Data sources
Every dataset the book draws on, with its years, weights and license, and the chapters that use it.
American Community Survey
American Community Survey 5-year estimates, 2019-2023, census tracts
- Id
- acs
- Producer
- U.S. Census Bureau
- Unit
- census tract (2020 boundaries)
- Years
- 2019-2023 (5-year)
- Weights
- Census Bureau survey weights applied upstream; the staged file carries no margins of error
- License
- Public domain (U.S. Census Bureau)
Cite as: U.S. Census Bureau (2024). American Community Survey 5-Year Estimates, 2019-2023.
The staged tract file holds point estimates only: margins of error were dropped upstream, so the precision of a small-tract value is unknown. Race shares are race-alone and include Hispanic residents, so they are not non-Hispanic shares. Utah, DC and Puerto Rico come from a Census API backfill, and DC and Puerto Rico lack some columns (the household count and the white-alone, child and households-with-children shares). The staged tract file's share of single-parent households uses a different denominator from the Census county tables, and the tracts of Utah and Puerto Rico, backfilled from the Census API, follow the county definition; chapters do not compare that measure between tracts and counties.
Used in Gender gaps in reading and math, School discipline and incarceration, College access, price and earnings, Race, income and mobility, Home value, income and redlining, Work, wages and access to jobs, Renting, affordability and eviction, Health behaviors and the supply of care, Air pollution, Multiple chronic conditions, Place and the length of a life, Gun homicide and the causes of early death, Overdose, alcohol and suicide, Violence and the life-expectancy gap, Civic life, loneliness and partnership, Life evaluation and daily feelings, Disability, retirement and Medicare, Well-being across the life span, Eight types of neighborhood, Change in the gradient over time (planned), Immigrants and the gradient (planned), Places that beat the model, Gradients, slopes and variance, The anatomy of a deprivation index, A deprivation index at two zooms, Families of places, Suppressed counties in death data.
Area Deprivation Index
Neighborhood Atlas Area Deprivation Index, version 4.0.1
- Id
- adi
- Producer
- University of Wisconsin School of Medicine and Public Health (Neighborhood Atlas)
- Unit
- census block group (rolled up to 2020 census tracts and to ZCTAs in the atlas files)
- Years
- 2023 release (version 4.0.1; built on ACS 2019-2023)
- Weights
- none (index); tract values are unweighted means of block-group ranks
- License
- Free for non-commercial use; cite the Neighborhood Atlas and Kind and Buckingham
Cite as: University of Wisconsin School of Medicine and Public Health. 2023 Area Deprivation Index v4.0.1. Downloaded from https://www.neighborhoodatlas.medicine.wisc.edu/. Kind AJH, Buckingham WR. Making Neighborhood-Disadvantage Metrics Accessible (NEJM 2018).
A national percentile rank, with a state decile alongside, built from American Community Survey measures of income, education, employment and housing; a higher rank means a more deprived neighborhood. The index understates deprivation in high-cost cities. Its inputs overlap with ACS-based outcome measures, which matters for any gradient plotted against it. Tract values average block-group ranks without population weights, and a tract whose block groups are all suppressed has no value. The terms allow non-commercial use only.
Used in Place across an American life, Birthweight and preterm birth, Infant mortality, Where births are falling, Smoking in pregnancy, School discipline and incarceration, College access, price and earnings, Teen births, Home value, income and redlining, Mortgage denials, Work, wages and access to jobs, Renting, affordability and eviction, Health behaviors and the supply of care, Air pollution, Multiple chronic conditions, Place and the length of a life, Gun homicide and the causes of early death, Overdose, alcohol and suicide, Mortality trends, 2018 to 2024, Violence and the life-expectancy gap, Civic life, loneliness and partnership, Life evaluation and daily feelings, Disability, retirement and Medicare, Well-being across the life span, Eight types of neighborhood, Places that beat the model, Gradients, slopes and variance, The anatomy of a deprivation index, A deprivation index at two zooms, Families of places, Suppressed counties in death data.
BLS Local Area Unemployment Statistics
Local Area Unemployment Statistics (LAUS), county annual averages
- Id
- bls-laus
- Producer
- U.S. Bureau of Labor Statistics
- Unit
- county of residence
- Years
- 1990-2024
- Weights
- none (one estimate per county)
- License
- Public domain (U.S. federal)
Cite as: U.S. Bureau of Labor Statistics (2026). Local Area Unemployment Statistics (LAUS), County annual averages, 1990-2024. https://www.bls.gov/lau/
Counts employed and unemployed residents by where they live, not where they work. County estimates come from a model that combines the Current Population Survey, payroll counts and unemployment insurance claims, so estimates for small counties lean on the model and carry no published margin of error. Annual figures are averages of the monthly estimates. Connecticut appears by planning region in every year of the file, so chapters spread its values to the legacy counties of the map. Puerto Rico municipios are in the file and are not drawn on the county map.
Used in Work, wages and access to jobs.
BLS Quarterly Census of Employment and Wages
Quarterly Census of Employment and Wages (QCEW), county annual averages by ownership and industry
- Id
- bls-qcew
- Producer
- U.S. Bureau of Labor Statistics
- Unit
- county of the workplace, by ownership and industry
- Years
- 1990-2024
- Weights
- none (jobs and wages reported by employers covered by unemployment insurance)
- License
- Public domain (U.S. federal)
Cite as: U.S. Bureau of Labor Statistics (2026). Quarterly Census of Employment and Wages (QCEW), County annual singlefiles, 1990-2024. https://www.bls.gov/cew/downloadable-data-files.htm
Counts jobs and wages at the place of work, from the reports of employers covered by unemployment insurance, so it describes the jobs located in a county and not the people who live there; the self-employed and most farm and military jobs are left out. Average annual pay is total wages divided by average employment, in current dollars, and mixes full-time and part-time jobs. Cells that would disclose an employer are withheld and stay missing. Connecticut appears under legacy counties or planning regions depending on the year.
