Roughly half of new private-sector business locations in the United States are still operating five years after opening, and about a third after ten, according to the Bureau of Labor Statistics. This article explains where those figures come from, how steady they have been across three decades, and why they say less about any single startup than the headlines suggest.
What the federal survival table actually measures
The Bureau of Labor Statistics (BLS) publishes a table of survival rates for private-sector establishments, grouped by the year they opened. An establishment is a single business location, such as a shop, an office or a plant, not a whole company. The table follows each opening-year group and records what percentage is still open one year later, two years later, and so on up to ten years.
The data come from the Business Employment Dynamics program, which BLS builds from the Quarterly Census of Employment and Wages. It covers private-sector establishments with employees. BLS’s latest Business Employment Dynamics release, dated October 28, 2026, covers the first quarter of 2026 and includes annual revisions. The survival table is current through March 2026.
One point matters for everything that follows: the table counts locations with employees. A company that closes one branch while its other branches keep trading shows up as a closure, and a business that shuts down and reopens under a new tax account can look like a closure followed by a birth. The BLS page does not discuss firms, so firm-level conclusions cannot be drawn from it alone.
The headline numbers: one year, five years, ten years
For establishments that opened in the year ended March 2016, BLS reports that 79.6% were still open after one year, 50.7% after five years and 34.2% after ten. The 1994 group looks strikingly similar: 79.6% after one year, 49.6% after five and 33.6% after ten.
| Opening year | After 1 year | After 5 years | After 10 years |
|---|---|---|---|
| 1994 | 79.6% | 49.6% | 33.6% |
| 2005 | 80.1% | 46.8% | 33.8% |
| 2016 | 79.6% | 50.7% | 34.2% |
| 2021 | 79.1% | 47.2% | not yet available |
| 2022 | 76.3% | not yet available | not yet available |
Source: BLS survival table for total private-sector establishments, as of March 2026.
Read plainly, the table says that about one in five new employer establishments does not reach its first anniversary, and that survival to year five hovers near the halfway mark. The oft-quoted claim that “90% of startups fail” is not what this federal series shows. Different datasets measure different things, and venture-backed technology startups are a small, unusual subset of all new businesses.
Does the risk fall as a business gets older?
Yes, and the table shows how. Compare the share that survives each stage with the share that had already survived the earlier one. Using the 2016 group as illustrative arithmetic:
- From year one to year two, 68.8% divided by 79.6% is about 86%, so roughly 14% of the establishments that made it through year one closed in year two.
- From year five to year ten, 34.2% divided by 50.7% is about 67%, so about a third of five-year survivors closed over the next five years.
- From year nine to year ten, 34.2% divided by 36.8% is about 93%, so only around 7% closed in that final year.
The pattern is consistent with what researchers often call the “liability of newness”: the youngest businesses close at the highest rates, and the yearly closure rate declines as the survivors accumulate customers, cash reserves and routines. The 1994 group tells the same story, with about 68% of five-year survivors reaching year ten (33.6% divided by 49.6%).
The pattern does not mean a ten-year-old business is safe. It means the proportion closing each year is lower than in the first two years.
Why have the rates barely moved in thirty years?
Looking down the table, the one-year and five-year columns move within a narrow band. One-year survival ranges from 75.2% (the 2008 group) to 80.9% (the 2020 group). Five-year survival ranges from 45.4% (the 2006 group) to 51.9% (the 2018 group).
Two features stand out. First, the groups that opened around the 2007–2009 financial crisis show some of the weakest results; the 2006 group’s five-year figure of 45.4% is the lowest in the table. Second, the 2021 and 2022 groups sit toward the weaker end so far: 47.2% five-year survival for 2021, and 76.3% one-year survival for 2022, the third-lowest one-year figure in the table after 2008’s 75.2% and 2001’s 75.7%.
