Real Estate

Real-estate data is organised by place and series, and the two are resolved separately. A place — "Austin, TX", a county, a ZIP code — becomes a stable geo_key. A geo_key plus a topic becomes a series_key. A series_key becomes observations. Skipping a step and guessing at a key is the one reliable way to get an unknown_entity back.

Geography is a graph, not a label. A match carries a bottom-up parents chain and outgoing crosswalks with a relation, a weight where one exists, a source, and a vintage — so an agent can move between a ZCTA and its county, or a county and its CBSA, and know how much of that mapping is approximate.

Resolve a place#

re_search_geographies(query="Austin, TX", limit=10)

{
  "data": {
    "query": "Austin, TX",
    "matches": [
      {
        "geo_key": "cbsa_12420",
        "level": "cbsa",
        "code": "12420",
        "name": "<CBSA title>",
        "state_code": "TX",
        "score": 1.0,
        "match_reason": "name_exact",
        "parents": [],
        "crosswalks": []
      }
    ]
  },
  "meta": { "source": "census_geo", "as_of": "2026-08-19T00:00:00Z", "notice": "..." },
  "pagination": null
}

Values are illustrative; field names are real. A blank query is a normal empty success rather than an error, and pagination is null because a relevance list is bounded rather than paged.

Find the series available for that place#

re_search_series(geo_key="cbsa_12420", query="house price", limit=10)

Matches sort by series_key and carry dataset and source metadata, units, frequency, seasonal adjustment, licence tag, and the observation coverage window — coverage_start and coverage_end, either of which can be null when nothing is ingested yet. Reading the coverage window before you request observations is how you avoid asking for a decade of a series that starts in 2019.

Either filter works alone, but at least one is required: a call with neither query nor geo_key is a bad_parameter, and an unknown geo_key is an unknown_entity with suggestions.

Read the observations#

re_get_series(series_key="fhfa_hpi_at_quarterly_nsa_cbsa_12420", start="2020-01-01", end="2026-06-30", limit=100)

{
  "data": {
    "series_key": "fhfa_hpi_at_quarterly_nsa_cbsa_12420",
    "dataset_key": "fhfa_hpi_at_quarterly_nsa",
    "source": "fhfa",
    "source_series_id": "<publisher series id>",
    "title": "FHFA House Price Index (all-transactions, NSA)",
    "attribution": "Data sourced from the FHFA House Price Index.",
    "units": "index",
    "base_period": "1995Q1=100",
    "currency": null,
    "frequency": "quarterly",
    "seasonal_adjustment": "not_seasonally_adjusted",
    "license_tag": "us-gov-public-domain",
    "geography": {
      "geo_key": "cbsa_12420",
      "level": "cbsa",
      "code": "12420",
      "name": "<CBSA title>",
      "state_code": "TX"
    },
    "start": "2020-01-01",
    "end": "2026-06-30",
    "observations": [
      {
        "period": "2026-04-01",
        "period_type": "quarter",
        "value": "412.250000",
        "as_of": "2026-08-19T00:00:00Z",
        "source_vintage": "2026Q2",
        "effective_date": "2026-08-19",
        "meta": null
      }
    ]
  },
  "meta": { "source": "fhfa", "as_of": "2026-08-19T00:00:00Z" },
  "pagination": { "limit": 100, "has_more": true, "next_cursor": "eyJ2IjoxLCJrIjpbLi4uXX0" }
}

Observations are newest-first and page with a cursor; keep series_key, start, and end constant while you walk backwards. Bounding the window with start and end is almost always better than paging a whole history you will discard.

• This is market-level data, not property-level. Everything here describes a geography — a CBSA, a county, a ZCTA, a state — over a period. There are no addresses, parcels, listings, or transactions in this category, and no tool that will synthesise them.
• Latest values only. re_get_series serves the current value for each period. It has no as_known_at parameter, so stored vintages are not queryable here; an observation's source_vintage and effective_date tell you which publication a value came from, but you cannot ask what the series looked like on a past date.
• Sources differ in kind, and it matters. Index levels, fair market rents, survey estimates, and hazard event counts are all in this category and none of them mean the same thing. Read units, base_period, and seasonal_adjustment off the series before comparing anything to anything.
• Survey estimates carry uncertainty. Census ACS observations come with 90-percent margins of error in the observation's own meta. An estimate quoted without its margin is a smaller-sounding number than the data supports.
• Geography vintages are dated. Delineations and relationship files are versioned, and a crosswalk carries the vintage it came from. A county-to-CBSA link has no weight; a ZCTA-to-county link can.