Used in Work, wages and access to jobs.
BRFSS
Behavioral Risk Factor Surveillance System (BRFSS), public-use microdata
- Id
- brfss
- Producer
- Centers for Disease Control and Prevention (CDC), with state health departments
- Unit
- respondent (adults in the states, DC and territories); geography stops at state and metro status
- Years
- 2014-2024
- Weights
- design weights: final weight, stratum and primary sampling unit (design-based estimates)
- License
- Public domain (CDC)
Cite as: Centers for Disease Control and Prevention (CDC). Behavioral Risk Factor Surveillance System Survey Data, 2014-2024. Atlanta, Georgia: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention.
An annual health survey pooled over 2014 to 2024 as a repeated cross-section, not a panel. Every estimate uses the survey design (final weight, stratum and primary sampling unit); unweighted means misstate prevalence. Questions rotate, so pooling needs the years each item was asked: firearm questions in 2017, marijuana in 2016 to 2019, and emotional support and life satisfaction in 2014 to 2017. Geography stops at state and metro status, and metro status is missing for 63 percent of respondents. Household income has a finer set of top categories from 2021 on, so income bands that span that year merge the new top categories. The 2024 file is CDC's full-year release; Tennessee did not collect enough interviews that year and is absent, and a few other states are missing in single earlier years. Respondent-level rows are not published.
Used in Health behaviors, person by person, Mental distress in young adults.
CDC injury and violence mapping
CDC Mapping Injury, Overdose, and Violence Dashboard, county and census-tract death rates
- Id
- injury-violence
- Producer
- Centers for Disease Control and Prevention (CDC)
- Unit
- county; census-tract cluster (one rate for each group of small neighboring tracts)
- Years
- 2019-2024 for counties, 2022-2024 for tracts (data as of 2026-05-14)
- Weights
- none (modeled rates with credible intervals; direct rates where counts allow)
- License
- Public domain (CDC, U.S. Department of Health and Human Services)
Cite as: Centers for Disease Control and Prevention (2026). Mapping Injury, Overdose, and Violence Dashboard, county and census-tract data (data as of 2026-05-14).
Rates are deaths per 100,000 residents of the area. An area-year with no deaths has a rate of zero from a count of zero; one with a few deaths, below CDC's reporting threshold, carries a CDC modeled estimate with an interval, or no rate where CDC marks it unstable; one above the threshold carries the direct rate. The tract file does not describe single tracts. CDC combines small neighboring tracts into groups of at least 10,000 residents (the median tract sits in a group of 5 tracts, and the largest group has 20), and every tract in a group repeats the group's value. In sparse rural areas a group can be larger than a county. The unit of any tract figure is the cluster, so tract counts and rates are never summed or averaged within one. Small counts are published as ranges, and rates CDC marks unstable are missing. The drug overdose measure counts unintentional and undetermined-intent drug poisonings. Deaths reach the data four to six months late, so the latest year is provisional. New York City is thinly covered at tract grain, Puerto Rico is absent from the tract file, and Connecticut is absent from the county file.
Used in Gun homicide and the causes of early death, Overdose, alcohol and suicide, Mortality trends, 2018 to 2024, Violence and the life-expectancy gap, Suppressed counties in death data.
CDC PLACES
PLACES: Local Data for Better Health, census tract and ZCTA estimates
- Id
- places
- Producer
- Centers for Disease Control and Prevention (CDC)
- Unit
- census tract (2020 boundaries) and ZCTA, by health measure
- Years
- 2024 release (tracts); 2025 release (ZCTAs, 2023 data year)
- Weights
- none (model-based small-area estimates)
- License
- Public domain (CDC, U.S. Department of Health and Human Services)
Cite as: Centers for Disease Control and Prevention (2024). PLACES: Local Data for Better Health, Census Tract Data 2024 release. Centers for Disease Control and Prevention (2025). PLACES: ZCTA Data, 2025 release.
PLACES values are modeled estimates built from BRFSS survey responses with census and ACS inputs, not counts of people. The 2024 tract release uses BRFSS 2022 for most measures and BRFSS 2021 for blood pressure and cholesterol, with ACS 2018-2022 and the 2020 census; the model uses county poverty and each respondent's age, sex, race and ethnicity and education, with state and county effects. The 2024 tract release has no blood pressure or cholesterol estimates for Florida's tracts. Gradients against an ACS-based index such as the ADI are therefore partly mechanical, and the text says so wherever PLACES is plotted against one. Values are crude prevalence among adults, not age-adjusted, so age structure affects comparisons. The tract release carries point estimates with no intervals; the ZCTA release carries intervals. A tract with no row for a measure had no published estimate, which is not zero. Puerto Rico is not covered.
Used in Health behaviors and the supply of care, Air pollution, Multiple chronic conditions, Place and the length of a life, Civic life, loneliness and partnership, Disability, retirement and Medicare, Eight types of neighborhood, Places that beat the model, Gradients, slopes and variance, Families of places.
CDC WONDER linked birth and infant death
CDC WONDER Linked Birth / Infant Death Records (current series), county and state
- Id
- wonder-lbd
- Producer
- National Center for Health Statistics (NCHS), CDC, National Vital Statistics System
- Unit
- county (250,000 residents or more) and state, by year
- Years
- 2018-2022
- Weights
- none (full counts; WONDER weights infant deaths, so totals can differ by rounding)
- License
- Public domain (NCHS, CDC); WONDER data-use restrictions apply
Cite as: Centers for Disease Control and Prevention, National Center for Health Statistics. National Vital Statistics System, Linked Birth / Infant Deaths on CDC WONDER Online Database. Data are from the Linked Birth / Infant Deaths Records 2007-2023, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program. Accessed at http://wonder.cdc.gov/lbd-current.html in June 2026.
