HUD Housing Choice Vouchers: The Section 8 Data Behind 2.3 Million Households and $30 Billion in Annual Rental Assistance

Every month, roughly 2.3 million American households pay a fraction of their income toward rent — and a federal check covers the rest. The Housing Choice Voucher program, widely known as Section 8, is the largest federal rental assistance program in the United States: $30 billion in annual outlays, 2,200 local administering agencies, and a waitlist system that has left the majority of eligible households without assistance for decades. HUD publishes the data. Almost nobody outside housing policy circles uses it systematically.

The HCV program sits at the intersection of federal housing policy, local real estate markets, and the structural gap between the supply of affordable units and the demand for them. Its administrative records — Fair Market Rents, Picture of Subsidized Households, project-level Multi-Family Housing data, and the CHAS affordability dataset — collectively constitute one of the richest windows into how housing unaffordability propagates across American cities and rural counties. This article covers the program mechanics, the datasets, and how to use them analytically.

Program overview

The Housing Choice Voucher program was created under Section 8 of the Housing Act of 1937 and substantially revised by the Housing and Community Development Act of 1974, which introduced the tenant-based voucher concept. The defining feature is portability: rather than tying federal assistance to a specific building (as public housing and project-based programs do), HCV gives subsidy authority directly to the household, which then finds a unit in the private market.

The mechanics are straightforward. A household with a voucher finds a private landlord willing to participate in the program and whose unit passes a Housing Quality Standards (HQS) inspection. The local Public Housing Authority — the administering agency — then pays the landlord the difference between 30% of the household's adjusted gross income and the applicable payment standard, which is typically set between 90% and 110% of the HUD-published Fair Market Rent for the area. The household pays the rest directly to the landlord. If the actual rent exceeds the payment standard, the household can pay the difference out of pocket, but HUD rules limit this overhang to prevent excessive cost burden.

Congress appropriates HCV funding annually. HUD allocates budget authority to the roughly 2,200 PHAs nationwide; each PHA then issues vouchers to households from its waitlist up to the limit of its budget authority. PHAs must maintain a utilization rate of at least 95% of their budget authority or face recapture — which creates pressure to keep waitlists active and to re-issue vouchers quickly when households exit the program.

The program cost in 2024 was approximately $30 billion in federal outlays, making it the dominant federal housing expenditure — larger than the entire LIHTC credit subsidy, larger than the HOME Investment Partnerships Program, and roughly six times the Community Development Block Grant appropriation. Despite that scale, it reaches fewer than one in four income-eligible households nationally.

Fair Market Rents

Fair Market Rents are the central data artifact of the HCV program. HUD publishes FMRs annually for approximately 2,600 FMR areas — combinations of metropolitan areas and non-metropolitan county groupings that together cover every county in the United States. Each FMR is expressed as a gross monthly rent (including utilities) by bedroom size, from efficiency through four-bedroom.

The FMR is set at the 40th percentile of gross rents paid by recent movers (households who moved into their current unit within the past 15 months) in the private unassisted rental market. The 40th percentile threshold was reduced from 45% in 1995 and then raised back in some jurisdictions as the program struggled to compete with tight rental markets in high-cost cities. The survey basis is the American Community Survey, supplemented by Random Digit Dialing surveys in markets where the ACS sample is insufficient to produce reliable estimates.

To illustrate the range: for FY2024, the 2-bedroom FMR was $2,765 in the New York – Newark – Jersey City metro area, $3,268 in the San Francisco – Oakland – Hayward metro area, $1,195 in Boise, Idaho, and $725 in rural Mississippi. A single national FMR would be useless as a payment standard; the local calibration is the whole point.

Small Area Fair Market Rents

The standard metro-wide FMR creates a structural problem in high-cost cities: the metro average masks enormous within-metro rent variation. A metro-wide 2BR FMR of $2,000 might represent a competitive payment standard in a lower-cost suburban submarket but be laughably inadequate in a central-city neighborhood where actual market rents run $3,000 or more. Voucher holders in such markets are effectively priced out of higher-opportunity neighborhoods and confined to wherever their payment standard will cover.

