Urbanism and YIMBYism outside big cities do not mean towers. What it does mean is neighborhoods dense enough to walk and oftentimes with enough destinations close enough together that a golf cart covers the rest for day-to-day, and better yet, pleasant enough that if you need a car, it could be a japanese style kei car or truck.

Golf carts? Well, being able to travel around town is so nice. So nice that we have places like Peachtree City, Georgia, and The Villages in Florida becoming popular because they are golf cart-friendly. In Peachtree, roughly eleven thousand registered golf carts carry its residents along more than a hundred miles of paths to the school and the store. The Villages is a 55+ master-planned community (with a big old waitlist) that is purpose-built for carts, featuring over 100 miles of dedicated multi-modal paths, tunnels, and bridges that connect homes to town squares, shopping, and medical centers.

These particular towns aren’t exactly what YIMBYs are selling but a (very) rough version of the thing. They are low-density and master-planned sprawl (one of them age-restricted). The golf cart is a fun way to get around, and they have enough foresight to plan for them. That said, it is still incredibly difficult and costly to plan around “golf carts,” let alone walkability with zoning and minimums on the books. Not to mention, the ride to the store still runs ten or fifteen minutes. The destinations still sit far apart, and the cart was added to cover the gap. We also have new developments like the Culdesac that’s based on walkability, not golf carts, with the biggest limit being how states and cities limit and restrict any sort of development.

The appetite is fifty years deep and almost entirely unmet. The product that meets it is the same legal freedom built at real density inside a city that already exists, where the destinations sit close together and the ten minute ride becomes three. Done right, the cart stops being a substitute for the car and becomes a small daily delight on top of a neighborhood you could also just walk.

One legal permission separates that product from getting built, and the movement is fighting for it in the wrong places. The fight runs from San Francisco and even to supposedly “technocratic” places like Berlin, where prices broke long before the movement existed. The places where the supply problem is now most fixable are the small and mid-size cities where demand is arriving, where land is cheap, and where the political opposition is real but not yet entrenched. The movement is not there. The demand already is.

Limited Frames

The U.S. housing crisis is routinely framed as a supply failure in the places where prices are already highest. That framing is not wrong on its own terms. The biggest issue it is very limited and based on where a lot of YIMBYs and urbanists live in.

I would argue that the secondary problem after NIMBYism is geographic demand concentration. Gyourko’s Superstar Cities identified the dynamic. A handful of metros combine inelastic land supply with a rising national share of high-income households, producing a bidding-up of land prices that sorts lower-income households out. Ganong and Shoag’s Why Has Regional Income Convergence in the U.S. Declined? supplied the consequence. From 1880 to 1980, U.S. state incomes converged at 1.8% per year. After 1980, the convergence collapsed. The mechanism is housing cost. Tightening land-use regulation in the highest-wage metros raised housing costs to the point where, after housing, the returns to migration for low-skilled workers turned negative.

Some YIMBYs’ response is to build more where the rich already are. Yes, I have written articles where I celebrate removing zoning restrictions in rich enclaves with utter glee. I do very much believe in getting rid of zoning, and even mid- and small-sized cities can benefit from this. I don’t share the idea that the concentration is fixed and focusing on economic of scales is the way to go. People love brands and having choices at the end of the day. A “monopoly” of a few superstar cities has shown that it is dangerous, especially if governance in said superstars is less than ideal. My home city of Houston is more affordable than most, isn’t because of Houston’s mayors. In fact, Houston mayors have historically tried to make housing more expensive via trying to implement zoning, which is defeated by local groups.

So, the alternative is to treat the concentration as a variable and change it by strengthening other small to mid-size brands/cities and promoting YIMBYism and urbanism.

The right to build is itself a product to sell, not an ideology

Conventional housing-policy frame treats zoning as a constraint on construction. Build new units and supply rises, thus so prices fall. The frame is correct but again a very incomplete and limited way of thinking. Legal entitlement to build is itself an asset whose value capitalizes into land prices the moment the entitlement changes, regardless of whether anyone builds anything.

Capozza and Helsley laid out the theory in two foundational papers. “The Fundamentals of Land Prices and Urban Growth” in the 1989 Journal of Urban Economics showed that in rapidly growing cities, the value of expected future rent increases (the growth premium,) can account for half of total land price. “The Stochastic City” added the option-value insight. Uncertainty about future rents imparts an option value to undeveloped land independent of any planned development. Glaeser and Gyourko’s “Zoning’s Steep Price” inverted this framing.

Gap between house prices and marginal construction costs is the zoning tax. It is the implicit price of the right to build that the regulatory system has confiscated. The zoning tax exceeded 50% of housing costs in San Francisco by the early 2000s, and Gyourko and Krimmel’s follow-up shows it is now moderate-to-high in nearly every American central city for which data exist.

This isn’t just a California thing. Greenaway-McGrevy and Phillips’s evaluation of Auckland’s 2016 Unitary Plan in Regional Science and Urban Economics in 2024 found land-price gradients shifting in upzoned versus non-upzoned areas consistent with a 23.7% increase in equilibrium floorspace, capitalized into land prices the moment the plan was announced. Cheung’s Urban Studies paper reconfirms this. Parcels where land dominated total value gained after upzoning, while parcels where the existing structure dominated lost relative price, because the new entitlement made redevelopment more likely.

We see this with California’s accessory dwelling unit reforms. Beginning with AB 2299 and SB 1069 in 2016, California systematically stripped local barriers to ADU construction on single-family parcels. The FHFA’s tracking of Enterprise-backed appraisals shows the share of California appraisals with an ADU rose from 1.8% in 2016 to 3.0% in 2023. Industry appraisal data shows ADU-equipped properties trading at 25-50% premiums to comparable non-ADU homes, with Los Angeles markets reaching 58%. UCLA’s Lewis Center has an ongoing parcel-level analysis explicitly designed to “isolate the impacts of ADU legalization (i.e., the option value created for existing parcels) on land values.” Every single-family lot in California gained the right to host a second dwelling unit and that right capitalized into land values before most owners exercised it.

São Paulo is another example. The city’s Certificates of Additional Construction Potential (CEPACs) are tradeable bonds granting the holder the right to build above the baseline floor-area ratio in designated zones. Between 2004 and 2009, São Paulo raised approximately $812 million by auctioning CEPACs. It’s not a policy I would prefer, but push comes to shove, development rights became a financial instrument with a quoted market price if we have to.

“Zoning reform produces new housing” is true and slow. “Zoning reform unlocks value already in your land but legally suppressed” is also true and fast. Value is a very subjective thing. It could show up in the comparable sales the week the plan is gazetted. It could show up in the fact that building an ADU will let the homeowner bring in their aging parents or just to rent out. That is a proposition you sell to a landowner, and there are many others. The affordability case turns on what happens next.

Yes, building more lowers prices

The NIMBY counter-argument arrives in two reliable examples.

The first is the denial of supply on prices aka supply skepticism. It is the claim that building more housing just attracts more demand, so prices won’t fall. Most YIMBYs are familiar with this.

The second example, called spillover skepticism, isn’t talked about as much. It is the claim that building in Buffalo cannot help San Francisco even if reform works locally.

Both objections are wrong. We many papers like the 2024 Furman Center Supply Skepticism Revisited synthesis gathered both theory and the evidence time and time again.

Locally, new construction lowers prices in the immediate neighborhood. Kate Pennington’s San Francisco analysis finds rents fall 2% for parcels within 100 meters of new construction, decaying to zero within 1.5 kilometers. Xiaodi Li’s New York study finds that for every 10% local stock increase, rents decrease 1% and sale prices decline within 500 feet. The “luxury housing causes gentrification” claim is the inverse of what the data show.

Within a metropolitan area, new construction creates moving chains that benefit lower-income households. Evan Mast’s JUE moving-chains paper found that 100 new market-rate apartments in 12 major American metros produced 45 to 70 openings in below-median-income neighborhoods within five years, with 19 to 39 of those openings reaching the bottom income quintile. Bratu, Harjunen, and Saarimaa’s Helsinki replication and Mense’s 2025 JPE Macro paper confirm the mechanism. Most demand redistribution happens within metropolitan footprints, which means the suburb is doing more supply work than the rival metro.

Across metros, the substitution channel operates at the national scale. Mondragon and Wieland’s Housing Demand and Remote Work estimates that an additional percentage point of remote work causes a 0.92% increase in local house prices after controlling for migration spillovers. The Bush Institute and SMU Build Homes, Expand Opportunity report runs the counterfactual. If all 250 largest U.S. metros had adopted Sunbelt-Mountain housing policies, 5.6 million more homes would have been built between 2010 and 2023, average home prices would be $115,000 lower, and rents would be $450 lower.

Let’s Talk About Houston

Houston is the only major American city without conventional use-based zoning. Three citywide referenda failed, but that hasn’t stopped the city from enforcing deed restrictions and minimum lot sizes. Still, despite the best efforts there is no zoning map. Both local groups and developers pushing, in 1998, Houston cut the minimum lot size within the I-610 Loop from 5,000 square feet to 1,400 square feet. They also removed the open-space and setback requirements that had made small-lot construction infeasible, and extended the reform citywide in 2013. Mercatus Center analyses document nearly 80,000 small-lot houses built since 1998. Not to mention more dense SFH areas with 34,000 detached townhomes added between 2007 and 2020 on lots that would have been illegal in any other major American city.

Just to give you idea on how much a superstar Houston is, in 2024, Houston issued 65,747 residential building permits. That’s the second-most of any U.S. metro. Texas as a whole issued 225,756 permits, more than any other state. The Bay Area, by contrast, issued 9,100 permits across all nine counties. Houston permitted more than seven times what the Bay Area permitted, on a population base roughly twice the size. Manhattan Institute analysis puts the Houston metropolitan median house-price-to-income ratio at 4.7.

It is worth noting that Houston does have its problems. It does have rising rents and a serious affordable-housing shortage at the lowest income tiers. Other cities have it worse. The reason I made the point is Houston’s model is not “housing is cheap.” It is “housing is buildable, so the price spiral does not run away.” The within-metro filtering mechanism documented by Rosenthal, Mast, Bratu et al., and Mense supplies the distributional payoff. The chain runs because the chain has units to consume.

Houston also shows what kind of reform wins. The conventional planning instinct is to find the optimal middle, a little zoning and a little density, both held back. Houston went the other way and deregulated use almost entirely, abolishing the rule dictating what you can put on a lot. The cities that tried to find a defensible middle produced the most disappointing outcomes. The optimal middle does not exist in housing policy.

A caveat about transferability deserves naming. Houston’s no-zoning baseline is path-dependent on three failed referenda, Texas property-rights doctrine, weak local government, and the absence of a state income tax. None of that package transfers cleanly to Ohio or Michigan or upstate New York. What transfers is the moves, meaning the lot-size cut, the by-right approval for small-lot construction, and the removal of open-space and setback rules. Texas proves the point inside its own borders, because the 2025 reform package kept zoning but reformed it in the same operational direction.

Texas Triangle and the Sunbelt

We all know about Houston. Why are we talking about Houston? It is the largest city in the Texas Triangle. DFW issued 47,045 single-family permits in 2024, Austin issued 16,521, and San Antonio issued 10,916. You are starting to see the structural difference between Texas and California. California has two superstar metros that absorb most of the in-migration and price most of the population out. Texas runs four mid-market metros, not to mention the many smaller markets inside and surrounding the Triangle. None of the Texas Triangle cities are really “dominant”. So, that lets prices set where wage-earning households can still buy.

What happens if that balance breaks? Well, we see the Texas system starts to revert to the superstar dynamic. Austin’s 2020-2022 absorption is the best-documented example. Home prices rose nearly 60% in 24 months, and smaller Texas cities including College Station absorbed demand shocks of 34% to 57%. By 2026, Austin has flipped to outmigration. Home prices are retreating from the former superstar, and inventory is recovering faster than any major U.S. metro. The demand reversal was not Austin’s doing. Higher interest rates and energy shocks are usual suspects. It didn’t help that the destruction of the hiring rates for white-collar professionals also helped cool every hot market in the country. Where supply is inelastic, a demand reversal shows up as frozen sales and prices that refuse to move. Where the building infrastructure already exists, it shows up as falling prices and recovering inventory.

In 2025, the Texas Legislature passed a bipartisan housing reform package explicitly modeled on the Houston template. SB 15 preempts minimum lot sizes in large cities, SB 840 authorizes multifamily by right on commercially-zoned land, HB 24 limits NIMBY protest power, and SB 2477 fast-tracks office-to-housing conversions. The coalition spanned business associations, free-market groups, faith-based organizations, racial and ethnic minority groups, and progressive housing advocates.

What’s in it for Cleveland and Atlanta

The empirical case for supply is settled. The political case is not, and it fails for a specific reason. The usual YIMBY argument does not explain why a Cleveland homeowner would vote for zoning reform when their property values are stagnant and their tax base is hollowing out. The option-value-of-buildability framework answers exactly that, and it is most powerful where it has been least applied.

The demand is already arriving. Zillow’s 2024 hottest-markets list put Buffalo first, with Cincinnati, Columbus, Indianapolis, and Providence rounding out the top five. Realtor.com’s 2026 forecast had Hartford, Rochester, Worcester, Toledo, and Providence as the top five “refuge markets” for buyers fleeing high-cost metros. The Philadelphia Fed’s great-reshuffle paper documents the welfare accounting. Originating cities saw housing-cost relief, and receiving markets absorbed the income shock. Hartford, Providence, and Chicago remain 55-76% below pre-pandemic inventory levels. As Austin showed, the supply infrastructure has to be in place before the demand arrives, not after.

Homeowners can be allies or adversaries. Either case, homeowners can still be customers. A homeowner in Cleveland watching their property value stagnate has the opposite economic interest from a San Francisco homeowner with seven-figure equity. Capozza-Helsley showed that the growth premium can account for half of urban land price, scaling with expected future rent growth. Buffalo, Cincinnati, Columbus, Indianapolis, and the recovering Connecticut and Rhode Island metros have demand fundamentals that produce larger option-value gains than Detroit, Toledo proper, or Cleveland city center, where the premium is thinner. The sign is positive in every case, and only the magnitude varies. Apply the mechanism to a Cleveland parcel assessed at $80,000 in a tract zoned exclusively for single-family use. The right to build a duplex, a triplex, or an ADU raises its value the day the entitlement is granted, and the homeowner who voted YES becomes wealthier. Even if the homeowner voted NO, that homeowner is still a customer. Often times, as we see with the ADU reform in California, NIMBY homeowners can and will build their own ADUs and other housing if it benefits them.

The obvious objection is that this cuts against affordability. If the right to build raises land prices, hasn’t housing just gotten more expensive? The resolution is in who gets what, and when. The option capitalizes into the owner’s land the day it is granted, but it pays out in cash only when someone exercises it, and exercised across thousands of parcels it produces the units that run the filtering chains. The owner is paid for the option. The renter is paid in the housing the option eventually builds. The two conflict only when the entitlement is scarce. Spot-upzoning one block while everything around it stays frozen lets capitalization run ahead of construction, because the lucky parcel becomes rare and worth hoarding. Broad, by-right upzoning does the opposite. It spreads the option value thin enough that holding out for the rezoning lottery stops making sense, and it maximizes the parcels that actually build. “Build everywhere” is not only fairer than “build on the rezoned block.” It is the configuration under which the land-price effect resolves into housing instead of into windfall. The renter in Cleveland is the person the chain is for.

The fiscal multiplier is largest for cities currently trapped. The National League of Cities’ fiscal conditions datashows cities with growing populations maintain a per capita ending balance of $566, against $438 in cities with declining populations. The mechanism is visible in Carmel, Indiana, whose mixed-use district generates land assessment values 20 times greater per acre than comparable single-use development, per Redevelopment Director Henry Mestetsky’s presentation to City Council. Carmel’s property tax rate is fifth-lowest in Indiana at $2.0354 per $100 of assessed valuation, and yet its high tax bills, a median $3,804 on a median home value of $455,500, reflect the assessed values the bundle produced rather than the rates. The commercial base expanded so far that low rates still funded the residential side. A city at the bottom of the assessment curve, which is to say Cleveland, gets there by allowing the parcel to be something other than a single-family home.

Atlanta is the in-between case. A high-growth Sunbelt metro with emerging affordability pressure, BeltLine-driven appreciation, and the beginning of zoning-reform discussion under “ATL Zoning 2.0,” but no Houston template installed. The political question is whether it adopts the Texas package now, while the coalition for affordability still exists, or waits until it has become San Francisco. Charlotte, Raleigh, Provo, and Greenville are running the Houston playbook. Atlanta has not yet.

There is a system-level case too. Braess’s paradox shows that adding capacity to a network and routing everyone through it can degrade performance for all users. The Hsieh-Moretti prescription of routing every worker to the highest-productivity metros is the housing-policy version of the same error, and the distributed answer scores higher on resilience and on the welfare of the median household.

The Rust Belt YIMBY argument is structurally simpler than the coastal version. Your property value rises, your tax base widens, and the affordability problem solves itself before the city becomes Austin in 2021. None of that is available in San Francisco. All of it is available in Cleveland, Buffalo, Toledo, Pittsburgh, Cincinnati, and Detroit.

Implementation bugs galore!

My position is maximalist one, get rid of zoning. That doesn’t mean I am not aware of the side effects it could cause. Thankfully, issues with implementation of zoning reform are documented, and each has a known patch.

Displacement. High-income remote workers arrive in secondary markets and bid up existing single-family inventory, pricing out incumbents before reform-enabled supply can respond. The patch is closing the window between “old restrictions removed” and “new supply online” through pre-approved building typologies and by-right approval. Minneapolis 2040 produced 16-34% lower home prices and 17-34% lower rents (Gu and Munro 2025) without a construction boom, through the demand-side effect of a credible reform announcement. But an announcement is a promise the supply has to keep. Minneapolis got the announcement right and the follow-through wrong, which is why it shows up twice in this playbook.

Backlash. Auburn, Maine’s Mayor Jason Levesque earned the title “YIMBYest City in America” and then lost reelection 62% to 38% in November 2023 to a deputy police chief who organized the opposition. Levesque’s own self-diagnosis names the failure mode in one line. He had “campaigned for the people who want to live in Auburn, not for the people who live in Auburn.” The patch is a designed consent structure that delivers the upside to the incumbent at the parcel level. Israel’s TAMA 38 program is the applied case. More than 100,000 apartments were reinforced between 2005 and 2024 because each building voted on its own reconstruction and the owners received the upside directly, an elevator, a balcony, a seismic retrofit, and 20-40% property-value uplift. It made the incumbent the primary beneficiary by construction, not by the hope that option value would capitalize after the fact.

Partial-implementation. Minneapolis 2040 eliminated single-family-only zoning in 2019, and the reform was widely framed as solving the binding constraint. It did not. The City Council explicitly directed staff to keep height, setback, and lot-coverage regulations closely matching existing single-family requirements. The Minneapolis Fed’s tracker shows only 46 new two-to-four-unit buildings permitted from 2020 to 2022 citywide, with just 20 in formerly single-family zones. Justin Fox’s analysis puts the end of single-family zoning at roughly 1% of new units permitted since January 2020. Eliminating the use category was the easier political move, and the built-form rules were the binding constraint left in place. The fix is to change the binding constraint, not the label. The Texas 2025 package is the operational template, with minimum lot sizes (SB 15), commercial-zone multifamily by right (SB 840), protest limits (HB 24), and conversion fast-tracking (SB 2477) shipped at once.

State preemption is still limited. Oregon’s HB 2001 in 2019 was the first statewide elimination of single-family-only zoning in towns above 10,000, and by the end of 2022 all 56 affected jurisdictions had complied, because the state coalition had been built across multiple prior legislative sessions. When West Linn and other suburban councils searched for workarounds. The coalition had the capacity to defend the law in each one. Auckland did not have that. The mandate arrived from above with no local advocacy infrastructure to defend it. Hence, the mandate had no constituency to hold it and keep the local town to account.

My point being?

Urbanism and YIMBYism are fundamentally products to be sold to towns and cities like SAP selling ERPs to enterprise. Different firms will use SAP’s systems in different ways. Same with cities and towns. Most likely way they are going to use YIMBYism and Urbanism is that golf cart community I described.

Then Textbook (a collection of local and state capacity from school systems) died. Quietly, the way most beautiful and doomed ideas do. No funeral, no eulogy, just a slow exhale. And rather than weep into my velvet shirt, I did the unthinkable!

I learned something. I took those lessons, pressed them like dead flowers between the pages of my own ambition, and have spent my idle hours building something. Something neat. Something that might even work, God help us all.

Here is a sneak peak! I hope you will enjoy the news when it comes out! The only thing I can guarantee is how gorgeous it looks.

Tencent’s then-president Liu Chiping told markets the company would “emphasize optimization” of costs (like personnel) in 2022. Chinese tech firms put hundreds of thousands of formal-sector white-collar workers out of work over the following three years. The Labour Contract Law was in force throughout (even with its grounds requirement and an active arbitration system). None of it stopped Tencent from performing layoffs.

“Socialist” California’s big tech firms cut entry-level hiring by more than 50% from 2019 to 2025 and ran successive layoff waves under a “notionally” opposite regime. California has at-will employment with no statutory severance and no just-cause standard. Google and Meta paid roughly 4 to 6 months of pay plus stock acceleration for the initial waves. Later waves of layoffs were not as generous to those unlucky souls.

You want to ask “Where were employers not firing anyone?” In Japan, the 2025 active job-openings-to-applicants ratio was 1.22 overall and 1.50 in manufacturing with unemployment held at 2.5% (some people don’t like that for some reason). Firms that wanted to reduce headcount did not invoke the four-requirement test for redundancy dismissal. They ran kibō-taishoku programmes (voluntary retirement at premium prices with 12 to 36 months of base pay on top of statutory retirement).

Tokyo Shoko Research found listed companies announced programmes covering 19,009 employees in 2024 (tripled from 2023).

What matters is whether local demand exceeds supply. When it does, employers cannot easily fire because they cannot easily replace (and the threat of layoffs loses its edge if people can find another job). When demand collapses you see that statutory protection becomes more like statutory guidelines (and, like all guidelines, employers find ways around it or just ignore them).

Hold up a moment? Earlier I imply that it is harder to fire someone in China than in California? Yes, with a heaping helping of caveats! Chinese employers face the Labour Contract Law‘s grounds requirement. This isn’t paper-only law the government quietly declines to enforce. The SPC’s August 2025 interpretation closed a common evasion route, and recent court rulings have invalidated performance-improvement-plan dismissals. That said, Chinese employers work around the formal regime through “optimisation” and pay N+1 as part of their “release”. SMEs in China often try to ignore the law and risk lawsuits.

You see incidents like the Jiangsu Suchun Liquor reorganization (SPC Guiding Case No. 164). A Chinese court, citing the local Party committee’s need to preserve jobs and prevent “mass incidents,” let an outside investor restart a bankrupt distillery’s production before formally taking it over. Essentially using bankruptcy law to keep workers employed.

Over the Pacific pond, Californian employers don’t really have any such requirements. Larger firms might pay out severance to reduce the likelihood of a wrongful-termination litigation. As a job market gets worse, severance packages often drop despite the threat of litigation.

You can guess whether the bottom half of each labour market has any formal protections. Despite the legal risks, a lot of Chinese SMEs ignore the Labour Contract Law, and 200 to 240 million flexible workers sit outside it entirely. Japanese non-regular workers (36.2% of the workforce) face yatoi-dome (the non-renewal of fixed-term contracts) as a normal exit mechanism. California employers fire at will and pay nothing in most cases.

You can guess what happens to young workers. China and California are both in genuine new-graduate hiring crises driven by different mechanisms. Japan’s labour market is tight. The protection that matters most for a 22-year-old graduate in 2026 is not Article 39 of the LCL or Tameny v. Atlantic Richfield. It is whether anyone is hiring.

Labour demand is the bedrock of protection

NBS reported a 16-24 urban unemployment rate (excluding students) of 18.9% in August 2025. It’s the highest since the methodology was revised in December 2023. Per China’s Ministry of Education data reported by The Economic Observer, the 2025 graduating class numbered 12.22 million and four consecutive cohorts have exceeded 10 million. A Zhilian Zhaopin survey put the 2024 graduate employment rate at 55.5% (down from 57.6% the year before). Zhang Dandan at Peking University (in a July 2023 Caixin article since removed from the original site) decomposed the cohort to argue the upper-bound rate could be as high as 46.5% if “lying flat” and “full-time children” populations are included.

California’s situation is hidden by the headline numbers. The state’s overall 5.5% unemployment rate (December 2025, per EDD) does not reflect the new-graduate cohort, which faces a different market. The Federal Reserve Bank of New York’s recent-graduate dashboard shows unemployment for ages 22 to 27 at 5.3% in Q2 2025, 5.7% in Q1 2026, which isn’t too bad. Then you see underemployment at 41.5%. For ages 20 to 24 with a bachelor’s or higher, BLS data put unemployment at 9.5% in September 2025. Recent graduates now run higher unemployment than the general workforce, reversing a thirty-year pattern. Oxford Economics attributes 85% of the rise in US unemployment since mid-2023 to new labour-market entrants. Handshake job postings for new grads fell 15 to 16% year-over-year and applications per posting rose 26 to 30%. Cengage Group found only 30% of 2025 graduates secured full-time jobs in their field.

Back to Japan, new graduates face the opposite condition. Recruit Career Future Research Institute’s Class of 2026 process survey found 39.3% of the cohort already held at least one informal offer by February 2025 (thirteen months before graduation). By December 1, 2025, the figure was 94.8%. JILPT’s official report put the Class of 2025 naitei rate at 92.6% (the highest since records began in 1999).

A new graduate in Japan is more than ten times as likely to have a job offer at graduation as a Chinese counterpart and substantially more likely than a Californian. Demographics, sectoral composition, and AI exposure explain almost all of it.

Dismissal regimes converge in practice

China’s Labour Contract Law permits unilateral termination only on “justified” grounds (Articles 39, 40, 41). Lawful no-fault terminations require N months of severance (one month per year of service). Unlawful termination requires 2x N (2N). Workers must be converted into open-ended contracts after ten years or two consecutive fixed-term renewals.

Japan’s Labour Contract Act Article 16 voids any dismissal “lacking objectively reasonable grounds and not socially acceptable.” Japan’s seiri-kaiko yon-yōken (four requirements) makes forced layoffs nearly unwinnable in court. The doctrine is set out in the Nihon no Jinjibu legal encyclopedia.

California (famously antibusiness) is fully at-will. The Labour Code §2922 reads, “An employment, having no specified term, may be terminated at the will of either party on notice to the other.” California’s only real exceptions are anti-discrimination protections.

How employers actually dismiss

China’s standard for white-collar dismissal is youhua (优化, “optimisation”). It’s a neat little euphemism for converting an unlawful termination into a mutually agreed departure under Article 36. The playbook (corroborated by Gank Interview’s January 2026 walkthrough) runs as follows. The employer issues a performance improvement plan with targets the employee is expected to fail. Once the employee fails, the company uses documented failure as Article 40(2) “inability to perform.” Most employees accept N+1 rather than pursue 2N through arbitration.

How can Japan get rid of headcount when the executives feel like it? Well, they have their own system, called kibō-taishoku (voluntary retirement). It’s still a lot more consensual than the Chinese and California examples. By mid-November 2025, Tokyo Shoko Research recorded 41 companies covering 11,045 employees. The Nikkei‘s analysiscalled it “the highest level since just after the Lehman Shock in 2009.” The more worrying trend is that the participants are not failing companies. The Nikkei now uses the term kuroji risutora (黒字リストラ, “profitable-company restructuring”).

When workers refuse the package, what happens? Employers use oidashi-beya (追い出し部屋, “chasing-out rooms”). The practice exists “to circumvent regulations preventing easy dismissal of permanent employees or to avoid paying enhanced retirement pay.”

California’s at-will baseline is heavily modified in the other direction. Per VentureBeat’s tracker and CNBC’s compilation, Google and Meta both paid 16 weeks base plus 2 weeks per year of service in the 2022 to 2025 layoff wave. A five-year mid-level engineer gets roughly 24 to 26 weeks. Employers pay any sort of severance is reducing chance of wrongful-termination litigation. Offering severance for a release is cheaper than a lawsuit. Especially, if the layoffs come at a suspicious time. As time goes on, Big Tech has been offering less and less severance per round.

Why the courts will not let employers exit the formal regime

It looks like these laws are nothing more than paper (with the exception of Japan of course). That’s not the whole truth. Japan is an aggressive case when it comes to enforcement but some cases are in the defense of the SMEs. On the other hand, Chinese and even Californian courts have started pushing back on the evasion mechanisms too.

China SPC’s Judicial Interpretation II (August 2025, effective September 1) directly attacks the standard playbook. Per the Jingtian Gongcheng law firm’s clause-by-clause analysis, “voluntary” social-insurance waivers are void and over-broad non-compete clauses on ordinary positions are void. The practice of “mixed employment” through affiliated companies triggers consolidated treatment of years of service. The Shanghai High People’s Court has invalidated PIP-driven dismissals where the employer skipped the required targeted training and transfer attempt. The SPC 32nd Batch Guiding Cases hold that termination grounds are locked in at the time of notice.

The Toyo Sanso case (Tokyo High Court, 1979) in Japan established the four-requirement framework still applied today. The Nihon Shokuen Seizō Supreme Court decision (1975) established the abuse-of-rights doctrine, codified verbatim into Article 16 of the Labour Contract Act in 2007. Voluntary retirement at premium prices exists because forced redundancy almost never survives judicial review.

California’s Supreme Court has slowly built four exception structures over four decades against at-will employment. Tameny v. Atlantic Richfield (1980) created a tort claim for wrongful discharge in violation of public policy. Foley v. Interactive Data (1988) recognized implied-in-fact contracts from handbook language and longevity. Guz v. Bechtel (2000) consolidated this framework. Green v. Ralee (1998) expanded public-policy sources to administrative regulations. At-will is still the law of the land in California. It’s just that the case law for litigation (that will undermine at-will) just keeps building.

The bottom half is easy to fire in all three places

What happens if you don’t have the time or the resources to get legal protections in China or Japan? Let alone take advantage of case law in California? The formal regime is simply inapplicable or unenforced, and dismissal is effectively at-will.

China, 200 to 240 million workers outside the Labour Contract Law

Per the State Council’s December 2025 report to the NPC Standing Committee, flexibly employed workers (think food delivery and rideshare apps) exceed 200 million. Caixin and Jinan University’s IESR put the figure at 240 million as of end-2024. As Qiushi puts it, “Some platform companies, in order to reduce labor costs, use franchise, agency, or contracting arrangements to lengthen the employment chain, or have workers register as individual industrial and commercial households, blurring the labor relationship.” Only 70.57 million flexible workers were enrolled in basic urban-employee pension insurance (roughly 30% coverage). These workers can be “fired” by app deactivation.

Small-firm Chinese employers routinely ignore the Labour Contract Law and take the risk. Migrant workers without proper contracts (which is more likely the case than not) are paid in cash and dismissed verbally with no recourse.

21 million non-regular Japanese workers face yatoi-dome

Per Q4 2025 labour force data, 21.26 million of 58.66 million employed workers are non-regular (36.2%). The share rose from 31.4% in 2004 to 37.1% in 2023 per MHLW data. Yatoi-dome (雇止め) is the term for non-renewal of fixed-term contracts. It is functionally a dismissal but is not legally treated as one. The Cabinet Office’s OECD EPL analysis shows Japan’s regular/non-regular protection gap is the 6th-largest in the OECD and has widened since the 1990s.

Japanese SMEs face an additional informal mechanism. Per the Nihon no Jinjibu encyclopedia, SMEs “face practical difficulty satisfying the four redundancy requirements” because they lack reassignment options and temporary-layoff capacity. Courts respond with “totality of circumstances” review which makes informal pressure more workable for SMEs.

California, at-will with no litigation tail at small firms

California’s at-will baseline is the default for almost everyone outside large firms. The wrongful-termination doctrines requires protected-class or protected-activity violation or an employer with enough assets that a contingency-fee plaintiff’s lawyer will take the case. Neither condition is typically met at small or even mid size employers.

More insecure forms of work like delivery apps are more explicitly carved out. California’s Proposition 22 classifies app-based drivers as independent contractors. They can be deactivated without cause and without recourse. Same as in China.

China’s welfare expansion as a response to the demand problem

Net it out. If you have the resources, you get more protection in China than California on dismissal (and only barely, when PIPs are in play). That said, China’s growing pro-employee case law isn’t a leftward turn. It’s a developmentalist state responding to several pressures at once. Demographic collapse. Weak household demand. A fiscal squeeze on the local governments that built the development model in the first place.

The demographic shock

Population fell for a third consecutive year in 2024. The NBS reported a year-end population of 1.408 billion, a decline of 1.39 million, with 9.54 million births against 10.93 million deaths. James Liang, in an essay on Sina Finance, notes the underlying decline accelerated. Dragon years average 6.4% more births than rabbit years, and 2024 only delivered 5.8%. A widely cited demographic report on Gelonghui puts China’s TFR around 1.02 in 2023. Per the China Association of Social Security, the 20 to 29 female cohort shrank by roughly 4.4 million per year between 2017 and 2022.

The State Council Development Research Center’s She Yu states the official framing. “The excessive cost of child-rearing has become the principal factor suppressing fertility intentions.” Ageing compounds it. The 2025 China Pension Finance White Paper reports 65+ at 15.6% of the population in 2024. CASS pension scholar Zheng Bingwen, at the 2025 Boao Forum, projects super-aged status (≥20% over 65) by 2032. Pension fund cash-deficit and reserve-exhaustion dates have been pushed from 2028/2035 to 2036/2044. Delay, not solution.

The childcare subsidy

Per the Implementation Plan jointly issued by the General Offices of the CCP and the State Council on July 28, 2025, every child born on or after January 1, 2025 receives 3,600 yuan per year until age three. The State Council Information Office press conference of July 30, 2025 confirmed Ministry of Finance budget of roughly 90 billion yuan for 2025. What’s unusual is the central government bearing about 90% of the cost rather than forcing provincial and local governments to pay for it. The Economic Daily called this “the first large-scale, universalized, direct cash benefit to the population since the founding of the PRC.”

China’s welfare state has historically operated through employers and hukou-linked entitlements. Almost never through a more universalist cash transfer. 3,600 yuan amounts to roughly 300 yuan per month.

Hukou liberalisation is a long trajectory, fiscally bottlenecked

A Peking University NSD working paper by Zhang Jipeng and Chen Zhu reconstructs the slow history of hukou liberalization. Small cities fully opened in 2014. Type-II large cities (1 to 3 million urban population) had restrictions abolished in 2019. The State Council’s July 2024 Five-Year Action Plan formalised removal of settlement restrictions in all cities below 3 million, substantial liberalisation for 3 to 5 million, and points-based settlement for the five megacities. A December 2025 State Council report to the NPC Standing Committee records that from 2019 to 2024 the resident-population urbanisation rate rose from 62.7% to 67.0%, and the urban-rural income ratio fell from 2.64 to 2.34. The State Council’s March 2025 acknowledgment put central incentive transfers for migrant-citizen public-service provision at 40 billion yuan in 2024 and 42 billion yuan in 2025.

40 billion yuan is a rounding error against what local governments have lost. Per the Sina Finance “China Land Finance Report 2024”, land-transfer revenue fell from a 2021 peak of 8.5 trillion yuan to 5.8 trillion in 2023, a 31.7% decline, and continued falling to roughly 4.87 trillion in 2024 per a Zhihu fiscal analysis. Land revenue averaged 34% of local government revenue from 2012 through 2023, peaking at 43.2% in 2020. Local dependence on central transfer payments rose from 28.6% of local revenue in 2021 to 36% by 2023, and Heilongjiang and Tibet receive over 80% of their fiscal spending from central transfers. Local governments lost roughly 3 trillion yuan in annual land revenue between 2021 and 2024. The 40-billion-yuan migrant-citizen fund is just over 1% of that loss. Beijing is filling a crater with a teaspoon.

This is why hukou reform is gradual. The cities most able to absorb migrants are precisely the cities that lost the most land revenue and now lack local fiscal capacity to expand public schools, public housing, and public health services.

Twenty-seven gas turbines run day and night without an air permit. 2 miles away are about six public schools, including an elementary school. They are trucked in to power xAI’s Colossus 2 data centre across the state line in Mississippi. The nearby Boxtown neighbourhood already breathes a cancer risk roughly four times the national average, and the turbines could emit more than 1,700 tons of nitrogen oxides a year.

Likely the largest industrial NOx source in a metro area already failing national smog standards. Mississippi regulators have nonetheless permitted an even larger facility at the same site. The turbines run because the regional grid takes five to seven years to deliver new interconnections, and AI capital-expenditure schedules do not.

A steady, high-load-factor customer spreads the grid’s fixed costs across a broader base. In theory, it should lower rates for the households around it.

It is also, in the United States in 2026, mostly just theory. Wholesale prices near significant data-centre activity have risen as much as 267% over five years, and household prices rose 7.4% in the year ending February 2026. PJM’s capacity-market clearing prices have hit their FERC-approved cap for three auctions running. Pew estimates that data-centre capacity costs alone are adding $18/month to residential bills in western Maryland and $16/month in Ohio, before transmission and distribution buildout. The industry isn’t doing what it says it could.

Ninety years earlier, in the same country, a federal agency that had not existed six years before was stringing electric lines to Tennessee farms whose owners had never owned a refrigerator. The Tennessee Valley Authority, chartered in 1933, employed more than nine thousand workers within a year, building dams and electrifying a region whose per-capita income was half the national average. The Rural Electrification Administration, founded in 1935, brought power to most American farms within fifteen years.

What could data centres do?

Data centre actors accounted for roughly half of U.S. corporate clean-energy procurement in 2024, and the IEA’s Energy and AI report documents them as the largest single class of corporate clean-power buyers in the world. You need demand for energy if you want to build out more energy infrastructure, especially clean energy infrastructure.

Electricity systems carry large fixed costs and recovered across whatever load shows up to pay them. A new subdivision adds load that peaks at dinner and disappears overnight. A data centre runs 24/7 at 90%-plus capacity factor. The data centre is the better customer on cost recovery alone.

What hyperscalers want is a lot of cheap power on timelines that match their capex schedules. They do not care if it’s clean energy or fossil fuels. They want compute, and they want it cheap. The cheapest compute in 2026 is powered by solar plus storage plus, where available, nuclear. You think that a rational hyperscaler in a working electricity system would invest in clean generation and sign contracts that anchor utility investment. It would absorb cost overruns so its neighbours’ bills did not move. Some of them do.

You also think that ones who have a bit of political influence would encourage the government to make the investments for them, in a way that doesn’t draw ire from the public.

So why don’t they all?

What do hyperscalers do instead?

The hyperscaler arriving in a new region can spend years on interconnection negotiations and absorb cost overruns so its neighbours’ bills do not move. Or the hyperscaler can route around the institution entirely. It can build behind-the-meter gas turbines, sited where permitting frictions are lowest, and run them while the permits are pending or without seeking them at all. xAI in Boxtown is the irresponsible example of this. It sits there not because the architecture forced it to, but because its principals do not believe the surrounding energy environment is their problem. The political economy so far signalled that they are right.

This isn’t just the xAI turbines in Boxtown. We have the 4.5 GW Crusoe gas expansion in Abilene, the 400 MW Williams “Socrates” facility serving Meta in Ohio, and Microsoft’s $7 billion Chevron-built gas plant in the Permian Basin.

Cleanview tracks 56 GW of behind-the-meter capacity across 46 sites and now projects 30% of new data-centre capacity will be served on-site. Almost all are up from effectively zero a year ago. It’s become enough of a problem that a bloc of Senate Democrats opened a formal investigation in March 2026 into twelve such projects.

It isn’t just all fossil fuel expansions. Microsoft signed a $10.5 GW renewables procurement with Brookfield in 2024, the largest corporate clean-power deal on record. Google has signed PPAs anchoring new utility-scale solar buildouts and a Kairos Power agreement for advanced reactors structured as a long-horizon clean-anchor contract rather than a turbine workaround. Emissions aren’t going to be much of a problem with solar panels and wind turbines.

It is a small set of operators who have decided that absorbing the procurement risk is cheaper, than the regulatory blowback the shortcut will eventually produce.

The queue is what makes the shortcut attractive. By the end of 2025, over 2,060 GW of generation and storage were seeking grid connection across the U.S., according to Lawrence Berkeley National Laboratory, more than twice the country’s existing fleet. The median wait from request to commercial operation has doubled, from under two years for 2000–2007 projects to over four years for 2018–2024 projects.

CenterPoint Energy in Texas reported a 700% jump in large-load requests in a single year. Other Texan utilities like ComEd and PPL (along with Oncor) now report more gigawatts of data-centre applications than their entire historical peak demand.

Lack of hardware makes it really hard to clear out the queue. By Wood Mackenzie’s Q2 2025 survey, lead times for standard power transformers have reached 128 weeks and 144 for generator step-up transformers. Roughly 80% of large U.S. power transformers are imported. Cleveland-Cliffs is the only domestic American producer of the grain-oriented electrical steel transformers require.

There are nearly $2 billion in announced manufacturing expansions. Which will not come online until 2028 or later. You need the grain-oriented-steel workforce and the high-voltage line crews. Along with substation engineers who connect new generation. We can’t scale up because the apprenticeship pipelines for these trades have been thinning for two decades while nobody was paying attention.

What do utilities want?

The other part of the problem is who the hyperscalers need to work with. American investor-owned utilities operate under a regulatory contract essentially unchanged since the early twentieth century. The state grants the utility a regional monopoly and an obligation to serve every customer in it.

In exchange, a commission sets rates that recover the utility’s operating costs plus a regulated return on its rate base (the depreciated capital sunk into plant and lines). This was a sensible design for extending wires to new neighbourhoods at predictable cost. It is also a disastrous design for coordinating a multi-trillion-dollar buildout against a state competitor that does have the capacity to do so.

This rewards rate-base growth above almost everything else. The Averch-Johnson effect (named after the 1962 paper that formalised it) is the standard result on this kind of regulation. A utility whose authorised return exceeds its cost of capital has a structural incentive to “over-invest”.

The American Economic Liberties Project’s January 2025 paper by former Sempra chief economist Mark Ellis estimated that authorised returns on equity exceed the actual cost of capital by enough to cost American utility customers roughly $50 billion a year, about $300 per household. The stock market capitalises that stream of excess returns at about $1 trillion. For 2024, the average authorised return on equity for U.S. regulated utilities was 9.7%, against a broad-market forecast of 6.7%. Over three decades, Ellis finds, IOU capital investment per kilowatt-hour delivered grew about 3.8% per year faster than demand. The legal threshold articulated by Justice Brandeis in 1923 and adopted by the Supreme Court in 1944 holds that returns above the cost of capital are “unjust and unreasonable“. Nobody enforces it, because nobody’s job description is to.

You can do this experiment yourself (well, almost). Two California utilities (San Diego Gas & Electric and the Sacramento Municipal Utility District) serving major metro areas under the same state energy policy. Same wildfire risk and demand growth. One answers to Sempra’s shareholders, the other to a board elected by the City of Sacramento. Between 2020 and 2023, SDG&E’s residential rate rose 65% and now exceeds 45¢/kWh. SMUD’s rose less than inflation and stayed near 17¢/kWh. The gap is more than two and a half times.

Financing explains some of it. Public utilities raise capital through municipal bonds at 4 to 5% interest and are exempt from corporate income tax. IOUs raise capital at nearly double the cost. But these account for perhaps a third of the gap. The rest is the management philosophy that follows from ownership. SMUD’s managers are not maximising earnings per share and its investors are not expecting double-digit returns. SDG&E’s are. Sempra’s CEO took home roughly $14 million in total compensation in 2023, while SMUD’s general manager made under $700,000. Sempra’s stock price tracks SDG&E’s rate base because that, by regulatory design, is the source of its profit. A larger rate base produces higher earnings, and the resulting share price feeds executive pay. Everyone in the chain responds rationally to the incentive structure. Nobody is behaving badly. The chain is the problem.

The pattern is industry-wide. The American Economic Liberties Project found that between 2020 and 2023, IOU residential rates rose 49% more than inflation while publicly-owned utilities’ rates rose 44% less. Investor-owned utilities used 2025 earnings calls to disclose $1.4 trillion in planned capital investment through 2030, a 21% upward revision in a single year, the rate base growing in real time with the profit stream attached.

The mechanisms range from outright bribery to rate-case manoeuvring. The defining recent case is FirstEnergy and Ohio’s House Bill 6, in which from 2017 to 2020 the company paid roughly $60 million through a dark-money nonprofit controlled by Larry Householder, then Ohio’s House Speaker, now a federal inmate. The payments bought $1 billion in subsidies for FirstEnergy’s struggling nuclear and coal plants. The company also paid $4.3 million to the man who would go on to chair the state utility commission.

The newest mechanism is secret data-centre contracting. You might think utilities at least file these things publicly. You can guess that they don’t.

A March 2025 Harvard Electricity Law Initiative paper, Extracting Profits from the Public by Eliza Martin and Ari Peskoe, reviewed nearly fifty regulatory proceedings and found a pattern. Utilities negotiate special contracts with hyperscale customers and file them with state PUCs under confidential treatment.

The infrastructure costs are then spread to all ratepayers through ordinary rate cases. Federal antitrust litigation against Duke Energy established that the company sought to shift the costs of a $325 million data-centre discount onto its other ratepayers. Similar patterns appear at AEP and Dominion, and at Entergy in the South. The rate structures that once spread the cost of reliable power across everyone are now forcing the public to pay for infrastructure built to supply a handful of wealthy corporations.

Utilities looking at a hyperscale interconnection is not looking at a problem. It is looking at the largest rate-base-growth opportunity in half a century. So, they will offer favourable terms and spread the infrastructure costs across its broader ratepayer base, filing the whole thing under confidential treatment. The hyperscaler, faced with grid lead times that cannot wait and just don’t care. Of course they take the deal.

Why doesn’t either side fix itself?

The deregulatory account is that American electricity got expensive because regulators got in the way. That Texas is the proof case (the most market-based major grid in the country) with the lowest industrial rates in the lower 48. Get government and the legacy utilities out of the way, the argument goes, and supply responds. The second Trump administration’s Executive Order 14154 and the “Speed to Power” agenda are its policy expression.

The progressive account runs the other way. American electricity got expensive because corporate concentration captured the public interest. IOUs extract excessive returns through regulatory capture. In this case, it is the hyperscalers externalise their power costs onto residential ratepayers. The alignment between Big Tech and the Trump administration is the visible quid pro quo. The ideal remedy is some sort of enforcement. It’s also a popular idea as Abigail Spanberger won Virginia by 14 points in November 2025 campaigning on this account. No to mention, 78% of Americans told Consumer Reports they are worried data centres will raise their bills.

Each account have a degree of truth. Permitting and interconnection lead times are just absurd. IOUs do extract roughly $50 billion a year through a comprehensively gamed system. But both share a deeper failure as they assume the architecture of American electricity governance is adequate. The operational (not political) problem is which actors operate within it.

ERCOT runs the most market-based major grid in the country. In twelve months it requested a federal emergency order during Winter Storm Fern. The legislature afterwards passed Senate Bill 6 to mandate large-load interconnection rules. The state also leaned on the Texas Energy Fund (a $9 billion voter-approved programme) to fund dispatchable generation. That is an energy and industrial policy in everything but name. Texas (reluctantly and in pieces) is reverting to state coordination because the market alone cannot do the planning.

Even the Texas Public Policy Foundation (ever hostile to coordinated state action) documents that the state’s winter capacity gap is now larger than during Winter Storm Uri in 2021. ERCOT added 31 GW of solar and 15 GW of batteries but only 3.1 GW of new natural gas between 2021 and 2026. While peak winter demand grew roughly 20%.

PJM (the largest grid operator in the country and serving 67 million people) hit its FERC-approved capacity price cap for the third auction in a row in December 2025. Still 6,625 MW short of its reserve-margin target. Absent the temporary cap negotiated with Pennsylvania Governor Josh Shapiro, PJM estimates the price would have cleared roughly 60% higher. It’s own market monitor attributed about $6.5 billion of the $16.4 billion auction cost to data-centre load.

The progressive account isn’t any better. If it was simply that Big Tech and IOUs extract from ratepayers. The cure would be to discipline them and nothing else. But hyperscalers choose gas turbines because no institution exists to deliver the alternative. State regulators are patching, late and unevenly. The Virginia State Corporation Commission approved a new GS-5 rate class in November 2025 shifting more buildout cost onto large data-centre customers. A the same time approving a $16/month residential rate increase.

Who got what they paid for?

So why do utilities and hyperscalers, whose deeper interests actually diverge, end up coordinating against the public anyway? American political economy is fragmented across layers of government and empowers local self-rule. Which privileges actors with long time horizons and the ability to operate across many venues at once. As JS Tan has argued, hyperscalers have both. It turns out, so do investor-owned utility holding companies. Almost nobody else does.

In practice this means venue arbitrage. A hyperscaler shopping a $5 billion project can play states and counties against one another, and play their regulators. A utility shopping a rate case can find a friendly PUC or legislature and play it against federal pressure. Neither needs the other’s enthusiastic cooperation. The short term incentives just happen to align.

Hyperscalers’ deeper interest is in cheap power, which is the same for you and. Given a working developmental institution, that interest would dominate. In the system we have now, it loses to something more proximate. Utility wants rate-base growth while the hyperscaler wants compute on capex timelines grid interconnection cannot meet.

You can see where this is going as the deeper interest is slow, but there is a shortcut. Most of the industry are going to take the shortcut.

They are coordinate with the utilities, sometimes through secret contracts. Other times through siting decisions that route load toward whatever utility offers the least scrutiny, not because their interests align but because nothing better is on offer.

The political surface is loud enough to mistake for the cause. Five major hyperscalers each gave $1 million to the Trump inauguration fund. Marc Andreessen and Ben Horowitz pledged $50 million to an AI super PAC.

Trump’s Executive Order 14148 fast-tracks any data centre over 100 MW, covering gas turbines and coal equipment while pointedly excluding wind and solar. I am not the “Big Tech bought the election” type. I am also not going to pretend the contributions and the EO are unrelated.

The temptation is to read this as straightforward corruption (I am not ruling it out). Hyperscalers bought the deregulation they wanted. I think that reading is too simple. Hyperscalers really did want was to scale. What they bought (and what the administration delivered) was permission to bypass the regulatory institutions that were not delivering it. The utilities were happy to receive their share.

The permission is not durable, and the political surface is already shifting. The backlash is not confined to the affluent-liberal-NIMBY caricature big tech finds convenient. As has put it, “Data centers are one of the only physical manifestations of ‘big tech’ encroachment on ppl’s lives and broad swaths of the public are very angry and resentful of these companies and the power they have,” adding that “AI cos have done a terrible job explaining why they need these data centers.” Some industry observers see the reckoning coming: predicts that “The next Dem admin will look to ban data center scale deployment of diesel gen sets.” puts the strategic failure more sharply: “If you want to maximize the rate of progress over the long run you need to take probable political opposition into account and take steps to mitigate it by making sure it won’t derail you.” The bypass strategy works only until it doesn’t.

There are good proposals out there to build as much energy (clean btw) to bring down prices to penny a kilowatt/hour. It’s just as said, the country has not lost the capacity to build. It has lost the capacity to build for the public.

When Zohran Mamdani won the Democratic primary for mayor of New York in July 2025, not everyone took the news well. The New York Post ran a piece headlined “Zohran Mamdani’s feeding off Gen Z’s misplaced doomerism.” Its argument leaned on Jean Twenge, who has spent fifteen years making the case that each new generation arrives carrying some fresh psychological pathology, rather than, say, responding to a decade that ran from the Great Recession through the COVID years, or anything else with real weight to it. Gen Z, she explained, votes “the bastards out” because depression makes the world look permanently grim.

Twenge is not wrong that perception drives behaviour. Indeed it does, and we would not argue otherwise. Where she goes wrong, in our view, is the perception itself, and the question of whether it tracks anything real. Young New Yorkers, and older ones, were looking at concrete, material things. The Second Avenue Subway cost $2.5 billion a mile, eight to twelve times what comparable lines cost in Paris or Madrid, by the count of NYU’s Transit Costs Project. A cap on food-vendor permits, in place since 1979, had sat untouched for forty-six years. On the black market a single permit now sublet for $15,000 a year, as one investigation found. Six thousand small-business rules had piled up over the decades, and we suspect nobody was keeping count.

Telling a voter that she is depressed, and that the cure is your candidate (one with a certain kind of “distinguished” history), is a way of not discussing the subway, or the economy that makes a plate of halal run over $10.

What about New York?

We are picking on Mamdani and New York for a straightforward reason. The city is thick with history and case studies, and not the dusty kind. Recent history that conservatives to progressives can learn from.

Two years before Mamdani won, New York drove out the one transit official in living memory, nicknamed Train Daddy, who had actually been fixing the system.

Andy Byford took over the New York City subway in January 2018, with on-time performance at 58.1%. He had come up from station foreman in 1989, through the London Underground and the Toronto Transit Commission. A self-described subway nerd, he spent his first weeks walking the platforms.

One of his first orders was for three vacuum cars to clean the track beds. Litter on a track bed delays no train. On any cost-benefit model a consultant would run, the vacuum cars are waste: money spent on something that moves no measurable number. Byford bought them anyway, and said why. “My customers shouldn’t have to look at that.”

That is not an especially technocratic line from Train Daddy. He had practical and safety arguments available to him, and used none of them. Instead he made plain that he was buying a subway that looked cared-for, on the theory that a rider who sees filth correctly concludes that nobody is in charge. Two equally competent engineers could walk the same platform, see the same track bed, and disagree about whether the vacuum cars are worth the money. Looking at the thing directly is necessary. It is not the whole of the skill. The other half is an imagination for what the thing feels like to the person using it.

Byford worked the trains in two ways. Contrary to the popular memory of him, much of it was incremental. A SPEED Unit audited the faulty signals that had been quietly slowing trains for years and raised limits where the engineering allowed. Not to mention, the Save Safe Seconds campaign shaved seconds off millions of daily trips. Eighteen months in, weekday on-time performance hit 81.5%, the highest since 2013.

His second way, and what gets remembered is the big structural projects. Fast Forward, a $40 billion plan to re-signal most of the system on a timeline 75% faster than the baseline. (A note before planning anything that bold, Byford had done the work for three decades across four agencies.) He knew how the procurement and the labour realities actually worked, not from a briefing but from doing them.

The point worth holding onto is that the vacuum cars belong to neither category. Clean track beds and re-signalled track are different kinds of project, and a manager who thinks only in measurable outputs will fund the second and cut the first. Byford funded both, because both mattered to how the system was experienced.

In January 2020 he resigned. Andrew Cuomo, Twenge’s beloved governor, had spent a year freezing him out: briefing Byford’s subordinates while Byford sat in his own office, and overruling his engineering judgement. He brought in AlixPartners to “transform” the MTA by trimming Byford’s authority down to “the day-to-day running of service.” Cuomo’s stated objection to Fast Forward was that it was too slow. A politician with no transit background was informing a four-continent professional that his timeline lacked ambition. Byford went on to run Transport for London, and now leads high-speed rail at Amtrak.

