US Census Housing Vacancies and Homeownership (CPS/HVS) Program

It is the program front end of the Housing Vacancy Survey, delivered as rows instead of releases. The Housing Vacancy Survey rides on the Current Population Survey - roughly 72,000 addresses interviewed for the federal government's official read on who owns, who rents and who sits empty - and the program packages its outputs three ways: the latest Residential Vacancies and Homeownership release, roughly twenty quarterly tables spanning rates, inventory components, median asking rent and median asking sales price, and an annual archive underneath them, plus an interactive time-series charting surface keyed to program code HV.

Why economists care: the rental vacancy rate this program carries is a component of the Conference Board's index of leading economic indicators. It is one of the few housing statistics that behaves as a forward input - a tightening vacancy series tends to lead rent growth rather than trail it. Current-cycle readings make the point concretely: Q2 2026 shows a 7.3 percent rental vacancy rate against a 65.0 percent homeownership rate.

Get a sample of this dataset cut to your regions and cuts, or read the rows below first.

Four delivery-shaped rows built from values captured during the August 2026 research pass:

quarter                        : 2026 Q2
geography                      : United States
total_housing_units_thousands  : 148163
vacant_for_rent_thousands      : 3547
rental_vacancy_rate_pct        : 7.3
homeowner_vacancy_rate_pct     : 1.2
homeownership_rate_pct         : 65.0

quarter                        : 2026 Q2
geography                      : United States - inside principal cities
rental_vacancy_rate_pct        : 8.0
geography                      : United States - outside principal cities
rental_vacancy_rate_pct        : 6.9

quarter                        : 2026 Q2
geography                      : United States
homeownership_rate_pct                 : 65.0
homeownership_non_hispanic_white_pct   : 74.5
homeownership_black_alone_pct          : 45.4
homeownership_hispanic_any_race_pct    : 58.6

quarter                        : 2026 Q2
geography                      : United States
median_asking_rent_usd         : 1531
median_asking_rent_west_usd    : 1941
median_asking_sales_price_usd  : 343800

Read them as one argument. Row one is the national frame: roughly 148 million housing units, about 3.5 million of them vacant for rent, producing the 7.3 percent headline. Row two dissolves the headline - the same national average describes an 8.0 percent principal-city market and a 6.9 percent suburbs market, which is why metro-status splits matter more than the topline. Row three is why tenure researchers keep the raw series: a 65.0 percent national homeownership rate conceals a 29-point spread between demographic groups. Row four prices the empty stock the first row counted - $1,531 median asking rent nationally, $1,941 in the West.

Six measures carry the spine, mapped during the August 2026 research pass against the survey's own published documentation, with every example value traced to a live reading from the current cycle.

The design logic is visible in the columns themselves: two vacancy rates measure friction in the market, one homeownership rate measures its structure, one inventory estimate measures what the rates are computed from, and two price series monetize the vacant stock. Everything else the program publishes - demographic cuts, geographic splits, precision margins - extends these six rather than replacing them, which is why deliveries hold a stable wide table instead of one schema per table.

The six-column spine covers every delivery; the extensions below fold under additional fields on request because they depend on the shape of the cut:

1. Inside/outside MSA and principal-city/suburb splits behind every headline rate - where one national figure separates into structurally different markets
2. Homeownership rates by race and ethnicity, family income band and age of householder - the demographic cuts running from 1994
3. Vacant-unit components in thousands: for rent, for sale only, rented or sold, held off market, seasonal
4. Monthly household estimates carrying the program's longest reach, from 1955
5. Standard errors and 90 percent confidence margins for the headline rates, published in the B-series tables
6. Vintage-2025-revised inventory rows covering 2000-present, republished under current population controls

Additional fields on request - name the layers your models need when you ask for a sample and they land in the same table, pre-joined on geography and period.

• Geography - United States and the four Census regions (Northeast, Midwest, South, West), with inside/outside MSA and principal-city splits in the tables and metropolitan-area breakdowns alongside. National is the headline; the splits are where the analysis lives.
• Temporal - a current-cycle release riding on deep archives: household estimates from 1955, rental and homeowner vacancy rates from 1956, homeownership rates from 1964, inventory from 1965, median asking rent and sales price from 1988, demographic cuts from 1994. Trend work can run longer than most firms have existed.
• Granularity - one observation per geography x measure x period. Quarterly at the flagship grain, monthly household estimates beneath, annual summary tables under those. No address-level microdata exists anywhere in the corpus; the survey publishes aggregates by design.

Two honesty notes worth knowing before modeling. Pandemic-era observations (2020-2021) carry an interpretive caution in the survey's own Source and Accuracy documentation - treat those quarters as measured under disrupted collection conditions. And the inventory series was rebenchmarked onto Vintage 2025 population controls back to Q2 2020, so pre-revision copies of recent inventory rows differ from the revised set.

API, files, or your warehouse. Daily, weekly, or hourly.

Seventy years of tenure history lands however your stack wants it: point queries over REST for the latest quarter, whole-archive files sized for overnight loads, or a direct pipe into Snowflake, BigQuery or Redshift. Cadence is yours to set and change - load the full history once, then let each new cycle accrue onto the same keys.

Every delivery ships the complete field dictionary above, sample rows for validation and any additional fields you requested pre-joined, so the first query you run is already shaped like the last one.

Ranked by how directly one row settles the day job:

1. Multifamily owners, lenders and REIT analysts - vacancy direction plus asking rent level is the earliest honest read on whether a rental market is tightening or softening; see demand forecasting.
2. Macro strategists and quant researchers - a leading-index component with history to 1956 builds housing factors that span multiple full cycles; methods continue on quant backtesting.
3. Market researchers and consultants - the official demand-side backdrop under any residential study, from national topline to principal-city cut; see market sizing.
4. Policy and demographic researchers - homeownership by race, ethnicity, income and age turns a percentage into a structural argument with documented margins.
5. Proptech and tenant-facing product teams - size the addressable renter base by metro status and region using the official frame.
6. Journalists and academics - cite the federal series directly instead of paraphrasing someone else's chart; see citation-grade research.

Investors and quants get a leading-index component and seven decades of tenure cycles engineered into model-ready factors. Market researchers and consultants get the official backdrop under every residential deck, national to principal-city. Data scientists and ML engineers get a six-measure spine that joins cleanly to price and permits panels without cleanup. Journalists, academics and students get citable federal numbers with margins of error attached. Developers and data-product builders get housing dashboards fed by flat rows instead of hand-parsed releases.

Scope note - this record is the program's living edge: the current release, the quarterly tables and the charting surface. Its full memory lives next door - the CPS/HVS Historical Tables hold about thirty workbooks running the same series back to the mid-1950s, and the two share one logical schema, so pairing them costs nothing and gains everything.

Grain note - the survey publishes aggregates by design. Anyone promising address-level tenure microdata derived from this program is selling something else; the honest play is joining these aggregates to your own property-level panel on geography and period.

Where to go next - the rail below collects the seventy-year archive, the price leg from FHFA, the traded-market counterpart from Nareit, both head-to-head comparisons, and the scoring note. Browse the shelf on the best diversified real estate activities datasets ranking or the diversified real estate activities data hub. For terminology, start with what rental vacancy rate means, what homeownership rate means or what the CPS/HVS is.