Small Area Fair Market Rents (SAFMRs) address this by computing FMRs at the ZIP code level rather than the metro level. HUD mandates SAFMRs for the highest-cost metros — those where the standard metro FMR is inadequate to access a reasonable share of the rental market — and makes SAFMR datasets available for all other jurisdictions that want to adopt them voluntarily. The SAFMR dataset is available at huduser.gov/portal/datasets/fmr/smallarea/index.html. In high-cost metros with SAFMRs, a voucher holder in a high-rent ZIP code receives a higher payment standard than one in a low-rent ZIP code within the same metro — substantially improving access to high-opportunity neighborhoods.

Payment standards and PHA discretion

PHAs are not required to set their payment standard equal to the FMR. Federal rules allow a payment standard between 90% and 110% of the FMR without HUD approval. PHAs in tight markets often push to 110%; PHAs in softer markets may use 90% to reduce per-unit costs and extend their budget authority to more households. HUD also allows exception payment standards above 110% for specific circumstances — households with disabilities needing accessible units, or areas experiencing severe market tightness — subject to HUD approval.

Waitlists and program gaps

The gap between HCV program capacity and eligible demand is one of the most significant structural features of American housing policy. Harvard's Joint Center for Housing Studies estimates that there are more than 17 million extremely low-income renter households (earning at or below 30% of Area Median Income) competing for approximately 5 million assisted units across all federal rental assistance programs. The HCV program serves roughly 2.3 million of those households — about 13% of the extremely low-income renter population.

The waitlist is the visible symptom of this gap. In most large metros, PHAs have waitlists of one to ten years. Many PHAs close their waitlists entirely when the backlog grows too large to manage — accepting no new applications for years at a time and reopening only briefly when throughput creates capacity. According to HUD's own estimates, only about 25% of households that would be income-eligible for HCV assistance actually receive it.

The disparity is not uniform. Rural PHAs in lower-cost markets sometimes have shorter waitlists and higher utilization rates, because FMRs are adequate to compete in those rental markets and fewer households are pursuing the same pool of assisted units. Urban PHAs in New York, Los Angeles, and other high-cost cities have waitlists so long that households entering the list today realistically have no prospect of receiving a voucher within a decade.

Emergency Housing Vouchers

The American Rescue Plan Act of 2021 authorized 70,000 Emergency Housing Vouchers (EHVs) targeted at specific high-need populations: individuals and families experiencing homelessness, at risk of homelessness, fleeing domestic violence or sexual assault, or recently homeless and at high risk of long-term homelessness. EHVs were a time-limited infusion rather than a permanent program expansion, but they represent the largest single addition to HCV program capacity in the program's history. HUD published separate data on EHV utilization rates and outcomes, showing the challenge of connecting the most vulnerable populations to the private rental market even when voucher funding is available.

Picture of Subsidized Households

The administrative record system for HCV participants is HUD's Public and Indian Housing Information Center (PIC). Every PHA with HCV units reports household-level data to PIC, covering demographics, income, employment, and housing unit characteristics for each assisted household. HUD aggregates these records and publishes the results annually as the Picture of Subsidized Households (PASH) at huduser.gov/portal/datasets/assthsg.html.

The PASH data is available at multiple geographic levels: national, by state, by PHA, and — most valuably for spatial analysis — by census tract. The tract-level file enables researchers to map voucher concentration, study the demographic composition of HCV households relative to surrounding neighborhoods, and analyze whether vouchers are clustering in low-opportunity tracts or dispersing across the metro.

Key variables in the PASH dataset:

• Number of units and people. Count of assisted units and total persons in assisted households.
• Average household income and income per person. Mean and median adjusted gross income, reported as an annual figure.
• Percent elderly. Share of households where the head or co-head is 62 or older.
• Percent disabled. Share of households where the head, co-head, or any member reports a disability.
• Percent female-headed household. A longstanding metric in assisted housing analysis, reflecting the demographic composition of the assisted population.
• Race and ethnicity breakdown. Share of households by race (White, Black, Asian, Native American, Pacific Islander) and Hispanic ethnicity, enabling fair housing analysis.
• Average months on waitlist. The mean time between initial application and voucher issuance, reported by PHA. This is among the most direct measures of program scarcity.
• Percent receiving welfare benefits. Share of households reporting TANF or other public assistance income.

