'Rested' vs. NEET: Why the Youth Stats Don't Match

Which Statistics Should We Use to Look at Youth Employment and Health

You’ve probably seen the headlines: more and more young people aren’t working. But a short column in the Financial Times last week raised a question about how we actually measure that. Data journalist John Burn-Murdoch asked something worth sitting with: how much of what shows up in the statistics reflects an actual change in health, and how much reflects a change in how we measure and diagnose it?

He grants that some signals are unambiguous and real — the number of young women showing up in emergency rooms for self-harm, for instance. But he argues that everything stacked on top of that baseline starts to pull in different directions. Depending on who a survey covers and how it measures things, you can end up with contradictory trends.

Reading his piece, I found myself pulling up Korea’s own statistics again.

Britain’s NEET and Korea’s “rested” status aren’t two names for the same group. NEET in the UK refers to 16-24 year-olds who aren’t in education, employment, or training. Korea’s “rested” (swieotseum, an activity-status category for people not economically active) is one activity status within the economically inactive population. The Korean figures I’ll use here also include people in their 20s and 30s, so the age range doesn’t line up either. With that difference on the table, I want to look at how health struggles and employment struggles show up differently across these statistics.

🇬🇧 The Health Problems Britain’s NEET Youth Are Reporting

As of January–March this year, the UK counted 1.01 million NEETs¹ aged 16 to 24. That’s the first time the figure has topped 1 million since 2013.

But what matters more than the headline number is why it’s happening. Among NEET youth, the share reporting a health condition that limits their ability to work rose from 26% in 2015 to 44% in 2025 — an 18-percentage-point jump. We should be careful to separate “reporting a health problem” from “becoming NEET because of it” — correlation isn’t causation here. Still, among NEET youth with a disability, the share citing mental health as their “main condition” nearly doubled, now topping 4 in 10.

These figures come from the “Youth and Work” review, commissioned by the UK government and led by Alan Milburn, the former Health Secretary. It’s an interim report, released this past May 28th, and what caught my attention was its diagnosis of the health section. For the first time in two centuries, it argues, shifts in health — particularly mental health — are constraining economic growth and shrinking the labor supply.

The report puts the social cost of NEET youth at £125 billion (~$156 billion), a figure it says is large enough to compare with the education budget. That said, it’s a composite estimate spanning economic losses, long-term effects, and fiscal line items — not actual government spending. It also includes items like taxes and benefits that, from a whole-of-society view, are really just transfers rather than net losses.

Leaving work for health reasons makes it hard to come back

You also need to consider how long this can drag on. Among people who left the labor market for health reasons between 2017 and 2019, nearly 8 in 10 were still NEET two years later.

The Milburn report locates the cause not in individual young people but in the system. It argues that young people are being classified as “unable to work” through the process of obtaining a medical certificate. Diagnosis, treatment, and discharge all happen — but there’s no pathway back to education or a job afterward, the report points out.

The UK’s Health Foundation examines several explanations: actual declines in health, access to diagnosis and treatment, shifts in awareness, and institutional incentives. Among these, the hypothesis that the system itself can shape how people respond goes like this:

• Some education and welfare support programs require confirming whether someone has a health difficulty.
• In the process of seeking out support, a person may end up receiving a diagnosis or assessment.
• This experience can then shape how they describe their own difficulties — including in later survey responses.

Not all support requires a medical diagnosis, and this process hasn’t been confirmed as the cause of rising response rates.

Greater awareness of health issues and better access to support can coexist with an actual worsening of health. This analysis alone can’t support the conclusion that the number of sick young people hasn’t actually grown.

🇰🇷 South Korea’s “resting” status classifies activity, not health

Now let’s look at South Korea.

As of 2025, the number of people in their 20s and 30s who are neither working nor job-hunting — classified as “resting”² — stands at 717,000. That’s the highest figure since record-keeping began in 2003. It represents 5.8% of the 20s-30s population, and among those in their 30s alone, the figure of 309,000 is likewise an all-time high.

