1. The Hour the Signs Come Down
Winter 2024, Mapo-gu, Seoul.
At the end of an alley, a crane truck is pulling down a sign. The four characters reading "○○ Chicken" take thirty minutes to separate from the wall.
Putting that sign up took ₩120 million in severance pay, ₩50 million in loans, and twenty-two months. Taking it down takes thirty minutes.
The sign is the franchisor's property. Only the name was his, and now the name has been removed too. A crew sent by the franchisor does the work. The cost of demolition falls to the shop owner.
In the final line of Chapter 4, we wrote: "Where signs come down, a different kind of displacement is beginning." This is where that displacement happens.
This alley's third closure of the year. The first was a café. The second was a snack bar. The third is the chicken place.
All three proprietors were in their fifties. All three started with severance pay. None of the three lasted three years.
A nail salon has moved into the café's vacated space. The snack bar's spot still has only a phone number for rental inquiries. What will take the chicken place's spot is not yet known.
Three to six months from now, a new sign will go up. The probability that another chicken place opens in the same spot is 20 to 30 percent. Everything else brings a different type of business — and half of those also disappear within three years.
In 2024, the number of business closures in Korea was 1.08 million. The highest figure recorded since statistical tracking began in 1995. A closure rate of 9.04 percent. Among self-employed workers, 44 percent said they were "considering closure within three years."
Nearly half are looking for a way out.
That figure of 1.08 million must be read differently. It means the assets of 1.08 million households evaporated. The weight of a single sign is not just metal and acrylic. Behind it lies someone's twenty-eight years, someone's severance pay, someone's daughter's tutoring fees.
The 1.08 million did not suddenly explode in 2024 alone. In 2020 and 2021, COVID-19 relief funds kept self-employed workers at the edge of viability alive. Government loan support and rent subsidies suppressed closures. After support ended in 2022, accumulated insolvency collapsed all at once.
The 1.08 million is not a temporary surge. It is the delayed reckoning of structural rot.
The faster the turnover, the faster the capital loss for individual operators accelerates. This is not a signal that the ecosystem is healthy. It is a signal that the structure is grinding people through.
2. The Chain of Displacement
This structure has a name. The chain of displacement.
In Book 5 we observed the archetype of this chain. Manufacturing automation displaces skilled workers; the displaced migrate to the service sector; excess competition in the service sector displaces them again. Book 5 analyzed this pattern at the national level. Chapter 5 traces it in the arc of one person's life.
The first link of the chain is manufacturing. Korean manufacturing employment peaked at 4.67 million in 2016 and has been declining since. By the end of 2023: 4.41 million. Over seven years, 260,000 jobs vanished.
Manufacturing's share of total employment fell from 20.4 percent in 2000 to approximately 15 percent in 2024. More serious than the aggregate decline is the qualitative shift in employment. Regular full-time work has contracted; fixed-term and dispatch positions have expanded. Real wages for production-technical workers have stagnated relative to white-collar workers. The number of manufacturing workers enrolled in Employment Insurance has also been declining continuously.
The second link is retirement. The average actual retirement age at large Korean corporations falls between fifty and fifty-three. Between the statutory retirement age of sixty and actual retirement, there is a gap of seven to ten years. Until the age at which National Pension benefits become available, there is another ten years still to go.
In that double gap — between actual retirement and statutory retirement, and again between statutory retirement and pension eligibility — the rate at which fifty-something male manufacturing retirees re-enter regular full-time work is 18 percent. Those who re-enter on non-regular or contract terms account for 25 to 30 percent. For the remaining half, it is either self-employment or withdrawal from job-seeking.
Even among those who do re-enter, fewer than 10 percent are estimated to maintain 80 percent or more of their previous wage. The issue is not whether re-entry happens but the quality of that re-entry.
The third link is self-employment. Between 35 and 40 percent of retirees make the transition to self-employment. With ₩150 million to ₩250 million in severance pay in hand, and a chicken franchise start-up cost of ₩80 million to ₩150 million, the figures align. Severance pay and start-up costs occupy the same bracket.
The wage gap between large and small companies reinforces this path. Average wages at small and medium-sized enterprises run at roughly 57 percent of large-corporation average wages. For someone leaving a large corporation, re-entering a small company means a wage reduction of 40 percent or more.
When re-employment is difficult, a lump sum is available, and the path of "I suppose I'll open a chicken place" has been socially normalized, entry into self-employment looks like a rational choice. It is not a choice. It is structural pressure.