Mamdani

Mamdani took office five years later. Before anyone reads what follows as cheerleading for a left-wing mayor, hold on for a moment. A Republican and an ex-Democrat independent are coming a few sections down, and they are also models of competency.

On January 25, 2026, three and a half weeks into his mayoralty, Zohran Mamdani got his first snowstorm. Winter Storm Fern dropped about a foot of snow on the city, and the streets were ploughed fast.

It was not smooth. Ploughs shoved snow into bus stops, the nearest open space, and nobody was tracking which stops had been cleared. Then sub-freezing cold turned the banks to ice. Riders posted photos of three-foot walls of frozen snow where their stops used to be, and crosswalks vanished. At least 20 people died of cold exposure in the weeks that followed. Some inside their own apartments. City Council Speaker Julie Menin said the deaths “were not inevitable.”

Four weeks later, the city’s first blizzard since 2016 buried all five boroughs under nearly two feet of snow. It was a worse storm by every measure. Mamdani’s response was markedly better. Between storms, the administration had geotagged every unsheltered bus stop and crosswalk so DSNY could track each location and the time it was last cleared. Emergency shovelers went from 550 to 1,300, their pay from $19 to $30 an hour. Warming centres more than doubled. An overnight travel ban kept private cars off the roads. By Monday morning, 2,200 bus stops were clear and no one had died on the streets. You do not often watch a mayor learn that fast from his own mistakes, or learn from them at all.

His administration brought the first enforcement action against a delivery app under New York’s food-delivery laws, an $875,000 settlement with HungryPanda and restitution to 380 immigrant-owned restaurants it had overcharged. The point is not the settlement figure. It is that the legal authority had been sitting there unused for years. Enforcing a junk-fee law against a well-capitalised tech platform was nobody’s idea of “what works”. It had no champion among the pundits who use and abuse that phrase. The Mamdani administration checked whether the law was being followed, found that it was not, and enforced it.

On the more “centre” or “moderate” side of things, Mamdani’s Executive Order 11 directed seven agencies to inventory the 6,000 rules and fees facing small businesses. The order set fixed deadlines and was built with the city’s chambers of commerce. The street-vendor reform is another example. The 1979 permit cap had created a black market charging $15,000 a year for a sublet, while vendors caught working without one faced criminal charges. The administration moved to lift the cap and decriminalise unlicensed vending. The aim was a more legitimate channel towards the 17,000 permits the city is legally required to issue and simply never got round to. When a cap is producing a black market, lift it. Then build the lawful channel, so that the legal path is also the easy one.

None of this is especially ideological, regardless of what the discourse on Twitter has decided. Which raises a fair question. Why did Eric Adams, the supposedly more “respectable” and “centrist” mayor, do none of it? The honest answer is that the credentialled-centrist style does not reward this work. It rewards what presents well to the foundations and the ratings agencies, and an unenforced junk-fee law presents as nothing at all. We are only months into Mamdani’s term, and anyone who tells you how it ends in three years is selling something.

Carmel and Detroit

So how do Mamdani and Byford actually connect? Two more cases, both rather more centrist, bring it round.

Jim Brainard became mayor of Carmel, Indiana in 1995 with no real public-office experience. Early on, a friend pointed at a pile of pipes by the road and asked him what they were. Brainard did not know, so he wrote to urban-planning departments around the country and asked for their syllabi. Twenty-eight years later Carmel had 152 roundabouts and a reputation as “the internet’s favorite small city,” on the strength of an urbanism that is genuinely walkable by American suburban standards.

Carmel’s traffic-fatality rate runs at roughly a fifth of the national one, and a downtown rebuilt from a sidewalk-less suburb now holds more than 150 corporate headquarters. A Republican, Brainard governed against half his coalition’s instincts, not out of heterodoxy, though in effect he was heterodox, but because he wanted Carmel to be a better product for its citizens and customers.

Mike Duggan inherited bankrupt Detroit in 2014. He had already run three Detroit-area institutions and knew them from the inside. As county prosecutor he had once reached back to a law-school class on common-law nuisance and used it to force absentee owners to fix or forfeit houses, an effort that saved an estimated 15,000 to 20,000 of them. As mayor he built a Department of Neighborhoods and watched Detroit post its first population growth since 1957. He has since left the Democratic Party to run for governor as an independent. On results alone, he would make a good governor for Michigan.

The method

Start with Circuit City, because it is the cleanest case of the failure this whole essay is about. By the technocratic playbook the company did everything right. Watching its costs, it fired 3,400 of its highest-paid store associates in March 2007, about 9% of the workforce, and replaced them with cheaper hires to bring down selling and administrative expense. On the spreadsheet the move was sound, sensible reallocation of capital.

But the spreadsheet was not being empirical. It was being empirical about the handful of things it had decided to count, and the most important fact about Circuit City was not among them: those associates were the company’s actual offer. The reason a shopper chose Circuit City over Walmart or Amazon is the same reason a confused customer walked out with the right television rather than empty-handed. Analysts warned at the time that knowledgeable service was one of the few things the chain had that Walmart did not. The layoffs were not an empirical decision. They were a prior commitment about what a retailer is, dressed in the costume of cost discipline. Knowledgeable service is simply a very hard thing to put on an asset sheet for the respectable types, and what cannot be put on the sheet does not exist.

That holiday, same-store sales fell almost 12%. Circuit City liquidated all 567 remaining stores less than two years after the firing.

What ties Circuit City to the mayors and to Byford? I argue it is two things called customer experience and process improvement. You treat the city as a system with users in it. You ask where the friction sits, and what the system feels like to the person inside it.

What we mean is fixing the process rather than blaming the people inside it. It is an old, unglamorous discipline, and for all its results nobody much embraces it. What respectable technocrat would rather stand near the factory floor and the quality audit than at the strategy offsite?

It is also not the credentialled-centrist mayoralty or some sort of McKinsey-trained reformer. It certainly not anything similar to Rahm Emanuel.

You would ask how a rider feels about track-bed litter, or a station that smells of urine. They ask how a vendor paying $15,000 for a permit the city is meant to issue free could possibly improve anything for the citizen buying lunch. It does not! It only makes the food cost more and restrict people from opening up shop.

It’s also important to not that incrementalism and big reform are not an either/or. The skill is knowing which the situation calls for. When the architecture is sound but clogged, you take a page from “Yes, Prime Minster” and salami-slice. That is hundreds of small cuts, each one beneath notice, and they add up further than you would think. When the architecture is broken, you sometimes have to rebuild.

And there is something for which neither the pundit’s grammar nor the consultant’s has a word. It is as some marketers would put it, changing how a thing is experienced without changing the thing itself. Often it is no more than spending the money to keep something clean more often. You would be shocked how frequently this is the cheapest real improvement available, and the quickest way to lift the very perceptions Twenge worries about. It is also, somehow, the first thing a manager cuts when he funds only what he can count.

Byford’s “my customers shouldn’t have to look at that” is the whole of it. The customer is the rider and the shop owner, not the donor and not the out-of-town activist. And the gap is not only one of interest. The donor and the activist have never stood on the platform at rush hour, and have never met the gig worker the junk fee was skimming. You cannot have an imagination for what a system feels like from inside it if you have never been inside it. A pundit will defend incrementalism as realism, and not just any incrementalism but the slow do nothing kind that ruins the broader reputation of incrementalism itself! The reply to that is constraint was never ambition. It was always capability, with a dab of will.

The pipeline

Why is it so rare? The generous answer is that the incentive structure rewards organised claimants. (On that note, many aren’t even the city’s customers and citizens, like commuters from New Jersey into New York.)

That answer isn’t the whole truth, but it isn’t wrong either. The activist machines are not the villains here, and, like it or not, they do real work. Someone has to gather the volunteers and carry a point of view. Their failing, left or right, is duller than villainy. It does not occur to most people to ask whether the thing actually runs, and they tend to assume, implicitly, that the civil service will do as it is told, which opens a whole other can of worms.

The gap they leave has been filled, for forty years, by a technocratic class. A class that most people, until recently, would have called Cuomo a part of, in spite of Cuomo’s contempt for Byford as “just a division head.”

Now consider what the class built. There is the $2.5-billion-a-mile subway, and the consulting capture of state functions alongside it. It did not merely neglect the unenforced fee cap and the uninventoried rules.

The commentary has the same blind spot. It will praise a mayor for lifting the city’s mood faster than it will praise him for auditing its signals. Morale is real, and worth having. But it is what riders feel when a system works. You cannot reach for it directly, no matter what some pundits will say.

So the question is not how to beat the activist machines. It is how to work with them, on the policy goals they and the city actually share, to cut the friction, and how to build a training pipeline for the operational craft.

We have multiple mayors and civil servants who express a high level of competence in improving cities, with many of them being amateurs on the mayor side. We know this isn’t a one-off. What’s also interesting is that it isn’t limited to partisanship. There is a case to be made for reproducibility, and the activists machines have the networks to actually spread any reproducible training or set of policies and procedures that reduce friction for goals.

Pundits, who have the audience to make that case, would rather produce more punditry.

On a clear September morning in 2024, in Baldwin Village, South Los Angeles, Mayor Karen Bass stood on a five-acre lot at the corner of Coliseum Street and La Brea Avenue and turned a ceremonial shovel. Behind her, renderings showed a six-storey building with 800 flats stacked above a Costco warehouse. A local resident, Karen Johnson, told NBC she was excited because “we only have a Ralphs right here.” Governor Newsom sent a statement. Councilmember Heather Hutt sent another. Of the 800 flats, 184 would be reserved for households at 30, 50, and 80% of area median income, with the lowest band renting at $1,040 a month. The project was privately financed. No Low-Income Housing Tax Credits. Just a wholesale club agreeing to be the load-bearing tenant.

A year later, Marina residents gathered in the parking lot of an iconic mid-century Safeway at the foot of the Golden Gate Bridge. The store is known to locals as the “Singles Safeway” for its Monday-night dating scene, has been there since 1959, and is a neighbourhood institution. A developer called Align Real Estate had proposed replacing it with a 25-storey, 790-unit residential tower designed by Arquitectonica, with the Safeway returning at 57% larger on the ground floor. Eighty-six of the flats would be deed-restricted to very-low-income households. The protest signs went up. Marina Community Association’s Erin Roach delivered the line: “We’re not anti-housing, we’re anti this development.” The mayor of San Francisco, Daniel Lurie, agreed with her. So did his ally on the Board of Supervisors, Stephen Sherrill, who had quietly removed the parcel from the citywide upzoning plan a month earlier. Both denounced the developer for “gaming the system.”

(On that note: Daniel Lurie is opposing this housing project in his city while polling his weakest on housing, with 78% of San Franciscans saying cost of living has gotten worse.)

Two projects. Roughly 800 flats apiece. Both anchored by a major retailer. Both rely on the same state law: California’s AB 2011, the Affordable Housing and High Road Jobs Act. One mayor turned the shovel herself. The other promised to fight.

Why does this matter?

The dominant story about American housing politics in 2026 is that the YIMBYs won. State legislatures have passed pre-emption laws. Pro-housing reformers have been elected in major cities. The zoning code is, slowly, being rewritten. We have been here before. We had Tesco’s Spenhill in London in the 2000s. We have had Hong Kong’s Housing Authority since the 1960s. We have had Gavin Newsom’s “3.5 million homes by 2025” pledge. The relevant question is not whether the rhetoric has shifted, but whether the rhetorical shift will translate into actual buildings.

This matters because the answer determines whether the United States can build out of its housing crisis or merely talk out of it. If state pre-emption laws hold, then several million flats become possible across underused commercial sites in expensive coastal metros within a decade. If the laws are quietly hollowed out by mayors who run as reformers (especially centrist reformers) and then govern as obstructionists, then the rhetoric was the policy and there is nothing else coming. Misdiagnosing why the Marina Safeway will or will not get built is the difference between betting on the law and betting on the politicians. Bet on the law and the politicians become weather. Bet on the politicians and the law becomes window-dressing. The Marina Safeway is one test of which bet is right. The Coliseum Costco is the other.

What’s wrong with the conventional story?

The conventional account of the Coliseum project, repeated by Mayor Bass and most of the friendly coverage, runs as follows: a pragmatic Democratic mayor of a large American city has used a clever new state law to deliver hundreds of units of affordable housing on a previously underused parking lot, in partnership with a beloved retailer and a private developer, all without federal subsidies. This is what good housing politics looks like. It is replicable. It is the future.

The strongest version of this story is genuinely compelling. AB 2011 is a remarkable piece of legislation. The Coliseum project is going up. Bass did champion it publicly. The 184 affordable units are at rents that working-class South Los Angeles families can afford.

It is the inference about the mayor that fails.

Karen Bass has, on closer inspection, a striking pattern of behaviour on housing that is hard to reconcile with the “pragmatic reformer” narrative. She stood up an ambitious permit-streamlining programme called ED1 and then rolled it back once development applications started flowing in. She opposed SB 79, the state bill allowing more apartments near transit stops, on the grounds that it deprived the city of granular control. She opposed allowing Palisades fire victims to build duplexes on their own lots, explaining to a podcast host that “the people in the Palisades didn’t want that…if you impose it on a neighbourhood, then you will have hardcore NIMBYs.” People has taken to calling her Mayor NIMBY (which is, on reflection, slightly unfair to the NIMBYs, who tend to be more honest about what they want). Daniel Lurie in San Francisco might follow the same pattern (in miniature so far), with the addition that his hostility to specific projects is more openly stated. Both happen at once: the headline embrace of one prestige project, and the quiet obstruction of the broader regime that would produce many more.

The pattern is not ideological inconsistency. It is loss aversion as urban policy. The anger of one miffed homeowner is felt by a mayor about twice as heavily as the gratitude of one new tenant. The new tenant doesn’t yet live in your district. Kahneman and Tversky’s old finding (losses register about twice as heavily as gains of equal size) predicts the entire mayoral career arc. The political incentive runs against new construction even when the policy preference runs for it. Which is exactly why the law has to do the work the mayor will not.

The story is not that good mayors deliver housing. The story is that even mostly-good mayors will obstruct housing whenever it is politically convenient, and the only thing even enables the production of buildings is the law and the political economy. Bass turned the shovel at the Coliseum because the project had already cleared AB 2011 review and at the time, the YIMBY coalition was strong enough to prevent Gavin Newsom from siding with her on this issue like he did with the duplexes. Had AB 2011 not existed, or had Bass possessed any local lever to slow the project down, I would not be writing this paragraph. The relevant agent is the statute, not the politician.

Why would anyone want to live above a Safeway?

Marina Safeway, as the San Francisco Standard put it in February 2026, “the NIMBYs’ last stand”: the mayor opposes it, the local supervisor opposes it, the neighbourhood opposes it, and the law still requires it to be approved. The Planning Department issued a notice of conditional AB 2011 eligibility on 24 March 2026. Under the statute, the city has 180 days from a complete application to issue a final decision. The deadline is 1 August 2026.

Yes, you get that and you are going to ask “So, what?”.

I have said almost nothing about why anyone would actually want to live above a Safeway. The politics is impossible to understand without the product, which is what the urbanism literature has done the least to defend.

The central observation of behavioural economics: human beings are far more affected by friction than by price. If you make something ten seconds harder and people stop doing it. Make it ten seconds easier and they start. Reshaping of daily life over the last twenty years has been about removing small frictions like faster free delivery or smartphones. The grocery-anchored mid-rise applies the same idea to the activity humans spend more time on than any except sleep: keeping themselves fed.

Consider what it is to forget the onions in America. Forgetting the onions is a delightful little sidequest. You might even ask yourself whether the dinner can survive without them. You consider takeout. You give up and get in the car. You might drive 1.6 or 5.6 miles to the nearest grocery store. Then you have to find the onions, most likely queue at self-checkout rather than a person. Finally, you drive home, and arrive forty-three minutes later with a slightly mutinous spouse.

In a grocery-anchored mid-rise, you take the lift down, walk past the rotisserie chickens, pick up the onions, take the lift back up. The whole excursion takes four minutes or maybe it’s ten minutes tops.

Industry analysis finds that ground-floor premium grocery stores increase the rent achievable above them by up to 5.8%, and that number is low. It reflects only what landlords currently know how to capture. Renters would pay considerably more if the typology were less unusual.

Rory Sutherland has a line that applies here. Modern luxury is not really about size or expense. It is about not having to schlep. The richest people in San Francisco don’t have larger refrigerators than the merely well-off. They have private chefs, which is a way of paying $300,000 a year to make the schlep someone else’s problem (a luxury good the IRS has not, to its credit, yet found a way to define as taxable compensation). You could say that the grocery-anchored mid-rise democratises a slice of that luxury. The lift in a Marina Safeway tower is, functionally, a private chef. Most likely a worse one, considering the state of cooking among my peers and Gen Z, admittedly. But available to everyone in the building, at no marginal cost beyond the rent.

Aesthetic objections is the part of this debate is something I find least persuasive. Marina’s opposition rests on the assumption that a one-storey supermarket plus a surface car park on prime waterfront land within sight of the Golden Gate Bridge is somehow attractive. It is not. It is a deeply ugly piece of mid-twentieth-century commercial sprawl that San Franciscans have learned to mistake for “neighbourhood character” because they have lived with it for so long. The cities the Marina’s defenders love when they go on holiday (you can guess) are entirely composed of flats above shops. Edinburgh’s New Town is a grocery-anchored mid-rise from 1770 with a Sainsbury’s at the bottom; you have nothing to teach the Edinburgh planning department (besides the fact that said planning department may also be NIMBYs ignoring history) about preserving neighbourhood character.

Grocery anchor is often the only genuinely good thing on the block. The Marina Safeway is iconic because Marina residents go there repeatedly, with affection. The proposal, recall, is to expand the Safeway by 57% while putting flats above it. It is puzzling that this counts as a controversy until you remember Sam Moss’s observation that wealthy neighbourhoods learned long ago to define their preferences as principles, and Veblen’s older observation that those principles tend to exclude the most people.

The most important fact about these buildings: almost everyone who has lived in one would do it again. The Wegmans-anchored apartments in Washington DC have waiting lists. The Whole Foods-anchored buildings in Los Angeles fill up the day they open. The Hong Kong Wah Fu Estate, opened in 1968, retains higher resident satisfaction scores than most British council estates of equivalent age. In Tokyo, the integrated Ito-Yokado towers are prestigious. The product works. The market knows. The mayor of San Francisco may not yet.

Why is this harder to stop than it looks?

The product love above is not separate from the politics. It is the politics. Three forces make grocery-anchored housing harder to kill than ordinary apartment development. Most opposing politicians haven’t noticed.

The typology is global and mature. Apartments stacked on a concrete or steel podium over a supermarket are the default mode of dense urban housing in much of the world. In Hong Kong, the Housing Authority has been building exactly this kind of building since the 1980s; about 2.13 million people, roughly 30% of the territory’s population, lived in Public Rental Housing in early 2025, most of it in podium-anchored estates. The ten-year supply target for 2025/26 to 2034/35 is 308,000 public units. In Singapore, 77.2% of the resident population lives in HDB flats, roughly 90% of large-format HDB shophouses are anchored by NTUC FairPrice, and the HDB built 21,000 such flats in its first three years of operation. In London, Tesco’s Spenhill subsidiary built more than 4,000 flats under Mayor Livingstone’s residential-led London Plan, until Tesco’s 2015 retrenchment killed the pipeline, of which more later. The American examples (Beckert’s Park on Capitol Hill, The Exchange in Wheaton, the various Whole Foods-anchored projects in DC and LA) are catch-up, not innovations.

The grocery anchor brings a constituency that ordinary apartment developments lack. A developer fighting NIMBYs alone is on hostile rhetorical terrain. The actual political economy is more complicated, but at community-meeting altitude “developer” is shorthand for “rich greedy person,” and that’s the altitude these fights are conducted at. A developer fighting NIMBYs with Costco or Safeway as co-sponsor is in an entirely different argument. Costco is consistently rated one of the most-trusted retailers in America. Its CEO Ron Vachris has described the Coliseum project as a way “to enter markets where traditional big-box development would be nearly impossible,” and the Coliseum alone is expected to create 400 permanent retail jobs plus thousands of construction jobs. The Marina Safeway has the kind of intergenerational neighbourhood loyalty no developer can manufacture by himself; opposing 800 flats above it is one thing, but opposing the expansion of the Singles Safeway (57% bigger, with a wider deli counter and more checkouts) is much stranger to argue at a community meeting.

This is the part that should be on every protest sign and conspicuously isn’t. The Marina opposition is asking that its favourite neighbourhood Safeway not be allowed to get bigger and better. The thing the protesters say they love (the daily-life utility of the store, the Monday-night singles tradition, the iconic mid-century building) would, under the proposal, get a substantial upgrade. The thing they say they want to preserve, the surface car park, would be the only thing lost. Try writing that on a placard.

Tesco in 2000s Britain mobilised the same dynamic: it was easier to argue with planners about flats than to argue with everyone who shopped at the Tesco. The grocery anchor is a constituency, with employees, customer loyalty, brand favourability, and UFCW representation behind it. When NIMBYs fight Costco or Safeway directly, they are fighting an organisation with more political capital than they have. They are, in many cases, fighting their own friends.

The desirability of the building disarms the standard objection. The standard NIMBY argument runs: a tall building here would destroy the neighbourhood character. The argument assumes the building is undesirable. When the building turns out to be exactly the kind of place people want to live, with a grocery store downstairs they already like, the argument collapses on its own. The Marina Safeway opposition is struggling to articulate a coherent harm. The Tenants Union explicitly declined to oppose the project on the grounds that “there’s nobody being harmed by that, except the shoppers who will temporarily have their Safeway closed.” When even the tenants’ lobby refuses to oppose your tower, you are no longer fighting NIMBYism. You are fighting nostalgia, and nostalgia is not a permitting category.

It can’t just be the policy?

You might say, “Hold up! We seen laws passed for other things, and only ADUs really get built at scale! How does groceries make a difference?”

Laws enable development, but it doesn’t actually build units. Consider what a developer has to assemble to finance affordable housing without an anchor tenant. The standard American capital stack runs to five separate sources: Low-Income Housing Tax Credits, allocated by state agencies on a competitive lottery with two-year wait times; state and local subsidies, each with its own application process; deferred developer fees, which don’t pay out until the building stabilises how many years later; CDFI loans, cheaper than commercial debt but slower to underwrite; and tax-exempt bond financing where available.

A typical “affordable” project takes three to seven years of predevelopment work before a shovel hits the ground, and roughly ~10% of total project cost is spent on assembling the financing itself.

Grocery-anchored typology shortcuts the entire stack. A Costco or Safeway lease underwrites the “affordable” component privately. The math works because grocery retail is one of the few credit-worthy commercial uses with stable long-term cash flows that can be securitised against the residential portion’s construction loan. Predevelopment shrinks from years to months. The Coliseum’s capital stack closed without a single LIHTC application, which the developer has described as a model for replication.

So NIMBYs are feeling with these projects are wider and better-financed than the neighbourhood-association opposition. Even where mayors object, the laws operate independently of them. The British experience is the cautionary case. In 2007, under Mayor Livingstone’s residential-led London Plan, Tesco set up its Spenhill subsidiary to develop housing above its own stores. By 2014, Spenhill had built more than 4,000 flats across a series of London sites and was the largest single grocery-anchored housing programme in Europe. Then, in 2015, under new CEO Dave Lewis, Tesco retrenched. The Spenhill sites were sold off, the development arm was wound up, and the pipeline died with it. The flats already built remain (residents seem perfectly happy with them, fifteen years on). The flats that would have followed never came. When the political economy rests on a single retailer’s strategic appetite, the policy lasts as long as the appetite does. AB 2011 is built differently: the supermarket is one supporting leg, the law is another, and the labour coalition is a third. Knock one out and the others hold. Tesco died, and the buildings stopped. AB 2011 cannot die without an act of the California legislature, which neither the building trades nor the grocers will permit.

In 2008, British media was talking about overpopulation and how the UK was at the start of a baby boom, especially during a time of low immigration and emigration.

Now? Boris Johnson opens his latest Daily Mail column with bells tolling in Tuscany for far more funerals than weddings, kindergarten playgrounds eerily deserted in Tokyo, and bus routes cancelled across Europe as elderly passengers take “their last subsidised freedom pass to the sky.” He’s clearly proud of the fact that his party, post the Great Recession, had smothered UK’s nascent baby boom in the crib. Can’t have those pesky brats having their own brats after all.

Johnson is still among the most effective political journalists writing in English, and this column is him at the top of his form. The prose is fluid. The imagery lands. The dark humour earns its place. He’s making a contestable claim feel obvious through pure rhythm, and the reader is enrolled before they notice there is one.

His argument is that the falling birth rate isn’t a crisis but a blessing. The planet will thank us. The doom-mongers should butt out. The state should let people decide. On a first reading, it’s persuasive.

Now picture two households Johnson’s prose doesn’t visit.

A couple in their early seventies, in a paid-off four-bedroom house in Surrey. They bought it in 1987. The mortgage cleared a decade ago. The house is worth, today, somewhere north of £900,000. Their state pension arrived this morning, indexed by the triple lock, up another 4.1 per cent on last year. They aren’t rich. They’re comfortable. Their lives are working.

Their grandson is twenty-eight, with a 2:1 from a redbrick and six years of post-graduation work history without a permanent contract in any of them. He works through an app; warehouse shifts in Park Royal, courier runs, occasional fortnights of nothing. Last month he made £1,420. The month before, £820. He shares a flat in Walthamstow (still hopes to find a proper gig one day) and pays £760 in rent. He hasn’t been on a third date in eighteen months; no personal flat to bring someone back to, no schedule to plan around, no future he can compress into a sentence on an app. He’s the same age his grandfather was when he had his second child. He doesn’t think he’s falling behind. He thinks the country has moved past the point at which what he’s doing counts as starting a life.

Both households are doing exactly what people in their situation do. One is going well. One is failing. They’re connected. Johnson’s bells toll in Tuscany. They also toll, more quietly, over every son, grandson, and great grandson who never quite started.

Britain isn’t the first country to have this argument with itself

How we explain why the grandson hasn’t started determines what we do next. If he’s freely chosen not to want what his grandfather wanted, the policy response is to do nothing. Accept the transition. Plan for a smaller, older Britain. Adjust at the margins.

If he wanted to launch and couldn’t, the response is the opposite. We have to ask which costs were imposed, by whom, and to whose benefit: whether the policies that made his grandfather’s house worth £900,000 are the same policies that hollowed out the floor under his entry-level career.

The two diagnoses produce opposite governments. The first sees a generation living differently. The second sees a generation taxed by an arrangement that was never put to a vote. Pension solvency, productivity, housing reform, the welfare state, and most concretely whether the grandson ever crosses into the life that surveys say he still wants: none of these stakes are abstract.

The cultural story, and why it doesn’t hold

The dominant explanation for falling Western birth rates is cultural. Women have entered higher education and the workforce in numbers their grandmothers couldn’t match. Marriage has been delayed or abandoned. Religious observance has weakened. Johnson gives this story its sunniest reading: the human population, he writes, is “organically self-regulating, seeking a better balance with Nature.”

Trouble is that the cultural-shift theory cannot absorb one specific, well-documented finding: desired fertility, what people say they want when surveyed, has barely fallen. Across the OECD, the modal answer to “how many children would you like to have?” remains two or three, almost identical to 1970. Achieved fertility has fallen sharply. The gap between what people want and what they get is the largest ever measured.

If people had simply changed their minds about children, the survey numbers would have fallen with the birth numbers. They haven’t.

Cultural change is also very suspect. British women were having multi-decade record breaking numbers of kids during the late 00s, with Gen X having more kids than their Boomer counterparts. The ‘00s was also a time of ever increasing education and women in the workforce. So, it clearly isn’t caused by women having degrees and careers.

What changed about the price?

France and Britain, side by side

France and Britain are, by every variable the cultural theory tracks, the same kind of country. Rich Western European democracies. Comparable women’s labour-force participation. Comparable secular trends. The same global currents. By the theory, they should be on the same demographic trajectory.

For two decades they weren’t. France’s TFR held above 1.8 from the early 2000s through the mid-2010s, among the highest in Europe. Britain’s drifted down toward 1.4. A large gap, and on the cultural theory inexplicable.

From the early twentieth century, consolidated in the post-war settlement under the Conseil National de la Résistance, France ran a deliberate structural pronatal regime. It wasn’t cash bonuses or Mussolini-style rhetoric. It was a coherent system. Two components carried most of the load. The first was universal heavily-subsidised crèche provision from infancy. The second was the quotient familial, a tax code that divided households by number of children, so a French family with three children paid significantly less income tax than the same family with one. Universal allocations familiales paid every family regardless of income. The French state treated family formation as infrastructure. Lyman Stone, the leading conservative demographer of pronatal policy effectiveness, titles his analysis plainly: Does Pronatal Policy Work? It Did in France.

The regime is currently being dismantled, and the dismantling has a name: Emmanuel Macron. The cuts began under François Hollande, where Macron served first as deputy chief of staff (2012–2014) and then as Minister of Economy (2014–2016). The quotient familial cap was lowered twice, from €2,336 to €1,500. The universality of allocations familiales, established in 1946, was broken in 2015 by income-based modulation. The Cour des comptes quantified the cumulative impact at €3.3 billion. As President from 2017, Macron extended these cuts. In 2024, despite proclaiming a réarmement démographique, he reduced congé parental from three years to three months. APLs were cut in 2017, frozen in 2018. The 2018 loi ELAN began the privatisation of HLM stock, gutting the SRU social-housing quota. The 2017 Ordonnances Macron expanded fixed-term contracts, capped prud’hommes wrongful-dismissal awards, and introduced the CDI de chantier, a project-based “permanent” contract paying no prime de précarité.

France’s TFR has fallen from 2.0 in 2014 to 1.62 in 2024, a 20 per cent collapse in a decade. 663,000 children were born in France in 2024, the lowest figure since the Second World War, and for the first time since 1945 France recorded more deaths than births in the year to May 2025. INED’s Generations and Gender survey shows ideal family size dropped from 2.5 children in 2005 to 1.9 in 2024. The regime worked while intact; France’s collapse is what the structural thesis predicts when you take it apart.

Britain ran the inversion for longer, and from earlier. Childcare costs rose from roughly 40 per cent of the total cost of raising a child in 2012 to around 60 per cent today, among the highest net rates in the OECD. The planning regime restricted housing supply while existing-home prices rose by an average of £100,000 per home in the seven years before 2008, and continued rising afterward. Real wages stagnated from 2007 onward. The under-thirty-five cohort, which owned 3.2 per cent of the country’s housing stock in 2008, hasn’t recovered share since. A young British family pays for the same inputs a young French family pays for, gets less help, earns roughly the same.

A country’s fertility mechanism operates at two stages. The first is the launch (whether a young couple ever has a first child), and the strongest individual-level predictor here is employment stability, with stronger effects for men than for women. Alderotti, Vignoli, Baccini and Matysiak’s Demography meta-analysis (2021) finds men’s unemployment the single most fertility-suppressive labour-market condition, with the effect intensifying since the Great Recession and biting hardest where welfare protection is weakest. German workers on fixed-term contracts have a twenty-percentage-point lower probability of childbearing than those on permanent contracts, and the gap persists more than a decade after labour-market entry. The McDonald’s shift, the zero-hours job, the CDI de chantier: these don’t just pay less. They suppress launch.

The second stage is parity progression (whether a couple has a second, third, or fourth child), and here the conventional wisdom is wrong. Doepke, Hannusch, Kindermann and Tertilt’s The Economics of Fertility: A New Era (CEPR summary) documents that the long-standing negative correlation between income and fertility has flattened or reversed across the rich world. Where structural conditions allow, higher income now predicts more children, not fewer. The facilitating conditions (supportive family policy, cooperative fathers, favourable norms, labour markets that don’t punish parenthood) are precisely the levers France pulled until 2014, and Britain hasn’t pulled this century. Cap benefits at the second child, freeze working-age welfare, leave childcare uncovered, end the universality of allocations familiales, and you suppress parity progression.

Now Hungary, the case Johnson names directly. He calls it a failure. The data don’t. Hungary’s TFR rose from 1.25 in 2010 to 1.59 in 2021 (when the COVID Crisis was in full swing and before Europe’s mini oil shock). Critics correctly note period TFR is noisy and can reflect timing rather than completed family size. Stone responds with tempo-and-parity-adjusted fertility (the gold-standard measure that strips out timing effects) and finds Hungary one of only two countries (Slovakia is the other) with documented increases in completed cohort fertility since 2014. His difference-in-differences analysis of seventeen countries’ pronatal interventions from 2000 to 2025 finds an average effect of 0.09 to 0.18 additional births per woman, persisting more than a decade. Mongolia, where pronatal incentives were expanded and later eroded with fertility moving correspondingly each time, is the cleanest natural experiment in the literature.

Cash baby-bonuses work; it’s more of a question if the support for said policy is sustained, structural family-policy regimes producing durable fertility differences over time. France-before-Macron is the long-run version. Hungary is the targeted-intervention version. France-after-Macron and Britain are these regimes running in reverse, producing the convergent outcome we now see.

The cultural theory can’t explain any of these gaps. The structural theory can. It’s also the one almost no one in the relevant coalitions is willing to act on.

Generational Asymmetry

The grandson in Section 1 isn’t a hypothetical, and he isn’t the dramatic case. He’s an example of a post-2008 young British man without an elite-credential pathway. The literature has spent thirty years explaining why his case is asymmetric: the man’s economic position is the load-bearing variable in whether a couple ever crosses into family formation.

Alderotti meta-analysis tucks this in carefully: men’s unemployment is the single most fertility-suppressive labour-market condition, with effects measurably stronger than women’s. The mechanism has a long pedigree, going back to William Julius Wilson’s When Work Disappears (1996), which argued that disappearing male wage jobs were the load-bearing cause of family disintegration in inner-city Chicago. Becker’s specialization model gives a complementary economic account.

Autor, Dorn and Hanson identified this cleanly. In When Work Disappears: Manufacturing Decline and the Falling Marriage-Market Value of Young Men (2019) (yes yes, I know it using the same name as WJW’s When Work Disappears), they exploited the gender-specific composition of the China trade shock. Regions hit by male-intensive industry shocks saw marriage fall, fertility fall, and the share of unwed mothers and below-poverty single-parent households rise. Regions hit by female-intensive shocks saw the opposite. Same external shock, opposite domestic effects, depending on which side of the household economy got hit. The asymmetry isn’t a curiosity. It’s the mechanism. Where you gut the male wage floor, the family doesn’t form. The same paper documents a differential rise in male premature mortality from drugs and alcohol in the affected regions: Case and Deaton’s deaths-of-despair trajectory, traced to the same labour-market shock that suppressed launch.

What did Britain do, post-2008, to the male wage floor?

Real average weekly earnings didn’t return to their April 2008 level until 2024, sixteen years of stagnation unprecedented in modern peacetime. Young men under thirty-five bore disproportionately more of this than other groups, because the contraction landed on the rungs they were trying to climb. Zero-hours contracts, by 2024 the main job for over a million UK workers, are concentrated among the young and overrepresented in male-dominated low-skill sectors. The Resolution Foundation reports job mobility (the engine of early-career wage growth) is lower than in the 1980s. Apprenticeship completions in male-coded trades fell across the 2010s as further-education funding for non-academic pathways was gutted under austerity. A redbrick degree, once a near-guarantee of a salaried entry-level rung, no longer reliably opens the door. Graduate vacancies in 2025 fell to their lowest level since 2018.

By February 2026, 957,000 young people in the UK were not in employment, education, or training, up by a quarter of a million in three years, with the increase concentrated among young men. Economic inactivity due to ill-health among working-age men under thirty-five has roughly doubled since 2019. Some is pandemic tail; most is structural: a generation taught to expect the labour market has no place for them.

This is darker than the wage data alone. Drug-poisoning deaths in England and Wales rose 81.5 per cent between 2012 and 2022, with men accounting for the overwhelming majority. The 2024 male suicide rate is the highest since 1999, and suicide remains the leading cause of death for British men under fifty. Scotland, hit hardest by deindustrialisation, now has the highest drug-death rate in Europe, with the excess concentrated in deprived male cohorts under forty-five. The pattern Case and Deaton documented in West Virginia is the pattern Wilson documented in 1980s Chicago is the pattern now visible in Blackpool, Glasgow, and the post-industrial North-East. Different accents, same shock.

The grandson is, statistically, on a less stable contract than female graduates of his cohort, and earns less in real terms than his grandfather did at the same age. He is more likely to be NEET than a female peer, and much more likely to die by suicide, drugs, or alcohol before forty than any recent cohort of British men. He is also the variable on which any future decision to have a child most heavily depends. Women want stable partners at the end of the day!

This is why he hasn’t been on a third date in eighteen months, and why telling him to relax (telling him the falling birth rate is “organically self-regulating”) is so glib. The architecture suppressed family formation by dismantling the floor under the people whose economic stability is the strongest individual-level predictor of whether a family ever begins. The grandson’s empty schedule is its signature.

A pronatalist column that doesn’t see this isn’t seeing the problem. Johnson reads the grandson’s stalled life as preference. It is, more often, the absence of a launch, and the launch was sabotaged at the male economic position. No British minister announced that the goal was to immiserate young working-class men. The two-child cap, the fixed-term contract regime, the housing shortage, the wage stagnation, the gutting of further-education funding for non-academic trades: each was defended on its own terms. Their joint product is a cohort of young men ground down at the point in life they would, in any previous British generation, have been earning enough to become someone’s husband. Macron’s France ran the same play in another tongue. The Ordonnances Macron read almost as if designed to suppress male launch directly: project-based contracts paying no precariousness premium, weakened wrongful-dismissal protection, hitting precisely the male-dominated trades.

There’s one more thing the structural literature won’t quite say. Taxing the inputs to family formation produces fewer families. It also produces a generation of men who don’t see what’s happening to them as a structural condition. They see it as personal failure. They shouldn’t blame themselves; the architecture is doing the work. But the architecture is invisible by design, and the failure is concrete. A number of them go looking for somebody to blame. The post-2010 turn in young male political behaviour across the rich democracies (toward Trump, AfD, Le Pen, Reform) reads at one level as a generation of men who can feel a floor missing under their feet and have been offered no honest account of who removed it. The far right tells them a story. It’s a wrong story, and the centre-left coalitions that nominally oppose it haven’t told them a better one. Johnson and his contemporaries are complicit in the demographic outcome, and in the political one downstream of it.

That’s the darker current under the falling birth rate. Underneath the demographic numbers, a generation of men is being processed by a policy regime that benefits older asset-holders into a condition the polite analyses don’t quite want to name. Demographers call it suppressed launch. Sociologists call it deaths of despair. The men themselves say the country has moved on without them. They are right.

Patterns repeat across the rich world

The France-Britain comparison doesn’t stand alone. It sits inside a pattern documented across the rich democracies.

The intergenerational wealth gap isn’t a British peculiarity. The Institute for Fiscal Studies and the Resolution Foundation arrive independently at the same finding: household wealth roughly doubled between the 1990s and 2020s, with gains going overwhelmingly to older asset-holding cohorts. UK 25-to-34-year-olds would need to become five and a half times wealthier over the next thirty years to match the wealth of the typical baby boomer at the same age. Those born in the 1980s are on track for lower rates of homeownership than any UK cohort since the 1930s.

The same shape appears across the United States, Australia, Canada, and the Netherlands. Wherever housing functions as the primary retirement asset and post-2008 QE was layered onto an asset-skewed economy while pensioner benefits were ringfenced, the redistribution followed. The British settlement is the most aggressive version. France is now the same regime in French.

Every country running this version reports the same thing about its young people: delayed family formation, widening gaps between desired and actual fertility, falling birth rates the cultural narrative cannot account for. There’s nothing anomalous about the British case. It’s what every other rich democracy is now becoming, in different accents.

Austerity architecture, admitted

George Osborne, who as Chancellor presided over the British post-2008 settlement, was once asked about the distributional effects of QE. His answer was unusually candid: “[QE’s] purpose is that people who already have something like a house or a pension or some shares will see the value of those things go up. That’s what the policy is designed to do and it feels unfair because the people who don’t have those things aren’t getting the benefits of the policy.”

That’s the architect explaining what he built. Were the admission about almost anything else, it would have made front pages; because it was about quantitative easing, the line lived briefly in the financial pages and disappeared. The policy was intended to inflate the asset base of those who already held one. We owe him the courtesy of believing him.

QE alone doesn’t produce the regime; QE was paired with austerity, and we should be more precise about what that austerity was. The cuts of the 2010s weren’t uniform. Pensioner benefits were ringfenced and indexed upward by the triple lock. Working-age welfare, social housebuilding, childcare subsidies, housing benefit, further-education funding, local-authority youth services: those were defunded. The LSE Centre for the Analysis of Social Exclusion showed services directed at young people absorbed the largest share of cuts while transfers to retired households were largely protected. The European Youth Forum’s Generation Austerity report documented the same pattern across Europe. What was sold as fiscal discipline was, in distributional fact, Youth Austerity: a contraction whose burden was selected by age. Greece, Spain, Italy ran their own variants under EU rules. The US ran a softer version through the housing crash and state-level cuts. France ran its own, slowly under Hollande and aggressively under Macron. Everywhere the same shape: asset support upward, contraction across the working-age life-course.

Why does this austerity architecture persist? Because the actors operating it respond to incentives. Older voters turn out at high rates, are more numerous in most rich democracies through at least 2040, hold assets whose value is sensitive to policy, and are the median voter on most distributional questions. They vote their interest. A homeowner who paid off her mortgage in 1995 has nothing to gain from cheaper housing in 2026 and a great deal to lose. This isn’t villainy on her part. It’s how coalitions work. Meloni’s pronatalist rhetoric is loud, her policy quiet, because the policy that would help would cost her own coalition. Johnson’s “let people decide” is the same trick in English.

Johnson’s record while in office is the most damaging thing about his article. On employment stability, the lever academic research identifies as the strongest predictor of whether a couple ever has a first child, his government oversaw a record expansion of zero-hours contracts: over a million UK workers were on them as their main job by 2024, the highest level on record, disproportionately concentrated among the young. On family benefits, the lever for parity progression, his government retained Osborne’s two-child benefit cap, the most explicitly anti-pronatal piece of legislation in modern British history. On housing supply, his 2020 Planning for the Future white paper, which would have introduced zonal planning to break the local-veto regime, was abandoned in 2021 under backbench pressure following the Chesham and Amersham by-election. The 300,000 homes annual target was scrapped under his successor for the same coalitional reason. The man writing in 2026 to tell young people to relax about birth rates is the man who held the levers academic research identifies as causal, and chose, at the political moment of decision, not to pull them.

Macron is the same story in another language. The pronatalist rhetoric is loud: réarmement démographique, the emphatic concern for la natalité française. The policy runs the other way: a quotient familial cap lowered twice while Macron was Hollande’s deputy chief of staff, the universality of allocations familiales broken under his watch as Minister of Economy, APLs slashed and frozen, the Ordonnances Macron expanding fixed-term and project-based contracts, congé parental cut from three years to three months. The French TFR responded by collapsing on the schedule the structural literature would predict.

The verdict

The British and French birth rates are now falling at the same rate, with at least four distinguishable causes. The deepest is structural: aging democracies, once their median voter passes a certain age, function as coalitions of asset-holders and produce policy that protects assets at the expense of the inputs to family formation. Below that, post-2008 monetary and fiscal policy converted the structural advantage into outcome: QE inflated the assets the older cohort owned, Youth Austerity cut the investment the younger cohort needed, and the cut fell asymmetrically on young men. Below that sit specific national choices: the triple lock, the two-child cap, the broken universality of allocations familiales, the Ordonnances Macron. Each could have been made differently. Hungary made some of them differently and got a different birth rate. The most contingent layer is rhetorical: Johnson’s column, Macron’s réarmement démographiquespeeches, a senior member of the winning cohort telling the losing cohort that everything is fine.

Fix only the rhetoric and you change nothing. Fix the structure and you fix the rest.

This is, finally, a piece about young people, so let me close by speaking to them directly.

What’s being denied to your generation is the oldest material precondition for romance: a flat with a second bedroom, a contract that lasts past Christmas, a future you can compress into a sentence for someone you might want to spend it with. The desire is real. Surveys across the rich democracies keep returning the same answer: young people still want partnership, still want children, still want roughly what their grandparents had. What’s denied is the material floor on which any of that gets built. The grandfather had his second child at twenty-eight because the labour market handed him the platform. The grandson on a zero-hours contract doesn’t get the platform, and so doesn’t get the second date.

You may be a conservative young man who reads Johnson and finds parts of his diagnosis on welfare correct. You may be a centrist young woman who voted Macron in 2017 because he seemed like the adult in the room. You may be a young progressive who turned out for Sánchez or Starmer or Harris because the alternatives looked openly cruel. The political differences between you are real. The material situation underneath them is identical, and the older members of each of your coalitions know it.

The conservative young man’s preferred mix won’t solve his housing problem if his coalition’s older voters don’t want lower house prices. Johnson has held the levers and didn’t pull them. Trump won 54 per cent of young American men in 2024 on the promise of letting them launch; eighteen months in, his approval among Gen Z sits in the mid-twenties, with job market, cost of living and healthcare cited as the breaks.

The centrist young woman’s preferred mix won’t fix her childcare problem if Macron is the centrist who lowered the quotient familial cap, broke universal allocations familiales, slashed APLs, and cut congé parental from three years to three months.

The young progressive’s preferred mix is, as it happens, what the structural literature actually endorses; the centre-left incumbents aren’t delivering it. Sánchez’s PSOE-Sumar government has presided over a “booming economy with falling unemployment” alongside an affordable-housing crisis that has gutted young Spaniards’ life prospects, and the young Spanish vote has fragmented across the radical left, Vox, and abstention. Germany’s SPD watched young men under twenty-four make AfD their preferred party in 2024. Starmer’s government is bleeding the under-thirty vote.

If you’re under thirty-five, you’ve watched your peers cycle through these vehicles. They voted left in 2017, centrist in 2019, hard right in 2024, and many have already turned on the right by 2026. The pattern isn’t stupidity. It’s a generation looking for the political vehicle whose elders will let them have what their grandparents had: a job, a flat, a partner, a child, a future they can put into a sentence. They keep moving, and the movement keeps being read as cultural noise. It isn’t noise. It’s the only signal a generation that has been told to relax has left to send.

The older members of every camp share something their disagreements obscure: they view your generation as economically expendable. The Tory pensioner, the PSOE pensioner, the SPD pensioner, the homeowner among American Democrats: different in their politics, identical in their economic incidence. They will tolerate your loss before their own. They are not lying when they tell you the falling birth rate is a blessing, that you should relax. They are telling you the truth. Why bother after all? Nothing the youth can do to stop them.

You are not in a coalition with the pensioners who out-vote you. The young Vox voter in Andalusia, the young Reform voter in Hartlepool, the young Mélenchon voter in Marseille, the young Trump voter who has already turned away from him: their differences are political and worth arguing about. Their material situation is the same. So is the future their elders are quietly content to let them lose.

There is a lamp lit above the firehouse on Main Street, USA, and it has been lit since 1955.

Walt Disney lived in the apartment above the firehouse during construction. The lamp signaled, by company convention, that he was in residence. He died in 1966. The lamp stayed on. It is still on.

I’m a bit too obsessed with this little detail. I mean, about fifty-six thousand people walk past it on a typical day. Past pastel storefronts, a horse-drawn omnibus, the smell of vanilla pumped through the doors of the Confectionery, gas lamps that came from Baltimore and were bought by the pound. Past the most successful piece of pre-war American small-town urbanism in continuous operation in the country. Tens of millions of Americans pay every year to walk through it.

It almost did not get built, because the project was irrational and a bit whimsical.

When Walt pitched Disneyland to his brother Roy and the board of Walt Disney Productions in 1952, the board said no. The company was already overextended on three other risky bets: high-budget live-action films, nature documentaries, and television. Estimated cost of the park, ten to twenty million dollars. Company net profit the prior year, $720,000. Banks told Walt the outdoor amusement business was “a cultural anachronism that had already declined into senility.” That was the actual phrase. Cultural anachronism. Declined into senility.

Walt did it anyway, in a way that today would be career-ending. He took a $60,000 loan against his life insurance through Commerce Bank in Kansas City, pledging the financial safety net of his wife and children as collateral. He sold his Palm Springs vacation home. He set up a personal corporation called WED Enterprises (his initials) outside the studio’s risk-management process so the board couldn’t kill the project. And he gave ABC television an equity stake in the park in exchange for the network distributing Disney programming and putting up roughly seventeen million dollars in construction money.

The park opened on July 17, 1955. By the end of its first year it had attracted 1.2 million visitors and was profitable. Disney bought ABC out five years later. The bankers were wrong about cultural anachronism.

Sixteen years later, Walt and Roy would build a place forty-six times the size. To pencil it, they had to convince the Florida state legislature to invent a county.

What does it take to build something big in America?

Most of what I write about is the small stuff: ADUs, infill, and sometimes reviving the downtowns of small cities and towns. The political economy supports that work right now, and what gets built matters. But it isn’t everything I want. I want more Main Streets across America, at a greater scale than I usually talk about. I want a country that still knows how to plan and build at Disney-World scale, places imagined (and even ridiculed for the absurdity of them) before they get reduced to spreadsheets. The political economy for that kind of ambition doesn’t exist in 2026, and I know it.

So. Government-engineered financing vehicles are how more ambitious urbanism actually gets built in this country. Private capital, by itself, cannot do it, and never has. Disneyland required Walt to mortgage his family’s safety net. Disney World required the Florida legislature to invent a special-purpose taxing district with bond-issuing authority. The 1968-73 apartment boom required a federal stack of Section 236 interest subsidies, FHA 221(d)(4) construction insurance, and Ginnie Mae securitization. The 1981-86 apartment boom required ERTA’s accelerated depreciation. Every time. The pattern is structural.

At the same time, YIMBY victories like California’s ADU laws (which are responsible for nearly a quarter of all new units in the state), Oregon’s HB 2001, Montana’s middle-housing reforms, and Massachusetts’s Section 3A are real wins. Without them, this country would be building less housing than we do now. To deny their impact is wrong, and the people doing the denying are doing more damage to American housing politics than they realize.

What if we had a better political economy? What if we could build on top of those victories? What are the next steps?

I’d like to start with Paul Williams’ recent report for the Center for Public Enterprise. American multifamily starts have been stuck at roughly 350,000 units a year for forty years, through every conceivable interest-rate regime, zoning regime, and political cycle. The two moments when production exceeded 500,000 sustained were the HUD boom of 1968-73 and the ERTA boom of 1981-86. Both came directly out of federal credit and tax infrastructure. When the machinery was on, supply boomed; when it was off, supply collapsed. The flat line is a policy artifact, not a market verdict.

So the next step is to turn the federal credit apparatus back on. Five levers. Most cost little (for the deficit hawks reading). All build on infrastructure that already exists. They would, plausibly, raise multifamily production to 500,000-700,000 units annually for the decade we need to work through the shortage.

Is zoning reform enough?

The YIMBY/urbanist argument, which I share, runs something like this. Zoning prevents apartments from being built in the metros where they’re most needed. Single-family-only districts, parking minimums, height limits, FAR restrictions, discretionary review, and weaponized environmental procedures collectively make it illegal or financially infeasible to build the housing demand exists for. Williams uses a container metaphor for this: your zoning, permitting, and planning regime is the size of the container you hold out to collect rainwater. An unpermissive regime is a tablespoon. The right reforms are an upgrade to a gallon jug.

But the metaphor itself half-admits the problem. The container does not summon the rain.

YIMBYism, urbanism, Strong Towns, whatever you want to call it: the working coalition has always been more than willing to talk about what zoning reform can and can’t do. ADU and infill reforms often outperform other zoning reforms because homeowners or small businesses who want to redevelop will do so for personal or local reasons, without reference to the broader market. It’s only the more market-optimistic “abundance” framing that argues you only need to reform the regulatory regime and the supply will come.

Look at Austin. Through the 2010s, Austin liberalized its zoning more aggressively than most peer metros. The pandemic added a one-time domestic-migration shock and zero-percent interest rates. Multifamily starts in Austin surged. National multifamily permits broke 500,000 in 2021-2022, the highest since the 1980s, and a large share of that bump came from Sun Belt metros where zoning had been preemptively cleared. Then the Federal Reserve raised rates from zero to over 5% in eighteen months. Construction loan rates followed them up. The bump unwound as interest rates increased and the tech industry that powered Austin’s growth shed workers. Permits are now down roughly 25% from peak.

But notice what YIMBYism and its organizers actually accomplished. They upgraded Austin to a gallon jug. The jug worked exactly as advertised: when the financing rain fell, the jug filled, and rents fell. They are still falling, and they need to fall further to get Austin’s rent burden back to pre-COVID levels. Without the organizers, the council members, and the years of work behind them, Austin would have fewer units and be more expensive today than it is. The constraint that arrived in 2024 was not the size of the jug. It was the rain.

The lesson here is not “zoning reform doesn’t work.” Zoning reform absolutely works, and it is a prerequisite for everything else. The lesson is that zoning reform alone gets cyclically captured by the financing environment. When finance is loose, supply surges. When finance is tight, supply collapses. Net production over a decade ends up barely distinguishable from a non-reformed regime, and the YIMBY-coalition victories of the 2020s end up looking, in 2035, like Austin in 2024.

The flat line is older than YIMBY. It tracks the disappearance of federal credit and tax infrastructure that historically made apartment construction financially attractive at scale. Forty years, one number, untouched by anything that happened at the local level.

To see why federal financing infrastructure is the structural answer, look at how Disney World actually got built.

How did the Disneys actually pull it off?

1955: how did Walt finance Disneyland?

Disneyland’s financing, at $17 million, was modest by today’s standards. A present-day developer would consider $17 million a routine project. And yet, in 1955, $17 million was approximately twenty-five times the parent company’s annual net profit. The board wouldn’t fund it. Walt had to invent his own vehicle.

The vehicle was personal. WED Enterprises, set up outside the studio. A life insurance loan from Commerce Bank in Kansas City, backed by his family’s safety net. The Palm Springs sale. A line of bank credit through Bank of America, where Walt had banked since the 1930s. And the ABC equity deal, which converted television rights into construction money.

By the time the park opened, Walt had pledged his personal financial existence against the project. The board’s caution had not been wrong, exactly. The project was a bet that could have ended Walt’s career and his family’s wealth. It paid off because Disneyland turned out to be a perpetual cash machine, and the cash machine became the funding source for the next attempt.

But notice what this means. The most beloved piece of small-scale American urbanism the country has produced in the postwar period was financed by a single individual borrowing against his own life. The capital markets said no.

1971: why did Walt need a county?

A decade later, Walt’s ambitions outgrew the personal capital stack. He wanted 27,000 acres in central Florida, forty-six times Disneyland’s footprint, including Magic Kingdom, hotels, transportation infrastructure, and the planned EPCOT city he wouldn’t live to see. The estimated cost was $400 million. In 2026 dollars, roughly $3 billion.

The Disneyland playbook would not scale. You cannot finance a $400 million project through life insurance loans. The bond markets would not lend on a swamp. Orange and Osceola Counties had no power, water, or sewer infrastructure within ten to fifteen miles of the site. And the governance infrastructure for a 27,000-acre, multi-municipality, master-planned development simply did not exist in 1965.

Walt’s solution, executed with Florida’s governor Haydon Burns and finalized after Walt’s death by Roy, was Reedy Creek. On May 12, 1967, Florida Governor Claude Kirk signed Chapter 67-764, creating the Reedy Creek Improvement District: a special-purpose taxing district with the same authority and responsibility as a county government. The district could levy property taxes. It could provide municipal services. Most importantly, it could issue tax-exempt municipal bonds.

The Florida legislature included an unusual pledge in the enabling act. The state pledged to bondholders that it would “not in any way impair the rights or remedies of the holders ... until all such bonds, together with interest thereon, and all costs and expenses ... are fully met and discharged.” This pledge was the financial keystone of the project. It made Reedy Creek’s bonds state-backed in all but name, salable at low interest rates to risk-averse institutional investors. By the time DeSantis tried to dissolve the district in 2022, Reedy Creek carried roughly $1 billion in outstanding bond debt, serviced annually by approximately $58 million in district taxes paid by Disney itself.

The state was challenged in court on whether a special-purpose district mostly serving a private corporation was a valid public purpose. The Florida Supreme Court upheld the district in 1968, on the explicit reasoning that “tourism, recreation and the conservation of natural resources” were valid public purposes.

Combined with retained operating cash flow from Disneyland and conventional corporate financing, Roy completed Disney World on time and on budget for $400 million without taking on additional corporate debt. Magic Kingdom opened October 1, 1971. By every measure (visitor count, longevity, cultural durability) it’s the most successful piece of master-planned American urbanism ever built.

The thing that gets me about Reedy Creek is what was actually required. The most ambitious piece of urbanism in modern American history needed a custom-built state-chartered taxing district with bond-issuing authority and an explicit state pledge to bondholders. Without Reedy Creek, the project does not pencil. Even with Walt and Roy. Even with Disneyland’s cash flow. Even with Florida cooperating.

1984: did building Disney World almost kill the company?

Disney World didn’t insulate the Walt Disney Company from the consequences of building it.

By the early 1980s, Disney was capitalized for content and theme-park operations but thinly defended on equity. In March 1984, the financier Saul Steinberg purchased 6.3% of Disney’s stock and triggered a takeover scare. By June, Steinberg controlled 12.2% and offered to acquire 49% for $1.3 billion. At the time, that was more than the company’s available cash, which was under $10 million.

To fend Steinberg off, Disney paid $297.5 million in greenmail plus $28 million in expenses, took on $850 million in new debt, and acquired the real-estate firm Arvida and the greeting-card maker Gibson Greetings as defensive bulk. The company that built Reedy Creek almost became a forced-sale target precisely because building Reedy Creek had left it financially exposed.

Think about what this means for the argument. Even Walt and Roy Disney’s playbook (decades of operating cash flow, a state-engineered financing vehicle, the most successful master-planned urbanism in American history) left the parent company close enough to the edge that a single hostile financier with $11 million in initial capital could nearly take it down a decade later. The state vehicle was necessary. It was also not quite sufficient. The parent corporation absorbed risk that no rational private capital structure would have absorbed unaided.

If the country’s most resourceful urbanist, with the State of Florida actively in his corner and a perpetual cash machine in California funding him, still ended up exposed to a hostile takeover in 1984, the proposition that present-day urbanism can be financed by private capital alone isn’t merely wrong. It’s senile.

2026: could anyone build Disney World today?

What would it take to build Disney World today?

I keep trying to imagine the capital stack and I cannot construct it. The Reedy Creek mechanism was partially dismantled in 2022, and its successor, the Central Florida Tourism Oversight District, is governor-controlled. No state legislature is currently in the business of inventing custom special-purpose districts for ambitious urbanism. The federal credit apparatus that produced the HUD boom of 1968-73 was dismantled after Nixon’s 1973 moratorium. The tax apparatus that produced the ERTA boom was dismantled by the 1986 Tax Reform Act. Construction lending from regional banks has collapsed since 2023, with the post-SVB regulatory environment and the $1.8 trillion CRE maturity wall hitting simultaneously. Conventional construction credit is nearly impossible to price under current input-cost inflation and tariff regimes.

A 2026 Walt Disney could not finance Disney World. No state would invent Reedy Creek for him (even Texas with its MUD system); the political environment for doing so has been actively poisoned by the 2022 fight. Conventional banks won’t lend at terms that make the project pencil. And without a federal pledge equivalent to Florida’s 1967 pledge to Reedy Creek bondholders, he could not securitize his way through the bond markets either.

Has the federal government ever done this for apartments?

The HUD boom of 1968-73 was the federal version of what Florida did for Disney one year earlier. The Housing and Urban Development Act of 1968 set a goal of 26 million new housing units over ten years. To hit that target, HUD deployed three integrated tools that, taken together, were essentially Reedy Creek for apartments. Section 236 interest-rate subsidies that reduced effective mortgage rates on qualifying rental projects to as low as 1%. FHA 221(d)(4) construction insurance, which provided non-recourse, fixed-rate financing covering up to 90% of project cost. And Ginnie Mae securitization, issued in 1970 for the first time in American history, which connected FHA multifamily lending to global capital markets and removed lender capacity as a binding constraint on production.

Multifamily starts surged from 400,000 in 1968 to an annualized 1,000,000 units in March 1973. Three million apartment units in four years. Then, in 1973, the Nixon administration imposed a moratorium on new commitments. Section 236 wound down. The boom ended.

The ERTA boom of 1981-86 was the tax-code version of the same idea. The Economic Recovery Tax Act of 1981 created accelerated cost recovery for real estate (15-year depreciation with front-loaded methods) that made apartment investment a tax shelter capable of generating paper losses sufficient to offset other income. Multifamily starts climbed from under 300,000 in 1981 to over 500,000 by 1985 and stayed there for five years. Then, in 1986, the Tax Reform Act extended depreciation to 27.5 years, switched to straight-line methods, and added passive loss limitations. The tax economics of rental housing development collapsed almost overnight. Multifamily starts fell to 140,000 by 1991 and have never sustainably exceeded 500,000 since.

Every time American urbanism or housing produces at scale, a custom financing vehicle is created to make it possible. Disneyland in 1955 ran on WED Enterprises, an ABC equity deal, and a personal life insurance loan. Disney World in 1971 ran on Reedy Creek and the state bondholder pledge. The HUD boom ran on Section 236, 221(d)(4), and Ginnie Mae together. The ERTA boom ran on accelerated cost recovery and passive loss treatment.

Every time we dismantle the vehicle, production falls back to a baseline that is a policy artifact rather than a market equilibrium. The 350,000-unit flat line for multifamily starts (forty years long, through every conceivable macroeconomic environment) is what happens after we turn the machinery off.

The pattern matters more than usual right now because the credit cycle is at exactly the wrong point. Multifamily loan maturities surge from $104 billion in 2025 to roughly $162 billion in 2026. Builder bankruptcies are accelerating. Construction lenders have widened spreads or pulled out of the market entirely. We are sitting in the sharpest construction-credit collapse since 2008, with the federal apparatus off, and a $1.8 trillion CRE maturity wall arriving on top of it.

What would turning the machine back on look like?

Five levers. What follows is the technical version.

The federal credit apparatus that produced the 1968-73 boom is, in modified form, still sitting on the federal balance sheet. It has just been allowed to atrophy. Reactivating it does not require inventing new agencies. It requires turning existing infrastructure back on.

Lever 1: why does an FHA loan take a year to close?

On paper, FHA’s 221(d)(4) is unbeatable as a construction financing product. Fixed-rate, non-recourse, up to 90% loan-to-cost, 40-year terms. It’s the lineal descendant of the program that financed the 1968-73 HUD boom. Today it finances roughly 10,000-15,000 units a year, about 4% of multifamily production, a rounding error against the three to four million unit shortage.

Why? Process friction. Loans routinely take 270-360 days to close, against 60-90 days for conventional construction financing. HUD launched a “single underwriter” reform in 2014 that was never fully implemented. A 2024 OIG auditfound HUD still tracking applications across five regional centers on manual spreadsheets.

Williams’ fix: codify single-underwriter in the MAP Guide, fund the IT modernization HUD has been promising for a decade, expand categorical exclusions for infill, and resource the program like the production platform it could be. Cost: negligible. FHA mortgage insurance is statutorily actuarially sound; the investment is in HUD’s operational capacity. Scale: 50,000-75,000 units a year in normal markets, with upside toward 100,000+ in crunches. For context, in 2010, during the GFC’s private-credit retreat, FHA across all programs financed more than 150,000 rental units.

Underrated, because it skips Congress and appropriations entirely. All it needs is an HUD secretary who cares about execution.

Lever 2: what are the states already doing?

While the federal apparatus has atrophied, state housing finance agencies have been quietly building the next generation of construction finance infrastructure. MassHousing’s BILD program pairs preferred equity with Freddie Mac permanent debt at debt-like returns, targeting 25-35% of total project equity. New York’s Housing Acceleration Fundprovides subordinate construction loans through CDFIs and banks at 3% interest-only. Michigan’s Housing Accelerator Fund is structured similarly.

These programs collectively represent about half a billion dollars in capital and have worked at small scale. MassHousing’s BILD has a pipeline of 33 projects representing 7,000+ units against an initial $50 million fund.

The match: a $2 billion one-time federal contribution capitalizing similar programs in 25-30 states, structured to revolve so the funds recycle into new projects after permanent conversion. Cost: $2 billion one-time, ongoing zero. Scale: roughly 30,000 additional units a year. The mechanism crowds in private capital, leverages existing state HFA expertise, and produces mixed-income housing rather than just deeply-subsidized affordable units. It also creates durable production infrastructure that can operate counter-cyclically, which is what we need given the post-2023 bank pullbacks.

Lever 3: why aren’t Fannie and Freddie in this market?

Fannie Mae and Freddie Mac dominate multifamily permanent lending, together roughly 41% of total origination volume in 2024, but are essentially absent from construction financing. This is a historical artifact, not an inherent limitation. The GSEs’ charters emphasize permanent mortgages because that’s where their underwriting expertise developed. Construction has always been left to banks and insurers.

The fix has three parts. Forward commitments on permanent financing, structured like MassHousing’s FORGE Loan, which uses a 36-month rate lock to let developers secure permanent financing terms before breaking ground. Mezzanine and preferred equity products that fill the gap between senior construction debt and developer equity, currently served at high prices by CDFIs and impact investors. And a secondary market for construction-to-perm loans, enabling true scale through MBS structures.

Cost: essentially zero from a federal-budget perspective. GSE products are priced to be self-funding. The policy lever is regulatory (FHFA scorecards, possibly a charter clarification from Congress), not fiscal. Scale: 50,000-100,000 units a year depending on portfolio share. This is the lever that does the most to address the post-2023 regional bank pullback, because it puts patient capital into the gap that private credit has vacated.

Lever 4: could the tax code do for housing what it did for solar?

The 1981 ERTA boom was driven by 15-year accelerated depreciation. The 1986 Tax Reform Act extended that to 27.5 years straight-line, and production collapsed within five years.

The proposal is to restore some version of accelerated treatment. Three options, ordered by aggressiveness: shorten the straight-line schedule from 27.5 years to 15-18 years; partially expense the first $100,000 per unit; or fuller expensing under tighter passive-loss rules. Today’s 37% top marginal rate is well below the 50% rate that prevailed during ERTA, which materially reduces both the per-dollar value to claimants and the revenue loss to Treasury. We can get much of the production effect without recreating the worst tax-shelter abuses that drove the 1986 backlash.

The closest analog is the IRA’s clean-energy production-tax-credit and investment-tax-credit structure. We just demonstrated, on the energy side, that this approach can mobilize hundreds of billions of private capital toward an industrial-policy goal in a few years. The mechanism translates to multifamily.

Cost: real but bounded. Shortening the depreciation window costs less than partial expensing, which costs less than full expensing, with corresponding production effects. Estimates range from a few billion annually for shorter straight-line to tens of billions for full expensing. Scale: 50,000-100,000+ units a year. This is the most fiscally substantive lever in the package and also the one with the cleanest recent precedent.

Lever 5: what is the FHLB system even for?

The Federal Home Loan Bank system is a $1.3 trillion government-sponsored enterprise that, per the CBO’s 2024 analysis, receives roughly $6.9 billion a year in implicit federal subsidy through the spread between FHLB consolidated obligations and Treasuries. Almost none of this flows into housing production. The 11 regional FHLBs primarily make collateralized advances to member institutions, totaling $737 billion in 2024. Affordable Housing Program contributions in 2024 were about $751 million, roughly 12% of net income.

In 2023, Silicon Valley Bank, Signature Bank, and First Republic Bank held tens of billions in FHLB advances days before failing, using the system as a lender-of-next-to-last-resort rather than a housing-finance institution. The FHLBs were repaid in full because their statutory super-lien gives them priority over the FDIC.

The fix: amend the FHLB Act to authorize construction-to-permanent multifamily lending, either directly or through member institutions. FHFA would set risk-based capital requirements and a 10% portfolio target for housing construction and permanent financing. The bipartisan CURB Act shows there is appetite for FHLB reform across the aisle.

Cost: zero federal-budget impact. The FHLBs are self-funding. Scale: 50,000+ additional units a year at a 10% portfolio share, supporting roughly $70 billion in housing investment. This creates permanent counter-cyclical infrastructure that lends into downturns when private capital retreats.

Why isn’t the technical case winning?

You can make a strong technical case for building lots of housing. The political case isn’t. “Build more apartments” has been the YIMBY ask for over a decade and has produced state-level wins, but at the federal level, where the credit apparatus actually lives, the slogan has been weak. There is no political coalition for “more apartments” that competes effectively with organized homeowner interests and the institutional capital that profits from the current product mix.

This is, at its root, a problem of register. At the national level, the case for housing supply has been argued in the language of supply, and the language of supply is technocratic, demographic, and a little bit boring. The arguments are correct. They have not produced federal credit reform. They have not produced a coalition broad enough to break the forty-year flat line.

The rest of the country isn’t listening because the rest of the country doesn’t read white papers. The rest of the country reads memory. Memory of how a place felt, of what your grandparents could afford, of the courtyard you played in when you were nine. The rest of the country has feelings about a 1925 courtyard in Pasadena, or about the lamp above the firehouse on Main Street USA, or about the 2BR their grandparents raised three kids in and that no developer in their current city would dream of building. Those feelings are politically active, and at the moment they’re being routed almost exclusively through reactionary politics, because reactionary politics has been more willing to speak the register of memory than YIMBYism has.

This is, plainly, a self-inflicted wound, but the wound belongs to the national policy conversation, not to the organizers who have actually been winning. Talk to anyone who knocked doors for California’s ADU bills, or who packed a city council hearing in Montana, and you’ll find people who already know how to speak the register of memory and loss and aspiration. The organizing has always been emotional. The white papers came after.

And I want to be specific about a particular dynamic, because it has gone uncalled-out for too long. There is a class of national-level housing commentator, usually writing for some big publication like the NYT or something small like a substack, occasionally from a think tank, rarely from a city where they’ve ever had to flip a council vote, who has spent any sort of time arguing that state and local YIMBY wins haven’t actually moved the needle. That California’s ADUs aren’t real urbanism. That Montana’s middle-housing reforms are theater. That Oregon’s HB 2001 didn’t change anything. That the whole movement needs to reset, get serious, get federal, get something. The argument has the causation exactly backwards. Those wins were the largest legalization expansions in fifty years of American housing policy. They were also the only consistent housing-policy victories of the last decade, full stop. They were earned by people who packed council hearings, knocked doors, got primaried, sat through six-hour zoning debates, and absorbed personal abuse from neighbors who showed up to call them slumlords for wanting to allow a duplex. The framing that “not enough is being built” is technically true and rhetorically poisonous, because it concedes the only working model American housing politics currently has to whoever shows up next claiming a better one. If you want more housing built, the organizers who have actually been winning are not the problem. They are the proof of concept. Defend their work like your political project depends on it, because it does.

And here is the thing the national discourse rarely says out loud. Local and state YIMBY organizing is, by every standard of rational self-interest, delightfully irrational. The expected personal benefit of any given organizer’s hours at a council hearing is somewhere south of zero. Show up, get yelled at, lose a Saturday, possibly lose a neighbor, possibly lose a friend, possibly get primaried, in exchange for what? A marginal bump in the probability that a duplex gets allowed three blocks from your house, in a city you might leave in five years anyway? Mancur Olson wrote a whole book about why this kind of organizing isn’t supposed to happen. The logic of collective action says every rational actor should free-ride.

And yet here they are, year after year, going to the hearings, knocking the doors, doing the thing the rational ledger says they shouldn’t (among other things), and producing the wins, despite the NIMBYs’ biggest support base is often the upper crust, in addition to the upper middle class Boomers. The largest developers consistently shown to be fair weather friends at best, and often opposed to the YIMBYesy candidates (Tom Steyer for example) .

The rational version of the movement, by contrast, is the national abundance commentariat: white papers, Substack posts, conference panels, no Saturdays lost. It has produced excellent analysis. It has not produced housing. The same was true of Walt Disney mortgaging his life insurance to build a park the bankers called a cultural anachronism. Big things get built by people willing to do what the rational ledger says they shouldn’t. So do small things.

The gap is at the national level, where the housing supply argument gets laundered into technocratic language before it reaches the people who might fund it or legislate it. Cost burdens, vacancy rates, permits per capita, whatever. None of that is how anyone ever got a state legislature to invent a county. The frames below are suggestions, not instructions. Organizers don’t need instructions from anyone. The frames are sketches, three of many possible ways to carry what organizers know into rooms where the conversation has lost its nerve. Treat them as starting points, not a definitive list.

What the frames share is the willingness to be moved. That posture is sometimes called nostalgia, sometimes whimsy, sometimes romance. I’m going to use whimsy and romance because they’re the most accurate. Whimsy is advocating for things because they’re charming. Romance is advocating for things because we used to have them and we miss them. Both are politically powerful in ways the supply argument has not been.

Frame 1: why can’t your daughter afford a 2BR?

You know a 28-year-old like this. She’s your sister, your daughter, your friend’s kid, the person at the next desk. She has a partner. They’ve been together for four years. They want to get married, they want to have kids, and they want to do it in the city where they actually live, which is the city where their jobs are. They make decent money. Combined, more than her parents made when they bought their first house. They cannot find a 2BR they can afford. They’ve been looking for two years. The 1BRs they can afford have a den that her partner uses as an office, which means the kid would be in their bedroom, which means they aren’t going to have the kid yet, which means they aren’t going to get married yet, which means the whole thing keeps getting pushed back, six months at a time, for what is starting to feel like the rest of their twenties.

Multiply her by twenty million. Under-35 household formation has cratered. The median first-time homebuyer is in their late 30s. Marriage and first-birth ages are at all-time highs. The classic American life script (meet someone, partner up, household, kid, equity, roots) is being delayed. Young people still want it. They cannot afford the shelter it requires.

Couillard’s 2025 paper finds that half the U.S. fertility decline since the 2000s is attributable to housing costs, with subsidies for 3+BR construction generating 2.3x more births per dollar than equivalent 1BR subsidies. The mechanism is not subtle. Families form in apartments with rooms for children. The apartments without rooms for children are what the current capital structure builds, because they pencil. The levers above can be tuned toward family product at near-zero administrative cost: a fast-track FHA processing tier for projects above some 2+BR threshold, GSE pricing tilts for above-baseline family-unit shares, depreciation schedules that favor larger units. Every lever, retuned slightly, builds the apartments where the 28-year-old can finally start her life.

The coalition this frame assembles is the broadest of the three. Young-adult voters who can’t launch. Their parents, who watch them not launch. Family-policy moderates on both sides of the aisle. Pronatalist conservatives who want the birth rate to go up. Affordability progressives who want renters to not be crushed. The frame is romantic about a particular thing, which is the American life script: the script that says you should be able to meet someone and marry them and have kids with them and put down roots in the city where your work is, and that the country owes its young people the conditions under which that script remains possible. The script is contested in the discourse. It is not contested in the actual preferences of under-35 Americans. They want it. They are being denied it. These levers are how we give it back to them.

Frame 2: do you remember when America built things?

Walt Disney was 51 years old, in declining health, with a board that opposed him and bankers who told him his project was a cultural anachronism. He built Disneyland anyway. Five years later he was already planning Disney World, forty-six times the size, and personally lobbying the Florida state legislature to invent a county to make the bond math work. Roy finished it on time and on budget for $400 million.

Walt is the patron saint of building big in America. Not “build a corporation.” Not “extract returns.” Not “exit at stabilization.” Build the place itself, at the scale the vision actually requires. And he wasn’t alone. The country we built before 1986 was a country full of Walts. We built the Interstate Highway System and the post-war suburbs. We built the master-planned new towns of the 1960s and 1970s, places like Reston and Columbia and Irvine and the Woodlands, designed from scratch and now home to hundreds of thousands of people. We built the HUD-boom apartment stock that put a roof over three million families in four years. The TVA dams. The L.A. aqueduct. The Brooklyn Bridge had been built two generations earlier, and yet here we still were, building. We built at scales that present-day America looks at and assumes must have been done by a different species.

We weren’t a different species. We were the same country, with the same union contracts, the same bureaucratic dysfunction, and the same political rancor. What we had that we no longer have is the machinery. Federal credit infrastructure. Tax policy that pointed at production. State legislatures willing to invent counties when counties needed inventing. The HUD secretary in 1968 had tools the HUD secretary in 2026 does not. The result is that Walt’s playbook, the playbook of all the postwar urbanists, doesn’t run anymore.

This is the romance of a lost capacity. It runs on the felt sense, which I think a lot of Americans have, that we used to be a country that could do things, and we’ve forgotten how. The forgetting is correctable. The correction is a tune-up of dormant infrastructure that previously produced, and could again produce, the construction curve that built modern America.

The coalition this frame assembles is the strangest and possibly the largest. Abundance liberals. Nostalgic conservatives. Infrastructure-state Democrats. The post-liberal right. The technocratic center. The people who read Marc Andreessen and the people who read Ezra Klein and the people who read both. Strong Towns sympathizers. Anyone who has stood in a piece of postwar civic infrastructure (a Carnegie library, a TVA dam, even a Robert Moses parkway, the lobby of a 1930s post office) and felt something that wasn’t there in the strip mall they drove past on the way over. The coalition is genuinely cross-partisan because the romance of national building capacity predates current partisan alignments. It belongs to FDR and Eisenhower and LBJ and the Reagan of accelerated depreciation. It belongs to the country before 1986.

Frame 3: why can’t we build bungalow courts anymore?

covers courtyard apartments better than I will over at Courtyard Urbanist, so I’m going to come at this from the precursor.

You know the form when you see it. A cluster of small one- and two-bedroom homes arranged around a shared landscaped courtyard. Four to ten units. Two or three stories. Wood windows, built-in cabinetry, decorative tile. Front porches that face the courtyard rather than the street, so when you come home you walk through a garden to your front door and your neighbor is on her porch and her kid is playing on the path between the units and the bougainvillea is doing what bougainvillea does in October light. The form is called the bungalow court. It was pioneered in Pasadena and Hollywood in the 1910s, spread across the Los Angeles basin in the 1920s, and was killed in the 1930s and 1940sby zoning ordinances, parking minimums, and the FHA underwriting standards that pushed American multifamily toward the double-loaded-corridor apartment block we have been building, more or less unchanged, ever since.

The bungalow courts that survived are beloved a century after they were built. They are nationally registered historic districts. They are the things people fly to Pasadena to walk through. They are also, as a building type, the most obvious example of missing middle housing: the form factor that everyone who has stood in one understands as what good American urbanism looks like, and that nobody has been able to build at scale for ninety years.

It isn’t because we forgot how. The form is well documented. Architects know how to draw it. Contractors know how to build it. Cities are starting to allow it again. The problem is that nobody can finance it. The Terner Center’s 2024 analysis finds that under current financing terms, missing-middle projects are not financially feasible in most California markets. The only category that pencils is for-sale duplexes in the East Bay and Sacramento. Harvard GSD’s researchconfirms the structural problem: the current financial system is designed primarily to support either single-family homes or large multifamily developments, leaving missing-middle projects at a disadvantage. The small-balance construction loan that finances a six-to-twelve-unit courtyard building is the loan regional banks have stopped making, that GSEs have never made, and that private credit cannot price profitably.

Enterprise Community Partners’ analysis is explicit about what’s needed: low-interest, low-cost small-balance loans that cover a larger share of construction cost (80% or higher), plus loan guarantees and pre-development support. This is exactly what the state HFA accelerator funds and streamlined FHA 221(d)(4) deliver.

So the frame, in plain terms. We used to build a kind of building that everyone who has ever stood in one loves. The zoning fight to allow it again is half-won; the financing fight to make it pencil hasn’t started. Pull the levers above and the bungalow court can pencil again.

The coalition this frame assembles is the most aesthetically self-aware. New Urbanists. Missing-middle advocates. Strong Towns sympathizers. Landscape architects. Anyone who has ever stood in a 1925 courtyard at 4pm in October and felt the light come down through the bougainvillea and wondered why we stopped building this. The pitch is concrete and small and specific in a way the other frames are not: these levers will let us build bungalow courts again.That’s a promise that can be drawn. It can be photographed. It is not a demographic abstraction. It’s a place you can stand in.

So what?

Government-engineered financing vehicles are how ambitious American urbanism actually gets built. WED Enterprises was Walt’s. Reedy Creek was Walt and Roy’s. The Section 236 stack was the country’s, for the 1968-73 boom. Accelerated depreciation was the country’s, for the 1981-86 boom. The pattern is structural. The flat line is the cost of pretending otherwise.

Williams is right that the apparatus exists, the levers are mostly low-fiscal-cost, and the production response would come fast because the entitled-but-unbuilt pipeline is already roughly 750,000 units deep. The infrastructure that produced the 1968-73 boom is sitting there. Turn it back on.

The technical case is settled. The political case isn’t. The agenda needs framing that builds coalitions broader than “more apartments” has historically built. I’ve offered three as starting points, not a definitive list: let young Americans start their lives, build big like Walt did, and finance the bungalow courts. Better frames almost certainly exist. The work is to find them.

And the coalition starts with crediting the people who have actually been doing the work. The state organizers. The local YIMBYs. The council members who took the votes that cost them friends. The tenants’ rights folks and the missing-middle architects and the Strong Towns chapters and the Pasadena preservationists who keep the bungalow courts standing. They built every inch of political ground the federal credit agenda could plausibly stand on. Anyone who shows up to the federal conversation and treats them as a problem to be routed around is not a serious participant in that conversation.

The bankers told Walt Disney that outdoor amusement was a cultural anachronism that had already declined into senility. They were wrong. The equivalent dismissal of multifamily today, that the market can produce what we need without federal credit infrastructure, is wrong the same way.

The lamp above the firehouse is still lit. The flat line is the choice we keep making, and the choice is reversible, if we want it to be.

The Steam page for Capcom’s Pragmata on launch week, April 2026. A million copies in two days. “Overwhelmingly Positive” user reception. Reviewers calling it “the next great Dad game.” The top user review reads: “You will want to have a daughter after this game, so be aware!” A father-blogger writing about My Neighbor Totoro mentions, almost in passing, that he and his wife used Mei’s name as their daughter’s middle name. A culture-transmission event captured in a sentence. The apparatus, visibly, is working.

The game itself: Hugh Williams, a spacefarer stranded on a lunar research station, finds an android girl named D-I-0336-7 and, in the first cutscene the studio considered important enough to bring its voice actors together for a table read, names her Diana. She points at herself, then at him. The relationship starts there and grows from there. The Guardian’s Tom Regan gave it four stars and noted the father-daughter bond is pulled off “with surprising deftness.” Kotaku’s reviewer called it “a Dad Game for fathers who actually like their kids” and observed that Hugh “subverts the usual angry video game dad trope by never once treating Diana like a burden.” The Gamer was sharper: in a triple-A prestige landscape where fatherhood has spent fifteen years being depicted as trauma expulsion (The Last of Us, God of War, BioShock Infinite), Pragmata “celebrates the positives surrounding fatherhood” rather than treating it as an elaborate working-out of one’s own grief. The gameplay itself is built around the relationship: Diana can hack enemy robots to reveal their weak points, Hugh covers her while she does, neither can progress without the other, and the base-camp mechanic has him hanging her drawings on the wall between missions. The game’s director, Chou Yonghee, described the theme directly in a March 2026 4Gamer interview: 「何を基準として人と呼べるのか」がテーマとなります。人間のヒューはアンドロイドのディアナを、どこまで友達や自分の娘として感じられるかを物語として描いています。The theme is “on what basis we can call something a person,” and the story is about how far Hugh can come to feel that android Diana is his friend, his daughter. Asked whether the relationship evokes something familial, Chou answered that what matters is not 「どこで生まれたのか、何でできているか」, “where you were born or what you are made of,” but 「どのような存在で、どんな考えをしているか」, what kind of being you are and what you think. Producer Oyama Naoto added that the game depicts, across many lines of dialogue, Diana learning 普遍的な愛情, universal love, from Hugh. The Japanese interview language is worth lingering on: Chou and Oyama are articulating, in their own terms, that kinship is not a function of biological lineage. They are not speaking to Japan’s fertility statistics. They are describing a thesis about what a family is. Asked specifically about Diana’s design, Chou told AUTOMATON that the most important thing in building the character was 「リアルな子供らしさ」, realistic child-like qualities, rather than an anime-style bishōjo character or a romantic interest (アニメ的な美少女キャラや恋愛相手としてではなく). The goal was a child the player would instinctively want to protect and feel attached to. Oyama added that the team aimed for ナチュラルな子供らしい可愛さ, natural child-like cuteness, explicitly rejecting あざとい可愛さ, manipulatively cute design. The distinction matters: the creators were aiming at the protective-parental response, not the sexualized one, and they said so in the interview. Fans have nicknamed the game “Dad Space” in English and spontaneously produced a parallel 自認お父さん (”self-identified dads”) phenomenon on Japanese Twitter, with players of both languages tweeting “俺がパパだ” (I’m the dad), “娘が欲しい” (I want a daughter), and the English “this is what having a daughter is like.” The cross-linguistic pattern is part of what is interesting about the phenomenon: the response is running the same way in two languages, on two continents, among players who have very little in common culturally apart from the game itself. The game peaked at nearly 60,000 concurrent players on Steam over its launch weekend and settled at a 92% positive user rating on a base of 2,000 reviews. Those are strong numbers for a single-player experience and a new IP.

You are tempted to say, it’s a game that doesn’t include the more “delightful” side of rasing a child. That a more realistic simulator wouldn’t provide the same results. Well, a high school in the Perth metropolitan area, 2003. The Virtual Infant Parenting program is in its third year. Fifteen-year-old girls are carrying home weighted dolls that cry at random intervals through the night, stop crying only when fed, burped, or rocked, and log each interaction for review by the teacher the following Monday. The curriculum wraps the weekend in workbooks on the financial and social costs of early motherhood. The program ran in 89 countries by various estimates, built on the intuitive theory that a weekend of simulated infant chaos would deter teenage pregnancy. In the only serious randomized trial ever run, across 57 schools and roughly 2,800 girls, the intervention group was 36% more likely to become pregnant by age 20 than controls. The designers were trying to deter teenage pregnancy and produced more of it instead.

Two interventions, both low-cost, both cultural, both producing measurable behavioral effects. One designed to celebrate family life and apparently succeeding. One designed to deter teen parenthood and apparently failing by exactly producing more of it. If media this cheap can jiggle fertility intentions up and down, either the channel the signal runs on doesn’t read cognitive framing, in which case no one knows what they are aiming at, or something else is doing the structural work and the media is a sideshow.

生まれてきてよかった / It’s Good to Have Been Born

The question matters because the fertility gap in rich countries is a misallocation problem, not a preference problem, and the two diagnoses imply very different policy responses.

Americans’ ideal family size sits at 2.7 children, roughly unchanged since the 1970s. Actual U.S. total fertility is around 1.6. Across 22 European countries, approximately 60 births occur for every 100 intended. Every single cohort of women born in the early 1970s across 19 European countries and the United States ended their reproductive years with fewer children than they had wanted at age 20–24. The demographic clock is not metaphorical. Cohort fertility locks in over a roughly twenty-year window and cannot be recovered by later policy. Getting the diagnosis wrong costs a cohort.

If the diagnosis is that preferences shifted, that young people no longer want children or no longer want them at replacement rates, the policy follows: restore the family-affirming culture and fertility recovers. If preferences held but execution failed, if young people want children but cannot get the conditions to have them, the policy follows differently: fix housing, labor-market entry, workplace architecture. The two prescriptions point at different institutions, different budgets, different coalitions, different timelines. Political incentives to pick one over the other have little to do with which one is right.

Pragmata is where the question gets compressed. If media can sell family to young men who are supposedly refusing it, the cultural diagnosis is at least partially vindicated and the structural diagnosis is overblown. If it cannot, if its observed effect is marginal and depends on conditions the media itself does not produce, the weight lives further down the stack. Most of the “discourse” arrives already certain which answer it wants. Outside the “discourse”, a small number of voices, Hayao Miyazaki most notably and also the team that made Pragmata, hold both positions at once and have done so for a long time, and noticing what they say when asked is a more useful starting point than trying to resolve the question through survey data alone. The evidence, when assembled, points in a specific direction and does not flatter the louder voices on either side.

The Fake New York

Before proposing an alternative, clear the field. Each of the four commentators discussed below was reacting, in the weeks around Pragmata’s launch, to either the game itself or to the wave of positive reception around it. Morgan’s thread quoted at length below was explicitly about the game and the response to it. Piker’s line about “unfuckable losers in the gaming sphere” was a reaction to a specific cultural moment that Pragmata had produced. Andrews and Yglesias hadn’t written about Pragmata by name but they were writing in the same weeks into the same fertility discourse that Pragmata was suddenly, unexpectedly, the center of. This is worth flagging because the familiar shape of this kind of argument is to use a cultural object as a door into a bigger conversation and then abandon the object. Pragmata didn’t invite itself to a fertility conversation. The commentary brought it there and then proceeded to make claims about its audience, its designers, and what the audience reaction meant. The takes below are the specific form that bringing took. Each makes an empirical claim. Each collapses on contact with the data, and the particular ways they collapse are worth reading.

The right-structural variant, associated with Helen Andrews, holds that “the girlboss lifestyle would not exist if it were not massively subsidized,” and that removing women’s economic options would restore fertility. I went through the empirical problems with this position at length elsewhere; the compact version is that the relationship between women’s income and completed fertility has been shifting from negative to positive across rich-country cohorts, which means the women with the resources to have larger families are increasingly the ones who do. Removing their options would redistribute the deficit, not close it.

The leftest variant, exemplified by Hasan Piker’s line that “unfuckable losers in the gaming sphere always talk about starting families, and stuff. And I don’t understand why this has become a thing that they care about,” holds that young men’s stated interest in fatherhood is either inauthentic, regressive, or both. The underlying cousin-claim, that Korean 4B-style political refusal is the productive feminist response to fertility conditions, has the same shape. The scale problem is severe. The Korean 4B movement had roughly 3,400 online members at its peak against a Korean birth collapse in the 25–29 age group of 87% between 2000 and 2023. The timeline doesn’t work; 4B post-dates most of the decline. The survey data doesn’t work either. Pew’s 2024 survey of U.S. adults aged 18–34 finds 57% of childless young men want to be parents someday, versus 45% of childless young women. Canadian GSS data has tracked a higher male pro-natal rate for three decades. The young men Piker is mocking for wanting fatherhood are, on the data, the majority of their demographic cohort. The joke lands on the median, not the edge case.

The center-left/technocratic-cosplay variant, associated with Matthew Yglesias, diagnoses that the binding constraint on marriage formation sits on the male side and then concludes that “someone needs to tell them to either stay in school or else take some other conscientiousness-demonstrating path.” Tell who, through which institution, with what sanction, on what timeline that matches the demographic clock.

The alt-right variant comes from Tim Pool’s Mary Morgan, whose viral thread was a direct response to the reception around Pragmata and the men publicly enjoying the father-daughter relationship it depicts. “Childless men do not have paternal instincts the way that childless women have maternal instincts… men first experience paternal instincts once they have their own children… if a man wants to possess a child for any reason other than it being a product of his own lineage, he is likely a predator.” This variant is the cleanest to falsify because it makes a direct neurobiological claim. The claim is that the circuitry the empathy-channel literature measures does not exist in men who are not fathers, or exists only for biological offspring. The specific trigger for the claim, it should be remembered, was male audience enthusiasm for a fictional androidic daughter who is not anyone’s biological lineage, which makes the position self-defining: any positive male response to Diana is, on Morgan’s logic, evidence of predation, because Diana is definitionally outside any player’s lineage.

The literature says otherwise. Ruth Feldman’s work on oxytocin and caregiving measured the response in fathers, in alloparents, and in unrelated adults; the neuroendocrine response to infant contact does not key on genetic relatedness.Kringelbach’s Oxford group clocked the orbitofrontal response to infant faces at roughly 140 milliseconds. Pre-conscious, graded by infantile features, present in both sexes regardless of parental status. Glocker’s 2009 fMRI work found reward-circuit activation in childless women viewing infant photographs; subsequent work extended the finding to men. The Galiani and Sosa 2026 NBER paper calibrates the empathy channel at 3–33% of observed fertility decline across rich countries, baseline 13.4%. The mechanism operates at population scale in both sexes. It would be unmeasurable if Morgan’s claim were correct.

Alloparenting (adult investment in non-offspring children) is a species-typical reproductive strategy in humans, documented across every culture in which it has been studied. A theory in which male interest in children outside direct lineage is coded as pathology would need to explain how any society ever staffed a classroom. Or a coaching bench. Or an uncle or godfather role. It would also need to explain the Pragmata phenomenon itself: a million-plus men engaging positively with a simulated father-daughter relationship and describing it as something they would want. Chou Yonghee, the director of that game, said the same thing the literature says, in plainer language, when an interviewer asked him about the theme: what makes someone a person, or a daughter, or a friend, is not 「どこで生まれたのか、何でできているか」, where you were born or what you’re made of, but 「どのような存在で、どんな考えをしているか」, what kind of being you are and what you think. The director of the game Morgan’s commentary is reacting to was making, in a games-press interview, the alloparenting argument her position denies is possible. Morgan treats the audience response as evidence of dysfunction. The neuroscience and the creator both treat it as evidence that the circuitry is working as designed.

Four takes, four failure modes, same pattern. Each variant solves for “who is to blame” and then reasons backward. None of the four tests its blame assignment against the available neurobiological, survey, economic, or demographic data. When the data is consulted, preferences are approximately where they have been for decades, the instinct is present in both sexes, and the gap between desired and actual fertility is large, persistent, and not produced by any of the mechanisms the commentary proposes. Something else is doing the work.

Shelter

The cleanest way to see what that something else is, is to run two parallels side by side and let the controlled comparison do the argumentative work.

The first parallel is Japan versus Korea. Nearly everything is held constant: East Asian cultural substrate, near-identical levels of Ghibli-and-K-drama family-affirming media saturation, similar education systems, similar gender-role trajectories, similar post-industrial development paths. If cultural messaging were load-bearing, the two should track each other. They diverge by a factor of nearly two.

Between 2000 and 2023, Korea’s TFR fell 51% while Japan’s fell 12%, and nearly all of that divergence concentrates in women aged 25–29, where Korean births collapsed 87% against Japan’s 35%. The age-banding rules out slow-moving explanations like value change or secularization, which would move all cohorts together. I walked through the full Han-Uchikoshi decomposition in a prior piece; the short version is that the variable that separates Korea from Japan in the 25–29 age band is male economic inactivity, which nearly tripled in Korea (from about 10% to about 30%) and barely moved in Japan (3.6% to 5.5%). Inactive Korean men drop out of the marriage market because Korean social norms still require men to clear an economic threshold (stable employment, housing resources, typically a substantial jeonse deposit) before marriage, and missing marriages translate almost directly into missing births in a society where out-of-wedlock birth is rare.

What the comparison does for the argument here is rule out the cultural-messaging candidate. Korea and Japan are saturated with similar quantities of family-affirming media, yet one collapsed and the other didn’t. 4B exists in Korea and has no Japanese analog, but 4B is three orders of magnitude too small to carry the divergence and post-dates most of it. Female labor-force participation rose in both countries at broadly similar rates. The only candidate variable that tracks the divergence is young male economic exclusion, which is a labor-market architecture problem, not a cultural one. Neither Ghibli nor 4B is doing significant work.

The second parallel runs a different comparison at smaller scale. The Australian VIP trial was designed to move fertility intentions downward via a simulated negative experience of infant care. The 2016 Brinkman et al. Lancet paper, reporting results from a cluster-randomized trial of 57 schools and approximately 2,800 girls (1267 actually receiving the class vs 1567 in control) in Perth, found the intervention group was 36% more likely to become pregnant by age 20 than controls. Brinkman’s hypothesis, as reported in The Conversation: the girls were “embraced with family support and positive attention” during the weekend; the social context made early motherhood feel manageable and socially affirming rather than punishing.

The channel the signal ran on read the inputs (infant-coded stimulus, positive social context, increased infant exposure in an age-group already near the relevant decision threshold) rather than the designers’ intent. This is exactly what the empathy-channel framework predicts. Glocker’s fMRI subjects, Kringelbach’s 140-millisecond response, and Feldman’s oxytocin measurements all identify a pre-deliberative system that reads environmental inputs regardless of the cognitive framing accompanying them. The VIP designers assumed they were operating on a deliberative system that would weigh a stated cost and avoid it. They were operating on a pre-deliberative system that registered another infant in the environment.

The same channel carries Pragmata and Totoro. These works do not convince anyone of anything. They add infant-coded and family-coded exposure to the ambient environment of players and viewers, which feeds a system that runs beneath deliberation and tracks environmental frequency. The effect is real, the mechanism is documented, and the magnitude is small, one input among thousands.

What’s worth noting about Pragmata specifically, and about the commentary around it, is the posture it takes against the dominant treatment of fatherhood in its own medium. For roughly fifteen years, the prestige triple-A game canon has rendered fathers as men processing grief: Joel in The Last of Us working through the death of his biological daughter by reluctantly taking on another, Kratos in the God of War reboot distant with Atreus because rage is the only mode he has, Booker DeWitt in BioShock Infinite constructing an elaborate metaphysical reason to fail at fatherhood. Fatherhood-as-trauma-expulsion is the shape the form settled into. Pragmata does not do this. Hugh names Diana in the first minutes of the game, asks her to keep telling him about whatever she is thinking about, hangs her drawings on the wall of their base camp, and promises to show her Earth. The Gamer’s review described the point directly: Pragmata understands that “sometimes it’s okay to spin a yarn that isn’t heavily defined by the trauma of bringing another soul into this world.” That a game with this posture, released into the most crowded AAA calendar in living memory, sold a million copies in two days and produced a spontaneous fan nickname (”Dad Space”) and a spontaneous meme-genre about the game inspiring parenthood, tells you something about what the audience was hungry for and wasn’t getting. The hunger is not evidence that media will move fertility. It is evidence that the audience’s stated preference for family-coded content exists, is broad, and was not being served.

Miyazaki himself has been unusually explicit about what he is doing. In a 2005 Guardian interview with Xan Brooks, he put it this way: “Personally I am very pessimistic. But when, for instance, one of my staff has a baby you can’t help but bless them for a good future. Because I can’t tell that child, ‘Oh, you shouldn’t have come into this life.’ And yet I know the world is heading in a bad direction. So with those conflicting thoughts in mind, I think about what kind of films I should be making.” At his 2013 retirement press conference, he compressed the same idea: “I wanted to convey the message to children that this life is worth living. This message has not changed.” In a Hiroshima Peace Media Center interview from 2009, asked about the “power of children,” he described the daycare next to his studio. アトリエの隣にある保育園の子どもたちは僕に力をくれます。泣き声や笑い声、叫んでいる声が聞こえてくると、僕は幸せな落ち着いた気持ちになります。年寄りを勇気づける力を幼い子たちは持っています。The children at the daycare give him strength; their crying, laughter, and shouting voices produce a happy, settled feeling; small children have the power to encourage the elderly.

Note what Miyazaki is saying. He is a pessimist. He believes children understand intuitively that the world they have been born into is not a blessed world. He makes the films anyway, as a deliberate act of pro-natal blessing in a civilization he thinks is failing. The taglines of his late films respond to a worsening environment in increasingly direct terms: 生きろ (”Live.”) for Princess Mononoke, 生まれてきてよかった (”It’s good to have been born.”) for Ponyo, 生きねば (”We must live.”) for The Wind Rises. He is not making the films under the illusion that they will fix the conditions. He is making them because, in his view, the gesture of blessing is the correct response to an environment the children did not choose and cannot yet change. This is the opposite of what the loud commentators are doing. They deny or weaponize the conditions. Miyazaki acknowledges the conditions and contributes the marginal exposure he can, understanding its limits. Totoro and Ponyo and the rest are not propaganda for a culture that has already succeeded. They are one specific artist’s offering against an environment he believes is depleting, delivered through a channel he understands from watching children out his window.

Both parallels converge on the same finding. In the Korea–Japan case, the deliberative/cultural frame underperforms and the structural/economic frame predicts. In the Pragmata–VIP case, the deliberative/persuasive frame underperforms and the pre-deliberative/environmental frame predicts. The channel carrying the fertility signal does not read intent. It reads conditions. The conditions are built in labor markets, housing, and marriage-formation norms, not in the cultural products that sit on top of them. Cultural products make marginal adjustments on top of conditions someone else set.

Mechanics

Zoom out. The specific cases are instances of a template I developed at length in a prior piece; the compressed version has three interlocking components.

First, Stephen Shaw’s 2025 decomposition across 314 million mothers in 33 countries shows that TFR is two independent numbers multiplied together. The share of women who become mothers at all (TMR) falls in discrete shocks and ratchets down; the average family size among mothers (CPM) stays roughly flat for decades. Most of the rich-country fertility decline over the past three decades has been TMR decline, not CPM decline. The problem is not, primarily, that mothers are having fewer children than they want. It is that a growing share of women who wanted to become mothers never reached the family-formation threshold in the first place.

Second, Galiani and Sosa’s 2026 NBER paper calibrates the empathy channel at a baseline 13.4% of observed fertility decline across rich countries, range 3–33%. The channel is dormant at high fertility and steeply active at low fertility, which means each cohort’s TMR produces the ambient infant environment of the next cohort, and each TMR shock depletes the exposure environment of the cohort behind. The trap closes itself without any cultural change required. Preferences can remain stable; the outcomes worsen because the inputs worsen. This is what Miyazaki was describing from his studio window when he said the children at the daycare next door gave him strength. He was noticing, without the vocabulary for it, that his own empathy-channel response was being sustained by an environment that a younger Japanese worker in a newer neighborhood with fewer children and fewer daycares would no longer receive. His instincts about what was happening to the society around him tracked the mechanism Galiani and Sosa would later put a coefficient on.

Third, Uchikoshi and colleagues’ 2026 Population and Development Review finding on child-driven marriage. Uncertainty about wanting children cuts the annual marriage transition rate by 30–50%, controlling for the stated desire to marry. The arrow runs opposite to the standard model: marriage does not cause children; the prospect of children causes marriage. Between 30 and 39 percent of never-married Japanese adults in their twenties and thirties report they are uncertain whether they want children. Not opposed. Uncertain. That uncertainty blocks marriage, which blocks births, which depletes ambient infant exposure in the next cohort, which feeds the uncertainty in the cohort after that. The three components chain together into a single propagating mechanism: the fertility gap is a TMR problem, running on a depleting empathy-channel amplitude, interrupted at the margin by child-driven marriage uncertainty, propagated cohort-to-cohort by environmental depletion.

Pragmata, for all that I have so far been using it as a headline fact about sales and reception, is also a fairly literal fictional rendering of this third mechanism. Hugh does not know Diana exists when the game begins. He encounters her, in the first minutes of play, as a small dependent being whose safety becomes a project he organizes the rest of his actions around. The naming scene is the game’s declaration that a relation of this kind can start from nothing and become the structure the rest of a life runs on. The reader does not have to agree that Hugh and Diana’s specific arrangement is a template for human reproduction to notice that the game is staging, in its opening cutscene, the phenomenon Uchikoshi’s paper names: a person’s relation to the possibility of a small being in their care, and the way that relation reorganizes their other decisions. That the staging worked on the audience at the scale it did is the piece of evidence Section I led with. That it worked cross-culturally, on Japanese and English-speaking audiences who had very little in common except the game, is mechanism-level evidence that the third component of the template runs on a channel that does not require cultural priming to activate. The mechanism and the game are describing the same thing. The game does not prove the mechanism; the mechanism explains why the game lands.

One observation before the pattern is complete. Darby Saxbe’s work on what gets called “the male loneliness crisis”notes that the loneliness pattern is not primarily gendered; it is generational, with young women tracking similarly to or slightly worse than young men on most measures, a finding corroborated by Richard Reeves’s American Institute for Boys and Men analysis of American Time Use Survey data. This is consistent with the empathy-channel depletion story. Fewer strollers, fewer playgrounds, less incidental contact with children and infants, fewer social occasions to form the kinds of relationships that produce such contact: all of it hits both sexes’ pre-deliberative priming equally. The youth-wide cohort effect is a symptom of the same environmental depletion the empathy channel measures. The gender-war framing of the fertility problem is looking at a shared symptom and calling it an asymmetric disease.

Infrastructure

Name the architecture that sustains the template. No villains here; the problem is structural precisely because everyone involved is behaving rationally. The load-bearing components are the ones where the fertility pattern emerges as the predictable response to an incentive structure nobody designed on purpose.

Housing markets. The supply constraint on family-sized housing is where family formation gets priced out. In Korea, the jeonse deposit architecture concentrates the threshold into a single lump-sum barrier; in the U.S. and U.K., exclusionary zoning compresses family-sized housing supply in the family-formation window. Doepke and Kindermann’s “bargaining over babies” model formalizes the downstream effect: when housing tightens, the lower-desire partner wins the intra-household bargain. The bargaining outcome trends toward fewer children than either partner’s stated ideal. Nobody is lying about what they want. The architecture pushes the realized outcome below both of their ideals.

Labor markets. We all know about the impacts of labor markets on men. What about women and “girl bosses” especially? Caregiving is professionally expensive because workplace architecture treats it that way. analysis of 1972–2024 Current Population Survey fertility supplements finds that women in the top decile of occupational status average approximately 1.4 children while working in those roles. That’s a real finding. The framing it often travels with, that demanding careers are a preference young women should revise downward, is a separate claim Stone himself rejects: “the path forward on the debate between work and family is not to ‘blame’ the girlbosses, nor is it to ‘white knight’ for them, or blame men for being unmarriageable.” The 1.4 number is telling us something about the workplace, not about what women want. The highest-status career tracks treat multi-year caregiving breaks as professional demotion, which is a choice downstream of labor-market conditions. When labor is abundant and workers are interchangeable, employers treat any extended break as a credibility problem. When labor is scarce and workers are hard to retain, employers build accommodations. Same firm, different market tightness, different workplace architecture.

This is where family-leave legislation runs into its ceiling. Parental leave laws exist on the books across most rich countries and in many U.S. states; actual uptake, and the career consequences of uptake, track local labor-market conditions more closely than the legal text. A law on paper is only a real protection to the degree that tight labor markets make it costly for employers to retaliate against workers who use it. The same logic extends to education: the extension of pre-career training pipelines from the early twenties into the early thirties (required master’s degrees, credential inflation, unpaid internships, extended postdoctoral chains) is itself a labor-market-slack phenomenon, which pushes family formation past critical biological windows for women and past the economic-threshold window for men. The Korean jeonse case is this at extreme intensity; a gradient of the same effect operates across the rich world. Aspirational media in a slack labor market is a drop of water on concrete. Aspirational media in a tight labor market lands in an environment where the ambient conditions for family formation are actually executable, and the marginal push the media provides can be acted on. Media sits on top of the conditions. It does not build them.

Recent examples illustrate the pattern, with a particular irony: the firms most aggressively penalizing women for family formation are often the same ones that promoted “girlboss” ideals. Deloitte Consulting faces a 2026 class-action lawsuit alleging the firm penalized employees for taking protected pregnancy and parental leave by scoring their annual performance evaluations against colleagues who worked the full year, without adjusting for protected absences. It means lower compensation and promotion prospects for those who used parental leave. The same firm announced plans to slash family benefits for non-client-facing staff, cutting parental leave from 16 weeks to eight and eliminating a $50,000 adoption/surrogacy/IVF benefit starting in 2027. Meta, Google, and other tech companies also have a rich history laid off employees while they were on medical and maternity leave. These same companies who promoted empowering women and “leaning in.” MrBeast’s companies face a lawsuit from an employee who alleges she was fired three weeks after returning from maternity leave, having been told during her pregnancy that she needed to “continue to grind” to have a job to return to.

Notably, these penalties occur even when benefits are government-funded rather than company-paid: in a weak job market, employers find ways to retaliate against leave-taking regardless of who bears the cost. Each case follows the same script: formal policies exist, workers take legally protected leave, and career penalties follow through mechanisms the law does not prevent without a labor market that can effectively punish bad actors.

Dating markets. The traditional meeting-and-sorting infrastructure has collapsed into dating apps, credential mismatches, geographic sorting by education, and the disappearance of workplaces and third spaces as matching venues. The Bodin et al. Swedish study of 191 men found men uncertain about fatherhood were three times more likely to cite concerns about a potential co-parent relationship than confident ones. The uncertainty propagating through Uchikoshi’s child-driven marriage mechanism is partly produced by a matching infrastructure that no longer reliably matches.

The commentary economy. Blame attribution is a product that can be manufactured cheaply and sold at scale. The cluster of takes identified earlier (Andrews, Yglesias, Piker, Morgan) isn’t a coincidence of four separate commentators having similar bad takes; it’s a structural equilibrium produced by the engagement architecture of online discourse.

cites NYU social psychologist Jay Van Bavel’s research showing that in real-life political-opinion surveys, belief distributions are approximately bell-shaped, with most people holding moderate views on most issues. Online, 97% of political posts on Twitter/X come from roughly 10% of the most active users, the most extreme end of the belief distribution on any given issue. The underlying distribution is a bell. The visible distribution is a U. The algorithm amplifies engagement and engagement is produced at the extremes, so moderate views become invisible while extreme views become the observed consensus.

This is the architecture that makes the equilibrium rational. A pundit who names a demographic scapegoat (the girlbosses, the unf*ckable gamers, the women who “waited too long”, the men whose paternal instincts “prove” they are predators) captures the 10% that produces 97% of posts. A pundit who walks through the Korean male-inactivity rate rising from 10% to 30% between the mid-1990s and early 2020s and interacting with the jeonse housing architecture to collapse the 25–29 marriage rate by 89% captures a polite nod and a small audience. The commentary market rewards the first; it neither rewards nor punishes the second; it simply does not amplify it. The market does not care whether a take moves fertility, because fertility movement is not the product. Attribution of blame is the product, and blame is a renewable resource.