South Korea’s Economically Active Population Survey has separate categories for “physical or mental disability” and “resting.” But being classified as “resting” doesn’t mean a person has no health issues at all. A person’s primary activity status and the reason they ended up in that status are simply two different questions.

The Answer Splits by Age

When last August’s supplementary survey asked people why they weren’t working, here’s what came back.

Age	Top reason	Share
15–29	Hard to find the job I want	34.1%
30s	Poor health	32.0%
60+	Poor health	38.5%

People aged 15–29 don’t rank health as their top reason. “Hard to find the job I want” tops their list, up 3.3 percentage points from the year before — the largest increase of any age group. Among people in their 30s, “poor health” was the most common answer.

The wording of the question itself is worth scrutinizing. Whether the phrase “poor health” adequately captures anxiety or depression is something a separate survey would need to confirm. Still, the wording alone doesn’t let us conclude that respondents excluded mental health issues.

The UK’s survey asks it differently. It asks “Do you have a health problem that has lasted, or is expected to last, 12 months or more, that limits your work?” — and if the answer is yes, respondents are asked to identify the main condition separately. That list of options explicitly includes “depression, nerves, anxiety.”

How you classify unemployment changes what the number means

The Korea Development Institute (KDI) found that rising numbers of “resting” people in their 20s is one factor behind the low unemployment rate. Under the assumption that this population grew less than it actually did, the estimate is that 45–71% of the drop in the unemployment rate over the study period can be explained this way. Only the upper bound of this assumption-dependent range should not be read as a confirmed contribution rate.

The unemployment rate is the share of unemployed people within the economically active population. If someone isn’t actively job-hunting, they’re generally classified not as unemployed but as economically inactive — which means the unemployment rate can fall even without any increase in actual employment. This isn’t just a matter of relabeling; it reflects how job-seeking behavior and statistical criteria are linked. You need to look at the unemployment rate alongside the employment rate, job-seeking intentions, and the reasons behind “resting.”

🔀 Two Statistics With Different Subjects and Questions

Let me put the two countries side by side.

Category	Statistic examined for the UK	Statistic examined for Korea
Primary group	16–24 year-olds not in education, employment, or training	20-somethings and 30-somethings classified as “resting” (a Korean labor-survey category for those neither working nor job-seeking); a separate age band, 15–29, is used in the survey on reasons for resting
Health-related figure	Share of NEETs reporting a health problem that limits their ability to work	Reasons given by the “resting” group for why they are resting
Caution when comparing	Having a health problem isn’t necessarily what caused someone to become a NEET	Being classified as “resting” doesn’t necessarily mean someone has no health problem

There may well be people in both groups facing similar difficulties, but the figures presented here don’t let us say that the same people were simply classified differently in each country. And statistical categories alone don’t automatically determine who gets support or how budgets are allocated.

We should also examine how support systems relate to diagnosis

The UK does have cash benefits tied to health and disability, but a diagnosis alone doesn’t automatically trigger payment. Korea, too, offers more than just counseling vouchers—there’s also treatment and regional mental health services, among other forms of support. Comparing the two countries’ entire support systems using only cash benefits versus 8 counseling sessions doesn’t really hold up.

Korea’s National Mental Investment Support Project provides a total of 8 professional counseling sessions. There are multiple paths to apply: people recognized as needing counseling not just at psychiatric institutions but also at mental health welfare centers, university counseling centers, and similar venues, or people who scored 10 or above on the PHQ-9³ during a national health checkup. A diagnostic certificate or checkup score isn’t the only route to eligibility.

South Korea Expands Mental Health Screening for Young Adults

Starting in 2025, the national mental health screening interval for young adults aged 20-34 shifted from every 10 years to every 2 years — five times more frequent. A new early psychosis screening component was also added.

The Ministry of Health and Welfare’s rationale is straightforward. South Korea’s mental health service utilization rate stands at 12.1%, far behind Canada (46.5%) or Australia (34.9%), and among young adults specifically it’s just 16.2%. The government has framed the goal as earlier detection and better access to services.

