Analysing pooled cross sectional data

[Hargun Kaur]

Hi, I am using cross sectional data (annual population survey) appended for multiple years and I wish to look at the relationship between wages for males and females based on whether they do a STEM or non-STEM degree. This is a pooled cross sectional dataset and not a panel dataset, I cannot use pooled OLS, right? Which econometric model can I use to analyse this data?

[Maxence Morlet]

Kind of a Mincerian regression, how rich is your dataset in terms of variables that could predict whether people work?

I would first go for a two-step Heckman, with the first step used to predict whether people work, and second step your actual outcome equation. Note that for this though you need an exclusion restriction: a variable that affects probability to work (and receive a wage) but does not impact actual wages.

You could try Poisson pseudo-maximum likelihood estimation as well.

But if possible I'd go with Heckman (1979).