The table does not explain why. Plausible factors include economic conditions, interest rates and the mix of industries entering, but BLS does not attribute causes in this table, and those explanations should be treated as hypotheses rather than findings. The 2020 group, which opened at the start of the pandemic, actually shows the highest one-year (80.9%) and two-year (72.3%) figures, a reminder that survival statistics can behave in counterintuitive ways when government support, moratoria and unusual demand patterns are in play.
How many new businesses start in the first place?
Survival rates describe the businesses that exist. A separate question is how many start. The Census Bureau’s Business Formation Statistics use applications for an Employer Identification Number (EIN), a tax identifier requested on IRS Form SS-4, as a signal of business initiation.
The Census Bureau is explicit that applications are not businesses. An application can come from someone who never starts trading, from a hobby, or from an entity that never hires. Census therefore also publishes high-propensity business applications, a subset with a high likelihood of becoming an employer business. An application qualifies if it comes from a corporate entity, indicates it will hire employees, gives a first wages-paid date, or falls in specified industries. Employer formations themselves are identified later, from the first payroll tax liability tied to an EIN.
Census also notes that the formation series are revised annually and that projections come from an econometric model with projection error. Its page lists the next annual update as part of the October 2026 monthly release, scheduled for November 12, 2026, so the most recent numbers are about to be recalculated. Starting with the January 2026 release, applications tied to internet sales are excluded from the high-propensity and corporation series, and the definitions were revised across the whole time series. Anyone comparing older published charts with current ones should check which definition each uses.
What the survival numbers cannot tell a founder
The federal table is a population statistic, and several limits stop it from working as a forecast.
- It mixes every industry. Restaurants, construction firms, medical practices and software companies are all in the “total private” group. Survival differs widely by sector, and the total hides that.
- Closure is not failure. A location can close because the owner retired, sold the business or merged it into another one. The table records that an establishment stopped appearing, not why.
- Employer businesses only. Many software companies begin with one or two founders and no payroll. Those never enter the table until they hire. See the discussion of how one person can run a software company for why that is increasingly common.
- Locations, not companies. A multi-site chain’s branch closing counts the same as a stand-alone shop closing.
For technology founders specifically, other measures matter more: how much cash is left and how long it will last, which is the subject of burn rate and runway; whether subscription customers keep paying, as covered in the SaaS business model explainer; and the financing path described in how startup funding works from pre-seed to Series A.
How job creation and destruction fit the picture
The same BLS program that produces the survival table also reports quarterly job flows. For the first quarter of 2026, BLS reported 7,381,000 gross job gains and 7,293,000 gross job losses in the total private sector. Gross job gains come from opening and expanding establishments, and gross job losses from contracting and closing ones.
Taking the difference gives a net gain of 88,000 jobs (7,381,000 minus 7,293,000), a small number set against turnover of more than 7 million jobs on each side. The point of the comparison is scale: in a typical quarter, millions of jobs appear and disappear as businesses open, grow, shrink and close. Failure and closure are a constant feature of the economy, not an anomaly, and the net change is a small difference between two very large flows.
Establishment counts tell a similar story. BLS reported 2,198,000 establishments gaining jobs and 2,243,000 losing jobs in that quarter, so slightly more locations were shrinking or closing than expanding.
What the long-run record suggests about expectations
The most reliable reading of the data is modest. Starting an employer business carries a meaningful chance of closure within five years, roughly a coin flip across the whole economy, and that chance has been stable for decades. The first two years are the most hazardous. Survival through year ten is a minority outcome, around a third, though again closure includes orderly exits as well as distress.
For readers evaluating a company, the lesson is to be wary of both extremes. Claims that nearly all startups fail overstate what federal data shows, and claims that careful planning removes the risk overstate what any dataset can promise. The table describes the economy’s average, and individual outcomes depend on the business, the market and the capital behind it.
This article is general information, not financial, investment, tax or legal advice.
Sources: Bureau of Labor Statistics; Census Bureau
