Every time a monthly report shows a jump in new business filings, headlines declare a startup boom. This guide explains what those filings actually count, how the U.S. Census Bureau tries to separate serious ventures from paperwork, and how to read the data sensibly.
What does a “business application” actually measure?
In the Census Bureau’s Business Formation Statistics, the core series is called business applications. According to the Bureau’s methodology page, it corresponds to “a subset of all applications for an EIN,” where EIN stands for Employer Identification Number, the tax identifier the Internal Revenue Service issues to businesses. The applications are filed on IRS Form SS-4, and the series excludes certain categories such as tax liens, estates, trusts and some industries.
That definition matters because an application is an intention, not an outcome. Someone can request an EIN to open a bank account for a side project, to set up a freelance practice, to sell goods online or to start a company that plans to hire fifty people. All of those appear in the same count. The number tells you how many people took a first administrative step, not how many companies came into being or how many will survive.
Why do filings rise and fall?
Application counts respond to many things besides entrepreneurial ambition. Economic conditions, job markets, tax rules, the ease of online registration and even outages can move the totals. The Bureau itself notes one example: an outage of the IRS EIN Assistant from December 30, 2025, to January 5, 2026, affected its December 2025 monthly data and the Week 53 2025 weekly data.
Methodology changes can shift series as well. As of the January 2026 monthly release, the Bureau says applications associated with internet sales are excluded from its high-propensity and corporation series, and the definitions were updated across the entire time series. Anyone comparing figures cited in older articles with current data should check whether the definitions have changed in between.
Seasonality is another factor. Filings tend to follow a calendar pattern, so the Bureau publishes seasonally adjusted figures that smooth out predictable swings. Mixing adjusted and unadjusted numbers in the same comparison is one of the most common mistakes in commentary on this topic.
Which series separate serious ventures from the rest?
The Bureau offers several related measures. They are easiest to compare side by side.
| Series | What it represents (per the Census Bureau) |
|---|---|
| Business Applications (BA) | The core count of EIN applications, with certain exclusions |
| High-Propensity Business Applications (HBA) | Applications “that have a high propensity of turning into businesses with payroll” |
| Projected Business Formations within 4 Quarters (PBF4Q) | A model-based estimate of employer businesses expected to originate from applications within four quarters |
| Projected Business Formations within 8 Quarters (PBF8Q) | The same type of projection over an eight-quarter window |
High-propensity applications are identified using traits on the Form SS-4 that are associated with higher formation rates, such as a corporate entity, planned hiring, or a planned first date for paying wages. In plain terms, an application from a corporation that says it expects to hire is more likely to become an employer than one from an individual with no hiring plans.
The projection series exist because actual outcomes take time to appear in data. They fill the gap between a filing today and what administrative records will show later.
How is a projection different from a count?
A count is an observed fact: this many applications were filed in a period. A projection is a modeled estimate: this many of those applications are expected to become employer businesses within four or eight quarters. The model draws on historical relationships between application characteristics and later outcomes.
That distinction affects how much weight to give each number. A count can be revised for technical reasons, as the Bureau did in its annual update applied in the January 2026 release on February 11, 2026, which revised business formation data and recomputed seasonal factors. A projection carries additional uncertainty because it assumes the future will resemble the past. If conditions change, for example if lending tightens or hiring slows, fewer applications may turn into employers than the model expects.
What happens to businesses after they open?
Formation is only the start of the story. Data from the U.S. Bureau of Labor Statistics on business establishments shows that survival rates have been fairly steady over time, even as the number of openings has grown. According to the BLS survival table, most annual cohorts retain roughly 75% to 81% of their establishments after one year. After five years, the share is lower: about 49% to 50% for mid-1990s cohorts, a low of roughly 45% to 46% for the 2006 cohort, and about 50% to 52% for cohorts opening between 2009 and 2019. The 2021 cohort stood at 47.2% after five years.
A few selected cohorts from that table illustrate the pattern:
| Opening year | Share surviving 1 year | Share surviving 5 years |
|---|---|---|
| 1994 | 79.6% | 49.6% |
| 2004 | 78.9% | 48.4% |
| 2014 | 79.7% | 50.8% |
| 2019 | 79.2% | 51.5% |
| 2021 | 79.1% | 47.2% |
Note that these figures cover private-sector establishments, which are individual business locations, and survival means still operating in a later March. They are not a direct measure of whether a founder’s particular idea worked out. For a deeper look at those figures, see how long new businesses last, according to BLS survival data.
A worked example: turning filings into expectations
The following is illustrative arithmetic, not real data. Suppose a region records 100,000 applications in a year, and 20,000 of them are flagged as high-propensity. Suppose also that analysts assume 40% of high-propensity filings and 5% of the remaining ones eventually become employer businesses.
- High-propensity: 20,000 × 0.40 = 8,000 employer businesses
- Others: 80,000 × 0.05 = 4,000 employer businesses
- Total: 8,000 + 4,000 = 12,000, or 12% of all applications
Now imagine the headline count rises 10% to 110,000, but the extra 10,000 filings are all low-propensity. The expected employers rise only by 10,000 × 0.05 = 500, to 12,500, an increase of about 4%. The headline grew more than twice as fast as the likely number of real employers. Those rates are invented purely to show the mechanism, but the logic holds: composition can matter more than volume.
Is a surge in filings good news for founders?
It depends on what you want to learn. A rise in applications can signal confidence, a loosening labor market that pushes people toward self-employment, or simply easier registration. More new entrants can also mean more competition for customers, talent and attention in a given niche.
For a founder deciding whether to enter a market, aggregate filings are a weak signal on their own. Better questions include whether customers already pay for something similar, how quickly competitors can be reached through ordinary channels, and whether the economics work at small scale. The resources on how to estimate startup costs and break-even and on what product-market fit really means address those more directly than any national statistic can.
How should you read the next monthly report?
A few habits make the monthly release more useful.
- Check the series. Is the quoted figure total applications, high-propensity applications or a projection? They answer different questions.
- Check seasonal adjustment. Compare adjusted figures with adjusted figures.
- Check definitions. Since the January 2026 release, internet-sales-related applications are excluded from the high-propensity and corporation series, so older comparisons may not line up.
- Look at trends, not a single month. Monthly figures can be affected by revisions and by events like system outages.
- Match the geography. National patterns may not describe your state, city or industry. The Bureau also publishes state-level and county-level application data, with its latest annual county update released June 10, 2026.
The Bureau has also scheduled its next annual update for the October 2026 monthly release on November 12, 2026, when revisions to formation data and seasonal factors may change earlier readings.
Reading startup statistics with the right expectations
Filing counts are a useful early indicator, but they measure intent, not companies. The more refined series, high-propensity applications and projected formations, get closer to the businesses most readers picture when they hear “startup,” and survival data from the BLS shows that opening is only the first hurdle. Treat each number as an answer to a narrow question, check how it is defined, and be wary of any headline that treats a record number of filings as proof of a record number of thriving companies. Founders weighing how to fund and structure their own ventures can continue with an overview of startup funding stages.
This article offers general information only. It is not financial, investment, tax or legal advice.
Sources: U.S. Census Bureau; U.S. Bureau of Labor Statistics
