The Cities Whose Schools Outperform Their Income

School ratings and household affluence track each other so closely that most rankings mostly measure income. Isolating what income does not explain produces a different list.

Across 3,265 cities in 42 states, mathematics proficiency falls as the share of students eligible for free or reduced-price lunch rises, at a correlation of -0.63 — the economic profile of the student body alone accounts for roughly 40% of the variation in how cities test when every state is pooled together. That is the uncomfortable fact underneath most school rankings: sort cities by proficiency and you have substantially sorted them by income.

A more useful question is which places test above what that relationship predicts. AreaProfile fits the line separately inside each state, because each state writes its own assessment, and then measures how far each city sits above or below its own state’s line. What follows is the residual, not the raw score.

Testing furthest above the state line

And furthest below

A residual is not a district rating and not a causal claim. It absorbs curriculum, funding formulas, teacher retention, how a district distributes students across buildings, and ordinary measurement error, alongside anything a school itself is doing. Proficiency is from the 2018–19 assessments, the last full year of testing before the pandemic. The free or reduced-price lunch share is from 2024–25, because that is the most recent year the federal collection publishes it. So the line is fitted across a six-year gap with the pandemic inside it, and a city whose economics moved between those years carries some of that movement in its residual rather than anything about its schools — a town that grew markedly poorer since 2019 will be predicted to test worse than its students actually did, and appear to outperform. This is the largest limitation on the list. Closing it would mean fitting the line against the lunch share as it stood in 2018–19; AreaProfile collects only the current year of that measure, so the matched-period version is not something the site can compute today. Fully virtual schools are excluded from every city figure, because they enroll students from across a state while reporting a single administrative address. Private and home schooling are outside the federal collection entirely.