The tract-level PASH file can be joined to ACS demographic data using the 11-digit census tract FIPS code, enabling spatial analysis of voucher concentration relative to poverty rate, school quality metrics, employment access, and other opportunity indicators. This is the primary data source for Affirmatively Furthering Fair Housing (AFFH) analysis of whether HCV program geography reflects or counteracts residential segregation patterns.

Project-Based Rental Assistance

Not all Section 8 assistance is tenant-based. Project-Based Rental Assistance (PBRA) — also called project-based Section 8 — is a separate program in which HUD contracts directly with private landlords to subsidize specific units in specific buildings. The subsidy attaches to the unit, not the household: when a tenant leaves, the next income-eligible tenant who moves in receives the subsidy.

PBRA covers approximately 1.2 million units in roughly 14,000 properties nationwide. HUD's Multi-Family Housing database at huduser.gov/portal/datasets/mfh.html documents every HUD-assisted multifamily property, including PBRA properties, with fields for program type, contract expiration date, unit count, year built, and geographic identifiers. HUD multifamily programs represented in the database include:

• Section 8 New Construction / Substantial Rehabilitation (NC/SR). Contracts executed in the 1970s and 1980s with newly built or substantially rehabilitated properties. These are the original project-based Section 8 contracts, many of which are now in their third or fourth renewal cycle.
• Section 236. An older interest-rate-subsidy program that predates Section 8; many Section 236 properties now carry layered subsidies including project-based Section 8 and LIHTC.
• Section 515 Rural Rental Housing. USDA Rural Development program for rural multifamily housing; sometimes included in HUD's multifamily tracking for portfolio completeness.
• Section 202 Supportive Housing for the Elderly. Direct loan and capital advance program for elderly housing; units are typically subsidized with project rental assistance contracts (PRACs) rather than Section 8.

The expiring-use problem

PBRA contracts are not permanent. They run for fixed terms — typically five to twenty years — and require renewal. At renewal, property owners can choose to opt out of the program if they believe market-rate rents would exceed the contract rents plus any HAP (Housing Assistance Payment). When owners opt out, the affordable units are lost and existing tenants are displaced unless they can secure HCV vouchers or find other assistance.

The risk of loss at contract expiration is the core of the expiring-use problem. The National Housing Preservation Database (NHPD), published by the Public and Affordable Housing Research Corporation, tracks all subsidized properties — PBRA, LIHTC, HOME, USDA — with contract expiration dates and preservation risk scores. It is the canonical tool for identifying which properties are at near-term risk of converting to market-rate use. Properties in high-rent markets where market-rate rents would substantially exceed contract rents, with contract expirations in the next one to three years, and with low preservation risk scores, are the highest-priority targets for preservation intervention.

Cross-cutting affordability data: AFFH and CHAS

Two cross-program datasets are essential context for HCV analysis.

CHAS: Comprehensive Housing Affordability Strategy data

The CHAS dataset, published by HUD at huduser.gov/portal/datasets/cp.html, is a custom tabulation of the American Community Survey designed to measure housing needs by jurisdiction. It covers every CDBG-eligible jurisdiction (states, counties, metropolitan cities, and urban counties) with tables breaking down households by income, tenure (owner/renter), housing cost burden, and housing conditions.

The key CHAS income categories are defined as fractions of Area Median Income: extremely low income (at or below 30% AMI), very low income (31–50% AMI), low income (51–80% AMI), and moderate income (81–100% AMI). For each income-by-tenure cell, CHAS reports the number of cost-burdened households (paying more than 30% of income on housing costs) and severely cost-burdened households (paying more than 50% of income on housing costs).

The CHAS numbers are stark. Among extremely low-income renters nationally, more than 72% are cost-burdened, and more than 50% are severely cost-burdened. This population — renters earning at or below 30% of AMI — is the primary target of the HCV program, and the CHAS data quantifies the structural mismatch between that population's size and the program's capacity. CHAS data is the appropriate denominator for calculating what share of cost-burdened extremely low-income renters in any jurisdiction currently receive HCV assistance — joining PASH unit counts to CHAS household counts produces that ratio directly.

