Thanks for the feedback! You are right that I attempted to write this as a "learning together" side-project analysis as my primary expertise is not in this domain or social science in general.

Can't agree too much that domain understanding matters in data science and hope I conveyed my healthy respect for how qualifications start to pile up when performing observational analysis in an unfamiliar domain (begs the question of if/when one should even go there at all!).

Also great thoughts on working at a more granular level to address the heterogeneity within groups. Unfortunately, I only found unemployment rates at a state and county level, but working with tracts would have been great if possible.