Infant deaths per 1,000 live births, as crude rates. County rows exist only for counties of 250,000 residents or more, and WONDER pools every other county in a state into one Unidentified Counties row. Cells under 10 deaths are suppressed and treated as missing, never zero, so summed county deaths understate the state total. The staged county panel has every year from 2018 to 2022. Connecticut appears under its legacy counties, and only 4 of them appear.
Used in Birthweight and preterm birth, Infant mortality, Suppressed counties in death data.
CDC WONDER natality
CDC WONDER Natality, 2007-2024 (current series), counts of live births by county
- Id
- wonder-natality
- Producer
- National Center for Health Statistics (NCHS), CDC, National Vital Statistics System
- Unit
- county by year (counties of 100,000 residents or more; smaller counties pooled within each state)
- Years
- 2018-2024
- Weights
- none (full counts of births)
- License
- Public domain (NCHS, CDC); WONDER data-use restrictions apply
Cite as: Centers for Disease Control and Prevention, National Center for Health Statistics. National Vital Statistics System, Natality on CDC WONDER Online Database. Data are from the Natality Records 2007-2024, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program. Accessed at http://wonder.cdc.gov/natality-current.html in June 2026.
The panel holds counts of live births by county, with birthweight, gestational age, mother's age and education and other cuts; rates come from dividing by a separate population or birth total. WONDER shows only counties of 100,000 residents or more and pools every other county in a state into one Unidentified Counties row, and a few counties move in or out of that pool between years. The threshold applies to an earlier population estimate than the 2020 census, so some fast-growing counties now above it are still pooled and a named county can sit below it. Chapters that need the same counties in every year use the 578 counties with births in every year from 2018 to 2024. WONDER suppresses cells under 10 births. The county totals have no suppressed cells, but the stratified cuts (birthweight, gestational age, mother's age) do, and a suppressed cell is missing, never zero. Connecticut appears under its legacy counties. 2024 is final in this pull.
Used in Place across an American life, Birthweight and preterm birth, Where births are falling, Smoking in pregnancy, Teen births, Gradients, slopes and variance, Suppressed counties in death data.
CDC WONDER underlying cause of death
CDC WONDER Underlying Cause of Death, 2018-2024 (single race), county and state
- Id
- wonder-ucd
- Producer
- National Center for Health Statistics (NCHS), CDC, National Vital Statistics System
- Unit
- county and state, by year and cause (drug overdose, alcohol-induced, suicide)
- Years
- 2018-2024
- Weights
- none (full counts of deaths; crude rates)
- License
- Public domain (NCHS, CDC); WONDER data-use restrictions apply
Cite as: Centers for Disease Control and Prevention, National Center for Health Statistics. National Vital Statistics System, Mortality 2018-2024 on CDC WONDER Online Database, released in 2026. Data are from the Multiple Cause of Death Files, 2018-2024, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program. Accessed at http://wonder.cdc.gov/ucd-icd10-expanded.html in June 2026.
Deaths per 100,000 residents, crude rather than age-adjusted, so a county's age structure affects comparisons. WONDER suppresses cells under 10 deaths, which removes 47 to 58 percent of county-years, depending on the cause. Suppression is uneven, so the small counties that remain visible are those with the most deaths, and a suppressed cell is missing, never zero. Summing visible counties understates the state total, so county figures are benchmarked against the state files. A county-year is published only when it reaches the threshold, so pooling a county's published years keeps its high years and drops its low ones, which biases a small county's pooled rate upward; county rates use counties published in every year. Connecticut's legacy counties have deaths but no population in 2022 to 2024, and those years leave both sides of a rate. The causes use different WONDER definitions (a drug and alcohol list for overdose and alcohol deaths, the ICD-10 113 cause list for suicide). Connecticut appears under its legacy counties. 2024 is final in this pull (CDC shared final 2024 deaths in February 2026). North Carolina amended many 2023 overdose death certificates after the final 2023 file, so its 2023 overdose count here is lower than the state's current count.
Used in Overdose, alcohol and suicide, Mortality trends, 2018 to 2024, Life evaluation and daily feelings, Immigrants and the gradient (planned), Suppressed counties in death data.
CDC/ATSDR Social Vulnerability Index
CDC/ATSDR Social Vulnerability Index (SVI) 2022, census tracts
- Id
- svi
- Producer
- Centers for Disease Control and Prevention and Agency for Toxic Substances and Disease Registry, Geospatial Research, Analysis and Services Program
- Unit
- census tract (2020 boundaries)
- Years
- 2022 release (ACS 2018-2022)
- Weights
- none (percentile ranks of tracts)
- License
- Public domain (CDC/ATSDR)
Cite as: Centers for Disease Control and Prevention / Agency for Toxic Substances and Disease Registry, Geospatial Research, Analysis, and Services Program. CDC/ATSDR Social Vulnerability Index 2022 Database, United States.
The index ranks tracts on ACS measures grouped into themes (socioeconomic status, household characteristics, racial and ethnic minority status, housing type and transportation); ranks are percentiles of tracts within the nation. It is a different index from the Area Deprivation Index, built from an earlier ACS vintage, and it includes race and ethnicity, which the ADI does not. The book also uses its group-quarters share to flag institutional tracts.
Used in The anatomy of a deprivation index.
