The new-business boom, and its caveat

Problem

"Is the startup boom real, or a headline?" Public data can answer it — if you read it honestly. The Census Business Formation Statistics track new-business applications weekly back to 2006. The trap: an application is an EIN request, not a business that hired anyone.

What I did

Pulled twenty years of Census BFS weekly applications, aggregated to full calendar years, and normalized per week so a 53-week year doesn't look like growth. Compared each year to 2019, split total applications from the high-propensity ones (Census's proxy for likely employers), and added a state cut — all reproducible from a keyless public source.

What the data says

Applications ran +58.8% above 2019 per week in 2025, and every year since 2021 stayed at least +44.5% above 2019 — a durable shift, not a one-year spike (2025 was the peak). But the honest part is the mix: the high-propensity share — applications statistically likely to become employers — fell from 37.8% in 2019 to 30.3% in 2025 (and from 59.1% back in 2006). More people are applying to start something; a smaller slice look like future employers. The full table, the state cut, and the chart are public. (BFS counts applications, not businesses or jobs; "high-propensity" is a proxy, not a promise.)

What this costs you

This is the shape of an analysis engagement — a real question answered from your data, caveats included. Analysis runs as a $1,000/mo retainer.