Fhfa House Price Index By County Data: Dataset Structure and Field Coverage

Twelve documented variables per master-file row, each defined with types and worked examples in hpi_dictionary.xlsx (17,456 bytes) alongside hpi_specifications.xlsx for file layouts:

• hpi_type - traditional, developmental, distress-free or non-metro
• hpi_flavor - purchase-only, all-transactions or expanded-data
• frequency - monthly or quarterly (period runs 1-4 for quarterly rows, 1-12 for monthly)
• level and place_name - geography tier plus its human-readable name
• place_id - state abbreviation, CBSA code or division code such as DV_ENC
• yr and period - observation date
• index_nsa and index_sa - the index value not seasonally adjusted, and seasonally adjusted where available
• rstderr and note - relative standard error where published, plus footnote or revision remarks

A worked sample row shows the shape: hpi_type traditional, 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. By period 3 of the same year the series reads 101.32 NSA against 100.90 SA — the gap is the seasonal adjustment doing its work.

Five products on one portal cover the county question from different angles, and they differ enough to pick deliberately:

The family starts deep: the House Price Index runs from January 1975, or first quarter 1975 for quarterly series, to present. Two caveats shape expectations. First, flavor and history interact — monthly purchase-only series begin January 1991, and only all-transactions data publish before that year, so a county appreciation model reaching into the 1980s leans on the all-transactions flavor. Second, the finest geographies are younger: annual census-tract indexes begin in 1998, and UAD appraisal aggregates start at 2013Q1.

The county-adjacent series reach back furthest of all: conforming loan limit tables run year by year back to the early 1970s, which lets you model buyer purchasing power over nearly the whole modern mortgage era. National Mortgage Database aggregates add new, outstanding and performing-mortgage totals from 1998 onward, though at national rather than county resolution.

Three checks keep a county model honest:

For appraisal-side context at finer geography, UAD Appraisal Aggregate Statistics zip county and census-tract appraisal distributions from 2013Q1 — useful cross-checks, not substitutes, for the price index itself.