How Census Counts New Businesses: EIN Applications vs Formations
Census’s monthly new-business figure counts tax-ID applications, not companies, and its “formations” count only firms that pay wages. What each series measures, with September 2026’s numbers.
Checked against Census release CB26-163 (September 2026 data, October 14, 2026) and the Bureau’s BFS methodology and definitions pages; the annual revision lands with the November 12, 2026 release
As of · September 2026, seasonally adjusted, US, Census Business Formation Statistics release CB26-1631
| Series | Counts | What it includes | September 2026 (count) | vs August 2026 | Source |
|---|---|---|---|---|---|
| Business Applications (BA) | Applications | All EIN applications on IRS Form SS-4, minus tax liens, estates, trusts, certain financial filings, applications outside the 50 states and DC, sectors 11 and 92 and some other industries | 534,947 | +0.5% | 1, 2, 3 |
| High-Propensity Business Applications (HBA) | Applications | BA that look likely to become employers: from a corporation, say they will hire, give a first wages-paid date, or are in listed industries | 147,331 | +1.3% | 1, 2, 3 |
| Applications with Planned Wages (WBA) | Applications | HBA that give a first wages-paid date on the SS-4 | 35,786 | +4.3% | 1, 2, 3 |
| Applications from Corporations (CBA) | Applications | HBA from a corporation or personal service corporation, per the legal form on the SS-4 | 40,951 | −0.2% | 1, 2, 3 |
| Projected Business Formations, 4 quarters (PBF4Q) | Projected formations | Employer businesses projected to appear within four quarters of the month of application | 29,022 | +1.4% | 1, 2, 3 |
| Projected Business Formations, 8 quarters (PBF8Q) | Projected formations | The same, within eight quarters | 39,703 | +1.0% | 1, 2, 3 |
Figures are Census’s seasonally adjusted national numbers as released October 14, 2026; the percentage is the release’s change from August 2026, which the release marks “statistical significance is not applicable or not measurable.”1 The series overlap: High-Propensity Applications are a subset of Business Applications, and the planned-wages and corporation series are subsets of High-Propensity, so the rows can’t be added together.2 The projected formations are Census’s model estimates (the two formation series are “spliced” from actual and projected values), not counts of companies already started.12 Not all series are shown: the release also breaks applications out by region and industry. This is a description of how the statistics are built, not a forecast.
How it works
The raw material is a tax-ID request
The Business Application series describe applications for an Employer Identification Number (EIN) made on IRS Form SS-4.1 An EIN is a unique 9-digit number the IRS assigns to entities for tax purposes, and Form SS-4 is the application.3 Census removes some applications before counting: tax liens, estates, trusts, certain financial filings, anything outside the 50 states and DC, sectors 11 and 92 (agriculture and public administration) and some other industries such as private households.3 What remains is the core series, 534,947 applications for September 2026.1
Why “applications” aren’t “businesses”
Census defines a business application as “an EIN application made predominantly for business purposes,” and an application can be filed long before, or without ever, becoming a firm with staff.3 That is why the Bureau publishes subsets. High-propensity applications are those whose SS-4 shows traits linked to a high rate of turning into an employer: a corporation, hiring plans, a first wages-paid date, or a listed industry such as accommodation and food services.3 Census says the traits are periodically re-evaluated against new formation data.3
A formation means payroll
The Business Formation series describe “employer business formations as indicated by the first instance of payroll tax liabilities” for the matching applications.1 Census’s own definition is “the formation of an employer (wage-paying) business from a business application.”3 By that definition, a business that never pays wages is not a formation, however many EIN applications it filed. That is why these series differ from the nonemployer counts in 32 Million: How Many US Businesses Have No Employees, which measure the firms with no payroll.
Four quarters, eight quarters, and a projection
Payroll data arrive quarterly, so Census looks ahead in quarters: how many applications from a given month show a first payroll within four quarters (12 months), or eight (24 months).2 Actual counts are only available for older months; for the recent ones Census projects the figure from an econometric model of how likely an application is to become an employer, and the headline formation series splices the actual and projected values together.2 In the October 14 release, actual formations within four quarters run through December 2022 and within eight through December 2021, with projections after.1
For September 2026, Census projects 29,022 formations within four quarters of that month’s applications, against 534,947 applications; our own division, 29,022 ÷ 534,947, is about 5.4%. Both are seasonally adjusted, and Census cautions that the formation figure “does not provide an estimate of the total number of business startups that appeared within a specific month.”1
Seasonal adjustment and revisions
BFS data come both not adjusted and seasonally adjusted using Census’s X-13 ARIMA-SEATS program, and the headline figures here are the adjusted ones.3 The Bureau also applies an annual update: NAICS industry codes for the previous five years are revised, the formation series are updated from the Business Dynamic Statistics, and seasonal factors are recomputed, which revises the adjusted series. The next one is scheduled for the October-data release on November 12, 2026, so these figures may change then.1
Which series to cite
The effort worth knowing about is which series to cite. Each is a different filter on the same IRS form, and Census publishes them from July 2004 onward at national, regional, state and county level, plus a limited-time weekly national and state view.2 A single sentence such as “new businesses rose” is therefore ambiguous until it names the series; the table above lists the exact name and definition of each.
Sources
- U.S. Census Bureau, Business Formation Statistics, September 2026 (release CB26-163, October 14, 2026), retrieved 2026-10-19; this address always shows the current release.
- U.S. Census Bureau, Business Formation Statistics: Methodology, retrieved 2026-10-19.
- U.S. Census Bureau, Business Formation Statistics: Definitions, retrieved 2026-10-19.