Employee Tenure Statistics 2026: 4.1 Years, the First Rise in 14 Years — and Why It Is Not Loyalty

Last updated: October 2026. BLS published Employee Tenure — 2026 on 24 September 2026, and it reversed the trend this page was originally built on. Every figure below is now the January 2026 reference period. This refresh rebuilt the headline, age, distribution, industry, occupation, sector, education and gender tables against the new release, recomputed all five derived metrics, added a section on why the median rose, and corrected one framing — the widening long-service gender gap — that the new data overturned. The correction is set out in full in section 9.

Employee tenure is the quietest number in HR and the one that most programmes are silently built on. Service awards assume it. Vesting schedules assume it. Onboarding payback models assume it. Every "cost of turnover" calculation has a tenure assumption buried inside it. For a decade those assumptions have been drifting out of date in one direction. As of January 2026 they are drifting in the other, and for a reason that should not reassure anybody.

This page assembles the verified tenure dataset: the headline median and its fourteen-year trend, the mechanism behind the 2026 increase, the full distribution, the age, industry, occupation, sector and education cuts, the gender gap that has just reversed, and what longitudinal data says about how long jobs actually last — which is a different question from how long people have currently been in them. Every figure carries its source and reference period. Where we derive something, the arithmetic is shown in the body rather than presented as a survey finding.

1. The headline: 4.1 years, and the end of a fourteen-year slide

The median number of years that US wage and salary workers had been with their current employer was 4.1 years in January 2026, up from 3.9 years in January 2024. For men the median rose from 4.2 to 4.3 years; for women it rose from 3.6 to 4.0 years, the largest two-year move in the table.

The survey is a supplement to the Current Population Survey, run biennially in January since 1996 and sponsored by the US Department of Labor's Chief Evaluation Office. That matters for two reasons: the sample is the CPS household sample of about 60,000 eligible households rather than a vendor panel, and the data is refreshed only every two years — which is why a great deal of content published between October 2024 and September 2026 reported January 2024 figures as current.

Median years of tenure with current employer. Source: BLS Employee Tenure, Table 1 (January 2026 release) and Table 1 of the September 2022 release for the 2012 and 2014 columns. Gap column derived.

Three things stand out. The first is that this is the first increase in the series since January 2012, when the median rose from 4.4 years in January 2010 to 4.6 — a fact taken from the BLS release of 18 September 2012. Every reading since had been flat or lower.

The second is that the recovery is overwhelmingly a women's recovery. Women's median tenure rose 0.4 years against 0.1 for men, which closed the gender gap in median tenure from 0.6 years to 0.3 — back to where it sat in 2016. Section 9 takes that apart, because it reverses what this page previously reported.

The third is that 4.1 years is still below every reading from 2012 to 2022. The 2026 increase recovers the 2024 drop and no more. Against January 2012's 4.6 years, the median is still down 0.5 years, or 10.9%.

2. Why the median rose — and why it is not a retention story

A median can rise for two completely different reasons. People can stay longer, which moves the distribution to the right at the top. Or the short-tenure population can shrink, which moves the middle person further up a distribution that is otherwise unchanged. These have opposite implications for anyone running a retention or recognition programme, and the BLS distribution tells you clearly which one happened.

Distribution of employed wage and salary workers aged 16 and over by tenure with current employer. Source: BLS Employee Tenure, Table 3, January 2024 and January 2026 releases. Change column derived.

The second half of the mechanism is that the short-tenure bucket is not being refilled. A worker enters the "12 months or less" bucket by being hired, and the JOLTS hires rate was 3.3% in August 2026 — 5.2 million hires in the month — with quits unchanged at 1.9%. Fewer hires means a smaller first-year population, which lifts the median mechanically. At 20.6%, the share of workers inside their first twelve months is below the 21% BLS recorded in January 2012.

Median tenure rose because the workforce stopped churning, not because it started staying. The share of workers with ten or more years of service moved by a tenth of a point.

3. Tenure by age: a threefold spread, and the end of the Gen Z story

Age remains by far the strongest predictor of tenure in this dataset — stronger than industry, occupation, sector or education. The median 20-to-24-year-old has been in post 1.5 years. The median 55-to-64-year-old has been in post 9.6 years.

Median years of tenure by age. Source: BLS Employee Tenure, Table 1.

