America files twice as many new businesses as it did in 2019. Most of them will never hire anyone.

What the data shows

The Census Bureau counts every application for an employer identification number, which is the closest thing the United States has to a real-time register of people trying to start something. In the twelve months to July 2026 there were 6,141,390 of them. In the same twelve months of 2019 there were 3,472,591. The increase is +77%, and it has not reversed in the five years since it began.

That is the number usually reported, and on its own it is close to meaningless, because an application is a filing rather than a business. The Bureau also publishes a subset it calls high-propensity: applications carrying the signals that historically precede actually employing someone, such as a corporate form, a stated payroll date, or an industry code where solo operation is rare. Nationally that subset is 28.2% of the total. Roughly seven in ten new filings are something else.

Two states are not doing what the other forty-nine are doing

Wyoming and Delaware sit far outside the distribution, and they are the two states whose corporate statutes are themselves an export product. Wyoming filed 145 applications per 1,000 residents and Delaware filed 58, against a national rate of 18 and a median state rate of 16. Neither state has a population that could plausibly generate those filings from inside its own borders: an entity registered in Wyoming is frequently operated from somewhere else, which is the service being bought.

This is worth stating precisely, because it is the single largest distortion in any state-level ranking of American business formation, and it is almost always reported as entrepreneurial dynamism. It is not evidence of fraud and it is not evidence of a booming local economy. It is evidence that a state has made itself the address of choice for entities formed elsewhere.

The relationship is real but moderate, and it survives removing Wyoming, the point that would otherwise be doing the work. States at the top of the growth table cluster around 26% high-propensity; New York, California and Delaware lead on that measure at 37.5%, 35.9% and 34.9%, and all three sit in the bottom half of the growth table.

5.9% of applications became an employer

The Census Bureau also follows applications forward, and reports how many became businesses with payroll within four quarters. That figure is a measurement rather than a forecast only where four quarters have actually elapsed, so the most recent cohort it can describe is the twelve months to December 2022. Of the 5,062,563 applications filed in that year, 296,531 had become employers a year later: 5.9%.

The spread across states runs from 3.0% in Wyoming to 11.3% in Maine, and it lines up with the high-propensity share rather than with the growth table. The states where filings grew fastest are not the states where filings turn into payroll.

Why this matters

A state that wants more employers is not measured by the first number. Application growth is easy to celebrate and easy to manufacture: it responds to filing fees, to registered- agent marketing, to gig-economy tax treatment and to whichever platform is currently telling people to form an LLC. The high-propensity subset moves much more slowly and is much harder to move on purpose.

For anyone selling to small businesses, the gap is a targeting problem with a number attached. Roughly 72% of new filings are unlikely ever to hire, which means a prospect list built from new business registrations is mostly a list of people who will never buy anything that is priced per employee. In Wyoming that share is 76%.

The full data

Rows marked with a dot carry a seven-year change that moves by more than 0.6 times its own size when the twelve-month windows are slid up to nine months in either direction. They are the smallest states by filing volume. The figure is shown rather than withheld, with its instability attached.