# Methodology and source notes

HousingStabilityIQ renter housing costs study | Prepared November 10, 2026

## Data and reference period

The main findings use the U.S. Census Bureau’s 2025 American Community Survey one-year detailed tables B25070, B25072 and B25074. The source files were retrieved and frozen on November 9, 2026. The estimates were released October 29, 2026 and describe the 2025 survey period. They are not a November 2026 measurement of rent.

The research also recalculated the corresponding 2024 tables to assess annual comparisons. Census permits comparisons of gross rent as a percentage of household income between these releases. The relevant table universes, labels and row structures matched across years. Table availability and other methodological conditions did not remain identical. Readable Census table-definition pages, downloadable raw files and release documentation appear in the source list below.

## Geographic scope

The study roster contains Florida, Georgia and North Carolina and 53 full metropolitan statistical areas touching at least one of those states, for 56 geographies. It is not a South Carolina statewide study. Cross-state metros retain their full Census-defined areas. Charlotte includes its South Carolina counties; other cross-state metros in the roster are likewise not restricted to their Florida, Georgia or North Carolina portions. The findings do not describe city limits.

Both releases use the July 21, 2023 metropolitan delineations. State and county legal-boundary reference dates advance from January 1, 2024 to January 1, 2025. Matching metro codes, names and county membership therefore does not prove identical geographic footprints.

Census boundary-change notes document Union County, Georgia, gaining territory from Lumpkin County effective May 1, 2024. Lumpkin is an Atlanta metro county; Union is outside the metro. Atlanta’s published year comparison is consequently not a strictly constant-footprint comparison. The amount of transferred territory and the number of affected households have not been quantified here. The within-2025 comparisons describe the published 2025 metro. The transfer is within Georgia and does not change the state’s external boundary. The documentary review found no additional relevant full-metro exception, but it was not a polygon-level geographic audit.

## Definitions and denominator treatment

The underlying universe is renter-occupied housing units, referred to here as renter households. Counts are estimates of households in the population, not numbers of survey respondents or individual people.

Gross rent includes contract rent plus applicable utilities and fuels. Burden is defined as gross rent of at least 30% of household income; severe burden is at least 50%. These thresholds include households exactly at 30% or 50%.

All shares use a denominator equal to the relevant total minus “Not computed.” This category includes no-cash-rent units and households with zero or net-loss income. Such cases cannot be treated as unburdened. The excluded counts are disclosed below so readers can distinguish the calculable subset from the full population in each group.

Householder age describes the person in whose name the home is owned or rented. It does not classify every person in the household and does not identify retirement status. B25072 ends with a category of 35% or more; it cannot establish severe burden by age. B25074’s dollar categories describe nominal annual household income, not individual earnings, constant purchasing power, area-median-income bands or the same households followed over time.

## Focal estimates and exclusions

All rows below refer to 2025. “Burdened” means at least 30% for the Florida and Atlanta rows, and at least 50% for the three North Carolina metro rows. Margins of error are at the 90% level. Percentage margins are approximate and expressed in percentage points.

| Group | Burdened households | Calculable households | Not computed | Burden share and MOE |
|---|---:|---:|---:|---:|
| Florida householder 65 or older | 353,397 | 491,734 | 57,164 | 71.9% ± 3.3 pp |
| Florida householder ages 35–64 | 805,761 | 1,386,842 | 80,105 | 58.1% ± 1.3 pp |
| Atlanta household income $35,000–$49,999 | 84,346 | 94,688 | 4,015 | 89.1% ± 3.8 pp |
| Atlanta household income $50,000–$74,999 | 109,807 | 158,457 | 2,151 | 69.3% ± 3.3 pp |
| Charlotte–Concord–Gastonia NC–SC | 88,029 | 376,269 | 19,387 | 23.4% ± 1.6 pp |
| Raleigh–Cary NC | 46,635 | 214,054 | 7,930 | 21.8% ± 2.5 pp |
| Greensboro–High Point NC | 23,936 | 110,594 | 6,128 | 21.6% ± 2.3 pp |

The full-precision downloadable data include total, numerator and denominator estimates and their margins of error. Printed estimates are rounded only for display. Counts have sampling uncertainty even where the prose omits a count margin for readability.

## Table cells used

The three-digit suffixes below identify rows within the named Census detailed table. The linked table-definition pages explain each cell’s label. Numerators are divided by their listed denominators and multiplied by 100.