The fourth link is closure. The three-year survival rate for chicken restaurants is 45.4 percent. Of 100 who start, 55 close within three years. The five-year survival rate falls to 20 to 25 percent. Two in ten survive.
The average recovery rate for start-up capital after closure is 30 to 40 percent. Invest ₩100 million and what remains in hand is ₩30 million to ₩40 million. The remaining ₩60 million to ₩70 million is sunk into franchise fees, interior fit-out, and equipment. Early termination of the franchise agreement draws penalties on top.
The fifth link is asset depletion. The time it takes for severance pay to evaporate is less than three years. When living expenses and private tutoring fees are covered from the remaining severance, ₩3 million to ₩4 million flows out each month. Thirty months and the account is empty.
Seventy-five to eighty percent of Korean household assets are locked in real estate. The OECD average is 50 to 60 percent — making Korea's concentration anomalous. Liquid assets are consumed first; then the household survives on mortgages against real estate; eventually, pressure to sell the home builds.
Self-employment closure feeds directly into housing insecurity. That is the chain.
And the chain extends to the next generation. The depletion of parental assets constrains educational investment in children — or else debt-financed educational investment leads straight into occupations with high AI exposure. Educational investment risks becoming not an upward ladder but an entrance ramp into the next round of displacement.
This is not business failure. It is one link in a structural chain.
3. A Map of Numbers
Lay the numbers out and they become a map. A map showing where displacement begins and where it flows.
Korea's self-employment rate is 23 to 24 percent of total employed workers. The OECD average is 15 percent. The United States is at 6 percent, Germany at 9 percent, Japan at 10 percent.
The OECD countries with higher self-employment rates than Korea are Greece, Turkey, and Mexico — countries at different levels of economic development, or with inadequate social safety nets. Korea's self-employment rate is an outlier among advanced economies.
The rate has been declining continuously from 35 to 37 percent in the early 1990s. The fact that despite a downward trend it still far exceeds the OECD average reflects structural factors — the combination of early retirement and an inadequate social safety net — that have become entrenched.
There are approximately 80,000 to 90,000 chicken restaurants. McDonald's has 40,000 locations worldwide. Korea has twice as many chicken places as McDonald's has outlets globally. In a population of 52 million, each chicken restaurant has a potential customer base of 580 to 650 people.
At that level of competitive density, the structure makes it difficult to turn a profit. The reason so many chicken places exist is that the barriers to entry are low. The cooking technique is relatively simple; delivery platforms have opened a distribution channel. Easy to start means competitors can enter just as easily.
When a chicken restaurant posts monthly revenue of ₩6 million — based on a mid-scale commercial district — subtracting ₩2 million in ingredients, ₩800,000 in rent, ₩600,000 in labor, ₩1.44 million in platform fees, and ₩660,000 in miscellaneous costs, what remains is ₩1.5 million.
The minimum-wage monthly salary as of 2025 is ₩2.06 million.
The owner earns less than minimum wage. After investing more than ₩100 million in severance pay, waking at five in the morning, working until eleven at night — the return is sub-minimum-wage income.
Subtract ₩800,000 in private tutoring fees for a daughter and ₩700,000 remains. Pay ₩200,000 in monthly loan interest and ₩500,000 remains. From that, a fifty-three-year-old man must cover his own living expenses, National Pension contributions, and health insurance premiums. Self-employed workers bear the full premium themselves.
Franchise systems are designed so that "someone without experience can get started." But when people without experience enter in mass, the survival rate of the entire market falls.
What the franchise has verified is "the franchisor's revenue model," not "the franchisee's viability." The franchisor suffers no loss when a franchisee closes. On the contrary, a new franchisee pays the same franchise fee and moves into the vacated spot.
All of it is written in the contract. Not fraud — structurally designed asymmetry.
These numbers are not a record of individual failure. They are coordinates that reveal the slope of the structure.
4. The Paradox of Overqualification
The chain of displacement does not operate only from below. It also operates from above.
Korea's university enrollment rate is 76 percent. First in the OECD. Germany's is only 33 percent. In Germany, more than 65 percent of high school graduates pass through the dual vocational education system. In Korea, not attending university is close to social stigma. In Germany, it is a respected alternative path.
The same years of education invested produce entirely different structures.
Total private tutoring expenditure is ₩29.2 trillion. A record high for the fourth consecutive year. An average of ₩470,000 per student per month goes into private tutoring. For high school students, the figure is approximately ₩700,000.