Census TIGER/Line tracts
U.S. Census Bureau TIGER/Line Shapefiles, census tracts, 2020 vintage
- Id
- census-tiger
- Producer
- U.S. Census Bureau
- Unit
- census tract (2020 boundaries)
- Years
- 2020
- Weights
- none (boundaries and internal points)
- License
- Public domain (U.S. Census Bureau)
Cite as: U.S. Census Bureau. TIGER/Line Shapefiles, Census Tracts: 2020 vintage (49 states + DC + PR) + 2024 vintage (CT planning regions only).
Used for tract boundaries and each tract's internal point, which locates a tract for distance measures; an internal point is not where residents live within the tract. Connecticut tracts use the planning-region coding of the 2024 vintage.
Used in College access, price and earnings, Place and the length of a life, Eight types of neighborhood, Places that beat the model, Gradients, slopes and variance.
Civil Rights Data Collection
Civil Rights Data Collection (CRDC), 2020-21 school year, district totals
- Id
- crdc
- Producer
- U.S. Department of Education, Office for Civil Rights
- Unit
- school district (local education agency)
- Years
- 2020-21 school year
- Weights
- none (counts reported by districts)
- License
- Public domain (U.S. Department of Education, Office for Civil Rights)
Cite as: U.S. Department of Education, Office for Civil Rights (2023). Civil Rights Data Collection (CRDC), 2020-21.
Districts report to the Office for Civil Rights themselves, and completeness varies. The 2020-21 year was disrupted by remote and hybrid schooling, so suspension counts are lower than in typical years; the data are used for gaps between groups within a district, not for typical-year levels. Suspension counts are students suspended at least once in a school (one or more times), not incidents, so a rate is a share of enrolled students; a student suspended at more than one school counts at each. Zero-tolerance expulsions are a subset of the expulsions with and without services and are never added to them. Some large districts report no suspension count and are left out of both counts and enrollment. Suppressed cells add nothing to district totals. Some reporting districts (charter, regional and single-grade-span districts) cannot be matched to the unified school district geography.
Used in School discipline and incarceration.
College Scorecard
College Scorecard, most recent institution-level data, May 2025 release
- Id
- college-scorecard
- Producer
- U.S. Department of Education
- Unit
- institution
- Years
- May 2025 release (most recent cohorts)
- Weights
- none (institution-level counts and rates)
- License
- Public domain (U.S. Department of Education)
Cite as: U.S. Department of Education (2025). College Scorecard, Most-Recent-Cohorts Institution-Level Data (05/19/2025 release).
Earnings cover students who received federal Title IV aid, pooled over several cohorts, so they do not describe all students; they are in constant dollars, while net price, cost of attendance and debt are nominal for their academic year. Branch campuses repeat their parent institution's earnings, so a count of rows is not a count of institutions. Each institution sits at its campus mailing ZIP code, which locates the campus, not the area its students come from. Small cells are privacy-suppressed and left missing, not zero. Completion rates are split by institution level (four-year versus shorter programs) and are not averaged together.
Used in College access, price and earnings.
County Health Rankings
County Health Rankings & Roadmaps, 2024 analytic data file
- Id
- chr
- Producer
- University of Wisconsin Population Health Institute
- Unit
- county
- Years
- 2024 release
- Weights
- none (published county measures; several are model-based)
- License
- CC BY 4.0 (University of Wisconsin Population Health Institute)
Cite as: University of Wisconsin Population Health Institute. County Health Rankings & Roadmaps 2024. www.countyhealthrankings.org.
Units differ by measure, so each value is read against CHR's own definition. Median household income is the Census Bureau's Small Area Income and Poverty Estimates value for 2022, as CHR reports it. Low birthweight, life expectancy and premature death are computed from NCHS vital statistics counts. The death measures cover different windows (homicide and motor-vehicle crashes 2015-2021, suicide and firearm deaths 2017-2021, child mortality 2018-2021, infant mortality 2015-2021, drug overdose, life expectancy and premature mortality 2019-2021) and different bases: premature mortality and suicide are age-adjusted, while the overdose, homicide, motor-vehicle, firearm, child and infant mortality rates are crude; infant mortality is missing for a county with too few infant deaths in the window. The school segregation index is Theil's index across all racial and ethnic groups (from the NCES Common Core of Data), not a Black-white measure. Many other measures are model-based estimates or come from other sources (ACS, CDC PLACES, CDC WONDER), so a CHR measure is not entered beside its own source in one model. Provider ratios (primary care physicians 2021, dentists 2022, mental health providers 2023) are blank for a county with no provider recorded, while the provider count stays in the file, so a blank ratio with a zero count means no provider, not missing data. Only some measures carry a reliability flag, and a value flagged unreliable or suppressed is left out. Racial subgroup values are missing where CHR does not publish them, which is not zero. Connecticut appears under its legacy county codes.
Used in Place across an American life, Birthweight and preterm birth, Infant mortality, Smoking in pregnancy, Test scores, class and race, Race, income and mobility, Health behaviors and the supply of care, Gun homicide and the causes of early death, Overdose, alcohol and suicide, Civic life, loneliness and partnership, Disability, retirement and Medicare, Suppressed counties in death data.
EPA EJScreen
EPA EJScreen 2024, tract-level environmental indicators (archived copy)
- Id
- ejscreen
- Producer
- U.S. Environmental Protection Agency; archived copy by the Environmental Data & Governance Initiative
- Unit
- census tract (2020 boundaries)
- Years
- 2024 (indicators); ACS 2018-2022 (demographics)
- Weights
- none (modeled indicators; percentiles are national ranks)
- License
- Public domain (U.S. EPA); archived copy on Zenodo
Cite as: U.S. Environmental Protection Agency (2024). EJScreen 2024 (Environmental Justice Screening and Mapping Tool), tract-level data. Zenodo mirror, DOI 10.5281/zenodo.14767363.