- Overall burden: B25070 rows 007, 008, 009 and 010 summed; denominator row 001 minus 011.
- Overall severe burden: B25070 row 010; denominator row 001 minus 011.
- Householder age 65 or older burden: B25072 rows 027 and 028 summed; denominator row 023 minus 029.
- Householder ages 35–64 burden: B25072 rows 020 and 021 summed; denominator row 016 minus 022.
- Household income $35,000–$49,999 burden: B25074 rows 033 through 036 summed; denominator row 029 minus 037.
- Household income $50,000–$74,999 burden: B25074 rows 042 through 045 summed; denominator row 038 minus 046.

## Uncertainty and comparison tests

Published Census margins of error are at the 90% confidence level. For a sum or difference of cells, the study approximates the margin of error as the square root of the sum of the squared component margins. This also supplies the denominator margin for total minus not-computed cases.

For a share, the study uses Census’s subset-proportion approximation. Let p be the numerator divided by the denominator. Square the numerator margin of error, subtract p squared times the denominator margin squared, take the square root, and divide by the denominator. Multiply by 100 for a percentage-point margin. If the quantity under the square root is negative, use the corresponding ratio formula with a plus sign instead of the minus sign and flag the result. Florida’s 2025 age-65-or-older estimate uses this fallback; its 3.3-point margin is retained.

Standard errors equal the 90% margins divided by 1.645. Difference tests divide the estimated difference by the square root of the sum of the squared standard errors, using a two-sided normal test. The fixed 50% threshold has no sampling margin. These calculations approximate variances and assume zero covariance across compared groups or years. They do not use survey replicate weights. The subtraction in the subset formula and the fallback formula switch mean a change in the calculated margin is not, by itself, evidence of a change in survey quality.

The study uses Holm adjustment at a 10% significance level within each stated comparison family. This reduces the risk of false findings when several tests are considered together, conditional on the validity of the individual approximate tests. It does not remove survey bias, composition changes or uncertainty about covariance. These are transparent analytical families, not a prospectively registered experiment.

The original 2025 descriptive families were retained in full:

- Florida householder-age gaps: 20 computable comparisons across the state and eligible full metros. The state’s 13.8-point gap passes; its approximate, unadjusted 90% margin is 3.6 points.
- Georgia and Atlanta income-band majority comparisons: four tests, covering two income bands in each geography. Both Atlanta shares pass the comparison with 50%.
- Charlotte, Raleigh and Greensboro severe burden: six tests, comprising all three pairwise share differences and all three pairwise count differences. All count differences pass; none of the share differences passes.

The chart intervals are marginal, unadjusted 90% intervals for individual estimates, not simultaneous intervals for the full test families. The comparison conclusions come from the stated adjusted tests, not from visually judging whether error bars overlap. A finding that is not statistically distinguishable does not establish equality.

## Sensitivity and annual comparisons

A conservative diagnostic applied the plus-sign ratio formula to every share in the unchanged 20-test Florida and four-test Georgia/Atlanta families. The Florida state age gap and both Atlanta majority findings still passed. The diagnostic margins were 3.9 points for the Florida gap and 10.2 and 7.3 points for the Atlanta shares in the lower and higher income bands, respectively. This is an alternative uncertainty calculation, not an official replacement for the subset method or a guaranteed design-based upper bound.

For 2024–2025 comparisons, none of the seven focal share changes passed its seven-test Holm adjustment, and none of the three North Carolina metro severe-burden count changes passed its three-test adjustment. These results do not establish that the underlying changes were zero. Atlanta also has the boundary limitation described above.

For transparency, a broader exploratory screen considered 483 computable share changes across the study roster. Under the standard approximation, two declines for householders ages 15–24, in North Carolina and Durham–Chapel Hill, passed that screen’s adjustment. Neither passed when the plus-sign diagnostic was applied to all 483 comparisons. These state and metro populations overlap, the Durham baseline had a relatively wide margin, and the survey’s age controls changed. The study therefore does not present these as established annual trends or public headline findings. They require further variance and household-composition analysis. They do not describe all young people or students; group quarters such as college dormitories are outside the renter-occupied housing-unit universe.

No independent replicate-weight variance estimate was computed. The Census variance-replicate product listed at review covers selected five-year tables, not these 2025 one-year tables. A public-use microdata reconstruction would be a separate analysis with different geography and potential tabulation differences, rather than an exact replacement for the published metro cells.

## Coverage and the 2025 survey

Of 56 geographies, B25070 was available for all 56 in both years. B25072 was available for 50 in 2024 and 51 in 2025; B25074 for 31 and 25, respectively. Of 728 potential geography-and-metric rows per year, 529 were computable in 2024 and 491 in 2025, with 483 matched share changes. Unavailable tables are not zero burden. No older five-year estimate was substituted, and missing, blank, suppressed or negative sentinel inputs were not converted to zero. The research did not establish why any particular table became unavailable.