Seoul runs 40 to 50 percent above the national average. High-income households spend ₩676,000 per month; low-income households spend ₩205,000 — a gap of 3.3 times.
From elementary through high school — eleven years — tutoring costs alone amount to ₩43 million to ₩66 million. Add university tuition and the total investment per child runs from ₩100 million to ₩300 million.
The destinations of this investment are three. Corporate white-collar positions, public service posts, and professional licenses. All three paths carry non-trivial AI exposure. As analyzed in Chapter 4, the AI task-replacement rate for clerical workers is estimated at 60 to 75 percent. The professional group — physicians, attorneys, accountants, teachers, journalists — falls at 45 to 65 percent.
The places in which Korean society has concentrated its greatest educational investment are the places most exposed to AI.
The wage premium for university graduates is contracting. The university-to-high-school wage ratio, which was 150 to 160 percent in the early 2000s, had narrowed to 130 to 140 percent by 2023. The premium has not disappeared, but investment costs rise every year while the return rate falls every year.
The competitive ratio for Grade 9 civil service examinations is 22 to 1. Hundreds of thousands of young people invest two to five years in preparing for the exam. The opportunity cost is tens of millions to ₩100 million won. The work performed by civil servants who clear that competition is entering the zone of AI automation.
Between the moment of investment and the moment of return, the environment has already changed.
At the same time, Korea's total fertility rate is 0.72. The lowest in the world. The pattern is one of concentrating greater educational investment in a single child while forgoing having children at all. Total private tutoring expenditure is at an all-time high while the student population is shrinking. Investment per capita is intensifying.
The destination of that concentrated investment is being shaken by AI.
Here lies the paradox: "majors with high employment rates also carry high AI exposure." Business administration employment rate 65 to 70 percent, AI exposure high. Computer science employment rate 75 to 82 percent, AI exposure medium to high. Medicine employment rate 95 percent, with AI pressure advancing rapidly in some specialties including radiology.
High employability is not the same as safety from AI. This is the core information deficit facing Korean parents who pay tutoring bills.
In Chapter 3 we analyzed the weight of the business card — the structure in which the title inscribed on a card becomes the anchor of identity. The ₩29.2 trillion in private tutoring is the investment made to obtain that card. When the value of the card wavers, the premise of the investment collapses.
The parents who conscientiously paid tutoring fees, who pushed their children to study for SKY universities, who got them into corporate white-collar positions. People who fulfilled everything the system demanded. The destination of that faithfulness is now in motion.
5. Platform Dependency
In the chain of displacement, there is another wall that self-employed workers encounter. The platform.
Three delivery platform companies control 97 percent of the market. Baemin at 59 percent, Coupang Eats at 24 percent, Yogiyo at 14 percent. An oligopoly. Sixty to 70 percent of chicken restaurant orders come through these platforms.
Operating without the platforms has become impossible.
In economic terms this is called platform dependency. Once dependent, the cost of leaving is too high, and the franchisee has no negotiating power over fee increases. When Baemin raised its brokerage commission from 5.8 percent to 6.8 percent in 2024, small-business owner associations protested strongly. But there was nowhere else to go.
The actual fee burden is approximately 24 percent of revenue. That figure combines the brokerage commission, delivery fees, payment processing fees of approximately 3 percent, and advertising fees. On monthly revenue of ₩6 million, ₩1.44 million goes to the platform.
Surviving on orders that bypass the platform is not possible. Without the fees, only 30 percent of revenue remains. That alone is not enough to stay open. With the fees, ₩1.5 million remains.
There is no exit.
Self-employed workers now have two landlords. They pay physical rent to the building owner; they pay digital rent to the platform. A double-rent structure. For a self-employed worker whose delivery revenue exceeds 50 percent of total sales, the platform is an invisible landlord imposing costs on par with the building owner.
Ranking within the platform is determined by advertising spend. Without achieving high exposure, finding the shop in the app becomes difficult. Spend on advertising and profits fall; profits fall and higher revenue becomes necessary.
At the same time, the platforms hold all the data — each shop's order records, customer reviews, reorder rates, popular menu items. The shop owner does not even have full access to their own store's data. Informational asymmetry creates asymmetry of power.
The other axis of platform dependency is the rider. Approximately 660,000 workers are employed on delivery platforms. 220,000 full-time, 440,000 as a side income. The median monthly income of the 400,000 to 500,000 delivery riders falls between ₩1.5 million and ₩2.5 million.
Enrollment in the four major social insurances is low; accident risk is high; algorithms determine dispatch and route. The rider is a "freelancer" dependent on the platform — doubly vulnerable, with neither independence nor protection.