EPA removed EJScreen from its website in 2025, so the book uses a public archive of the 2024 release. The indicators are EPA model output (air concentrations, proximity scores and toxicity-weighted releases), not monitor readings. Fine particulate matter and ozone are missing for Puerto Rico, Hawaii and Alaska, and diesel particulate matter for Alaska, and missing does not mean zero exposure. Ozone is EPA's peak metric, an average of the year's highest daily maximum concentrations, not a summer mean. Percentile columns rank tracts nationally. The demographics inside the file are ACS 2018-2022, a different vintage from the book's ACS 2019-2023. Several indicators are heavily right-skewed.
Used in Air pollution, Places that beat the model.
Eviction Tracking System
Eviction Lab Eviction Tracking System, monthly eviction filings by census tract
- Id
- eviction-lab
- Producer
- Eviction Lab, Princeton University
- Unit
- census tract by month, in the cities and metros the system tracks
- Years
- 2020-2026 (monthly)
- Weights
- none (counts of court filings)
- License
- Eviction Lab data license; cite Hepburn, Louis and Desmond; not public domain
Cite as: Peter Hepburn, Renee Louis, and Matthew Desmond. Eviction Tracking System: Version 2.0. Princeton: Princeton University, 2020. www.evictionlab.org.
Coverage is the 35 cities and metros in the staged file, not the nation, so a tract that is absent was not observed and did not have zero filings. Filings are landlord court petitions, not completed evictions or displacements. Eviction moratoria in 2020 and 2021 lowered filings by policy, so a dip in those years does not measure housing distress. Values are counts, not rates, and comparing tracts needs a household denominator. Tracts with sealed identifiers were dropped and cannot be recovered.
Used in Renting, affordability and eviction.
Gallup U.S. Daily Poll
Gallup U.S. Daily tracking poll, well-being extract (respondent level)
- Id
- gallup
- Producer
- Gallup, Inc.
- Unit
- respondent (one interview); only aggregates are published
- Years
- 2008-2017
- Weights
- none (the extract has no survey weight; every estimate is unweighted)
- License
- Proprietary Gallup microdata; respondent-level rows are not published
Cite as: Gallup, Inc. Gallup U.S. Daily tracking poll, 2008-2017, well-being extract. Licensed respondent-level data; only aggregates are published.
A repeated cross-section of 340,604 to 356,599 interviews a year, not a panel. The extract has no survey weight, so every figure is unweighted and levels can differ from the weighted figures Gallup publishes; it supports patterns across people and places, not national prevalence. Evaluative well-being (the Cantril ladder) and experiential well-being (feelings yesterday) measure different things and are never pooled. In 2008 political questions came just before the ladder and lowered its scores; Gallup removed them from part of the sample in January 2009 and added a transition question in April 2009, so ladder estimates leave out 2008 or show that they do not depend on it. County comes from the respondent's ZIP code and is approximate for ZIP codes that span counties. County context columns were joined from other sources and were measured once. Item coverage varies by year. Respondent-level rows, ZIP codes and respondents' county identifiers are not published.
Used in Civic life, loneliness and partnership, Life evaluation and daily feelings, Well-being across the life span.
Health Inequality Project
Health Inequality Project, county life expectancy by household-income quartile, 2001-2014
- Id
- health-inequality
- Producer
- Health Inequality Project (Chetty, Stepner, Abraham and coauthors)
- Unit
- county (1,559 larger counties), by income quartile and sex
- Years
- 2001-2014 (pooled)
- Weights
- none (estimates from linked tax and mortality records, with bootstrap standard errors)
- License
- Open (cite Chetty and coauthors, 2016)
Cite as: Chetty R, Stepner M, Abraham S, et al. (2016). The Association Between Income and Life Expectancy in the United States, 2001-2014. JAMA 315(16):1750-1766. Data: Health Inequality Project, healthinequality.org.
The measure is expected age at death for people who reach age 40, by household-income quartile, not life expectancy at birth. Each county has one pooled estimate for 2001 to 2014, so the file has no year dimension and gives no trend. Only 1,559 larger counties are included; a missing county was not estimated, so the file does not represent the whole country. Unadjusted and race-adjusted columns are separate and are never mixed. County codes follow the 2010 vintage, so Connecticut does not join, and commuting zones are the 1990 definition.
Used in Disability, retirement and Medicare.
HMDA mortgage lending
Home Mortgage Disclosure Act (HMDA) Loan/Application Register, 2024, aggregated to census tracts
- Id
- hmda
- Producer
- Consumer Financial Protection Bureau and the Federal Financial Institutions Examination Council (FFIEC)
- Unit
- census tract (2020 boundaries), by loan purpose
- Years
- 2024
- Weights
- none (counts of applications; tract cells under 10 applications suppressed in staging)
- License
- Public domain (U.S. federal)
Cite as: Consumer Financial Protection Bureau / FFIEC (2025). Home Mortgage Disclosure Act (HMDA) Loan/Application Register (LAR), 2024. https://ffiec.cfpb.gov/data-browser/
Covers lenders above the HMDA reporting thresholds, which is most but not all of the mortgage market. A denial rate is denied applications divided by originated plus denied applications and is not adjusted for credit risk, so it is not a measure of discrimination on its own. The application count includes purchased loans. Race and ethnicity groups are applicants' reported groups and do not sum to the all-applicant total. A tract cell with fewer than 10 applications is suppressed in staging (a rule of this book, not an HMDA flag). Median income is in thousands of dollars and median loan amount is in dollars.
Used in Mortgage denials.
