Real Estate Operating Companies Data: Starts, Prices and Credit · Head-to-head
Census New Residential Construction — Housing Starts, Completions & Sales vs FHFA Data Portal — House Price Index, Conforming Limits & Mortgage Market Data
Which real estate operating companies data: starts, prices and credit data fits your job: Census New Residential Construction — Housing Starts, Completions & Sales, or FHFA Data Portal — House Price Index, Conforming Limits & Mortgage Market Data. API, files, or your warehouse. Daily, weekly, or hourly.
Census New Residential Construction — Housing Starts, Completions & Sales
FHFA Data Portal — House Price Index, Conforming Limits & Mortgage Market Data
Where the fields line up
No shared field names. These two answer different questions.
| Field | Census New Residential Construction — Housing Starts, Completions & Sales | FHFA Data Portal — House Price Index, Conforming Limits & Mortgage Market Data |
|---|---|---|
Year | Calendar year of the annual observation. The historical series begins with 1959. | not in this set |
Month | Observation period label on the monthly worksheets, carried in both seasonally adjusted and not-seasonally-adjusted forms. | not in this set |
United States | National total of housing units for the measured stage, expressed in thousands. | not in this set |
Northeast / Midwest / South / West | Census-region totals in thousands of units; every region column repeats beside the national figure on the same row. | not in this set |
Total | All housing units within a geography, regardless of structure size. | not in this set |
1 unit | Single-unit structures - the closest published proxy for traditional single-family detached supply. | not in this set |
2 to 4 units | Small multifamily structures. Not shown separately before August 1963, because through July 1963 the 5-units-or-more column also carried 2-to-4-unit buildings. | not in this set |
5 units or more | Large multifamily structures, five or more units per building; absorbs the 2-to-4-unit class through July 1963. | not in this set |
(NA) | Not-Available marker printed where the requested detail does not exist - a structural absence, not a suppressed value. | not in this set |
(S) | Flagged because the value does not meet publication standards: tests for identifiable and stable seasonality fail the reliability bar. | not in this set |
hpi_type | not in this set | Type of index series: traditional, developmental, distress-free or non-metro. |
hpi_flavor | not in this set | Flavor of HPI based on the underlying transaction data: purchase-only, all-transactions or expanded-data. |
Coverage, side by side
| Census New Residential Construction — Housing Starts, Completions & Sales | FHFA Data Portal — House Price Index, Conforming Limits & Mortgage Market Data | |
|---|---|---|
| Geographic | Fixed regional columns: United States, Northeast, Midwest, South, West | `level`, `place_name` and `place_id` carried on every row, nation to census tract |
What each contains
Pick by fit, not by loyalty.
| Census New Residential Construction — Housing Starts, Completions & Sales | FHFA Data Portal — House Price Index, Conforming Limits & Mortgage Market Data | |
|---|---|---|
| Reference period | `Year` plus `Month` label each observation | `yr` plus `period` integers (1-12 monthly rows, 1-4 quarterly) |
| Observed value | Count columns in thousands of housing units (national, regional and structure-size bands) | `index_nsa` and `index_sa` index values |
| Geography | Fixed regional columns: United States, Northeast, Midwest, South, West | `level`, `place_name` and `place_id` carried on every row, nation to census tract |
| Adjustment basis | Parallel worksheet tabs: Not Seasonally Adjusted and Seasonally Adjusted | Paired columns on one row: `index_nsa` beside `index_sa` |
| Market segmentation | Structure-size bands: `Total`, `1 unit`, `2 to 4 units`, `5 units or more` | `hpi_flavor` purchase-only / all-transactions / expanded-data, plus `hpi_type` |
| Quality flags | `(NA)` marks unavailable detail; `(S)` flags unstable seasonality | — |
| Frequency dimension | Implied by worksheet layout: annual, NSA monthly, seasonally adjusted monthly | Explicit `frequency` field: monthly or quarterly |
What each does better
Census New Residential Construction
Absolute unit counts, summable across 65-plus years. The values are thousands of housing units, not index points: the 1959 annual row prints 1,517 thousand total US starts with 1,234 thousand of them single-family, falls to 1,252.2 thousand in 1960, recovers to 1,313 thousand in 1961. The Northeast column runs 268.7, 221.4, 246.3 across the same years, so a regional share model needs nothing else; see housing starts.
Time-to-build economics nobody else in this pairing measures. Four length-of-time workbooks quantify months from permit authorization to start and from start to completion, in average and percentile forms - direct inputs to operating-company construction-timeline models. Methodology matters at the edges: month-level precision means same-month authorization-to-start counts as zero, and pre-permit ground breaking can push values negative.
the FHFA Data Portal
A geography ladder that ends at neighborhoods. The level dimension steps from the nation through Census divisions, all 50 states plus DC, Puerto Rico, MSAs, counties, 3-digit ZIP codes, 5-digit ZIP codes and finally census tracts. The flagship master HPI file holds roughly 184,000 rows - about 17 MB - and a sample row reads like a dimensional fact table: traditional type, purchase-only flavor, monthly frequency, East North Central Division (DV_ENC), 1991 period 1, index_nsa and index_sa both 100.0. Census tops out at four regions.
The verdict
Verdict: sample both, pick by fit. If your question counts homes - starts turning into completions, single-family separating from five-or-more-unit multifamily, the West building faster than the Midwest, how long a permitted home takes to finish - Census New Residential Construction — Housing Starts, Completions & Sales answers in units you can quote.
With identical 9/10 scores, let returned rows decide rather than brand familiarity: cut each sample to your geographies, stages and date windows, and keep whichever matches the shape of your model.
Sample both, pick by fit. See Census New Residential Construction — Housing Starts, Completions & Sales · See FHFA Data Portal — House Price Index, Conforming Limits & Mortgage Market Data
Fair questions
Is Census New Residential Construction better than the FHFA Data Portal?
Neither is better outright - both score 9/10 with verified field dictionaries. Census wins counting questions: starts, completions and sales in thousands of units by region and structure size back to 1959.
Which dataset reaches smaller geographies?
The FHFA Data Portal, decisively. Its level dimension steps from the nation through divisions, states, MSAs, counties, 3-digit and 5-digit ZIP codes to census tracts, with about 184,000 rows in the master HPI file alone. Census publishes national totals and the four Census regions, though its Survey of Construction microdata does reach individual sampled buildings.
Do the two datasets overlap anywhere?
On frame rather than metrics: both cover the American residential market, both publish national figures, and both carry monthly and quarterly observation frequencies. Their measured quantities never intersect - Census reports counts of homes in thousands of units, FHFA reports index values and lending parameters such as county conforming loan limits.
How far back does each dataset's history run?
Census reaches furthest: annual series begin in 1959 - 65-plus years - with monthly unadjusted and seasonally adjusted sheets and press releases archived since 1995.
Can Datadory deliver both datasets together?
Yes. Both records merge onto one delivery calendar, with geography joined where the two frames meet - Census regions to FHFA divisions or states - and periods reconciled between Year-Month labels and yr-period integers. Request a sample of either record first, or ask for both merged and cut to the geographies and date windows you model.