The biggest gain in the table, by some distance, is 25-to-34-year-olds, up 0.3 years to 3.0 — an 11.1% increase, and the highest reading for that age band since January 2014. This is the cohort that did most of the job-switching in 2021 and 2022, and it is the cohort whose tenure has recovered most. Meanwhile every band from 45 upwards is flat or within a tenth of a year of where it was.

At 1.5 years, 20-to-24-year-olds post the highest reading anywhere in the published 2012–2026 series for that age band, which has otherwise sat at 1.2 or 1.3 years. Young workers have always had short tenure, because tenure is a function of age: you cannot have ten years of service at 23. But the direction of travel among young workers is up, and has been for two consecutive releases.

Tenure among 25-to-34-year-olds is at its highest since 2014, and 20-to-24-year-olds are at a series high. The "Gen Z won't stay" story is the one part of this dataset the data has never supported.

The age gradient also explains most of what looks like industry or occupation variation later on. A sector that hires young has short tenure almost by construction. Keep that in mind when you reach the industry table — and when someone benchmarks your company against a national median without adjusting for the age of your workforce.

4. The distribution: where your workforce actually sits

A median tells you where the middle person is and nothing about the shape around them. The distribution is the more useful artefact, and it is the one almost nobody quotes.

Cumulate those and you get the numbers that actually drive programme design. They are arithmetic on published BLS shares, not a survey finding, so we label them as derived.

That is the structural case for celebrating early. It is also why the one-year and three-year marks do more work than the twenty-year one: the twenty-year award reaches 10.2% of the workforce, the five-year award reaches 47.2%, and a first-anniversary award reaches nearly everyone who is still there. We wrote about designing that ladder in milestone rewards for work anniversaries and birthdays.

5. The number that breaks every international comparison

You have almost certainly read a version of this claim: American workers stay with an employer for about four years, while the OECD average is around ten and Greece, Italy and Slovenia are over twelve. It appears in think-tank commentary, consultancy decks and a great deal of HR content.

It is a category error. BLS publishes a median. The OECD job-tenure series is an average. Job tenure is a strongly right-skewed distribution — a long tail of thirty-year employees pulls the mean well above the middle — so comparing one country's median to another's mean overstates the gap by a large and entirely artificial margin.

6. Tenure by industry, and the milestone multiple

Median tenure ranges from 2.4 years in leisure and hospitality to 7.0 years in federal government — a factor of nearly three. The final column below is ours: how many times the sector's own median an employee must stay to reach a conventional five-year first milestone.

Median years of tenure by industry. Source: BLS Employee Tenure, Table 5. Final column derived: 5 divided by the January 2026 median.

The practical fix is not lowering the bar everywhere; it is setting the first rung relative to the population it applies to. We looked at the operational side of this in frontline employee recognition, where the distribution problem is at its sharpest. If the reward itself has to work across a dozen countries and a dozen pay bands, our reward catalog is built for exactly that spread.

Two further notes on this table. Utilities rose from 4.9 to 6.3 years and wholesale trade from 4.6 to 5.4, the two largest increases; both are small CPS subsamples and both sit squarely inside the reclassification caveat above. Local government fell a full year, from 6.4 to 5.4 — the largest decline in the table, and one that reappears in the occupation data as a 0.4-year fall for education, training and library occupations. Those are substantially the same people: public-school staff are counted as local government.

7. Tenure by occupation

Against January 2024, the largest gains are arts, design, entertainment, sports and media (3.5 to 4.1 years), life, physical and social science (3.7 to 4.2) and installation, maintenance and repair (3.9 to 4.4). Management occupations rose from 5.7 to 6.1 and now top the table outright.

The declines are concentrated in public-facing public-sector work: farming, fishing and forestry fell from 5.1 to 4.3 years, and both education, training and library occupations and protective service occupations fell 0.4 years, to 4.9 and 4.8 respectively. Education, training and library is the standout, because it had been one of the most stable occupation groups in the series — 5.3 years in both 2022 and 2024 — and because it moves in step with the one-year fall in local-government tenure noted above.

8. Sector and education: the public-sector premium is collapsing

Public sector median tenure was 5.6 years in January 2026 against 3.9 years in the private sector. Public employees still stay longer — but by 44%, not the 77% recorded in 2024. This is the single largest structural change in the 2026 release and almost nobody has written about it.

