FHFA Data Portal — House Price Index, Conforming Limits & Mortgage Market Data

It is the federal government's ledger of American house prices and mortgage-market rules, written by the agency that regulates Fannie Mae, Freddie Mac and the Federal Home Loan Banks. The FHFA Data Portal gathers the agency's downloadable datasets in one place, and its centerpiece is the FHFA House Price Index - a weighted repeat-sales measure of single-family house price change built from purchase and refinance mortgage transactions on properties financed through the Enterprises since January 1975. Because it re-measures the same houses across sales, the index isolates pure price movement from the mix of what happens to sell in a given month.

Two dimensions lift it above the headline number everyone quotes. First, altitude: the index publishes monthly for the United States and the nine Census Divisions and quarterly for states, metros, non-metro areas, Puerto Rico, manufactured homes, three-digit ZIPs, five-digit ZIPs, counties and census tracts - so a national claim can be tested against the geography your portfolio occupies. Second, flavors: purchase-only series use repeat purchases alone, all-transactions series add refinance appraisals, expanded-data series fold in FHA and county recorder records to push coverage below the conforming loan limit, and distress-free variants screen out foreclosure-driven transactions.

Around the indexes sit the rest of the portal: county conforming loan limits recalculated each November under the HERA formula and published back to the pre-HERA era, Uniform Appraisal Dataset aggregate statistics at national through census-tract level, National Mortgage Database aggregates on new, outstanding and performing residential mortgages, Duty to Serve eligibility and performance files, Enterprise Housing Goals, FHLBank membership and stress-test data, the NSMO Public Use File, Underserved Areas Data, a MIRS Transition Index, Public Use Databases and interactive dashboards for HPI, Duty to Serve, Borrower Assistance, UAD and loan limits. For real estate operating companies, that combination answers the questions that move net operating income: how fast collateral values are moving, in which geographies, and where the financing ceilings sit.

Two observations from the consolidated HPI master file exactly as they arrive:

# FHFA House Price Index -- master file rows (traditional / purchase-only)
# one row per geography x frequency x period; index base per file conventions

hpi_type   : traditional       hpi_flavor : purchase-only
frequency  : monthly           level      : USA or Census Division
place_name : East North Central Division          place_id : DV_ENC
yr         : 1991              period     : 1
index_nsa  : 100.0             index_sa   : 100.0

hpi_type   : traditional       hpi_flavor : all-transactions
frequency  : monthly           level      : State
place_name : Abilene, TX*      place_id   : 10180
yr         : 2013              period     : 10
index_nsa  : 370.65            index_sa   : 369.12

markers    : *illustrative pairing of the dictionary's worked examples;
             rstderr carries relative standard error on expanded-data
             estimates, note carries row footnotes

Read the first row and the mechanics show themselves: level and place_id identify the geography, frequency and period locate the observation inside the year, and the two index columns sit beside each other so seasonal adjustment becomes a column choice rather than a separate download. The 1991 print reads exactly 100.0 - the file's baseline convention - which makes every later value directly interpretable as percent appreciation off that base.

The asterisk matters more than it looks. The second row pairs the two worked examples the published dictionary itself uses - the place name and the CBSA code - to demonstrate the join keys rather than to quote a specific metro print. The full master file runs to roughly 184,000 such rows across flavors, frequencies and geographies, and the same keys repeat on every flavor, so a division-level backtest joins to a tract-level drill-down without reshaping anything.

Twelve fields carry every row of the HPI master file: hpi_type, hpi_flavor and frequency describe which series you are reading, level, place_name and place_id locate it in space, yr and period locate it in time, and index_nsa and index_sa hold the not-seasonally-adjusted and seasonally adjusted values side by side. rstderr publishes the relative standard error where an estimate warrants one, and note carries row footnotes and revision remarks as data rather than PDF margin notes.

Everything beyond this verified core folds out under additional fields on request - the county conforming loan-limit tables, UAD appraisal aggregate statistics, NMDB mortgage aggregates, the annual tract, ZIP and county index files, volatility and median-price parameters, the Duty to Serve and Housing Goals families and the NSMO and Public Use Databases. Name the groups you want when requesting the sample and the extract arrives carrying exactly those tables.