More frequent screening will surface difficulties that went undetected before. Going forward, any statistics on this should be read alongside how many people were tested, who the target population was, and what the actual results showed.

📏 The share above the PHQ-9 cutoff isn’t the same as the clinically diagnosed rate

The PHQ-9 is a tool used to gauge the severity of depressive symptoms and flag people who need further evaluation. If you read the share of people who cross the score cutoff as the clinically diagnosed prevalence of depression⁴, you’ll end up with a gap.

A large-scale meta-analysis published in the Journal of Clinical Epidemiology in 2020 pinned this down precisely. Pooling individual participant data from 44 studies covering 9,242 people, it compared the PHQ-9 cutoff of 10 points against results from structured clinical interviews⁵, and here’s what it found:

• Pooled estimate of the share scoring 10 or above on the PHQ-9: 24.6%
• Prevalence by structured clinical interview: 12.1%
• Average ratio across studies: a 2.5x overestimate

The paper concludes that using this score cutoff as a stand-in for depression prevalence calls for caution. Note that dividing the pooled rates directly and averaging the ratios across individual studies are two different calculations.

But this finding doesn’t mean the PHQ-9 is unfit for screening purposes or as a threshold for offering counseling support. The goal of estimating prevalence and the goal of getting counseling to people who need it are two different things.

Korea’s three statistics measure different subjects and metrics

The three sources measure different things. They’re not the same indicator that lets you directly compare an increase against a decrease.

Measurement tool	Result from each source	Caution when interpreting
HIRA (Health Insurance Review and Assessment Service) treatment statistics	Number of people in their 20s treated for depression up 127.1% over 5 years	Counts people who received treatment
Ministry of Health and Welfare’s 2021 Mental Health Survey	1-year prevalence of mental disorders among adults: 8.5%	Covers ages 18–79, includes disorders other than depression. Methodology also differs from earlier surveys
Economically Active Population Survey	717,000 people in their 20s–30s reported as “resting” (not working, not job-seeking) in 2025	Measures activity status, not a health diagnosis

The first counts people who came to a hospital. The second measures actual condition through diagnostic interviews. The third asks whether someone is working or not. The three are counting different things. You can’t simply combine these three sources and conclude that depression among young people has risen or fallen.

When HIRA’s statistics were released in 2022, an executive of the Korean Neuropsychiatric Association offered this comment: if the prevalence rate hadn’t risen notably but the number of patients in their 20s had jumped sharply, that likely reflects a lower barrier to seeking psychiatric treatment, not necessarily a rise in the underlying condition. In other words, access to treatment can itself shape the statistics.

Even government documents mix up the terms

One more thing worth flagging. According to the “10-Year Depression Prevalence Status” data that the Ministry of Health and Welfare submitted to the National Assembly, depression prevalence rose from 1.16% in 2014 to 2.3% in 2023. But the number of patients treated in that same dataset went from 584,948 to 1,043,141.

If you divide 1,043,141 by South Korea’s population of roughly 51.7 million, you get about 2.0%. That doesn’t match the 2.3% figure in the data. We’d need to check what population was used as the denominator. This discrepancy alone isn’t enough to pin down whether it’s a denominator issue or an outright error.

Either way, this is treatment utilization, not epidemiological prevalence. It’s the share of people who went to a hospital, not the share of people who have the condition. Figures based on treatment records and prevalence estimated through clinical interviews need to be clearly labeled as separate things.

💔 Real Risk and the Need for Support Must Also Be Checked

The reason I’ve been pointing out the limits of measurement is to more accurately judge what kind of support is actually needed.

In 2024, emergency room visits for self-harm or suicide attempts totaled 35,170 cases. This is a count of visits, not unique individuals, so the two should be kept distinct. Teenagers and people in their twenties accounted for 39.9% of these. Women accounted for 21,479 cases, far more than men’s 13,691, and among women, those in their twenties made up 26.6% and teenagers 20.6% — concentrated heavily among the young.

