The Address Index
How the numbers are made

Greenwich methodology and revision history

The Address Index publishes Greenwich market measures from public, source-linked records. This page explains what each measure is, how much evidence sits behind it, and every material revision to a published reading.

Index revised Sep 19, 2026. The 2026 Q2 Address Index reading was revised from 105.7 (-1.7% YoY, nowcast) to 120.7 (+11.4% YoY, reconstructed).

Newer public parcel sale histories added 447 usable sales and extended the fitted repeat-sales index from 2025 Q3 through 2026 Q2. Quarters after the latest state transaction file (2025 Q3) are now estimated from those reconstructed repeat-sale pairs instead of the monthly-report nowcast, and 2025 Q2 and Q3 were revised up as more pairs were added.

Methodology and full revision history →

The Address Index

A weighted repeat-sales index: it compares sales of the same Greenwich single-family properties over time, so a month with unusually large or small homes does not masquerade as a price change. Quarterly estimates are smoothed over four quarters and scaled so the 2024 average equals 100.

Observed, reconstructed, nowcast

Observed quarters have direct state transaction coverage. Reconstructed quarters rely on public parcel sale histories or interpolation. A nowcast extends the index past the latest complete transaction period using official monthly reports, with a deliberately wider uncertainty band. Direct transaction coverage currently runs through 2025 Q3; the fitted index runs through 2026 Q2.

Uncertainty

The shaded range is a 95% statistical interval. It widens when direct observations are sparse and widens further for a nowcast. The center line is an estimate, not an appraisal.

12-month market level

A trailing 12-month price signal for single-family homes. Monthly medians are shown separately, as unconnected dots, because each month describes a different basket of sold homes.

Recorded sales

Every sale shown is an observed amount and recorded date from a public government source. Non-use transfers, unresolved price conflicts, street-only addresses and ambiguous parcel matches are excluded or left visibly unmatched. No sale is modeled or invented.

Property files

Owner-free parcel facts, assessments, recorded sales, planning filings and map screening. A mapped intersection is a screening flag, not a legal, engineering, insurance or zoning determination; a source that is not connected is shown as not connected.

Privacy

No owner names are collected or displayed anywhere on the research desk, in page metadata or in the downloadable data.

Data and validation

The dashboard snapshot and its validation record are published as validation JSON. Sources: State transactions, Town assessor archive.

Revision policy

Index revisions

Late public records can change a quarter we have already published. Each new build is compared with the reading in our latest published edition. When a published quarter moves by 2 index points or more, or its year-over-year change moves by 2 percentage points or more, the dashboard shows an “Index revised” note with the date and reason, and the revision is logged here.

RevisedQuarterPreviously publishedRevised toWhy
Sep 19, 20262026 Q2105.7 (-1.7% YoY)
nowcast · Aug. 6, 2026 edition (the dashboard showed 106.2, +0.4% YoY, until Sept. 19)
120.7 (+11.4% YoY)
reconstructed
Newer public parcel sale histories added 447 usable sales and extended the fitted repeat-sales index from 2025 Q3 through 2026 Q2. Quarters after the latest state transaction file (2025 Q3) are now estimated from those reconstructed repeat-sale pairs instead of the monthly-report nowcast, and 2025 Q2 and Q3 were revised up as more pairs were added.
  • 2025 Q4 to 2026 Q2 moved from nowcast to reconstructed evidence.
  • Direct state transaction coverage still ends in 2025 Q3.
  • The methodology version (address-index-1.0) did not change.

Directional market intelligence only — not an appraisal, brokerage opinion, or investment recommendation.