Reading the patterns

The earlier parts follow one outcome at a time. This part reads them together and asks what kind of evidence they are. It describes places in three ways and then turns to research of a different kind. A deprivation ranking, the Area Deprivation Index that most earlier chapters use, puts every tract on one scale. A typology sorts tracts into groups that look alike on many measures at once. A model residual is the part of an outcome that a few census measures leave unexplained. Each describes places and not the people who live in them, and none of them, by itself, identifies a policy that works.

Chapter 9.1, How far one ranking goes will set the outcomes of the book side by side, each in its own units and on the ranking its chapter uses, and classify each by a stated rule: whether it changes steadily from the least to the most deprived places, changes unevenly, shows no detectable difference between the two ends, or is better at the deprived end. It will report the unit, period and ranking of every row, because the chapters compare different places in different years.

Chapter 9.2, Eight types of neighborhood sorts census tracts into eight types on measures of income, poverty, education, family structure, tenure, age and density, then maps the types by county and compares them on outcomes and on the Black and Hispanic shares of residents, none of which is an input. It sets the types beside the first edition's typology, which had life expectancy, mobility, diabetes, smoking and the two shares among its inputs and so spread life expectancy more widely across its types. It finds that the types are not sharply separated and reads them as regions of a continuum. A second sort, built from deprivation measures alone, is in Chapter 10.2, Families of places.

Chapter 9.3, Places that beat the model fits upward mobility, the adult income rank of children raised in low-income families, on five census measures and averages the residuals to counties. Counties below the prediction cluster in the South, the largest counties above it lie around New York, and the counties above the prediction lost tax filers on net in filing years 2022 and 2023. A residual is not a measure of what a place does for a child. It mixes noise in the estimates, measures the model leaves out and the gap between the decades of the outcome and of the predictors.

Two planned chapters extend the comparison. Chapter 9.4, Change in the gradient over time (planned) will ask whether the gap in outcomes between more and less deprived neighborhoods widened or narrowed between 2010 and 2023. Chapter 9.5, Immigrants and the gradient (planned) will ask whether neighborhoods with more immigrants follow the same gradient between deprivation and health, and where they depart from it.

Chapter 9.6, What moving can change will close the book with research that the comparisons cannot replace: randomized offers of housing vouchers, studies of children whose families moved at different ages, and studies of policies that changed a place or an institution. Each study is reported with its design, population and period, and the chapter says what each can and cannot tell a reader about the patterns in the earlier parts.