Methodology

How the score is built

The composite is a weighted average of five national percentile ranks. Nothing is proprietary and nothing is smoothed by hand: the weights are below, the inputs are named, and every market page prints the arithmetic that produced its own number.

The five components

ComponentWeightMeasuresDirection
Job growth0.30Employment, trailing 12 months vs the 12 beforehigher is better
Rent trend0.2512-month change in market renthigher is better
Affordability0.15Rent as a share of median household incomelower is better
Supply pressure0.15Permits per 1,000 residents and months of unsold inventorylower is better
Climate risk0.15Expected annual building loss as a share of building valuelower is better

Rules that matter more than the weights

Nothing is ever stubbed at a neutral value

A component with no data is absent, not 50. Absent components have their weight redistributed across the ones that exist, and each page records how much of the full scale was actually measured. Filling a gap with a neutral number invents data and flatters thin markets.

A score is published only when every buildable component exists

An earlier version scored any market with at least two components. The leaderboard immediately filled with tiny micropolitan areas measured on two inputs, because a two-component index has nothing to pull it back toward the middle and so produces more extreme values than a five-component one. Markets missing a component now get a page, all their raw figures, and an explicit reason instead of a number that implies a comparability they do not have.

Job growth is ranked within its own source pool

BLS publishes payroll employment (CES) for only about 431 metro areas. The rest use LAUS resident employment, which covers every metro and micropolitan area. The two measure different things — jobs located in a metro versus employed people living there — and their growth rates correlate at about 0.65, not 1.0. Each metro is therefore ranked against markets measured with the same instrument, and every page says which one it used.

County labor evidence is descriptive, not part of the metro score

County pages use the BLS Quarterly Census of Employment and Wages to show annual covered jobs, average weekly wages and the largest disclosed private supersector. QCEW measures workplaces located in the county, while LAUS measures employed residents and CES measures payroll jobs for selected metros. These series answer different questions, so the county evidence is labelled separately and never substituted into the five-component metro score.

Outliers are clipped before ranking

Rent growth, job growth and rent burden are winsorised at the 1st and 99th percentile before percentile ranks are computed. Small markets throw occasional impossible prints — one metro showed rent up 41% in a year — and one bad figure should not be allowed to define the top of a ranking.

Climate risk is a loss ratio, not a dollar total

FEMA’s headline National Risk Index score is built from expected annual loss in dollars, so it scales with how much property exists: Los Angeles County scores 100 while a county of 64 people scores 0.03. Scoring on it would punish every large metro for being large. We use the annualised loss ratio against building value, which is size-independent and is what an insurance premium tracks. The two correlate at only 0.40.

Invariants

Before any page is written, the build checks that component contributions sum to the published score, that weights sum to 1.000, that gross yield equals rent × 12 ÷ price, and that the rent burden and price-to-income ratio imply the same household income. A violation fails the build; it is never a warning. Each market page prints its own checks so a reader can repeat them.

Worked example — Crawfordsville, IN scores 88: job 25.20 + rent 23.25 + afford 14.55 + supply 11.85 + climate 13.65.

What the score is not

It is not a prediction, a recommendation, or an underwriting model. It ranks markets against each other on measurable conditions today. Property-level economics — the actual building, its condition, its financing and its insurance quote — decide whether a deal works, and none of that lives in public data.