Chapter 4.1 · The deed and the ledger

Home value, income and redlining

Life expectancy is 2.2 years higher for each doubling of a ZIP code area's median home value, and tracts graded A on the 1930s HOLC maps have a life expectancy of 80.9 years against 76.8 years in tracts graded D, for deaths in 2010 to 2015.

Author: Vishal Singh
Affiliation: NYU Stern
Published: 9 October 2026

Life expectancy at birth is 2.2 (95% CI 2.1 to 2.3) years higher for each doubling of the median home value across 19,655 ZIP code areas when each area counts once, and 2.3 (95% CI 2.2 to 2.5) years higher when areas count in proportion to their population. Life expectancy is the USALEEP estimate for deaths in 2010 to 2015 and home values are from the American Community Survey for 2019 to 2023. The slope describes places. It does not say that a household that moved to a costlier area would live longer.

The chapter then turns to two other records that follow the address: the shares of tax returns that claim the earned income tax credit or report a high income, and the grades that the Home Owners' Loan Corporation gave neighborhoods in the 1930s. Every comparison is between places (ZIP code areas, counties and census tracts), and none describes individual households. Eviction filings and the housing cost burden of renters are in Chapter 4.4, Renting, affordability and eviction.

Home values rise with life expectancy

Each ZIP code area gets the population-weighted mean life expectancy of the census tracts for which it is the largest-overlap ZCTA, with institutional tracts such as prisons and nursing facilities left out. Across the areas, life expectancy has a correlation of +0.59 (95% CI 0.56 to 0.61) with the logarithm of the home value and +0.53 (95% CI 0.50 to 0.55) with the home value in dollars. It moves more closely with the logarithm, so the slope is stated per doubling of the home value.

Ranking ZIP code areas by home value and averaging life expectancy within twentieths of the areas, weighted by population, gives 74.1 (95% CI 73.6 to 74.8) years in the lowest twentieth and 82.6 (95% CI 82.2 to 82.9) years in the highest, a difference of 8.4 (95% CI 7.7 to 9.0) years. Economists call the price of a bundled feature a hedonic price: the price of a house is the sum of what buyers pay for each part of the address, such as bedrooms, schools and the commute.[1] Life expectancy is not a feature that a buyer can see or purchase, so the slope is not a price that anyone pays. It records that homes cost more where the people nearby live longer, and the data do not separate what an area provides from who can afford to buy there.

Inverting the slope within a metro gives a local exchange rate. In each of the 37 metros with at least 60 ZIP code areas and 1.5 million residents in the sample, the median home value of its ZIP code areas divided by the within-metro slope of life expectancy on the logarithm of home value is the housing dollars that separate two areas one year of life apart near the metro median. A year costs $18 thousand (95% CI $16 thousand to $22 thousand) in Cincinnati, the metro with the lowest point estimate, and $140 thousand (95% CI $120 thousand to $169 thousand) in the San Francisco-Oakland metro, the highest, a ratio of 7.6 (95% CI 6.1 to 9.7) between those two estimates (Figure 1). That ratio compares the extremes of 37 noisy estimates, and its interval holds those metros fixed, so it understates how far apart the extremes are. The point estimate is above $100 thousand in 4 metros, New York, Los Angeles, San Diego and San Francisco-Oakland, and the intervals for New York and San Diego include that amount. Cleveland has the largest within-metro correlation between life expectancy and the logarithm of home value, +0.91 (95% CI 0.86 to 0.94), and a year costs $22 thousand (95% CI $18 thousand to $25 thousand) there. The interval for Detroit, the next largest, overlaps Cleveland's, so the order is by point estimate.