AFFH: Affirmatively Furthering Fair Housing

The Fair Housing Act requires HUD-funded jurisdictions to affirmatively further fair housing — to take active steps to reduce residential segregation and expand access to opportunity. HUD's AFFH rule, reinstated in 2021 after being suspended in 2020, requires jurisdictions to submit Equity Plans documenting their fair housing analysis and program responses.

HUD publishes the AFFH Data and Mapping Tool at hudgis.hud.gov/AFFH, which integrates the PASH tract-level voucher data with ACS demographics, HUD-defined opportunity indices, and school quality metrics. The tool is designed for grantees preparing AFFH submissions, but the underlying datasets are also available for download. The AFFH opportunity indices — covering school proficiency, job access, poverty exposure, environmental health, and transit access — can be joined to PASH tract-level voucher concentration data to test empirically whether voucher holders are concentrated in low-opportunity tracts.

Data access

HUD's primary public data portal is the HUD User Data Portal at huduser.gov. Key datasets and access paths:

• Fair Market Rents. Annual bulk download at huduser.gov/portal/datasets/fmr.html. Includes metro-area summary CSVs, Small Area FMR ZIP-code-level files, and historical data back to 1983. An unofficial API at hudapi.com/fmr provides FMR lookup by FIPS or ZIP code.
• Picture of Subsidized Households. Annual bulk download at huduser.gov/portal/datasets/assthsg.html. Available at national, state, PHA, county, and census tract levels. File format is Excel or CSV with a codebook PDF.
• Multi-Family Housing (PBRA properties). Download at huduser.gov/portal/datasets/mfh.html. Property-level records with contract details, expiration dates, and geocoded locations.
• CHAS data. Download at huduser.gov/portal/datasets/cp.html. Available at census tract, county, and jurisdiction levels. The codebook is extensive; the relevant tables for HCV analysis are tables 7 (renters by income and cost burden) and 9 (renters with severe housing problems).
• HUD Geospatial Data. Shapefiles and GeoJSON for HUD programs at hud.gov/program_offices/comm_planning/datasets. Includes CoC boundaries, PHA service area boundaries, opportunity zone overlays, and FMR area boundaries.
• FRED macroeconomic context. The Federal Reserve's FRED database includes series TNTHOUSSINPVT (tenant-occupied housing units in private structures) and related rental market aggregates that provide national-level context for HCV program trends.

Code: ranking metros by FMR-to-income affordability gap

The following script downloads HUD's FY2024 Fair Market Rents for all metropolitan areas, fetches median renter household income from the Census ACS 5-year estimates, merges the two on CBSA code, and ranks metros by the ratio of the annualized 2BR FMR to median renter income. Metro areas where that ratio exceeds 50% are in severe affordability crisis — a voucher holder paying 30% of income toward rent still requires a payment standard that consumes more than half of what the typical local renter earns. A free Census API key is required for the ACS query; register at api.census.gov/.

import pandas as pd
import requests

# ---------------------------------------------------------------
# Step 1: Download the HUD Fair Market Rents for all metro areas
# FMR data is published annually at:
# https://www.huduser.gov/portal/datasets/fmr.html
# The summary file includes all FMR areas in a single CSV.
# We use the FY2024 small area and metro summary file here.
# ---------------------------------------------------------------
FMR_URL = 'https://www.huduser.gov/portal/datasets/fmr/fmr2024/FY2024_FMRs_revised.zip'

import urllib.request, zipfile, io

print('Downloading HUD FY2024 Fair Market Rents...')
with urllib.request.urlopen(FMR_URL) as resp:
    zf = zipfile.ZipFile(io.BytesIO(resp.read()))
    # The archive contains metro-area and non-metro summary CSVs.
    # Select the file that covers metro (CBSA-level) FMR areas.
    csv_names = [n for n in zf.namelist() if n.endswith('.csv')]
    print('Files in archive:', csv_names)
    # Use the first CSV; adjust the index if the archive structure changes.
    fmr_df = pd.read_csv(zf.open(csv_names[0]), encoding='latin-1', low_memory=False)

print(f'FMR records: {len(fmr_df):,}')
print(fmr_df.columns.tolist())

