Homebuilding Data Guide: 15 Datasets & Sources (2026)

The Homebuilding data landscape in 2026

The U.S. Census Bureau anchors the macro side through four programs. The Building Permits Survey (quality score 9, the highest in the slice) publishes monthly, year-to-date and annual unit counts down to county and place level. New Residential Sales carries median prices, months' supply and inventory by stage of construction back to 1963. The annual Characteristics of New Housing series quantifies square footage, bedrooms, bathrooms and wall material. And the Survey of Construction program hub distributes the microdata behind all of them, reaching individual sampled units from 1999 to 2025.

Where should a homebuilding data project start?

Start with the pipeline you are trying to model. Authorization-to-completion tracking pulls from the Census programs, where every file is free and machine-readable. Cycle timing adds the NAHB Housing Market Index as a sentiment overlay. Product and floor-plan work adds Characteristics of New Housing, and machine-learning projects branch into the Hugging Face furniture corpora. The use-case walkthrough below maps each workflow to named sources.

Government construction statistics (US Census Bureau)

US Census Building Permits Survey (BPS) (quality score 9) is the public-domain backbone: monthly, year-to-date and annual counts of new privately-owned housing units authorized by building permits, broken out by structure size (1-unit, 2-units, 3-4 units, 5+ units) at national, Census region, state, CBSA/metro, county and place level. Excel files cover 2019 to present, comma-delimited ASCII history runs from 1995 to 2019, and the Time Series/Trend Charts tool extends U.S. and region series back to 1960. Expect hundreds of ASCII/Excel files per year across four regional folders, each place-level monthly file roughly 100-250 KB covering thousands of jurisdictions. Island Areas geography was added in the 2025 annual release.

US Census New Residential Sales (quality score 8) tracks what happens after authorization: new single-family houses sold and for sale, median and average sales prices, and inventory by stage of construction, at national and regional level. The historical record is ten Excel files of roughly 35-85 KB each holding annual series from 1963 to present with monthly detail in recent years; price-range files cover 1999 onward across three vintages of band definitions, and quarterly price-and-financing cross-tabs round out the set. Coverage is new, privately-owned units only — existing homes are excluded per Survey of Construction scope.

US Census Characteristics of New Housing (quality score 8) is the physical-product record: roughly 90 Excel summary tables per year (100-150 KB each) plus one annual microdata file of individual housing records, quantifying square footage, bedrooms, bathrooms, wall material, sales prices and financing. Summary tables reach back to 1973, microdata covers 1999 through 2025, and the 2025 data was released July 1, 2026. One access quirk: SAS-format files require a phone request.

US Census Survey of Construction (SOC) Program Hub (quality score 8) is the program hub behind New Residential Construction, New Residential Sales, Construction Price Indexes and Characteristics of New Housing. Its distinguishing asset is record-level microdata — one file per year from 1999 to 2025 covering tens of thousands of sampled units with Census division and metropolitan area identifiers, containing no address-identifying information and flagging modified values.

Trade-association sentiment

Retail and discovery layers

Each major homebuilding workflow maps to specific named sources rather than generic categories:

1. Track the single-family pipeline from authorization to sale. Join US Census Building Permits Survey counts at county and place level with New Residential Sales absorption and months'-supply figures; the related-pool record New Residential Construction (Building Permits, Housing Starts & Completions) covers the same Census pipeline and serves as the natural cross-check for anyone feeding single family starts forecast inputs.

1. Read builder sentiment as a leading indicator. The NAHB/Wells Fargo Housing Market Index publishes monthly 0-100 values by nation and Census region back to January 1985 — about 500 observations of housing market index history for cycle-timing models.

1. Set product and floor-plan strategy against measured outcomes. Characteristics of New Housing distributes square-footage, bedroom, bathroom and wall-material distributions across roughly 90 tables per year, with summary series reaching back to 1973.

1. Screen metro-level demand for land acquisition. The Building Permits Survey geography spans state, CBSA, county and place, with Island Areas added in the 2025 annual release — fine enough to rank individual jurisdictions rather than whole metros when searching building permits by county.

1. Separate builder pricing power from market appreciation. Set the new home sales median price series from New Residential Sales against the FHFA House Price Index from the related pool — the spread isolates what builders capture versus market-wide movement.

Four kinds of teams drive most homebuilding data searches, and each starts from a different source:

Data scientists build pipeline and sentiment models. Their queries read like "housing market index history" and "months supply of new homes," and the answers are the NAHB Housing Market Index's monthly 0-100 series back to January 1985 and New Residential Sales' inventory-by-stage-of-construction files alongside its median-price history. Our data scientists page collects these modeling workflows in one place.

Market researchers size demand geographically before anyone breaks ground. They search "building permits by county" and "characteristics of new housing data," landing on the Building Permits Survey's county- and place-level counts by structure size and Characteristics of New Housing's roughly 90 annual tables. The market researchers page maps the sizing workflow end to end.

Investors and quant researchers use sentiment as a cycle signal. The Housing Market Index's regional series and moving averages, joined with New Residential Sales counts and prices back to 1963 and cross-checked against the related-pool FHFA House Price Index and American Housing Survey, let them separate builder pricing power from market-wide appreciation.

A practical homebuilding data stack stacks four layers, cheapest and most durable first:

1. Pipeline layer. Anchor everything to the Building Permits Survey — monthly, YTD and annual unit counts by structure size from national level down to place, free and commercial delivery terms — and cross-check authorization-to-completion tracking against the related-pool Census New Residential Construction record. Join New Residential Sales for counts sold, months' supply and inventory by stage of construction, budgeting for revisions as later releases add survey returns.

Every figure below comes straight from the cataloged source records, as of August 2026:

Continue into the connected pages: