South Korea's Birth Rate One Of the Lowest on Earth. The Missing Piece Might Be Better Urbanism
The macro shocks are coming. The local environment may decide who has children
On February 28, 2026, war between the United States, Israel, and Iran effectively shut the Strait of Hormuz. Brent crude surged past $82 intraday; Barclays warned of $100, JPMorgan modeled $120, Deutsche Bank’s worst case reached $200. By Monday, South Korea’s KOSPI had suffered its steepest single-day plunge in history, eclipsing the post-9/11 crash. An economy that grew just 1% in 2025 absorbed another blow at precisely the moment the Bank of Korea had projected a fragile recovery to 2%.
Almost nobody (or at least on Substack besides myself) is discussing what this shock will do to births.
The pattern is by pretty familiar for long time readers: 1973, 2008, 2022. Each energy & economic shock collapsed total maternity rates (% of women who become mothers) across affected economies while CPM barely moved. The most expensive pronatalist programs in Europe could not withstand two years of energy-driven inflation. Hungary’s TFR, raised from 1.23 to 1.59 over a decade, collapsed to 1.39 by 2024. The Czech Republic’s fell from 1.83 to 1.37, the lowest since records began in 1806.
South Korea enters the 2026 shock from the lowest baseline on Earth. Its TFR stands at 0.72 nationally, 0.55 in Seoul. More than half of Korean women are projected to remain childless. The government has spent upward of $280 billion on pronatalist policy since 2006: subsidized housing, paid parental leave, cash bonuses for newborns, government-sponsored dating events. The TFR continued to fall. One study found that 74% of baby bonus disbursements went to births that would have occurred anyway.
And at these levels, every shock cuts deeper. A 5-point TMR decline from 90% leaves 85% entering motherhood. The same decline from 48% leaves 43%.
Countries cannot control energy shocks. They can control the physical structure of their cities: how far the daycare is, how long the commute takes, whether a clinic is walkable when a toddler spikes a fever at midnight. These spatial relationships are not linear. They exhibit thresholds and cliffs that most pronatalist policy ignores entirely.
A recent study by Choi, Lee, Jung, and Sohn (2026), published in Cities, maps these thresholds across Seoul. Daycare within 400 meters (1/4 of a mile) is associated with higher births; the effect vanishes beyond. Commuting past 65 minutes turns sharply negative. Neighborhood clinics within 200 meters (1/8th of a mile) are positive. Large hospitals within 1.5 kilometers (~1 mile) are negative.
What the study says
Method in brief
Most prior studies of residential environments and fertility rely on density-based indicators (hospitals per 1,000 residents, daycare centers per district). These capture quantity but miss physical accessibility. An area might have high hospital density on paper, but if the hospitals cluster in one corner of the district, residents on the other side face long travel times and effectively low access.
The deeper problem is the assumption that these relationships are linear. They aren’t. Tamefuji (2010) found that maternal employment drops sharply beyond 15 minutes from childcare, not gradually, but at a cliff. Chen et al. (2022) documented similar distance-decay patterns for subway access, with the premium confined to roughly 500 meters to 1.2 kilometers. Smith Jr. et al. (2013) found the relationship can reverse entirely: excessive hospital proximity suppresses fertility through noise and congestion. The terrain is cliffs, plateaus, and reversals. Linear models can’t map it.
The Choi et al. study adopts an N-minute city framework, evaluating accessibility by travel time rather than facility counts, across five functional categories: working, supplying, caring, learning, and enjoying. Its unit of analysis is the 250m × 250m grid cell, a resolution that aligns with actual walkable distances and mitigates the Modifiable Areal Unit Problem inherent in irregular administrative boundaries. Birth counts, available only at the administrative level, are distributed to grid cells proportionally based on women aged 20–39 (who account for 92.8% of Korean births). Independent variables measure network distance along actual road paths to the nearest facility in each category, with sociodemographic controls including age diversity, elderly ratio, young adult ratio, marriages, and average monthly income.
Among models tested, XGBoost achieved an R² of 0.8578, substantially outperforming linear regression at 0.7827. SHAP was applied for interpretability: feature importance rankings, summary plots for effect direction, and dependence plots for threshold identification. Geographically Weighted Regression then captures how relationships shift across Seoul’s geography. Together, these tools reveal that relationships are non-linear, where the inflection points fall, and how they vary across space.
Shorter Commutes Means Higher Children Per Mother
Commuting time exhibits the clearest threshold in the entire analysis. SHAP values are positive in the 50–65 minute range, then turn sharply negative beyond 65 minutes. Not a gradual decline. A cliff, a point at which the daily time budget for parenting appears to become unviable. The GWR results add texture: the negative association is most pronounced in Seoul’s southwestern districts (Gangseo, Yeongdeungpo, Dongjak) and the steep-topography areas of Changsin-dong, where transit bottlenecks compound the commuting burden. For the 46% of Korean households that are dual-income, every minute beyond 65 is a fertility disincentive that no cash subsidy can easily offset.
People Love Living Near Trains
Subway proximity reveals a different problem. Within 220 meters of a station, SHAP values are positive: extreme convenience supports work-family balance. Between 220 meters and 1 kilometer, values turn negative. Beyond 1 kilometer, the effect converges to zero.
The mechanism is housing prices. Transit-adjacent areas in Seoul command steep premiums. Young families, newly married, carrying education debt, facing the enormous costs of private tutoring, are precisely the households priced out of the 220m-to-1km zone. For them, the housing premium overwhelms the commuting benefit. That’s a spatial equity problem masquerading as a transportation one.
Daycare, clinics, and the hospital paradox
Public daycare shows the sharpest single threshold: SHAP values strongly positive within 400 meters, dropping off sharply beyond. The GWR confirms this effect is especially pronounced near central business districts (Seoul Station, Gwanghwamun) where dense office complexes make walkable childcare a critical lifeline.
The policy implication cuts against current practice. Childcare policy in Seoul has focused on quantity: more facilities, more slots. The data say placement matters at least as much. A daycare center 600 meters away, however well-staffed, falls outside the threshold. For a parent rushing between a morning shift and evening pickup, those extra 200 meters are the difference between a schedule that works and one that doesn’t. But here’s the equity problem: areas where daycare, clinics, and schools cluster within walkable thresholds tend to overlap with higher-income neighborhoods. The 400-meter radius doesn’t just correlate with higher fertility. It correlates with higher housing prices, which price out the young families who would benefit most.
Neighborhood clinics tell a similar story at finer scale: positive SHAP values within 200 meters, diminishing toward zero beyond 400 meters. Infants and toddlers visit clinics an average of 6.5 times per year, two to three times the frequency of other age groups. A walkable clinic isn’t a luxury for these families. It’s a near-daily necessity.
Large hospitals present the opposite pattern, and it is one of the study’s most striking findings. SHAP values are negative within 1.5 kilometers, positive between 1.5 and 3.5 kilometers, declining again beyond. The GWR confirms this around major institutions: Seoul National University Hospital, the National Medical Center, Korea University Anam Hospital. The mechanism is straightforward: sirens, congestion, transient populations erode the residential tranquility families seek. Different scales of healthcare infrastructure relate differently to residential life.
Education, leisure, and null findings
Elementary schools within 820 meters and private academies within 200 meters is associated with higher fertility. A notable GWR finding: educational infrastructure shows largely insignificant spatial variation across Seoul, implying that schools have reached spatial saturation. Fertility effects come not from regional disparities in supply but from whether individual cells fall within the walkable thresholds. The relevant planning unit is the neighborhood, not the district.
Cultural facilities within 780 meters are associated with lower fertility. This isn’t counterintuitive once you consider where cultural facilities cluster in Seoul: intense commercial zones where noise, congestion, and nightlife dominate. In Hongdae and Hyehwa, youth-oriented entertainment districts, fertility rises with distance from the cultural core. These aren’t family amenities. They’re signals of an environment oriented toward childless urban lifestyles.
Parks showed no consistent directional trend. Restaurants and grocery stores had no significant effect. The study’s authors think it’s likely because South Korea’s advanced food delivery infrastructure (72.5% usage, reaching 84–87% among twentysomethings and thirtysomethings) has decoupled food access from physical proximity.
Sociodemographic patterns
The sociodemographic controls largely confirmed expected patterns. Income shows positive SHAP values above approximately 3.27 million KRW, close to the national average wage of 3.64 million KRW, consistent with risk theory: economic insecurity leads people to delay or forgo childbirth. Areas with greater age diversity (entropy index above 0.94) show higher fertility, consistent with research on intergenerational childcare support networks. Marriage is the single most powerful predictor of birth counts, unsurprising given that only 2% of Korean births occur outside marriage, compared to the OECD average of 40%. The study appropriately conducted a secondary SHAP analysis excluding marriage to reveal the residential environment’s independent contribution.
“Family-friendly” is a luxury good
This is where the analysis becomes most consequential.
The thresholds identified above (400 meters for daycare, 200 meters for clinics, 820 meters for schools, under 65 minutes for commuting) are not distributed equally across Seoul. Areas that meet them are desirable. Desirable neighborhoods command higher prices. Higher prices push out young families. The infrastructure gap between neighborhoods that meet these thresholds and those that don’t isn’t just inconvenient. It’s a mechanism driving housing price differentiation, and through that, reproductive inequality.
In a city where the price-to-income ratio stands at 13.9 years, where average apartment prices have surpassed $1 million, where 82.6% of young people rent, this cycle is the operating reality for millions. A “fertility-friendly environment” has become a commodity accessible only to those who can already afford it.
The $280 billion in pronatalist spending has targeted reducing the private costs of children for individual families. Almost none has been directed at ensuring that the public infrastructure associated with higher fertility (walkable daycare, accessible clinics, manageable commutes) is equitably distributed across income levels and neighborhoods. The money targeted the wrong margin.
What this means for urban policy
The study recommendations below is grounded in an empirically identified threshold.
Childcare placement over quantity. The current approach increases facility counts per district. The evidence argues for a different standard: ensure every residential area falls within 400 meters of a public daycare center. Walkable distance, not per-capita ratio. This almost certainly requires more facilities in underserved neighborhoods rather than additional slots in areas already within the threshold.
The 65-minute commuting constraint. Land-use and transit planning should treat this as a hard ceiling for residential zones targeted at young families. This means integrating childcare and affordable housing into transit-accessible nodes, not merely building transit lines and hoping families benefit. The subway paradox underscores the point: access to transit is necessary but not sufficient. It must be paired with affordability measures that prevent young families from being priced out of the very zones where the commuting benefit exists.
Differentiated healthcare proximity. Neighborhood clinics should be sited within 200 meters of residential concentrations. Large hospitals require buffer planning, a recognition that their negative externalities within 1.5 kilometers may discourage family formation. Not an argument against hospitals. An argument for understanding that institutional scale matters for residential environments.
Threshold-based school placement. Educational infrastructure is spatially saturated in Seoul; fertility effects are driven by micro-level proximity, not district-level supply. Future placement should ensure individual neighborhoods fall within 820 meters of an elementary school and 200 meters of supplementary educational facilities.
Caveats
This study identifies associations, not causal effects. Families with children may sort into neighborhoods with better infrastructure, rather than better infrastructure causing higher fertility. Disentangling selection from treatment would require longitudinal data tracking neighborhoods before and after infrastructure changes, data not yet available at this spatial resolution. The dasymetric mapping used to distribute births to grid cells improves on raw administrative data but inevitably introduces estimation error. And the nearest-neighbor metric assumes residents use only the closest facility; a neighborhood with three daycare centers within 600 meters might offer more effective accessibility than one center at 350 meters. Future research using Two-Step Floating Catchment Area methods could capture these dynamics.
Several features of the Korean context (the 2% non-marital birth rate, the dominance of private academies, the delivery infrastructure that decouples food access from proximity) limit direct generalizability. But the structural dynamics (threshold effects in walkable accessibility, the tension between transit proximity and housing affordability, disamenity effects of large-scale institutions) are likely to operate in any high-density city where young families face time poverty and financial constraint. The specific numbers will differ. The logic should travel.
Getting the questions right
South Korea (or most countries) cannot close the Strait of Hormuz. It cannot control oil prices or global inflation. What it can control is whether the physical environment makes parenthood logistically survivable for households on the fence.
Every threshold in this study is a policy lever. Every one is currently left to the market, which prices young families out of the neighborhoods where the thresholds are met and traps them in neighborhoods where they aren’t. The evidence from Seoul tells us where to start. The evidence from 1973, 2008, 2022, and now 2026 tells us we are running out of time.