Logos52
wiki / Worldviews & the Political Order / East Asian Exams The Arms Race

East Asian Exams - The Arms Race

Concept updated 2026-09-16

One exam at about age 18 decides a Korean or Chinese student's university, and the university decides pay and senior posts for decades. Because the exam ranks students against each other, tutoring by one family forces tutoring by the rest, a child costs about $275,000 to raise, few couples have a second, and 45 years of bans have not stopped it.

East Asian Exams - The Arms Race

In China and South Korea, one national exam taken at about age 18 decides which university a student enters. Graduates of the top universities fill a large share of senior posts in government and business. At Samsung, an employee’s pay grade depends on which university they graduated from. A score from age 18 therefore still sets pay decades later.

An arms race, in everyday use, is two countries building more and more weapons because each is afraid of the other’s weapons. Families sit in the same position over this exam. The exam does not pass every student who reaches a set standard. It ranks students against each other, and only the top of the ranking reaches the universities that lead to those posts. When one family buys more tutoring, its child moves up the ranking and other children move down, so the other families buy more tutoring to catch up. Each family would be better off if every family stopped at the same time. No family stops alone, because a family that stops moves its child down the ranking while the others keep paying.

Core takeaways

  • The exam ranks students against each other, so tutoring bought by one family forces tutoring on the rest, and no family can stop without giving up its own child’s place.
  • Raising a child to 18 in South Korea costs about $275,000, which is 7.8 times Korea’s output per person. About a third of it goes to private cram schools.
  • Because a place costs that much, few Korean couples have a second child. Korea’s fertility rate was 0.72 in 2023, against the 2.1 a population needs to stay the same size.
  • Attempts to stop the race have all failed. China abolished the exam for 11 years under Mao and brought it back in 1977. Korea banned private tutoring in 1980 and a court struck the ban down in 2000.
  • The exam was built so that rich and poor sat identical papers, and rich families now take most of the top places. In 2019, 41% of students at Korea’s three top universities came from high-income families and 5% from the poorest.
  • The same race produces more graduates than there are graduate jobs. China grew university enrollment 47% in the single year 1999, 12.22 million students graduated in 2025, and at least 70,000 Chinese food delivery drivers hold master’s degrees.

What the exam decides

Korea’s three top universities hold about 5% of the country’s undergraduate places and supply 46.3% of its senior government officials. The ranking decides which students get those places. Families pay for private tutoring to move their children up that ranking. In South Korea much of the tutoring happens in private cram schools called hagwons, and nearly half of Korean children under 6 already attend one.

One family buys more tutoring, its child moves up and other children move down, those families buy more tutoring, and the ranking ends unchanged Why the spending keeps rising one family buys more tutoring its child moves up the ranking other children move down those families buy more tutoring the ranking is unchanged and every family is paying more A family that stops moves its child down the ranking while the others keep paying, so no family stops alone.

Tutoring spending in Korea reached a record 29.2 trillion won in 2024, in a year when the number of school-age children fell again. More than 80% of Korean schoolchildren now take some form of private tutoring.

What a place costs a family

Raising a child to 18 in South Korea costs about $275,000. That is 7.8 times Korea’s output per person, against 4.1 times in the United States. About a third of the Korean figure goes to hagwons. A family that wants its child to keep up spends that on one child, and few Korean couples have a second.

cost of raising a child to 18, as a multiple of output per person Cost of raising a child to 18, as a multiple of output per person South Korea 7.8 times 7.8 times United States 4.1 times 4.1 times About a third of the Korean figure goes to private cram schools. The Korean total is about $275,000.

Korea’s fertility rate, the average number of children per woman, was 0.72 in 2023, 0.75 in 2024 and 0.80 in 2025. The 2025 rise came from weddings delayed during the pandemic and from an unusually large group of women reaching their early thirties, and it is expected to fade after about 2028. A population needs about 2.1 to stay the same size. Economists estimate that Korean fertility would be 28% higher if families did not compete over status and schooling. In the poorest fifth of Korean families, they estimate childlessness would fall from about 5% to under 1%.

children per woman Children per woman replacement level 2.1, the level that holds a population steady 2.1 China about 1.0 1 Singapore 0.87 0.87 South Korea, 2025 0.80 0.8 South Korea, 2024 0.75 0.75 South Korea, 2023 0.72 0.72 Taiwan 0.695 0.695 The 2025 rise came from delayed weddings and a large group of women in their early thirties, and is expected to fade after about 2028.

The same pattern holds across the region. Taiwan, which has over 54,000 cram schools, is at 0.695. Singapore, where about 85% of pupils in the last years of primary school take private tuition, is at 0.87. China is at about 1.0, and its childcare subsidy of 3,600 yuan a year covers 0.67% of what raising a child there costs.

