Skip to content
INDEPENDENT FINANCIAL EDUCATION Clear explanations. Better questions.
EasyLoanWorld Your guide to financial clarity
Explore guides
Menu
FINANCIAL GUIDE Research & Data

Business Applications Are Not Business Openings: Read the Census Series

Separate business applications, nested application groups and projected employer formations before using a headline to assess competition.

A red stapler with a silver mechanism sits against a red background.
Illustrative office photograph. Photo by Mike Meyers on Unsplash.

Estimated reading time: 4 minutes

A rise in business applications isn’t a count of stores that opened their doors. Identify the series before using a headline to estimate new competitors. The Census Bureau’s Business Formation Statistics overview distinguishes applications for employer identification numbers from actual and projected employer-business formations.

Those are different stages of activity. A state-level application figure alone doesn’t establish how many operating competitors entered your town, industry or customer market. Treat it as background evidence, not a substitute for researching those businesses directly.

Read the series name before adding counts

The Census definitions distinguish these application measures:

  • BA: the core business-application series, after specified filtering of EIN applications.
  • HBA: high-propensity applications, a subset associated with a greater likelihood of becoming businesses with payroll.
  • WBA: applications with planned wages, a subset of HBA indicating a first wages-paid date.

These aren’t three separate groups to total. A planned wage date is an application characteristic, not confirmation that wages have been paid. Census defines business formation here as the transition from an application to an employer business.

Worked example: one thousand applications, not fifteen hundred

Use an invented, complete collection of 1,000 qualifying applications for one place and one application month. Assume exactly 400 are high-propensity, including 100 with planned wages. These are raw classroom counts, not Census observations or seasonally adjusted estimates.

Three nonoverlapping parts of the same fictional application collection
Part of collection Count
Planned wages, within high-propensity 100
Other high-propensity applications 300
Outside the high-propensity subset 600
All qualifying applications 1,000

The parts reconcile: 100 + 300 + 600 = 1,000. High-propensity applications are 100 + 300 = 400. Adding the displayed headline categories instead—1,000 + 400 + 100—produces 1,500 category inclusions, not 1,500 distinct applications. The extra 500 is repeated counting.

The 400 high-propensity applications are 40% of the 1,000 applications. That percentage describes the composition of this collection. It isn’t a measured 40% success rate, and it supplies no count of businesses that actually began paying employees.

A projection isn’t another batch of openings

Census’s methodology links employer formations to the first payroll-tax liability. BF4Q tracks formations originating from an application month’s applications within four quarters; PBF4Q projects that outcome when actual results aren’t yet available. The projection uses a statistical model, not a list of confirmed openings.

SBF4Q combines actual and projected segments into one series. Check which segment you’re quoting. A four-quarter application-cohort measure shouldn’t be relabeled as businesses that opened during the application month.

Our raw-count example isn’t a test that every published adjusted table must pass: Census notes that seasonal raking and rounding can produce industry-level HBA above BA. Consult the current methodology before treating an apparent inconsistency as an error.

Keep the projected outcome attached to its starting group

Now stipulate that an invented projection for the same 1,000 applications is 120 employer formations within four quarters. This is a supplied exercise input, not an estimate produced by our arithmetic or Census’s actual model.

Dividing 120 by 1,000 gives 12%: the projected count relative to its application base. Adding 120 to 1,000 would instead mix an anticipated outcome with the applications from which it originates. It wouldn’t establish 1,120 new businesses.

Nor can this exercise assign all 120 projected formations to the 400 high-propensity applications. It supplies no outcome breakdown by subgroup. Reporting 120 ÷ 400 = 30% as their projected formation rate would introduce an unsupported allocation.

A later observation of 110 formations for that same collection and horizon would be ten below the supplied projection. That comparison would describe forecast error in this invented case—not ten business closures. No closures or operating-business stock were supplied.

Use the statistic for the question it can answer

Save the series code, application period, geography, release version and adjustment label. Separately identify whether the figure is an application count, an observed employer formation or a projection. For a local expansion decision, add evidence about actual competing locations and customer demand; the series alone doesn’t establish either.

For other data-reading distinctions, see payroll jobs versus employed people and Small Business Credit Survey percentages.

General U.S. research education. All numerical examples are invented, not current market estimates or business forecasts. Census sources checked October 28, 2026.