The statistical reality is that a person who closes a chicken restaurant and transitions to delivery riding is a common path. Those displaced from self-employment move into platform labor. But that path, too, is a structure of subjugation to algorithms.
The chain of displacement does not end at self-employment.
In Book 1 we observed the process by which handloom weavers were displaced by the power loom. To the weaver, the power loom was not a tool but a structure. Accept it and wages fell 82 percent; refuse it and the job was lost. The platform is the same structure. Accept it and 24 percent of revenue must be paid; refuse and 70 percent of customers are lost.
The power loom of two hundred years ago, and the delivery app of 2024. The tool has changed. The logic of the structure is the same.
6. The Anxiety That Has Arrived, the Institution That Has Not
This series has followed a single formula.
Technological innovation produces concentration of capital; concentration of capital produces social instability; social instability triggers institutional redesign. In Book 6, this formula has been extended. Individual adaptation and informal institution-building intervene between social instability and formal institutional redesign.
In Chapter 5, the formula has reached its third term.
Technological innovation. Korea's robot density is 1,012 units per 10,000 manufacturing workers. First in the world. The gap with second-place Singapore at 730 units and the global average of 151 units is overwhelming.
The corporate adoption rate for generative AI, measured among large enterprises, is 45 to 55 percent. Small and medium-sized enterprises remain at 10 to 15 percent, but indirect exposure through large corporate supply chains is spreading rapidly. The number of AI medical devices approved by the Ministry of Food and Drug Safety has exceeded 180. From 21 in 2019, that figure grew ninefold in five years.
The first wave of automation — the stage at which robots replace simple repetitive work — is already complete. The second wave — the stage at which AI absorbs judgment, analysis, and coordination — is underway.
Concentration of capital. Three delivery platform companies hold 97 percent market share. Kakao Bank manages 26.7 million customers with a few thousand employees — more than ten times the per-employee efficiency of a major commercial bank branch.
The beneficiaries of that efficiency are the platforms and capital; the bearers of its cost are self-employed workers and middle-tier labor. In Book 2 we analyzed the mechanism by which technological innovation accelerates the concentration of capital. That mechanism has arrived — in the form of delivery apps and internet banks — in the alleys of Seoul.
Social instability. 1.08 million closures. An 18 percent regular employment re-entry rate for fifty-something male manufacturing retirees. A youth extended unemployment rate of 20 to 24 percent. The OECD estimates that 38.8 percent of Korea's current jobs face high automation risk — approximately three times the OECD average of 14 percent.
These numbers are the concrete content of the abstraction called "social instability."
The fourth term — institutional redesign — has not yet arrived. A cap on platform commission rates is under discussion but has not been legislated. The AI Basic Act (인공지능 기본법) is delayed in the National Assembly. The Tomorrow Learning Card (Korea's public reskilling subsidy) provides ₩3 million to ₩5 million in annual training expenses, but the connection between completing training and actual re-employment is weak.
Korea's unemployment benefits last a maximum of nine months at 60 percent of previous wages. Denmark guarantees up to two years at 90 percent. Self-employed workers are largely excluded from Employment Insurance.
Social instability has arrived. Institutional redesign has not yet departed. In the gap between the two, individuals must construct their own safety net from severance pay. That safety net is the chicken place.
When the chicken place fails, the safety net disappears and the individual stands on bare ground.
7. Two Chairs — The 1997 Retiree and the 2024 Retiree
December 1997, Seoul.
Immediately after the IMF bailout, a mid-sized auto parts company adds a name to its restructuring list. Kim Yeong-su (a pseudonym), forty-seven, department chief of production management. He had done everything the company asked. He worked late nights, came in on weekends, memorized the quality-control manual.
The criterion for restructuring was not diligence. It was total labor cost. Eliminating one department chief freed up the combined annual salary of two assistant managers. He was diligent, and so he was expensive; he was expensive, and so he was cut.
With his severance pay and a loan, he started a small parts-machining shop. Three lathes, two employees. The hands that had been cutting metal for twenty years were still alive. He held on in a market where parts unit prices had fallen 30 percent in the wake of the IMF crisis. He lasted three years. He closed in 2001.
But during those three years, he worked with his own skills.
Autumn 2024, Asan.
Lee Jung-hoon, fifty-three, department chief of production technology. He, too, had done everything the company asked. He listened to the sounds, sensed the tremors of the mold, dedicated twenty-eight years to reducing the defect rate.