Figure 1

Life expectancy is 2.2 years higher for each doubling of ZIP-area home value, and a year of life costs $18 thousand in housing dollars in Cincinnati and $140 thousand in San Francisco-Oakland

Life expectancy at birth (deaths in 2010 to 2015) against median home value (ACS 2019 to 2023) across ZIP code areas, and the housing dollars per year of life in each large metro

Life expectancy is higher in ZIP code areas with higher home values, a correlation of +0.59 with the log of home value

Life expectancy at birth in years against median home value in dollars on a log scale, one dot per ZIP code area (random sample), with population-weighted means for twentieths of ZIP areas

Loading data

Dots drift up from left to right, and the markers for twentieths of ZIP areas by home value rise at every step, from 74.1 years in the lowest twentieth to 82.6 years in the highest. The dashed line is fitted across all 19,655 ZIP code areas, each counted once, and has slope 2.2 years per doubling of home value.

Life expectancy rises across the whole range of home values, and a year of life costs the fewest housing dollars in Cincinnati, Cleveland and Columbus and the most in San Francisco-Oakland, San Diego and Los Angeles. The first tab plots a random sample of 5,000 ZIP code areas, with 3 areas outside the plotted range left out of the sample; the second tab gives, for each of 37 metros, the median ZIP-area home value divided by the within-metro slope of life expectancy on the log of home value.Note: The sample is 19,655 ZIP Code Tabulation Areas, built from the tracts that have a life expectancy estimate, with institutional tracts such as prisons and nursing facilities left out. Connecticut has no ZCTAs in the tract file and is absent. Home values are top-coded by the Census Bureau at about one million dollars.Note: The dots are drawn at random from the ZIP code areas inside the plotted range with seed 12345. Every estimate in the chapter uses all ZIP code areas.Note: Twentieths hold equal numbers of ZIP code areas. Intervals on the markers resample whole counties; intervals in the second tab resample the ZIP code areas of one metro and treat them as independent, so they are narrower than they should be. A metro enters with at least 60 ZIP code areas and 1.5 million residents in the sample.Note: Life expectancy is the estimate for deaths in 2010 to 2015 and home values are for 2019 to 2023, so the two describe different years.Source: USALEEP life expectancy 2010-2015 (non-institutional tracts, area-weighted onto 2020 tracts) and ACS 2019-2023 ZCTA median home value; population-weighted twentieths, unweighted fitted line, modeled estimates

The exchange rate follows the price level more than the strength of the link. Across the metros, dollars per year of life correlate +0.91 (95% CI 0.81 to 0.96) with the metro's median ZIP-area home value, so a year is cheap where homes are cheap and dear where homes are dear. Part of that correlation is built in, because dollars per year is the median home value divided by the slope. The figures say more about what housing costs than about how steeply longevity rises with value. Read them as orders of magnitude, because the ZIP code areas of one metro resemble one another and the interval for a single metro is narrower than the uncertainty.

Income brackets change shape with deprivation

A median summarizes a distribution of incomes with one number, and income tax returns describe it by bracket. Across 3,129 counties, with Connecticut's planning regions standing in for its counties, the share of individual income tax returns for tax year 2022 that claim the earned income tax credit rises from 12.0% (95% CI 10.4% to 13.2%) in the least deprived tenth of counties, ranked by the Area Deprivation Index, to 28.9% (95% CI 26.4% to 31.2%) in the most deprived tenth (Figure 2). The share of returns with an adjusted gross income of $200,000 or more falls from 11.1% (95% CI 10.3% to 11.9%) to 1.9% (95% CI 1.7% to 2.1%). Nationally, 14.9% of returns claim the credit and 7.8% are in the top bracket. Across counties weighted by population, the EITC share correlates +0.57 (95% CI 0.43 to 0.72) with deprivation and the top-bracket share correlates −0.74 (95% CI −0.82 to −0.69).

The ranking and the shares overlap in what they measure, because the Area Deprivation Index is built partly from income and poverty, so some of the gradient is built in. Tenths hold equal numbers of counties, not equal numbers of people: the least deprived tenth holds 130.1 million people and the most deprived tenth 3.8 million.

The brackets change shape across the range and not only in their average. In the least deprived fifth of counties, 26.7% of returns report an adjusted gross income under $25,000 and 10.0% report $200,000 or more. In the most deprived fifth the shares are 37.6% and 2.3%. The lowest income group is the largest of the five groups in every fifth of counties, and the data do not say who is in it.