Miyazaki is the useful control case. He has held roughly the same position for thirty years, pessimist about the world and maker of blessings for children anyway, through four decades of changing discourse environments, without converting it into a brand or a niche. He says the same thing when the Guardian asks him in 2005, when the Hiroshima press asks him in 2009, and when he retires in 2013. The commentators identified above have to generate fresh takes because the engagement function decays. Miyazaki can repeat himself because his position was never optimized for engagement in the first place; it was optimized for whatever he believes the actual problem is. The four takes, whatever their ideological differences, converge on the same architectural behavior: treat the demographic group whose life conditions are most constrained by the problem as the cause of the problem, propose no delivery mechanism, harvest engagement, move on. The commentators are not stupid. They are responding to an incentive structure that rewards this specific behavior regardless of intent. The underlying distribution of views in the general population, which the real-life surveys find, is a bell of shared concern and ambivalence. The visible discourse is a minority-opinion product with outsized distribution.

Let’s go outside

Let me step out of the formality for a minute, because I’ve been at this for a while now and the register I’ve been writing in, architecture and channel and threshold and depletion, has a way of making the actual thing I care about sound “cleaner” than it is. Apologies in advance if the gear shift feels abrupt. It is.

The actual thing is this. I know men (young, old, Gen Z, and millennial) who want to be dads and are starting to think they won’t get to be, and I know women who want to be moms and are starting to think they won’t get to be, and the reasons in every case are specific and boring and heartbreaking. The rent. The job. The partner who wanted different things. The degree that took too long. The loneliness that wouldn’t break. The year, and then the next year. None of them are stupid (well, they are still human and humans are stupid to a degree). None of them are selfish (that’s not true, we are all selfish to a degree). None of them are refusing anything. They are a generation of people who have been told in one register that they should want this, in another register that wanting it is suspect, and in a third register that the conditions under which the wanting could be acted on are, at present, beyond the scope of anybody’s concern.

The reason I wrote thousands, if not tens of thousands, of words on this isn’t that I find demographic dilemma interesting, though I do. It’s that there is a specific kind of grief that comes from wanting a life the world won’t arrange itself to let you have, and the people I care about are starting to carry that grief, and I don’t want to watch them be lectured about whose fault it is when what they need is for somebody to notice the weight.

Miyazaki watches children from his studio and decides to make something for them anyway, knowing the world is heading somewhere bad, a take I very much agree with despite recent innovations. Chou and Oyama spent six years building Hugh and Diana and told an interviewer, when asked, that what makes a family isn’t biology, it’s what kind of being you are and how you treat the ones who need you. The artists, when you ask them, say the simple thing. Most of the commentary, when you ask it, says something else. I’ve tried in this piece to say the simple thing in the register the commentary accepts, which is a long walk to take to get to a place any honest parent could have described in a sentence. The walk was necessary because the room is loud, and the architectural vocabulary is what lets the argument survive inside the room.

If anything here was useful to you, use it. If something was wrong, tell me. I am not trying to convince anyone to want a family they do not want. I am trying to clear some air around the people who do want one, so that the work of actually helping them, which is real work, doable work, and work that a small number of places have been quietly doing for decades, has somewhere to land that isn’t already occupied by a take. The wanting is there. The loud voices have been misdescribing it. That is most of what I came here to say.

The game I started this essay with ends, like all games, with the credits rolling and the player putting down the controller. The father-daughter relationship the million-plus players spent their weekend inside goes back into storage on the console. The players go back to their lives. Some of those lives contain daughters and some don’t, and some of the ones that don’t would, under different conditions, contain them. A game cannot build the conditions. A few of my fellow pronatalists like Lyman would say games and media like cartoons are bad. Others like myself takes more joy at modern media. A game can, at most, let the players know that the conditions are the problem, by briefly showing them what it feels like when the conditions aren’t there and the relationship is allowed to exist anyway. Some people call it escapism, but the lack of choice people have it’s a gift.

That is the specific gift Chou and Oyama made, and Miyazaki has been making some version of it for forty years, and the people with children to protect or children to want have been receiving those gifts and using them to sustain themselves through an environment that is, as Miyazaki keeps saying, heading somewhere bad. The gifts are not a substitute for changing the environment. They are a reason to keep going until someone does.

Why does Jiangsu build and California can’t?

In November 2013, six months after Xi Jinping announced the Belt and Road Initiative at a podium in Astana, the Jiangsu Provincial Development and Reform Commission stood up a dedicated leading small group to handle implementation. The group had a budget and full-time staff. Within three years, Jiangsu had launched 47 infrastructure projects across partner countries worth USD 8.3 billion, formed 23 international partnerships, and sent its senior leadership on 15 overseas promotional trips. The province’s 13th Five-Year Plan, adopted in 2016, treated the initiative not as a directive from Beijing but as an opportunity Jiangsu intended to shape around its own industrial priorities.

In January 2008, five years before Jiangsu even heard of BRI, California voters approved Proposition 1A, a $9.95 billion bond issue to build a high-speed rail line connecting San Francisco to Los Angeles. Governor Schwarzenegger signed the authorizing legislation the following year. The line was projected to open in 2020 at a total cost of $33 billion. As of February 2026, not a single train operates on any portion of the California High-Speed Rail system. The projected cost has risen to approximately $135 billion. The current governor, Gavin Newsom, announced in 2019 that the full line would not be built, pivoted to a partial segment in the Central Valley, and then in 2025 claimed credit for entering the “track-laying phase” of that reduced segment. In the same state, over the same period, over $37 billion has been spent on homelessness programs; California now has more unhoused residents than any other state in the country. Newsom ran for governor in 2018 promising 3.5 million new housing units by 2025. The actual total is 640,873 homes built across seven years, less than a fifth of his target.

Both of these happened. Both are the work of senior sub-national officials in major economies, operating under democratic or authoritarian institutions as the case may be, given large budgets and clear mandates.

One set of officials built things. The other set did not.

Which half is the West copying?

The misdiagnosis of why this gap exists is already shaping Western policy, and the policies being produced will make Western economies worse rather than better.

In September 2024, former European Central Bank president Mario Draghi delivered his report on European competitiveness to the European Commission. The report is now the official economic doctrine of the EU, endorsed by every member state and the European Parliament, with Renew Europe minister Stéphane Séjourné declaring that “everything proposed since has been aligned with it.” Draghi’s diagnosis: Europe is falling dangerously behind the United States and China, its productivity growth has collapsed, its firms are too small to compete at global scale, its fragmented markets prevent the emergence of European champions. Draghi’s prescription: relaxed merger rules, industrial consolidation, capital market integration, the construction of “European champions” in key sectors, €800 billion per year of investment partly financed by joint EU borrowing.

Consider how specific this gets. The EU currently has 34 mobile network operators compared with 3 in the United States and 4 in China. Draghi treats this as a self-evident problem, blaming “overly cautious competition authorities” for the “plurality of players.” The Commission’s Competitiveness Compass, released in January 2025, goes further than Draghi himself in loosening merger rules to support European champion creation. The telecoms sector is the first target. Defense is next. Automotive is in the queue.

American commentators make the same pattern-match. Jennifer Lind’s Foreign Affairs essay from February 2026, drawn from her book Autocracy 2.0, frames China’s rise as a story of “smart authoritarianism” in which the CCP adapted its tools of control to foster innovation while preserving central command. The Washington Monthly in October 2025 praised China’s “state-managed capitalism” for combining “political repression with a coherent economic plan.” The American Compass, on the right, characterizes China as a state-controlled economy whose successes flow from CCP direction over industry. The implicit lesson across these accounts, regardless of whether the author approves or disapproves, is that China is winning because of central command, and that Western states must therefore either copy the command or wall themselves off from its results.

Chinese centralization is real, and in specific domains it is valuable. Beijing directs state bank credit into strategic sectors at scales the West has never matched. Beijing holds industrial electricity prices below market through state-set benchmark tariffs. Beijing disciplines its own billionaires when they step out of line, most famously when Jack Ma criticized financial regulators in October 2020 and saw Ant Financial’s IPO suspended within weeks. These are desirable uses of state power. A Western observer who looks at Chinese success and concludes that the center’s strategic capacity matters is not wrong.

What that observer is wrong about is which uses of central power the West will actually copy.

The EU is not about to cap industrial electricity prices. It has spent the last five years deliberately raising them. The IEA’s 2025 analysis shows EU industrial electricity prices at roughly twice US levels and 50% above China’s in 2025, up from 50% above the US and 20% above China in 2019. The gap doubled in half a decade, driven by EU ETS carbon pricing, deliberate phase-out of domestic generation, and LNG dependence. Germany shut down its last operational nuclear reactors in April 2023 while importing LNG at roughly three times pre-2022 prices. By country, industrial rates run roughly 18 US cents per kWh in Germany, 16 in Japan, 12 in France, and 7 in China. The EU has the formal authority to hold industrial electricity prices low through state aid, cross-subsidization, or domestic production mandates. It refuses to use it.

The United States, which has the world’s largest shale gas reserves, imposed a moratorium on new federal oil and gas leases in January 2021 through Executive Order 14008. A Louisiana federal judge ruled the moratorium exceeded presidential authority, but lease sales kept falling by orders of magnitude through the rest of the term. The blocked Q1 2021 Wyoming sale covered 476,000 acres across 383 parcels; the Q1 2024 sale covered 13,417 acres across 30 parcels.

The leasing freeze was the visible half; the financial architecture was worse. The Net-Zero Banking Alliance launched in April 2021 under UN auspices with JPMorgan, Citigroup, Bank of America, Morgan Stanley, Wells Fargo, and Goldman Sachs as founding members committed to align lending portfolios with a 1.5°C pathway by 2050. Reserve-based lending to small and mid-cap independents dried up. Institutional LPs backed away from oil and gas private equity under ESG pressure. When Russia invaded Ukraine and WTI peaked near $130 in March 2022, US producers with breakevens around $60 still could not grow. The Dallas Fed Q1 2022 Energy Survey found nearly 60% of executives cited investor pressure to maintain capital discipline as the primary reason they were not drilling; 29% said growth was not dependent on oil price at any level. Big integrated producers returned cash to shareholders. Biden’s response was historic SPR releases, pressure on OPEC to pump more, and Russia sanctions that pushed prices higher.

Four years later the pattern repeats with the rhetoric inverted. The 2026 Iran war closed the Strait of Hormuz on March 4, pushing Brent past $120 and OPEC production down a record 7.56 million barrels per day that month. Trump’s response is the same playbook: SPR release structured as a loan from a reserve still at 58% of capacity a year after he pledged to fill it to the top, pressure on Saudi Arabia and OPEC to ramp output, and deregulation of LNG export terminals whose first shipments arrive in 2029-2030. Small and mid-cap producers remain capital-starved. Marshall Adkins of Raymond James says Trump’s push for lower prices is bad for US producers; the EIA now predicts US production will fall to 13.5 million barrels per day in 2026, the first annual contraction since 2020. The Net-Zero Banking Alliance dissolved in November 2025; institutional capital remains absent anyway, because the alliance was the visible surface of allocation preferences that persist without it.

Biden had $130 oil, a producer base with $60 breakevens, and the full federal toolkit: Defense Production Act loans, Treasury loan guarantees, SPR floor-price purchases, bank regulator guidance. He deployed none of it. Trump inherited the same toolkit, inverted the rhetoric toward deregulation, and is producing the same outcome: falling domestic production during a price shock, capital-starved small and mid-cap producers, strategic dependence on OPEC goodwill. When Beijing’s strategic sectors need capital, state banks lend and municipal guidance funds take equity. When American producers needed capital during two separate price shocks in four years, the federal response across both parties has been to release the emergency reserve and ask OPEC for help.

These are not limits on Western central authority. These are exercises of Western central authority, in directions opposite to what Beijing chose. Both Berlin and Washington possessed the formal tools to do what Beijing did and refused. The EU could direct state aid at industrial energy costs; it preferred carbon pricing that raises them. The US could have directed capital to domestic producers through multiple price shocks across two administrations; it preferred emergency reserve releases and calls to OPEC.

Now consider capital direction. Beijing routes strategic capital through state banks and guidance funds into semiconductors, batteries, renewables, and EVs at scales the West has never matched. The CHIPS Act of 2022 authorized $52 billion in US semiconductor funding; Chinese state guidance funds deployed around $1.52 trillion into strategic sectors between 2015 and 2021 alone, with state-bank lending pushing the total substantially higher. The EU’s Important Projects of Common European Interest framework has cumulatively spent less than €40 billion since 2018, compared with €190 billion in uncoordinated national aid in 2023 alone. Scale matters. Direction matters more.

Now consider discipline of incumbents. When Jack Ma criticized regulators, Beijing suspended Ant Financial’s IPO, launched antitrust investigations against Alibaba, and ultimately extracted multi-billion-dollar fines and structural changes. Similar discipline hit the private education tutoring sector (essentially dismantled in 2021), online gaming (forced time limits and new-game approval moratoriums), and real estate speculation (forced deleveraging through the “three red lines” policy). Contrast this with the West’s relationship to its own incumbents. The EU has issued antitrust fines against US tech firms totaling well over €10 billion since 2017, and none of these have produced structural change. The US Department of Justice’s antitrust case against Google, decided in 2024, has yet to produce remedies. The political will to discipline incumbents does not exist in Western systems at the level Beijing routinely exercises.

These are the categories where Chinese central authority acts decisively against commercial incumbents on behalf of strategic priorities: cheap energy, directed capital, and disciplined platforms. These are also the categories where Western states consistently refuse to exercise their own authority. Subtract those three and look at what remains. The one category where Chinese central authority routinely fails (forced industry consolidation, as the Dongfeng-Changan merger collapse of June 2025 will show) becomes, in Western discourse, the primary lesson to copy. Draghi wants fewer telecom operators. The Competitiveness Compass wants looser merger rules to build champions. American commentators want scale through consolidation. The West is about to import the single habit that even Xi cannot fully execute, while refusing to import the habits where Beijing excels.

Western centralization without the political will to hold energy prices, direct capital at scale, or discipline billionaires will produce the costs of centralization without the benefits. It will leave the West with the worst possible combination: fewer competitors, none of them strategically directed, all of them protected.

Is China really a command system?

The monolithic picture rests on real evidence. China is a Leninist party-state. The Central Organization Department appoints every senior official in the country. Xi has spent a decade consolidating personal power: abolishing term limits, enshrining “Xi Jinping Thought” in the party constitution, purging hundreds of thousands of officials through anti-corruption. Lind is not wrong that the regime has maintained rigid control while achieving innovation. Fang, Li, and Lu’s analysis of 3.7 million policy documents confirms real recentralization after 2013: the central share of local policy portfolios rose from roughly 30% to above 40%.

The monolithic story is not wrong to identify central strategic capacity as part of what makes China hard to compete with. What it misses is that central strategic capacity does not execute itself, you still need people and subsystems. Every directive Beijing issues must be translated into operational activity by sub-national administrations. The center sets the strategic direction; the provinces and municipalities build the semiconductor fabs, finance the battery makers, run the tech regulatory investigations, and implement the benchmark industrial tariffs. Remove the deep provincial bench and Beijing’s strategic directives become press releases. This is exactly what they become in Qinghai.

The sharpest recent empirical work on the gap between central directive and provincial execution is a 2026 study by Zheng Gong and Tim David in the Journal of Chinese Political Science, Provincial Governance and Variation in China’s Belt and Road Initiative. Gong and David assemble panel data on BRI implementation across all 31 Chinese provinces from 2015 to 2019 and run a hierarchical linear model to decompose the variance in implementation effectiveness. Their finding: about 34 percent of the variation sits between provinces rather than within them over time, and provincial characteristics explain 58 percent of that between-province variance. The single strongest predictor, even after controlling for GDP per capita, population, coastal location, and distance from Beijing, is provincial bureaucratic capacity.

This is not what a pure command system produces. In a pure command system, the same directive produces comparable outcomes, with variation attributable to geography or noise. Gong and David find something else. The same directive produces radically different outcomes across provinces, tracking each province’s administrative depth. The center is the senior partner, not the sole actor. Take away provincial capacity and the center has authority with nothing to point it at.

The second piece of evidence is the Dongfeng-Changan merger collapse of June 2025, which reveals what happens when central direction confronts provincial resistance on a policy where Beijing itself is ambivalent. In February 2025, the State-owned Assets Supervision and Administration Commission, reporting directly to the State Council, announced a plan to merge Dongfeng Motor Group and Chongqing Changan Automobile into a single central state-owned automaker with roughly five million annual vehicle sales, a scale competitive with BYD. Xi Jinping’s Central Financial Affairs Commission had made industrial consolidation an explicit priority.

116 days later the merger collapsed. Caixin‘s reporting identified the cause as intervention from the municipal government of Chongqing and the provincial government of Hubei, each refusing to accept the terms on offer. The companies, with that political cover, walked away. Analysts had predicted this before the announcement was even official, noting that auto consolidation in China “has failed to happen due to reluctance by government stakeholders to give up control of such important contributors to their local economies.” The fiscal architecture makes the refusal rational rather than obstructionist. If Chongqing loses Changan’s headquarters, it loses the tax revenue, the employment numbers, the cadre credit, and the center’s willingness to tolerate the province’s LGFV debt. No local leader who wants promotion acquiesces to that without extraordinary compensation.

The monolithic story cannot accommodate this without contortion. The distributed-competence story explains it immediately, and it also explains why Beijing wins on some domains and loses on others. Where central strategic priorities align with provincial incentives (strategic capital allocation, cheap industrial electricity, disciplining outside-system incumbents like Alibaba), the center wins. Where central priorities cut against provincial fiscal interests (consolidating automakers means some provinces lose their local industrial base), the center loses, even to Xi. The distributed structure is not a bug in the Chinese system. It is a feature whose costs and benefits are distributed unevenly across policy domains.

What happens when three provinces get the same directive?

Three provincial cases, same directive. Gong and David study Jiangsu, Henan, and Qinghai, selected to represent high, medium, and low bureaucratic capacity respectively. All three provinces receive the same BRI directive in 2013. All three face the same central policy intensity over 2015-2019. All three are subject to the same Central Organization Department appointment system, the same nomenklatura, the same target-responsibility evaluation framework. The variable that differs is provincial administrative capacity. The outcomes track that variable with almost embarrassing clarity.

Jiangsu is the high-capacity case. Its Provincial Development and Reform Commission moved within six months to stand up the dedicated BRI leading small group mentioned at the opening of this piece. The 13th Five-Year Plan of 2016 treated BRI as an opportunity to align provincial industrial policy with overseas market expansion. Jiangsu’s execution figures cited at the opening (47 projects, $8.3B, 23 partnerships, 15 overseas promotional trips) came from a deep SOE bench in construction and engineering that allowed the province to assemble consortia for overseas tenders. The mechanism Gong and David describe is “bureaucratic entrepreneurship,” in which provincial officials do not merely implement the central directive but actively shape it around what their province is organizationally capable of delivering. Jiangsu was not told to do any of this specifically. It chose to, because its administrators could.

Henan is the medium-capacity case, and the most instructive of the three. It is inland. It has no obvious BRI advantages. Its economy is smaller and less internationally integrated than Jiangsu’s. By the logic of the monolithic story, Henan should have either followed Jiangsu’s template (if Beijing imposes uniformity) or done very little (if provincial resources determine outcomes). It did neither. Henan’s administrators negotiated. They identified aviation logistics as the province’s one potentially BRI-relevant asset, then spent three years lobbying the National Development and Reform Commission to designate the Zhengzhou Airport Economy Zone as a regional BRI hub. They drafted feasibility studies. They bargained with central agencies. They did not ask for permission to participate on standard terms; they asked to participate on terms their province could actually meet. By 2019, Henan had launched 23 BRI projects worth roughly USD 3.1 billion, less than Jiangsu but more than any low-capacity Western or interior province. The province got there through policy entrepreneurship, deployed by officials most foreign observers could not name, in a province most foreign observers could not place on a map.

Qinghai is the low-capacity case. It issued an official BRI framework in 2016. It established coordination bodies similar to those in Jiangsu and Henan. It incorporated BRI into its provincial work reports. It launched eight projects worth roughly USD 0.4 billion by 2019, less than 5 percent of Jiangsu’s volume. Gong and David document that many of these were pre-existing provincial programs relabeled as BRI initiatives without new resource commitments. The monolithic story would predict either uniform execution (which did not happen) or principled refusal (which also did not happen). What happened instead was symbolic compliance: the forms of participation, without the substance, because the substance exceeded the province’s administrative capacity.

Same directive. Same national context. Same five-year window. Three outcomes, each tracking the administrative capacity of the receiving province rather than the intensity of the central directive. I don’t know about you, but this clearly does not look like the centralized command system we were told.

The point becomes sharper in Gong and David’s cross-level interaction analysis. The authors estimate that central policy intensity interacts with provincial bureaucratic capacity at γ = 0.187, p < 0.01. When Beijing pushes harder on BRI, high-capacity provinces deliver substantially more, while low-capacity provinces barely respond. A one-standard-deviation increase in central pressure produces a 0.44-unit change in implementation effectiveness in high-capacity provinces and only a 0.09-unit change in low-capacity ones. Central pressure is an amplifier of provincial capacity, not a substitute for it. When Beijing leans on Jiangsu, Jiangsu builds more. When Beijing leans on Qinghai, Qinghai issues more press releases.

Beijing’s willingness to hold electricity prices low means nothing without provincial grid operators executing the tariff structure; state-bank credit directed to semiconductors means nothing without municipal guidance funds co-investing and local SOE managers building the fabs; discipline of Alibaba means nothing without provincial and municipal regulators who can run the investigations and enforce the remedies. The center is the senior partner, not a substitute.

Who actually writes Chinese industrial policy?

The BRI case is not an anomaly. It is the operating system.

Fang, Li, and Lu’s 2025 NBER analysis, summarized accessibly by VoxDev, identifies 768,000 documents as industrial policy, roughly a quarter of all government policy output. Of these, 13 percent come from the central government. 45 percent come from provinces. 39 percent come from cities. 3 percent come from counties and townships. Eighty-seven percent of Chinese industrial policy, by document count, originates below Beijing. A 2025 Stanford SCCEI brief puts the figure for Chinese policy overall at more than 80 percent originating locally. The authors attribute this to tournament competition among officials seeking promotion through growth-oriented policy innovation, the same incentive structure that Li and Zhou described in their 2005 analysis of political turnover and economic performance.

Provincial capacity drives not just policy volume but policy quality. Wang and colleagues, writing in Management and Organization Review in 2024, examine industrial specialization choices by all 31 provincial governments and find that provinces with greater organizational efficacy, measured by access to better-resourced local SOEs in their focal industries, make smarter specialization decisions. The finding mirrors Gong and David’s finding for BRI: capacity predicts both the quantity and the intelligence of provincial policy output.

The theoretical framework that organizes all of this is Chenggang Xu’s 2011 Journal of Economic Literature article, which names the Chinese system “regionally decentralized authoritarian.” Xu’s formulation: the center controls personnel and sets strategic direction, subnational governments run the bulk of economic execution. The personnel pipeline is centralized. The policy machinery is not. What looks from the outside like a single commanding intelligence is actually several thousand provincial, municipal, and county-level administrations running their own industrial strategies, competing with each other for growth, and negotiating with Beijing over the terms on which they participate in centrally-branded initiatives.

The same pattern appears in finance. The 1994 tax-sharing reform centralized most tax revenue with Beijing but left local governments responsible for the bulk of public expenditure. Local revenue fell below 40 percent of the national total; local expenditure now approaches 85 percent. The instruments developed to bridge that gap are themselves evidence of distributed administrative sophistication. The Local Government Financing Vehicle, pioneered in Wuhu, Anhui in 1998, became ubiquitous after 2008. By 2023, total LGFV debt reached roughly 50 trillion yuan, or 41 percent of GDP; Victor Shih’s UCSD analysis puts the total including shadow credit at 75 to 91 percent of GDP. More than twelve provinces now carry LGFV debt exceeding half their own provincial GDP. Revealingly, CKGSB reportingconfirms that LGFV financial distress is concentrated precisely where Gong and David’s paper identifies low-capacity provinces: Guizhou, Inner Mongolia, Ningxia, Liaoning, Qinghai.

Since around 2014, Chinese industrial policy financing has also flowed through Government Guidance Funds, public-private investment vehicles combining state capital with private VC and PE expertise. Georgetown’s CSET reportprovides the authoritative English-language overview. Between 2015 and 2021, roughly 2,000 guidance funds deployed around £850 billion; they now account for about 30 percent of all PE and VC capital raised in China. The architecture is deliberately tiered, as Boston University’s Global Development Policy Center explains. Beijing sets broad strategic priorities. Provincial and municipal authorities refine them with local specificity. Sub-funds typically invest primarily within their own jurisdictions.

The Hefei municipal guidance fund apparatus is the case that best illustrates the pattern. Hefei, the capital of Anhui province, is not a Tier-1 Chinese city. It does not have the fiscal resources of Shanghai or Shenzhen. Its guidance fund nevertheless took early equity stakes in both NIO and BOE Technology before either was a dominant firm, and helped transform Hefei into a national center for electric vehicles and display technology. It was not Beijing that picked NIO. It was municipal officials in Anhui, making allocation decisions with enough sophistication to pick winners that national-level funds had passed on, operating within a strategic framework Beijing had set but Beijing had not executed.

The answer has two parts. The center sets strategic direction, holds input costs low through state-set tariffs, directs credit at scale toward priority sectors, and disciplines incumbent commercial interests that threaten strategic coherence. The provinces and municipalities execute, competing with each other under a tournament incentive structure, financed through a tiered capital-allocation system that places real decision-making at the sub-national level. Remove either half and the system stops working. What looks like authoritarian command is the interaction of strategic central authority and deep distributed execution.

What makes it hold together?

The architecture that sustains this pattern is not a conspiracy and does not require anyone in it to be virtuous. It is a set of incentive structures that make both strong strategic center and distributed provincial competence the rational equilibrium.

Start with personnel. The Central Organization Department controls appointment and promotion for every senior cadre in the country. Pierre Landry’s 2008 book Decentralized Authoritarianism in China documents the core mechanism: officials are selected young, rotated deliberately across jurisdictions and functional portfolios, and promoted on the basis of measurable performance in their current posts. Li and Zhou’s 2005 Journal of Public Economics article shows that provincial leaders’ promotion probabilities track GDP growth in their provinces, creating a tournament in which officials compete for advancement by delivering measurable outcomes. Jia, Kudamatsu, and Seim (2015) find that competence and factional connections operate as complements rather than substitutes. Both performance and political reliability are rewarded, but the performance component is real.

Forty years of rotation-and-promotion tournament, operated at continental scale with real career consequences, produces a population of senior administrators who have run multiple jurisdictions, handled multiple portfolios, and been selected repeatedly for their ability to deliver results. The system’s pathologies are real: statistical falsification (Wallace 2014), anti-corruption purges that disproportionately target factional rivals, metrics gaming that produces visible outputs while obscuring substantive failure. But even the pathologies operate within a machine that has built administrative capacity at the provincial level. That is why Jiangsu can stand up a BRI leading small group in six months and why Hefei can pick NIO before anyone else does.

The fiscal architecture reinforces this. The 1994 tax reform created a structural imbalance between local revenue (below 40 percent of national) and local expenditure (now approaching 85 percent). Local governments had to develop financing mechanisms or default on their obligations. The LGFV system layered on top of state-bank lending is essentially a parallel municipal-bond market routed through land collateral, land-use-rights capitalization, and implicit central guarantees. It is not elegant. It has produced the distressed-LGFV crisis concentrated in the lowest-capacity provinces. But it works because the incentives reward local governments that can finance and deliver infrastructure, and punish those that cannot. The guidance-fund system layered on after 2014 does the same for industrial equity: reward provincial and municipal administrations that can identify and back winners, disadvantage those that cannot.

Put together, the deep provincial bench plus strategic central authority willing to act against incumbents is what produces results. Western attempts to copy Chinese centralization consistently pick the wrong half.

The United States consolidated its auto industry decades ago into the Big Three (GM, Ford, Chrysler) and has spent forty years losing global market share, punctuated by repeated bailouts, the most recent in 2008-2009 when two of the three required federal rescue. Japan, by contrast, kept its industry fragmented. Toyota, Honda, Nissan, Mazda, Subaru, Suzuki, Mitsubishi, Daihatsu, and Isuzu all still operate as independent firms, with cross-shareholdings but no merger pressure from Tokyo. Japan has dominated global auto markets for most of the last fifty years. China’s current EV industry structure (BYD, Geely, NIO, Xpeng, Li Auto, Zeekr, all operating as distinct competitors backed by different provincial coalitions) looks structurally closer to Japan than to Detroit.

The EU’s response to Chinese EV dominance is to propose consolidating its own fragmented auto sector into fewer, larger European champions. This is moving from the Japan model toward the Detroit model while citing Chinese success as the justification. The direction is inverted. China is not winning because of consolidation. China is winning because its center holds strategic input costs low, directs capital to strategic sectors, disciplines platforms that threaten strategic priorities, and its provinces compete intensely on execution within that strategic framework. Remove the provincial competition and you do not have China. You have Detroit in the 1970s, plus some extra paperwork.

The American side of the misreading is worse, because it is domestic. US commentators across the ideological spectrum praise Chinese centralization while their own sub-national governments are visibly failing. California is governed by Gavin Newsom, whose record includes the high-speed rail debacle described at the opening, $37 billion spent on homelessness while California’s unhoused population remains the largest in the nation, and an Employment Development Department that lost $33 billion to unemployment fraud during the pandemic. Texas is governed by Greg Abbott, whose tenure includes the February 2021 power grid failure that killed hundreds after the Public Utility Commission under Abbott’s appointees dismantled oversight mechanisms that could have required winterization, followed by a years-long political posture of claiming to have fixed a grid that independent assessments still find vulnerable. These are not minor officials. California is the world’s fifth-largest economy. Texas is the second-largest US state by GDP.

Compare this to the governors of Jiangsu and Guangdong. You almost certainly cannot name them. They run provincial economies larger than most OECD countries, coordinate tens of billions of dollars annually in guidance-fund equity, anchor overseas infrastructure initiatives that span continents, and do so within a tournament structure that would remove them from office if they failed to deliver. The anonymity is the tell. The American system does not produce provincial executives of comparable depth, and the commentators praising Chinese centralization are praising the theater visible to foreign observers while ignoring the machinery that actually does the work.

What is the West getting wrong?

China’s industrial performance comes from two machines running in combination, not one. The first is a personnel and fiscal architecture that produces a deep provincial administrative bench: rotation across jurisdictions, promotion by tournament on measurable outcomes, local fiscal autonomy routed through LGFVs and guidance funds. The second is a political economy in which central authority is willing to use its power against powerful domestic commercial interests when strategic priorities require it: industrial electricity prices held below market, state-bank credit directed to strategic sectors at scale, billionaires disciplined when they step out of line. The 87% of Chinese industrial policy that originates below Beijing implements the 13% that originates at the center, and neither half works without the other.

The Western mirror lacks both halves. It lacks deep provincial and municipal competence not because it needs a CCP-style cadre tournament (it doesn’t), but because Western parties reward factional politics over governing results. A party that actually wanted to deliver housing, trains, and grid capacity would produce competent provincial executives within a decade. None has tried in a generation. It lacks the central strategic will because its incumbent commercial interests are wired into the political coalitions that make central authority legitimate. US oil majors and European Green coalitions each hold veto power over their respective energy policies. The formal toolkit exists; the political economy prevents its deployment.

The inversion matters. When the West “centralizes” in response to Chinese success, it centralizes in the one category where even Xi regularly fails (forced consolidation, champion-building, merger liberalization), while refusing to centralize in the categories where Chinese centralization actually delivers. The EU’s Competitiveness Compass wants to consolidate 34 mobile operators into fewer champions while industrial electricity prices have widened to nearly double US levels and Apple and Google remain untouched. American industrial policy advocates want national champions while the US has not passed serious antitrust reform in a generation and has reduced rather than expanded domestic hydrocarbon production across two administrations.

The Western failure is not a missing Xi. Centralizing without both halves produces only protected inefficient monopolies like the American car industry, a political class slightly more empowered to defend them, and an industrial electricity bill that keeps going up. If you want the benefits of centralization, you need to govern and focus on cutting energy prices and the costs of capital. If you will not do those things, centralization will make you weaker, not stronger. This is what the West is about to learn, and the learning is going to be expensive.

Cities around the world keep losing their commercial cores. In America, suburban malls make the news when Macy’s announces another closure. In the United Kingdom, high streets bleed quietly, one boarded shopfront at a time, until the council notices that the business rates projections have gone hollow. In France, the cœurs de ville of mid-sized towns lose their merchants to out-of-town retail parks. In Japan, shotengai vacancy accelerates faster than municipal budgets can respond. In Korea, second-tier cities watch their downtowns empty while national urban-regeneration funding rises and falls with each election cycle.

The standard response in every country is familiar: do a very limited rezone the site (if we are lucky), subsidize the developer, build something maybe mixed-use (or luxury apartments with a brewery), declare victory, move on. This usually fails (in part not aggressive enough). When it works, as it did in Lakewood, Colorado, or in Margate, Kent, or in a growing list of French communes that have taken Action Cœur de Ville seriously, it works for reasons that have almost nothing to do with zoning and everything to do with the thing that is hardest to talk about: whether the local government involved is actually capable of executing a complex multi-year project with discipline. Most are not.

The reframe this piece turns on: commercial decline is a governance problem just as much it is a zoning/urbanist problem, which means it is also the best opportunity most cities will ever have to build real state capacity. Not a nuisance to be managed. A (dangerous) gym to build up local institutional muscle. We’ll come back to why.

Zoning reform (or just get rid of it) is necessary, and more than likely the prerequisite. Spend five minutes with the evidence on parking minimums, mixed-use prohibitions, and discretionary review timelines and the case becomes obvious. But zoning reform increases the number of attempts without changing the survival rate of each one. At the end of the day, businesses are probabilistic, not deterministic. A perfectly deregulated commercial district with great foot traffic and cheap rents will still see roughly half of its new restaurants fail within five years.

Which is bad for you, the policymaker, because you need tax revenue and visible results. Churn undercuts both: the storefront that closes after eighteen months doesn’t show up in the ribbon-cutting photos, and every visibly empty window is ammunition for the less development-friendly opposition you just beat to pass the reform in the first place.

So, if the odds are bad, can policy move the probability distribution: make more attempts possible, improve the survival rate of each attempt, catch the survivors before they fail for reasons unrelated to their underlying viability. What policy cannot do is make the distribution disappear.

What do YIMBYs and urbanists get from caring about what happens after a unit gets built? Everything. Both movements came out of substantive goals (affordability, jobs, quality of life) not procedural ones, and better local governance is what makes hard-fought reforms survive the next administration. Carmel (for urbanists) and Austin (for YIMBYs) didn't happen by accident.

The stakes are fiscal, not just about aesthetic

Before going further: this actually matters, and not for the reasons people who write about urbanism usually give.

Consider two shapes the damage takes. One fast, one slow.

The fast version is Pittsburgh Mills, in Frazer, Pennsylvania. One analysis documents assessed value falling from $148 million in 2018 to under $15 million by 2024. A 90% collapse of a local property tax base, from a single asset, in six years. Each dying lower-tier American mall also removes something in the range of $1.8 to $3 million in annual sales tax revenue from its host municipality. For some cities where one mall represents 20 to 30% of the local tax revenue. This is not a rounding error, and it’s clear that malls and downtowns are slow motion fiscal emergencies.

The slow version is Detroit, the case in any advanced economy where downtown and urban-core decline produced outright municipal bankruptcy, the largest in U.S. history, filed in 2013. By the mid-2010s, 23% of housing units were vacant, 36% of commercial parcels were vacant, and the property tax delinquency rate had skyrocketed to 54%. The bankruptcy was the visible event; the governance crisis had been developing for fifty years. Yes the city went bankrupted, and the main reason is the Great Recession. But before the recession, Detroit has a property tax base eroded to the point that even extraordinarily high rates cannot fund basic services, commercial vacancy that becomes structural rather than cyclical, and a fiscal situation that makes it easy for the Great Recession to break the city.

Detroit is an extreme example, but the mechanism is not unusual. In the UK, a high street with 30% vacancy generates substantially less in business rates than a healthy one, and the rates system punishes retail while exempting online commerce, so the distressed physical property pays more per pound of turnover than its online competitors. In France, communes rely on taxe foncière and the Cotisation Foncière des Entreprises; both erode with vacancy, and the TaSCom layer adds further exposure to peripheral retail that draws revenue away from the center. This is part of why the Opération de Revitalisation de Territoire framework, which lets participating communes block peripheral commercial permits, exists at all. In Japan, the fixed asset tax and city planning tax erode as shotengai vacancy accelerates, compounded by population decline and aging demographics that raise per-capita service costs. In Korea, property tax and local income surcharges follow the same curve, which is part of why the Urban Regeneration New Deal was framed as an emergency response to municipal fiscal deterioration rather than as ordinary urban policy.

The consequences are concrete. Schools, police, roads, transit, all funded from tax bases that this decline is dismantling.

The damage spreads outward. Cleveland research found that residential properties within 500 feet of tax-delinquent or foreclosed commercial sites experienced a 9.4% value loss between 2004 and 2009. We do not have Cleveland-grade contagion studies for every jurisdiction, and we want to be honest about that, but the qualitative pattern is consistent across UK high street research and Japanese shotengai work: a commercial core in visible decline depresses nearby residential desirability, slows new household formation, and accelerates the outflow of those who can leave. The residential tax base follows the commercial one, on a lag.

When a commercial district collapses, private service density collapses with it: not just retail, but clinics, pharmacies, banks, post offices, grocery stores, childcare. Residents still need these services, and local government either accepts degraded access or backfills at public cost from a tax base that’s simultaneously shrinking. Meanwhile the vacant properties themselves impose costs with no corresponding revenue: code enforcement, fire calls, policing, utility maintenance, planning department time absorbed by proposals that come and go. Across a high street with 30% vacancy or a downtown with more boarded plywood than open signage, staff time gets absorbed into reactive management rather than proactive anything.

Finally, the thing hardest to measure but that may matter most: when a city cannot keep its main commercial core functional, residents stop believing the local government is competent at anything else. That perception is downstream of the fiscal failures and upstream of everything the government wants to do next. A mayor trying to pass a transit bond, build affordable housing, expand school funding, or raise any tax for any purpose is negotiating from a weaker position if the commercial core of her city is visibly dying. A functioning commercial core is evidence of a functioning government. A dead one is evidence of the opposite, regardless of whose fault it actually was. Politics does not grade on fairness.

So why bother? Because this is the best forge you have.

Commercial decline is a serious governance crisis. It is also, paradoxically, an opportunity, and one most cities do not recognize while they are standing in the middle of it.

Three reasons.

First, the results are visible. A mayor who builds a pension fund’s long-term solvency is doing essential work nobody will ever thank her for. A mayor who revitalizes her downtown is doing work that every voter, every local business owner, every visiting relative can see and touch. The political economy of state capacity is usually cruel: the boring work that matters gets no credit, and the flashy work that does not matter gets all of it. Commercial revitalization is the rare case where the political incentive and the substantive outcome point in the same direction.

Second, the problem forces you to build capacity across every front at once. Most governance problems build one muscle. Passing a transit referendum builds lobbying capacity and not much else. Running a successful summer youth program builds contracting and program-management capacity in a narrow domain. Commercial revitalization is different. To do it well, a city has to build planning capacity (diagnostics, rezoning, urban design), financial capacity (grant writing, tax-credit stacking, developer negotiation), land-holding capacity (land banks, community land trusts, legal tools for moving hostile owners), business development capacity (the thing Singapore builds and almost nobody in the West has), and coalition capacity (anchor institutions, community organizations, state and national partners, successor administrations). This is the broadest-spectrum institutional workout any mid-sized city is ever going to encounter. And the capacities built here transfer. A land bank built to handle commercial properties can be retooled for housing. A coalition built to align hospitals and universities around commercial procurement can be redeployed for neighborhood health work. Anchor relationships outlast the specific project.

Third, if you fail, you still learn. The honest truth is that many cities will not fully succeed even when they try. The probabilistic nature of business formation, the hostility of absentee owners, the vagaries of national economic conditions, and the simple slowness of institutional change mean that perfect success is rare. But attempted revitalization produces institutional learning even when the specific project underdelivers. A city that tries and partly fails at a mall redevelopment has a planning department that now knows how to negotiate with developers, a finance office that has learned to work with CDFIs, and a place manager who understands retail. A city that does not try has none of those.

One caveat. This argument applies to cities that already have the problem. A city without commercial decline should not manufacture a commercial project to build state capacity. Different problems make different forges.

One more thing worth naming: the ownership asymmetry

Before turning to what the conventional response gets wrong, one structural observation that runs underneath everything else.

American enclosed malls typically have one owner. Downtowns, high streets, Singaporean strata malls, and most European commercial districts do not. This changes the problem fundamentally, and not in a simple direction.

A single-owner commercial district with an aligned owner is the fastest-moving case in the entire revitalization literature. One negotiation produces site control. No 80% strata consent. No hundreds of individual landlords to chase. One counterparty, one term sheet, one closing. When that alignment exists, nothing in the distributed-ownership literature moves as fast.

A single-owner commercial district with a hostile or indifferent owner is among the hardest cases any mayor faces. A REIT holds the mall through bankruptcy. A successor lender takes it back in foreclosure. An absentee investor holds it for tax loss, or for the option value of eventual land appreciation, and has decided that doing nothing is the most profitable choice. The city then finds itself with one very powerful stakeholder who cannot be moved by any land-use tool the American system currently provides. Eminent domain exists in theory but is politically radioactive and legally uncertain for commercial property. The UK’s High Street Rental Auctions do not exist in America. French préemption commercial does not exist in America. The city is stuck.

A downtown, a high street, or a shotengai is neither: always a collective action problem, slower than the aligned-owner case and more workable than the hostile-owner case. Slower because coordinating many owners takes time, and the tools for forcing coordination are weak in most jurisdictions. More workable because the city can build coalitions with owners who want to cooperate even when others refuse to engage. Progress is incremental but not blocked.

The tools that work in one case often don’t work in the others. Land banks, HSRAs, préemption, community land trusts: these are distributed-ownership tools that consolidate what the market cannot. At a single-owner mall they are largely beside the point; the question there is whether the counterparty can be moved, and if not, whether the city has the legal and financial staying power to make inaction more expensive than cooperation over time. Conversely, the kind of negotiation discipline that makes single-owner mall redevelopment work is not the skill a downtown revitalization actually needs. A downtown needs coalition capacity, patience, and the ability to hold many small wins together long enough to add up to something.

Much of what makes mall redevelopment look successful in the American literature is selection bias: the cases we read about are the ones where the single owner was eventually aligned. The cases where the owner was not aligned do not become cases. They become permanent vacancy, because nothing in the toolkit can move them.

Why the conventional response doesn’t work

The standard move (rezone, subsidize, build mixed-use, attract tenants) has two problems. First, it treats the upstream cause as if it were the downstream symptom. Second, it assumes that someone competent is going to execute.

Commercial decline is downstream of institutional failure, not upstream. Fragmented ownership, absent governance, thin planning capacity, weak business development, disinvested social fabric: these produce the vacancy, not the other way around. Fixing the buildings without fixing the institutions produces what the UK High Streets Task Force, after working with 149 English high streets over five years, honestly called activity without transformation.

The evidence on delegated private management is not subtle. Between November 2020 and June 2021, three major publicly traded mall REITs filed for Chapter 11: CBL & Associates, Pennsylvania Real Estate Investment Trust, and Washington Prime Group. CBL alone was managing 107 properties totaling 66.7 million square feet across 26 states. PREIT filed for bankruptcy a second time by 2024. Mall management firms are doing exactly what they were designed to do. That is the problem. Their obligation runs to investors and creditors, which translates into optimizing net operating income. No accountability to the municipal tax base, the surrounding neighborhood, or the small businesses that depend on foot traffic. And when one entity owns 100-plus malls and its financial model fails, the whole portfolio fails at once: fast, visible, bankruptable. The cities hosting those malls also lose control at once; bankruptcy reorganizations are slow, and cities that had been negotiating with CBL suddenly found themselves negotiating with a lender, then with a creditor committee, then with whichever entity emerged from reorganization. Each handoff reset the relationship. A city without sustained capacity across years loses the thread entirely.

The downtown equivalent is less visible because it rarely produces a single failure event. Instead, you get the slow distributed version: hundreds of absentee landlords making individually rational decisions that collectively produce vacancy, decay, and no one to hold accountable. The principal-agent problem is harder to name when it is distributed across consultants, developers, and BIDs rather than concentrated in a REIT. It is the same problem, less legible.

Business Improvement Districts are the institutionalized version of the delegation move, and the evidence is more differentiated than either side of the BID debate tends to allow. The NYU Furman Center found that large BIDs in dense Manhattan predicted a 15% commercial property value gain over ten years, though Ellen, Schwartz, and Voicufound the gain disappears when a third of the BID budget goes to the executive director, and mid-sized and smaller BIDs showed no discernible effect at all. The deeper critique, from a fifty-year reappraisal of BID scholarship: BID governance excludes renters, workers, and residents, so BIDs optimize for what property owners value (security, cleanliness, brand attractiveness) rather than what communities need (affordable space, diverse business mix, support for incumbent independents). Our point is narrower. BIDs can clean streets. They cannot improve the survival probability of a merchant whose business model has been disrupted.

Japan offers the most instructive cautionary tale. The machizukuri sanpo, a package of three laws enacted between 1998 and 2000, was Japan’s attempt to address commercial core decline at national scale. The research is sobering: these laws did not revitalize city centers. An MIT study found Town Management Organizations made negligible contributions, largely because TMO boards were dominated by commercial stakeholders, funded by government without performance pressure, and participated in by residents only nominally. This is, almost exactly, the accountability failure of American BIDs translated into a different national context. When the same institutional design fails in two countries for the same reasons, the design is the problem.

The conventional response fails because the conventional framing is wrong. The question is not how to fix this mall, this downtown, or this high street. It is what capacities the decline is asking us to build.

What the reframe implies

If commercial decline is a capacity problem rather than just a simple real estate one, then the question becomes: what, concretely, does a city have to build?

Five things. They are not a toolkit. A toolkit implies the tools are the point. They are not. The capacities are. Each of these capacities, once built, transfers to every other hard problem a city faces. Zoning reform sits inside the first capacity. It is necessary. It is also one part of one capacity.

Capacity 1: Planning, with urbanism as a cross-cutting layer

Planning capacity is the ability to diagnose, design, and execute changes to the built environment that survive legal challenge and political turnover. Urbanism (how blocks, streets, uses, and transit interact to generate foot traffic) lives inside this capacity. It is one lever among several, or even the prerequisite for all other levers. Foot traffic is necessary for commercial activity; foot traffic alone produces tourists, not functioning commercial districts.

What the capacity actually includes: the ability to run a commercial district diagnostic, with every vacant property, every owner, every vacancy duration, and every business quality assessment tracked. The ability to execute rezoning that survives legal challenge: upzoning, mixed-use overlays, form-based codes, all of which most cities in every jurisdiction we examine have to relearn. The ability to reform (or eliminate) parking minimums, which in most jurisdictions is administrative rather than legislative and yet remains one of the most underused year-one levers available. And the ability to steer public facilities to where they do the most commercial good.

That last one is the Japanese contribution, and most planners outside Japan haven’t absorbed it. Field research returns to a simple finding: concentrating city hall offices, hospitals, clinics, libraries, and kindergartens in or adjacent to the downtown core creates the resident demand commercial activity requires. When institutions relocate to peripheral sites, they take their customers with them. Japan has built this into planning law through the Location Optimization Plan under the 2014 revision of the Law on Special Measures for Urban Revitalization, which requires municipalities to designate urban function concentration zones and residential guidance zones. Toyama City is the best-documented case. Its “dumplings and sticks” concept (residential areas as dumplings, transit lines as sticks connecting them to the core) has concentrated population and commercial activity around downtown. A mayor anywhere can borrow the discipline without borrowing the statute: think about where public investment locates, what it draws around it, and how transit ties the result together.

Villa Italia Mall in Lakewood, Colorado closed in 2001 with over 50% vacancy, and was rebuilt as Belmar: 80 shops, 20 restaurants, 300,000 square feet of office, 3,500-plus residents, nine acres of parks. It now generates roughly $17 million annually in tax revenue, more than four times what the old mall produced. Three things made this work, none of them zoning alone. The city acted early, before the tax base had collapsed. It recruited the developer rather than responding to a developer proposal, which preserved negotiating leverage. And it broke the 104-acre site into 22 city blocks with a conventional street grid, which turned the urban design into its own governance discipline.

Chattanooga, Tennessee is the downtown equivalent, over forty years rather than ten. By the 1980s, Chattanooga had a deteriorating downtown and an industrial pollution problem that had earned it the label of America’s dirtiest city. What followed (a ULI panel in the mid-1980s, the Vision 2000 initiative, the River City Company seeded with $12 million from foundations and banks, the Tennessee Aquarium, the pedestrianized Walnut Street Bridge, a 22-mile riverwalk) produced downtown retail vacancy below 2% by 2016 and 23% downtown residential population growth from 2000 to 2015. Belmar and Chattanooga are not copies of each other. They share a deeper pattern: accountable governance with political stakes, simultaneous attention to container and content, and sustained institutional commitment across election cycles rather than within them.

When this capacity is absent, it looks like Youngstown. Youngstown, Ohio lost 60% of its 1930 population and in 2005 adopted planned contraction as its governing strategy, the first major American city to formally accept that it would be smaller going forward. The plan was visionary. MIT research found it was only partially implemented as zoning, because of political, legal, social, and economic resistance. The gap between the plan and the zoning is the capacity gap. It is common, and it is expensive.

Capacity 2: Financial capacity

Financial capacity is mostly an administrative problem, not a money problem. The federal toolkit exists. Most small cities cannot access it. The same logic applies, with different specifics, across the OECD.

The most underused lever we found is the Federal Historic Tax Credit. Since 1976, it has leveraged $257.8 billion in private investment and produced 3.4 million jobs. Half of all projects are in neighborhoods at or below 80% of area median income. The National Trust reports a $1.20 return in tax revenue per dollar invested. For a mayor with an eligible historic building at risk of demolition (a downtown hotel, a Main Street bank, a former department store), the HTC is often the most powerful tool available, because it makes private rehabilitation financially viable without requiring public capital. No new legislation needed. The State Historic Preservation Office is already staffed to walk developers through the certification process. The reason the HTC is underused is not policy; it is that most municipal planning departments do not know it exists. European and Asian equivalents (UK Historic England grants and Architectural Heritage Fund loans, French monuments historiques fiscal advantages, Japanese and Korean cultural property designations) are narrower in scope but exist.

Community Development Block Grants are the most flexible U.S. federal program for commercial revitalization, and Iowa’s Downtown Revitalization Fund is a well-documented state deployment. The catch is that accessing CDBG requires a Consolidated Plan, HUD compliance, and reporting capacity; for smaller cities, the first investment is often staff, not projects. New Markets Tax Credit offers a 39% federal credit over seven years but requires a CDFI partnership most cities do not have. Both tools are real; both require partner capacity that is itself the thing to invest in.

TIF deserves honesty about its record. It is authorized in 49 of 50 U.S. states, with at least 10,000 districts nationwide, and David Merriman’s Lincoln Institute review of 30-plus studies concluded that in most cases, TIF has not accomplished the goal of promoting economic development. The “but-for” test gets gamed. Retail TIFs shift activity without creating it. In Chicago, roughly $660 million (nearly a third of all city property taxes) flows to TIF districts, shielded from standard budget oversight. TIF is defensible when paired with a specific project and rigorous “but-for” discipline, as in Belmar. It is least defensible as a general downtown subsidy program. The capacity question is whether your city can run the “but-for” test with integrity. If the answer is no, TIF is not the right tool, regardless of what neighboring cities are doing with it.

Main Street America remains the most evidence-backed template for downtown commercial district management in the U.S., though its reported statistics are self-reported by participating communities rather than independently evaluated. The $115 billion reinvested, the 815,000 net jobs, the $18.03 per dollar return: treat these as pattern evidence rather than causal estimates. The finding that holds up across independent research bodies is narrower and more important: Texas Main Street is explicit that effective downtown revitalization requires paid staff. Volunteer management underperforms paid management. This appears in the Main Street literature, the BID literature, and the UK High Streets Task Force findings. When three independent bodies of research agree on something this specific, it is worth believing.

Capacity 3: Land-holding capacity

Land-holding capacity is the ability to acquire, hold, and redirect distressed commercial properties when private markets will not, whether those properties are dead anchor stores, vacant downtown blocks, or individual high street shops an absentee landlord refuses to lease. Fragmented ownership is the single largest structural barrier to adaptive reuse we encountered in every jurisdiction, and it is also where the international evidence diverges most sharply. Different countries have given their local governments very different powers to act on it.

The American tradition is the land bank, where state enabling legislation permits it. Dan Kildee’s Genesee County model has inspired over 250 similar efforts nationwide, and land banks have done particularly important work in legacy-industrial cities with large downtown and near-downtown vacant inventories, including Detroit itself, where the land bank is now one of the largest landowners in the city. Where enabling legislation is absent, advocating for it at the state level is a medium-term investment. The Champlain Housing Trust, seeded with a $200,000 city grant under then-Mayor Bernie Sanders, now stewards 3,000-plus housing units and 160,000-plus square feet of nonresidential space, and over 200 other CLTs have followed the Burlington template. Commercial CLTs are a newer and less-evaluated model. The Dudley Street Neighborhood Initiative in Boston’s Roxbury, uniquely granted limited eminent domain authority, now stewards over 30 acres including commercial buildings. Systematic evaluation of commercial CLT outcomes is essentially absent; the case rests on governance logic and on the stronger housing CLT track record.

The French tools are the ones Americans and British mayors should genuinely envy. The droit de préemption commercial, codified in 2005, lets a commune designate a périmètre de sauvegarde, receive automatic notification of any sale of a fonds de commerce within that perimeter, and exercise preemption within two months. The commune must then rétrocède the property within two years to a merchant whose activity preserves commercial diversity. A 2023 Conseil d’État ruling clarified that this has to be used for a genuine development project, not as a blocking mechanism. The constraint is that the commune needs capital to buy and manage until rétrocession. The program is embedded in Action Cœur de Ville, literally Action for the Heart of the City, which since 2018 has mobilized more than €5 billion across 244 mid-sized communes. Phase 1 results include a 15% footfall increase and 91% elected-official satisfaction. The Opération de Revitalisation de Territoire framework lets signatory communes suspend peripheral commercial permits that undermine the center. A direct anti-peripheral mechanism. No Anglo-American equivalent exists.

The UK’s new power is High Street Rental Auctions, operational since December 2024. A property vacant for 365 or more days in a 24-month window can be designated, noticed, and, if the landlord fails to let within eight weeks, auctioned for a 1-to-5-year lease. Early adopters include Bassetlaw, Darlington, Mansfield, and Rugby. Industry analysis flags real capacity constraints around surveying, legal fees, and judicial review risk. In Broxtowe, the threat alone was sufficient to prompt a landlord to act without the council completing the process. The carrot-and-stick effect is real.

The most reproducible move in the distributed-ownership case is activation before renovation. Japanese field research consistently finds that creating reasons for people to come to the commercial center (farmers markets, festivals, concerts, whatever) before investing in physical improvements works without large capital budgets. The related move is cheap space for new entrepreneurs, removing the financial barrier to entry; the MAD City project in Matsudo subleased nearly vacant buildings cheaply without restoration obligations. Germany arrived at the same insight under harder conditions: Leipzig’s population fell from 700,000 to under 500,000, housing vacancy hit 20%, and Stadtumbau Ost, launched 2002 with €2.7 billion in federal and state funding, allowed demolition as a legitimate tool while managing interim vacancy through Zwischennutzungen: community gardens, artist spaces, green infrastructure. Leipzig was growing again by 2008.

Returning to the single-owner mall question. When the owner is aligned, none of this matters. Land-holding capacity is beside the point, because the city can negotiate directly for site control. When the owner is not aligned, none of this works either, because the distributed-ownership tools assume fragments to consolidate. What the city needs instead is a different set of tools: receivership frameworks that can take properties from negligent owners, mandatory maintenance codes that create holding costs, land-value taxation that makes sitting on vacant commercial property expensive, and legal staying power across multiple administrations to pursue these remedies to their conclusion. Few cities have built these well. This is among the largest gaps in municipal capacity anywhere, and one of the hardest to close, because most of the fixes require state or national legislation that varies by jurisdiction. We return to this in the advocacy agenda.

Capacity 4: Business development capacity

Business development capacity is the ability to improve the survival probability of the businesses operating in a commercial district, not just to rehabilitate the buildings around them. This is the most underbuilt capacity in every Western city we examined, and it is where international comparison yields the clearest transferable lessons. Physical rehabilitation creates the conditions for commercial improvement. It does not produce it. A renovated storefront occupied by a low-quality business is still a low-quality business.

The institutional state of the art is Singapore’s Heartland Enterprise Centre, established in 2019 and fully funded by Enterprise Singapore. HECS assisted more than 2,000 heartland enterprises in 2024 alone. The Heartland Enterprise Placemaking Grant funds merchants to organize promotional events. The Enhanced Visual Merchandising Programme helps shops refresh storefronts. The Heartland Innovation and Transformation programme places selected merchants in incubator spaces (Sprout@AMK in Ang Mo Kio is the first) for twelve months of mentorship and concept testing, after which graduates can apply for subsidized HDB shop space. The 2026 Singapore budget expanded the program further as a deliberate instrument of neighborhood commercial quality uplift. This is not generic small business advice. This is the machinery of moving the survival probability distribution.

Two of HECS’s four components are within reach of most mid-sized Western cities within a mayoral term, a third requires foundational land-holding work first, and the fourth is probably not coming. Placemaking grants for merchant-organized events are fully transferable: any Main Street organization or chamber of commerce can administer something equivalent on a modest budget. Visual merchandising technical assistance is partially transferable with one dedicated hire paired with an existing place-management program. Incubator space with subsidized post-graduation retail depends on public ownership of commercial space (HDB, in Singapore’s case), which requires a commercial CLT, a land bank, or structured public-private leasing to replicate. And the coordinated inter-agency support across planning, housing, and heritage (Enterprise Singapore acting with URA, HDB, and the National Heritage Board under a single ministry) is the component that genuinely requires Singapore-level state capacity. No American or British analogue exists. Planning around its absence is more useful than pining for it.

Seoul’s Sewoon Sangga project shows the same dual-capacity pattern at a different scale, in a downtown megastructure rather than a neighborhood heartland. When Mayor Park Won-soon reversed the demolition decision in 2015, the plan attended to both container and content. The Suri Cooperative Association linked Sewoon’s craftspeople to new customers: business development organized explicitly to protect existing commercial ecology during regeneration. Per Brookings: visitors tripled, vacancy fell from 30 to 18 units, and the new tenant mix (VR, robotics, CNC manufacturing alongside traditional electronics repair and printing) reflects deliberate ecology-preservation strategy. The caveat is obvious: Seoul Metropolitan Government has professional staff that Daejeon, Gwangju, and Changwon do not, but the institutional move is portable even if the scale is not.

Can’t we just simply regulate businesses? Well, San Francisco’s formula retail ordinance is the closest the Western literature gets to moving the distribution through regulation (which is going as well as you can expect). In controlled neighborhood commercial districts, 10% of businesses are formula retail, versus 25% elsewhere, though the causal attribution is genuinely contested and the Planning Department itself recommended raising the threshold from 11 to 19 locations. Formula retail restriction works best in high-demand markets where vacant storefronts fill quickly; in struggling districts with chronic vacancy, it may leave you with empty storefronts rather than local ones. It does not improve the odds of the individual business.

Capacity 5: Coalition capacity

Coalition capacity is the ability to convene, align, and sustain the institutions whose cooperation determines whether any of the other four capacities produce durable outcomes. Anchor institutions. Community organizations. Business associations. State and national partners. Philanthropic funders. Successor administrations who will inherit whatever you leave behind.

The most documented American template is the Cleveland Greater University Circle model, which is a near-downtown case, not a mall case. Cleveland Clinic, University Hospitals, and Case Western Reserve University (convened by the Cleveland Foundation) aligned hiring, procurement, housing, and transportation commitments across seven surrounding neighborhoods. The outcomes: 859 new housing units, the Evergreen Cooperatives with 225 employees in worker-owned businesses linked to institutional supply chains, $44 million in transportation infrastructure. The Affordable Care Act requirement that nonprofit hospitals conduct community health needs assessments every three years created a policy lever U.S. mayors did not previously have. European equivalents exist but are patchier; the UK’s NHS Integrated Care Systems and the French contrats locaux de santé framework both open convening opportunities that local governments underuse.

Margate shows sustained council commitment at the UK scale: a Victorian seafront town with a decaying core rather than either a mall or a classic downtown, but structurally the same problem. The Turner Contemporary gallery, opened 2011, plus Thanet District Council’s compulsory purchase and restoration of Dreamland, plus more than a decade of partnership between Thanet and Kent County Council. Reported outcomes: 3.9 million gallery visits, 19% tourism sector growth between 2013 and 2015, £47 million added annually, 300 jobs at the two anchors, £22.2 million in Levelling Up funding. Disentangling Turner Contemporary from concurrent Kent tourism trends is genuinely hard, and the gallery itself acknowledges that not all benefits have been shared by local residents; pockets of deprivation remain. Margate generates visitors and property investment. It does not, by itself, reduce incumbent-resident poverty. The institutional lesson is the sustain itself: more than a decade of coordinated council commitment, not a flagship intervention.

Chattanooga works at this level too, the same case from Capacity 1, doing different work. The coalition half of the Chattanooga story is the 40-year commitment: ULI panel, Vision 2000, the River City Company with its seed capital from eight foundations and seven banks, held together across administrations. Critics reasonably note the benefits flowed unevenly, and this is not a poverty-reduction case. But the institutional sustain is real, and it was produced by coalition capacity, not any single tool.

What happens without this capacity is summarized most clearly by the UK High Streets Task Force. The task force engaged 149 high streets over five years. Its final finding: the main barrier to high street revitalization was the lack of suitable local governance structures. Not ideas. Not funding. Not expertise. Governance. 40% of places visited needed stronger partnerships before any substantive work could begin. You can recommend as much as you like, but it often does not make any difference if the resources aren’t there to deliver. That is the single most important sentence in the five-year task force report.

Korea institutionalizes this well in principle. The Urban Regeneration Special Act 2013 requires each designated urban regeneration area to have a Support Center, a local technical assistance body staffed by planners, designers, and community organizers, with a five-year local activation team that keeps projects alive. The institutional structure is sound. Korea’s problem is that national funding has fluctuated sharply across political cycles, and coalition capacity at the local level is the only thing that partially protects against that.

A (admittedly cliché) call to action

Most cities are not short of space. They have enormous amounts of it. Dead mall pads. Hollow downtown blocks. Half-empty high streets. Vacant shotengai. Shuttered cœurs de ville. What they are short, is ideas, which at the institutional level means capacity. For the cities that have this problem (and hundreds across every country in this piece do), the commercial district is not just a crisis to manage. It is the best opportunity those cities will ever have to build the state capacity they need for everything else. The results are visible, which means voters will reward the work. The problem demands five capacities at once, which means the institutional muscle transfers to housing, transit, climate, and whatever comes next. Even partial success produces real learning.

A mayor who thinks she is fighting to save a mall is fighting the wrong war. A mayor who thinks she is building the institutions her city will need for the next thirty years is fighting the right one, and using the mall as the forge.

Every dead mall, every hollowing downtown, every emptying high street, every struggling shotengai, every shuttered cœur de ville, these are the training ground for every other hard problem a city will face.

The question is not whether to intervene. The question is whether the intervention leaves the city more capable than it found it.

Appendix: the five-minute country cheat sheet

For practitioners who want the jurisdiction-specific starting point.

• United States: Main Street America, Center for Community Progress (land banking), LISC (CDFI and financing), NPS Technical Preservation Services (HTC), Democracy Collaborative (anchor institutions), National League of Cities.

• United Kingdom: Institute of Place Management, High Streets Task Force legacy, Power to Change, BID Foundation.

• France: ANCT (Action Cœur de Ville, Petites Villes de Demain), Banque des Territoires, CEREMA.

• Japan: Cabinet Office Urban Revitalization HQ, MLIT.

• South Korea: MOLIT, URIS, Seoul Solution.

• Singapore (reference model): URA, Heartland Enterprise Centre

• The desire gap is the opportunity. People already want more children than they’re having. The job isn’t persuasion; exposure to kids alone increases fertility ideals, it’s building environments where existing desires can be acted on. Moralizing about young men needing to shape up or young women not having the right values doesn’t close the gap and won’t close the gap.

• Build the coalition and infrastructure subnationally Local governments can move completed family sizes; national governments control entry into parenthood. Stop waiting for national reform. Build working ecosystems at the municipal and provincial level now, capture the gains available there, and assemble the political coalition of places that have actually delivered.

• This is a product/service delivery problem, not a culture war Every place that sustained above-average fertility built an integrated system, held it across administrations, and made it visible in everyday life. They didn’t change anyone’s values. They built an environment where existing values could be executed on. Integration and constancy of purpose are the product. Everything else is downstream.

The mothers in Nagi who arrive with plans for one child and leave with plans for two aren’t being persuaded of anything. They wanted more children before they got there. The previous environment didn’t let them have them. Nagi did. The wanting was already there. The system finally caught up to it.

Hold that in mind. It’s the shape of the argument.

When Steve Jobs walked on stage in January 2007 and introduced the iPhone, he wasn’t selling a device. He was selling a proposition: that the Mac would still work with it, that the App Store would be stocked next year, and that whatever Apple announced in 2012 would still run on the thing you bought in 2007. And then there was always “one more thing.” The phrase was pure theater, but it worked because there was always actually one more thing, and the one more thing always connected to every other thing he’d already shown. The showmanship and the system were the same act. But the showmanship only sold because underneath it was an obsessively integrated system. Hardware talked to software talked to services talked to retail, and every seam was somebody’s job. The phone was worth more because the App Store was stocked. The App Store was worth more because the phone was in your pocket. The Mac was worth more because both synced to it. None of the components did much alone; together they produced something that felt like magic to the customer, and the magic was real because the machinery underneath it was real. People who didn’t know they wanted a smartphone discovered they wanted one because the environment made smartphones feel like a thing to have, and concentration of early adopters changed what the next wave of customers could imagine.

Apple is the one widely understood case of what an ecosystem does. The ecosystem isn’t the Mac, the App Store, the iPhone, or the repair network. It’s the way each piece makes the others more valuable. It requires both: the razzle & the dazzle that makes people want to enter the system, and the tick-tick-tick of a system that actually delivers once they’re inside.

Fertility policy has almost no equivalent. It has components: baby bonuses, parental leave, childcare subsidies, housing grants. Most countries spend real money on some of them. The aggregate numbers keep falling. But here’s the thing: the components mostly work. Korean baby bonuses produce babies. German parental leave reforms produce conceptions. Italian childcare expansions, above a threshold, produce births. Evaluated individually, the programs do what they’re designed to do, with some doing much better than others. Evaluated as a system, they produce the world’s lowest fertility rates. The reason is that nobody designed the system.

A baby bonus that lands in a city with affordable family housing, available childcare, and neighbors who used the same bonus last year is not the same intervention as a baby bonus that lands in a city without those things. Same check. Same amount. Completely different product. The check didn’t change. The environment it landed in changed everything about what the check could do. That’s the alchemy, and it’s the same alchemy Apple was selling: the pieces make each other work, and the felt experience of the pieces working together is the thing that actually shifts decisions.

But alchemy that works has machinery underneath it. The baby bonus lands right because someone built the housing, someone staffed the childcare, someone made sure the programs talked to each other instead of each department optimizing its own line item while the family fell through the gaps between them. The central government runs cash transfers. The province runs its own programs. The city runs still others, with funding from local taxes. No one is responsible for whether they cohere. Each layer optimizes its own components. The family doesn’t experience the components. The family experiences the gaps.

On top of this, the whole arrangement has to survive an election, and a new mayor/governor/president/PM, and a budget cycle, and the moralist on the op-ed page who thinks the problem is that young people have the wrong values (oftentimes the right accusing women, the center/left accusing men, and sometimes the inverse), and the minister who wants to announce something visible before the next vote rather than sustain something invisible that compounds over fifteen years. The alchemy requires the machinery. The machinery requires the politics. A few places have figured out how to hold all three within the layer they control, and the difference shows up in fertility numbers that sit two to five times apart under identical national laws.

Same country, same laws, five times the children. Why?

Two young Korean women, same age, same income, same education, same national fertility policy. One lives in Yeonggwang County. One lives in Busan’s Jung District. Yeonggwang’s fertility rate is 1.71. Busan Jung’s is 0.32. Economic models don’t predict this. Cultural models don’t predict it. Ideological and religious models don’t predict it. What’s the variable nobody is measuring?

In 2009, a research group put childless women in fMRI scanners and showed them photographs of infant faces. The women’s reward circuitry (the nucleus accumbens, the part of the brain that fires when someone you love walks into the room or when you eat a perfect peach) lit up before they consciously processed the image. Later work using EEG clocked the response at roughly 140 milliseconds. The women didn’t decide to find the babies cute. The brain had already responded by the time the deciding part woke up. The response was graded: more infantile features produced stronger responses, in women who had never been pregnant, who weren’t thinking about pregnancy, who in some cases didn’t want to be pregnant. Something in them was keeping score of how many babies they had encountered. They didn’t know they were keeping score.

The women didn’t decide (she did, but you get the point). That’s the part I keep coming back to. The empathy response is decision-making before the deliberation, the part that runs on perception before the reasoning wakes up. It’s hard to build policy around a 140-millisecond response. It’s also hard to build policy while pretending it isn’t happening.

And the 140-millisecond response doesn’t stop at desire. It reaches all the way into marriage. In Japan, somewhere between 30 and 39 percent of never-married adults in their 20s and 30s say they are uncertain whether they want children. Not opposed; uncertain. brought my attention to Fumiya Uchikoshi and colleagues at Harvard, Princeton, Tokyo, and Gakushuin’s paper. They tracked what those uncertain adults did over the following years, in a paper published this year in Population and Development Review. Controlling for whether they wanted to marry (which in conventional models would be the variable doing the work), they found that uncertainty about fertility cuts the annual marriage transition rate by 30 to 50 percent. People who don’t know whether they want children marry later, or don’t marry.

The arrow runs the other way. Marriage doesn’t cause children; the prospect of children causes marriage. They call it child-driven marriage, and the reframes where the causation you assumed was backward tend to be where the actual leverage lives.

So the woman in Yeonggwang and the woman in Busan aren’t just in different policy environments. They’re in different neurological environments. The woman in Yeonggwang walks past children in playgrounds, parents pushing strollers, neighbors visibly managing two and three kids. Her brain is accumulating 140-millisecond responses before she deliberates anything, and the accumulation shapes her fertility desires, and the desires shape whether and when she marries. The woman in Busan walks past convenience stores and study cafés. Same brain. Same circuitry. Different inputs. The variable nobody is measuring is the density of the environment the circuitry is responding to.

What if TFR isn’t really one number?

The density is measurable, but the measurement most people use throws the signal away. When Stephen Shaw decomposed the Total Fertility Rate across 314 million mothers in 33 countries, he found that what looks like one number is actually two numbers multiplied together, and the two don’t move together. He checked it three ways (breakpoint co-occurrence, wavelet coherence, mutual information) and confirmed statistical independence. The share of women who become mothers at all (call it TMR, the Total Maternal Rate) drops in sharp shocks and doesn’t recover between them. The average family size among mothers (CPM, Children per Mother) stays roughly flat for decades. TFR averages them together and throws away 49% of the information in the process. A single TFR value of 1.5 corresponds, on average, to thirteen distinct combinations of the two underlying numbers. The thing every fertility paper for sixty years has been measuring is half a measurement, and the half that’s missing is the half that tells a mayor what to do.

Here’s the simplest way to hold it: TFR is an average of two different decisions made by two different people at two different life stages. The 27-year-old deciding whether to become a mother at all is deciding TMR. The 33-year-old mother of one deciding whether to have a second is deciding CPM. Policies that work for one often don’t work for the other. Forces that move one often don’t move the other. Sixty years of policy has been optimizing against an aggregate that hid both failure modes.

Shaw’s shock pattern has the right time signature for an environmental mechanism, with a wrinkle the smooth-decline reading misses. TMR doesn’t drift downward; it drops in discrete shocks and ratchets down. Italy, Japan, and the UK all show sharp TMR falls around 1974. The US shows the same shock pattern in 1971 and again in 2008. Shaw’s prefecture-level analysis of Japan is the cleanest single piece of evidence: TMR fell in all 47 Japanese prefectures simultaneously between 1974 and 1975, with a probability under the null of less than 0.0001, as improbable as flipping a coin 47 times and getting heads every time. The breakpoint cluster in monthly birth data begins in October 1974 and peaks in November, which, accounting for conception and gestation, lines up almost exactly with the Japanese Oil Shock of October 1973. CPM stayed roughly flat across the same period. The shock altered who became a mother at all. It barely touched how many children mothers ended up having.

This gives the transmission mechanism its time signature. A shock (economic crisis, labor-market upheaval) disrupts fertility-desire formation in the cohort experiencing it. That cohort’s lowered birth rate depletes the next decade’s ambient infant exposure. The empathy channel runs at lower amplitude on the following cohort. Their fertility desires form against a thinner environment. Their TMR ratchets down again at the next shock and doesn’t recover between shocks because the environment never refills. CPM stays stable because it’s a different decision, made by people already inside parenthood, responding to local rather than national conditions. The smoothness of the aggregate decline was an artifact of TFR aggregation.

Put Shaw’s ratchet together with Uchikoshi’s child-driven marriage, and East Asia’s marriage collapse stops being mysterious. Atoh’s work and recent estimates by Tan, Cui, and Uchikoshi attribute 40 to 70 percent of TFR decline in Japan, Korea, Taiwan, and Singapore to falling marriage rates. The conventional reading is that fewer people are marrying, so fewer children are born. The better reading is that an earlier generation’s TMR shock depleted the ambient infant environment of the next cohort, which produced uncertain fertility desires in unmarried young adults, which depressed their marriage rates, which produced still fewer children (the next TMR step down), which further depleted the environment for the cohort after that. The marriage collapse is downstream of a fertility-desire collapse, which is downstream of a depleted environment, which is downstream of TMR shocks the previous generation absorbed and never recovered from. The trap closes itself.

Is the whole problem just economics?

None of this is to say the empathy channel is the only thing going on. Rising education costs, women’s labor-market opportunities, contraception, the genuine economic burden of raising children: all do real work in the fertility decline. Galiani and Sosa are explicit that the empathy channel is one mechanism among several, additive to economic pressures rather than replacing them. Shaw is similarly careful that his shock-and-ratchet pattern shows association rather than definitive causation. What the combination makes visible, and what sixty years of TFR-anchored policy never had to confront, is that one of the actual drivers of the decline has an environmental leverage point that fiscal instruments cannot reach. The economic factors still matter. They are not the whole picture, and the part they don’t cover is the part where felt environment compounds over generations.

TMR and CPM respond to different forces and require different interventions. In most developed countries both have been declining, meaning the aggregate decline reflects two different decisions made by different people at different life stages. The 27-year-old deciding whether to enter parenthood at all is a different customer from the 33-year-old mother of one deciding whether to have a second. Any policy portfolio that works has to work for both, and most don’t. Most serve the existing-parent customer through benefits that escalate with each child. The prospective-parent customer, who doesn’t yet have a child to trigger any of them, is served by cash she hasn’t yet decided to earn.

Who can actually reach any of this?

The two customers’ frictions live at different scales, and this matters for what work can be done now. The split is rough. Most frictions sit at one layer, but individual frictions cross between them. TMR’s binding constraints are mostly national, and Shaw’s shock pattern shows why: TMR moves with national-scale disruptions that hit cohorts simultaneously, and the disruptions ratchet rather than reverse. Whether a 25-year-old can form the fertility desires that drive her marriage decision depends on whether she has stable employment, whether her education will complete on a reasonable timeline, whether her labor market will penalize her 66 percent in maternal earnings, whether her employer will force her to sign an illegal contract promising to resign if she becomes pregnant. No mayor can fix any of these. CPM’s frictions are mostly local, consistent with Shaw’s finding that CPM stays flat across periods of national turbulence; it’s not insulated from the national environment, but it’s tracking something more local. Whether a mother of one has a second depends on whether local childcare works, whether her neighborhood contains other families managing two or three children in ways that shift what a livable family size feels like. A mayor can absolutely reach these.

Housing is the clearest friction that refuses the split. A young couple that wants to marry and start a family can’t if there’s nowhere in the city they can afford a family-sized unit. That’s a TMR friction, and the lever is overwhelmingly municipal: zoning, parking minimums, permitting timelines, all decided at the city and sometimes state level. The federal government has essentially no direct authority over what gets built where. Housing is also a CPM friction at the same layer (the mother of one in a two-bedroom can’t have a second if the three-bedroom doesn’t exist because zoning prevents it from being built) and it’s the substrate the empathy channel runs on (a neighborhood of family-sized units near transit puts families in walkable range of each other; a neighborhood of studios above retail doesn’t). Three jobs, one lever, one layer. The political coalition for this already exists in YIMBY organizing across the English-speaking world, which has proven that cross-ideological housing coalitions are buildable at the municipal and state level in a way that almost no other pronatalist intervention can claim. Pronatalists who aren’t already in that coalition should be.

The local cases below do most of their measurable work on CPM (higher concentrations of families with children raising completed family sizes among residents who were going to become mothers somewhere) while also producing TMR effect through two channels: concentration raising ambient infant exposure for residents who weren’t already fertility-minded, and housing-stock decisions directly enabling household formation for couples who currently can’t start one. But the bulk of the TMR problem is national and isn’t being reached by any of the local work.

Yeonggwang. What the young couple sees walking through their county is most of this argument in concrete form. Children in the playgrounds, parents pushing strollers, neighbors who took the same marriage incentive the county offers, used the same baby care package, claimed the same child-raising subsidy. Six consecutive years as Korea’s highest fertility rate, more than double the national average, in a rural county with no subway, no university, and no tech industry. Busan Jung-gu, three hundred kilometers away under the same national framework, has none of the visible design and none of the children. Same country, same laws, more than five times the children per woman in Yeonggwang. Section I returns to what Yeonggwang actually does.

A skeptical reader will say: much of this is selection. Fertility-minded families move to Yeonggwang. Haenam County, also in South Jeolla, once held Korea’s top TFR and then lost it. (Section I returns to Haenam as the anti-case.) The critique is real and applies in varying degrees to most cases below. It’s also less damaging than it looks once the chain is in hand. Families moving into Yeonggwang aren’t reshuffling a fixed stock of fertility; they’re concentrating in a place where they are visible to each other and to everyone else who lives there. Concentration does CPM work directly, by raising completed family sizes among residents who were going to become mothers somewhere, and it does some TMR work indirectly, by raising infant exposure for residents who weren’t already fertility-minded and letting the empathy channel run on them before the child-driven marriage mechanism shifts their decisions. Concentration is one of the channels through which production happens.

What the local cases are showing

Don’t people already want more children?

The Nagi pattern, wanting already present and the system finally catching up, is the shape of the problem almost everywhere. Korean women, surveyed across multiple cohorts, say their ideal is between 1.8 and 2.16 children. They produce 0.72. Italian women say their ideal is around 2; under “ideal living conditions” they say 2.3. They produce 1.18. Across 27 EU member states, 87% of women say two or more children is their personal ideal, and a third of women aged 40 to 54 report having fewer children than they wanted. The desired-actual gap is enormous and it is not closing on its own. People want children. The system isn’t delivering on the wants people already have.

This is the first-order claim, and it belongs in front of every other one. Most of the available fertility recovery in the next decade lives inside this gap. Closing the gap doesn’t require convincing anyone of anything. It requires building the environment in which the children people already say they want become plannable. Cash transfers, ecosystems, and moral exhortations all bid for the same job. They are not equally good at it, and the reasons they differ matter.

What makes this more than a one-time transfer is what happens after the gap closes. Children produced by gap closure don’t just exist. They occupy the environment: playgrounds, sidewalks, train platforms, supermarket aisles. The 22-year-old commuter who hasn’t formed a family preference yet now lives in a denser child environment than her older sister did. Her brain is keeping score in 140-millisecond increments, before she consciously deliberates anything at all. By the time she’s deciding, she’s deciding on top of accumulated unconscious priors that were shaped by the previous round’s gap closure. Her ideal number of children, when she’s surveyed, will be slightly higher than her sister’s was at the same age, because the environment that formed it was slightly denser.

This is what Sebastian Galiani and Raul Sosa formalize as the empathy channel in their April 2026 NBER paper. Each child in an environment generates a positive externality on the desired fertility of surrounding adults, mediated by the reward circuitry that fires before conscious processing. Konrad Lorenz identified the underlying baby schema in 1943; Glocker and colleagues confirmed in 2009 that infant faces with high baby-schema intensity activate the nucleus accumbens in women who have never been pregnant. Kringelbach’s group established that the response runs within roughly 140 milliseconds. Ruth Feldman’s research showed that oxytocin tracks infant contact regardless of whose child it is, in fathers and alloparents and unrelated adults. The channel is dormant when fertility is high, where additional infants barely move the signal, and steeply active when fertility is low, where each additional infant (or its absence) shifts the operating point. Galiani and Sosa calibrate the channel at between 3 and 33% of the fertility decline, with 13.4% at baseline.

The implication is that gap closure compounds. Closing the desired-actual gap for the current cohort puts more children into the environment, which lifts desired fertility in the cohort behind them, which means the next round of gap closure operates on higher desires, which puts still more children into the environment. The same loop ran in reverse for fifty years and produced the trap. Fewer children meant less exposure meant lower desired fertility meant fewer children. The local ecosystems that work are running the loop forward.

This is the difference between a cash transfer and an ecosystem, and it explains why Korea’s modest cash-transfer gains don’t compound. Cash arrives at the moment of birth in a household already inside parenthood. It does some gap-closure work for that household, but it doesn’t densify the environment in concentrated, visible ways. The 22-year-old commuter sees nothing she didn’t see before. The next cohort’s desired fertility is unmoved. Each year’s transfers do roughly the same work as the previous year’s, and the base doesn’t grow. Local ecosystems concentrate. Yeonggwang’s higher fertility produces a denser child environment, which lifts desires in residents who weren’t already family-oriented, which produces still more children, which densifies the environment further. Same gap-closure intent. Very different long-run trajectory.

Whose preferences have to change?

Two recent thought leaders clarify why the frame matters by getting it wrong. Helen Andrews, on the right, argues that the “girlboss lifestyle would not exist if it were not massively subsidized” and treats the removal of women’s economic options as the implicit pronatalist program. Something that doesn’t match the research on the topic, especially with a greater correlation of income and CPM.

Matthew Yglesias, from the center-left, identifies that the binding constraint on marriage formation is on the male side (educated women marry at higher rates than working-class women, and the men who aren’t getting college degrees are the weak link). He then concludes that “someone needs to tell them to either stay in school or else take some other conscientiousness-demonstrating path.”

At time of Ygelsasis’s writing, the unemployment rate for college graduated young men and non college grades are at the same, alongside the fact that post 2008. Men saw persistently higher unemployment rates than women. On top of that, learn to plumb is a dismal failure as blue collar job market collapse.

Tell who, with what mechanism, through what institution, on what timeline that matches the demographic clock. The prescription doesn’t survive contact with the question of who would deliver it.

I find both arguments frustrating (well, disgusting and repulsive) in similar ways. Both are about whose preferences need to change. The data says preferences don’t need to change for the first-order gain. The preferences are already there, in the form of a desired-actual gap that no amount of moralizing will close, as if moralizing haven’t been tried. Andrews wants women to revise downward what they want from their lives so they will accept what the current system delivers. Yglesias wants working-class men to revise upward what they expect of themselves so they will become deliverable to women whose preferences would then be met, regardless of current economic realities.

Neither asks what environment would let the preferences people currently hold actually be executed on, or what the same environment would do to the next cohort’s preferences as a compounding effect. Andrews is comfortable making women bear the cost of a transition she isn’t responsible for designing. Yglesias is comfortable making working-class men bear the diagnosis without offering them anything that resembles a path. Both treat the people whose lives are at stake as objects to be hectored or modified, rather than as people responding rationally to an environment that someone built and someone could rebuild.

Research like Han and Uchikoshi that Yglesias gestures at without naming deserves direct treatment, because the talk about economics being the reason why men are the binding constraint. Among Korean men aged 25 to 29, the share neither working nor seeking work nearly tripled between the mid-1990s and the early 2020s, from roughly 10% to 30%; in the same Japanese age group, the rate barely moved.

The study attributes about a third of Korea’s fertility decline in this age group to that rise. Male economic inactivity is a national-scale TMR friction that excludes the inactive from the marriage market almost entirely, and no mayor can fix it. Naming it doesn’t change who can act on it. It does change what a thinkpiece is supposed to do.

Where can we see the chain working?

Sarah Hrdy’s cooperative-breeding research adds the historical dimension: girls in traditional societies arrived at first pregnancy with hands-on experience caring for younger siblings and neighbors’ children, which exposed the neurobiological systems that generate maternal desire to the stimulus they evolved for. Modern societies dismantled that apprenticeship by age-segregating residential life and privatizing childcare. Local ecosystems that concentrate children in visible public space are partly rebuilding what age segregation took apart.

The population-level test is clean. A quasi-experimental study of Italian municipal childcare expansion found that municipalities starting with childcare coverage above 20 percent and expanding further showed measurable fertility gains over four years. Municipalities starting below 10 percent and expanding just as aggressively showed nothing. The authors attribute this to a minimum cultural threshold for using educational services; below it, expansion doesn’t register as a meaningful option for couples making family decisions. The shape is exactly what the compounding loop predicts. Visibility has to cross a threshold before additional supply registers in decisions, because what registers isn’t the supply but the normalization of using it. Same population, same selection, different births.

How do you build an ecosystem in a rural county?

Yeonggwang. What the young couple sees walking through their county is most of this argument in concrete form. Six consecutive years as Korea’s highest fertility rate, reaching 1.71 in 2024 against a national 0.75, in a rural county of 52,000 people with no subway, no university, and no tech industry. What Yeonggwang has is a dedicated Population and Jobs Policy Office (인구일자리정책실) established in 2019 as the first such department at Korea’s basic-local-government level, and more than 80 lifecycle support programs running across marriage, pregnancy, birth, childcare, and youth retention. The county frames itself as a saenghwal-miljakhyeong dolbom dosi, a “life-embedded care city,” and uses the older Korean phrase “from cradle to grave” (요람에서 무덤까지) to describe the setup: every child is enrolled in the support network from birth and the services follow them through school, marriage, and eventually their own children.

The individual programs are unremarkable as components. A ₩5 million marriage grant, rent-interest support for newlyweds, a ₩30 million transportation card for expecting mothers, neonatal support scaling from ₩5 million for a first child up to ₩35 million for a sixth, fertility-treatment subsidies, and a monthly birth allowance of ₩200,000 payable for 216 months until age 18 (₩43.2 million in total). More unusual is the father’s parental leave subsidy (아빠 육아휴직 장려금), which pays men who take leave and which the county frames as a tool for both gender equality and paternal bonding. Most unusual is the adjacent infrastructure: e-mobility manufacturing at Daema Industrial Complex, Hanbit Nuclear Power Plant employment, and Neulpum Village young-professional housing that makes Yeonggwang a place a young person can both move to and stay. Marriages jumped 40 percent between 2023 and 2024 (from 177 to 248), which will feed the TMR of the next cohort. Mayor Jang Se-il describes the sequence directly: youth support, marriage, birth, child-rearing, in that order, as the chain the county is trying to operate on. The design is the ecosystem. The specific programs are the visible pieces.

Sejong, Korea’s purpose-built administrative capital established in 2012, operates on a different principle. Sejong held Korea’s highest TFR at the metropolitan level for seven consecutive years starting 2015, peaking at 1.89 and still at 1.03 in 2024 even as the national figure fell to 0.75. Sejong didn’t layer subsidies. It assembled three things at once: stable public-sector employment as central ministries relocated from Seoul, relatively affordable new housing (a local policy choice enabled by Sejong’s being built without legacy zoning politics), and public kindergartens accounting for 93 percent of early-childhood provision against a national figure closer to a quarter. The housing did TMR work directly: it was what let young government workers form households. The childcare did CPM work: it was what let those households have second children. The employment anchored both. Concentration raised infant exposure for everyone else who lived there. The empathy channel activated. Fertility desires formed.

Lee Sang-rim, a demographer at the Korea Institute for Health and Social Affairs, puts the principle directly:

“We need to approach fertility from the perspective of the youth life course, independence, employment, housing, family formation, child-rearing, rather than through fertility-boosting subsidy programs.”

When the first three work, the fourth and fifth can happen. Sejong’s advantage is narrowing as neighboring Daejeon competes for the same newlyweds with its own ₩300,000-per-month subsidy, which illustrates how fragile municipal success is when the national framework doesn’t recognize and protect what works.

When does a train-station daycare beat a cash bonus?

Nagareyama, a Tokyo commuter town that elected Yoshiharu Izaki mayor in 2003, ran a formal SWOT analysis of the city before taking office, picked dual-employed households with children as the target customer, and built infrastructure around them. The headline intervention was train-station childcare: at Nagareyama-Otakanomori and Minami-Nagareyama stations, parents drop children between 7 and 9 a.m., staff bus them to participating daycares across the city, and the buses return them by 6 p.m., at 100 yen per day. The logistical problem it solved was real. The less obvious thing it did was put parents and children together in the most visible public space in the city twice a day, for years. Nagareyama reached a fertility rate of 1.50 in 2022 against a national 1.26, and was Japan’s fastest-growing city for six consecutive years.

Matsudo runs a similar model with different instruments. The city has won or placed in the top three of Nikkei’s dual-career-friendly cities ranking every year since 2018, and has achieved 10 consecutive years of zero waitlist by building 121 small-scale childcare facilities positioned at all 23 train stations in the city. Matsudo treats the commuter-parent as the customer: every station has childcare within walking distance of the fare gates. The insight travels within a prefecture willing to organize around it.

Akashi, in Hyogo Prefecture, demonstrates what happens when a city government chooses what it is. Mayor Fusaho Izumi raised the child-policy budget from 125 billion yen to 297 billion yen while the overall budget stayed flat, and tripled the staff on child policy, recruited lawyers and welfare specialists onto the municipal payroll, consolidated kindergartens and libraries under a new Children’s Future Department, and moved authority over preschool policy from the school board to the mayor’s office. The visible “five free things” for children rested on organizational restructuring that took years and that no one outside city hall noticed. The reorganization was the intervention; the free services were its downstream expression. Akashi hit a TFR of 1.7 in 2018 against a national 1.42 and ran nine consecutive years of population growth. Critics point out that much of the gain is in-migration from Hyogo municipalities that didn’t match the policy. Right, and it matters less than they think. In-migration concentrates families in an environment where they’re visible to Akashi’s existing residents, and concentration activates the chain.

Two smaller cases extend the pattern. Hinode-machi in western Tokyo produced the largest TFR improvement of any Japanese municipality in Nikkei’s survey, +0.57 from 1.02 to 1.59, through coupons and allowances. Furubira-cho in Hokkaido raised its TFR by 0.47 to 1.62 on a retention logic its mayor stated plainly: “Childcare support has a defensive aspect against out-migration. It’s worth the cost.”

What happens when you get the numbers without the ecosystem?

Haenam County, in the same South Jeolla province as Yeonggwang, is the story of success that wasn’t. Haenam held Korea’s top TFR for seven years running, peaking at 2.46 in 2015. In 2017 it became the only Korean region to exceed 2.00 since Korean below-replacement fertility set in. The county received a Presidential Award on Population Day in 2016. It was the global poster child for municipal fertility success, until it wasn’t. By 2022, Haenam’s TFR had fallen to 1.04. Annual births fell from 839 in 2015 to 295 in 2021. The Korean press coined a term for what had happened: 먹튀 출산, “eat-and-run birth.”

There’s something bleak about a town winning a Presidential Award for something it wasn’t actually doing. Neighboring counties started offering bigger cash subsidies. Fertility-minded families moved for the subsidies, collected them, and moved on. The design Haenam had was transactional, not ecological. There was nothing to stay for after the cash arrived. A 2023 analysis by the Korea Institute for Health and Social Affairs showed the deeper mechanism: Haenam’s “high fertility” years corresponded to a low share of women aged 15 to 49 in the population (28 to 35 percent), compared to 44 to 54 percent in genuinely growing high-fertility places like Jinju, Asan, and Dangjin. Haenam had a small denominator that cash could inflate temporarily. It did not have an ecosystem that shifted desires in the next cohort. When the cash moved elsewhere, so did the numbers.

Is there a single variable that matters most?

Kasuga, a Fukuoka commuter town of roughly 115,000, is the exemplar of the scale-of-commitment principle. Kasuga allocates approximately 41 percent of its municipal budget to child welfare spending (児童福祉費), roughly triple what comparable Japanese municipalities allocate. The town positions itself as family-friendly for Hakata-area commuters, operates six train stations connecting to Fukuoka’s center, and pushed its child-focused investment at a level that a recent analysis of all 1,741 Japanese municipalities by Kawasaki and Morimoto identified as the single most powerful municipal-level predictor of desired birth rates, with a coefficient of 0.47 that dwarfs density, transit access, income, facility counts, and commute patterns. The point isn’t that every municipality should allocate 41 percent. The point is that the share of the budget going to children is the most consequential variable any municipality controls, and most municipalities don’t know this because nobody has told them.

Can a province do what a country can’t?

Bolzano, Trentino, and Vorarlberg sustain fertility meaningfully above their national averages over decades at the provincial scale, and they do it through three different designs.

Bolzano (Alto Adige / Südtirol) is the consumer-facing brand. Italy’s national fertility has fallen to 1.13 in 2025; Bolzano holds 1.55 and is the only Italian province whose fertility rate actually rose between 2024 and 2025. The province runs an 8.7-billion-euro annual budget, with roughly 200 million euros going to family policies specifically, split across direct transfers, the 0-to-3 childcare network, and employer incentives for work-family conciliation. The visible elements are the Family+ card giving retail discounts to families with three or more children, retailer partnerships, trilingual welcome backpacks for newborns, and the Tagesmütter system of certified home-based childcare that covers rural valleys where centralized daycare would be uneconomic.

The institutional elements are more load-bearing than the visible ones. A single unified Department for Family has coordinated the work since 2013, an Agency for the Family acts as the policy interface with municipalities and employers, a Consulta per la Famiglia of 19 members shapes direction, and every municipality and district community has a family-referent in its executive council. This is the organizations-don’t-optimize-separately principle in action: the province broke down the barriers between departments that would otherwise each run their own family program badly. The housing component is equally important and usually overlooked: Bolzano builds more housing per capita than Milan or Rome (roughly 0.22 square meters per capita in 2022 against 0.14 in Milan, and 0.31 against 0.11 pre-Covid), which means a young couple can actually find a family-sized apartment. And the Audit certification for family-friendly workplaces now covers 55 certified companies, bringing the employer into the ecosystem rather than leaving the conciliation problem to households alone.

Trentino, just south, runs a more institutional setup. Provincial Law No. 1 of 2011 established an integrated family system with a dedicated agency run by Luciano Malfer since its founding. The province runs a Family Audit certification for employers, a “Comune Amico della Famiglia” designation for municipalities, a network of Distretti Famiglia (family districts), and 93.4 percent municipal coverage of early-childhood services. Regional TFR for Trentino-Alto Adige combined is 1.39 against the national 1.18, with Trentino’s own number somewhat below Bolzano’s but well above the Italian baseline. The contrast with Bolzano is instructive: Bolzano does consumer-facing branding backed by services; Trentino does institutional structure backed by a provincial statute. Different instruments, same outcome.

Vorarlberg, Austria’s westernmost Land, runs a third model with almost no centralized brand. Vorarlberg’s 2023 TFR was 1.50 against Austria’s 1.32. The province offers a Familienpass that every Vorarlberg family automatically receives, good for discounts at over 200 partner businesses and reduced public transit fares, renewing automatically until children turn 18. On top of this sit the familieplus program recognizing family-friendly municipalities, and Vorarlberg lässt kein Kind zurück (”Vorarlberg leaves no child behind”), a comprehensive child-welfare framework coordinating federal and provincial supports. Municipalities administer most services; the province provides the shared infrastructure; the brand is almost invisible compared to Bolzano’s. This matters because Vorarlberg demonstrates that the specific visibility design is substitutable. What isn’t substitutable is sustained institutional commitment to families as the unit around which other provincial decisions orient themselves. (Vorarlberg’s trajectory is fragile: the regional family association reports a 9.4 percent birth-rate decline between 2021 and 2023, reflecting Austria-wide housing and cost-of-living pressures. Even strong provincial ecosystems don’t insulate against national-scale shocks.)

Malfer puts the shared principle in his own words:

“Money isn’t what counts. In these policies the winning card is a management model that conceives of the family as a resource for society.”

The management model is the product. Everything visible downstream is just the management model becoming visible. Michela Morandini, Bolzano’s director of Social Cohesion, Family, Elderly, and Cooperatives, says the same thing from the other province:

“There is broad consensus across all local parties on maintaining these policies, which have lasted for years and are not continually changed like national ones.”

A family pulling out the Family+ card today has reasonable confidence it will mean the same thing in ten years. That confidence is the product. The card is its visible manifestation.

What does twenty years of commitment actually buy?

Nagi, a rural town of roughly 5,700 people in Okayama Prefecture, is the closer on commitment. Nagi’s TFR climbed from 1.41 in 2005 to 2.81 in 2014, then to a peak of 2.95 in 2019, which the town built and sustained for more than two decades through ordinary interventions any municipality could copy: cuts to its own council seats and staff to free up child-policy budget, volunteer aiiku-iin delivering personal gifts to newborn families, and a suite of supports sustained long enough that residents learned to plan around them. Eighty-five percent of Nagi’s child-rearing households have two or more children; 48 percent have three or more. That’s a CPM distribution: completed family sizes shifted upward by an environment that made second and third children plannable.

A young couple in Nagi believes the support will still be there in five years because it has been there for twenty. That belief is part of the product, and it can’t be manufactured through a single visible intervention. It’s the cumulative output of a town that kept doing the same thing long enough that the doing became part of what the town is. Call it constancy of purpose. Institutions that change direction every administration cannot build the trust that lets families plan around them, which is the failure mode every municipality trying the visible interventions without the underlying commitment eventually demonstrates: a mayor announces, the press covers, fertility ticks up, the next administration arrives with different priorities, the budget tightens, the program disappears. The customer base that had started to believe the city was committing learns that institutional commitments cannot be trusted.

What the municipal cases share isn’t generosity. None of them spent dramatically more than comparable places. What they share is the alchemy and the machinery held together long enough that the compounding started, and the politics to keep it running across administrations. Visibility shapes desires, which shape marriage decisions, which shape fertility. Each link is now empirically established.

Why the national conversation is stuck

What if the French exception isn’t fading evenly?

France is the case Shaw’s framework reframes most sharply. The conventional reading is that France’s century-long fertility exception is fading: French fertility is closer to European averages than a decade ago and falling. The decomposition says something more specific. Between 2013 and 2023, France’s CPM remained stable between 2.3 and 2.4 children, while TMR fell from 85.0% to 72.8%.

French mothers are still having French-sized families. What’s collapsing is entry into motherhood at all. This matters because French policy has been organized for a century around the CPM side: generous family allowances that scale with parity, school and childcare systems that make two and three children manageable for existing mothers, tax incentives (the quotient familial) that lower the marginal cost of additional children inside a family. These instruments still work for the population they reach, which is why CPM held flat through a decade when most of Europe saw it slipping. None of them do anything for the 27-year-old who hasn’t started. Whether she enters motherhood depends on whether she can form a household, whether the labor market will penalize her for doing it, whether the housing she can afford has room for a child.

The CPM-supporting ecosystem, the part that works locally on existing parents, is doing roughly what it always did. The TMR-supporting national conditions are deteriorating. The French exception isn’t fading in some general sense; it’s fading in one component while the other holds, and the failing component is the one that sits before any local ecosystem can reach.

This is the diagnostic payoff of the decomposition. A country can spend more, sustain the spending longer, build the most institutionally continuous family policy in the developed world, and still watch the aggregate number slip, because the aggregate hides which half is slipping. If you can’t see that France’s TMR is falling while its CPM is stable, you can’t see that French policy needs a different kind of intervention now than the one it has historically delivered well.

Why do cash transfers keep hitting ceilings?

Poland’s 500+ program raised fertility from 1.29 to 1.45 within two years through a near-universal cash transfer; Polish fertility has since fallen back. Germany’s 2007 parental allowance reform produced a measurable increase concentrated among highly educated women, but current German fertility is where it was before. Targeted components can move TMR for the segment they target. They don’t compound without an ecosystem around them. said it in 2020 with more directness than most pro-natal literature allows:

“Any pro-natal policy agenda will have to be about more than just a child tax credit. It’s vital that policymakers also think about how they can remove obstacles to marriage, facilitate access to decent housing, and accelerate completion of education, all vital elements in the modern economic life cycle leading up to childbearing.”

Cash at the moment of childbirth arrives after most of the consequential choices have been made.

What does x% percent of a national budget buy when the system is broken?

Korea is the anti-case, and Korea’s decline is heavily TMR. Korean marital fertility has been stable; the collapse is in whether people marry at all, which means the binding constraints are all the national conditions shaping whether young Koreans form fertility desires in the decade before marriage. The country spends roughly 6 percent of its government budget on fertility policy and has the world’s lowest national rate, which is what spending looks like when it’s flowing through a system that doesn’t work.

Korean policies, evaluated rigorously, have measurable effects in their domains. Kim’s analysis of Korea’s regional baby bonus variation finds Korean TFR would have been 4.7 percent lower without the transfers, equivalent to roughly 562,000 fewer children over the lifecycle of the 2015 female cohort, with elasticities from 0.34 to 0.58 depending on birth order. The same program dropped the third-child sex ratio from a counterfactual 124.7 boys per 100 girls to 105.3, cash reshaping a deeply embedded cultural preference. A 2011 reform expanding childcare leave subsidies for higher-earning mothers produced a 2.3 to 2.5 percentage point increase in conception rates with no significant effect on employment. These are real findings. Korean policies work in the places where they can reach; they do modest CPM work on families who were already going to have at least one child. What they don’t do is reach the TMR frictions producing the aggregate decline, because those frictions sit before any component-level policy can touch them. No marriage bonus reaches a 25-year-old whose employer will force her to sign an illegal contract promising to resign if she becomes pregnant, whose housing is unaffordable on non-regular employment, whose education ran six years past the age her mother was when she married, and whose ten-year earnings penalty for motherhood is seven times what a Swedish mother faces.

The vertical fragmentation makes this worse. Korea runs national cash transfers, metropolitan housing rules, ward-level childcare allocation, and a private hagwon market that none of those layers regulates. Each layer optimizes its own component. Nobody is responsible for whether they fit together for the family trying to use them. The principle that organizations fail when departments optimize separately applies equally across levels of government. The central, provincial, and municipal layers are running independent strategies on the same families.

The Korea Institute of Public Administration said substantively the same thing in 2023: current regional support policy “lacks long-term performance indicators and performance management systems; insufficient project identification and evaluation systems for the regional extinction response mission; and insufficient awareness and capacity at the local government level”. The country’s successful municipal outliers get no structural recognition: as National Assembly member Lee Gae-ho noted in August 2025, Yeonggwang has produced six consecutive years of top-ranking fertility and measurable net in-migration, yet “actual fiscal support is lacking” because the central government has no incentive system that rewards municipalities that succeed. The center can’t see what its own municipalities have built, which means it can’t amplify what works or replicate it elsewhere.

The scale of the TMR frictions deserves naming because the numbers are staggering. Korean workers put in 1,865 hours a year against 1,431 in Sweden. Raising a child to 18 costs $275,000, or 7.8 times GDP per capita, against 4.1 times in the US. Eighty percent of Korean children attend a hagwon, and Koreans spent $19 billion on shadow education in 2023. None of this is reachable by any subsidy a mayor can offer. This is what the environment looks like when the chain has been running in the wrong direction for a generation. The TMR shocks that drove the chain into this position arrived discretely and ratcheted down. Each ratchet depleted the next cohort’s exposure environment further.

Doesn’t Korea’s recent rebound change the picture?

Korean TFR rose slightly in 2024 and 2025: 0.72 in 2023, 0.75 in 2024, an estimated 0.80 in 2025, the first sustained increase in a decade. The rebound is structural timing. The echo-boom generation born in the early 1990s is entering peak childbearing age, and pandemic-deferred marriages are happening with the usual two-year lag before births. None of the structural conditions has changed. The aggregate moved without the environment moving, which is exactly the failure mode the decomposition was built to expose: TFR can rise on cohort timing while the underlying ratchet stays where it was, and the next cohort will discover the gain didn’t compound.

National-scale policy is mostly stuck. Multiple ministries with separate budgets, short electoral cycles, slow demographic feedback, and structural reforms that are politically contentious in ways component-level programs aren’t. France stayed the course for a century because institutional continuity held, and even France is now showing the TMR component slipping while the CPM component holds, which is what slipping looks like before the aggregate number tells you it’s slipping. Korea and Germany and Italy haven’t, and there’s no credible mechanism in their current political economies for them to start.

The local work is the infrastructure

What can’t a mayor reach?

The bulk of the fertility problem in most developed countries is TMR (fewer young adults entering the parenthood pathway at all), and the bulk of TMR’s binding constraints are national. No mayor can reach them. Housing is the major exception, and its politics are moving. Most of the others are not. Energy policy shapes the real cost of running a household and therefore the economics of forming one. Industrial policy determines whether the labor market absorbs young men at stable wages, which determines whether they become marriageable on the timeline their biology and their desires suggest. Education finance shapes both the direct cost of raising a child and the timeline on which a young adult finishes training and enters the family-formation window.

Each of these deserves careful work, close attention to political economy, and mapping of the coalitions that could move them. None of it exists yet in the form pronatalists need. Someone has to do it. This essay is not that essay. This essay is about what to do in the meantime. Shaw's ratchet pattern makes the stakes legible: the 1970s oil shock didn't just raise prices, it triggered the economic disruption that stamped an entire cohort's TMR down permanently and never let it recover. Energy and industrial policy are pronatalist policy, whether pronatalists treat them that way or not. That essay still needs writing.

The claim here is that while the national conversation remains stuck, useful subnational work should proceed now. The municipal and provincial cases do most of the achievable work on CPM (completed family sizes among women who were going to become mothers), direct TMR work on the frictions local authority can reach (housing the main one), and additional TMR work through local concentration raising ambient infant exposure for residents who weren’t already fertility-minded. None of this solves the national TMR problem. But together these build the infrastructure national TMR reform will need to land on if and when it arrives, and they’re the work doable inside the political constraints mayors and provincial administrators actually face.

What are mayors actually doing when they succeed?

A mayor building Nagareyama’s station childcare or Akashi’s Children’s Future Department isn’t building a scale model of what a prime minister should do. She’s doing two things at once: building CPM infrastructure, and working around the missing vertical coherence her national government won’t supply. The mayor can’t make the central government cohere with the city, but she can at least make the city cohere with itself. Akashi consolidated kindergartens and libraries under a new Children’s Future Department and moved preschool authority from the school board to the mayor’s office. Trentino passed Provincial Law No. 1 of 2011 to put a single agency in charge. Nagi cut its own council seats to free up child-policy budget. These are vertical integration moves inside the layer the actor controls.

The empathy channel runs in an environment. Environments are built locally, by entities small enough to act and sustained enough to hold. A country with no local ecosystems is a country where, even if national TMR reform eventually arrives, the young mothers it produces stop at one child because the local environment doesn’t make a second or third livable. A country with a hundred Nagareyamas is a country where national TMR reform, when it comes, has CPM infrastructure to land on. The local work complements national reform; it doesn’t substitute for it.

Yeonggwang and Sejong and Nagareyama and Akashi and the Alpine provinces aren’t demonstrations of what Korea or Italy or Japan should do at the center. They show what mayors and provincial administrators can do now, inside the political constraints they face. Every one of these places shifted ambient exposure, then desires, then marriage and fertility decisions, at the scale where they had authority to act. The aggregate numbers for their countries didn’t move much, because most other cities and provinces didn’t do the same thing and because the national TMR frictions continued to depress entry to parenthood across the full cohort. Building the ecosystem in your jurisdiction makes the national number movable when the center eventually turns, and captures the CPM gains available at your scale regardless of what the center does.

The Uchikoshi findings sharpen this in a specific way, and Shaw’s decomposition sharpens it further. If fertility desires shape marriage decisions rather than the other way around, and if marriage decisions drive fertility in East Asia, then the TMR leverage point is the environment shaping fertility desires in unmarried young adults. The bulk of that environment is national. But some of it is local: the ambient infant exposure a 25-year-old experiences in her actual neighborhood is a municipal-scale variable, and local ecosystems that sustain high exposure do activate the empathy channel for residents who weren’t already fertility-minded. This is the TMR work local ecosystems can do. It’s real but limited. What it isn’t limited on is CPM: the same local environments that produce some TMR activation are also the environments where existing mothers decide whether to have second and third children, and that decision is almost entirely a local question. National cash transfers, by contrast, deliver modest CPM effects and essentially no TMR effects because they don’t touch the binding national constraints that shape young adults’ fertility-desires environment. Uchikoshi’s own policy conclusion lands here, where the operational decision gets made: “policy efforts that increase fertility desires among young people contemplating family formation by reducing uncertainty around childbearing may also contribute to higher fertility via their positive effect on marriage.” Reducing uncertainty around childbearing is exactly what a local ecosystem does. It makes the environment clear enough that a young adult can see what her thirty-five-year-old self might plausibly live, and the empathy channel takes care of the rest.

