Hand a clustering algorithm thirteen facts about every census tract — income, age, race, density, health, mobility — and ask it to sort America into kinds of neighborhood. It returns eight recognizable worlds, ten years of life expectancy apart at the extremes, each with its own version of an American life.
Most of this series sorted neighborhoods along one axis at a time — deprivation, density, price. But neighborhoods are bundles, and the bundles repeat: the same combination of density, age, income, and race appears in Cleveland and Memphis and Milwaukee, recognizable on sight. Cluster analysis formalizes the recognition. We standardized thirteen tract features and let k-means find the natural groupings, with no labels, no geography, and no outcome data about where a tract sits — only what it is like.
A statistical confession before the tour: the silhouette scores are modest (~0.15), which means these are regions of a continuum, not islands. The boundaries are drawn by the algorithm, but the neighborhoods near them shade into each other. The types earn their keep not by being sharply separated but by being instantly recognizable — and by how much of a life's dashboard the label alone predicts.
Ordered by life expectancy. The names are ours — editorial shorthand for statistical centroids — but every number is the population-weighted mean of the type.
Income = median household income; density = people/sq mi; LE = life expectancy at birth; mobility = mean adult income percentile of children raised in low-income families in these tracts.
Color every county by its most common type and the algorithm — which never saw a coordinate — redraws a familiar map: the Quiet Countryside and the struggling Rural Heartland divide the interior between them (roughly, the upper Midwest vs. the South and Appalachia); Standard-Issue Suburbia rings every metro; the Working Cities and Abandoned Cores trace the industrial Midwest and the urban South; the Barrios follow the border; the Gilded Enclaves hug the coasts and the college towns.
Now run the life course through the typology. Each row below is an outcome from a different stage of life; each dot is one type's population-weighted mean. The types were built mostly from who lives there now — yet they sort children's adult outcomes, measured decades later, almost perfectly. The dashboard is the book's thesis in miniature: tell me the kind of neighborhood, and I will tell you — statistically, never individually — the shape of the life.