Figure 2

The share of tax returns that claim the EITC rises from 12.0% to 28.9% across tenths of counties by deprivation, while the share with adjusted gross income of $200,000 or more falls from 11.1% to 1.9%

Shares of individual income tax returns by tenth of counties on the Area Deprivation Index, and the distribution of returns across five income groups by fifth of counties, tax year 2022

Across tenths of counties, the EITC share rises with deprivation and the top-bracket share falls

Percent of individual income tax returns, county-population-weighted mean, tax year 2022, by tenth of counties on the Area Deprivation Index (1 = least deprived)

The EITC line rises at nearly every step and the top-bracket line falls at every step; the two lines start at nearly the same level in the least deprived tenth and end far apart in the most deprived. Bands are 95% intervals from a bootstrap over states.

Show data

Data for The share of tax returns that claim the EITC rises from 12.0% to 28.9% across tenths of counties by deprivation, while the share with adjusted gross income of $200,000 or more falls from 11.1% to 1.9%, By deprivation tenth
conf_high	conf_low	counties	decile	series	share
13.19	10.39	312	1	Returns claiming the EITC	12.03
16.07	12.76	313	2	Returns claiming the EITC	14.75
17.66	14.05	313	3	Returns claiming the EITC	15.93
18.40	15.68	313	4	Returns claiming the EITC	17.01
20.10	16.71	313	5	Returns claiming the EITC	18.34
22.56	17.61	313	6	Returns claiming the EITC	20.18
21.23	17.57	313	7	Returns claiming the EITC	19.32
26.24	18.63	315	8	Returns claiming the EITC	22.40
26.09	21.03	312	9	Returns claiming the EITC	23.69
31.16	26.41	312	10	Returns claiming the EITC	28.93
11.87	10.34	312	1	Returns with adjusted gross income of $200,000 or more	11.06
8.08	6.73	313	2	Returns with adjusted gross income of $200,000 or more	7.46
6.31	5.30	313	3	Returns with adjusted gross income of $200,000 or more	5.91
5.24	4.55	313	4	Returns with adjusted gross income of $200,000 or more	4.88
4.38	3.66	313	5	Returns with adjusted gross income of $200,000 or more	4.06
3.69	3.20	313	6	Returns with adjusted gross income of $200,000 or more	3.43
3.20	2.80	313	7	Returns with adjusted gross income of $200,000 or more	3
2.94	2.50	315	8	Returns with adjusted gross income of $200,000 or more	2.69
2.64	2.24	312	9	Returns with adjusted gross income of $200,000 or more	2.43
2.13	1.73	312	10	Returns with adjusted gross income of $200,000 or more	1.92

The EITC share rises and the top-bracket share falls across the tenths, and the bars show the same pattern for five income groups: the lowest group grows and the highest group shrinks toward the most deprived fifth. The first tab plots population-weighted means of county shares with 95% intervals from a bootstrap over states; the second tab pools the returns of the counties in each fifth.Note: Tenths and fifths hold equal numbers of counties, not equal numbers of people, so the most deprived tenth holds far fewer returns than the least deprived. The Area Deprivation Index is built from income, poverty, education and housing measures, so a gradient of income-based shares against it is partly built in.Note: The IRS file covers people who file a return. The EITC share is the share of returns that claim the credit, which depends on eligibility and on filing, and the top-bracket share is the share of returns with adjusted gross income of $200,000 or more. Both are counts of returns with no sampling error; the intervals in the first tab come from differences between states.Note: 14.9% of all returns claim the EITC and 7.8% are in the top bracket. Counties with no Area Deprivation Index are left out; Puerto Rico is not in the file.Source: IRS Statistics of Income county file, tax year 2022, and Area Deprivation Index 2023; population-weighted county means, full counts of returns

Both shares exceed the national shares in only 15 of the 447 counties that have at least 150,000 residents, among them Fulton County and DeKalb County in Georgia and Essex County in New Jersey. In most of the other large counties one tail is heavier than the nation's and the other lighter.