# ---------------------------------------------------------------
# Step 2: Keep relevant columns and rename for clarity
# Column names vary slightly by release year.
# fmr_2br = 2-bedroom Fair Market Rent (the standard benchmark unit)
# fmr_area_name = human-readable FMR area label
# ---------------------------------------------------------------
# Adjust these column names to match the actual release if they differ.
fmr_df = fmr_df.rename(columns={
    'fmr_2': 'fmr_2br',           # 2BR FMR in dollars/month
    'area_name': 'fmr_area_name', # FMR area label
    'fips2000': 'fips_code',       # FIPS identifier (varies by release)
})

# Drop rows without a 2BR FMR value
fmr_df = fmr_df.dropna(subset=['fmr_2br'])
fmr_df['fmr_2br'] = pd.to_numeric(fmr_df['fmr_2br'], errors='coerce')
fmr_df = fmr_df.dropna(subset=['fmr_2br'])

print(f'FMR areas after cleaning: {len(fmr_df):,}')

# ---------------------------------------------------------------
# Step 3: Fetch median renter household income by metro area
# from the ACS 5-year estimates.
# Table B25119: median household income in the past 12 months
#   by tenure (B25119_003E = median renter income)
# We query at the metro/CBSA level.
# ---------------------------------------------------------------
CENSUS_KEY = 'YOUR_CENSUS_API_KEY'  # Register free at api.census.gov/

print('Fetching ACS median renter income by metro area...')
r = requests.get(
    'https://api.census.gov/data/2022/acs/acs5',
    params={
        'get': 'NAME,B25119_003E',
        'for': 'metropolitan statistical area/micropolitan statistical area:*',
        'key': CENSUS_KEY,
    },
    timeout=60,
)
r.raise_for_status()

rows = r.json()
headers = rows[0]
acs_df = pd.DataFrame(rows[1:], columns=headers)
acs_df = acs_df.rename(columns={
    'metropolitan statistical area/micropolitan statistical area': 'cbsa_code',
    'B25119_003E': 'median_renter_income',
    'NAME': 'metro_name',
})
acs_df['median_renter_income'] = pd.to_numeric(acs_df['median_renter_income'], errors='coerce')
acs_df = acs_df.dropna(subset=['median_renter_income'])
# Exclude areas where the estimate is -666666666 (not computed)
acs_df = acs_df[acs_df['median_renter_income'] > 0]

print(f'ACS metro areas with valid renter income: {len(acs_df):,}')

# ---------------------------------------------------------------
# Step 4: Merge FMR data with ACS renter income on CBSA code.
# HUD FMR files include a cbsasub or cbsa field in most releases.
# Check the column list from Step 2 and adjust the key accordingly.
# ---------------------------------------------------------------
# Attempt a merge on the CBSA numeric code (5 digits, zero-padded).
# HUD uses different column names by release year; common options:
#   'cbsa', 'cbsasub', 'metro_code', 'msasub'
# Try each in order and use the first that exists.
cbsa_col_candidates = ['cbsa', 'cbsasub', 'metro_code', 'msasub']
cbsa_col = next((c for c in cbsa_col_candidates if c in fmr_df.columns), None)

if cbsa_col is None:
    print('WARNING: No CBSA code column found in FMR file. Columns available:')
    print(fmr_df.columns.tolist())
    raise SystemExit('Cannot merge without a CBSA code column.')

fmr_df['cbsa_code'] = fmr_df[cbsa_col].astype(str).str.zfill(5)
acs_df['cbsa_code'] = acs_df['cbsa_code'].astype(str).str.zfill(5)

merged = fmr_df.merge(acs_df, on='cbsa_code', how='inner')
print(f'Matched metro areas: {len(merged):,}')

# ---------------------------------------------------------------
# Step 5: Compute FMR as a percentage of annual median renter income.
# FMR is monthly; annualize (x12) then divide by median annual income.
# Ratio > 0.30 means the FMR alone exceeds the standard affordability
# threshold for a household earning the local median renter income.
# Ratio > 0.50 = severe affordability crisis by FMR alone.
# ---------------------------------------------------------------
merged['fmr_annual'] = merged['fmr_2br'] * 12
merged['fmr_pct_income'] = (merged['fmr_annual'] / merged['median_renter_income']) * 100