When the AI system was introduced, his title changed to "AI System Monitoring Supervisor." The hollowing-out structure we saw in Chapter 4 — the title remains, the content disappears — was applied to him as well.
The man who pressed the confirmation button. Then he retired. He opened a chicken restaurant with his severance pay. He lasted twenty-two months. The sign came down.
Twenty-seven years apart. The structure is the same. Those most faithful to the previous system are the first to be pushed aside. In 1997, what did the pushing was the foreign-exchange crisis. In 2024, what does the pushing is AI.
Only the name of the force has changed. The profile of the person being pushed aside is strikingly the same. Fifties, manufacturing, management, diligent. Someone who gave their best in the role of breadwinner.
One thing is different. Kim Yeong-su in 1997 had an option — parts machining — through which he could apply his own skills, even in a shrinking market. For Lee Jung-hoon in 2024, the chicken restaurant was an option unconnected to his twenty-eight years of accumulated expertise.
What AI replaced was not only the job. The transition path after retirement has also narrowed.
Kim Yeong-su could carry his skills out with him. Lee Jung-hoon had to leave his skills behind. Because AI had absorbed them.
The chain of displacement has grown shorter. The distance from retirement to asset depletion has drawn closer. The buffer zone has contracted.
8. Where Are You Standing?
AI exposure and capacity for transition. Plot these two axes and four quadrants appear.
The first quadrant is high AI exposure combined with high transition capacity. IT developers in their thirties and forties, young financial analysts, freelance designers. People for whom the situation is dangerous but movement is possible. Moving now is the best course.
The second quadrant is high AI exposure combined with low transition capacity. Manufacturing managers in their fifties, bank tellers in their fifties, middle-aged general office workers. People for whom the situation is dangerous and movement is also difficult.
This is the most vulnerable group. Age, finances, network, psychology — four walls blocking simultaneously. A structural problem in which individual-level resolution is nearly impossible.
The third quadrant is low AI exposure combined with high transition capacity. Physicians, architects, strategy consultants. People whose licenses provide protection, whose work centers on judgment, and who have resources to move.
A relatively advantaged position. But this position is not permanent either. As Chapter 4 showed, radiology imaging AI is already reporting performance on par with specialists.
The fourth quadrant is low AI exposure combined with low transition capacity. Care workers, caregivers, basic service workers. AI does not replace them. What the hands do, what the body does, remains.
Yet low wages and social isolation create a different form of vulnerability. Stable but isolated.
Lee Jung-hoon occupies the second quadrant. His core work of twenty-eight years — process quality judgment, production schedule coordination, team management — has been replaced by AI and IoT systems. AI exposure is high. At fifty-three, with most of his severance pay spent, he has a network inside manufacturing but finds it difficult to apply toward an external transition. Transition capacity is low.
And yet Lee Jung-hoon's decision to go to Hai Phong, Vietnam, is an attempt to shift quadrants. Automation process technology that has already been displaced in Korea remains scarce in Vietnam. Change location and the exposure level changes.
In Book 5 we analyzed that "indispensability is a function of position." The same skill that is surplus in Korea is scarce in Vietnam. In the prologue we wrote "Incheon–Hai Phong, one way." That one-way ticket is an attempt to move from the second quadrant to the first or third.
In Chapter 4, Jeong Min-ho, forty-five, was experiencing the hollowing-out of white-collar work. In this matrix he falls close to the second quadrant. AI is absorbing his work, but he is eight years younger than Lee Jung-hoon. Eight years makes a difference in transition capacity.
Even within the same quadrant, the coordinates are not identical.
This matrix is not a prescription. It is a diagnosis. A self-help book would say, "Change your positioning." But to tell someone placed in the second quadrant "move your coordinates" is also, simultaneously, a society that has not redesigned its structure transferring responsibility onto the individual.
Lee Jung-hoon's decision to board the plane for Hai Phong is an expression of courage — and at the same time the outcome of Korean society's failure to provide him with other options. It is not those who predict the future who survive. It is those who build their own narrative within uncertainty who turn to the next chapter.
In Chapter 6 we open the drawers of two people. Lee Jung-hoon and Kim Su-jin — the one displaced, and the one who crossed over. We ask whether the difference between them is a matter of individual capacity or structural condition.
Threshold Question: Among the things you did today, distinguish those that will still need to be done by a person next year from those that will not. If you cannot make the distinction — determine whether that is because of your own capability, or because no one yet knows the answer. That distinction is the hardest question right now.