The shares also line up with the mobility of children from low-income families (see Chapter 3.3, Race, income and mobility, which defines the income rank and owns that finding). In a county-population-weighted regression of standardized variables, the top-bracket share alone has a coefficient of 0.41 (95% CI 0.33 to 0.50). With both shares and the deprivation index in the model, the coefficient on the top-bracket share is −0.03 (95% CI −0.27 to 0.20), an interval that includes zero, and the coefficient on the EITC share is −0.49 (95% CI −0.75 to −0.23), which stays away from zero. The two shares correlate −0.65 (95% CI −0.70 to −0.60) across counties, so the regression cannot split their common signal cleanly. All of this is association across counties.

Grades drawn in the 1930s and neighborhoods today

Between 1935 and 1940 the Home Owners' Loan Corporation drew residential security maps for cities across the country and graded neighborhoods from A, labeled best, to D, labeled hazardous.[2] The area descriptions that came with the maps show that the race and national origin of residents were among the factors in a grade, next to the housing stock and incomes.[3] The maps have long been out of use, and the grades still line up with today's neighborhoods.

Overlaying the graded polygons on 2020 census tracts gives 10,952 tracts in 222 HOLC cities and city areas where graded polygons cover at least 50% of the tract. Each tract takes the grade that covers most of its area. Life expectancy at birth, for deaths in 2010 to 2015, averages 80.9 (95% CI 80.1 to 81.6) years in grade A tracts and 76.8 (95% CI 75.5 to 77.8) years in grade D tracts, a gap of 4.1 (95% CI 3.3 to 5.1) years with tracts counted in proportion to their population (Figure 3). The gap is 3.6 (95% CI 2.6 to 4.9) years when graded polygons must cover at least 90% of a tract, so it moves with the matching rule and stays positive. Grade A tracts are in cities whose graded tracts have a lower average life expectancy than the cities of grade D tracts, so comparing tracts within the same city, with a separate baseline for each city (fixed effects), widens the gap to 5.8 (95% CI 5.3 to 6.4) years.

Other measures step in the same direction. The adult income rank of children from low-income families, from the Opportunity Atlas for children born in 1978 to 1983, averages 46.3 (95% CI 45.3 to 47.1) in grade A tracts and 37.9 (95% CI 36.3 to 39.2) in grade D tracts, in percentile points of the national household income distribution. Homeownership is 67.8% (95% CI 65.0% to 70.6%) of households in grade A tracts and 35.3% (95% CI 31.7% to 39.7%) in grade D tracts, and the mean of tract median household incomes is $127 thousand (95% CI $115 thousand to $138 thousand) and $71 thousand (95% CI $63 thousand to $78 thousand) in the two grades. These comparisons describe neighborhoods today, and the two grades differ in housing, income and population from before the maps were drawn.

Figure 3

Tracts graded A on the 1930s HOLC maps have a life expectancy of 80.9 years, against 76.8 years in tracts graded D, for deaths in 2010 to 2015

Tract outcomes by 1930s HOLC grade: means of five measures, the spread of life expectancy, and the gap between grade A and grade D tracts city by city

Life expectancy, adult income rank, homeownership and household income are lower in each step from grade A to grade D

Population-weighted mean of tract values by HOLC grade, each panel in its own units

Life expectancy at birth (years)

Adult income rank of children from low-income families (percentile)

Homeownership rate (% of households)

Median household income (thousands of dollars)

Area Deprivation Index (national percentile, 100 = most deprived)

Mean of tract values (units in each panel title)

Life expectancy, adult income rank, homeownership and household income fall at each step from grade A to grade D, and the intervals for grades A and D do not overlap in any of those panels. The deprivation index is close to the same value for grades B, C and D, and the interval for grade A overlaps theirs.

