In July 2026, 28 of America’s 51 labour markets reported unemployment at or below the national 4.1%. Those 28 hold 36.6% of the country’s workers.The four largest markets are all above the line.

Most states beat America. Most Americans do not.#

For July 2026 the United States published an unemployment rate of 4.1%. Each of the fifty states and the District of Columbia published one too, on the same definition and the same seasonal adjustment. Twenty-seven printed a rate below the national figure, one printed exactly 4.1%, and twenty-three printed a rate above it. On a count of places, the country looks better than its own headline.

On a count of people it does not. The 28 areas at or below the national rate hold 36.6% of the American labour force. The 23 above it hold 63.4%. Nearly two workers in three live where unemployment is higher than the number they read in the news.

The mechanism is not subtle once the table is sorted. The four largest labour markets in the country are California at 5.1%, Texas at 4.5%, Florida at 4.6% and New York at 4.4%, and every one of them is above the national rate. Together they hold 33.3% of everyone in the American labour market. The largest area at or below the line is Pennsylvania at 3.9%, then Ohio at 3.4%, Georgia at 3.3%, North Carolina at 3.6% and Virginia at 3.7%. Good rates, in states that hold fewer people.

None of that makes the national figure wrong. It is a rate over the whole country, so it weights by workers, not by borders, and it is the right number for the question it answers. It is simply not the number describing the typical American place, and the two get quoted as if they were the same thing. The median area printed 3.9%.

The distance inside one country#

South Dakota reported 2.0% and the District of Columbia reported 5.9%. That is 3.9 percentage points, and the top is 2.95 times the bottom. One currency, one federal labour law, one month, one definition.

The bottom of the table is thinly populated and the top is not. South Dakota, North Dakota, Nebraska and Iowa together hold 2.2% of the country’s workers, and all four report between 2.0% and 3.2%. California alone holds 11.5% and reports 5.1%. A reader who has only seen the national rate has been told the average of these places weighted by how many people live in them, which is a true number and a poor description of any of them.

It is worth saying what the spread is not. It is not a ranking of which economies are strong, because a rate is a share of the people who are looking, and the next section shows two areas printing very different rates on very different participation. And nothing here tracks a state over time: this is one month, and a month is a photograph.

A rate counts the people who are looking#

Every area in the table publishes a second rate, and it is the one that says how much of the adult population is in the market at all. Nationally, 61.4% of Americans aged 16 and over were working or looking in July 2026. Across the fifty-one areas that figure runs from 53.8% in West Virginia to 70.2% in the District of Columbia, a spread of 16.4 points, which is more than four times the spread of the unemployment rate itself.

Put the two side by side and the rate stops being a league table. West Virginia printed 4.1%, exactly the national rate, on the lowest participation of the fifty-one. The District of Columbia printed 5.9%, the highest rate of the fifty-one, on the highest participation of the fifty-one. Read only the first column and West Virginia is average and the District of Columbia is the worst place in America to look for work. Read both and they are not comparable at all.

The arithmetic behind that is in the exhibit below, and it is worth stating plainly because it is where most readings go wrong: a person who gives up looking leaves the unemployment count and the labour force at the same moment. The rate falls. Nothing improved. That is not an accusation against any state in this table, and the piece names no cause for any of the participation figures, because age structure, retirement, disability, study and opportunity all sit inside that one number and this data cannot separate them.

Why the states do not add up, and why that matters#

There is an obvious next step with a table like this: add the states together and check the national figure. It does not work, and the size of the failure is worth printing. Summed across the fifty-one areas, the state series give 7,235,216 unemployed people. The national series for the same month gives 6,916,000. The difference is 319,216 people, and the rate implied by the state sum is 4.26% against a published 4.1%.

Neither figure is wrong. They are produced by different instruments. The national number comes from the Current Population Survey, a monthly household survey. The state numbers come from a modelled programme that uses the same survey, plus unemployment insurance claims and payroll data, and is benchmarked on its own schedule. The two are designed to be internally consistent, not mutually additive, and the gap between them is a known property rather than an error in either.

Which is exactly why this edition prints the difference rather than quietly using one number where the other belongs. Every comparison above is state against state, or a state against the published national rate, and never a state against a national figure rebuilt from states. If you take one thing from this piece other than the finding, take that: the moment a table of American states is summed to check Washington’s arithmetic, the answer is already wrong by about three hundred thousand people.