Census’s 2025 user notes identify three relevant changes:

- A six-week appropriations lapse disrupted August–November collection panels, affecting sample size, response and precision. Each year’s published margins were used.
- Revised population controls incorporated age corrections and boundary updates. Weighting changes may affect within-year subgroup estimates as well as annual comparisons. Their specific effect on this study’s age gap has not been quantified.
- Household record swapping was discontinued and some release rules changed. Population thresholds and data-quality rules remained. These changes should not be assumed to explain an individual table’s absence without evidence.

The estimates are descriptive survey findings. They do not establish causes or measure eviction, homelessness, current asking rents, retirement status or policy effects.

## Reproducibility

The source manifest records official download URLs, retrieval timestamps, byte counts and SHA-256 hashes. The data files accompanying this study retain unrounded estimates and margins, focal and full comparison results, coverage and the geography roster. Sources were frozen on November 9, 2026 rather than silently mixed with later revisions.

A separate implementation reproduced all 1,020 computed metrics across both years and all 553 computed comparisons. National arithmetic reconciliations matched totals, burden counts and not-computed counts across the 2,224 common geographies in 2024 and 1,998 in 2025. Fresh source downloads matched all ten stored source files. These checks support reproducibility and arithmetic accuracy, not a claim of external peer review or exact survey covariance.

## Official source guide

The following Census pages explain the variables and methods in readable form. Raw files are supplied for readers who want to reproduce the calculations. The study’s source freeze is November 9, 2026.

1. [2025 ACS one-year release schedule](https://www.census.gov/programs-surveys/acs/news/data-releases/2025/release-schedule.html) and [2025 summary-file overview](https://www.census.gov/programs-surveys/acs/data/summary-file.2025.html). Release timing and access to the table-based files.
2. [B25070 table definitions](https://api.census.gov/data/2025/acs/acs1/groups/B25070.html) and [2025 B25070 raw data](https://www2.census.gov/programs-surveys/acs/summary_file/2025/table-based-SF/data/1YRData/acsdt1y2025-b25070.dat). Gross rent as a percentage of household income.
3. [B25072 table definitions](https://api.census.gov/data/2025/acs/acs1/groups/B25072.html) and [2025 B25072 raw data](https://www2.census.gov/programs-surveys/acs/summary_file/2025/table-based-SF/data/1YRData/acsdt1y2025-b25072.dat). Age of householder by gross rent as a percentage of household income.
4. [B25074 table definitions](https://api.census.gov/data/2025/acs/acs1/groups/B25074.html) and [2025 B25074 raw data](https://www2.census.gov/programs-surveys/acs/summary_file/2025/table-based-SF/data/1YRData/acsdt1y2025-b25074.dat). Household income by gross rent as a percentage of household income.
5. [2025 comparison guidance](https://www.census.gov/programs-surveys/acs/guidance/comparing-acs-data/2025.html) and [2025 table changes](https://www.census.gov/programs-surveys/acs/technical-documentation/table-and-geography-changes/2025/1-year.html). Comparability and table structure.
6. [Census worked examples for derived estimates and tests](https://www2.census.gov/programs-surveys/acs/tech_docs/accuracy/2024_ACS_Accuracy_Document_Worked_Examples.pdf). Approximate margins of error and significance testing.
7. [Collection disruption note](https://www.census.gov/programs-surveys/acs/technical-documentation/user-notes/2026-04.html), [population controls note](https://www.census.gov/programs-surveys/acs/technical-documentation/user-notes/2026-05.html), and [disclosure and release rules note](https://www.census.gov/programs-surveys/acs/technical-documentation/user-notes/2026-06.html). Changes affecting interpretation of the 2025 release.
8. [2024 geography vintages](https://www.census.gov/programs-surveys/acs/geography-acs/geography-boundaries-by-year.2024.html) and [2025 geography vintages](https://www.census.gov/programs-surveys/acs/geography-acs/geography-boundaries-by-year.html). Boundary reference dates and metropolitan delineations.
9. [2024 national boundary-change notes](https://www2.census.gov/geo/docs/reference/bndrychange/national_gcn_2024.txt) and [July 2023 metro delineation workbook](https://www2.census.gov/programs-surveys/metro-micro/geographies/reference-files/2023/delineation-files/list1_2023.xlsx). The Lumpkin-to-Union territory transfer and Atlanta metro membership.
10. [Variance-replicate table availability](https://www.census.gov/programs-surveys/acs/data/variance-tables.html). Scope of the published replicate-variance product.