Show data

Data for Tracts graded A on the 1930s HOLC maps have a life expectancy of 80.9 years, against 76.8 years in tracts graded D, for deaths in 2010 to 2015, Means by grade
conf_high	conf_low	grade	label	mean	n	outcome
81.57	80.13	A	Grade A	80.93	629	Life expectancy at birth (years)
79.81	78.30	B	Grade B	79.10	2,133	Life expectancy at birth (years)
78.74	76.88	C	Grade C	77.95	4,710	Life expectancy at birth (years)
77.79	75.50	D	Grade D	76.84	2,671	Life expectancy at birth (years)
47.15	45.26	A	Grade A	46.26	672	Adult income rank of children from low-income families (percentile)
44.14	41.99	B	Grade B	43.11	2,245	Adult income rank of children from low-income families (percentile)
42.56	39.57	C	Grade C	41.20	5,051	Adult income rank of children from low-income families (percentile)
39.15	36.27	D	Grade D	37.93	2,974	Adult income rank of children from low-income families (percentile)
70.58	64.96	A	Grade A	67.81	672	Homeownership rate (% of households)
56.70	48.44	B	Grade B	52.55	2,246	Homeownership rate (% of households)
47.75	39.33	C	Grade C	43.45	5,052	Homeownership rate (% of households)
39.73	31.66	D	Grade D	35.32	2,973	Homeownership rate (% of households)
138.4	114.9	A	Grade A	127.4	669	Median household income (thousands of dollars)
96.19	82.24	B	Grade B	89.54	2,236	Median household income (thousands of dollars)
78.97	68.80	C	Grade C	74.16	5,015	Median household income (thousands of dollars)
78.02	62.77	D	Grade D	71.12	2,920	Median household income (thousands of dollars)
38.43	25.53	A	Grade A	31.55	670	Area Deprivation Index (national percentile, 100 = most deprived)
49.21	33.19	B	Grade B	40.78	2,241	Area Deprivation Index (national percentile, 100 = most deprived)
52.72	33.35	C	Grade C	41.77	5,040	Area Deprivation Index (national percentile, 100 = most deprived)
52.98	31.56	D	Grade D	40.64	2,953	Area Deprivation Index (national percentile, 100 = most deprived)

On four of the five measures the dots step down from grade A to grade D, the deprivation index is close to the same value for grades B, C and D, the distributions of life expectancy overlap across grades, and the grade A dot lies to the right of the grade D dot in every city. The tabs plot population-weighted means with 95% intervals from a bootstrap over counties, the interquartile range of tract life expectancy, and the two grades' means for each city with enough tracts.Note: A tract enters when graded polygons (A to D) cover at least 50% of its area and takes the grade that covers the most of it. The comparison is descriptive: the grades followed the neighborhoods' housing, incomes and the race and national origin of residents, so differences today do not isolate the effect of the maps.Note: The tracts are in 222 HOLC cities and city areas, matched by overlay. With a stricter rule that graded polygons cover at least 90% of a tract the life-expectancy gap between grades A and D is 3.6 years.Note: Life expectancy is USALEEP for deaths in 2010 to 2015, area-weighted onto 2020 tracts, with institutional tracts left out. The adult income rank is the Opportunity Atlas rank of children born 1978 to 1983 to parents at the 25th income percentile, interpolated from 2010 to 2020 tracts by area. Tract values are weighted by ACS 2019-2023 population.Source: Mapping Inequality HOLC maps (1935-1940) overlaid on 2020 tracts; USALEEP 2010-2015, Opportunity Atlas (1978-1983 birth cohorts), ACS 2019-2023 and ADI 2023; population-weighted, modeled estimates

The Area Deprivation Index is close to the same value in grade B, C and D tracts (41, 42 and 41) while life expectancy falls from grade B to grade D. Several of the census measures behind the index are home values, rents and mortgage payments, so places with high housing costs read as less deprived. These data do not test whether that explains the flat pattern.

The grades shift the middle of the distribution and do not sort tracts into separate groups. In 75% of random pairs of one grade A tract and one grade D tract, drawn in proportion to population, the grade A tract has the higher life expectancy, and in the remaining pairs the grade D tract does. The interquartile ranges of neighboring grades overlap, and so do those of grades A and D.

Within cities the gap has the same sign wherever it can be measured. Among the 27 cities with at least 5 grade A and 10 grade D tracts that have a life expectancy, the gap is positive in every one, with a median of 5.7 years and a range from 3.4 to 10.1 years between the 5th and 95th percentiles of cities. A city's grade A mean rests on few tracts, so single cities are noisy.