Geography: national totals and nine Census Divisions, all 50 states plus DC and Puerto Rico, metropolitan statistical areas, counties, three-digit ZIPs, five-digit ZIPs and census tracts - the deepest published altitude of any federal house price series. Level availability varies by product: the monthly series stop at divisions, while the annual all-transactions developmental series reach tracts. Conforming loan limits resolve to individual counties, which is the geography origination actually turns on.

Temporal: the repeat-sales panel starts in January 1975, giving five decades that span every modern housing cycle; the county loan-limit ladder reaches back to the early 1970s; UAD appraisal aggregates run from 2013Q1 onward; NMDB mortgage aggregates reach back to 1998.

Granularity: monthly index observations for the US and divisions, quarterly for everything below them, annual values in the county loan-limit tables and the tract and ZIP files, and quarterly-to-annual cuts in the appraisal and mortgage aggregate sets. Set against the wider Datadory catalog, few price panels combine fifty years of depth with census-tract resolution - this is the reference standard we deliver against.

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

Pick the channel your team already works in: flat files sized for batch loads of the full fifty-year index history, structured payloads for products that surface live valuations inside an underwriting tool or dashboard, or a direct pipe into Snowflake, BigQuery or Redshift. Cadence is yours to set - daily, weekly, or hourly - and to change when your models change.

Every delivery ships with the field dictionary, validation rows and a schema that holds steady between deliveries, so a 1975 division row and this month's print load from one table.

• Collateral valuation and AVM calibration - model teams anchor valuations and desk reviews to repeat-sales levels at the metro, ZIP and tract geographies portfolios actually occupy.
• Loan-limit-aware origination - lenders read each county's conforming maximum before pricing, so jumbo-versus-conforming boundaries come from the regulator's current table.
• Home price risk research - quants build factor exposures and drawdown scenarios from the monthly national and division series, then localize them with state and metro flavors.
• Portfolio marking - owners push regional appreciation through hold-period models to project asset values and refinancing windows.
• Appraisal-quality monitoring - compliance teams benchmark valuation distributions with UAD aggregates matched to their own geography and quarter.
• Mortgage market structure work - economists trace origination, balance and performance trends with NMDB aggregates from a nationally representative sample.
• Affordability and policy commentary - analysts pair five decades of index history with the loan-limit ladder to size how much housing stock sits above the conforming ceiling.

Workflow-level applications continue on the investors and quants use cases and market researchers use cases pages.

Investors and quant researchers get the official price input behind every MBS and REOC valuation model - index depth down to census tracts, long enough to span several housing cycles. Market researchers and consultants get the regulator's own home price measures and loan-limit tables ready for client decks. Data scientists and ML engineers get a documented schema with explicit flavor and geography keys - clean labels for valuation and credit models without hand-cleaning. Risk and compliance leaders get appraisal-quality baselines and origination limits from the agency that regulates the Enterprises. Journalists, academics and students get the housing indicator the Fed already quotes, citable with its methodology visible. Developers building data products get a schema that has survived methodology changes since the Ford administration, wired into proptech dashboards without babysitting.

Three things worth knowing upfront.

First, mind your flavors. Purchase-only, all-transactions, expanded-data and distress-free series answer different questions and move differently at turning points - a purchase-only index excludes refinance appraisals, and an expanded-data index reaches further below the conforming loan limit by folding in FHA and recorder records. Say which question you are answering and the sample arrives carrying the right flavor, not just the default.

Second, match the geography to the cadence. Monthly publication exists only for the US and Census Divisions; everything below that - states, metros, ZIPs, counties, tracts - publishes quarterly or annually. A request for "monthly tract-level indexes" has no matching series, and we would rather flag that boundary in writing than deliver a resampled proxy.

Third, know what lives outside the master file. The county conforming loan-limit tables, UAD aggregates, NMDB statistics, Duty to Serve files and the NSMO Public Use File are distinct products with their own layouts, and the exact schemas inside some zipped aggregate sets are documented in separate dictionaries - those details fold out under additional fields on request. Get a sample of this dataset and we route rows scoped to your geographies, flavors and date range, with the field dictionary attached.