HUD CHAS housing affordability
Comprehensive Housing Affordability Strategy (CHAS) data, American Community Survey 2018-2022, the table of income, cost burden and tenure
- Id
- hud-chas
- Producer
- U.S. Department of Housing and Urban Development (HUD), Office of Policy Development and Research
- Unit
- census tract (2020 boundaries) and county, by tenure (owners, renters, all households)
- Years
- 2018-2022
- Weights
- none (special tabulation of American Community Survey household counts)
- License
- Public domain (U.S. federal)
Cite as: U.S. Department of Housing and Urban Development (2025). Comprehensive Housing Affordability Strategy (CHAS) Data, 2018-2022 ACS, County. https://www.huduser.gov/portal/datasets/cp.html
A special tabulation of the American Community Survey that counts households by tenure, income relative to HUD's area median family income (HAMFI) and housing cost burden: housing costs above a set share of household income (moderate burden) or above a higher share (severe burden). HAMFI is HUD's own benchmark and differs from the Census median household income used elsewhere in the book. HUD rounds every cell independently, so a detail row can exceed its subtotal in small tracts and shares are kept as published. Count columns carry margins of error; the shares built from them do not, so tract shares are noisy. Connecticut appears by planning region.
Used in Place across an American life, Renting, affordability and eviction.
HUD Fair Market Rents
HUD Fair Market Rents, fiscal year 2025, by FMR area
- Id
- hud-fmr
- Producer
- U.S. Department of Housing and Urban Development (HUD), Office of Policy Development and Research
- Unit
- HUD FMR area (a metropolitan area or a nonmetropolitan county; towns in New England)
- Years
- fiscal year 2025 (October 2024 to September 2025)
- Weights
- none (one published rent per area)
- License
- Public domain (U.S. federal)
Cite as: U.S. Department of Housing and Urban Development (2024). Fair Market Rents for the Section 8 Housing Assistance Payments Program, Fiscal Year 2025. https://www.huduser.gov/portal/datasets/fmr.html
Fair Market Rents are monthly rent standards for the housing voucher program, set by bedroom count for each FMR area and applied to every county in it, so counties in the same metropolitan area share one value and differ in this file only across areas. New England areas are towns, and Connecticut appears under its legacy counties. The file is a single release, so a chapter that uses it for earlier years treats it as a proxy for relative rent levels across places and says so.
Used in Renting, affordability and eviction, Life evaluation and daily feelings.
IPEDS
Integrated Postsecondary Education Data System (IPEDS), 2023 survey files
- Id
- ipeds
- Producer
- National Center for Education Statistics (NCES), U.S. Department of Education
- Unit
- institution
- Years
- 2023
- Weights
- none (counts reported by institutions)
- License
- Public domain (U.S. Department of Education, NCES)
Cite as: U.S. Department of Education, National Center for Education Statistics, Integrated Postsecondary Education Data System (IPEDS), 2023 (HD2023 directory; EFFY2023 fall enrollment; SFA2223 student financial aid).
Combines the 2023 institutional directory, fall enrollment and 2022-23 student financial aid files. County and coordinates give the institution's own address, not where its students live. Negative codes in the categorical columns mean not applicable or not available, and they are not numbers. Enrollment and Pell shares are missing for institutions absent from those survey files.
Used in College access, price and earnings.
IRS migration data
IRS Statistics of Income, county-to-county migration data, 2022-2023
- Id
- irs-migration
- Producer
- Internal Revenue Service, Statistics of Income Division
- Unit
- county (inflows and outflows of returns, individuals and income)
- Years
- 2022-2023 (filing years)
- Weights
- none (counts of returns and exemptions)
- License
- Public domain (IRS Statistics of Income)
Cite as: Internal Revenue Service, Statistics of Income Division (2025). County-to-County Migration Data, 2022-2023.
Flows come from address changes between consecutive tax returns, so people who do not file are missed, disproportionately the very poor and some retirees. Individuals are the exemptions claimed on returns, a proxy for household members. Income is in thousands of dollars and comes from the second year's return, after the move. A county whose inflow or outflow is withheld has no net figure. The IRS suppresses small flows, and a suppressed flow is missing, never zero. This file uses Connecticut planning-region codes, not the legacy counties.
Used in Places that beat the model.
IRS Statistics of Income
IRS Statistics of Income, individual income tax statistics by county and ZIP code, tax year 2022
- Id
- irs-soi
- Producer
- Internal Revenue Service, Statistics of Income Division
- Unit
- county and ZIP code, by adjusted-gross-income bracket
- Years
- tax year 2022
- Weights
- none (counts of returns, not people)
- License
- Public domain (IRS Statistics of Income)
Cite as: Internal Revenue Service, Statistics of Income Division (2024). County Data and ZIP Code Data, Individual Income Tax Statistics, Tax Year 2022.
Dollar amounts are in thousands of dollars, and counts are returns, not people. The files have no all-returns row, so a county or ZIP total is a sum over the income brackets. Charitable giving appears only for filers who itemize, so it understates giving among filers who take the standard deduction. The ZIP file is keyed by ZIP code and joined to ZCTAs approximately.
Used in Home value, income and redlining, Disability, retirement and Medicare, Places that beat the model.