Causal evidenceBoundary design with propensity-score weighting; comparison of cities around a population cutoff

Whether the HOLC maps changed the neighborhoods they graded

The comparisons above cannot separate the effect of the maps from the neighborhoods the maps graded, because surveyors graded areas on housing, incomes and race. Aaronson, Hartley and Mazumder study the maps with decennial census data.[3] They compare changes over time in outcomes between the two sides of a grade border, within a few blocks, using census data from before the maps as well as after. Because borders already differed before the maps, they weight comparison borders by propensity scores so that the earlier patterns in race and housing match, and they separately use borders that were unlikely to have been drawn along neighborhood differences. A third analysis compares cities just below and just above the population cutoff that decided whether a city was mapped.

They find that the maps reduced homeownership rates, house values and rents and increased racial segregation in later decades, and the comparison of cities around the cutoff gives broadly similar results. The paper does not study life expectancy or children's incomes, and its border estimates are local to areas beside a border.

What these data cannot show

Every comparison in this chapter describes places. A ZIP code area, a county or a tract is not a person, and a pattern across places need not hold for the people inside them (ecological inference). The residents of a ZIP code area differ from movers in income, health and age, and the data cannot separate selection from what an area provides.

The price of a year joins life expectancy for deaths in 2010 to 2015 to home values for 2019 to 2023, so the two describe different years. The life expectancy estimates are modeled for small areas and area-weighted from 2010 to 2020 tracts, and Connecticut has no ZIP code areas in the tract file, so it is absent. The dollar figures divide a price level by a slope that is itself estimated with error.

The IRS file covers people who file a return, so the shares are shares of returns and leave out people who do not file. The EITC share depends on eligibility and on whether eligible households claim the credit, and the top-bracket share is a share of returns, not of people. Connecticut is absent from the mobility regressions because its planning-region codes do not match the Atlas's counties.

The grades followed the neighborhoods' housing, incomes and populations in the 1930s, so the staircase today combines what the neighborhoods were with what happened after. The boundary study in the box above finds that the maps themselves mattered for housing outcomes and segregation next to borders, and it does not show that the grades explain today's differences in life expectancy. Tracts are matched to grades by area, the Atlas ranks are interpolated by area from 2010 tracts, and the maps cover the cities that HOLC surveyed, so a city that is absent was not surveyed, which does not mean it had no redlining.

Data sources

USALEEP life expectancyLife expectancy at birth by tract, deaths in 2010 to 2015, area-weighted onto 2020 tracts and averaged to ZIP code areasYears: 2010-2015 · Weights: ZCTA population; each ZIP code area once for the unweighted slope · License: Public domain (NCHS, CDC)
American Community SurveyACS 2019 to 2023: median home value by ZCTA, tract population, homeownership, household income and renter householdsYears: 2019-2023 (5-year) · Weights: 2019-2023 tract population · License: Public domain (U.S. Census Bureau)
IRS Statistics of IncomeIRS Statistics of Income county file, tax year 2022: returns, returns claiming the earned income tax credit and returns by adjusted gross income bracketYears: tax year 2022 · Weights: none (full counts of returns) · License: Public domain (IRS Statistics of Income)
Area Deprivation IndexArea Deprivation Index 2023, national rank, tracts averaged to countiesYears: 2023 release (version 4.0.1; built on ACS 2019-2023) · Weights: ACS 2019-2023 tract population · License: Free for non-commercial use; cite the Neighborhood Atlas and Kind and Buckingham
Opportunity AtlasAdult household income rank of children born 1978 to 1983 to parents at the 25th income percentile, county and 2020-tract filesYears: 1978-1983 birth cohorts (2018 release) · Weights: county population; tract population · License: Public use (Opportunity Insights and the U.S. Census Bureau)
Mapping Inequality (HOLC maps)Mapping Inequality HOLC residential security maps, 1935 to 1940, graded polygons overlaid on 2020 tractsYears: 1930s-1940s maps (2023 digitization) · Weights: ACS 2019-2023 tract population · License: CC BY-NC-SA 4.0 (cite Nelson and coauthors)