Inside the public sector the movement is not uniform, and the federal line is the odd one out. State government fell from 5.7 to 5.4 years and local government from 6.4 to 5.4, but federal government rose from 6.5 to 7.0 — the only level of government to gain, and a reversal after four consecutive declines from 8.8 years in 2016.

Education moves the needle less than either age or sector, but slightly more than this page previously allowed. Across attainment groups for workers aged 25 and over, median tenure runs from 4.5 years (less than a high school diploma) to 5.6 years (doctoral or professional degree) — a spread of 1.1 years against an overall 4.9. Bachelor's-only sits at 4.9 and a master's at 5.1, so the familiar claim that a degree buys tenure is carried almost entirely by the doctoral and professional band, where men reach 6.7 years.

Age moves median tenure by a factor of more than three. Sector moves it by 44%. Education moves it by about thirteen months from bottom to top. Benchmark on age and sector, or do not benchmark at all.

9. The long-service gender gap has narrowed — a correction

BLS Table 2 tracks the share of workers aged 25 and over who have ten or more years with their current employer. It is the cleanest available proxy for a long-service population. When this page was published in September 2026, that gap had widened at every reading for a decade, and we said so. The January 2026 data reverses it.

Percent of employed wage and salary workers aged 25+ with ten or more years of tenure. Source: BLS Employee Tenure, Table 2. Gap column derived.

The underlying gap has not closed, and the overall long-service share is still deteriorating: 30.1% of workers aged 25 and over have ten or more years, the lowest reading in the published series and down from 33.2% in 2016. The composition of that decline simply shifted from women to men in this release.

10. How long jobs actually last (a different question)

Here is the methodological trap at the centre of this topic. Median tenure is a point-in-time measure: it asks people currently in work how long they have been there. Most of those jobs have not finished. It is not, and cannot be, a measure of how long a job lasts.

For that you need longitudinal data, and BLS has it. The National Longitudinal Survey of Youth 1979 followed 9,964 people born between 1957 and 1964, first interviewed in 1979 when they were 14 to 22. The most recent results — released 26 August 2025 and reissued 21 July 2026 — cover jobs held through age 58.

Source: BLS, National Longitudinal Survey of Youth 1979, cohort born 1957–64, n = 9,964. Dashes indicate BLS does not publish the duration split for that age band. This series was not updated by the September 2026 tenure release.

The two duration rows are the most useful numbers on this page for anyone designing a milestone programme. Of jobs started between 18 and 24, 87% ended within five years — so roughly 13 in 100 early-career jobs ever reach a five-year award at all. Of jobs started between 45 and 54, 56% ended within five years, so 44 in 100 do. A five-year milestone is not one programme applied to one workforce; it is a lottery with a 13% hit rate for your graduates and a 44% hit rate for your senior hires.

It also reframes the first year. If 61% of early-career jobs end inside twelve months, the twelve-month anniversary is not a minor rung on the ladder — it is the rung most of the population will never reach. The onboarding evidence points the same way; see our employee onboarding statistics for the first-year attrition and hiring-cost benchmarks that sit behind it.

11. Tenure rose while hiring and quitting stayed at decade lows

These facts are usually treated as contradictory. They are the same fact.

• Median tenure rose to 4.1 years in January 2026, the first increase since January 2012.
• The JOLTS quits rate was 1.9% in August 2026 — 3.1 million quits, unchanged on the month — against a series high of 3.0% in November 2021, when 4.5 million people quit in a single month. The JOLTS series begins in December 2000.
• Hires were 5.2 million, a rate of 3.3%, and job openings 7.1 million at 4.3% (JOLTS, August 2026, published 29 September 2026).

A median is a stock and a quits rate is a flow. Tenure in January 2026 reflects the accumulated hiring of the preceding decade, including the enormous churn of 2021 and 2022: that cohort has now aged into the three-to-four-year bucket, which is exactly where the distribution gained the most ground. Meanwhile a low hires rate means fewer people entering the first-year bucket to replace them. Both effects push the median up, and neither is a change in how long people choose to stay.

Two consequences follow. First, do not use tenure as a real-time retention indicator — by the time it moves, the decision that moved it is two to four years old; use quits, regretted attrition and first-year survival for that, as set out in our employee turnover statistics. Second, a rising median in a frozen labour market is not good news. It means your people have fewer places to go, and the share of them with genuine long service has not moved. When hiring resumes, the mechanism runs in reverse. The engagement data says the same thing from the other side: only 46% of US employees now think it is a good time to find a job, down ten points, while engagement sits flat at 31% — disengaged and staying, as our employee engagement statistics for 2026 set out.