One caveat. Some fraction of the fertility-desires shift is autonomous, genuine preference change that ecosystems cannot reach. A 25-year-old who has thought it through and concluded she doesn’t want a life that includes children isn’t going to be empathy-channeled back into wanting one. That fraction exists, and the ecosystem argument makes no claim on it. What the ecosystem argument claims is the fraction shaped by environment, which Shaw’s reading of the Fertility Gap puts at the center of the problem: the gap between desired and actual fertility stems less from mothers having fewer children than desired, and more from a growing share of women who hope to become parents but ultimately remain childless. That’s a TMR phenomenon, women who wanted to become mothers and didn’t, and it’s exactly the territory where ecosystems do work, because the fraction of TMR loss attributable to depleted environment rather than autonomous preference change is large, and it’s largest in places where the ratchet has been running longest.

So what should pronatalists actually do?

The pronatalist response to national policy stagnation shouldn’t be despair. It should be aggression at every layer below national, because that’s where the work that compounds is actually getting done. I am still going to write essays on why energy and industrial policy is pronatalist policy and vice versa at the end of the day, because no pronatalist policy is immune from oil shocks, as Shaw’s paper makes perfectly clear. Pronatalists need to think about this stuff, but we are still limited by the political economy.

Build the ecosystem where you have authority. Capture the CPM gains available at your scale. Start with housing, the YIMBY coalition already exists, the municipal authority is real, and the intervention does TMR, CPM, and empathy-channel work simultaneously.

The local work also builds the political capital and institutional expertise that pronatalists will need for national reform. A country with a hundred Akashis has a bench. A country with zero has a press release.

Jobs didn’t convince anyone they wanted a phone. He built the system, and the system delivered, and the delivering changed what the next wave of people could imagine wanting. The demographic version is harder and slower, but it’s the same thing: alchemy on top, machinery underneath, and the politics to keep both running long enough that the compounding kicks in. The jurisdictions that have figured out how to hold all three together are municipalities and provinces. The countries that end the next four decades with that infrastructure in place will be in a different position from the countries that don’t.

On March 26, 2026, the Austin City Council unanimously voted to legalize Accessory Commercial Units and Front Yard Businesses: small commercial spaces, up to 200 square feet, that homeowners can operate from residential lots. A bakery in a converted garage. A corner store stocking what surrounding blocks actually buy. A coffee window cut into the side of a bungalow. All now legal, by-right, in a major American city. The vote went almost unnoticed outside planning circles. It should not have.

Here is what Austin just did, in terms the real estate market will recognize. Walkable urban neighborhoods (the Georgetowns, West Villages, Pearl Districts, Beacon Hills) command price premiums of 50% to 200% over comparable non-walkable neighborhoods in the same metros. Christopher Leinberger’s research at George Washington has documented that walkable urban places occupy roughly 1% of land in major U.S. metros but capture disproportionate shares of office and retail rent premiums. Demand clearly exceeds supply. One of the most beloved things about Disney World, for what it’s worth, is its walkability.

And yet, in the entire 21st century, the U.S. has produced nearly zero new walkable neighborhoods, and even fewer through conventional development. The few exceptions don’t inspire confidence, even though urbanists and residents love them. Culdesac Tempe, the first car-free neighborhood built from scratch in modern America, required a startup, a zoning negotiation, and about five years to deliver one neighborhood. Carmel, Indiana required a 28-year mayoralty under Jim Brainard and hundreds of millions in public investment to produce one walkable downtown district inside an otherwise well-designed but still car-dependent suburb of cul-de-sacs and roundabouts. Walkability has a scaling problem.

Austin’s vote is the first serious attempt by a major American city to fix it through a different mechanism. ACUs are a refinement of what ADUs already proved with their alchemy. ADUs, at the end of the day, added housing to (mostly) desirable neighborhoods one parcel at a time, faster than any conventional project. ACUs do the same thing for the other half of the walkable-neighborhood equation: destinations.

The premium and the bottleneck

The walkable urban premium is well documented and routinely ignored by policy. The pattern is the same across markets. Office rents in walkable urban places run 75% above their drivable suburban equivalents. For-sale residential and retail command comparable premiums, with neighborhood retail sustaining rents big-box formats simply cannot match. A small fraction of the metropolitan footprint produces a disproportionate share of economic output and an even larger share of land value.

The supply side is the puzzle. American cities know how to build subdivisions, office parks, big-box retail, and strip malls. They don’t know how to build walkable neighborhoods, and the evidence is in the production stats. Two prominent new examples in the last quarter-century. One required a startup-backed developer and a desert. The other required a generational mayor and a downtown island floating in a sea of conventional suburbia.

Carmel is a real accomplishment. The few hundred acres of walkable urbanism it produced are valuable both as places and as proof that retrofit is possible. But even the modern gold standard of American suburban retrofit, run by the most committed mayor of his generation, produced walkability for a small fraction of the city’s footprint. If 28 years and a singular political career buys you one walkable district inside one suburb, the conventional pathway is in worse shape than the conventional praise of Carmel suggests.

The building across the street

Walk almost any older American neighborhood and you’ll pass it: a small commercial use operating inside a residential zone, legally, profitably, without complaint. A corner store. A neighborhood bakery. A barbershop in what used to be a front parlor. These exist because they were grandfathered when the zoning changed in the mid-20th century. Strong Towns documents these “clearly used to be a corner store” houses as architectural fossils: the chamfered corner, the slightly-too-large front window, the faded ghost-sign on the brick.

You can guess where this goes. An identical building, proposed today, three blocks over, would be illegal. Same footprint. Same use. Same neighborhood. Different answer. The grandfathered store doesn’t generate traffic chaos. It certainly doesn’t depress property values. The cities most celebrated for their charm (Portland, New Orleans, Charleston, older parts of Boston) are precisely the cities where these uses survived. The Bywater corner store profiled in Planning magazine is one of thousands of examples sitting in plain sight.

The control group and the treatment group sit across from each other on the same block. The grandfathered corner store anchors a neighborhood that, partly because of the store, commands a premium. The block with the store is worth more than the block without. We’re arguing to legalize a value-creation mechanism that already works in every major American city. You can see it in operation. You don’t have to speculate about it.

Two columns of math

The economic case for ACUs is clearest in two columns of numbers.

Column one is what an ACU bakery costs to launch, all in. A homeowner converting an existing garage or front room pays only for equipment and inventory: $10,000 to $25,000 per the standard home-bakery cost data, with some operators starting on as little as $5,000. A homeowner building a new accessory structure adds the shell cost. Austin ADU construction runs $80,000 to $150,000 for a full residential unit; a 200-square-foot commercial accessory structure without full residential plumbing should run materially less, maybe $20,000 to $50,000, though direct ACU construction data is thin. All in for new construction: roughly $30,000 to $75,000.

Column two is the alternative. A small leased brick-and-mortar bakery typically runs $26,600 to $36,700 at the low end before lease deposits, and that’s for borrowed space the operator doesn’t own. A medium commercial bakery, per Sharpsheets data from 1,600+ bakeries, requires $125,200 to $393,000 (leasing mind you). A franchised Panera runs nearly $1 million. This is all before ACUs.

You most likely aren’t going to build an ACU from scratch in this example (unless your neighbors really, really love your baked goods). So for the more realistic case, a garage-conversion ACU bakery launches somewhere between roughly one-third and near-parity with a small leased bakery, depending where you land in each range. Against a medium commercial format, though, even the high end of the garage conversion comes in at a fraction of the cost. The new-construction ACU bakery is messier: its top end ($75K) overlaps with the bottom end of a medium commercial build ($125K), so the savings are real but not dramatic at the margin. And critically, in both ACU cases, the operator owns the asset rather than renting it, which changes the long-run economics entirely. Add lease deposits and monthly rent to the column-two numbers and the gap widens over time.

But the more interesting number is the one that doesn’t appear on the bakery’s books. The bakery is not just selling coffee (well, it probably is selling coffee, alongside the baked goods). It’s producing value, and almost none of that value shows up on its P&L. The operator captures the coffee margin and builds equity in a small commercial structure. The neighborhood captures the bigger prize (if you run a good business): amenity capitalization, because people genuinely do want to grab a fresh bagel or croissant or whatever before heading out to work. Leinberger’s work and the broader literature on amenity capitalization suggest the value uplift on surrounding parcels from successful neighborhood-serving commerce is measurable and significant. Direct ACU-specific studies don’t yet exist. The inference rests on the broader walkable-premium evidence, but it’s a strong one.

That’s the point of the ACU as a policy instrument. It lets a small operator produce a large public good, financed by their own willingness to run a business and own a small structure. The city writes one ordinance and gets a value-creation engine in return.

What a storefront is for

There’s a question worth asking before we go further: in an age of Amazon next-day delivery and ghost kitchens, why does the visible storefront matter at all? (Besides the fact that Amazon didn’t really kill retail. Private equity did. RIP Joann Fabrics.) Why not just let the bakery operate from a hidden production kitchen and ship?

Because the storefront is the product, not just a delivery mechanism for goods. A residential block punctuated by a hand-painted bakery sign, a corner store’s chalkboard listing today’s tacos, a coffee window with two stools tells a passerby this is a place where things happen. It tells them it’s reasonable to walk here. The grandfathered corner store on a Boston side street is photographed by tourists. The identical-on-paper block four miles out is not. The difference shows up in property values, in tax base, in the simple fact of people choosing to live there.

Austin’s resolution rejected the existing legal preference for what it called “hidden” commerce and embraced “the active, visible neighborhood engagement that defines Austin’s character.” Visible commerce is a public good, and the public pays for it in rent, home prices, and neighborhood premiums.

This is also why ACUs can’t be replaced by their digital substitutes. Walkability isn’t a market for goods. It’s a market for the presence of small commerce within walking distance, and presence requires visibility.

ACUs were only possible because of ADU normalization and the demand for denser housing (for whatever reason, people want to live near each other). The 2010s and the HOME initiative’s three-units-per-lot reform created the cultural and procedural baseline. The 2026 ACU vote arrived on top of a decade of that work. Berkeley, Spokane, and Burien are following similar paths, and Texas H.B. 2464 and S.B. 541 already protect “no-impact” home businesses statewide.

The hardest objection

The objection that matters isn’t traffic, parking, or gentrification. If you’re reading this, chances are you’re on the urbanist or YIMBY side of things, so your real objection is probably this: a 200-square-foot bakery is a curiosity, not a policy. ACUs at this scale won’t move the needle on walkability or housing affordability at any scale that counts.

ACUs, like ADUs, are easier to set up despite being more expensive per square foot to build. And one corner store per residential block in a typical Austin neighborhood, plus a few specialty operators (the bakery, the coffee window, the bicycle repair) layered on top, would replicate the destination density that the grandfathered pattern already produces and that already commands a premium. It’ll take time. But like ADUs, which genuinely surprised everyone with how many units they ended up producing, ACUs may transmute neighborhoods in ways we don’t fully see yet.

Each successful ACU also makes the next one politically and culturally easier, the way each successful ADU did. The 200-square-foot cap is annoyingly low, but it’s a floor, not a ceiling. The trajectory of these reforms (visible already in Burien’s draft framework) is upward, toward 1,500-square-foot neighborhood corner stores.

I should also address Pomona, because it’s the most important counter-data point. Pomona legalized ACUs by-right and has not yet seen homeowners apply. The lesson is humbling. Legalization is necessary but insufficient. Cities serious about ACUs need to build the supply chain alongside the ordinance: financing pathways, pre-approved design templates, a permit concierge for the first wave of applicants, demonstration projects that show neighbors what a successful ACU actually looks like. Austin’s parallel investments in ADU lending pathways suggest the city has learned this lesson. Other cities should learn it before passing the ordinance, not after.

NIMBYs, at the end of the day, are going to NIMBY. But it’s going to be fun pointing out the grandfathered rebuttals. Traffic? Point across the street at the grandfathered store that hasn’t generated any. Property values? Point at the premium attached to the block with the bakery. Neighbor opposition? Point at the surveys showing that residents of blocks with small commercial spaces describe them, in Berkeley’s case, as the reason the neighborhood “feels like a community.” Revealed preference is what people pay premiums to live near. Strong NIMBYs, just like with ADUs, will always oppose the new whatever, but they also have a nasty habit of building out what’s good for them personally: some want their own ADU. All of them just don’t want others building one.

A note on what ACUs will and won’t do. A 200-square-foot homeowner-operated corner store is not about to disrupt 7-Eleven. The format will stock a curated selection because it has to. It’s small. That’s the point. The ACU corner store wins on walking distance and curation, not on selection. It fills a gap that convenience-store retail abandoned years ago: the half-block, daily-trip, recognize-the-owner gap. It doesn’t compete with the corridor formats. It complements them, or ignores them.

The production process for the premium

The reforms that change cities most are rarely the ones that cost the most. They’re the ones that remove a small obstacle to something people already want to do. ACUs cost a city nothing to legalize. They generate tax revenue, foot traffic, social capital, entrepreneurship, and the street-level texture no master plan can manufacture from above.

The next Georgetown won’t be built. It’ll be grown, parcel by parcel, by homeowners who were finally allowed to. Austin started growing one in March. Cities that legalize the growing will capture the premium. Cities that wait for the next Culdesac probably won’t get one.

If you’re a developer, treat ACU progress as a leading indicator for land acquisition. The amenity capitalization will accrue to surrounding parcels well before it accrues to the ACU operator. If you’re a mayor, write the ordinance now instead of waiting for a Brainard-grade political career to materialize. If you’re a franchise strategist, design the 200-to-1,500-square-foot neighborhood concept before your competitors do, the way Starbucks colonized commercial corridors a generation ago. And if you just care about the kind of neighborhood worth living in: the question isn’t whether the production process works. Austin proved it does. The question is which city builds the second one.

Tokyo, April 1, 2026. JOGMEC, the Japan Organization for Metals and Energy Security, publishes a news release on its website. New equity investment instruments for LNG projects. Revised debt guarantee fee schedules, with discounts calibrated to the volume of LNG destined for Japan. These are two distinct instruments with different risk profiles: equity capital for exploration and debt guarantees for development. The announcement follows December 2025 deliberations at METI’s Advisory Committee on Resources and Fuels and implements the 7th Strategic Energy Plan’s target of raising Japan’s self-development ratio (自主開発比率) to 50 percent by 2030 and 60 percent by 2040, up from roughly 37 percent in FY2023.

The bureaucratic language is unremarkable. The timing is certainly not if you are on the up and up with current events. While the largest energy supply disruption since the 1970s is actively unfolding, while Japan’s LNG reserves are measured in weeks, while the yen has lost more than a third of its value against the dollar since 2021, the institution responsible for Japan’s energy security is announcing revised guarantee fee schedules for upstream equity positions. This is the latest turn of a ratchet that has been clicking forward for six decades. The question we explore is whether the ratchet is turning fast enough. And whether it is turning in the right direction.

Midland, Texas. Same week. Rigs idle in stockyards. The Permian Basin frac crew count is down 20 percent from its January peak. The Q1 2026 Dallas Fed Energy Survey captures the other half of the paradox: business activity turned positive for the first time in nearly a year, driven by the crisis-induced price spike, but nearly 70 percent of large E&P firms reported they had not changed their 2026 drilling plans. Small firms were far more responsive; nearly 60 percent planned to increase drilling. But large firms control more than 80 percent of total production, and they are not moving. One executive’s survey comment distills the mood: “Oil prices are high but will fall dramatically as soon as the new government in Iran is announced.”

The world’s largest energy importer and the world’s largest oil producer are both failing to respond to the same price signal, for different reasons that turn out to be connected. Tokyo cannot secure supply at any cost. Midland cannot deploy supply at any price. Between these two failures sits a market structure that enriches a narrow group of aligned producers and imposes costs on everyone else.

Welcome to the cost that I call the discipline tax.

The Architecture of Capital Discipline

The mechanism has been well documented. The shale boom of 2010 to 2014 was, to a first approximation, a capital destruction event disguised as a growth story. Operators outspent cash flow for a decade. Saudis and OPEC responded with increasing their own production. Production kept climbing higher and higher until investors revolted. The 2014 to 2016 crash and the 2020 pandemic collapse cemented the shift: Wall Street demanded returns, not growth. We acknowledge, emphatically, that people need to make a profit at the end of the day and a correction was needed.

Our point is different.

The correction evolved into a production restraint regime that, whether by explicit coordination or structural convergence, aligns the behavior of publicly traded U.S. producers with OPEC+ output management. The FTC’s complaint against Scott Sheffield, Pioneer Natural Resources’ co-founder and former CEO, provided the documentary evidence: hundreds of text messages with OPEC officials discussing crude oil market dynamics, pricing, and output. Public and private communications aimed, in the FTC’s language, “to organize tacit (and potentially express) coordination of capital investment discipline and oil production levels in the Permian Basin and across the United States.” Sheffield’s own summary of the effort: “I was using the tactics of OPEC+ to get a bigger OPEC+ done.” The parallel Hess finding in the Chevron merger review documented similar communications and similar endorsements of supply restrictions. The FTC characterized Sheffield’s behavior as a “sustained and long-running strategy to coordinate output reductions”, not a one-off event.

The posture is now industry-wide and spans every major basin. In North Dakota’s Bakken, rigs have fallen from 217 at peak to 27, with frac crews dropping from 13 to 7, yet production holds near 1.17 million barrels per day because longer laterals substitute for activity. A HighPeak Energy SEC filing states the logic in plain English: “Although recent events in the world have caused a surge in near-term oil prices, we are committed to developing our assets at the appropriate cadence, one that reflects sustained market conditions, capital discipline and long-term value creation.” The worst supply disruption in fifty years, and the plan does not change.

Varied forms of the NOPEC bill, designed to strip OPEC’s sovereign immunity and expose cartel behavior to U.S. antitrust law, have been introduced some 16 times since 2000. In 2007, it passed the House 345 to 72 and the Senate 70 to 23, overwhelming bipartisan margins, only to die under a Bush veto threat. In 2022, it advanced through the Senate Judiciary Committee 17 to 4. Never brought to the floor. Who lobbied against it? The American Petroleum Institute and the U.S. Chamber of Commerce, representing the very companies whose production restraint aligns with the cartel’s output management. API explicitly urged a “no” vote.

The predecessor asked why the American industry doesn’t respond to price signals. We ask a different question: what does that non-response cost everyone else?

The international dimension is what the domestic debate almost always misses. When U.S. shale was growing at breakneck speed, adding 5 million barrels per day between 2010 and 2019, it was the only non-OPEC supply elasticity in the global market. The only counterweight to cartel output management. Discipline removed it. OPEC+ members now hold virtually all global spare capacity, roughly 3.5 million barrels per day, concentrated in Saudi Arabia and the UAE, the same countries currently absorbing Iranian missile strikes. The discipline regime didn’t just restructure the American oil industry. It restructured the global energy market, removing the only supply-side competition that importing nations could rely on.

To see who benefits from this arrangement, consider the IMF’s fiscal breakeven oil prices for OPEC member states, the minimum price per barrel each country needs to balance its government budget. For 2025, the IMF estimated Iran at $124.12, Algeria at $118.95, Iraq at $92.43, Saudi Arabia at roughly $91, Kuwait at $81.84, and the UAE at $49.95. The GCC median fiscal breakeven is projected at about $70 for 2025, declining to $62 by 2030 as diversification proceeds. Before the Hormuz crisis, Brent averaged $65 in 2025. That means even the discipline-elevated price was not high enough for most OPEC treasuries to balance their books. The war has, for the moment, solved their problem: Brent above $100 generates massive surplus revenues for Saudi Arabia and comfortably covers Iraq’s $92 threshold. The discipline regime and the cartel’s output management together hold prices in a range that serves OPEC fiscal needs and major-company shareholder returns simultaneously.

The tax this arrangement levies falls on everyone outside the aligned producer group: importing nations, independent operators, service companies, energy communities, and consumers. And it is paid not in a line item on a bill but in currency depreciation, trade deficits, inflation, and idled human capital.

But here is a question worth posing to every audience, including the producers who believe they benefit: do you really want structurally elevated oil prices as a permanent feature of the global economy? The importing nations obviously do not. But even for the producers, the math is less favorable than it appears. High prices accelerate the transition to alternatives. They incentivize every non-OPEC government to invest in renewables, nuclear, and efficiency. They drive Asian economies toward electrification at a pace that, over a decade, permanently destroys oil demand. China’s own state media has described the Hormuz crisis as a “historic opportunity” for its renewable energy industry, with analysts projecting solar export surges comparable to the 120 percent increase that followed the 2022 Russia-Ukraine shock. CNPC News, the outlet of China’s largest oil company, called for accelerating wind, solar, storage, and hydrogen development as the structural lesson of the crisis. The discipline regime extracts maximum rent today at the cost of accelerating the obsolescence of the asset base.

But the damage from structurally elevated prices is not only about accelerating alternatives. It is more immediate and more fundamental than that. Cheap, reliable energy supply is the foundation on which industrial economies are built. Korea’s steel mills, shipyards, semiconductor fabs, and petrochemical complexes do not run on renewables. They run on hydrocarbons, and they compete globally on margins that structurally elevated input costs erode quarter by quarter. Japan’s manufacturing export machine, already strained by a 37 percent currency depreciation, cannot absorb permanently higher energy costs and remain competitive with Chinese producers who have pipeline access to Russian gas and domestic coal-to-chemicals capacity. Every dollar added to the per-barrel cost of energy is a dollar subtracted from the capital budget of a shipyard in Geoje, a chip fab in Kumamoto, a petrochemical complex in Chiba. For economies whose global competitiveness depends on transforming imported energy into exported manufactured goods, the price of that energy is not a macroeconomic abstraction. It is the determining variable.

It is worth noting, in this context, that the shale boom of 2010 to 2019 was the only force in the global market that ever broke OPEC’s pricing power. Five million barrels per day of new U.S. production, brought online by independent operators funded by American PE and debt markets, drove prices low enough to strain every OPEC fiscal budget and force multiple rounds of production cuts. Capital discipline, enforced by Wall Street and aligned with OPEC+ output management, removed that force. The independent operators who created it are still there. Their acreage is still there. Their geological knowledge is still there. What is not there is the capital. And the capital that replaced American PE in the Haynesville, the $10 billion from Japanese trading houses, operates on a different logic: longer time horizons, lower return thresholds, willingness to fund through price cycles. The question of what that logic could do if applied beyond Tier 1 acquisitions, to the broader independent sector where the supply elasticity actually resides, is one that the current moment makes difficult to avoid.

Who Pays

The typical account of capital discipline treats it as an internal industry matter, a question of corporate governance and investor preferences. But discipline is not a closed system. It radiates outward. The restraint that enriches shareholders in Houston imposes costs on importers in Tokyo, locks independent operators out of capital markets, and liquidates the service infrastructure that any new supply response would need. Three groups. Three different kinds of damage. One structural cause.

The Import Bill

A Japanese trading company executive calculates the cost. His government’s fossil fuel import bill nearly doubled in a single year, from JPY 17 trillion in 2021 to JPY 33.7 trillion in 2022. The trade deficit hit a record JPY 20 trillion, roughly $155 billion, the largest since records began in 1979. The yen depreciated from 109.78 to the dollar in 2021 to 131.37 in 2022, to 140.5 in 2023, and past 150 in 2024, a decline of more than a third. The mechanism is straightforward: dollar-denominated energy imports feed current account deterioration, which drives depreciation, which raises the yen cost of the next shipment, which worsens the current account further. A vicious cycle with no natural brake.

The Bank of Japan spent an estimated $60 billion in currency interventions in September and October 2022, the first yen-buying interventions since 1998. The effect on the exchange rate trajectory was not lasting. GDP growth slowed to 0.9 percent in 2022 and just 0.1 percent in 2024. Government debt stands at 235 percent of GDP, the highest among advanced economies. Energy subsidies from 2022 to 2025 totaled JPY 13.4 trillion. IEEFA projects that a prolonged Hormuz closure could reduce Japan’s GDP by up to 3 percent in 2026, reversing the modest recovery underway.

The standard response, diversify suppliers and sign more LNG contracts, has reached its limits. Japan’s LNG import bill increased 98 percent in yen terms between 2021 and 2022 even as import volumes declined by 3 percent. Diversification of supply sources does not insulate against global price shocks. What would insulate is equity ownership of production capacity, the self-development ratio that JOGMEC exists to raise. The ratio stands at 37 percent. The 60 percent target for 2040 implies an enormous volume of new upstream investment. JOGMEC’s budget for oil and gas exploration and asset acquisition is expected to double in FY2026 to 108.2 billion yen, roughly $733 million. Japanese public institutions provided $93 billion in support for overseas oil and gas projects between FY2013 and FY2024, with 45 percent concentrated in upstream investments. The capital exists. The institutional infrastructure exists. The question is what it buys.

Korea’s exposure is, if anything, more acute. The country imports more than 95 percent of its crude oil from the Middle East and more than 20 percent of its LNG through routes that transit the Strait. The won lost 20 percent of its value in 2022, moving from 1,189 to the dollar in January to 1,428 by October. Energy imports consumed $145 billion in eight months. The depreciation constrains the Bank of Korea’s ability to cut rates, because cheaper money would accelerate the won’s decline, which in turn suppresses domestic demand. A macroeconomic trap in which energy import costs simultaneously cause inflation and prevent the monetary response to it. The Korea Institute for International Economic Policy (KIEP) projects that even in an early ceasefire scenario, oil prices will not return to the pre-war level of $63, with $90 as the floor and $117 if the blockade extends. The Korea Research Institute for Industrial Economics estimates that manufacturing costs would rise 11.8 percent if the blockade lasts three months or more. The OECD cut Korea’s 2026 growth forecast from 2.1 percent to 1.7 percent, the largest downgrade among major countries. Korean commentators describe the situation as a “triple trap” (삼중 함정): energy crisis, U.S.-China strategic competition, and the North Korean nuclear threat, all converging at once.

Korea has the institutional pieces to address this. KOGAS, the state gas monopoly, simultaneously functions as monopsony importer, pipeline operator, and regasification terminal owner, an organizational structure that gives it enormous latent buyer power and operational capability for managing an integrated value chain from wellhead to burner tip. KNOC, the national oil corporation, has operated upstream projects in 17 countries. Korea’s chaebols, Samsung C&T, SK Group, Hyundai, POSCO, frequently invest in overseas energy projects, funded by KEXIM, K-Sure, Korea Development Bank, and the sovereign wealth fund KIC. The downstream infrastructure is world-class. What is missing is the upstream half.

And unlike Japan, where upstream energy security policy has been pursued with institutional consistency across administrations, Korea has witnessed shifts in policy priorities with every change of government. As one comparative analysis noted, Japan’s consistent approach produced an upward trend in its oil and gas self-sufficiency ratio. Korea’s inconsistency produced a decline. KNOC’s debt-to-equity ratio climbed to 529 percent by 2016. KOGAS’s debt-to-equity exceeded 450 percent, with uncollected receivables reaching 14.5 trillion KRW by end-2024, the result of regulated domestic prices lagging global spot rates for years. Since 2013, Korean energy policy moved away from self-sufficiency targets altogether, focusing instead on reducing debt ratios at the state-owned enterprises. The Blue Whale project, Korea’s flagship offshore exploration in the East Sea with potential reserves of 3.5 to 14 billion barrels, was defunded by the government in September 2025 and BP was brought in at 49 percent to share costs that Korea could no longer bear alone.

The institutional capability exists. The financial capacity to deploy it does not. And the policy consistency to sustain a long-term upstream investment program has never been achieved.

The contrast with China is instructive, though not in the way Western commentary usually frames it. China imports more Middle Eastern oil than Japan and Korea combined: over 55 percent of its crude comes from Middle Eastern producers, with roughly 45 percent of total oil imports transiting Hormuz. Oil imports hit record highs in 2025 at 11.55 million barrels per day. Yet while Japan’s yen collapsed 37 percent and Korea’s won fell 20 percent, China’s currency and bond markets have been comparatively stable. Not because China is less exposed. Because it spent two decades building buffers that Japan and Korea did not. Overland pipelines from Russia and Central Asia that bypass maritime chokepoints entirely. Strategic and commercial reserves estimated at 1.3 to 1.6 billion barrels, covering three to four months of demand, well above the IEA’s 90-day recommendation. Domestic production at roughly 27 percent of consumption. And a preferential transit arrangement with Iran that has allowed Chinese-flagged vessels to continue passing through the strait, with the IRGC reportedly assessing fees in yuan.

But even China’s buffers have limits. A Renmin University economist quoted by Xinhua described the IEA’s record strategic reserve release as “a painkiller, not surgery” (止痛药, 不是手术刀): it can relieve symptoms but cannot solve structural supply dependence. If the closure extends beyond three months, even China’s reserves and pipeline capacity face serious strain. China’s advantage is not that it solved the vulnerability. It is that it diversified supply routes and built redundancy into the system while the strait was still open. Japan and Korea, almost entirely dependent on seaborne imports through contested chokepoints, did not build the equivalent on the supply side. The question is whether they still can.

The numbers, we think, speak for themselves. But the implicit question deserves stating: why is there no countervailing supply capacity? Why are these countries still price-takers in a market that is structurally organized to extract maximum rent from importers? And what does it mean that the country with the deepest buffers, China, is now resuming large-scale purchases of U.S. crude and LNG to resell into Asian markets as a tool of political influence, while the countries with the shallowest buffers have made no comparable move toward upstream equity ownership?

The Locked-Out Driller

A private operator in the Anadarko Basin has 200 undeveloped locations on Tier 2 acreage. Breakeven at his current cost structure: about $65 per barrel, the Dallas Fed survey average. Large firms average $61; small firms average $66. Compare that to Diamondback’s ~$37 per barrel breakeven. That $28 gap is not geology. You know it and I know the reason, it’s organization: the accumulation of longer laterals, automated drilling analytics, continuous pumping, microgrids, advanced water recycling, developed by well-funded operators with deep engineering benches. With those improvements, the operator’s breakeven could be $48. But the improvements require engineering teams, data systems, and capital the operator does not have.

PE firms are not interested: exit markets are frozen, and a Tier 2 independent does not offer the return profile that institutional LPs demand. The majors are not interested: they are buying each other’s Tier 1 inventory, not developing Tier 2. Bank lending has tightened. Thirty-nine percent of E&P firms expected capital expenditures to decline in 2026; among large producers, the figure was higher still. One Dallas Fed respondent captured it with the weariness of four decades: “I have never felt more uncertainty about our business in my entire 40-plus-year career.” Another put it more bluntly: “’Drill, baby, drill’ does not work with $50 per barrel oil.”

The operator is viable but capital-starved. The geology works. The problem is that every domestic capital source has either exited, consolidated, or tightened. The irony is that his Tier 2 acreage is viable at $48 with the right operational improvements. JAPEX paid $1.3 billion for Verdad Resources’ DJ Basin assets in December 2025, demonstrating that the rock in analogous basins works for buyers with different return expectations. But the only capital that would fund a multi-year development program through price cycles, without demanding a 20 percent IRR and a three-year exit, is the kind of patient, balance-sheet-backed capital that American PE structurally cannot provide.

And even if the capital appeared tomorrow, the operator would face a second problem. The engineering teams that could implement the operational improvements, the people who know how to run automated drilling analytics, optimize pad development, manage simul-frac completions, are being laid off by the service companies at the same time he needs them most.

The Idled Engineer

A completions engineer in Houston is experiencing the consequences. Her frac crew was laid off when the operator dropped a rig. The U.S. frac crew count is down roughly 15 percent year-over-year, with the Permian down 20 percent from its January peak. ConocoPhillips announced plans to eliminate up to 25 percent of its global workforce; Chevron cut 15 to 20 percent, roughly 8,000 people. Petroleum engineering graduates at U.S. universities collapsed from a peak of 2,615 in 2017 to roughly 623 bachelor’s degrees annually, a 76 percent decline. Some programs were nearly wiped out: Louisiana State down 89 percent, University of Oklahoma down 90 percent, Colorado School of Mines down 88 percent from their peaks. Her students are choosing software engineering, not because demand for energy is falling, but because capital discipline has throttled the demand for her expertise.

The techniques she knows, the ones that drove the best operators’ breakevens down by 8 percent in two years through ultra-long laterals, AI-driven well spacing optimization, continuous pumping, 90-plus percent produced water recycling, are transferable. The question is: transferable to whom?

If a different capital allocation logic were applied, the demand for petroleum engineers would not look like it does today. The total upstream workforce has shed 252,000 core jobs while producing substantially more energy. Those 252,000 people did not forget how to drill. But they are leaving the industry permanently, and the universities are no longer replacing them.

What’s Already Moving

It would be tempting, at this point, to conclude that the situation is simply stuck, that capital discipline has locked the system and no countervailing force exists. That would be a mistake. Something is already happening. Over the past twelve months, Japanese companies have deployed more capital into American shale gas than at any point in history. To understand what they have done, and more importantly what they have not done, we need to look carefully at the deals.

In twelve months, Japanese companies deployed more than $10 billion into American shale gas, part of a $550 billion investment framework between Washington and Tokyo.

JERA, Japan’s largest power generator, paid $1.5 billion for Haynesville assets in October 2025, closing in February 2026. Production of 500 MMcf/d, with plans to double to 1 Bcf/d through operational improvement. JERA holds 6.5 million tons per year of LNG offtake from U.S. projects due online by the end of the decade. JAPEX followed in December with $1.3 billion for Verdad Resources’ DJ Basin assets. Tokyo Gas secured positions in East Texas.

Then the capstone. In January 2026, Mitsubishi Corporation acquired Aethon Energy Management for $7.5 billion, the largest acquisition ever made by a Japanese trading house in the American upstream sector. Aethon’s Haynesville assets produce 2.1 Bcf/d of natural gas, equivalent to roughly 15 million tons per year of LNG. The sellers: Ontario Teachers’ Pension Plan and RedBird Capital Partners, PE backers exiting exactly the kind of independently managed, operationally excellent shale positions that discipline has rendered unfashionable for domestic capital. The new entity, Adamas Energy, will be a wholly owned Mitsubishi subsidiary, with Aethon founder Albert Huddleston’s son serving as CEO.

Mitsubishi’s stated rationale merits careful attention. The investment will “accelerate efforts to build an integrated value chain in the United States, from upstream gas development to power generation, data center development, chemicals production, and related businesses.” Mitsubishi’s North American energy platform already includes partnerships in Canadian shale gas, midstream operations via CIMA Energy in Houston, LNG exports through LNG Canada and Cameron LNG, and power generation via Diamond Generating. The Aethon acquisition was not a standalone bet on natural gas prices. It was the addition of an upstream position to a value chain that already stretches from wellhead to power plant.

These deals represent a genuine and, in some ways, unprecedented shift. Japanese firms historically preferred minority stakes in overseas E&P assets; the 2026 wave marked a pivot to full operational control. And the movement is not only Japanese. KOGAS signed a 10-year supply agreement with Trafigura for 3.3 million tons per year, mostly from U.S. LNG sources. POSCO signed a heads of agreement for 1 million tons per year from Alaska LNG over 20 years.Hanwha Aerospace signed a 20-year deal with Venture Global for 1.5 million tons per year. And in a development that received little English-language coverage, KOGAS and JERA, the world’s two largest LNG importers, signed a cooperation agreement for cargo swaps and coordinated supply management in direct response to the Hormuz crisis. A KOGAS official stated that they maintain “response preparedness including inter-country coordination such as planned cargo exchanges with JERA within the year.” The two biggest buyers in the global LNG market are already cooperating at the operational level, quietly, without fanfare.

And then there is Alaska. The Alaska LNG project, a $44 billion development led by Glenfarne, has secured preliminary commercial commitments from LNG buyers in Japan, Korea, Taiwan, and Thailand for 11 million tons per year, including agreements with JERA, Tokyo Gas, CPC, PTT, and POSCO International. POSCO is providing steel for the 807-mile pipeline and making an equity investment. The project’s stated competitive advantage: “short shipping distance to Asia, featuring canal-free routes avoiding contested waters.” That last phrase is the Hormuz hedge stated in six words. LNG from Alaska’s North Slope, piped to the Pacific Coast and shipped to Tokyo or Busan, never transits the Strait of Hormuz, the Strait of Malacca, or any other contested chokepoint. On the oil side, ConocoPhillips’ $9 billion Willow project is halfway to completion with first oil expected in 2029, producing conventional crude with decades of production life, and a record $163 million federal lease sale in the National Petroleum Reserve in March 2026 drew eleven interested companies. But we should note the caveats: Alaska state lawmakers have stressed that Alaska LNG has no binding gas sales agreements and no agreements for financing. As one senator put it, “You know you have a project when you have take-or-pay contracts and you have access to capital.” The preliminary commitments are real. The final investment decision is not.

That said, we should be precise about what these deals, across all the basins, are and what they are not.

They are acquisitions of already-operational assets from competent producers on proven acreage. They bought Aethon, “quietly the largest private natural gas producer in the United States”, not a struggling wildcatter on Tier 2 ground. Aethon was built by Albert Huddleston and his team over decades: veterans who understood the Haynesville intimately, who had assembled acreage positions through the bankruptcies and restructurings of the 2015 to 2020 period, and who had achieved operational results that justified the $7.5 billion price tag.

This is not an accident. It reflects a deep structural preference in the way Japanese trading houses approach upstream risk. They do not take geological risk. They buy operational assets after someone else has proven the geology, established the production history, and demonstrated the decline curves. This preference is rational, culturally embedded, and unlikely to change because the Strait of Hormuz closed. Mitsubishi bought Aethon because Aethon was already working, not because Mitsubishi suddenly became a wildcatter. And it is, in part, why the self-development ratio has remained at 37 percent after six decades of JOGMEC effort. The institutional machinery is designed to acquire de-risked assets, not to take the exploration risk that creates them.

We think this preference, however rational in its origins, requires re-examination in light of what it is actually buying and what it is leaving on the table. Because here is the part of the story that the preference obscures: the “risky wildcatters” that Japanese institutions have historically avoided are the people who built the assets they are now paying billions to acquire. Aethon was assembled from the wreckage of Chesapeake Energy’s bankruptcy. Verdad Resources, which JAPEX bought for $1.3 billion, was built by a PE-backed team that took the geological risk in the DJ Basin. Every de-risked asset in the current acquisition pipeline exists because an independent operator, funded by American PE or debt markets, drilled the wells, proved the reserves, and established the production base that makes the asset investable by risk-averse institutions. The wildcatter is not the opposite of the de-risked asset. The wildcatter is the person who created it.

And the wildcatter model did not fail because the geology was bad. It failed because it was too successful. The shale boom added 5 million barrels per day to U.S. production between 2010 and 2019. That surge crashed prices, triggered the investor revolt of 2014 to 2016, and produced the capital discipline regime that now prevents the independent sector from operating. The wildcatters drilled so much oil that they destroyed their own business model. The discipline regime is the scar tissue from that success. What it left behind is not a landscape of unproven geological risk. It is a landscape of proven basins, known formations, established production histories, and demonstrated well results, operated by hundreds of independent companies that lost their capital source, not their geological knowledge.

This distinction matters enormously for the risk calculus. The operators currently sitting capital-starved across the Permian, the Bakken, the Anadarko, the DJ Basin, and the Haynesville are not wildcats in any meaningful sense. They have drilled wells. They have production data. They have decline curves. They have breakeven calculations published in the Dallas Fed survey every quarter. Many of them have acreage positions that were assembled and high-graded over decades. What they lack is not geological proof. It is capital. And the capital they need operates on a time horizon and return threshold that American PE can no longer provide.

The risk of waiting for the next Aethon to appear, fully formed and ready for a $7.5 billion acquisition, is that inaction does not preserve optionality. It destroys it. Every month the discipline regime persists, the operators who proved up the basins shed staff, defer maintenance, let leases expire, and sell equipment at liquidation prices. The engineers who could optimize the assets leave the industry. The service companies that would support development programs downsize or close. The longer a risk-averse institution waits for the geological risk to be fully retired before deploying capital, the more the operational infrastructure required to realize the value of that geology degrades. What looks like a de-risked opportunity today will look like a greenfield rebuilding project in three years if nobody funds the operators in the interim.

The question is not whether Japanese or Korean institutions should become wildcatters. They should not. The question is whether they can recognize that the landscape they are surveying is already far more de-risked than their institutional memory suggests, that the people who de-risked it are available and identifiable, and that the window in which those people and their knowledge remain in the industry is closing.

The Efficiency Gap Nobody’s Closing

This brings us to the question that, in our view, ought to be central to the conversation about American energy production but almost never is. Not “should we drill more?”, a political question with no analytical content. Not “will OPEC cut?”, a geopolitical question beyond any single country’s control. But: why is the gap between the best operators and the rest so large, and who has the capability to close it?

The drilling productivity gap between the best operators and the rest is not small. Diamondback breaks even at $37 per barrel. The Dallas Fed survey average is $65. Large firms average $61; small firms, $66. That gap, roughly 43 percent, is not primarily geological. It is organizational. Diamondback’s CEO put it plainly: “Never underestimate the American engineer.” The techniques driving the gap are documented: ultra-long laterals exceeding 15,000 feet, AI-driven drilling analytics, continuous pumping, microgrids for power cost reduction, 90-plus percent produced water recycling. Coterra is installing microgrids in West Texas to reduce power costs. These are organizational capabilities, not proprietary physics. Data systems, workflow management, crew training, equipment standardization. They are, in principle, transferable.

But to whom? And by what mechanism?

The wildcatter we described earlier faces a specific operational problem. His breakeven is $65 because he independently contracts drilling, completion, sand supply, water management, and midstream transport without scale economies or standardized data systems. The engineering knowledge to solve each of these problems exists. It is the same knowledge that drove Diamondback’s breakeven down by 8 percent in two years. But it is embedded in organizational systems that the independent cannot access or afford to build from scratch.

In certain industrial traditions, this exact problem, how to transfer the operational capabilities of a lead firm to a fragmented supplier base, has been solved systematically. The lead firm embeds engineers in supplier operations. It teaches its production system from the inside. It shares cost savings. It builds relationships that last decades, not quarters. The result is not charity; it is a supply chain whose quality and efficiency reflect the standards of the lead firm rather than the baseline of the market. The same trading houses now buying U.S. shale assets, Mitsubishi and Mitsui among them, operate exactly such supplier networks in their manufacturing and logistics businesses.

The fragmentation runs deeper than any single operator’s cost structure. Three hundred or more Permian independents each independently negotiate OFS contracts, drilling schedules, sand procurement, water management, pipeline capacity, and marketing arrangements. The collective purchasing power represented by that fragmented base is substantial but unrealized. Each operator pays retail for services that, if aggregated, could command wholesale terms. Mitsubishi’s stated strategy, upstream development to power generation to data centers to chemicals, is precisely the kind of integrated value-chain organization that has historically solved this kind of coordination failure. Some industrial economies have built entire competitive advantages by wrapping value chains into single corporate ecosystems, eliminating the coordination costs that fragmented competitors cannot overcome on their own.

And beyond integration, there is another model worth noting: network-based industrial coordination. Dense clusters of specialized small firms sharing intelligence, pooling procurement, and collectively negotiating with larger counterparties. Not by merging, but by cooperating. This model drove rapid industrialization in parts of East Asia, transforming thousands of small manufacturers into globally competitive production ecosystems without requiring any single firm to achieve scale on its own. The 300-plus Permian independents represent exactly this kind of fragmented base: individually subscale, collectively formidable, lacking only the coordination mechanism that other industrial traditions developed decades ago.

We are not suggesting that any of these models should be imported wholesale into the Permian Basin. Industrial traditions are not modular components. But the gap we have documented, between the $37 breakeven of the best operators and the $65 breakeven of the rest, is an organizational gap. The engineer being laid off in Houston has the techniques to close it. The wildcatter in the Anadarko has the acreage where it could be closed. The trading houses in Tokyo have the organizational tradition of closing exactly such gaps across fragmented supplier bases. And the capital to fund the effort operates on a time horizon, decades rather than quarters, that the problem requires.

We are not assembling a blueprint. That is not our job. Our job is to document that the pieces exist, and that as of this writing they are scattered across three continents, in the hands of three groups that have not yet recognized each other as relevant to their own problems.

The Math

The predecessor article documented what $100 oil used to buy in Houston: 457,500 jobs, marriages, mortgages, children. It documented what $100 oil buys now: buybacks, dividends, and a fertility decline that shows up in the demographic data a decade later. The discipline tax is the gap between those two outcomes, paid not only by Houston’s roughnecks and their families, but by every importing nation whose trade deficit, currency depreciation, and inflation are the arithmetic consequence of a supply side that refuses to respond.

The numbers documented above require only one comparison. Japan’s fossil fuel import bill increased by JPY 16.7 trillion in a single year. The BOJ burned $60 billion in currency interventions that achieved nothing lasting. Against that: JERA’s Haynesville acquisition cost $1.5 billion. Mitsubishi’s Aethon deal cost $7.5 billion. Total Japanese shale investment over twelve months: $10 billion. JOGMEC’s cumulative upstream investment over six decades of operation amounts to less than what the Ministry of Finance spent in a single weekend trying to defend a currency collapsing under the weight of energy import costs it had no supply-side mechanism to reduce.

The discipline tax is paid every year. The alternative is a one-time capital deployment that, even in a worst case, costs less than a single quarter’s currency intervention.

Close

The Strait of Hormuz remains closed. Asia’s reserves are measured in weeks. Brent is above $100. Iran is now demanding sovereignty over the strait and proposing tolls on foreign shipping, potentially $800 million per month.

In Midland, Texas, rigs sit in stockyards. Nearly 70 percent of large producers have not changed their drilling plans. Equipment auctions run at steep discounts. A tool store owner, “this is my sixth boom-bust”, waits for customers who are not coming. But this is not a boom-bust. An engineer with twenty years of frac experience is updating her LinkedIn profile, and the petroleum engineering program that trained her has lost 88 percent of its enrollment since the peak. Across the basin, operators with viable acreage and manageable breakevens sit capital-starved, not because the rock is bad, but because the capital allocation regime sends money to buybacks instead of wells. A PE fund manager in Houston could produce a list of investable independents, with acreage maps, well results, and breakeven calculations, in a week. Anybody can find them. Nobody with the right capital structure has looked.

In Tokyo, a bureaucratic news release on a government website announces revised guarantee fee schedules for LNG investment. JOGMEC’s budget is doubling. The self-development ratio, after six decades of institutional effort, sits at 37 percent. The target is 60 percent by 2040, fourteen years away. Closing a 23-percentage-point gap in fourteen years would require a pace of upstream equity acquisition several times faster than anything Japan has sustained in the preceding six decades. The $10 billion deployed into U.S. shale over the past twelve months is the most aggressive burst in JOGMEC’s history. At that rate, sustained annually, the math might work. Whether the institutional machinery built for a slower tempo can operate at that speed is a question the news release does not address.

In Seoul, the situation is starker. The institutional debts and policy reversals documented earlier have produced no corrective response. Korea’s chaebols have the organizational capability to wrap entire value chains into integrated ecosystems. Korea’s state institutions have the downstream infrastructure and the procurement relationships. What they have not had is the political consistency to sustain an upstream investment program long enough for it to compound. Japan’s recent $10 billion burst shows what is possible when the institutional commitment is there. Korea has not yet had its $10 billion moment. The OECD’s growth downgrade, the largest among major countries, suggests that the cost of continued delay is no longer abstract.

The rigs, the engineers, the acreage, the geological knowledge, the organizational expertise, the patient capital, the institutional infrastructure: all of it exists. All of it is available. The operators are identifiable. The acreage is mapped. The breakevens are published in the Dallas Fed survey every quarter. The service companies holding the equipment and the skilled crews are listed in trade journals anyone can read. The information required to evaluate the opportunity is not hidden. It is sitting in public filings, investor presentations, and industry databases, waiting for someone with the right time horizon to act on it.

To understand what is being lost, it helps to understand what it costs. A drilling rig contracts for roughly $15 million to $20 million per year. A frac crew runs about the same. The equipment being auctioned at 30 percent discounts in Midland stockyards today was drilling wells last year. It does not need to be designed, manufactured, or shipped from overseas. It needs a contract. The engineers being laid off were, until recently, implementing the operational improvements that drove the best operators’ breakevens below $40. The formations they worked on have already been drilled, tested, and proved. These are not speculative costs for unproven capabilities. They are the current liquidation prices of a functioning industrial ecosystem. For context: retaining a frac crew for six months costs less than what the Bank of Japan spent in a single hour of currency intervention in October 2022. The entire annual cost of a drilling rig is a rounding error on a single weekend’s sovereign debt defense. The mismatch between what is being spent to treat the symptoms of energy dependence and what it would cost to maintain the operational capacity that could address the cause is, at this point, almost farcical.

But not forever. Every month that passes, another frac crew disperses. Another engineer leaves the industry. Another equipment auction clears at pennies on the dollar. Another university program shrinks. A rig that sits idle for six months needs refurbishment before it can drill. A frac crew that breaks apart takes months to reassemble, and the experienced hands do not come back from the construction job or the data center that hired them. The geological knowledge is published. The operational knowledge is embodied in people, and those people are walking away. The window in which the idle capacity described in this article remains available is not indefinite. It is closing, visibly, one layoff and one liquidation at a time.

Most of it is still idle. Not all of it will be there when someone finally comes looking.

Everyone agrees that a lot of new American buildings are ugly. Scan any neighborhood forum, any city council hearing, any urbanist Twitter thread, and you will find near-universal consensus on this point: the blocky five-over-ones, the black-and-white “gentrification buildings,” the unornamented apartment blocks that have colonized commercial corridors from Seattle to Charlotte. Nobody likes them. The “ugly is fine” YIMBYs concede the premise when they tweet “I don’t care if it’s ugly, people need homes.” The NIMBYs concede it when they invoke design review boards as the last line of defense against aesthetic ruin. Both sides agree on the diagnosis. Both sides are wrong about the cause.

The standard explanation for the ugliness of contemporary American development points to developer greed, architectural laziness, or the inevitable tradeoffs of building at scale. The more important answer is regulatory: zoning codes, parking mandates, setback requirements, use separations, and the discretionary design review process have collectively created a narrow corridor of permissible building forms, and most developers are rationally optimizing within it.

If the ugliness is a product of regulation, then the remedy runs in the opposite direction from what NIMBYs propose: abolish the zoning codes and design review processes that constrain form, invest in the urban systems that produce collective beauty, and let the emergent intelligence of thousands of small actors do what committees cannot.

I should be clear about my commitments. I want to abolishing zoning and discretionary design review, to the point that it makes some YIMBYs uncomfortable on that front. I also want to stress that we urbanists and YIMBYs, part of a tradition that stretches from Jane Jacobs through New Urbanism through Strong Towns, a tradition that has always cared deeply about what cities look like. The “I don’t care if it’s ugly” posture adopted by a vocal faction within the movement is a deviation from that tradition, and it is both empirically unsupported and strategically counterproductive.

I should be precise about what I am criticizing. YIMBYism is an internet-native movement, and that is one of its great strengths. As a commenter said before, the internet is what allowed people who experience the diffuse benefits of new housing, future residents, priced-out renters, to find each other and organize in a way that previous generations of social scientists assumed was impossible. The Minneapolis YIMBYs made memes. Online discourse built the coalition.

What I am objecting to is not online advocacy but a specific annoying rhetorical concession: the claim that aesthetics do not matter, that ugliness is an acceptable price for units. That concession is loud but not representative of the broader movement. , a professor at Soka University of America who writes the newsletter , offered us a better frame for the common enemy in our conversation with him: not “NIMBYs” as people, but “freezing our towns in amber and endless sprawl.” The aesthetics question is central to both halves of that formulation.

College Towns

College Towns, where urbanism meets higher education.

By Ryan M Allen

If you care about building more housing, you should care about what it looks like. Not because beauty is more important than shelter (it isn’t) but because the evidence suggests you don’t have to choose.

The tools designed to produce beauty are producing ugliness

Since 1994, Seattle has subjected most new apartment construction to “design review,” in which building proposals must win approval from a volunteer citizen board. Dan Bertolet of the Sightline Institute undertook an exhaustive analysis of how this process actually functions. What he found should trouble anyone who believes that committee oversight produces better buildings.

Seattle’s design review board postponed approval of a 400-unit, transit-oriented development with 168 subsidized homes, objecting to the color, the presence of a ground-floor daycare, and the shape of the building. The East Design Review Board delayed 110 affordable homes and a daycare above Capitol Hill Station because they considered the design “really good” but not “great.” An ECONorthwest analysis found that projects in 2018 took 84 percent longer to move through the permitting process than comparable projects in 2010.

These are not isolated anecdotes. Architects in Seattle report that developers routinely choose worse designs to avoid triggering the design review threshold. As one architect told Sightline: developers’ first question is whether they can avoid design review, and firms regularly see “good clients building indifferent projects because the ‘better’ idea that we are pitching triggers a requirement for design review.” The process does not merely delay good buildings. It actively incentivizes bad ones.

In New York, the Landmarks Preservation Commission rejected a proposal to replace a crumbling two-story brick building with a superior design featuring a glass-and-metal storefront and a delicate brick screen. The original building, the commission determined, was “reflective of the period of significance.” Housing that survives this pageant arrives in a diminished state: delayed, downsized, and bearing rents that reflect months or years of regulatory compliance rather than the cost of beauty.

Design review was built in good faith. The intuition behind it (that aesthetic quality matters to communities, that citizen input should shape the built environment) is not unreasonable. We take that intuition seriously. The problem is that the process does not deliver on its own premises. Jack Nasar’s research on urban design aesthetics identifies the formal variables that drive positive aesthetic responses: order, moderate complexity, and elements of familiar styles. But design review boards optimize for consensus, not for any coherent aesthetic principle. The result is architecture designed to offend no one, which is to say architecture that delights no one. As of the early 1990s, 83 percent of surveyed American cities had adopted some form of design review. We could not locate a single peer-reviewed study demonstrating that these boards produce measurably better aesthetic outcomes than unreviewed construction.

, an urban planner and the writer of , showed us what the alternative to committee review looks like. In our interview, he described Giralda Avenue in Coral Gables, Florida, a formerly struggling two-lane street with parallel parking that was pedestrianized through a design-driven process. “Each component of the project was considered a piece of art, from the planters to the pavers to the color of the tiles,” Brown told us. The street is now the most vibrant row of restaurants in the city. After construction, a rezoning allowed residential uses and balconies, further activating the space. The aesthetic success was not incidental to the project’s political success. It was the engine of it.

Car Free America

City Planning and Transportation Policy.

By Mark R. Brown, AICP, CNU

But Giralda Avenue was not produced by a volunteer design review board evaluating whether a building was sufficiently “great.” It was produced by a Business Improvement District and City Commission with a specific design vision and the professional capacity to execute it. Design vision works. Design-by-committee does not.

The zoning trap

The design review problem sits atop a deeper structural issue: the underlying zoning codes that constrain architectural form in ways that make individual buildings worse and urban systems ugly.

Parking minimums alone distort form dramatically: ground-floor parking podiums, setback requirements for surface lots, and the sheer volume of building dedicated to car storage rather than human habitation. Use separations prevent the mixed-use ground floors that give streets their vitality. Height limits in residential zones force density into narrow commercial corridors, creating the abrupt transitions between single-family homes and mid-rise apartment blocks that generate the most visceral opposition.

Adrian Pietrzak and Tali Mendelberg of Princeton have shown through survey experiments that people react negatively to buildings that are architecturally “out of place” relative to the height or style of surrounding buildings. They call these “contextual development preferences,” and they are real and consequential. But zoning itself creates the mismatches. By restricting height in some areas and concentrating density in narrow corridors, zoning forces the very contextual mismatches that generate opposition. The system produces the disease it claims to cure. Exclusionary zoning created the built form, the built form created aesthetic expectations, and those expectations now drive opposition to anything that departs from them.

Allen’s work on college towns confirms this at the ground level. He observed that the best urbanism in many American cities survived around universities, not because of design review, but because campuses accidentally preserved the pre-war built environment. “A lot of the best urbanism that we can see in some of the small towns comes from university towns,” he told us, “because they sort of act as a natural city.” Students don’t have cars. The campus maintained walkable density. Pre-war structures that were demolished elsewhere survived near universities because there were always people walking there, because you couldn’t easily sprawl students out.

The aesthetic quality that people love in college towns like Ann Arbor, Chapel Hill, and Madison was not produced by design review. It was produced by walkable density, mixed use, and the absence of parking-driven sprawl, and it survived despite the zoning regime that destroyed the same quality everywhere else.

Allen crystallized the absurdity with a single observation about Disneyland. Walt Disney, he told us, was “inspired by his hometown and his wife’s hometown, sort of these ideas of an idyllic version of a city or a town that was no longer existing.” So Disney built a replica. “What’s funny is it costs $120, $130, whatever it costs now... to go in and walk down this town. And that feeling used to be in every town across the country.” We regulated away the beauty. Now we pay to visit a facsimile.

Ugly buildings, beautiful city

Tokyo is filled with plain, unremarkable, even ugly individual buildings. The concrete is often bare. The facades are often flat. The signage is often cacophonous. And yet Tokyo is, by wide consensus among urbanists, architects, and ordinary visitors, one of the most beautiful and livable urban environments on earth. This paradox contains the most important lesson in the entire aesthetics debate: beauty in cities is an emergent property of urban systems, not a feature stamped onto individual buildings by review committees.

Jorge Almazán, a Tokyo-based architect and professor at Keio University, spent years studying what makes Tokyo work. His book Emergent Tokyo: Designing the Spontaneous City, written with Joe McReynolds and his Studiolab research team, identifies five distinctive urban typologies that produce Tokyo’s collective beauty from individually unremarkable parts. Yokochō alleyways: narrow lanes of micro-bars and restaurants, chaotic and cramped, that are among Tokyo’s most beloved spaces. Zakkyo buildings: mixed-use vertical structures stacking small businesses on upper floors in combinations that would be illegal under virtually every American zoning code. Dense low-rise neighborhoods of small wood-frame houses on tiny lots, where residents create micro-gardens on their doorsteps that beautify the streetscape emergently. The beauty arises not from any of these typologies in isolation but from the system that connects them.

In an interview on the Compounding Podcast, Almazán described how Tokyo’s street network creates what he calls “pocket neighborhoods.” Efficient arterial roads form the exterior boundary; inside is a dense maze of small streets that cars avoid because they are inefficient to navigate. “Without forcing people,” Almazán explained, “just because of the structure of the city itself, what you have is that basically people or cars crossing those areas tend to go slowly.” The result is naturally walkable, car-light neighborhoods, without any explicit pedestrianization or design mandate.

And the walkability generates beauty. When you walk slowly through a neighborhood, you notice: the micro-gardens, the shop fronts, the varied building forms, the textures and asymmetries that would be invisible from a car at 45 miles per hour. Almazán described residents contributing “a little bit to the community by putting small pots, even cultivating small trees... When you walk through many of these neighborhoods... you see the personality of everyone.” No committee mandated it. It emerged from thousands of individual decisions within a permissive regulatory environment.

Almazán himself cautions that the emergent urban spaces he documents cannot simply be dropped into other cities. Some of what makes Tokyo’s plain-building beauty work is culturally specific: the micro-garden behavior reflects norms of collective public stewardship that are not easily legislated, and the jijikai (neighborhood self-governance associations) operate through social consensus mechanisms that are not how Americans resolve land-use conflicts.

But the culturalist explanation can be taken too far. What is replicable is structural: Japan’s national zoning framework with 12 cumulative use categories, which constrains nuisance without constraining form; the small-lot structure that allows individual landowners to build at varied scales; the mixed-use permissions that make zakkyo buildings legal; the transit investment that creates walkable catchment areas; and the absence of mandated aesthetic review. No design review board evaluates whether Tokyo’s buildings are beautiful. The beauty emerges from the system.

The American comparison, and the Houston objection

The contrast that should trouble us is not between Tokyo and Kyoto, or Tokyo and Singapore. It is between Tokyo and American cities that have individually more beautiful buildings but collectively uglier urban environments, precisely because they police individual facades through design review while neglecting the urban systems that produce collective beauty.

San Francisco has some of the most individually beautiful residential architecture in the United States: the Painted Ladies, the Victorians, the Edwardian rowhouses. It also has one of the most aggressive design review and historic preservation regimes in the country. The result: a housing crisis so severe that median rent exceeds $3,000, tent encampments within sight of million-dollar homes, a city that has frozen its beauty in amber while the human cost mounts. San Francisco has beautiful buildings. It does not have a beautiful city, because beauty at the system level requires dynamism, affordability, and the presence of actual people, not just facades.

Santa Fe, New Mexico, mandates a Pueblo Revival aesthetic: every building in the historic district must conform to a specific adobe style. The result is not beauty but uniformity, a theme-park homogeneity that visitors find charming and residents find stifling. Meanwhile, outside the mandated zone, the city sprawls in the standard American pattern. The distinction between mandated local style (Santa Fe) and organic local style is crucial. The former produces uniformity. The latter produces genuine beauty.

Georgetown, in Washington, D.C., enforces historic preservation so strict that individual window replacements require commission approval. Meanwhile, the District’s housing affordability crisis deepens. The aesthetic preferences of existing homeowners have been codified into law, at the direct expense of people who need homes.

In every case, the pattern is the same: beautiful buildings in a small preserved zone, surrounded by sprawl, car dependency, and housing scarcity. Tokyo inverts this entirely. It prioritizes systemic qualities (walkability, mixed use, density, architectural freedom) and produces a city that is collectively beautiful even though most individual buildings are plain.

At this point a skeptical reader will raise the obvious question: What about Houston? Houston has no zoning ordinance, the closest thing to what we are proposing that exists in a major American city. And Houston’s aesthetic reputation is not Tokyo’s.

Houston eliminated the wrong half of the equation. The city removed use restrictions but did not invest in the system-level qualities that produce collective beauty. Houston has no comprehensive transit network creating walkable catchment areas. Its streets are engineered for automobile throughput, not pedestrian experience. And Houston retains extensive private deed restrictions that function, in many neighborhoods, as de facto zoning.

But Houston’s failure is not simply a matter of neglect. It is structural. The same fragmented governance that prevented Houston from adopting restrictive zoning also prevents it from building coherent systems: coordinated transit, competent water management, walkable infrastructure. And the Texas Department of Transportation operates as a parallel authority that widens highways regardless of what local residents want, actively degrading the walkability that our argument identifies as the aesthetic precondition. Houston didn’t just choose not to build the system. Houston’s governance makes system-level investment structurally difficult, and the state highway authority works against it.

This does not mean Houston’s freedom produced nothing beautiful. It did. The Beer Can House, the Orange Show, the folk-art vernacular that could never survive a design review board: these are exactly the kind of individual creative acts that architectural freedom enables. But they remain isolated curiosities you drive to, not part of a collective streetscape you walk through. Freedom produced the beauty. The missing system is what would have connected it.

The Houston case does not disprove the thesis. Eliminating restrictions on what you build is necessary but insufficient. You must also invest in the system around what you build, and you need governance capable of doing so: the walkable streets, the transit, the human-scale infrastructure that enables the emergent beauty we find in Tokyo, in college towns, in the pre-war American main streets that Disney had to replicate. Houston removed the zoning without building the system.

The beauty of working with what exists

The Tokyo case demonstrates this pattern at the city scale. But the evidence extends further, into adaptive reuse, local design revival, and the uncurated spaces that fragmented ownership produces.

Adaptive reuse, the repurposing of existing buildings for new functions, is one of the most reliable pathways to aesthetically successful development. It is also 15 to 30 percent less expensive than new construction and can be completed up to 30 percent faster. The World Economic Forum reported in 2025 that reusing built assets emits 50 to 75 percent less carbon. And multiple projections estimate that 90 percent of real estate growth within the next decade will involve adaptive reuse rather than new construction.

The aesthetic case is equally strong, though less often articulated. Reused buildings retain proportions, materials, structural details, and spatial relationships that took decades to develop, what we might call accumulated aesthetic capital. Older buildings predate modern climate control and were designed with thick walls, awnings, high ceilings, and large windows for functional reasons: temperature control and natural light. These features are also precisely the qualities people find beautiful.

Calvin Chua, an architect whose Singapore-based firm Spatial Anatomy works across challenging contexts in Asia, showed us how adaptive reuse operates in practice. In our interview, Chua drew a sharp distinction between high-capital “landmark” reuse (converting a power plant into a museum) and the community-focused reuse that he finds more compelling. An example that was brought up was Karl Bengs, a German architect who has spent years renovating abandoned kominka (traditional Japanese houses) in dying rural villages. Bengs is not building tourist attractions. He what he wants to do is building actual homes, based on local designs. He applies modern techniques to preserve traditional forms, embracing local design vocabulary honestly, not through committee mandate (the Santa Fe model) but through an architect responding to place, materials, and community need. The aesthetic quality is central to the project’s success: people move to these renovated villages partly because the buildings are beautiful, and partly because they are real.

Chua’s own research on Singapore’s strata malls revealed a different dimension of uncurated beauty. Strata-titled malls let individuals own shops outright, not rent them. Any building change requires 80 percent owner approval. “Because the ownership structure is so fragmented,” Chua explained, “it’s very hard to curate what shops get to exist in there. But at the same time, because of the lack of curation, this subculture or unique shops start to emerge in these strata malls.”

The uncurated interiors are among Singapore’s most culturally vibrant spaces. Some owners refuse million-dollar buyout offers because the malls function as what Chua describes as “retirement villages,” places with their own ecosystems, where the sense of purpose and community is embedded in the fragmented, unmanaged structure of the space. The spirit emerged because the space was uncurated. No property management company dictated what went where.

A comparison that you can draw between Singapore’s strata malls and American strip malls, both of which have fragmented ownership and eclectic tenants. But American strip malls are car-oriented and surrounded by parking lots, while strata malls are embedded in walkable, transit-rich neighborhoods. Same ownership structure. Radically different aesthetic outcomes. The variable, once again, is the urban system.

Why urban systems produces beauty

We have now surveyed the evidence across multiple contexts: Seattle’s design review failures, Tokyo’s emergent beauty, Japanese rural revival through adaptive reuse, Singapore’s uncurated strata malls, American college towns, and American cities that froze individual beauty in amber while their collective systems decayed. The pattern is consistent. The mechanism operates through five channels.

First, diversity of form. When thousands of individual actors build on small lots with minimal regulatory constraint, the result is a varied streetscape with different heights, materials, setbacks, and uses. Nasar’s research identifies moderate complexity as a key driver of positive aesthetic evaluation. Zoning and design review suppress complexity by mandating uniformity.

Second, walkability as aesthetic enabler. The aesthetic experience of a city requires pedestrian-speed movement. Walkability is not merely a transportation policy. It is an aesthetic precondition. Almazán’s pocket neighborhoods, Allen’s college towns, and Brown’s Giralda Avenue are all places where people walk, and therefore where people see.

Third, temporal dynamism. Tokyo rebuilds its average building every 30 years. The built environment constantly adapts to current needs and tastes; unsuccessful buildings get replaced, successful ones get imitated. A regulatory regime organized around preventing change is an aesthetic straitjacket.

Fourth, bottom-up curation. The most beloved urban spaces (Tokyo’s yokochō, Singapore’s strata malls, the pre-war American main streets) were not designed by committees. They were curated from below: individual shop owners, residents, and small builders making independent decisions that accumulated into coherent streetscapes. This is the process that Cozzolino (2021) describes as “the joint yet individual action of numerous actors on various scales” producing “rich, coherent and harmonious built environments.”

Fifth, existing structures as aesthetic capital. Adaptive reuse preserves proportions, materials, and spatial relationships that took decades to develop. Each generation builds on what the last created, rather than starting from zero under new regulatory constraints.

We can already hear the objection from the “ugly is fine” faction: that we are romanticizing Japanese farmhouses, backyard cottages, and micro-gardens over the thing that actually matters, which is units. The strongest test of this framework comes from the most unlikely source: Soviet commie blocks. Nobody romanticizes them. Gray prefabricated concrete panels, assembled like industrial Lego, repeated across thousands of sites from Budapest to Vladivostok. And yet, as urbanist Adam Something has documented, the residential areas built around these buildings routinely outperform brand-new private housing developments in livability. The reason is the system. Soviet-era planning placed kindergartens, schools, doctors’ offices, parks, supermarkets, and public transit within walking distance of every apartment. Street plans excluded through traffic. The buildings were ugly. The neighborhoods were livable. New luxury developments built nearby, with individually superior architecture but no investment in surrounding services, cannot match them.

If the system-level argument holds even for commie blocks, the ugliest mass-produced buildings on earth, it holds everywhere. Get the system right and even ugly buildings produce good neighborhoods. Get the system wrong and even beautiful buildings produce broken cities.

We are not arguing against units. We are arguing that the system around the units determines whether those units produce neighborhoods people actually want to live in. The commie blocks delivered massive quantities of housing, fast and cheap. But the achievement endured because the surrounding system was good. Units matter. The system around them matters more.

To be clear: we are not endorsing the political system that produced commie blocks. But the urban planning insight, that neighborhoods need services, walkability, green space, and transit within reach of every front door, was correct. The Soviets achieved it through authoritarian mandate. Our proposal achieves it through architectural freedom and capital incentives.

The political science: why this is urgent

Everything we have argued so far could be dismissed as an aesthetic preference. A wave of recent political science research makes that dismissal untenable. Aesthetic judgments causally determine whether people support housing development.

David Broockman, Christopher Elmendorf, and Joshua Kalla, political scientists at UC Berkeley, UC Davis, and Yale, published a working paper in 2025 demonstrating that aesthetic concerns about housing are, in their words, “widespread, not pretextual, and causally affect support for development.” Their experimental evidence shows that manipulating the aesthetic quality of proposed buildings changes support levels: ugliness reduces support even when it poses no threat to quality of life. These are not NIMBY rationalizations. They are sincere aesthetic judgments that drive real political behavior.

Their most striking finding cuts against the standard NIMBYism narrative. Homeowners in already-dense areas supportdense development in their neighborhoods, indeed more so than homeowners in single-family neighborhoods. People who live in collectively beautiful dense environments support more density. That is the Tokyo effect translated into American political science. The market confirms what the research suggests: walkable, aesthetically coherent neighborhoods command the highest property values in their cities. Poundbury’s home prices consistently run above the surrounding Dorset market. The pre-war college-town neighborhoods this essay celebrates are among the most expensive real estate in America. People vote with their wallets for pretty, walkable places. That is not sentimentality. It is revealed preference at scale.

Pietrzak and Mendelberg’s study, published in Urban Affairs Review in 2025, adds a crucial finding to the contextual preferences we discussed earlier: fit matters more than height itself. A taller building that matches its surroundings is often more acceptable than a shorter one that clashes. And buildings that don’t fit prompt intentions to engage in costly political behavior, like attending public meetings, pressuring officials, organizing neighbors. Support rarely mobilizes symmetrically. This asymmetry explains why public hearings skew toward opponents of new housing.

The strongest real-world test of whether beauty can overcome opposition is Poundbury, King Charles’s New Urbanist development on the outskirts of Dorchester in Dorset. Poundbury gets the system right in almost every way our argument demands. It is mixed-use, with over 260 businesses and 2,750 jobs integrated alongside housing. Thirty-five percent of homes are affordable, built to the same standard as market-rate units and indistinguishable from them on the street. There is no use-separation zoning; homes, shops, offices, and factories share the same streets. The architecture draws on local Dorset vernacular, designed by Léon Krier to create walkable density at human scale. Even The Guardian, which once ridiculed the place as “fake, heartless, authoritarian and grimly cute,” later conceded that “a growing and diverse community suggests it’s getting a lot of things right.” The Sunday Times named it one of the best places to live in Britain. Home prices have steadily risen, demand consistently outstrips supply, and the model has been replicated at Nansledan in Cornwall and Tornagrain in Scotland. On the merits, Poundbury is a success.

And yet. When the Duchy of Cornwall attempted to replicate the Poundbury model at South East Faversham in Kent, proposing 2,500 homes designed by the same architect (Ben Pentreath) to the same standards, the project was met with fierce local opposition. The Faversham Society called it “an existential threat to the very nature of the town.” National newspapers ran headlines about a local revolt. What is significant is what the objections were not about. They were not about aesthetics. As Building Design reported, design was barely mentioned in the objections; the opposition focused on traffic, loss of farmland, and strain on services. Even the original Poundbury was not immune: the Ramblers called it a “ghetto,” and existing Dorchester residents opposed its expansion. The British Social Attitudes Survey found that about seven percent of people are more likely to support new housing because it is better designed. In a fight decided by who shows up to the meeting, seven percent is not a rounding error. But it is not a majority, and it cannot do the work alone.

The Faversham case refutes the hypothesis, adopted as UK government policy under Housing Secretary Michael Gove, that “communities will welcome development when it is beautiful.” Beauty alone will not accomplish this. But notice what beauty did accomplish at Faversham: it took design off the table as an objection. Nobody at the public meetings complained that the buildings were ugly. The opposition had to find other grounds, and the grounds they found (traffic, services, infrastructure) are precisely the system-level failures our argument identifies as the real source of opposition. Beauty cleared one obstacle. It cannot clear them all, and treating it as a substitute for system investment is the error. Poundbury itself succeeded because it paired beautiful buildings with system-level investments: mixed use, walkable streets, integrated employment and affordable housing. But Poundbury also required thirty years of patient capital from a royal estate with no shareholders, a model its own estates director acknowledges volume housebuilders cannot deliver. And its beauty is enforced from above, not emergent from below: residents need Duchy permission to change their front door color. This is mandated beauty, closer to Santa Fe than to Tokyo. Strip the system and keep the facades and you get Faversham’s opposition. Build the system and let the beauty emerge and you get Tokyo.

Allen’s college-town observations confirm the dynamic at the local level. In Lubbock, Texas, neighbors fought a student housing development proposed across from the Texas Tech campus. The university is going through with it, Allen told us, but the opposition wasn’t really about the students. It was about the high-speed stroad between the neighborhood and the campus, which “looks like a highway drag strip.” The ugliness is in the infrastructure, not the building. Fix the system and the aesthetics follow.

The feedback loop that emerges from this research is the key to the entire argument. Zoning constrains building form, producing ugly buildings. Ugly buildings generate aesthetic opposition. Aesthetic opposition justifies more zoning. Around and around. But breaking the cycle requires understanding who is in it, because the research describes two distinct populations, not one.

The hardline NIMBYs, Allen’s “freezing in amber” crowd, will oppose any development regardless of what it looks like. Faversham proves this. These opponents are not making aesthetic arguments in good faith. They are using aesthetics as a socially acceptable wrapper for opposition that is really about preventing any change at all. You will never convince them. The BIMBY hypothesis (Beauty In My Back Yard, the idea that beautiful buildings will melt opposition) wastes political capital trying. And it hands the hardliners a new veto: “It’s not beautiful enough,” which is exactly the logic that let Seattle’s design review board reject affordable housing for being merely “good.”

Then there are the persuadable: people who are not reflexively opposed to development but who care sincerely about what their neighborhood will look like afterward. This is the population that Broockman’s research captures. Their aesthetic concerns are real, not pretextual. They respond to contextual fit (Pietrzak and Mendelberg). They are more supportive of density when they already live in attractive dense environments. For this group, aesthetics is a valuable tool in a larger set: not the argument that wins the fight on its own, but the one that tips the balance, clears design objections off the table, and builds places people will fight to protect once they exist.

The YIMBY who tweets “I don’t care if it’s ugly” is making a specific strategic error: handing the hardliners a weapon to use on the persuadable. The hardliner can now say, “See? They admit it’ll be ugly. They don’t care about your neighborhood.” The “ugly is fine” faction thinks they are being tough-minded and pragmatic. What they are actually doing is collapsing the persuadable into the hardliner camp by conceding the aesthetic premise. It is not a principled stand. It is an unforced error, and the hardliners are grateful for it.

What’s the idea?

The case for abolishing zoning and design review is well established in the YIMBY literature. We will not rehearse the standard arguments. What I want to focus on is what comes after abolition: the specific development forms and financing structures that can produce the emergent beauty we have been describing.

Accessory dwelling units and accessory commercial units

The closest thing to the emergent process in America already has a name. Two names: Accessory Dwelling Units (ADUs) and Accessory Commercial Units (ACUs). I know it’s not the same, but in practical terms, it’s what most communities can use.

ADUs (the backyard cottages, garage conversions, and above-garage apartments that cities from Portland to Los Angeles have begun to legalize) add architectural diversity at the lot level. A neighborhood of identical ranch houses with a scattering of backyard cottages and converted garages has more visual complexity, more of the moderate complexity that Nasar’s research identifies as a driver of positive aesthetic response, than the same neighborhood without them. The American Planning Association notes that research does not support fears about ADUs degrading neighborhood character; there are indications they do the opposite.

ACUs are the less familiar but arguably more transformative cousin. As the APA’s December 2025 Zoning Practicearticle defines them, ACUs are “small-scale, often homeowner- or tenant-operated businesses integrated into primarily residential lots”: corner coffee kiosks, backyard salons, garage bicycle repair shops. ACUs are the American re-legalization of what Tokyo’s zakkyo buildings already do and what American neighborhoods did naturally before use-separation zoning made it illegal. As Garlynn Woodsong wrote for the Congress for New Urbanism, “We used to build mixed-use neighborhoods in our cities that freely mixed the retail destinations people need with the residences where they live, in very close proximity.” Zoning ended this. ACUs would restart it at the smallest, least disruptive scale.

Both ADUs and ACUs produce collective beauty through the same channels we identified: diversity of form, bottom-up curation, and mixed use. They also avoid the “contextual misfit” problem that Pietrzak and Mendelberg show triggers political opposition. A backyard cottage fits because it is small-scale and owner-driven. A garage-front salon fits because it faces the sidewalk at human scale. These are not five-over-ones dropped into single-family zones. They are incremental, bottom-up, fine-grained additions that produce the closest thing to the Tokyo effect at the neighborhood level.

The cost paradox and the case for cheaper capital

Here is where the policy problem becomes acute. ADUs and ACUs are the cheapest total-cost form of new development. A detached ADU typically costs $150,000 to $350,000, a fraction of conventional multifamily construction. An ACU built from a converted garage or a prefab shed can cost as little as $10,000 to $30,000.

But they are also, paradoxically, the most expensive form of development per square foot. The reason is structural: every ADU, regardless of size, requires a kitchen, a bathroom, utility connections, a foundation, permits, and design work. These are fixed costs that do not shrink with the unit. When you spread them across 500 square feet instead of 2,000, the per-square-foot figure climbs steeply.

This cost paradox means that the development form with the greatest potential to produce collective beauty is also the form most punished by the current financing environment. Conventional mortgage products and construction loans are designed for conventional buildings on conventional lots. ADUs and ACUs fall into what one practitioner calls a regulatory liminal space: not quite a home, not quite a commercial building, not quite a renovation, and therefore not quite eligible for the financing products designed for any of those categories.

This is where the capital-incentive proposal becomes essential. The existing federal incentive structure already differentiates by policy goal: the Low-Income Housing Tax Credit makes capital cheaper for affordable housing (only in a permissive zoning environment); the Historic Tax Credit makes it cheaper for preservation; Opportunity Zones steer investment to low-income communities (whenever investment is effective is a different story). We propose two layers of capital reform.

The first layer: make capital cheaper for all small/mid-scale development. ADUs and ACUs are made possible because homeowners’ own access to capital. Adaptive reuse and incremental infill should have access to financing products designed for their scale. This means grants, low-interest construction loans, and tax credits specifically targeting the fixed-cost problem: subsidizing the permitting, design, and utility-connection costs that make small units disproportionately expensive per square foot. California’s $40,000 ADU grant and New York’s Plus One ADU Program are steps in this direction, but they remain piecemeal and underfunded relative to the scale of the opportunity.

The second layer: make capital cheapest for development that meets objective, system-level aesthetic criteria based on the fact that they is a premium created by said criteria. Not criteria evaluated by a committee, but criteria checkable by staff on a site plan: mixed-use ground floors with active frontage, pedestrian-oriented building placement (entrance on street, no setback parking), use of local or regionally sourced materials, adaptive reuse of existing structures, contextual height transitions, and green or activated public frontage. These criteria incentivize the system-level qualities that the research identifies as producing collective beauty, without requiring anyone to evaluate whether a facade is “good” or “great.”

We anticipate the objection that even “objective” aesthetic criteria add a compliance layer. The distinction is categorical, not cosmetic.

Design review is a veto gate. You cannot build until the committee approves. The standards are subjective. The process adds months or years. And because the gate is binary (approved or rejected) the rational developer minimizes risk by designing below the review threshold, which means designing worse.

A capital incentive is an optional bonus. You can build regardless of whether you meet the aesthetic criteria. Meeting them unlocks cheaper financing. Not meeting them imposes no penalty beyond forgoing the bonus. The homeowner who builds a standard ADU is no worse off than today. The homeowner who orients the ADU toward the sidewalk with an active frontage and uses local materials gets a lower interest rate or a larger grant.

Veto gates reduce housing supply. Optional bonuses do not. Veto gates concentrate power in committees. Optional bonuses distribute it to homeowners, small builders, and neighborhood entrepreneurs. The mechanisms are opposites. But the criteria must genuinely be objective and checkable by staff. “Mixed-use ground floor with active frontage” can be verified on a site plan. “Contextually appropriate massing” cannot. The moment the criteria require subjective judgment, the bonus becomes a gate. The line is real and it must be held.

Allen’s observation that the YIMBY conversation is too centered on New York and San Francisco has direct implications here. Smaller cities and college towns have lower development costs, more existing building stock suitable for adaptive reuse and ADU conversion, and neighborhoods where ACUs could reintroduce the walkable commerce that zoning eliminated.

What the evidence demands

Tokyo is not beautiful because its individual buildings are beautiful. It is beautiful because its urban system (walkable, mixed-use, human-scale, architecturally free) allows thousands of small decisions to compose into something collectively splendid. San Francisco has beautiful buildings and a broken city. Houston has freed its buildings but is limited by it’s fragmentation and being a plaything of TxDOT. Tokyo has plain buildings and a beautiful, functional, livable whole. The variable is the system. It has always been the system.

Every YIMBY who tweets “I don’t care if it’s ugly” is giving the hardliners ammunition to spend on the persuadable. Every NIMBY who invokes beauty to block all development is using aesthetics as a veto, not a value. Both positions are wrong, but they are wrong in different ways, and the difference matters. The hardliners cannot be converted and should not be catered to. The persuadable can be won, and aesthetics is one of the tools that wins them, but only as part of a larger set. The error is treating beauty as a substitute for system investment, not treating beauty as a tool worth using. The way to win is to build the systems that make beauty emerge: walkable streets, mixed use, architectural freedom, and cheaper capital for the small-scale development that produces it. The argument that collective beauty is an emergent property of free urban systems, not a product of committee oversight, is not a compromise between these positions. It is the position that makes both of them obsolete.

College Towns

College Towns, where urbanism meets higher education.

By Ryan M Allen

Car Free America

City Planning and Transportation Policy.

By Mark R. Brown, AICP, CNU

As Easter is coming up at time of writing, I decided to be a little festive with this article. Buckfast Abbey, founded in 1018 during the reign of King Cnut, dissolved by Henry VIII in 1539, and refounded in 1882 by exiled French Benedictine monks, is today best known to the British public for two things, neither of which the monks are entirely comfortable discussing in the same breath.

The first is Buckfast Tonic Wine, a caffeinated fortified wine (15% ABV, 281 milligrams of caffeine per 750ml bottle, roughly eight cans of cola) that the monks began producing in the 1890s as a medicinal tonic. I can personally guarantee that if you want to put a little pepper in a party, this wine/tonic does the job. Originally marketed with the charmingly antique slogan “Three small glasses a day, for good health and lively blood,” Buckfast has become, improbably and somewhat horrifyingly, one of the most crime-associated beverages in Scottish history. A BBC Scotland investigation found it mentioned in 5,638 crime reports in the Strathclyde police region between 2006 and 2009. The Scottish Prison Service reported in 2015 that Buckfast was a significant factor in over 40% of its inmates’ arrests. The drink has earned itself an extraordinary litany of folk nicknames: “Wreck the Hoose Juice,” “Commotion Lotion,” “Cumbernauld Rocket Fuel,” and, perhaps most evocative, “a bottle of What the Hell Are You Looking At.”

The second thing Buckfast Abbey is known for, or rather should be known for, is bees.

For seventy-eight years, the Abbey housed one of the most consequential applied-science programmes in the history of entomology: a breeding operation that produced a disease-resistant honeybee now raised in more than twenty-six countries. The man who ran it, a German-born Benedictine monk named Brother Adam, logged over 100,000 miles searching for breeding stock, kept pedigree records spanning half a century, and retired at ninety-three. He was awarded the OBE. He received honorary doctorates from Uppsala and Exeter. He is, by any reasonable measure, one of the most important figures in modern agriculture.

The public discourse around Buckfast Abbey has overwhelmingly focused on the wine. The bee work receives a fraction of the attention. That imbalance is this article in miniature. What happened to the bees at Buckfast is now happening to bee conservation writ large. The moral panics get the headlines. The practical work gets the results. And if you care about keeping bees alive, you have to understand why.

The Scope of the Crisis

In January 2025, commercial beekeepers across the United States began reporting something catastrophic. As they prepared their colonies for the annual migration to California’s almond orchards, the single largest pollination event on earth, they discovered that their bees were simply gone. Not dwindling. Not struggling. Gone. Colony after colony, dead or abandoned, at rates that dwarfed anything in the modern record. By the time the surveys closed, the numbers told a grim story: an estimated 1.6 million colonies lost, commercial operations absorbing average losses of 62%, and total economic damage exceeding $634 million.

The 2024–2025 U.S. Beekeeping Survey, conducted by Auburn University and the Apiary Inspectors of America, recorded estimated annual colony losses of 55.6%, the highest since annual tracking began in 2010–2011, and 14.2 percentage points above the fourteen-year running average of 41.4%. Winter losses alone hit 40.2%, exceeding all historical averages. State-level annual losses ranged from 34.3% to 90.5%. The triage survey administered by Project Apis m., which covers beekeepers managing roughly 68% of the nation’s colonies, found that commercial operations reported average losses of 62% between June 2024 and February 2025. Honeybees pollinate more than 70% of major global crops. Their pollination services are valued at roughly $15–20 billion annually in the United States alone.

One detail from the surveys deserves particular emphasis. For the second consecutive year, and contrary to the first fifteen years of survey data, commercial beekeepers experienced more severe losses than hobbyists. Commercial operations (those managing more than 500 colonies) lost an estimated 62%, compared to 51% for hobbyists. Commercial beekeepers have more resources, more experience, and more sophisticated management practices than most hobbyists. They are the backbone of pollination-dependent agriculture. When the professionals are losing more bees than the amateurs, something odd, something structural, has shifted.

And around these beekeepers, as they counted their dead and calculated their ruin, the internet did what it always does. Environmentalists blamed neonicotinoid pesticides. Agrochemical defenders blamed Varroa mites. Left-leaning commentators accused Big Ag of ecocide. Right-leaning commentators accused green activists of Luddism. On social media, the bees became, once again, a mascot for everyone’s priors.

How Bees Became a Culture-War Football

Left’s Story: Ban the Neonics and Punish Big Ag

We should be generous with this camp, because they identified something real. Neonicotinoid pesticides, a class of systemic insecticides introduced in the early 1990s typically applied as seed coatings, are toxic to bees. That is not in serious dispute. Laboratory studies have repeatedly demonstrated sublethal effects on navigation, immune function, and reproduction. Environmental organisations (Friends of the Earth, the Center for Food Safety, the Sierra Club, the Pesticide Action Network) identified a genuine stressor and pushed it into public consciousness at a time when the agrochemical industry was doing its best to minimise it. The EU’s 2018 ban on three neonicotinoids (imidacloprid, thiamethoxam, clothianidin) reflected genuine, if contested, scientific concern. These organisations deserve credit for sounding an alarm.

The trouble is what happened next. The narrative calcified into a monocausal story: pesticides are the villain, and a ban is the solution. The political campaign became an end in itself, disconnected from the messy, multicausal reality of bee decline.

The evidence for that disconnect is now substantial. More than a dozen large-scale field studies across North America and Europe have reached broadly similar conclusions: no conclusive observable adverse effects on bees at the colony level from field-realistic exposure to neonicotinoid-treated crops. The EU ban, now years old, has not demonstrably reversed European bee declines. Meanwhile, the campaign absorbed enormous political energy, media attention, and activist resources that might have been directed toward threats where the scientific consensus is considerably stronger. Varroa mites, the single most devastating parasite of managed honeybees worldwide, received a fraction of the public attention devoted to neonicotinoids, precisely because mites do not come with an easily identifiable corporate villain attached.

The politicisation dynamic is worth examining directly. Leaked documents reported by The Times of London described a 2010 meeting at which four senior European scientists discussed a coordinated plan to produce papers with predetermined conclusions and shepherd them through peer review with pre-selected reviewers, explicitly in order to obtain a ban. Whether one regards this as scientific manipulation or legitimate advocacy depends somewhat on one’s priors. But it illustrates a broader pattern: when environmental questions become political campaigns, the incentive structure shifts from “what is true” to “what is useful.”

Right’s Story: It’s All Mites, and Regulation Is the Real Threat

The opposing camp gets the entomology more right. Varroa destructor is, by broad scientific consensus, the single most significant threat to managed honeybee colonies worldwide. The mite feeds on developing and adult bees, transmits deadly viruses (deformed wing virus A and B, acute bee paralysis, chronic bee paralysis), weakens immune systems, and shortens worker lifespans. More importantly, Varroa has developed resistance to amitraz, the most commonly used miticide, a finding confirmed in the 2024–2025 USDA field sampling of failing colonies. If one had to point to a single proximate cause of mass die-offs, Varroa would be the strongest candidate.

But acknowledging Varroa’s primacy should not become a permission structure for dismissing all pesticide concerns or opposing all regulation. Here is where the right-wing version of the story becomes, in some ways, more dangerous than the left’s. Agrochemical industry actors have demonstrably invested in reframing the debate around mites precisely to deflect attention from legitimate pesticide questions. Internal corporate communications reported by The Interceptdescribe Bayer’s “Bee Care” programme as a crisis-PR strategy, with staff celebrating its success in shifting media coverage away from neonicotinoids. CropLife America compiled lists of search-engine terms to manipulate, and consulting firms worked to decouple Google results for “bee decline” from “neonicotinoids.” The mite narrative is scientifically stronger than the pesticide narrative. But it has also been instrumentalised by corporations with obvious financial interests.

And there is a worse problem. The left’s monocausal story is wrong about the mechanism but right that the situation is a crisis requiring public investment. The right’s version (”it’s just mites, stop regulating, let the market sort it out”) provides ideological cover for defunding the very research, extension services, and breeding programmes that are actually solving the problem. Every practical intervention we describe later in this article depends on sustained public funding: USDA research budgets, university grants, extension-service staffing, nonprofit operating costs. The conservative position, followed to its conclusion, eliminates the funding that makes the work possible. Misguided activism that directs resources to the wrong target is bad. A political movement that argues against directing resources at all is worse.

The right answer is not “it’s only mites” any more than it is “it’s only neonics.” Multiple stressors interact in complex ways: mites, pathogens, nutritional deficiency from habitat loss, climate disruption, pesticide exposure, and amitraz resistance. But acknowledging that complexity does not require another round of blame-assignment. It requires work, and it requires someone to pay for the work.

Neither side, through all its fundraising emails, congressional hearings, protest marches, op-eds, and viral social-media content, has bred a single Varroa-resistant bee, developed a nutritionally complete pollen substitute, trained a beekeeper in mite-monitoring technique, or kept one colony alive through a hard winter.

We should be honest, though, that the line between “activist” and “practitioner” is not always clean. Some organisations do both. Project Apis m., a nonprofit funded largely by the beekeeping and almond industries, is the clearest example: it funds applied research, coordinates the triage surveys that produced the loss data cited above, connects commercial breeders with USDA scientists, and advocates for federal research funding, all at once. The Inflation Reduction Act’s conservation funding, the USDA’s breeding programmes, the extension services that train beekeepers in mite monitoring: none of these exist without political advocacy. The question is not whether activism matters. It is whether the activism is pointed at something that will actually keep bees alive, or whether it is pointed at something that will generate donations and media coverage. Those are different targets, and most of the loudest voices in the bee debate have chosen the second.

Which brings us back to the monk.

Brother Adam and the Dartmoor Bees

In the early twentieth century, a mysterious epidemic known as the Isle of Wight disease, later identified as acarine disease caused by the tracheal mite Acarapis woodi, swept through the British Isles, devastating native honeybee populations. At Buckfast Abbey, 29 of 45 colonies were destroyed. The entire native British black bee (Apis mellifera mellifera) was effectively exterminated in the region. The only colonies that survived were headed by Italian queens (A. m. ligustica) crossed with native drones.

Karl Kehrle, the sickly German boy sent to the Abbey at age eleven, joined the Benedictine order, took the name Brother Adam, and in 1915 began assisting Brother Columban in the apiary. By 1919, aged just twenty-one, he was placed in charge of the entire bee operation. He would remain in that role for seventy-three years, retiring in 1992 at the age of ninety-three. He died in 1996, in his ninety-ninth year.

His approach was selective breeding carried out with exceptional rigour and patience, the same logic that livestock farmers had employed for centuries, applied with unprecedented discipline to insects. He established an isolated mating station on Dartmoor, a treeless, windswept moor where no feral bees could interfere with controlled matings. He inbred drone lines and headed drone-producing colonies with sister queens, ensuring the drones were near-genetically identical. Every queen, worker, and drone at Buckfast Abbey had a known descent on both the maternal and paternal side, with records spanning more than fifty years.

He operated on a three-year breeding cycle. Year one: produce and introduce at least thirty queens of each genetic combination, distributed across multiple apiaries for fair comparison. Year two: evaluate mature colonies under real-world field conditions using a standardised scoring system for swarming tendency, aggression, comb stability, and honey yield. Year three: graft larvae from the best-performing queens for the next generation.

Over his lifetime, Brother Adam logged more than 100,000 miles searching for breeding stock: Turkey, the Sahara, the Near East, East Africa, the Kilimanjaro region. He crossed Italian, Carniolan, Anatolian, Greek, and African subspecies into his programme. His selection criteria were precise and practical: low swarming tendency, lack of aggression, comb stability, and what he called “a boundless capacity for foraging work.”

Most importantly, he prioritised disease resistance, not through chemical treatment but through selecting for inherited traits. As he wrote in 1950: “We do not believe greatly in the various treatments generally recommended for bee diseases... by means of careful selective breeding throughout a period of twenty years we have overcome the inherent susceptibility to this disease to such an extent that it practically never occurs.”

That sentence, written seventy-five years ago, remains radical today.

The result was the Buckfast bee: productive, gentle, disease-resistant, and adaptable to diverse climates. In 1920, his colonies averaged 87 kilograms of surplus honey per colony; individual colonies exceeded 152 kilograms. A 1986 BBC documentary reported a single Buckfast colony producing over 181 kilograms, roughly 400 pounds. Today, Buckfast bees are bred across more than twenty-six countries. Norwegian and Finnish Buckfast strains have shown Varroa resistance, achieved through the same patient selection methods Brother Adam pioneered, extended to address a threat that emerged after his time.

He did not campaign for the banning of any substance. He did not write op-eds blaming the government. He identified the problem, formulated a principle (that resistance should be bred, not merely chemically suppressed), developed a rigorous method, executed it for seventy-eight years, and produced a tangible output that continues to benefit beekeepers worldwide.

The question is whether anyone is doing the same thing today.

The New Practical Frontier: Better Feed, Better Breeding, Better Practice

Brother Adam’s methods were brilliant but constrained by the tools of his era: phenotypic observation, physical isolation for mating control, paper record-keeping. What is encouraging about the present moment is not that we have abandoned his approach, but that we can accelerate and extend it. The most promising developments in bee health are practical improvements to the three things that have always mattered: what bees eat, how we breed them, and how we manage them.

Better Feed

Bees rely on pollen for essential lipids called sterols, six specific compounds (24-methylenecholesterol, campesterol, isofucosterol, β-sitosterol, cholesterol, and desmosterol) that are critical for growth and development. Climate change and intensive agriculture have reduced the diversity of available flowers, leaving colonies nutritionally deficient. Existing commercial pollen substitutes (protein flour, sugars, oils) provide calories but lack these sterols. The analogy one researcher offered: it is comparable to the difference for humans between eating balanced meals and eating meals missing essential fatty acids.

A team led by the University of Oxford, in collaboration with Royal Botanic Gardens Kew, the University of Greenwich, and the Technical University of Denmark, set out to close this gap. First, they figured out which sterols bees actually need by painstakingly dissecting individual nurse bees and analysing pupal tissues, identifying the six key compounds that dominate bee biology. Then they developed a way to produce those sterols at scale using precision fermentation: growing yeast (Yarrowia lipolytica) in bioreactors and drying the output into a powder that can be mixed into standard bee feed.

The results, published in Nature in March 2025, were dramatic. In controlled glasshouse trials over three months, colonies receiving the sterol-enriched diet produced up to fifteen times more larvae that reached the pupal stage compared with colonies on standard diets. Supplemented colonies continued raising brood throughout the entire study period; unsupplemented colonies stopped producing brood after approximately ninety days. The nutrient profile of larvae in the supplemented group matched that of bees feeding on natural pollen.

Why does this matter structurally? Because most of the specific sterols bees need do not exist in commercially harvestable quantities in nature. No amount of wildflower planting or pesticide banning could produce a nutritionally complete artificial feed. If field trials confirm the laboratory results, the supplement could reach beekeepers within two years. It would also reduce competition for limited wildflower pollen, indirectly benefiting wild bee species.

Better Breeding

What Brother Adam did by observation and paper records over seventy-eight years, modern breeders can do faster using molecular tools, not to modify bees, but to identify desirable traits more quickly and select for them more efficiently. The breeding is still conventional. The selection is informed by better data. This is the same logic as using a blood test to identify which cattle carry a trait for disease resistance, rather than waiting years to see which ones get sick.

The most consequential breeding work being done today centres on a trait called Varroa-sensitive hygiene (VSH): the behaviour of detecting and removing mite-infested brood from the hive. Bees with strong VSH behaviour can smell the mites inside capped brood cells, uncap them, and remove the parasitised pupae before the mites can reproduce. It is a natural defence. Asian honeybees (Apis cerana), which co-evolved with Varroa, possess it instinctively. The question is whether European honeybees, which did not co-evolve with the mite, can be selected to express it strongly enough to suppress mite populations without chemical treatment.

The answer, after a decade of work, is yes. And the evidence comes from three distinct programmes, operating independently, on two continents. No organism has been genetically modified in any of them. Bees have been selected, with better measurement tools and coordinated effort, for traits they already possess.

Arista Bee Research, founded in 2014 in the Netherlands, has coordinated a network of over 300 beekeepers across seven countries in a systematic VSH-selection programme. In their 2022 assessment, the first year they passed the milestone of 1,000 test colonies, 130 breeders across 31 groups evaluated 1,065 colonies. The results: 34% of these colonies (367 in total) were classified as high-VSH, meaning they can suppress mite populations without any chemical treatment. Of those 367, fully 43%, or 149 colonies, were 100% VSH: after researchers deliberately introduced 100–150 mites into each hive, not a single reproducing Varroa mite could be found in the brood. They opened 300 to 600 brood cells per colony to verify this. By their ten-year review in 2024, Arista could report that in their own hives, there is “no longer bee mortality due to Varroa, and this without any chemical treatment.” At their Luxembourg mating station, more than half of all colonies showed high levels of Varroa resistance.

Meanwhile, in the United States, the USDA’s Honey Bee Breeding, Genetics, and Physiology Research Laboratoryin Baton Rouge has been developing Pol-line bees since the late 1990s, a stock selected specifically for VSH behaviour. In head-to-head commercial trials across four states (Mississippi, California, North and South Dakota), Pol-line colonies that received no mite treatment in the fall had a winter survival rate of 62.5%, compared to just 30% for standard untreated colonies, more than twice as likely to survive. Even when both groups received treatment, Pol-line bees maintained a significant advantage: 72% survival versus 56%. They also showed significantly lower levels of deformed wing virus A, deformed wing virus B, and chronic bee paralysis virus. As the lead researcher put it: “We would like to replace reliance on chemical controls with honey bees that have high mite resistance of their own.”

And the breeding logic is now reaching the commercial sector. Wes Card in Louisiana and Ryan Lamb in Texas have been independently selecting for mite resistance in their production stock for years. In February 2025, during the worst colony-loss season on record, Project Apis m. and the USDA’s Bob Danka visited both operations to evaluate their potential breeder queens. Colonies untreated since April 2024, through an entire production season and a full winter, showed very low mite infestations in the brood. During a year when 62% of commercial colonies nationwide were dying, these bees were thriving. They were also among the top honey producers in their respective operations.

One of the selection methods used across all these programmes deserves mention for its endearing hands-on quality. Researchers collect mites from the hive floor, put them under a microscope, and count how many legs are missing or partially chewed off. High levels of mite damage indicate colonies where worker bees are actively grooming and attacking parasites, a heritable behavioural trait. By breeding queens from these high-grooming colonies, breeders develop more resistant lines.

This is Brother Adam’s vision, scaled and diversified. More breeders. Better measurement. Faster feedback loops. The same fundamental logic: observe, select, breed, evaluate, repeat.

Better Practice

The 2024–2025 triage surveys revealed something the culture-war narratives completely obscure: colony survival varied enormously depending on management practice. Differences in mite-monitoring frequency, protein and carbohydrate feeding schedules, miticide timing, and overwintering methods all correlated with significantly different loss rates. The headline losses are real and alarming. But they are not uniform. Some beekeepers, using the same stock in the same regions, lost dramatically fewer colonies because they managed better.

The single most impactful intervention, according to extension specialists, is regular Varroa monitoring: counting mites per hundred bees at scheduled intervals and treating when thresholds are crossed. This is not high technology. It is disciplined practice. Many beekeepers, including some commercial operators, do not do it consistently. The gap between best practice and average practice is wide, and closing it requires training, not legislation.

An increasingly significant management innovation is indoor cold storage: placing colonies in climate-controlled facilities (typically around 7°C and 25% relative humidity) during winter months. The practice, long used by some Canadian beekeepers, has been rapidly gaining adoption among U.S. commercial operations. Colonies in cold storage stop producing brood, which creates a “brood break” that prevents Varroa from reproducing. Bees that cluster indoors rather than flying age more slowly and consume fewer resources. USDA-funded research combining Varroa-resistant Russian bees with cold-storage overwintering found that the approach produced survival rates and colony sizes comparable to outdoor apiaries in warmer climates, at lower cost per colony. In an era when warmer autumns are extending Varroa’s reproductive window, cold storage is a practical adaptation to climate change that requires no new legislation, no bans, and no breakthroughs. It requires a shed, a thermostat, and good planning.

On the landscape side, hedgerow and habitat restoration near farmland has shown measurable pollinator benefits. One study found bee abundance increasing by 8% within a single year of growing hedgerows, with species richness continuing to improve over seven years. Programmes like the Pollinator Partnership’s Bee Friendly Farmingcertification and the Xerces Society’s farmland conservation work have begun translating these findings into practical guidance for growers: flowering cover crops during fallow periods, reduced-mowing regimes on field margins, retention of scrubby marginal land that costs more to farm than the pollination benefit it provides by being left alone.

Training infrastructure exists and is growing. In the UK, BeeBase has grown from 19,000 registered beekeepers in 2010 to 47,000 today. In the U.S., state extension services, university programmes, and nonprofits like Project Apis m. offer free or reduced-cost courses for both commercial and hobbyist beekeepers. The infrastructure for improving practice exists. What it lacks is the political sex appeal of a pesticide ban or a corporate takedown.

Why the Discourse Fails when Work Succeeds

Bee decline is multicausal. Political narratives require identifiable villains. The mismatch between the complexity of the problem and the simplicity demanded by activism generates a predictable pathology: each side selects the causal factor that best fits its pre-existing ideological commitments, amplifies it, and dismisses the rest.

Applied breeding programmes, nutritional research, and extension education operate on different incentive structures. They are optimising for measurable outcomes: larval survival rates, mite loads per hundred bees, overwinter survival percentages, honey yield per colony. The feedback loop between intervention and outcome is tighter, more honest, and more productive than the feedback loop between a Twitter campaign and bee survival. One better queen this year. A slightly more resistant line next year. A measurably healthier apiary the year after that.

We should be careful, though, not to let our own argument become a comfortable distortion. “Activists bad, practitioners good” is a satisfying story, and it has the same structural flaw as “neonics bad, Bayer evil”: it is too clean. The practical work we have described does not fund itself. The USDA Pol-line programme exists because Congress appropriated money for it. The Inflation Reduction Act’s $20 billion in conservation funding exists because advocacy organisations pushed for it. Extension agents have jobs because state legislatures decided to pay for them. The 2008 and 2014 Farm Bills expanded federal bee research funding partly because beekeepers showed up and made noise. If the takeaway from this article is “ignore the political process and just do the work,” we have failed. The work requires political support, and political support requires advocacy.

The real failure is not activism per se. It is activism that optimises for the wrong objective function, measuring success in media coverage, petition signatures, and corporate-villain narratives rather than in research funding secured, extension programmes staffed, and breeding networks supported. It is also, and perhaps more damagingly, a conservative counter-narrative that uses “it’s just mites” as a reason to cut public spending on the very programmes that are producing results. The worst outcome is not that we fight about pesticides. It is that we fight about pesticides instead of funding the USDA, and then cut the USDA’s budget anyway.

There is the question of opportunity cost, but it runs in both directions. Every dollar of activist funding spent litigating a neonicotinoid ban is a dollar not spent training beekeepers in mite-monitoring techniques. But every dollar cut from USDA research budgets in the name of fiscal conservatism is a dollar that could have funded the next Pol-line. Every news cycle consumed by “beepocalypse” rhetoric crowds out the unglamorous truth that better feeding schedules and community-based breeding programmes are producing concrete results right now. But the silence that replaces it is not better, because the public learns nothing at all, and the appropriations committees have no political reason to act.

Objections

“You’re dismissing activism.” I hope the preceding section made clear that we are not. Without advocacy, the practical work has no funding. Our problem is with activism that measures its success in headlines and petition signatures rather than in research dollars secured and extension programmes staffed. The energy is not the problem. The targeting is.

“You’re letting corporations off the hook.” Agrochemical companies have engaged in documented efforts to shape scientific discourse and deflect blame. That behaviour warrants criticism and, where appropriate, legal accountability. It’s not even dead and buried events that drive this distrust, just Roundup alone more than justifies distrust and hostility again and again. But corporate malfeasance, real as it is, does not change the entomological reality: Varroa mites are the single largest proximate threat, nutritional deficiency is a growing crisis, and the most effective interventions available right now are better breeding, better feed, and better management. Holding both truths simultaneously is not corporate apologetics.

“You’re letting fiscal conservatives off the hook.” This is the objection we take most seriously, because our essay’s rhetorical structure (”stop fighting and start working”) can easily be co-opted by people who want to stop fighting and stop funding. Every programme I have praised depends on public money. The USDA breeding lab, the extension services, the university research grants, the nonprofit networks: all of them require sustained appropriations from legislatures that are perpetually looking for things to cut. If this article provides ammunition for anyone arguing that bee conservation is a solved problem that no longer needs government support, I have written it badly. It is not solved. It needs more support, not less.

“You can’t scale a monk.” Brother Adam had institutional backing from a monastery, no family obligations, a lifetime tenure, and the personality of a single-minded obsessive. His story is inspiring but not reliably replicable.

Fair enough. And yes, we would happily fund more monks (even if they are making violence in a wine). Institutions that provide long-term stability and patient capital for multi-decade applied research are invaluable. Brother Adam’s monastic context was not incidental to his success; it gave him something almost no modern researcher enjoys.

But most of the work now being done to save bees is being done by non-monks. The Arista Bee Research network is three hundred amateur beekeepers in seven countries, sharing data in spreadsheets and checking mite-damaged legs under microscopes on weekends. The USDA Pol-line programme is government scientists working within constrained budgets. Wes Card and Ryan Lamb are commercial beekeepers running businesses, not contemplative orders. The Oxford nutrition team are university researchers on fixed-term grants. Extension agents across the United States and the UK are civil servants offering free courses to anyone who shows up.

The method is replicable. The discipline is replicable. I are not arguing for more monks, though I wouldn’t turn that idea away. I am arguing for more of what the monk did: systematic, patient, craft-based work, carried out by communities of practitioners who share methods and data. That is already happening. It deserves more support than it currently receives.

The Monk, the Moor, and the Powder

Brother Adam spent seventy-eight years on Dartmoor, in the wind and the cold, scoring queens and recording pedigrees. Today a network of amateur breeders across Europe checks mite-damaged legs under microscopes and shares data in spreadsheets. A commercial beekeeper in Louisiana carries untreated hives through a catastrophic winter because he selected the right queens. A team at Oxford dries yeast into powder that may keep colonies alive through pollen-poor winters. A shed full of bees in Idaho sits dark and cool at seven degrees, breaking the mite’s reproductive cycle while the beekeeper saves twenty-three dollars per colony.

None of them need a hashtag. They need time, rigour, funding, and the recognition that in environmental crises, as in most things, it is work, not outrage, that bends the curve.

A 2025 college graduate applies to forty jobs, hears back from three, gets ghosted after two rounds of interviews with each. Her LinkedIn feed tells her why: AI is coming for entry-level work. Dario Amodei predicts half of entry-level white-collar roles will vanish by 2030. Jack Dorsey fires 40 percent of Block’s workforce and says intelligence tools have changed what it means to run a company. She starts to wonder if her degree was a mistake.

Marc Andreessen, on the 20VC podcast last week, told her she is being lied to. Most large companies are overstaffed by 25 to 75 percent thanks to pandemic-era hiring binges, he said, and AI is “the silver bullet excuse” to clean house without admitting the real reason. “AI literally until December was not actually good enough to do any of the jobs that they’re actually cutting. It just can’t have been AI.”

He is mostly right. The evidence we lay out below supports him. But his version has a convenient blind spot.