All data sources

Methods

This chapter compares places (ZIP code areas, counties and census tracts), not households, and each comparison has its own sample, weights and interval method, set out below. Life expectancy is for deaths in 2010 to 2015, while home values and the tract measures are from the ACS 2019-2023, so the outcome and the predictor come from different periods. The metro exchange rate covers only metros with many ZIP code areas and residents in the sample, the HOLC comparison covers only tracts in graded cities, and the Area Deprivation Index that ranks counties is built partly from income, so some of the income gradient is built in.

ZIP code areas, life expectancy and home value

Life expectancy is the USALEEP estimate for deaths in 2010 to 2015 on 2020 tracts (area-weighted from 2010 tracts), with institutional tracts, where at least 25 percent of residents live in group quarters, left out.

A ZIP Code Tabulation Area gets the population-weighted mean of the tracts whose largest-overlap ZCTA it is, and the ACS 2019-2023 median home value.

Twentieths of ZIP code areas hold equal numbers of areas.

Figure explained here: Figure 1

Years per doubling and the metro exchange rate

Years per doubling of home value is the slope of ZCTA life expectancy on the base-2 log of home value, with each ZCTA counted once and weighted by population, from the book's shared routine, with intervals from a bootstrap over counties.

The metro exchange rate is the median of the ZIP areas' median home values divided by the within-metro slope of life expectancy on the natural log of home value, for metros with at least 60 ZIP code areas and 1.5 million residents in the sample, with intervals from a bootstrap over ZIP code areas.

Figure explained here: Figure 1

Income shapes and the county ranking

The IRS county file for tax year 2022 gives returns in eight adjusted-gross-income brackets; the EITC share is returns claiming the credit divided by all returns, and the top-bracket share is returns with adjusted gross income of $200,000 or more divided by all returns.

Counties are ranked on the Area Deprivation Index (tract ranks averaged with ACS 2019-2023 population weights); tenths and fifths hold equal numbers of counties. Means and correlations are weighted by county population, with intervals from a bootstrap over states.

Mobility regressions use weighted z-scores of the Opportunity Atlas county rank, with errors clustered on state; Connecticut drops out because its planning-region codes do not match the Atlas's counties.

Figure explained here: Figure 2

HOLC grades and tract outcomes

Graded polygons (A to D) are intersected with 2020 tracts in an equal-area projection; a tract with at least 50 percent of its area graded takes the grade covering most of it, and the city of the polygon covering most of it.

Tract outcomes are weighted by ACS 2019-2023 population, with intervals from a bootstrap over counties; the within-city gap uses a weighted regression with a separate baseline for each city.

Figure explained here: Figure 3

Bootstrap settings and city comparisons

The city comparison resamples tracts within a city and grade and treats them as independent.

Bootstrap intervals use 2,000 replicates and seed 12345.

References

Rosen, Sherwin (1974). Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition. Journal of Political Economy 82(1): 34-55. doi:10.1086/260169
Nelson, Robert K.; Winling, LaDale; Marciano, Richard; Connolly, Nathan; et al. (2023). Mapping Inequality: Redlining in New Deal America. American Panorama, ed. Robert K. Nelson and Edward L. Ayers, Digital Scholarship Lab, University of Richmond. dsl.richmond.edu/panorama/redlining/
Aaronson, Daniel; Hartley, Daniel; Mazumder, Bhashkar (2021). The Effects of the 1930s HOLC "Redlining" Maps. American Economic Journal: Economic Policy 13(4): 355-392. doi:10.1257/pol.20190414

How to cite

@misc{singh2026homevalueincomeandredlining,
  author = {Singh, Vishal},
  title = {Home value, income and redlining},
  year = {2026},
  url = {https://vishalsingh.org/zipcode/home-value-income-and-redlining},
  note = {In ZIP Code Destiny. Published 9 October 2026.}
}

Reading paths

Money and the institutions that follow the map