Who wins the places now

The exam was set up so that rich and poor students sat identical papers, and for a time it worked that way. It no longer distributes places evenly. In 2019, 41% of students at Korea’s three top universities came from high-income families and 5% came from the poorest. At Seoul National University the share from the top income brackets rose from about 40% in 2017 to 55.5% in 2021. Korea’s top-income households spend nearly seven times what the poorest spend on tutoring. In China, urban students are five times as likely as rural students to reach a top-tier university.

Wealth in Korea is also moving by inheritance rather than by work. Between 2002 and 2022 national income grew 2.7 times, while the value passed down through inheritance and gifts grew 8.3 times.

What the exams measure

In Seoul, five-year-olds sit a two-hour exam to enter English academies, which parents choose so that their children will speak English fluently. The exam has no speaking section. A pass shows that the child can sit through a two-hour adult test and that the family paid for the years of coaching needed to get there.

The Korean exam at 18 contains questions, known as killer questions, that deliberately cover material outside the school curriculum. They exist to separate the top 1% from everyone else. A student cannot prepare for them at school, so the family buys the preparation. One Korean tutor who specialises in them earned about $14 million in a year, more than the country’s best-known footballer.

Every attempt to stop it has failed

Five attempts to stop the race, each followed by the tutoring returning or growing What governments tried, and what happened China, under Mao abolished the exam for 11 years → back in 1977 Korea, 1980 banned private tutoring → struck down, 2000 Seoul, 2009 10 p.m. curfew on cram schools → lessons on buses China, 2021 banned for-profit tutoring → 70% to 80% still attend Korea, 2023 removed the killer questions → families spent more In 1979, 6% of Korean students attended cram schools. By 2008, 75% did.

Under Mao, China abolished its university entrance exam for 11 years. The exam came back in 1977, when 5.7 million people sat it and 5.4% passed. South Korea banned private tutoring in 1980, and officials whose children were caught with a tutor lost their posts. Tutoring went on in secret, with tutors posing as housekeepers and lessons held in cars, and in 2000 a court struck the ban down as unconstitutional. In 1979, 6% of Korean students attended hagwons. By 2008, 75% did.

Seoul then tried a 10 p.m. curfew on cram schools in 2009, and inspectors drove the academy district looking for light behind the blinds. Some hagwons re-registered as self-study libraries, and some taught lessons on chartered buses. In 2023 the government removed the killer questions from the exam, and families spent more on tutoring the following year than before. China banned for-profit tutoring in 2021, which wiped out $77 billion of market value and closed 700,000 centres. Between 70% and 80% of Chinese high school students still take extra classes.

Many Chinese parents defend the exam itself. Without it, they say, the best places would go to the families with the best connections, and a hard test applied to everyone is the fairer of the two.

More graduates than graduate jobs

China grew university enrollment by 47% in the single year 1999, on the argument that families would spend savings on tuition and students would stay off the job market. The ministry’s own planning chief objected at the time that graduates would come to expect jobs the economy could not supply. In 2025, 12.22 million students graduated, and unemployment among Chinese aged 16 to 24 reached 18.9% in August that year. At least 70,000 Chinese food delivery drivers hold master’s degrees, and more than a fifth of drivers on the two largest delivery platforms hold a degree of some kind. Civil service applicants per post rose from 40 in 2021 to 98 in 2026.

China's 1999 enrollment rise, 12.22 million graduates in 2025, 18.9% youth unemployment, 70,000 delivery drivers with master's degrees, and 98 applicants per civil service post What the race turns out China, 1999 university places up 47% in one year 2025 12.22 million graduates August 2025 18.9% out of work, ages 16 to 24 at least 70,000 delivery drivers hold a master's 2026 98 applicants per civil service post Training more people for top positions than there are top positions has a name, elite overproduction.

In South Korea, 504,000 young people were neither working, looking for work, nor studying in February 2024, a record. Pay at a small Korean firm is 54% of pay at a large one. Small firms hold 80% of Korean jobs, and 4% of young Koreans say they want to work at one.

Training more people for top positions than there are top positions has a name, elite overproduction. In one count of historical cases, three quarters of them ended in the breakdown of the state rather than in reform.

Judging people by evidence of ability instead

A different way to sort people is to look at what they have actually done. One company group asks applicants for a few bullet points giving evidence of exceptional ability, and hires on a conversation rather than on the paper record. Its owner says a resume may look impressive and still be worth less than the conversation. If twenty minutes of talk is not remarkable, he takes the talk as the better evidence. He also says he has been fooled by pedigree himself, hiring people because they came from a famous employer.

Evidence of that kind is easier to see than it used to be. Work that people make is published where anyone can look at it, so the record of what a person can do is often available without a test or a diploma. There is a caution against this in the material above. Parents defend the exam because they expect any other method to be captured by families with connections. A method that runs on personal judgment gives connections more room.

Open questions

Parents defend the exam because they expect the alternative to be captured by connections. If places were handed out on evidence of ability, what would stop that capture?

Korea removed the hardest questions in 2023 and families spent more the next year, not less. What change inside the system has ever lowered the spending?

Evidence of ability is visible for people whose work is published. What happens to a person whose work leaves no public record?

Sources