The trend is also clear. Between 2018 and 2022, self-harm and suicide attempts among teenagers rose from 95.0 to 160.5 per 100,000 people — a 68.9% increase. Among those in their twenties, it rose from 127.6 to 190.8, a 49.5% increase. Over the same period, the increase across all age groups combined was just 11.8%. In other words, the spike was uniquely steep among young people.

A study analyzing emergency rooms in Seoul over eight years found that among patients aged 24 or younger who came in for suicide attempts or self-harm, 75.4% were women.

Emergency room data is crucial evidence for grasping serious risk. The survey methods and this measurement approach differ, but this data, too, can be shaped by hospital usage patterns and the scope of reporting and counting. Rather than treating any single source as a complete measurement, we need to look at them together.

The effectiveness of policy is also hard to judge from a single indicator. We need to check symptoms, healthcare utilization, crisis situations, and return to education or employment separately.

For reference, Tyler Cowen, the economist who introduced John Burn-Murdoch’s column, attached this caveat: this interpretation isn’t decisive, but neither is the opposing interpretation. I hold the same view. What’s certain is this — the measurement tools we’re currently using aren’t precise enough to capture this problem accurately.

Oswarld’s Lens

There’s a failure pattern I’ve run into again and again while building GTM strategies: when you turn a metric into a KPI, people start optimizing for the number itself, and that can pull them further from the original goal.

The case I’ve seen most often is MQLs. When a marketing team sets verified leads as its KPI, the number can climb without ever translating into revenue. What changed wasn’t people’s behavior — it was the definition of a “lead.” Once a number becomes the target, the fastest thing to get optimized is the method for producing that number.

Reading the UK analysis, this experience came right back to me. That said, my experience with marketing metrics can’t prove that health diagnoses are being inflated. I think we need to check whether support programs and measurement methods might be shaping how people respond, while separately examining whether symptoms are actually worsening and what treatment is actually needed.

What worries me in particular is whether we’re gathering enough information to design proper support.

Korea’s “swieoteum” population — people who report they’re “resting,” neither working nor job-seeking — is already captured in statistics, and there’s a supplementary survey asking why people are resting, along with related support programs. But the total headcount alone can’t tell us enough about each person’s health, willingness to seek work, caregiving burden, or the support they actually need. Like the UK’s estimates of social cost, Korea should also make explicit what’s being calculated and how, and have a concrete discussion about the costs and benefits of support.

I think what matters more than renaming categories is understanding people’s actual situations in greater detail, and tracking how things change after support is provided. Employment and health should be examined together — but one statistic should never be used to explain away the meaning of another.

Closing

The UK’s NEET category and Korea’s “resting” (swieosseum) classification differ in both scope and definition. Treatment counts, screening scores, and clinical interview results each tell a different story. As screening and support expand, we need to record these distinctions clearly if we want to properly judge which support and policies are actually working.

Next time you see a headline about a “youth mental health crisis,” check what instrument produced that number. Whether it’s a treatment count, a screening score, or a diagnostic interview changes the story entirely.

If you’ve run into unclear guidance or a difficult application process while trying to find support, I’d like to hear about it.

---

Today’s piece dealt with statistics and institutions. But if you’re going through a hard time right now, you don’t have to get through it alone. You can call the Suicide Prevention Counseling Line at 109 for 24-hour support, or if you’d rather text or use KakaoTalk, the SNS counseling service “Madlen” is also open around the clock.

Ministry of Health and Welfare Counseling CenterDial 129 (no area code needed) for welfare counseling services covering emergency welfare, welfare support, suicide prevention, and alcohol addiction.129.go.kr

---

📨 If someone around you could use this, please share it.

