The Starting Line

Infant mortality — deaths in the first year of life, per thousand live births — is the oldest summary statistic of a society's health. It captures the things that have to go right at once: a healthy pregnancy, a safe delivery, clean water, warmth, a doctor within reach, a mother who survived to raise the child. Where any of those is scarce, the rate climbs.

The United States buries a larger share of its infants than any other wealthy nation. But "the United States" is a fiction here. The national rate is an average of countries-within-the-country that a baby has no say in being born into.

A baby's odds, by state

Rank the states by infant mortality and the spread is enormous for a single rich country: from 2.8 per 1,000 in New Hampshire to 8.4 in Mississippi — a child three times as likely to be buried before its first birthday depending on which line of the map it crosses on the way out of the womb. The South anchors the bottom; the Northeast and Upper Midwest the top.

Infant deaths per 1,000 live births, 2018, by state. Bars colored from low (blue) to high (red). State rates pool every county, including the small-county aggregate, so they are stable even where individual counties are suppressed. Source: CDC WONDER, Linked Birth / Infant Death Records, 2018.

Disadvantage is a dose — from the very first day

Drop below the state to the county and the book's central pattern appears at the youngest possible age. Sort America's large counties into four groups by neighborhood deprivation and the infant-death rate climbs at every step — 4.0 in the least-deprived quarter, 7.2 in the most. There is no safe range and no plateau; each dot below is a county, and the cloud tilts steadily upward as deprivation rises.

County infant mortality vs neighborhood deprivation, large counties (≥10 infant deaths in 2018). Each circle is a county, sized by its number of births; the heavy steps are the birth-weighted mean within each deprivation quartile. Deprivation is the population-weighted Area Deprivation Index of the county's tracts. Sources: CDC Linked Birth / Infant Death 2018; Neighborhood Atlas ADI 2023; ACS 2019–2023.

The color line

The same counties, re-sorted by the share of residents who are Black, redraw a related but distinct gradient (correlation 0.58). Black infants die at far higher rates than white infants — a gap that decades of research ties to the cumulative stress of racism, unequal prenatal care, and the deprivation already shown above, not to anything intrinsic. This is a county-level association, not a measure of any individual's risk; but it is the demographic signature of the same deprivation gradient, wearing a different label.

County infant mortality vs percent Black, same large counties, sized by births. The line is an unweighted least-squares fit shown only to mark the tilt. Ecological correlations like this one describe places, not people, and overlap heavily with the deprivation gradient in §02. Sources: CDC Linked Birth / Infant Death 2018; ACS 2019–2023.

The first verdict

Everything else in this project happens to people who survived this page. The gradients that follow — in schooling, mobility, longevity — are measured among the children who lived. Infant mortality is where the sorting starts: the same deprivation axis that will later cost a decade of life expectancy is already, in the first twelve months, deciding who gets a life to shorten.

It is also the gradient most clearly open to policy. Infant mortality fell by more than half in the United States over two generations, and the places that closed it did so with prenatal care, Medicaid, and clean delivery — not with anything that required the deprivation map to change first. The dose is real; the dose is also, here, unusually treatable.

Notes & data

Vintage and coverage. CDC WONDER Linked Birth / Infant Death Records, 2018 — the most recent year with complete 51-jurisdiction county coverage in our staged pull (2019 was a partial re-pull of 8 states and is excluded). State rates use all counties including the small-county aggregate; the county-level charts in §02–03 keep only the 253 large counties with at least ten infant deaths, which is where WONDER reports an unsuppressed figure.
Suppression, not imputation. WONDER suppresses any county-year with fewer than ten infant deaths. Those counties are pooled into a state-level "unidentified" total (kept for the state rate) and otherwise dropped — they are disproportionately small and rural, so the county charts describe metropolitan America. We never impute a suppressed cell.
Deprivation. Area Deprivation Index, Neighborhood Atlas 2023, population-weighted from block groups to 2020 tracts to counties. As elsewhere in the project, ADI understates deprivation in expensive coastal cities; the infant-mortality gradient is robust to that.
Ecological correlations. The §03 association between percent Black and county infant mortality is a property of counties, computed across places, and must not be read as an individual probability. It overlaps substantially with deprivation; both are shown because the book's argument is that place summarizes accumulated exposure, and race and deprivation are two of the names that exposure goes by.
Rates. Deaths per 1,000 live births, computed as summed deaths ÷ summed births within each group, so larger counties weight more — the same convention as the rest of the project.