Housing Data Sources: How to Measure Housing Need

Almost every claim about housing need in the United States traces back to about a dozen public datasets, most of them produced by HUD or the Census Bureau. No single one measures “housing need.” HUD supplies the income and rent benchmarks, Census surveys supply cost burdens, program inventories count subsidized homes, a one-night count measures homelessness, and building permits track new supply.

Measuring need means combining them, and knowing four things about every number: what it counts, what year it describes, what geography it covers, and how much error surrounds it. This guide covers each source, then works through a local housing needs assessment.

The Datasets at a Glance

Income and Rent Benchmarks

HUD Income Limits

HUD estimates a median family income, usually called area median income, for every metropolitan area and non-metropolitan county each year. From it HUD derives the income limits that decide eligibility for public housing, Section 8, and many other programs.

The source data is the ACS, so the limits lag. The FY 2026 limits took effect on May 1, 2026, a month late because the Census Bureau delayed its 2024 five-year ACS release. HUD described the equivalent gap for FY 2025 as a two-year lag.

HUD’s areas often span several counties, so a county’s limits may reflect a larger metro. And properties financed with tax credits or tax-exempt bonds use a separate table, the Multifamily Tax Subsidy Projects (MTSP) income limits. For a specific place, use HUD’s income limits lookup on HUD User.

Fair Market Rents

A Fair Market Rent (FMR) is HUD’s estimate of the 40th percentile gross rent, meaning rent plus utilities, for standard-quality units in an area, based on what recent movers pay. HUD publishes FMRs each year for metropolitan areas, some HUD-defined parts of metro areas, and non-metropolitan counties. They take effect at the start of the federal fiscal year, generally October 1. Small Area Fair Market Rents are calculated by ZIP code.

FMRs exist to run programs. In the Housing Choice Voucher program, agencies generally set the payment standard between 90% and 110% of FMR.

An FMR is an estimate projected forward. For FY 2027, HUD started from 2024 ACS rents, inflated them with a blend of Consumer Price Index data and private rent data drawn from six commercial sources, and then applied a trend factor to reach the fiscal year.

Other indicators are built on FMRs. The Housing Wage published by the National Low Income Housing Coalition (NLIHC) is the hourly wage a full-time worker needs to afford the FMR at 30% of income.

Household Surveys

American Community Survey

The ACS is the Census Bureau’s ongoing household survey and the raw material for most other sources here. It comes in two main forms:

• One-year estimates cover areas with 65,000 or more people. They are the most current and the least precise. The 2024 estimates were released on September 11, 2025. As of August 2026, the Census Bureau had not set a release date for the 2025 estimates.
• Five-year estimates pool 60 months of responses and cover every area down to census tracts and block groups. They are the most reliable and the least current. The 2020–2024 estimates were released on January 29, 2026.

Every ACS estimate is published with a margin of error, which can be large for small places. The Census Bureau also advises comparing only non-overlapping five-year periods, since 2018–2022 and 2019–2023 share four years of responses.

CHAS

CHAS data are custom ACS tabulations that the Census Bureau prepares for HUD. They sort households by tenure and by HUD’s own income bands, at 30%, 50%, 80%, and 100% of HUD Area Median Family Income (HAMFI), and count how many in each band have a housing problem.

CHAS recognizes four problems: incomplete kitchen facilities, incomplete plumbing, overcrowding (more than one person per room), and cost burden (housing costs above 30% of income). Costs above 50% of income are a severe cost burden.

Local governments use CHAS for the needs assessment in the Consolidated Plan that HUD requires of its grantees. Federal rules call for that assessment to rest on Census data supplied by HUD, as updated by any properly conducted local study, or on another reliable source that the jurisdiction identifies.

Three limits matter:

• Lag. The release of December 23, 2025, is based on 2018–2022 ACS data.
• Counts only. CHAS reports household counts, so users calculate percentages and choose the denominator themselves.
• Comparability. HUD urges caution with year-to-year comparisons, because variable names and definitions may have changed. Some estimates have also been suppressed since the 2018 tables.

American Housing Survey

The AHS is sponsored by HUD and conducted by the Census Bureau, which describes it as the most comprehensive national housing survey since it began in 1973. Unlike the ACS, it returns to the same housing units over time and asks about their physical condition and costs.

Its limit is geography. The AHS covers the nation and selected large metropolitan areas, not counties or small cities. It has been fielded every two years, with 2023 the latest published, and HUD is shifting it to continuous data collection.

Inventories of Assisted Housing

Picture of Subsidized Households

Picture is HUD’s annual summary of households in its rental assistance programs: public housing, Housing Choice Vouchers, project-based Section 8, Section 202, Section 811, and several older programs. It reports unit counts, occupancy, incomes, and household characteristics at every level from the nation down to census tracts and individual projects.

Each year’s file is a snapshot of households reported as of December. HUD suppresses any cell covering fewer than 11 families and marks entries as non-reporting where fewer than half of households were reported. Picture leaves out USDA rural housing, Indian housing, HOME, and CDBG unless a property also has a covered HUD subsidy.

