Friends Across the Divide

Chetty et al. (2022) analysed 21 billion friendships between 72 million Facebook users and found that friendships across class lines are among the strongest predictors of upward mobility. I set out to reproduce five of their headline figures myself.

I rebuilt the figures in Python: an interactive Plotly map of economic connectedness across US counties, connectedness versus mobility in the 200 largest counties, binned scatters by ZIP code, and friending bias in high schools and by racial diversity. Most of the work was making ten data files in four formats agree: re-padding FIPS and ZIP codes that lost their leading zeros, matching county names, and binning with deciles and weighted ventiles to match the paper's method.

All five figures reproduced from public data, with an interactive map added on top. The results hold: more cross-class friendships go with more upward mobility, and sharing a school or neighbourhood is not the same as sharing friends.