Single-Family Residential REITs Data: Vacancy History, ZIP-Level Prices and the Listed Tape · Head-to-head
Census Housing Vacancies and Homeownership (CPS/HVS) vs FHFA House Price Index Datasets
Which single-family residential reits data: vacancy history, zip-level prices and the listed tape data fits your job: Census Housing Vacancies and Homeownership, or FHFA House Price Index Datasets. API, files, or your warehouse. Daily, weekly, or hourly.
Census Housing Vacancies and Homeownership (CPS/HVS)
FHFA House Price Index Datasets
Where the fields line up
No shared field names. These two answer different questions.
| Field | Census Housing Vacancies and Homeownership | FHFA House Price Index Datasets |
|---|---|---|
year | documented | not in this set |
rental_vacancy_rate_q1_q4 | documented | not in this set |
homeowner_vacancy_rate_q1_q4 | documented | not in this set |
inventory_category | documented | not in this set |
inventory_estimate_units | documented | not in this set |
margin_of_error | documented | not in this set |
percent_of_total | documented | not in this set |
year_quarter_historical | documented | not in this set |
vacancy_rate_geography_column | documented | not in this set |
cell_value | documented | not in this set |
seasonally_adj | documented | not in this set |
hpi_type | not in this set | documented |
Coverage, side by side
| Census Housing Vacancies and Homeownership | FHFA House Price Index Datasets | |
|---|---|---|
| Granularity | Aggregates by geography and householder characteristic - age of householder, race/ethnicity, family status, units-in-structure | Aggregates by place code and series flavor - no transactions, loans or properties identified |
What each contains
Pick by fit, not by loyalty.
| Census Housing Vacancies and Homeownership | FHFA House Price Index Datasets | |
|---|---|---|
| Documented fields | 11 | 13 (the master file itself is twelve columns) |
| Field types | Year and quarter row headings, percent-rate columns by quarter, inventory categories as rows, thousands-of-units counts, margin-of-error and percent-of-total figures, seasonally-adjusted yes/no flag | String enums for index type and flavor, frequency and level markers, place name and place code, integer year and period, paired unadjusted and adjusted index levels, relative standard error, dollar median price |
| Signature fields | `Rental Vacancy Rates (Q1-Q4)` (7.3); `Homeowner Vacancy Rates (Q1-Q4)` (1.2); `Type (Table 4)` inventory categories; `Margin of Error`; `seasonally_adj` | — |
| Shared concepts | Geography key, period stamp, numeric estimate, seasonal-adjustment treatment, reported uncertainty | Geography key, period stamp, numeric estimate, seasonal-adjustment treatment, reported uncertainty |
| Entity granularity | Aggregates by geography and householder characteristic - age of householder, race/ethnicity, family status, units-in-structure | Aggregates by place code and series flavor - no transactions, loans or properties identified |
| Overlap verdict | Structural overlap only: both reduce to place x period x estimate. Occupancy percentages versus 100-based price levels never join natively. | — |
What each does better
Census Housing Vacancies and Homeownership
FHFA observes transactions, never households.
A demand-side signal with a macro pedigree. The rental vacancy rate is a component of the Conference Board's index of leading economic indicators, making the series a standard input for residential demand analysis beyond REIT screens - rental vacancy rate has a vocabulary around it that a repeat-sales index does not touch.
Depth of history. The first row of the historical rental vacancy table reads 1956 1st quarter: 6.2 percent nationally, 4.8 inside metro areas, 3.3 in the Northeast. Inventory estimates begin in 1965.
Sampling honesty built into every table. Each estimate ships with its margin of error, including for year-over-year differences, so a 0.1-point move can be told apart from noise.
FHFA House Price Index Datasets
Geography all the way down. Annual developmental indexes descend to counties, five-digit ZIP codes and individual census tracts - the sample tract file carries rows like 01001020100, AL, 1998, index 100.00 against a 100.08 year-2000 baseline - with quarterly series for states, metros and three-digit ZIPs. Census HVS stops at the 75 largest metros - decisive for a single-family rental buy-box screened block by block.
A finer time axis. Purchase-only indexes run monthly for the US and the nine census divisions from January 1991, where the survey produces strictly quarterly estimates.
Disclosure rules are explicit: an MSA needs at least 1,000 total transactions with at least 10 in a quarter, and eleven large metros split into OMB metropolitan divisions.
