The best US city to live in is the one with nothing wrong with it

We built a score out of four federal datasets, published the weights before we looked, and ran all 95 of the biggest US metros through it. The winner is not the best at anything. It is the only place in the country that is never bad at anything.

Every “best places to live” list is somebody’s opinion with a number bolted onto it. That is not a complaint about the taste involved — it is a complaint about the arithmetic. The weights are almost never published, and a ranking whose weights you cannot see is a ranking you cannot disagree with.

So we did it the other way round. Pick the metrics first. Write the weights down. Apply them to every metro identically, from federal data, and let the answer be whatever it turns out to be.

It turned out to be Fayetteville–Springdale–Rogers, Arkansas, and the reason why is more interesting than the name.

The ruler

Four metrics. The weights were fixed before the data was pulled, which is the only thing that stops this from being a search for a headline.

Each metro is scored on its percentile rank within the 95, not its z-score. The raw units are not comparable to one another, and two of them are badly skewed — San Francisco’s price level and New York’s commute would drag everybody else’s z-score toward the middle and quietly compress the entire field. A percentile is immune to that.

It costs something, and the cost is worth stating: a percentile throws away the size of a gap. Finishing 1st on housing burden earns the same 25 points whether you beat second place by a hair or by ten points. The composite is then the weighted sum of the four percentiles, times 100.

The metros, and why they are metros

The 95 largest US metros by 2024 population — floor about 464,000 people, together 223.5 million, roughly two thirds of the country. The selection is by size alone. That matters more than it sounds: a ranking of places somebody already decided were nice is not a ranking, it is a re-ordering of a shortlist.

And metros, not cities. “Austin” here is the Austin–Round Rock–San Marcos CBSA, because a commute time and a price level are properties of a labour market, not of a city limit.

What is deliberately not in it

Violent crime. It belongs in this score. It is not here, and the reason is worth being specific about rather than hiding.

The FBI’s Crime Data Explorer API works without a paid key. But it exposes state and agency endpoints only — there is no metro endpoint. A metro crime rate means aggregating thousands of individual police departments through a county→CBSA crosswalk, and the demo key’s rate limit makes that impossible. It needs a free api.data.gov key and about a day.

Four measured metrics beat five with one invented.

So the ruler is four, and the video says four on screen. Nothing was substituted in the gap. The temptation to drop in a plausible-looking fifth column is exactly how a ranking of 95 real cities becomes fiction that passes every check you have, because a fabricated chart is indistinguishable from a real one.

Also absent, and on purpose: schools, weather, culture, healthcare. Each is either not measured at metro level at all, or measurable only as somebody’s index of somebody else’s index.

The field is tighter than anyone admits

Before the winner, the shape. Almost every metro piles into one narrow band, and in that band the score barely separates anything:

Ten places of rank costs about seven points of score. That is the real reason every published best-places list disagrees with the next one. In the middle of this distribution a trivial change of weights — move 5% from commute to cost — reshuffles twenty cities. The distinction between #12 and #19 is not a finding. Anyone presenting one is selling the precision of their own opinion.

Which is what makes the top of the table worth looking at, because one metro pulls clear of that band by a distance the band cannot explain.

The winner

96.4 out of 100. That is 4.6 points clear of Wichita and 2.25 standard deviations above the mean of the 95. Here is where the points came from:

Why it wins, which is not why you would guess

The obvious reading is that Fayetteville won on housing — it is first in the country, 23.1% of households paying 30%+ of their income to be housed, in a country where the median metro is around 30% and Miami is at 45%. That is a real result and it is not the mechanism.

The mechanism is the absence of a weakness.

Fayetteville never falls below 8th of 95 on any of the four metrics. No other metro in the set manages that. And a weighted sum of capped percentiles is asymmetric in a way that makes this decisive: you can only ever earn 30 points on cost of living no matter how cheap you get, but you can lose all 30. A hole costs more than a peak pays.

The whole top of the table has that shape. None of these places is spectacular at anything. They are the ones that never gave anything back. Austin is 7th in the country for job growth and pays for it at 65th on housing burden and 72nd on commute. San Jose is 14th on jobs and 91st on cost of living. One hole is enough.

The test: where people actually move

A score nobody can disagree with proves nothing, so it has to be checked against something measured that is deliberately not one of the four. We used Census net domestic migration for 2024, per 1,000 residents, over the same 95 metros — where Americans physically went.

Spearman ρ = 0.50. Genuine agreement, and a long way from the same thing. The disagreements are the finding:

Both ranks are computed over the same 95 metros. That sounds like pedantry and is not: ranking a 24-city subset against a 95-city full set and calling the difference a finding is the single easiest mistake in this genre, and we made it in the first version of this analysis.

This is not the ruler being wrong. It measures cost, housing, jobs and commute. It does not measure weather, or family, or the fact that a place is where you are already from. Migration is people optimising a different function — and if you want the honest version, a lot of the Florida rows are people optimising for January.

Three ways this produced correct-looking charts that were wrong

Every one of these is invisible in the output, which is the only reason they are worth writing down.

1 · The ranking that depended on file order

BEA publishes the price index to one decimal, so ties are common. Sorting to get a percentile handed two metros with byte-identical measured values scores up to 1.5 points apart — depending on nothing but where they sat in the file. A ranking that depends on file order is not a ranking.

Fixed with the midrank, so tied metros share the average of the ranks they span, and the displayed rank is inverted back out of the percentile rather than re-sorted separately — which means the rank shown can never disagree with the score that produced it. It moved the winner from 96.8 to 96.4 and its worst per-metric rank from 7th to 8th, both of which are quoted out loud in the video.

It was caught only by writing the scoring a second time, independently, and diffing all 95 scores. No chart, no gate and no amount of looking would have found it. The two implementations now agree exactly.

2 · A sign error that crowned Miami

One flipped direction ranked Miami first — on a cost index of 111.8 (89th of 95) and a housing burden of 45.2% (94th of 95). The chart looked completely normal. Bars in a sensible order, a plausible-sounding winner, nothing to see. Direction is now declared once, in one table, and used in exactly one place.

3 · Real numbers for the wrong real city

Indexing CBSAs in file order matched “Las Vegas” to Las Vegas, New Mexico, and “Cleveland” to Cleveland, Tennessee — micro areas that share a principal city name with a major metro and happen to appear earlier in the file. Nothing errors. You get real, correct, verifiable federal numbers for a city nobody was asking about. Fixed by matching off the official Census geography file and printing the entire mapping for a human to read.

What would change the answer

The weights are a judgement. They are published so you can disagree with them, and the file below has every raw value and every per-metric rank in it, so disagreeing is a spreadsheet exercise rather than an argument. If your weights put Fayetteville fourth, that is a real answer too — and you will be able to show your working, which is more than the lists can.