12. What this means for service awards

The best available employer-side benchmark on service awards is also, awkwardly, a survey from 2003. The WorldatWork and National Association for Employee Recognition 2003 Recognition Survey (September 2003, 413 responses from a 13% response rate) found 87% of responding organisations used length-of-service awards, more than any other formal programme type, and that 87% of those programmes had already been in place for more than five years. It also found that companies gave a length-of-service award to an average of 28% of their employee population annually.

That last figure is worth holding against the BLS distribution, because the two cannot both describe a five-year ladder.

13. Benchmark cheat sheet

14. Seven tenure statistics to handle with care

Tenure has a worse citation-hygiene problem than most HR topics, because the underlying data is released only every two years and the gap fills with approximations. These are the seven we see most often — and the first one is brand new, created by the September 2026 release itself.

1. "Employee tenure is rising, so retention is improving." The median rose from 3.9 to 4.1 years, but the share of workers with ten or more years of service moved by 0.1 points and the share aged 25+ with ten or more years actually fell, to the lowest reading in the published series. The gain came entirely from the under-two-year population shrinking by 3.4 points in a labour market with a 3.3% hires rate. It is a composition effect, not a loyalty effect. See section 2.

2. "Americans stay 3.9 years." That figure is now two releases out of date; the current median is 4.1 years, referencing January 2026. Anything published after 24 September 2026 quoting 3.9 as current is quoting the January 2024 release.

3. "Americans stay four years, Europeans stay 10+." A median against a mean. The BLS figure is a median; OECD job-tenure figures are averages. Our derived US mean is around 7.6 years. The gap is real but roughly a quarter of the size the comparison implies. A widely shared May 2025 ITIF commentary sets a BLS median directly alongside OECD and JILPT averages without distinguishing them, and much of the HR content repeating those numbers inherits the error.

4. "The average job lasts four years." It does not. Median tenure measures how long current employees have so far been in post — most of those spells are unfinished. For completed job lengths use the NLSY79 figures in section 10: 61% of jobs started between 18 and 24 ended within a year.

5. "The average person will hold 12 jobs in their lifetime." The 12.9 figure describes people born between 1957 and 1964, counting jobs from age 18 to 58. BLS is explicit that it cannot be used to forecast the job counts of younger cohorts, whose careers are not finished. Quote it with the cohort and the age range or not at all.

6. "Gen Z job-hops more than any previous generation." Not in this data, and less so than ever. Tenure for 25-to-34-year-olds rose from 2.7 to 3.0 years, the highest since 2014, and 20-to-24-year-olds reached 1.5 years, a high for the published series. Young workers have always had short tenure — boomers in the NLSY79 cohort held 5.6 jobs between 18 and 24 — because tenure is a function of age, not of generation.

7. "87% of companies have length-of-service awards." That figure comes from the WorldatWork and NAER 2003 Recognition Survey: September 2003, 413 responses, a 13% response rate from an association membership skewed to large North American employers. It is 23 years old and still circulating as current. WorldatWork's most recent public Trends in Employee Recognition summary dates from February 2019 (n = 445) and puts only the overall "have a recognition programme" figure at 87%, with the programme-type detail behind a members-only wall. No current free primary source publishes service-award prevalence.

Methodology and sources

Every figure on this page is attributed to a named source with its reference period and, where the publisher disclosed it, its sample size. Where a statistic is derived, the inputs and the arithmetic are shown in the body rather than presented as a survey finding. Percentages are reproduced as published and not re-based. Where a derivation rests on an assumption — the open-ended top bucket in section 5, the uniform-within-bucket spread in section 12 — the assumption is stated and its sensitivity given. Where a previous edition of this page reached a conclusion the new data overturned, the change is recorded in section 9 rather than quietly edited out.

Citation. If you use these figures, please cite the underlying source rather than this page, and link here as the compilation. The six derived metrics — the three-band decomposition showing the 2026 rise happened entirely below ten years, the service-award reachability ladder, the approximately 7.6-year US mean tenure, the milestone multiple by industry and occupation, the 57.5% collapse in the public-sector tenure premium since 2016, and the 8.4% annual milestone rate under a 5/10/15/20 ladder — are original to this page and should be attributed to Rewordin together with their stated inputs and assumptions.