First Difference Regression - age categories

[Sophia Neumann]

Dear All,

I'm writing my master's thesis and would like to analyze how flexible working agreements influence the work-family conflict (WFC).
I use a panel dataset with t=2 and would like to conduct a first difference regression.

I control the number of children in the household (nkidsliv), the working hours (job7), and if they live in a relationship (relstat), etc.

Now I'm at a point where two questions arose:

1. How can I control the age of the children? So I have 3 categories stored in the variable ykage.
2. In general, when doing the first difference regression, do I interpret for example the number of children in the household as a change? So if there are more children the WFC would go up?

This is my stata code so far:

reg D.(WFC flextime job7 nkidsliv relstat), noconstant vce(cluster id)


Thank you so much for your help!

[Andrew Musau]

Is your variable "WFC" continuous? Using linear regression suggests that it is.

1. How can I control the age of the children? So I have 3 categories stored in the variable ykage.

As it is a categorical variable, just include it in the regression as such. So your command will be:

xi: reg D.(WFC flextime job7 nkidsliv relstat i.ykage), noconstant vce(cluster id)

2. In general, when doing the first difference regression, do I interpret for example the number of children in the household as a change? So if there are more children the WFC would go up?

In this case, you are using FD regression to eliminate the time-invariant household effect. As a matter of fact, with \(T=2\), FD regression is equivalent to fixed effects. Compare with

xtset hhid
xtreg WFC flextime job7 nkidsliv relstat i.ykage, fe robust

So yes, the interpretation remains exactly as in linear regression except that you add "after controlling for time-invariant household effects."

[Sophia Neumann]

Hi Andrew,

thank you very much, that helped and solved my issues!