---

References & Further Reading

Primary sources

• John Burn-Murdoch, “What’s really going on with mental health?”, Financial Times, July 2026. : This is where today’s piece started. Since it only covers UK and US data, I added the Korea angle myself.
• Alan Milburn, “Young people and work: interim report”, UK Government, May 28, 2026. : This is where the 44% and £125 billion figures come from. The introduction is especially worth reading. The final report comes out this September.
• Brooke Levis, Andrea Benedetti, John P. A. Ioannidis, Brett D. Thombs et al., “Patient Health Questionnaire-9 scores do not accurately estimate depression prevalence: individual participant data meta-analysis”, Journal of Clinical Epidemiology, 2020. : This paper addresses the limits of using screening scores to estimate prevalence. Pooling 44 studies, it shows that a PHQ-9 cutoff of 10 inflates prevalence estimates by a factor of 2.5.
• The Health Foundation, “Why are a growing number of young people who are NEET reporting work-limiting health conditions?”, March 2026. : This piece weighs multiple explanations — actual health status, access to diagnosis, and shifts in awareness and institutions. It’s the source I leaned on most for today’s issue.

Korea data sources

• KDI, Causes and Implications of the Recent Low Unemployment Rate, 2025. : Analyzes the relationship between rising “resting” numbers and the unemployment rate under different assumptions.

• Ministry of Health and Welfare, Guide to the National Mental Health Investment Support Program, 2024. : Shows the various pathways through which a need for counseling can be officially recognized.

• National Data Agency, 〈August 2025 Supplementary Survey of the Economically Inactive Population〉: The raw data behind the age-group breakdown of reasons for “resting.” Looking at the table directly, the gap between people in their 30s and those in their 20s really jumps out.

• Ministry of Health and Welfare, 〈2021 Mental Health Survey〉: The source of the 8.5% one-year prevalence figure. Be sure to read the footnote noting that the survey methodology changed from the previous round.

• Health Insurance Review & Assessment Service, 〈Five-Year Statistics on Depression and Anxiety Disorder Treatment〉, June 2022: The source of the 127.1% increase among people in their 20s.

• Ministry of Health and Welfare, 〈Expansion of Mental Health Screening for Young Adults〉 press release, October 17, 2024: The document announcing the decision to shorten the screening interval from 10 years to 2 years.

Background

• Ben Baumberg Geiger, Melanie Jones & Victoria Wass, “Disability prevalence and disability-related employment gaps in the UK 1998-2012: Different trends in different surveys?”, Social Science & Medicine, 2015. : Evidence that this debate isn’t new. The same authors found an identical discrepancy in physical disability statistics 11 years ago. The title says it all: different surveys, different trends.
• Christoph Henking & Ben Baumberg Geiger, SocArXiv preprint, 2026. : The original source of the “institutional medicalization” hypothesis. Keep in mind it’s a preprint that hasn’t yet been peer-reviewed.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

---

📝 Glossary

Footnotes

1. NEET: A status describing someone who is not enrolled in school, not working, and not in job training. The UK counts this for ages 16-24. It differs from unemployment — the unemployed are actively looking for work, while NEET also includes people who have stopped job-seeking entirely. ↩

2. “Resting” (swieotseum): A category in Korea’s Economic Activity Population Survey. When asked “What did you mainly do last week?”, people who answer something other than childcare, housework, schooling/coursework, old age, or physical/mental disability are classified here. It isn’t identical to “other reasons” as a whole, and can reflect anything from temporary rest to health or job-search issues. ↩

3. PHQ-9: A nine-item self-report questionnaire for assessing depressive symptoms. A total score of 10 or above is considered “moderate or greater depression.” It’s a screening tool for flagging people who need further evaluation, not a diagnostic instrument. ↩

4. Prevalence: The proportion of people who actually have a condition at a given time or over a given period. This is a completely different concept from the rate of clinical utilization — the share of people who actually go to the doctor. Even if 100 people have a cold, only 30 might see a doctor. Conflating the two badly distorts the numbers. ↩

5. Structured clinical interview: A diagnostic method in which a trained interviewer conducts a direct interview following a fixed sequence and set criteria. It’s far more accurate than self-report questionnaires, but it’s costly and time-consuming. That’s why large-scale surveys often substitute questionnaires instead — which is exactly where the error creeps in. ↩