What does money do to a neighborhood's health habits? We ran the same regression — behavior on income, with composition controls — separately in all fifty states plus D.C., for six behaviors: 306 small models. Five behaviors answered in unison. One answered backwards, almost everywhere.
Story № 2 drew health behaviors along the national deprivation axis. The factory method from story № 11 lets us ask the sharper question: does income's grip on behavior hold inside every state — inside Mississippi's policy environment and Utah's and Vermont's — once you control for who lives where? For each state we regress the tract-level prevalence of a behavior on log household income, controlling for the tract's age and racial composition, and keep one number: the coefficient on income, scaled to a doubling.
Each row is a behavior; each dot is one state's own regression coefficient. Five behaviors live entirely on the left of zero: doubling a tract's income predicts about five points less smoking, seven points less physical inactivity, four points less frequent mental distress — in every single state. Then there is the row that breaks the pattern. Binge drinking rises with income in 50 of 51 jurisdictions. Money does not buy abstinence; it changes the vice. Deprivation taxes the body through cigarettes and immobility; affluence celebrates itself at happy hour.
The smoking coefficients agree on sign but not size — and the spread is itself orderly. Plot each state's slope against its smoking level and the correlation is −0.67: income's protective grip is strongest exactly where smoking is most common. In high-smoking states (Kentucky, West Virginia), a doubling of tract income predicts seven or eight points less smoking; in low-smoking states (Utah, California), there is little smoking left for money to prevent. Disadvantage and the local norm compound each other.
These are descriptive elasticities across neighborhoods, not effects of handing anyone a raise: richer tracts differ in education, stress, prices, and norms, and PLACES estimates are themselves partly modeled from demographics. What the zoo establishes is narrower and stronger — the direction of every relationship survives fifty-one separate policy environments, which no single national regression could show. And the binge-drinking row is the book's cleanest warning against moralizing the gradient: the poor do not simply "behave worse." Each class drinks from its own bottle; only one of them is taxed in years of life.