Where they're equivalent
Both are federal statistical producers, not vendors. The U.S. Census Bureau fields the survey; the Federal Housing Finance Agency models the indexes from enterprise mortgage records. Neither resells someone else's work.
Both carry verified field dictionaries and the top rubric score. Two of only 145 datasets scored 10/10 across the 1,744-record catalog, whose average sits at 7.81 - and both dictionaries were confirmed rather than inferred in the same August 21, 2026 research pass.
Neither observes a transaction or a household directly. One publishes sample-based aggregate rates; the other anonymized transactions compressed into index levels. No addresses, no owners, no loan identities appear in either dictionary.
Both are single-country by construction. All 50 states plus DC on the Census side; the same footprint plus Puerto Rico on FHFA's. For a second country, neither record helps - though the slice's secondary pool (the American Housing Survey National Microdata among them) extends the domestic picture.
Both wear seasonality on their sleeve. Adjusted and unadjusted values travel side by side in each, so you decide which basis your model sees.
The verdict
Verdict: sample both, pick by fit - tied on score, divided by question.
Take Census Housing Vacancies and Homeownership (CPS/HVS) if your unit of analysis is occupancy or tenure. Absorption forecasting for lease-up models, market-tightness dashboards, owner-versus-renter demand shifts, demographic cuts of who owns (65.0 percent owned in Q2 2026) - that is this dataset's home turf, and the investors quants use cases lean on exactly that shape. Accept a quarterly rhythm and a 75-metro ceiling.
ZIP- and tract-level valuation comps, buy-box screening, collateral-risk monitoring, appreciation studies back to the mid-1970s, volatility-adjusted return models - the developers builders use cases live here.
Rule of thumb: if the question begins "how many units sit empty", take Census HVS; if it begins "what did values do", take FHFA HPI. Both hang off the single family residential reits data hub.
Sample both, pick by fit. See Census Housing Vacancies and Homeownership · See FHFA House Price Index Datasets
Or take both in one feed
Yes - this pairing works because the two series measure opposite sides of the same trade. A defensible loop: screen markets with FHFA's tract and ZIP appreciation series (repeat-sales method), then qualify demand with Census HVS - where the Q2 2026 gap between the 7.3 percent rental vacancy rate and the 1.2 percent homeowner vacancy rate spans 6.1 points, a reminder that rental supply sits looser than for-sale supply in the same national market. The CPS/HVS housing vacancy survey entry unpacks how those rates are constructed.
Three cautions straight from the records. First, no join key: FHFA keys rows to coded places (place_id, CBSA codes, tract FIPS), Census HVS keys tables to named geographies and quarter headings, so the crosswalk is yours to maintain. Second, units collide: percent rates against 100-based index levels - never average them, ratio them. Third, periodicities disagree: monthly divisions against quarterly metros, so resample before aligning.
Datadory ships either record alone or both merged onto one calendar, normalized so those cautions are handled upstream, delivered daily, weekly, or hourly - your call. Or take both in one feed.
API, files, or your warehouse. Daily, weekly, or hourly.
Fair questions
Do the two datasets cover the same ground anywhere?
Structurally, yes. Both schemas solve the same four problems - a geography key, a period stamp, a numeric estimate and an explicit seasonal-adjustment treatment - and both add an uncertainty column, Margin of Error on one side and relative standard error on the other. Substantively they barely intersect: one measures whether units are occupied, the other what sold homes were worth.
Which dataset reaches further back in time?
Depends on the series you name. Census HVS carries quarterly rental and homeowner vacancy rates back to 1956 and housing inventory to 1965. FHFA HPI reaches the mid-1970s for its expanded-data national series, starts monthly purchase-only indexes in January 1991, notes all-transactions series extend earlier, and runs annual tract indexes from 1998 with 1990 and 2000 baseline columns.
Which dataset covers smaller geographies?
Its annual developmental indexes descend to counties, five-digit ZIP codes and individual census tracts, with quarterly state, metro and three-digit ZIP series alongside. Census HVS tops out at the 75 largest metropolitan areas plus states from 2005 and four census regions, with metro and non-metro breakdowns below the national line.
Can Datadory deliver both datasets together?
Yes. Either record arrives alone or merged onto one delivery calendar, normalized so percent rates stop colliding with 100-based index levels and monthly periods stop colliding with quarterly ones, delivered daily, weekly, or hourly - your call. Name the geographies, the household cuts and the date windows when you request the sample and it lands pre-cut, with field definitions attached.