LEHD Origin-Destination Employment Statistics
LEHD Origin-Destination Employment Statistics (LODES): residence area, workplace area and origin-destination tables
- Id
- lehd-lodes
- Producer
- U.S. Census Bureau, Longitudinal Employer-Household Dynamics (LEHD) program
- Unit
- census tract, 2020 boundaries (residence and workplace tables); pairs of counties (origin-destination table)
- Years
- 2002-2022
- Weights
- none (counts of jobs, with noise added for confidentiality)
- License
- Public domain (U.S. federal)
Cite as: U.S. Census Bureau, LEHD Program (2024). LODES8 Origin-Destination Employment Statistics, 2002-2022. https://lehd.ces.census.gov/data/
Counts jobs, not people: a worker with jobs in several places is counted once for each job. The Census Bureau adds noise to job counts for confidentiality, so counts for small tracts are imprecise. Earnings come only in broad bands of monthly earnings in current dollars, so the share of jobs in the top band rises over time with prices as well as with pay. Coverage follows state unemployment insurance records and federal employment, and some state files are stubs or incomplete in some years (Arkansas in 2002; Massachusetts through 2010; Alaska in 2019; Alaska and Michigan in 2022), so a state-year whose workplace table is nearly empty must be left out rather than read as jobs outside the state. Because second jobs count and the workplaces of employers with several sites are assigned by model, the share of jobs located outside the home county is higher than the share of workers who commute out of their county in a survey of workers. The residence table places a job at the worker's home tract and the workplace table at the job's tract.
Used in Work, wages and access to jobs.
Mapping Inequality (HOLC maps)
Mapping Inequality: Redlining in New Deal America, digitized HOLC residential security maps
- Id
- holc
- Producer
- Digital Scholarship Lab, University of Richmond
- Unit
- HOLC-graded neighborhood area (216 surveyed cities)
- Years
- 1930s-1940s maps (2023 digitization)
- Weights
- none (map polygons)
- License
- CC BY-NC-SA 4.0 (cite Nelson and coauthors)
Cite as: Nelson, R.K., Winling, L, Marciano, R., Connolly, N., et al. (2023). Mapping Inequality: Redlining in New Deal America. American Panorama, ed. Robert K. Nelson and Edward L. Ayers. https://dsl.richmond.edu/panorama/redlining/
Surveyors of the 1930s and 1940s graded neighborhoods from A (best) to D (hazardous). Coverage is the 216 cities HOLC surveyed, so a city that is absent was not surveyed, which does not mean it had no redlining. The areas do not follow census boundaries, so tract values come from a spatial overlay. The digitized grade field has anomalies (blank, E and F) that are set aside before any grade is used. The license allows non-commercial reuse with attribution and share-alike.
Used in Home value, income and redlining, Mortgage denials, Air pollution.
Medicare Geographic Variation
Medicare Geographic Variation Public Use File, county level, 2024
- Id
- cms-geo-variation
- Producer
- Centers for Medicare & Medicaid Services (CMS)
- Unit
- county
- Years
- 2024 (from the 2014-2024 file)
- Weights
- none (beneficiary counts; CMS price-standardizes payments)
- License
- Public domain (CMS, U.S. federal government work)
Cite as: Centers for Medicare & Medicaid Services. Medicare Geographic Variation Public Use File, 2014-2024 (county-level, 2024 cross-section).
The file covers Original Medicare beneficiaries, so people enrolled in Medicare Advantage plans are not counted. Payments are CMS price-standardized per-capita amounts that remove local wage and payment-rule differences, so they are not raw dollars of billed care; standardization does not adjust for beneficiaries' health, and a figure that uses CMS's risk-adjusted version says so. Small-county values are suppressed by CMS and left missing, never zero. The preventable-hospitalization columns are empty in this release. Connecticut appears under its planning-region codes.
Used in Disability, retirement and Medicare.
Opportunity Atlas
The Opportunity Atlas, adult outcomes of children by census tract and county
- Id
- opportunity-atlas
- Producer
- Opportunity Insights and the U.S. Census Bureau
- Unit
- census tract (2010 boundaries; a file interpolated to 2020 tracts is also used) and county
- Years
- 1978-1983 birth cohorts (2018 release)
- Weights
- none applied in the staged tables
- License
- Public use (Opportunity Insights and the U.S. Census Bureau)
Cite as: Chetty, Friedman, Hendren, Jones, Porter (2018). The Opportunity Atlas. U.S. Census Bureau CES public-use tract outcomes.
The outcomes are for children born in 1978 to 1983 and measured around 2014 to 2015, so a place describes where children grew up decades ago. Income ranks are shown as percentiles. The suffixes p25 and p75 name the parent income percentile, not the child's. The staged files carry no standard errors, so single-tract values are noisy in small tracts. Income ranks are staged by gender with races pooled, and by race with genders pooled at p25 only; teen birth is staged for women by race. Income ranks by race and gender together are not staged. The file on 2020 tracts is interpolated by land area, not population, and Connecticut tracts in the native 2010 file use legacy county codes.
Used in School discipline and incarceration, Race, income and mobility, Teen births, Boys and mobility (planned), Home value, income and redlining, Civic life, loneliness and partnership, Eight types of neighborhood, Immigrants and the gradient (planned), Places that beat the model, The anatomy of a deprivation index, A deprivation index at two zooms, Families of places.
Published estimates from cited studies
Estimates reported in the studies a chapter cites, quoted with their citation
- Id
- published-estimates
- Producer
- The authors of each cited study (see the chapter's references)
- Unit
- as reported in each study
- Years
- as reported in each study
- Weights
- as reported in each study
- License
- Quoted from published research for commentary
Cite as: See the references of the chapter that quotes the estimate.
A fact with this source is a number reported in a published study and quoted beside its citation, usually inside a causal box. The book does not recompute it, and its estimand names the study, the sample and the design. Such numbers describe the study's sample and period, not the places in this book.
Used in Smoking in pregnancy, Teen births, Mortgage denials, Multiple chronic conditions, Place and the length of a life, Mortality trends, 2018 to 2024, Violence and the life-expectancy gap, Life evaluation and daily feelings, Eight types of neighborhood.