LIHTC and Housing Finance Agency Data

HUD’s database for the Low-Income Housing Tax Credit lists 55,345 projects and 3.9 million units placed in service between 1987 and 2024. Each record gives the address, unit counts, placed-in-service year, credit type, and other financing. Data for 2025 will be added in spring 2027.

Treat it as a cumulative record of what was built, with gaps. HUD notes that nearly every variable is missing for at least a few projects. State housing finance agencies hold the underlying records, and many publish their own award lists and property inventories. They also report tenant data to HUD, which releases state-level tables and cautions that the demographic fields are not collected consistently.

National Housing Preservation Database

The National Housing Preservation Database (NHPD) was created in 2011 by NLIHC and the Public and Affordable Housing Research Corporation (PAHRC). It links federal subsidy records at the property level and shows when each property’s affordability restrictions end, which makes it the main tool for tracking expiring-use risk. Using 2025 NHPD data, NLIHC found restrictions set to expire within five years for 417,515 federally assisted homes.

An end date is not a loss. NLIHC notes that many of these properties renew their assistance or secure new funding.

Homelessness: The PIT Count and AHAR

The point-in-time count is a tally of people experiencing homelessness on a single night in the last 10 days of January. Local planning bodies under the Continuum of Care program carry it out. They count people in shelters every year and must count unsheltered people at least every other year. HUD publishes the results in Part 1 of the Annual Homelessness Assessment Report (AHAR). The report on the January 2025 count came out in May 2026.

HUD notes several limits. Weather, local policies, and changing methods affect the unsheltered count, and the true number could be larger than reported. The count covers shelters, transitional housing, safe havens, and unsheltered locations, so people doubled up with friends or relatives fall outside it. A Continuum of Care may cover a city, a county, or a whole state, so its figures often do not match one jurisdiction.

Supply and Vacancy

Building Permits Survey

The Census Bureau’s Building Permits Survey counts new privately owned housing units authorized by permit, by structure size, monthly and annually. It covered about 19,900 permit-issuing places as of January 2023 and publishes data for states, metro areas, counties, and individual places.

A permit is an authorization, not a finished home. The Census Bureau notes that construction is sometimes abandoned after a permit is issued, and homes that are built are finished well after the permit date.

Housing Vacancy Survey

The Housing Vacancy Survey reports rental vacancy, homeowner vacancy, and homeownership rates each quarter. It covers the nation, regions, states, and the 75 largest metropolitan areas. For a county or town, vacancy estimates come from the ACS instead.

Common Misuses

• Treating FMR as the market rent. It is a 40th percentile benchmark for a whole area.
• Mixing income measures. HUD’s median family income is not the Census median household income, and HAMFI includes HUD adjustments.
• Using the wrong limits. Tax credit properties use MTSP limits, not the Section 8 table.
• Ignoring dates. A release published in late 2025 can describe 2018–2022.
• Adding inventories together. NLIHC and PAHRC found in 2024 that 41% of federally assisted homes rely on more than one subsidy program.
• Reading permits as completions, or one night’s homeless count as a year’s total.
• Quoting small-area estimates without margins of error.

Worked Example: A Local Housing Needs Snapshot

Suppose a planning commission in a fictional county of 90,000 people wants a short, defensible picture of housing need. All figures below are invented.

1. Fix the geography. The county is large enough for one-year ACS estimates. Its towns are not, so they need five-year data.
2. Set the income bands. Staff look up the county’s HUD income limits and note whether HUD groups it into a larger metro area.
3. Count need by band. Suppose CHAS shows 5,000 renter households at or below 50% of HAMFI, of whom 3,800 are cost-burdened and 2,400 severely so. That is 76% and 48%. The report labels the figures “2018–2022.”
4. Check the workforce band. The summary CHAS tables separate households at 80% to 100% of HAMFI and group everyone above 100%. Measuring need up to 120% takes HUD’s detailed download tables, some of which use finer bands, or ACS income tables.
5. Compare rents with incomes. Suppose a household earns $48,000. Thirty percent of that is $1,200 a month. If the two-bedroom FMR were $1,500, the gap would be $300. Staff add the ACS median gross rent as a second reference.
6. Inventory assisted housing. Staff pull Picture, the LIHTC database, and the NHPD, match properties by address so none is counted twice, and list restrictions that end within ten years.
7. Add homelessness. The county sits inside a multi-county Continuum of Care, so the report gives that figure and says it cannot be isolated.
8. Add supply. Five years of permits by structure type show what is being authorized. Vacancy comes from the ACS.
9. Label everything. Each number carries its source, year, geography, and, for survey data, its margin of error.

The assisted inventory should not be subtracted from the cost-burden count to produce a “gap,” because assisted households are already in the CHAS totals. The result is a snapshot. It does not replace the market study that funders typically require for a specific project.

The Bottom Line

Housing need is measured by assembling sources that were each built for a different purpose. HUD’s income limits and Fair Market Rents set the benchmarks, CHAS and the ACS describe households, program inventories describe the assisted stock, the PIT count covers homelessness, and permits track new supply. A number is only as good as its label, so state the source, the year, and the place every time.