“About one in three new graduates leaves within three years. Young people these days…”
That sounds as if a uniquely modern generation suddenly discovered the exit door. Japan’s long-run data tell a much less dramatic story.
For university graduates from March 2021, 34.9% had left their first employer within three years. That really was the highest figure in the latest 15-year window. Move the graph left, though, and the 2000 cohort is at 36.5% and 2004 at 36.6%.
34.9%. Wow, high.
→ Highest in 15 years.
→ Scroll left.
→ “…Wait, it was higher 20 years ago.”
The “one in three” figure mainly refers to newly employed university graduates leaving the employer they joined after graduation. It does not mean one-third of all young people permanently stopped working; many simply changed jobs.
The long-run conclusion is closer to “young workers have been leaving early for a long time” than “today’s youth suddenly lost the ability to endure work.”
What happens when 34.9% is viewed over 30-plus years instead of 15?
The Ministry of Health, Labour and Welfare series reaches back to the 1987 graduating cohort. Selected university rates are 28.4% in 1987, 23.7% in 1992, 32.0% in 1995, 34.3% in 1999, 36.5% in 2000, 36.6% in 2004, 35.9% in 2005, 28.8% in 2009, 31.9% in 2013, 32.8% in 2017, 31.5% in 2019, 34.9% in 2021, and 33.8% in 2022.[1][2][3]
The rate moves with economic conditions and cohorts, but it is not a line from “almost nobody quit” to “suddenly one-third quit today.” It has spent decades around roughly three in ten. Teikoku Databank made the same broad point in a 2022 analysis.[5]
So “highest in 15 years” is true. “A uniquely modern youth crisis” is a different claim, and the 30-plus-year graph does not establish it.
High-school graduates make the old-days myth even harder to sustain
For newly employed high-school graduates, the same series shows 46.2% in 1987, 48.1% in 1996, 50.3% in 2000, and 49.5% in 2004. The 2022 cohort was 37.9%.[1][2]
You start with “one in three quits today” and run into periods where the historical data basically answer: “Try one in two.”
University and high-school graduates are different populations, so the rates should not be mixed as if they were directly interchangeable. Industries, company size, education rates, hiring systems, and the economy differ. But the figures are enough to break the caricature that young people in earlier decades almost never left their first employer.
Japan’s 1984 Labour White Paper already stated that roughly 40% of newly employed high-school graduates left within three years. It even noted that youth turnover can be part of finding work better suited to one’s aptitude and is not automatically negative.[6]
The 1980s were already having this argument. Generation Z did not invent it.
Go back to people who are almost 80 today, and early movement still appears
The most comparable modern series begins in 1987, so it would be wrong to draw one perfectly uniform line from today’s twenty-somethings to today’s late-70s and 80s.
But older evidence exists. In a 1970 Diet hearing, a Labour Ministry official reported that among March 1966 high-school graduates, 25.7% had left or changed jobs after one year and 53% had done so after three years.[7] Someone around 18 in 1966 would be about 78 in 2026.
A cohort now close to 80 had roughly half leaving or changing workplaces within three years.
That 53% is not methodologically identical to the modern employment-insurance separation rate, so it should not be subtracted directly from today’s 33.8%. It is historical context, not a decimal-perfect comparison. But it makes the claim “young people barely moved jobs back then” very hard to defend.
Why does someone still remember “our generation didn’t quit”?
As a joke, call this “old-guy☆bias.” It is not an academic term and it is not a claim about all older people. The underlying sampling problem is real: if you interview someone who has worked at the same company for 30 years, the peers who left decades ago are no longer standing in that office.
One relevant idea is survivorship bias: judging the whole group from the cases that remain visible.
Imagine 100 young hires. About 30 leave within three years. More disperse later through job changes, self-employment, retirement, or career changes. Decades later, a camera visits the original company and interviews a worker in their sixties with 30-plus years of service.
Of course you found a long-tenured worker. You searched among the people who were still there.
The peers who left after three years may be at another employer or in another industry. They are structurally absent from the sample “30-year veterans of this company.”
So “I stayed even when it was hard” may be completely true as personal history. The error comes when it expands into “young people in my generation generally stayed.”
Collect only the people who stayed, and you discover they stayed. Astonishing.
Long-run statistics put the missing people back into the denominator.
“Young people these days” may have a longer tenure than the workers it criticizes
A 2019 paper ran five preregistered studies with 3,458 participants. Adults tended to judge today’s youth more negatively than youth of the past, especially on traits in which the adults viewed themselves as strong.[8]
People who valued respect for authority were more likely to say young people lacked respect; avid readers were more likely to say youth did not read enough. Memory is not a perfect recording. The present self helps reconstruct the younger self and the remembered “youth of the past,” making “we were better back then” easier to generate.