Rural-Urban Continuum Codes
USDA Economic Research Service Rural-Urban Continuum Codes, 2023
- Id
- ers-rucc
- Producer
- U.S. Department of Agriculture, Economic Research Service
- Unit
- county
- Years
- 2023
- Weights
- none (classification)
- License
- Public domain (U.S. federal)
Cite as: U.S. Department of Agriculture, Economic Research Service (2023). Rural-Urban Continuum Codes, 2023. https://www.ers.usda.gov/data-products/rural-urban-continuum-codes
The codes run in 9 classes: 1 to 3 are metropolitan counties, grouped by the size of the metro area, and 4 to 9 are nonmetropolitan counties, grouped by urban population and by adjacency to a metro area. It is one classification as of 2023, not a series. It classifies counties, and the tract-level Rural-Urban Commuting Area codes are a different scheme, so they are never compared as equivalents. The 2 counties with no code are uninhabited American Samoa islands.
Used in College access, price and earnings, Work, wages and access to jobs, Overdose, alcohol and suicide, Suppressed counties in death data.
SSA beneficiaries by ZIP code
Social Security Administration, OASDI Beneficiaries by State and ZIP Code, 2024
- Id
- ssa-oasdi
- Producer
- U.S. Social Security Administration, Office of Retirement and Disability Policy
- Unit
- ZIP code (a ZIP code that crosses a state line appears once for each state)
- Years
- December 2024
- Weights
- none (full counts from the Master Beneficiary Record)
- License
- Public domain (U.S. federal)
Cite as: U.S. Social Security Administration, Office of Retirement and Disability Policy (2025). OASDI Beneficiaries by State and ZIP Code, 2024. Master Beneficiary Record, 100 percent data. https://www.ssa.gov/policy/docs/statcomps/oasdi_zip/
The file is keyed by USPS ZIP code and joined to ZCTAs through a crosswalk, which is approximate. Benefit totals are monthly dollars in thousands for December 2024, not annual and not per person, so a per-person figure divides by the beneficiary count. SSA controlled-rounds counts and leaves out ZIP codes below a disclosure threshold, so small cells are approximate. Connecticut ZIP codes do not join to the ZCTA geography used here, so Connecticut is not drawn on ZIP-level maps. Beneficiaries living abroad are excluded.
Used in Disability, retirement and Medicare.
Stanford Education Data Archive
Stanford Education Data Archive (SEDA), version 6.0, pooled achievement by county and school district
- Id
- seda
- Producer
- Stanford Center for Education Policy Analysis (The Educational Opportunity Project at Stanford University)
- Unit
- county and geographic school district, by student subgroup
- Years
- 2009-2019 (version 6.0)
- Weights
- none (empirical Bayes estimates; assessment counts are given)
- License
- CC BY 4.0 (Stanford CEPA); cite SEDA
Cite as: Reardon, S.F., et al. (2024). Stanford Education Data Archive (Version 6.0). Stanford University.
Scores are in standard-deviation units of the national cohort (the cs columns) or in grade levels (the gcs columns), and the scales are never mixed. Each estimate pools the tested grades over 2009 to 2019, so the file is one pooled vintage, not a yearly series. Virginia has reading but no math estimates at both county and district grain, DC is missing, and Hawaii has no districts; every SEDA map says so. Counties with few assessments have noisier estimates. Connecticut counties carry legacy codes.
Used in Test scores, class and race, Gender gaps in reading and math, Learning growth and starting level (planned).
USALEEP life expectancy
U.S. Small-area Life Expectancy Estimates Project (USALEEP), life expectancy at birth by census tract
- Id
- usaleep
- Producer
- National Center for Health Statistics (NCHS), CDC, with the Robert Wood Johnson Foundation and NAPHSIS
- Unit
- census tract (2010 boundaries); a file interpolated to 2020 tracts is also used
- Years
- 2010-2015
- Weights
- none (modeled estimates with standard errors)
- License
- Public domain (NCHS, CDC)
Cite as: National Center for Health Statistics. U.S. Small-area Life Expectancy Estimates Project (USALEEP): Life Expectancy Estimates File, 2010-2015.
Life expectancy at birth is a modeled small-area estimate from deaths in 2010 to 2015, and NCHS has not updated it, so every statement carries that period. Estimates exist for 67,148 tracts of 2010; a tract with no estimate is absent, never zero years. Institutional tracts (a prison or a dormitory) produce artifacts such as Roosevelt Island, so a single tract extreme is not reported on its own. The file on 2020 tracts is interpolated from 2010 tracts with land-area weights, not population weights, and its standard errors are good for mapping only.
Used in Place across an American life, Home value, income and redlining, Multiple chronic conditions, Place and the length of a life, Gun homicide and the causes of early death, Violence and the life-expectancy gap, Life evaluation and daily feelings, Eight types of neighborhood, Places that beat the model, Gradients, slopes and variance, The anatomy of a deprivation index, A deprivation index at two zooms, Families of places.
USDA Food Access Research Atlas
USDA Economic Research Service Food Access Research Atlas, 2019 release, census tracts of 2010
- Id
- food-access
- Producer
- U.S. Department of Agriculture, Economic Research Service
- Unit
- census tract (2010 boundaries)
- Years
- 2019 release (population from the 2010 census)
- Weights
- none (census counts)
- License
- Public domain (U.S. Department of Agriculture, Economic Research Service)
Cite as: U.S. Department of Agriculture, Economic Research Service (2021). Food Access Research Atlas, 2019.
The book uses this file only for the 2010 census population of each 2010 tract, which weights comparisons of Opportunity Atlas estimates on the same tracts. Its food-access flags are not used. Tracts follow 2010 boundaries, so Connecticut appears under its legacy tracts.
Used in School discipline and incarceration, Race, income and mobility, Teen births.