The business boom is real — mostly

"Record numbers of new businesses" makes a great headline. The Census data behind it is real — and it says something the headline leaves out. Here's the honest read, from a keyless public source anyone can reproduce.

Yes, the surge is real — and durable

US business applications ran +58.8% above 2019 per week in 2025, and every year since 2021 held at least +44.5% above 2019. This isn't a single pandemic-year blip that faded; it's a level shift that has held for years (2025 was the peak).

The part the headline skips

Not every application is a business-in-waiting, and the Census flags the ones statistically likely to become employers ("high-propensity"). That share fell from 37.8% in 2019 to 30.3% in 2025 — and it's been sliding since 2006 (59.1%). High-propensity applications did rise (+27.2% vs 2019), just far less than the total. So: more formation, proportionally less of it employer-shaped.

Why the honest read matters

"Applications" is an intent signal, not an outcome — not a jobs number, not GDP. Aggregate to full years, normalize per week (a 53-week year isn't growth), and split the mix, and "booming" becomes "booming, with a caveat." That caveat is the useful part.

Key takeaways

• Applications +58.8% vs 2019 and durably so — real level shift, not a blip.
• Likely-employer share fell to 30.3% (from 37.8% in 2019) — count the mix, not just the total.
• Applications ≠ businesses or jobs — BFS is an intent signal; say what the data says.
• Reproducible from public data — keyless Census BFS, full method in the repo.

Keep reading: the full analysis.