bysort id (year): gen kid0 = 1 if kid1[_n+1]==1

forvalues y=0/8 {
replace kid`y' = 0 if kid`y'==.
}

Then I ran a panel regression analysis
xtset grwage kid1 kid2 kid3 etc..

but I did not come up with the results that I expected. After further researching the literature I came to the conclusion that I should execute an event study to analyze the impact of child birth. I do not know how to do that yet. I have read all there is about the estudy STATA command but that seems to be only useful for stock returns. Not sure if and how I could use that.
My goal is to create a figure like the one I have attached. This is NOT my article, it is from: Parenthood and gender inequality: Population-based evidence on the child penalty in Finland - Antti Sieppi, Jaakko Pehkonen.

Maybe anyone familiar with event studies who could shed a light on this?

Kind regards,
Elisabeth