METHODOLOGY
How Birdseye indexes a place
A Birdseye index is a place's position relative to every other place we measure, not a property of that place on its own.
Every number behind it traces to a documented public source, a stated standardization method, and a recorded vintage.
One harmonized geography
Birdseye holds data at four levels — tract, county, metro (CBSA) and ZIP — and the map can be viewed at any of the four, with the tract as the finest unit. Sources publish at whatever level suits them: county, metro, ZIP, school district, permit jurisdiction. Each is crosswalked onto that common set, with the interpolation or roll-up method recorded per metric and shown in the tool.
Where a value has to move between geographies, it is allocated by whichever share fits what is being measured — population counts by population share, housing measures by housing-unit share, tenure measures by occupied-unit share, genuinely spatial measures by land area. Where no allocation is meaningful, such as a county-wide rate, the tract simply takes the value of the larger geography containing it, and the tool labels which geography each number actually came from.
Sources and vintages
Birdseye is built from documented public sources, ingested on their own release schedules and harmonized onto one geography — Census, FHFA, HMDA, LEHD, NCES, FBI/state UCR, BLS and more.
Every metric carries its own vintage rather than the tool having one as-of date; the full list of sources, what each publishes and how current it is, is on the sources page.
See the full sources list.
From a measure to an index
Each metric is converted to a z-score, which states how far a place sits from the national average in standard deviations. Dollar-denominated and other right-skewed measures are logged before that comparison, so a handful of extreme places cannot compress every ordinary one into a narrow band; which metrics receive this treatment is recorded per metric.
A growing set of sources uses a robust form instead — centered on the median and scaled by the 5th-to-95th-percentile spread — so a single outlier cannot inflate the scale for everyone else; a metric uses one form or the other at every geography it is served from, never a mix. Every z-score is then capped at ±4 standard deviations, loose enough to preserve real differences between genuinely extreme places.
Direction — which end of a metric counts as better — is decided once per metric and held in one authority rather than re-guessed by each chart; a metric's default is stated on its own page in the metric catalog, and a user can still choose which end they are rewarding. Wherever the tool shows a value, it is placed next to a benchmark for the metro and the nation, so a number is never presented with nothing to compare it to.
Indexes and Topics
Six Indexes answer a plain question a buyer actually has. Each blends a handful of the metrics above, at a published default weight:
| Index | Default weight | Members | Promise |
|---|---|---|---|
| Affordability | 100% | 5 | What it costs to buy or rent here, against local incomes |
| Safety | 0% | 2 | Reported crime and the visible distress that goes with it |
| Schools | 0% | 4 | How the local public schools perform |
| Economy | 0% | 4 | Jobs, incomes and work within reach |
| Appreciation | 0% | 3 | How much home values have risen here, over one, three and five years |
| Accessibility | 0% | 5 | How easily you can reach work and get around from here |
| Rental Yield | 0% | 3 | What a rental here earns relative to what it costs |
| Growth | 0% | 5 | Whether the people, jobs and incomes here have been increasing over the past five years |
| Buyer's Leverage | 0% | 4 | How much room a buyer has here right now, against this market's own normal |
| Stability | 0% | 5 | How steady this market and its job base have been |
| Lending Health | 0% | 3 | How readily, and how cheaply, mortgage credit is reaching buyers here |
That default mix is a stated editorial choice, not a byproduct of how many metrics happen to sit behind each Index, and it is fully adjustable — down to opening an Index and reweighting the individual metrics inside it. The six weights are set by the user and sum to 100.
An Index is computed for a geography only when it clears the member coverage gate — a 0.60 fraction of that Index's members have to actually be present there, i.e. at least 60%. A geography too thinly covered for an Index is marked rather than guessed at.
Separately, eleven Topics hold every metric the tool shows, including the many that belong to no Index at all — a Topic is a filing system, not a scoring construct.
Full definitions are in the metric catalog; the formulas above are documented in more depth on the scoring page.
Suppressed, not estimated
On the map, grey means a tract is suppressed or below that coverage threshold — never filled in with an estimate. Suppression happens where a metric is unpublished for a geography, withheld for confidentiality, or absent because the underlying program does not cover that area.
Where a value is missing, it is excluded from a place's blend rather than replaced with an average; substituting the mean would make sparse data indistinguishable from complete data. A report page marks exactly which of its numbers are suppressed rather than silently omitting them.
What the numbers are not
These are estimates derived from public sources — Census, HUD, FBI, BLS and others — not official statistics issued by those agencies.
Nothing here is a scraped listing of unknown provenance, and no proprietary model sits between a source file and an index: every step from raw file to z-score is one of the transformations described above.
See it on the map
Open the country, weight what matters, and drill into any tract — or read the underlying docs.