The Business Formation Statistics released by the U.S. Census Bureau on September 11 showed that in August, seasonally adjusted business applications totaled 531,728, down 7.8% from the revised 576,512 in July. The size of this drop is striking, but the month-on-month pullback occurred after a 8.1% month-on-month gain in July, making it look more like volatility at high levels rather than a signal that entrepreneurial activity has entered a downturn on its own.
Business application statistics come from employer identification number (EIN) tax forms. They are fast and wide-ranging, so they can be used to observe potential entrepreneurial activity earlier, which is why they are often treated as a leading indicator of economic vitality. However, there is still a long way between “applying for a number” and “starting a business with ongoing employment of staff.” Some people cancel plans, some entities are used only for asset holding or sole proprietorships, and some applications take months before they generate payroll tax records.
530,000 applications are expressions of intent, not 530,000 new employers that have already opened
The Census Bureau’s Business Formation Statistics include four types of application series and eight types of formation series. Total applications exclude tax liens, estates, trusts, some financial reporting, agriculture, and public administration, among other categories. High-propensity applications further filter by industry, legal form, and whether the applicants plan to pay wages, selecting the subset more likely to become employer firms. Different definitions answer different questions; you can’t simply take the largest total applications figure and treat it as the number of new companies.
Seasonal adjustment is also crucial. Filing schedules for annual tax returns, changes in academic terms, holidays, and policy timing points all affect EIN applications. The statistical agency uses historical patterns to remove predictable seasonal fluctuations. The August month-over-month decline of 7.8% refers to the adjusted change, not a synchronized drop in every state and every industry. To assess a trend, at least three-month averages, high-propensity applications, applications where the applicant plans to pay wages, and the forward-looking expected business formations over the next four or eight quarters should be observed at the same time.
The July data were revised to 576,512 applications, up 8.1% month over month. Looking at the two months together, part of August’s decline offset the jump from the prior month. Monthly series are inherently sensitive to concentrated filing and calendar effects. If the media only highlights one negative number, it’s easy to write “a cool-down in momentum” as “a collapse in startups.” A more accurate framing would be: application flow pulled back from the July peak, but whether it will continue trending downward depends on additional months of data.
The statistical definitions also change in 2026. Starting in January, the Census Bureau reissues time series using the 2022 industry classification and excludes some high-propensity and company application categories related to internet sales. Annual updates will also revise industry codes and actual business formation using more complete administrative records. When comparing over longer periods, you must use the same version of the series; you can’t misinterpret changes in definitions as true economic turning points.
To judge startup quality, continue to examine hiring, financing, and survival rates
Application volume reflects willingness to enter. What truly affects the economy is whether these intentions turn into actual operations, investment, and jobs. When interest rates are high, the cost of registering a business doesn’t necessarily rise sharply, but rent, inventory, equipment, and working capital get more expensive, making applicants more cautious about starting real operations. Digital platforms also reduce experimentation costs, allowing many people to apply first and validate the market later—widening the gap between application counts and the number of traditional brick-and-mortar firms.
Industry structure also affects what the numbers mean. Capital needs, hiring speed, and failure rates vary greatly across professional services, construction, food service, transportation, and e-commerce. With the same total applications, if more come from high-propensity applicants who plan to pay wages, the signal for future employment is usually stronger. If they’re concentrated in low-capital activities with no employees, the pattern is closer to changes in self-employment and project-based income.
Regional data also shouldn’t be masked by the national average. In states with faster population growth, active housing construction, or concentrated industry investment, application volume may remain strong. Regions dependent on a single industry may be more sensitive instead. When comparing changes across states, you should also use population or working-age population as the denominator; otherwise, larger states naturally have more applications. Disasters, tax filing deadlines, and state-level registration policies can also create short-term peaks and troughs.
From the business-owner perspective, a decline in applications for “reduced” reasons could mean expected returns are falling, or it could simply be that earlier demand was pulled forward. After a sharp rise in July, some applications that would otherwise have been filed in August may already have been completed. Without data on the distribution of application dates and the eventual formation of results, it’s impossible to separate these two explanations. Therefore, a headline may describe the pullback, but it shouldn’t directly attach a confident label about entrepreneurs’ sentiment or the business cycle.
Investors should also combine this set of data with small business confidence, bank lending standards, job vacancies, and local consumer activity. If applications rise but lending tightens, it may mean more creativity or ideas but less access to financing. If applications fall while sales improve among existing firms, it doesn’t necessarily imply the economy is weakening. The advantage of BFS is that it’s early; the cost is that it’s far from the final outcome. It’s suitable for providing directional clues, but not for drawing independent growth conclusions.
The Census Bureau clearly defines “expected business formation” as a forward-looking estimate: it projects how many applications from a given cohort will generate payroll tax obligations over the next four quarters, not the total number of companies that actually open in a specific month. News writing should preserve this caveat. The 531,728 applications in August and the 7.8% month-over-month decline are real statistics, but they only indicate changes in entry flow. Whether the startup surge persists is ultimately answered jointly by high-propensity applications, the emergence of real employer firms, job creation, and business survival.