# ---------------------------------------------------------------
# Step 6: Rank metros by affordability gap (FMR as % of income).
# Top 10 where FMR exceeds 50% of median renter income.
# ---------------------------------------------------------------
severe = merged[merged['fmr_pct_income'] > 50].copy()
severe = severe.sort_values('fmr_pct_income', ascending=False)

top10 = severe.head(10)[['metro_name', 'fmr_2br', 'median_renter_income', 'fmr_pct_income']]

print()
print('Top metros where 2BR FMR exceeds 50% of median annual renter income:')
print(top10.to_string(index=False, float_format=lambda x: '{:.1f}'.format(x)))

# ---------------------------------------------------------------
# Step 7: Summary statistics across all matched metros
# ---------------------------------------------------------------
print()
print(f'Metros where FMR > 30% of renter income: {(merged["fmr_pct_income"] > 30).sum():,}')
print(f'Metros where FMR > 50% of renter income: {(merged["fmr_pct_income"] > 50).sum():,}')
print(f'National median FMR-to-income ratio:       {merged["fmr_pct_income"].median():.1f}%')
print(f'Highest ratio (most unaffordable):         {merged["fmr_pct_income"].max():.1f}%  '
      f'({merged.loc[merged["fmr_pct_income"].idxmax(), "metro_name"]})')
print(f'Lowest ratio (most affordable):            {merged["fmr_pct_income"].min():.1f}%  '
      f'({merged.loc[merged["fmr_pct_income"].idxmin(), "metro_name"]})')

Expected results: the most unaffordable metros by this metric tend to cluster in California coastal markets (Santa Barbara, San Luis Obispo, Santa Cruz), Hawaii, and resort/tourism markets (Jackson Hole, Nantucket, the Florida Keys) where median renter incomes are moderate but FMRs reflect premium market rents. The most affordable metros — lowest FMR-to-income ratio — tend to be smaller Midwestern metros with moderate rents and relatively high manufacturing wages. The analysis also surfaces a structural feature of the HCV program: in the most unaffordable metros, a payment standard at 110% of FMR still falls short of covering the median 2BR unit, which means voucher holders are systematically excluded from much of the private rental market even with maximum program parameters.

Research applications

The HCV datasets support a range of analytical applications beyond the affordability gap analysis above.

Voucher concentration and opportunity access

Joining the PASH census tract file to AFFH opportunity indices produces a direct measure of whether voucher holders are concentrated in low-opportunity neighborhoods. Studies using this approach consistently find that vouchers are overrepresented in high-poverty, low-school-quality tracts relative to the distribution of available rental units — suggesting that low payment standards, landlord participation barriers, and search constraints effectively funnel voucher holders into lower-opportunity areas. The SAFMR policy was specifically designed to counteract this by raising payment standards in high-opportunity ZIP codes.

Landlord participation rates

HUD's Small Area FMR datasets, combined with American Community Survey rental vacancy data and PASH utilization rates, can support analysis of landlord participation barriers. In markets where FMRs are adequate relative to market rents but voucher utilization rates are low, the constraint is often on the supply side — landlords declining to accept vouchers due to HQS inspection requirements, paperwork burden, or legal barriers to source-of-income discrimination. Several states have enacted source-of-income anti-discrimination laws; their effects can be estimated by comparing PHA utilization rates and waitlist lengths before and after enactment.

Race and geography of the HCV program

The PASH race and ethnicity breakdown, combined with the tract-level geographic identifiers, enables analysis of whether Black and Hispanic voucher holders are systematically more likely than White voucher holders to use their vouchers in high-poverty, high-segregation tracts — even within the same PHA's service area. This is an AFFH question with significant fair housing implications, and the PASH data provides the administrative record basis to study it at scale.

PHA performance and budget utilization

HUD publishes PHA-level performance data including budget utilization rates, which measure what share of each PHA's allocated budget authority is actually spent on housing assistance payments. PHAs with utilization rates below 95% face recapture; PHAs consistently below 90% may face other oversight actions. Cross-referencing low utilization rates with high FMRs relative to local market rents — where the PHA's payment standards may be too low to secure units — reveals the structural sources of utilization failure.