A meta-analysis of 100 studies with 7,129 participants also found an age-related relative preference for positive over negative information in attention and memory under some conditions. That does not mean older adults always romanticize the past; it is a qualified cognitive tendency, not a universal diagnosis.[9]
Workplace “generation differences” are also much less dramatic than popular stereotypes suggest. A recent meta-analysis found few systematic, practically meaningful generational differences across workplace outcomes.[10]
Comparing a 60-year-old today with a 25-year-old today does not isolate generation. Age, period, education, economic conditions, role, and institutions are mixed together. To test whether youth changed, compare similar ages across eras—roughly a 25-year-old now with a 25-year-old decades ago. The turnover series then delivers the punchline: the earlier young workers quit too.
Did journalists really fail to notice the long-run graph?
We cannot know what every reporter or editor saw, so there is no evidence for either “everyone knew and deliberately hid it” or “nobody noticed.”
What is clear is that the long-run data were available. The ministry materials around the 34.9% figure include the historical series. Widen the time window and 36.5% in 2000 and 36.6% in 2004 appear.[2][3]
NTV’s wording was “highest in 15 years,” not “highest ever.”[4] That wording is accurate. But the story changes when the baseline changes: a 15-year window produces a local high; a 30-plus-year window produces a high point inside a decades-old band.
News also competes for novelty, conflict, people, visuals, simplicity, and attention. A 2023 study analyzed roughly 105,000 headline variants and more than 370 million impressions; for an average-length headline, one additional negative word increased click-through rate by about 2.3% on average.[11]
An experiment with 136 Belgian political journalists found that stories shown with better audience analytics tended to be ranked more prominently on a simulated homepage, while traditional news values such as negativity also mattered.[12]
That helps explain why “three-year university turnover has hovered around 30% for decades” is less television-friendly than “one in three quits! 15-year high! Generational clash!”
The useful lesson is not “all media are bad.” It is: check the comparison window and the long-run baseline.
The right edge of a graph makes news; scrolling left makes history.
Young people are also perfectly capable of thinking well; age is not a certificate of correctness
Flipping the stereotype into “young smart, old stupid” would repeat the same mistake in reverse. But “less life experience, therefore shallow reasoning” does not follow automatically either.
In the OECD Survey of Adult Skills, Japan’s 25–34 group scored higher on average than the 55–65 group in literacy, numeracy, and adaptive problem solving; the literacy gap was 33 points. Japanese 16–24-year-olds also scored above the OECD average in all three domains.[13]
That does not prove younger cohorts are biologically smarter. Ageing, education, training opportunities, and cohort effects are mixed together, while older adults can have major advantages in accumulated knowledge and experience.
Still, consider the method:
Older worker: “I’ve worked for 40 years, so I know.”
Younger worker: “Does your experience represent the whole cohort? Where are the peers who left? What was the base rate then?”
The second set of questions does not become wrong because the speaker is younger. It is checking the sample and the baseline.
Maybe the real record-holder for retention is the “kids these days” lecture
The 2021 university cohort’s 34.9% really was a 15-year high. But 2000 was 36.5% and 2004 was 36.6%. University turnover has spent decades around roughly 30%, and high-school rates in some earlier periods were in the 40–50% range. Historical evidence from the 1966 cohort also records heavy early movement, though with a different measure.
Interviewing only long-tenured workers invites survivorship bias. The perception that “today’s youth are worse” is itself a researched psychological phenomenon. Giant labels such as “Gen Z” or “Shōwa generation” are crude tools for judging an individual.
So the best summary is not:
“Young people today quit quickly.”
It is:
“Young people have been quitting their first jobs fairly often for a very long time.”
A young worker may leave after three years.
The “young people these days…” lecture has been refusing to resign for centuries.
Sources
[1] MHLW, March 2022 cohort: https://www.mhlw.go.jp/stf/houdou/0000177553_00010.html
[2] MHLW, historical series: https://www.mhlw.go.jp/content/11805001/001318971.pdf
[3] MHLW, March 2021 cohort: https://www.mhlw.go.jp/stf/houdou/0000177553_00007.html
[4] NTV NEWS NNN: https://news.ntv.co.jp/category/society/ed2b354f1643443a8cfbf84a421100f7
[5] Teikoku Databank: https://www.tdb.co.jp/report/economic/5rdo_0wah/
[6] 1984 Labour White Paper: https://www.mhlw.go.jp/toukei_hakusho/hakusho/roudou/1984/dl/02.pdf
[7] Diet record, 2 Apr 1970: https://kokkai.ndl.go.jp/simple/detail?minId=106314410X00919700402
[8] Protzko & Schooler (2019): https://doi.org/10.1126/sciadv.aav5916
[9] Reed, Chan & Mikels (2014): https://pubmed.ncbi.nlm.nih.gov/24660792/
[10] Ravid, Costanza & Romero: https://doi.org/10.1002/job.2827
[11] Robertson et al. (2023): https://doi.org/10.1038/s41562-023-01538-4
[12] Lamot & Van Aelst (2020): https://doi.org/10.1080/1461670X.2019.1686411
[13] OECD, Survey of Adult Skills 2023: Japan: https://www.oecd.org/en/publications/survey-of-adults-skills-2023-country-notes_ab4f6b8c-en/japan_91adbde1-en.html
