Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Korea’s fertility rate went up. Two years running. Births in 2025 came in at 254,457, a 6.8 percent increase on the prior year, and the total fertility rate lifted from 0.75 to 0.80. Headlines followed. Officials made careful noises about a turning point. Somewhere in a ministry building, a paper is being drafted that says the policy finally worked.
It did not work. And understanding precisely why it did not work is the most valuable thing a global investor can extract from Korea right now, because the same arithmetic is queued up behind it across Italy, Spain, Japan, Taiwan, Poland and eventually China itself. Korea is not an outlier. Korea is simply first in the queue, running the experiment at speed, with the results published quarterly.
What the rebound actually is
Korea’s own health and welfare researchers have already done the decomposition, and the answer is unflattering. The Korea Institute for Health and Social Affairs attributed the increase mainly to demographic composition, not to behavioural change. There is a temporarily larger cohort of women in their early thirties passing through prime childbearing years. That bulge is expected to hold only until around 2030. Layered on top of it is a wave of weddings that were postponed during one of the strictest pandemic regimes in the developed world, released all at once. In a country where roughly 95 percent of births occur inside marriage, a marriage backlog mechanically becomes a birth backlog eighteen months later.
Neither of those things is a policy outcome. Both are timing effects with a known expiry date. The underlying question, which is whether a typical Korean woman in her thirties now expects to have more children than she did five years ago, has not moved in any direction that shows up in the data.
The cash did not do it either. Korea has spent somewhere north of 280 trillion won across sixteen years on baby bonuses, childcare subsidies, fertility treatment support and housing incentives. That is a very large number producing a fertility rate of 0.80 against a replacement level of 2.1. If you were running this as a capital allocation programme you would have shut it down a decade ago.
The cost curve nobody wants to price
Here is the number that explains the failure better than any survey. In 2025, total household spending on private education in Korea fell 5.7 percent to 27.5 trillion won, while monthly spending per participating student hit a record 604,000 won. The aggregate went down only because there were 120,000 fewer children in the system. Intensity per child went up.
That is the whole story compressed into one data point. The hagwon system is not a discretionary luxury that families trade off against a second child. It is a positional arms race, and positional arms races do not deflate when participants exit. They concentrate. Fewer children competing for the same narrow set of university slots means more spend per remaining child, not less.
A cash transfer of a few thousand dollars against a commitment measured in hundreds of thousands is not an incentive. It is a rounding error with a press release attached. Any government anywhere proposing to solve fertility with transfer payments should be made to explain why Korea’s result will not be theirs.
The automation thesis has a distribution problem
Korea holds the world’s highest robot density at 1,220 industrial robots per 10,000 manufacturing employees, growing around 7 percent annually since 2019. Second place, Singapore, sits at 818. Germany is at 449. This is not a plan. It is already built.
The optimistic framing says automation makes the demographic question moot. Output per worker rises fast enough that a shrinking workforce still funds an expanding retiree base. Korea is the cleanest live test of that proposition anywhere on earth, and the result so far is that productivity gains have accrued overwhelmingly to capital rather than to labour or to the tax base.
That is the part the abundance narrative skips. Robots do not pay contributions into a pay-as-you-go pension system. Their owners pay corporate tax at whatever rate the political system can enforce on globally mobile chaebol balance sheets. The engineering question was solved years ago. The distribution question has not even been seriously posed. Investors treating automation as a demographic hedge are assuming a fiscal transfer mechanism that does not currently exist in any advanced economy.
Immigration is going backwards, not forwards
The consensus assumption is that ageing rich countries will eventually be forced into large-scale immigration by economic necessity. Korea is running the counterexample.
The country’s main channel for lower-wage foreign labour is the E-9 visa under the Employment Permit System. The government cut the 2026 quota to 80,000, down from 130,000 in 2025 and 165,000 in 2024. That is a 52 percent reduction across two years, in a country with the developed world’s most severe labour supply problem. The stated reason was normalising post-pandemic demand.
[Also See: Five Vassal States America Can No Longer Protect]
Structurally, the E-9 is a rotation scheme rather than a settlement scheme. Terms run three to five years, family accompaniment is not permitted, and there is no default path to permanent status. Even at maximum throughput it adds workers temporarily without adding a population. To offset the projected contraction in the working-age base through migration alone, Korea would need to admit foreign residents and their families in numbers that would restructure the country’s ethnic composition within a generation. No electoral coalition anywhere in Korean politics supports that, and the quota trajectory tells you the direction of travel is the opposite.
Take the general lesson: economic necessity does not automatically generate political permission. Anyone modelling European or Japanese labour supply on the assumption that migration scales with need should look at what Korea did when the need was greatest.
The clock that actually matters
Forget the fertility rate for a moment and look at the fiscal calendar. Korea passed its first pension reform in nearly two decades in March 2025, lifting contributions from 9 percent to 13 percent by 2033 and the replacement rate from 40 to 43 percent. The effect was to push the National Pension fund’s depletion date from the mid-2050s to roughly 2064, or 2071 under an optimistic return assumption.
Set that against the OECD’s assessment that by 2064, Korea’s population aged 20 to 64 is projected to halve. The fund runs dry at almost exactly the moment the contributor base is at its thinnest. The reform bought a decade. It did not change the shape of the curve.
This is the real deadline for every optimistic thesis about Korea. Automation dividends, unification windfalls, migration reversals, fertility recoveries: all of them have to arrive and compound before roughly 2064 or they are irrelevant to the fiscal outcome. The 2025 reform did not solve the problem. It moved the invoice.
The export bet is narrowing to two companies
The final fallback is that Korea does not need people because it has champions. Sell more to the world, run the economy on Samsung, SK hynix, Hyundai and LG. That works right up until the champions face competition, and they now do, from both directions.
At the leading edge, Samsung Foundry holds roughly 6.5 percent of the global foundry market against TSMC above 70 percent, and its 2nm yields have been reported in the mid-50 percent range against a 60 percent threshold generally required for stable mass production. That gap compounds, because yield determines who wins the volume that funds the next node.
From below, CXMT has taken roughly 8 percent of global DRAM with first-quarter revenue growth above 700 percent, and is building toward high-bandwidth memory production. Korean incumbents still own the high-margin AI memory tier, and SK hynix in particular is having an extraordinary run. But the commodity base underneath that tier is exactly where a state-backed competitor with domestic demand and no return-on-capital constraint does its damage.
Concentration is the risk nobody discounts. When a national economy’s growth story reduces to two firms in one cyclical industry, the sovereign correlation stops being diversifiable.
What this means for capital
Korea will not disappear. It will get smaller, older, more indebted per worker, and considerably more dependent on alliances it cannot reciprocate militarily or demographically. That is a slow story, and slow stories get systematically mispriced because markets discount what happens in the next eight quarters, not the next four decades.
Three practical takeaways. First, treat any headline about fertility policy succeeding as noise until someone shows you the cohort decomposition. Second, in ageing economies the pension depletion date is a harder catalyst than the birth rate, because it forces a legislated tax or benefit event on a known schedule. Third, the countries currently telling themselves they will fix this later are watching, in real time, a wealthy and competent democracy try everything and fail. The evidence is in. The question is whether anybody trades on it before it is priced.
By 2050 the developed world will be a shrinking minority of global population and a shrinking minority of global consumption growth. Korea is showing what the transition looks like from the inside. The classrooms are the leading indicator. In 2026 the number of first graders fell below 300,000 for the first time, and more than 4,000 schools have already closed permanently. Those buildings are not coming back, and neither is the assumption of demographic expansion that the entire postwar asset pricing model was built on.