If these companies are so bloated, where are the good products? Microsoft has over 220,000 employees, yet Windows 11 shipped with a Start menu that broke basic workflows, an ad-infested lock screen, and a Settings app that still cannot do half of what the Control Panel could in 2009. Google has 180,000-plus people and has been trying to fix Google Assistant for a decade. Apple spent years and roughly 165,000 employees on a Vision Pro that sold to almost nobody. These are not organizations with 50 to 75 percent surplus headcount. They are organizations where management layers, internal politics, and misallocation have made it hard to ship anything well. The people are not unnecessary. The structure wastes them.

There is also something Andreessen will not say, because it implicates his own industry. If cost of capital were still near zero, as it was from 2009 to 2022, Andreessen Horowitz would be writing bigger checks into more portfolio companies, each hiring aggressively to chase growth before profitability. Venture capital ran on that model for over a decade: fund companies, tell them to hire fast, grow into the valuation later. The people now called “bloat” were, three years ago, called “scaling.” What changed was the federal funds rate. When capital is expensive, every investor suddenly discovers their portfolio companies have too many people. Andreessen is diagnosing a symptom and calling it a disease.

Meanwhile the graduate’s problem is real, and it is worse than either narrative acknowledges. The share of unemployed Americans who are new workforce entrants hit a 37-year high in 2025, peaking at 13.3 percent in July, exceeding anything recorded during the Great Recession. It settled to 10.6 percent by February 2026, still worse than the worst of 2008-09. The New York Fed reports that the underemployment rate for recent college graduates climbed to 42.5 percent by the fourth quarter of 2025, meaning almost half the people who just spent four years and six figures on a degree are working jobs that never required one. Finance and information services, the traditional on-ramps for college graduates, have been shedding an average of 9,000 jobs per month since 2023. Before the pandemic, those same industries were adding 44,000 jobs per month.

She is right that the ladder is broken. She is wrong about what broke it. And Andreessen, for all his bluster, is wrong about why.

It is not AI. But it is not simply a correction of pandemic overhiring, either. The true culprits run deeper: a job ladder that has lost half its rungs over four decades, rising employer concentration, the quiet spread of noncompete agreements into low-wage work, and the sharpest monetary tightening cycle in forty years, now about to be compounded by an oil shock that will freeze what little entry-level hiring remains. The wrong diagnosis leads to the wrong treatment, and young workers cannot afford the delay.

Forty-Years of Collapse

To understand what is happening to the graduate, we need to go back much further than ChatGPT. Four decades further.

Labor economists have used the metaphor of the job ladder for so long that we sometimes forget it describes a real mechanism. The job ladder is not a metaphor for ambition. It is the institutional infrastructure through which labor markets allocate talent and distribute gains. A young worker enters the labor market and takes a job at a firm that may not pay particularly well. Over the next several years, she receives offers from competing employers. Some pay more. She accepts the better ones and moves up. Robert Topel and Michael Ward documented in 1992 that the average American worker has seven employers in the first ten years. That churn is not dysfunction. It is the mechanism by which wages grow. About 60 percent of the wage growth in the first decade of a career comes not from raises within a firm but from moving between firms toward better-paying employers.

That is how it is supposed to work. A March 2026 working paper from Niklas Engbom, Aniket Baksy, and Daniele Caratelli shows how badly it has broken down. Using Current Population Survey microdata from 1982 to 2023 and a job-ladder model calibrated to match the data, they estimate that employed workers today are roughly half as likely to receive a better-paying outside offer as they were in the 1980s. Net upward job mobility, their summary measure of ladder strength, fell by 51 percent between the early 1980s and the 2010s.

Half the rungs are gone.

The consequences for wages are severe. After more than doubling between 1940 and 1970, real hourly earnings in the United States have increased by only about 25 percent since 1980. If you hold the demographic composition of the workforce constant at its 1980s level, stripping out the effect of the shift toward older and more educated workers, real wage growth since 1980 is close to zero. Engbom, Baksy, and Caratelli estimate that the decline in the job ladder accounts for about one-third of that slowdown, reducing annual real wage growth by 0.68 percentage points. Over four decades, that compounds into a very large number.

Only about one-third of the wage effect is mechanical, meaning workers climb the ladder less often and therefore end up at lower-paying firms. The remaining two-thirds arises because firms choose to pay less when they face less competition for workers. We tend to think of the job market as a place where employers set wages and workers accept or decline. The Engbom paper shows that a large share of what determines wages is not the employer’s productivity or the worker’s skill but the threat of departure. When that threat recedes, as the Nobel laureate Peter Diamond predicted, firms converge toward offering the minimum workers will accept. The ladder is the invisible discipline that makes the whole system pay what it should. And the discipline has been quietly collapsing for forty years.

The decline is broad-based. Men and women, white and nonwhite workers, college graduates and non-graduates all saw large drops in upward mobility. But the decline is especially pronounced for young workers, whose net upward mobility fell from 1.286 in the 1980s to 0.482 in the 2010s. Women experienced a larger decline than men. Nonwhite workers fell from 1.099 to 0.423. There is no demographic group for which the ladder is working as it once did.

The graduate applying to forty jobs in 2025 would recognize the symptoms even without seeing the data. A Rezi.ai analysis of job postings found that 35 percent of positions labeled “entry-level” now require three or more years of experience. Entry-level roles requiring zero to two years have dropped by 29 percentage points. This “experience inflation” is what the collapsed ladder looks like from the ground. When firms face no competitive pressure to invest in developing junior talent, they redefine entry-level to mean mid-career. The label stays. The opportunity vanishes.

Engbom, Baksy, and Caratelli rule out several plausible alternative explanations. The decline is not driven by lower matching efficiency: job-finding from nonemployment has declined only modestly over forty years, while job-finding from employment has fallen sharply. If the matching technology or labor demand were the problem, both rates would have moved together. The decline is not driven by house lock-in: renters experienced larger declines in on-the-job search efficiency than homeowners. It is not driven by dual-career constraints: single-career households experienced larger declines than dual-career households. This is not a story about workers who stopped climbing. It is a story about an economy that pulled the ladder away.

Who Pulled the Ladder Away

Two forces show up consistently in the cross-state evidence: rising employer concentration and the proliferation of noncompete agreements.

States where employer concentration increased more between the 1980s and 2010s experienced larger declines in upward job mobility. States where a higher share of workers report being bound by noncompete agreements show the same pattern. A back-of-the-envelope calculation by Engbom, Baksy, and Caratelli implies that rising concentration and noncompete use together may account for roughly 60 percent of the national decline in the efficiency of on-the-job search.

The concentration story is about arithmetic. Fewer employers in a labor market means fewer competing offers for employed workers. Fewer competing offers means less pressure on wages. You can see the same mechanism in industry after industry. In oil and gas, as documented in a previous article, roughly $200 billion in mergers since late 2023 has consolidated half of the Permian Basin’s most productive formation under two companies. ExxonMobil acquired Pioneer for $60 billion, Chevron acquired Hess for $53 billion, and Diamondback merged with Endeavor for $26 billion. The wildcatter ecosystem that once converted price signals into production and employment, chaotically and wastefully but at enormous scale, was replaced by capital-disciplined giants that return cash to shareholders regardless of what prices do. The labor market consequences follow: fewer firms competing for workers, less upward mobility, lower wage pressure.

The same dynamic, less dramatic but more pervasive, operates across the broader economy. It does not require conspiracy. It requires only that the number of independent employers shrinks and that the remaining employers face less competitive pressure to bid for workers.

The noncompete story is more galling. Legal experts describe the period from 1990 to roughly 2010 as the golden age of noncompete enforcement in America. What started as a tool for protecting senior executives’ trade secrets metastasized into a blanket restriction applied to hourly workers, sandwich shop employees, pet cremation technicians. An estimated 30 million Americans, nearly one in five workers, are bound by a noncompete agreement. These agreements directly suppress the mechanism through which the job ladder operates: they prevent employed workers from accepting better offers.

The FTC, under the Biden administration, attempted a nationwide ban. The estimated effects were large and specific: $400 to $488 billion in increased wages over the next decade, $524 per worker per year in additional earnings, 8,500 new businesses annually, and 17,000 to 29,000 additional patents per year. The rule was struck down in federal court. The Trump administration formally vacated it in September 2025. The FTC has shifted to case-by-case enforcement, including a February 2026 consent order against a pet cremation company that had imposed blanket noncompetes on 1,780 employees, including hourly laborers and drivers. The bipartisan Workforce Mobility Act, reintroduced in June 2025 by Senators Murphy, Young, Cramer, and Kaine, would ban most noncompetes nationwide. It has been referred to committee. No further action has been taken. Over 150 bills have been introduced in more than 35 states, creating a patchwork that varies by jurisdiction. The patchwork is the opposite of the clear nationwide signal that would restore competitive dynamics.

The graduate scrolling LinkedIn is not competing against chatbots. She is competing against four decades of eroded mobility, in a labor market where the companies that might hire her face less pressure to do so than at any point since the data began. We all spent two years worrying that AI will trap young workers in obsolete careers, if they every get a career in the first place. Meanwhile, noncompete agreements have been legally trapping workers in underpaying jobs for decades, and we barely noticed.

AI Alibi

So if the ladder was already broken, why does everyone keep blaming AI?

Partly because there is a credible-looking academic case. Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen’s 2025 working paper, aptly titled “Canaries in the Coal Mine,” finds a 16 percent relative decline in employment for early-career workers ages 22 to 25 in the most AI-exposed occupations since the public release of ChatGPT in November 2022. Brynjolfsson himself reiterated the finding at a Stanford panel in March 2026: about a 20 percent decline for entry-level software developers, about 15 percent for call center workers. Mid-career people are doing fine. Senior people are doing well. The narrative has cultural momentum. A Harvard survey found that 44 percent of young Americans believe AI will reduce job opportunities, while only 14 percent expect it to create more.

Let’s take this argument seriously. But taking it seriously means subjecting it to scrutiny.

The central weakness is the timeline. The “Canaries” paper documents an employment inflection point beginning in November 2022, immediately following ChatGPT’s public launch. By June 2023, roughly half of the total observed decline had already materialized. For this to be an AI story, we need to believe that within six months of a consumer chatbot’s release, firms across the economy decided that AI could replace junior staff, built the necessary enterprise infrastructure, redesigned complex workflows, ensured data security and regulatory compliance, and executed staffing changes at national scale.

That did not happen.

The OpenAI API, a prerequisite for building any custom application, only launched on March 1, 2023. ChatGPT Enterprise, which offered the data privacy guarantees that corporations require before touching sensitive operations, launched on August 28, 2023. The generative AI models available in late 2022 and early 2023 fabricated information freely enough that the term “hallucination” entered common use. U.S. Census data shows that fewer than 10 percent of large businesses were even planning to use AI in the next six months as late as the fourth quarter of 2023. By the third quarter of 2025, actual adoption among large businesses had only climbed to 12 percent. The tools that would have been necessary for the displacement story to work did not exist during the period the displacement supposedly occurred.

What did exist was a hiring freeze, and it started before the chatbot arrived. An analysis of job postings data from Lightcast, published by Zanna Iscenko and Fabien Curto Millet of Google’s economics team in their January 2026 paper for the Economic Innovation Group, shows that vacancies for the highest AI-exposure quintile of occupations peaked in March and April of 2022 and declined sharply throughout the rest of the year, a full six months before ChatGPT launched. The decline in employment levels starting in November 2022 is the predictable, lagged consequence of that earlier hiring freeze: when you stop bringing in new graduates, routine attrition is no longer offset by new hires, and the headcount of 22-to-25-year-olds mechanically shrinks.

Iscenko and Curto Millet illustrate this with a simple thought experiment. Take a hypothetical occupation that starts with an equal number of workers at each age from 22 to 65 and impose a complete hiring freeze but zero layoffs for one year. When you study this occupation a year later, entry-level employment for the 22-to-25 age band has shrunk by 25 percent. Employment for every other age group has not changed at all. The statistical pattern perfectly mimics targeted AI displacement. It is not AI displacement. It is arithmetic.

The job postings data also contradicts the AI story on its own terms. Within occupations most exposed to AI, postings for junior roles have not declined faster than postings for senior roles. Both fell roughly in parallel from their spring 2022 peak, with junior postings actually stabilizing faster. If AI were selectively automating entry-level tasks, the divergence between junior and senior postings would be the first place you would see it. It is not there.

Why, then, did AI-exposed occupations get hit harder than others? Because “AI exposure” and “interest rate sensitivity” overlap almost completely. Occupations in the top quintile of AI exposure are overwhelmingly concentrated in information, finance and insurance, and professional and technical services. Approximately 38 percent of workers in the most AI-exposed quintile are in these sectors, compared with less than 2 percent in the least-exposed quintile. These are the sectors most sensitive to capital costs and economic uncertainty. When the Federal Reserve began its most aggressive tightening cycle in forty years in March 2022, the timing lined up exactly with the decline in job postings across these sectors. Research by Zens, Böck, and Zörner has found that workers in tasks rated as easily automated are also disproportionately affected by conventional monetary policy shocks. The correlation is confound, not coincidence.

And the same differential pattern appears in the hiring slowdown of early 2020, when generative AI could not even theoretically be the explanation. If the same occupations decline relative to others during every downturn, AI is not required to explain the pattern. Monetary policy and sector sensitivity are sufficient.

The finding is not isolated. Gimbel, Kinder, Kendall, and Lee find no discernible break in aggregate U.S. employment trends since ChatGPT. Humlum and Vestergaard, using extremely detailed Danish microdata, find that AI’s impact on individual earnings and hours is “precise zeros,” with workplaces adopting AI showing no shifts in job creation or destruction. They do find a decline in early-career employment, but their difference-in-differences analysis shows AI is not driving it. A Yale Budget Lab report described the notion of AI roiling the job market as “largely speculative.” The Vanguard finding is the sharpest of all: their December 2025 economic outlook reports that the roughly 140 occupations most exposed to AI automation are actually outperforming the rest of the labor market in both job growth and real wage increases since the second quarter of 2023. If AI were displacing workers, these occupations would be the worst place to look for a job. They are not.

AI may yet reshape entry-level work in ways we cannot currently measure. But the claim that it has already done so at scale is not supported by the evidence. So why do companies keep saying it?

Because it works. The practice has acquired a name: AI washing. Of the 1.2 million job cuts U.S. companies announced in 2025, nearly twice 2024’s total, AI was cited in only about 55,000, or 4.5 percent, according to the research firm Challenger, Gray and Christmas. A January 2026 Forrester report found that “many companies announcing A.I.-related layoffs do not have mature, vetted A.I. applications ready to fill those roles,” describing the trend as companies attributing financially motivated cuts to future AI implementation that does not yet exist. Nearly 60 percent of hiring managers surveyed by Resume.org said they emphasize AI’s role in reducing hiring or cutting jobs specifically because it is viewed more favorably than admitting to financial constraints.

The cases are specific enough to be embarrassing. Amazon’s CEO Andy Jassy initially linked the company’s rounds of layoffs to AI and generative tools, then walked it back, saying the cuts were “not really AI-driven, not right now at least.” Jack Dorsey’s Block fired over 4,000 employees in February 2026 citing AI, but the company’s own filings show it had grown from 5,477 employees in 2020 to over 10,000 by 2025 during a pandemic-era hiring spree. The cuts brought headcount roughly back to where it was before the over-expansion. Meta reportedly planned to cut 20 percent of its workforce not because AI does those workers’ jobs, but to pay for the servers that run AI. As Bloomberg Opinion put it in March 2026: “Don’t blame AI for poor management decisions.”

The most telling datapoint is the simplest. Since March 2025, New York State has given employers the option to cite “technological innovation or automation” in the legally required WARN Act layoff notices they file before mass reductions. Of the 160 companies that filed notices since then, including Amazon and Goldman Sachs, companies that freely cite AI efficiencies in their investor communications, not one checked the box attributing layoffs to AI. When it matters legally, the AI explanation disappears.

Why would executives do this? Peter Cohan, a management professor at Babson College, told Built In that AI is “the least bad reason companies can use” for layoffs. Blaming tariffs risks political retaliation. Blaming revenue shortfalls spooks investors. Blaming pandemic over-hiring is an admission of strategic failure. Blaming AI sounds forward-looking and market-friendly. Shares go up.

The reputational cost, as Mercer’s Global Talent Trends 2026 report documents, falls on workers: employee concerns about AI-related job loss jumped from 28 percent in 2024 to 40 percent in 2026, and 62 percent of employees feel their leaders underestimate the emotional impact of AI on the workforce. When companies AI-wash their layoffs, they corrode the trust needed for workers to engage productively with AI tools. The wrong diagnosis does active damage.

The companies that ghosted our graduate after two rounds of interviews are, in many cases, the same ones citing AI efficiencies in their earnings calls. Not every company is following that script. In February 2026, IBM announced it was tripling entry-level hiring, including for software developers and other roles “we’re being told AI can do.” IBM’s chief human resources officer, Nickle LaMoreaux, explained the logic at the Leading with AI Summit: the entry-level jobs of two to three years ago can largely be performed by AI, so IBM rewrote the job descriptions. Junior developers now spend less time on routine coding and more time working directly with customers and building new products. The entry-level job description changed. The entry-level job survived. Her warning to competitors who are cutting instead was equally direct: companies that forego entry-level hiring will have to poach mid-level employees from competitors at a 30 percent premium, people who don’t know the internal culture and take longer to get up to speed. Axios called it a “narrative violation.”

The Oil Shock Will Finish What the Fed Started

That is the structural story: a ladder that has been losing rungs for four decades, weakened by concentration and noncompetes, with AI taking the blame for damage it did not cause. Now comes the cyclical crisis that will make all of it worse.

The proximate cause of the entry-level hiring collapse of 2022 to 2025 was the Federal Reserve’s tightening cycle, which began in March 2022, the exact month when job postings in those same rate-sensitive sectors began falling. The tightening was necessary. Inflation was running well above target. But the distributional consequences were predictable: the sectors where young, college-educated workers concentrate are the sectors most exposed to rising capital costs. As John Haltiwanger, Henry Hyatt, and Erika McEntarfer demonstrated in their 2018 research, the job ladder is highly procyclical, and more so for younger workers. When the economy tightens, hiring stops, workers stop quitting, and entry points to the market disappear.

Indeed’s Hiring Lab describes the current environment as “low-fire, low-hire.” Employers are not expanding but not cutting either. Comfortable for incumbents. Devastating for anyone trying to get a first foothold. ZipRecruiter data shows that employee turnover dropped from 177 percent in 2023 to just 50 percent in 2025. Workers are clinging to positions. Vacancy chains are frozen. The people who suffer most from a frozen market are, by definition, the people who don’t yet have a place in it.

There were, as of early 2026, cautious signals of a thaw. ZipRecruiter found that 63 percent of businesses plan to increase hiring, with emphasis on entry-level roles. Some forecasters saw rate cuts on the horizon.

Then the Strait of Hormuz closed.

On February 28, 2026, the war between the United States, Israel, and Iran began. Within 48 hours, tanker traffic through the Strait, which carries roughly 20 percent of the world’s seaborne oil, had collapsed by 81 percent. Seven of twelve major marine insurance clubs cancelled war risk coverage. Brent crude surged above $82, with JPMorgan warning of $120 and Deutsche Bank modeling a full closure at $200. European diesel futures jumped 23 percent in a single day.

We laid out in detail what this oil shock means for energy markets, inflation, interest rates, and fertility. What matters here is the implication for an entry-level labor market already hanging by a thread.

The transmission runs through a short chain. Higher oil prices feed into broader inflation. The SF Fed found in December 2025 that two-year Treasury yields now respond more than three times as strongly to oil supply news as they did before 2021. Markets are already pricing in no rate cuts this year and possibly further hikes. Each increment of tightening falls hardest on the sectors already hemorrhaging entry-level positions. An oil-driven inflation spike does not create new jobs for graduates anywhere in the economy. It guarantees that the interest rate environment remains hostile for the sectors that would otherwise absorb them.

In a prior era, an oil shock at least had a compensating mechanism. High prices meant pain at the pump, but they also meant a drilling boom in West Texas, the Gulf Coast, and the Intermountain West. Young men without college degrees could earn $80,000 to $120,000 on a rig crew. That income supported marriages, mortgages, and children. Between 2010 and 2014, Houston alone added 457,500 jobs.

That industry no longer exists. As we documented, capital discipline and consolidation have produced an oil sector that converts high prices into share buybacks rather than employment. The U.S. rig count fell from 750 in December 2022 to 550 by late February 2026, while production held near record levels. The total upstream workforce has shrunk by 252,000 from its peak while producing substantially more energy. Petroleum engineering degrees have collapsed 76 percent from their 2017 peak. When EOG Resources held its fourth-quarter earnings call the day before the Iran strikes, it committed to returning 90 to 100 percent of free cash flow to shareholders and keeping production flat. Scott Sheffield, formerly of Pioneer Natural Resources, said it as plainly as anyone: “Whether it’s $150 oil, $200 oil, or $100 oil, we’re not going to change our growth plans.”

So: the oil shock raises inflation, which delays rate cuts or forces hikes, which prolongs the hiring freeze, which extends the entry-level drought. The sectors that would historically have absorbed displaced workers are either structurally configured not to expand (energy) or exposed to diesel and input cost increases (construction, manufacturing). And the broader uncertainty, compounded by tariff escalation and immigration policy changes, ensures that employer risk aversion persists.

The graduate entering the labor market in the spring of 2026 faces a convergence with no close parallel in the postwar period. A ladder weakened over four decades. A tightening cycle that has frozen what remains. An oil shock extending the freeze. And a public conversation that remains fixated on a chatbot.

Compounding (Economic) Scars

If this were merely a matter of waiting a few years for conditions to improve, none of what follows would matter much. It is not.

The research on labor market scarring says that what happens to our graduate in the next two years will follow her for decades. Till von Wachter’s 2020 survey of the evidence in the Journal of Economic Perspectives documents what happens to young workers who enter the labor market during a recession. For a typical downturn, where unemployment rises by 4 to 5 percentage points, college graduates experience an initial earnings reduction of about 10 percent that takes 10 to 15 years to fade. High school graduates fare roughly twice as badly. Nonwhite entrants experience larger earnings losses, mostly driven by reductions in weeks worked in the first four years.

Schwandt and von Wachter (2019) study puts the cumulative cost in dollar terms: among all labor market entrants, entering during a large recession reduces the present discounted value of earnings over the first ten years by 9 percent, rising to 13 percent for those without a high school diploma and 11 percent for nonwhite workers. These losses amount to three-quarters of mean annual earnings for the average entrant over the first decade.

The effects go past paychecks. Adverse labor market entry persistently increases alcohol consumption, leads to higher obesity and more smoking in middle age. College graduates who entered during the 1980s recession showed higher rates of heart attacks decades later. By their late 30s, according to Schwandt and von Wachter (2020), unlucky entrants start dying at higher rates than their luckier peers, a gap driven by heart disease, liver disease, lung cancer, and drug overdoses. The gap keeps widening through their 40s.

We keep calling this a labor market problem. By their late 30s it is a mortality problem.

Unlucky cohorts have fewer children, divorce more often, and are more likely to live alone in middle age. Criminal activity rises for at least 15 years after entry, especially for men and high school dropouts. Self-esteem erodes. None of this is temporary. A young worker entering a weak labor market in the 1980s could still recover, because the structural machinery of upward mobility remained intact. Today that machinery is broken.

And the safety net was never designed for people at the start of their careers. Unemployment insurance, job search assistance, and retraining all require work history that new entrants, by definition, do not have.

The effects on family formation deserve particular emphasis, not least because they have downstream consequences that last generations. Drawing on Stephen Shaw’s 2025 analysis of 314 million mothers across 33 higher-income countries, that economic shocks suppress entry into motherhood rather than reducing family size among women who are already mothers. The oil shock of 1973 drove simultaneous declines in the Total Maternal Rate across all 47 Japanese prefectures. The 2008 financial crisis drove a sustained decline in U.S. first births that continued well after the economy recovered. Hungary’s TFR collapsed from 1.59 to 1.39 between 2021 and 2024 as the energy crisis erased a decade of pronatalist policy gains. Czechia’s fell from 1.83 to 1.37, the lowest since records began in 1806.

The mechanism runs through male employment. When young men cannot find stable work, marriages do not form, and first births do not happen.

A 2025 study in the Chinese Sociological Review found that approximately a third of Korea’s fertility decline in the 25-to-29 age group traces to a near-tripling of male economic inactivity. The collapsing job ladder is the slow-motion version of the same mechanism. And each successive shock operates on a lower baseline, which means the same-sized disruption does more damage.

When the graduate expresses cynicism about the economy’s capacity to reward her effort, that cynicism is rational. She is responding to an economy that has, in measurable and documented ways, stopped rewarding effort as reliably as it once did.

The Ladder Is the Point

We are drawn to dramatic explanations because they are legible and because they absolve us of responsibility. If a machine took your job, no human decision is to blame. Andreessen’s version is only slightly better: if pandemic overhiring is the culprit, the correction is mechanical and self-resolving. Neither account requires anyone to do anything differently.

But the evidence assembled here points to something less cinematic and more damning: policy choices, market structures, and institutional failures that have systematically weakened the mechanism through which young workers build careers and economies distribute prosperity. Consolidation eliminated competitive ecosystems and noncompetes legally prevented workers from climbing. Concentration reduced the pressure that made firms pay well. The Fed’s tightening cycle froze what remained of the ladder, and the oil shock will extend the freeze.

If the diagnosis is wrong, the prescription will be wrong too. If you believe AI is displacing entry-level workers, you prescribe AI literacy training, prompt engineering curricula, reskilling programs oriented around technology adoption. These may be useful in their own right. But they do not address a collapsing job ladder, rising employer concentration, or noncompete agreements that legally prevent workers from accepting better offers. Prescribing AI training to address a structural labor market crisis is like prescribing swimming lessons to someone who is living in a drought blighted area.

The structural diagnosis implies different interventions. On employer concentration, antitrust enforcement oriented not only around consumer prices but around labor market competition and the ability of workers to receive competing offers. On noncompete agreements, finishing what the FTC started. The estimated $524 per worker per year in additional earnings from a nationwide ban is not a theoretical exercise. Oregon’s 2008 ban on noncompetes for hourly workers raised job-to-job mobility by 12 to 18 percent. The mechanism works. On social insurance, programs that do not condition eligibility on prior work history, which new entrants by definition lack. On the oil shock, recognizing that monetary tightening in response to a supply-side price spike is, as Bernanke and Blanchard’s own research has shown, the wrong tool for the problem.

None of these are mysteries. The Workforce Mobility Act has bipartisan sponsors. It is stalled in committee. The evidence on labor market concentration has been accumulating for a decade. The scarring research has been published and replicated. The bill is written. The papers are cited. The data is not ambiguous.

And nothing is moving.

That is the part that is hard to write about calmly. Not the structural decline, which at least has the dignity of being a long process driven by large forces. The part that is hard to sit with is the gap between what we know and what we are doing about it. We know that the job ladder is broken. We know what broke it. We know what the scarring research predicts for the cohort entering the labor market right now. We have interventions with bipartisan support, proven track records, and detailed cost-benefit analyses gathering dust in committee.

The graduate we started with is not an abstraction. She is one of millions people who finished college in 2025 (not just in the US) and walked into the worst entry-level market in four decades. The scarring literature says that what happens to her in the next two years will shape her earnings, her health, her family, and her life expectancy for decades. That should bother us more than it does.

Here is a fact that should not be true.

A detached accessory dwelling unit in Los Angeles costs $200--$600 per square foot to build. A unit in a conventional apartment building costs $150--$250. On every metric of construction efficiency, the apartment wins. The ADU is a bespoke backyard project with fixed costs that barely budge whether you build 400 square feet or 800. The apartment is an industrial product: 200 identical units sharing one foundation, one elevator core, one set of plans. Any first-year real estate finance student could tell you which is the efficient way to house people.

And yet.

California permitted 1,269 ADUs in 2016. By 2019: 14,702. By 2023, Los Angeles County alone permitted more than 45,000. One in three homes now permitted in the city of LA is an ADU, in neighborhoods where apartments hadn’t been built in forty years. Meanwhile, national multifamily production has been stuck at roughly 350,000 units a year for four decades. Through booms and busts, through low rates and high. The line barely moves.

So the expensive, inefficient option is the one that’s actually scaling, while the cheap industrial product can’t get out of its own way. Understanding how this trick works is the key to understanding how housing actually gets built in America. And the YIMBYs reading this (who have spent years fighting zoning battles and are understandably impatient) may find the mechanism delightfully underhanded.

Are You Asking (The Right) Question?

It is easy for housing advocates to hyper-focus on this question: How do we reduce the cost of producing a housing unit? That question leads to debates about modular building, materials innovation, labor costs, and construction productivity. Reasonable debates, sometimes very fun and certainly very important, but almost entirely beside the point.

The key question: What does it actually take to get a unit built, occupied, and on the market?

Ask that, and something interesting happens (as we see in real life). A $300,000 ADU where the homeowner already owns the land, already has equity to finance it, and can get a ministerial permit in 60 days has a total cost to the system that is far lower than a $200,000-per-unit apartment that took 3 years to entitle, 12 months to secure a construction loan, 24 months to build, and required $50 million in institutional capital. The ADU’s construction cost premium is the price of bypassing every bottleneck that makes housing expensive to produce in America.

That premium turns out to be a bargain.

Paul Williams’s Raising the Housing Investment Level documents the bottlenecks with precision. FHA’s 221(d)(4) program, on paper one of the best construction financing products ever designed, takes 270--360 days to close versus 60--90 for conventional financing. Banks have tightened construction lending since 2022. And the result is visible in a single statistic: Williams estimates that roughly 750,000 multifamily units sit in a “shadow pipeline”. Entitled by local planning departments, having survived the multi-year gauntlet of public hearings and environmental review, but not yet applying for building permits because the financing isn’t there. Beyond that, another 117,000 units are formally authorized but haven’t broken ground. The pipeline is clogged at every stage. Multifamily starts fell 25% in 2024 to just 354,000 units, on top of a 14% decline in 2023, even as completions from the 2021--2022 boom continue to deliver, outpacing new starts by over 220,000 units in the past year. The pipeline is draining and not refilling.

ADUs skip all of this. The homeowner is the capital source (home equity), the decision-maker, and the developer. One person, one kitchen-table conversation, one HELOC application. 53% of California ADU builders paid cash. The average Massachusetts homeowner sits on $337,000 in tappable equity. There are 70 million owner-occupied single-family homes in this country. After legalization, each one is a pre-entitled development site with a pre-qualified capital source and a motivated builder.

A System Designed to Produce Nothing

Here is a sequence a typical apartment project must survive: land acquisition → discretionary zoning entitlement (public hearings, design review, political negotiation) → environmental review (NEPA, CEQA) → architectural design → construction financing (its own underwriting, appraisal, committee process) → building permit (plan review: 6--12 months) → construction (18--36 months) → certificate of occupancy → permanent financing takeout. Every stage involves different actors, different requirements, different timelines, and projects die in the handoffs between them.

Williams describes the FHA version: applications move through a fragmented chain, different specialists reviewing the appraisal, market study, cost estimate, sponsor financials, with no single person owning the deal. A 2016 Inspector General audit found the “single underwriter” reform was never actually codified. A 2024 follow-up found HUD still tracked assignments on manual spreadsheets across five regional centers. This program finances 10,000-15,000 units a year, roughly 4% of multifamily production, in a country short 3 to 4 million homes.

The ADU system: one decision-maker, one site, one ministerial permit, one contractor, one financing decision. Decision to occupancy: 6-12 months. The system is lean not by design but by accident. The unit of production is too small for complexity to accumulate. The homeowner feels every cost overrun and every rent check directly. No principal-agent problem. Especially in cities with pre-approved plan programs, fewer moving parts means fewer ways to fail.

The per-square-foot premium is an escape tax. You’re paying to leave a system so broken it can barely function.

It’s Not Housing Policy. It’s a Product.

A common theme among marketing and advertising is that value is perceived, not objective. A $2 cup of coffee becomes $6 in a ceramic mug with good lighting. Some marketers calls this alchemy.

In that sense, ADUs are alchemy.

The homeowner deciding to build one is not performing a pro forma. She’s thinking: Mom is getting older and I want her close but not in my kitchen. My daughter graduated and can’t afford rent. My property taxes went up and $2,500 a month would change everything.

It is a family decision, a retirement decision, a keeping-up-with-the-Joneses decision. The housing unit is a byproduct. FHFA data shows California properties with ADUs appraise at $1,064,000 versus $715,000 without, a gap growing at 9.34% annually since 2013. Homes with ADUs sell for 20-35% more. One study found a 40 to 60% average increase in property value. Rents run $2,000-$4,000 per month. And the land, the single largest cost in multifamily (often 30--50% of total project cost), is already paid for. She already owns it.

Her comparison isn’t “ADU versus apartment per unit.” It’s “ADU versus doing nothing and missing the rental income, the appreciation, or it’s a solution for Mom.” At almost any construction cost, the ADU wins.

I am puzzled by, well, actually I laugh at economists puzzled by the cost premium (yes, I am aware that some economists are aware of this, some just really want to stick with the purity of Econ 101 or just plain hate Econ 102 or 202 ie microeconomics). They assume all actors optimize on the same variable. The homeowner is optimizing on a bundle (financial, emotional, practical) that’s invisible to macro analysis. The premium looks irrational only if you ignore everything she actually cares about. Housing production is just a side effect.

The implication for YIMBYs, and this is the part that should make you grin, is that we have more ways on top of political advocacy to increase the number of units in a city. Once legalized, tactics like figuring out how to market ADUs are on the table or working with local and state-level developers and banks on something that can get units online ASAP.

Preying on NIMBY Self-Interest

Here is where the alchemy gets unsettling in the best possible way.

The NIMBY who opposes the apartment building on her block will never stop being a NIMBY. She will show up to every hearing, file every appeal, fund every lawsuit. NIMBYism is a political identity, not a misunderstanding, and she’s not going to grow out of it.

But that same NIMBY will build an ADU in her backyard without blinking. Not because she’s been converted (she hasn’t) but because what’s good for her is, in her mind, different from what’s good for everyone else. The apartment threatens her neighborhood and brings strangers she didn’t choose. The ADU enriches her property and houses Mom, or a tenant she screens personally, who pays $2,500 a month. Same unit of housing. Completely different calculus, because self-interest overrode principle.

People cheerfully exempt themselves from rules they impose on others when doing so is profitable. A crafty YIMBY can prey upon that (which is another article in itself!).

The play: make it as easy, as profitable, and as frictionless as possible for everyone , including NIMBYs to build ADUs after ADUs are legalized, knowing they’ll keep fighting every apartment, every triplex, every upzoning. The pitch is: Your home is your biggest asset. Here’s how to make it work harder for you. Here’s the pre-approved plan. Here’s the HELOC calculator. Here’s what your neighbor earns in rent. Here’s where Mom could live. The word “density” never appears, and the housing unit gets built anyway.

You don’t need converts, just to appeal to the more self-interested actors inside a permissive regulatory framework. The NIMBY builds her ADU for the same reason Willie Sutton robbed banks: that’s where the money is.

And every ADU she builds makes the next affordability fight easier. Not because she changes her mind, but because the neighborhood changes around her. One ADU on a block is invisible. Five normalize density. Ten make the “character of the neighborhood” argument sound ridiculous, because the neighborhood already has ADUs and I want one. Greed, not ideology, is the transmission mechanism. California ADU permitting grew 42-76% every year from 2016 onward, driven largely by neighbor-to-neighbor diffusion. She built one. I want one too.

Why the Battle Stays Won

ADU legalization is fought once, at the legislative level, and once it’s law, individual projects are ministerial: no hearing, no council vote, no neighbor veto. The YIMBY movement fights the one battle that matters, and then self-interest handles the rest.

Auburn, Maine proves this beautifully. Mayor Jason Levesque, “The YIMBYest City in America”, spent six years pushing ADU legalization, form-based code, and zoning reform. Permits spiked from 26 units in 2021 to 237 in 2022. Then in November 2023, he got crushed 62--38 by Jeff Harmon, a retired state police deputy chief who had literally sued a housing development near his property.

The NIMBYs won the election, but ADUs stayed. The ADUs are standing, the apartments are occupied, and the reforms were not repealed, because a new mayor can’t unbuild a house. Levesque lost, but the production outlasted his political career. And while Auburn is now certainly going after other types of housing, Levesque-era ADU regulations are left mostly alone.

(YIMBYs: if you take nothing else from this piece, take Auburn. Win the legislative fight. Let self-interest handle the rest. The housing survives the backlash.)

Jersey City: The Sophisticated Version

Auburn is the optimistic case: a blunt NIMBY who can’t unbuild houses. Jersey City under Mayor James Solomon, who took office in January 2026, is the worrying one.

Solomon is no Jeff Harmon. He doesn’t oppose development. He knows better. He concedes openly that “rents would be higher if a lot of those buildings hadn’t been built.” He understands supply and demand. He understands that Jersey City produced 13 homes per 1,000 residents in 2024, twice as many apartments as all of Queens, and that this production is why the city remains livable for working families. He is pursuing policies that will reduce future building anyway, and he knows it.

Within six days of taking office, Solomon signed an executive order auditing all 100+ active tax abatements in Jersey City, framed against a $250 million budget deficit he inherited. He campaigned on raising the inclusionary zoning mandate to 20% affordable units in all major new developments. He voted against every luxury tax abatement brought before him on the council, including what he called “the Pompidou, a $150 million giveaway to the Kushner family”(Kushner Companies, Jared Kushner’s family development firm, is a major Jersey City builder). The rhetoric is progressive, but he knows what the math does.

Jersey City’s current inclusionary zoning ordinance requires 10-15% affordable units. No project has yet broken ground at the 15% rate. Not one. Solomon wants 20%. An independent analysis by Better Blocks NJ, using the financials from an actual audited Jersey City project, calculated that making a universal inclusionary mandate work would require tax abatements for every covered project, costing roughly $184 million annually in foregone property tax revenue. The analysis concluded that the most probable outcome is “no affordable housing is built under the Solomon plan in the short run, while in the long run, housing development grinds to a halt.” Solomon has a Harvard public policy degree and can read a pro forma. He knows this.

The small-developer squeeze is where the cynicism becomes structural. Solomon’s proposals target “projects of 50 or more, 100 or more” units, with the exact threshold left deliberately vague. The PILOT audits, the inclusionary mandates, the compliance overhead: these are costs that Kushner Companies, with its legal department and government affairs team, can absorb. A small or mid-sized developer building their first 30-unit apartment project cannot. The large firm negotiates carve-outs and restructures deals around new requirements. The small firm looks at the regulatory uncertainty, the audit threat, the inclusionary math that doesn’t pencil, and builds in Hoboken instead.

Notice what is absent from Solomon’s housing platform: any mechanism to actually fund new housing production. No revolving construction loan fund like MassHousing’s $20 million ADU program. No state HFA acceleration fund like the ones Williams describes in Massachusetts, New York, and Michigan. No subordinate financing to unstick stalled projects. No streamlined permitting. His entire framework is extractive: audit existing abatements, mandate inclusionary units from developers, cap rent increases, collect development fees. Every policy takes from the production pipeline and none puts capital into it. For a man with a Harvard public policy degree who concedes that supply matters, that’s not an oversight.

The construction trades noticed. During the 2025 mayoral race, the Hudson County Building and Construction Trades Council endorsed Bill O’Dea, not Solomon. So did the Hudson County Central Labor Council, the Teamsters, and IATSE (the stagehands’ and technicians’ union). The workers who pour concrete and frame walls looked at Solomon’s platform and backed someone else. Solomon’s labor support came from SEIU 32BJ, which represents building serviceworkers (janitors, doormen, security guards), and the Working Families Party. The distinction matters. The people who clean finished buildings endorsed Solomon; the people who actually build them did not. They understood what his policies would do to their pipeline of work.

The effect is to consolidate Jersey City’s development market among the very “big developers” Solomon campaigns against, while squeezing out exactly the small and mid-sized builders who represent the farm system for housing production. The progressive posture produces a market that is more concentrated, not less. The man who rails against Kushner creates a regulatory environment that only Kushner can play in. We suspect this is the intended outcome, not merely the predictable one. Solomon is not stupid, and he is not confused about how development economics work. He told us so himself.

This matters for our argument because it illustrates the type of political risk that multifamily housing faces and that ADUs do not. Solomon doesn’t need to ban development or repeal zoning. He just needs to make the deal worse for every project in the queue. Higher inclusionary mandates. PILOT audits with a $250 million hole to fill. Vague thresholds that create uncertainty. Each lever individually sounds reasonable. Together, they freeze the pipeline.

ADUs are structurally immune to this. No abatement to audit, no PILOT to revoke, no inclusionary percentage to ratchet up. The homeowner’s HELOC doesn’t need the mayor’s approval. A city that chills its multifamily pipeline through regulatory accumulation cannot touch the ADU pipeline at all, because the ADU pipeline doesn’t run through city hall. It runs through kitchen tables, one self-interested homeowner at a time.

The irony, which Solomon knows perfectly well even if his supporters don’t: the homeowners building ADUs in Jersey City’s residential neighborhoods are adding housing supply in exactly the distributed, incremental, politically invisible way that his policies make impossible for apartment builders. If he succeeds in choking the multifamily pipeline, ADUs become not a complement to apartment production but its substitute. Less housing, produced less efficiently, at higher per-unit cost. ADUs because of ease of capital and distributed development are the sort of housing cockroaches we can learn from.

Where It’s Working, and Where It’s Going

Nationally, ADU permits hit 200,000+ annually, with 2.8 million cumulative permits tracked. Oregon (72% ADU-to-construction ratio), Washington (62%), Hawaii (54%) show California-consistent adoption. Massachusetts legalized statewide in 2024 and saw 1,600+ applications in year one, slower than California, but Governor Healey is pushing hard on design programs and financing. New York City’s “City of Yes” targets 80,000 ADUs, which would represent the largest single municipal ADU program in the country.

Austin: Where ADUs Meet Walkability

Austin is doing something nobody else has tried: layering ADU liberalization on top of a comprehensive density reform that already produced the steepest rent declines of any large U.S. city from 2021 to 2026.

The HOME Initiative allows three units per single-family lot, slashed minimum lot sizes to 1,800 square feet, and in year one, permits in single-family zones jumped 86%. Homeowners can build two ADUs per lot, on lots as small as 2,500 square feet, with zero parking. At Austin’s lower costs ($80,000-$300,000) and rental rates of $1,500-$3,000/month, the payback is faster than California.

But the genuinely interesting development isn’t housing. It’s commerce.

Accessory commercial units (ACUs) are the commercial cousin of the ADU: a small-scale business on a residential lot. A corner coffee kiosk. A backyard salon. A garage bike shop. The Congress for New Urbanism calls them a return to pre-Euclidean zoning, the streetcar suburb where a corner grocery was woven into the residential fabric before mid-century zoning laws banished retail to strip malls.

Ten ADUs on a block make a denser suburb. Ten ADUs plus three ACUs make a walkable neighborhood. The difference matters. And ACUs run on the same self-interest engine: the homeowner doesn’t open a kiosk because she read Jane Jacobs. She opens it because she can rent 400 square feet to a coffee roaster for $1,500 a month. The walkability is a byproduct. Each ACU makes nearby ADUs worth more (closer to amenities, higher rents). Each ADU makes nearby ACUs more viable (more customers walking distance). The whole thing is a self-reinforcing loop powered entirely by greed.

ACU startup costs can be as low as $10,000, an order of magnitude below conventional commercial rents of $20-$70/sqft. Austin, with its density reforms and transit-oriented overlays along the Project Connect light-rail corridor, is positioned to test ADU-plus-ACU neighborhood formation at scale. YIMBYs: push for ACU legalization alongside ADUs. The two together produce neighborhoods, not just bedrooms.

San Diego: When “ADUs” Had Elevators

San Diego tested what happens when you remove the ADU density cap entirely. The Bonus ADU Program (2020) allowed one market-rate bonus ADU for every affordable ADU. In Transit Priority Areas: no cap. Permits surged from under 500 in 2019 to over 2,700 in 2023. 20% of new homes built in the city were ADUs by 2024. Developers built 17-unit projects on single-family lots. Someone proposed 120 units in Pacific Beach. At a certain point you have to admire the audacity of calling a 120-unit development an “accessory dwelling.”

The council imposed caps in a 5-4 vote in June 2025. But three things matter:

One, the program proved the thesis stone cold. The Terner Center found developers chose Bonus ADUs over traditional multifamily because the pathway was faster, cheaper, and less encumbered. Rational actors picked the “expensive” unit type because total system cost was lower.

Two, the backlash shows you can’t shoehorn apartment-scale production into the ADU frame. ADUs get their political cover from being small, incremental, homeowner-controlled. At 17 units, that cover evaporates. Use the ADU pathway for what it does best (one to four units, distributed, homeowner-driven) and build a separate streamlined pathway for bigger projects. That’s Williams’s federal investment agenda.

Three, even after the rollback, San Diego remains more ADU-friendly than almost anywhere. The program overreached (not really, but you get the point), got slapped back, and still left the city with a permitting regime most metros would envy.

If a more straightforward approach fails, we may be able to take advantage of San Diego’s little experiment to help other cities “upgrade” their own ADU laws.

Trick’s Limits, & Why the Apartment Agenda Matters

ADU alchemy is real, it’s magical, but it has boundaries. 200,000 ADU permits a year won’t close a 3-4 million unit shortage alone. A third of ADU builders use them for family or personal space rather than rental housing. Production skews toward wealthy homeowners who can self-finance. You can’t build a 200-unit apartment with a HELOC.

Here is what the YIMBY movement needs to reckon with: the zoning fights are necessary, and they are not sufficient. We have spent a decade getting very good at legalizing housing. We have not spent nearly enough time figuring out how to fund the housing we’ve already legalized. The 750,000 units in the shadow pipeline, the 117,000 with permits and no shovels in the ground, the 25% drop in multifamily starts even as we kept winning zoning victories: these are not zoning failures. They are financing failures. We are legalizing units faster than the financial system can produce them.

This is the gap Williams’s Raising the Housing Investment Level fills, and I think every serious YIMBY should read it. It is the clearest articulation I’ve seen of what comes after the zoning fight. The proposals are not exotic: streamline FHA so it closes in 90 days instead of 360, match state revolving loan funds so stalled projects get subordinate financing, let the GSEs participate in construction lending, accelerate depreciation so the after-tax math improves for investors, redirect the FHLB system toward its original mission. Together they could add 230,000-355,000 units to annual multifamily production. Most cost the federal government little or nothing. Several require regulatory changes, not appropriations. And history says they work: the HUD boom of 1968-1973 produced 3 million units in four years; the tax relief boom of 1981-1986 produced 2.6 million in five. When the feds made apartments pencil, apartments got built.

For small developers and community banks, ADUs are the farm system. ADU lending is underexploited by community lenders. MassHousing authorized $20 million for ADU construction loans. The San Diego Housing Commission offers construction loans and technical assistance. A developer who delivers twenty ADU projects has the track record a community bank needs before underwriting a 30-unit apartment building. The on-ramp to the multifamily market runs through backyards. But if the multifamily market itself remains frozen because the financing system is broken, the on-ramp leads nowhere.

ADUs are housing cockroaches: they survive everything because of easy capital and distributed development. We should learn from them. But the lesson isn’t “just build ADUs.” The lesson is that capital access, having multiple developers on hand, and system simplicity produce housing, and their absence kills it. Apply that at the ADU scale and you get backyard cottages popping up. Apply it at the institutional scale, through the reforms Williams lays out, and you get the 500,000 multifamily units a year we need to actually end the shortage.

In June 2017, a small team from a Chinese biotech company that almost nobody in the West had heard of walked into a presentation hall at the American Society of Clinical Oncology annual meeting and showed their data. ASCO is the biggest oncology research conference in the world. The data were from an experimental CAR-T cell therapy for multiple myeloma. The response rates were near 100% in heavily pre-treated patients. The room noticed.

Within months, Johnson & Johnson’s Janssen subsidiary signed a worldwide collaboration, paying $350 million upfront. It was one of the largest payments ever made for a China-originated therapeutic asset. The company was Legend Biotech, a subsidiary of a gene synthesis company called GenScript, operating from limited space in Nanjing. The therapy would eventually become Carvykti, the first and only cell therapy to show a statistically significant overall survival benefit in multiple myeloma. By mid-2025, it was generating roughly $439 million per quarter.

That moment would have been nearly unthinkable five years earlier. It has since become routine. In 2024, China's innovative drug license-out transactions reached 94 deals totaling approximately $51.9 billion. In 2025, the numbers more than doubled: 157 cross-border outlicensing deals worth a record $135.7 billion. In the first quarter of 2026 alone, deal value hit $60 billion, a 73 percent jump from the same period a year earlier and nearly half of the entire 2025 total. GSK signed a $12 billion alliance with Hengrui. Novartis paid $5.36 billion for a deal with Argo Biopharmaceutical. AstraZeneca agreed to a weight-loss drug collaboration with CSPC Pharmaceutical worth up to $18.5 billion, with $1.2 billion upfront. The direction of pharmaceutical deal flow has reversed entirely. A decade ago, Chinese firms overwhelmingly licensed drugs in from the West. Now nearly half of transactions flow the other way.

But here is what the deal figures don’t capture: Unlike other aspects of Chinese industrial policy (like cosmetics), China’s medical industrial policy that produced this innovation may not be able to sustain it. China’s domestic venture investment in biotech peaked at $15.7 billion in 2021 and plunged to $4.2 billion in 2024, in spite of surging cross border deals. Only four biotech firms listed on Chinese exchanges that year, the lowest since 2008. The government’s own pricing regime compresses margins on innovative drugs so severely that every major Chinese pharma company now depends on Western licensing revenue to fund its R&D. And in December 2025, Congress passed the BIOSECURE Act, a law that restricts how U.S. pharmaceutical companies receiving federal funds can do business with designated Chinese biotechnology companies, putting new pressure on exactly the cross-border integration that makes the whole system work.

This is not a simple success story. It is a story about a system, built deliberately over decades, that moved a country from the periphery of pharmaceutical R&D toward the frontier, and that now faces contradictions rooted in its own design. Four companies illuminate different pathways through that system. A recent working paper from the National Bureau of Economic Research adds precision about which forces drove what. And the structural tensions reveal whether any of it can last.

Assembling the Machine

Before the companies, before the deal flow, before the insurance reform that everyone is now trying to understand, something more basic had to happen. China had to build the capacity to innovate. That took decades, and it was not an accident.

The most important infrastructure change was the August 2015 State Council policy document that restructured the Chinese drug regulatory system. Before 2015, the system was rigid, redundant, and slow. The reform overhauled the Drug Administration Law, introduced a new Vaccine Administration Law, and rewrote the core regulations. Then, in June 2017, China joined the International Council for Harmonisation, committing to align its regulatory standards with the FDA, EMA, and other leading agencies. The NMPA adopted all 71 ICH guidelines and compressed review timelines for innovative drug applications from roughly 14 months to approximately 30 working days. A reduction of more than 90%.

The physical infrastructure came alongside it. “Made in China 2025,” “Healthy China 2030,” the 14th Five-Year Plan, and the construction of more than 100 state-financed biotechnology parks built a physical R&D base at a scale few countries can match. Capital market reforms opened pathways for biotech firms to list and raise funds. The Hong Kong Stock Exchange’s Chapter 18A rules allowed pre-revenue biotech companies to list on the main board. Shanghai’s STAR Market attracted innovative drug makers seeking domestic investors.

Then there were the people. China’s biopharma boom has been driven substantially by returnee scientists: researchers trained at Western pharmaceutical companies and academic labs who went back to China, bringing expertise, networks, and institutional knowledge about how drugs actually get developed and approved. Government recruitment programs actively cultivated this talent pool. The sector now draws from 30,000 active AI researchers at the intersection of AI and drug discovery, and China produces roughly 5 million STEM graduates a year.

None of this was self-assembling. Regulatory reform, physical infrastructure, capital markets, human capital. Each was the product of deliberate state investment. This was industrial policy: the state building the preconditions for an industry to exist.

But supply-side capacity alone does not explain the timing. China’s scientific output and talent base grew steadily throughout the 2010s. The clinical trial surge didn’t. Something happened around 2016 that activated the capacity. To understand what, you first need to see how four companies rode the wave, and how each of them ran into the same wall.

The 30-year grind

In 1970, the Chinese government opened a small pharmaceutical factory in Lianyungang, a mid-tier coastal city in Jiangsu Province. For two decades, it produced pharmaceutical ingredients and generic copies of Western drugs. Exactly the kind of operation that confirmed the conventional wisdom: developing countries consume medicines invented elsewhere.

When Sun Piaoyang took over in the early 1990s, he had a theory. Generics were a commodity business with vanishing margins. The future was innovation. But you couldn’t jump to the frontier overnight. So Sun pursued what he called “secondary innovation”: me-too drugs that improved upon existing Western molecules. Critics dismissed it as generics with extra steps. Sun’s argument was different. Secondary innovation was scaffolding. It built the capabilities that genuine novelty would eventually require. The company listed on the Shanghai Stock Exchange in 2000 and began a long, deliberate pivot.

Three decades later, the scaffolding has produced a building. Hengrui’s R&D spending increased 33.8% year-over-year in 2024, with over 29% of revenue going to research. Innovative drug revenue rose 30.6% to roughly RMB 13.9 billion ($1.9 billion). In 2024, Hengrui overtook AstraZeneca as the most prolific clinical trial sponsor on earth. A state-owned factory in a city most Western pharma executives couldn’t find on a map, out-trialing a century-old British-Swedish multinational.

But Hengrui’s story also reveals the tension at the heart of this system. The government’s pricing regime subjects drugs to negotiated cuts of 50–63%. Those cuts compress domestic margins on innovative drugs. Hengrui’s response has been to license assets to the West: a $12 billion GSK alliance, three obesity drugs to a Bain Capital/RTW-backed startup for up to $6 billion. The company doesn’t license directly to Western pharma. It licenses into separately capitalized Western entities, receiving upfront payments while retaining equity. The structure de-risks geopolitical exposure. It has been widely copied. But it also means that Hengrui, the company that most fully embodies China’s pharmaceutical transformation, cannot sustain its R&D from its home market alone.

Luxury of Time: What patient capital makes possible

Drug development grinds against the clock of capital markets. Most VC-backed biotechs don’t have 15 years. HUTCHMED could.

The company was established in 2000 as a subsidiary of CK Hutchison Holdings, Li Ka-shing’s conglomerate. That parentage gave it patient capital and freedom from quarterly fundraising pressures. Fruquintinib, its flagship molecule, was discovered entirely in-house, a highly selective VEGFR inhibitor designed to deliver anti-tumor efficacy with fewer off-target effects than competitors. The development timeline is itself the point: well over a decade from discovery to NMPA approval in 2018, making it the first China-discovered novel oncology drug (for metastatic colorectal cancer) to receive full approval in China. Then the global FRESCO-2 trial across roughly 130 sites in 10 countries. FDA approval November 2023. European Commission approval June 2024. Projected global sales exceeding $1.5 billion.

HUTCHMED built a hub-and-spoke partnership model: discovery in-house, Eli Lilly for China distribution, AstraZeneca for savolitinib, Takeda for global commercialization. In 2025, it sold a 45% stake in its non-core drug distribution business for $608.5 million to concentrate on next-generation programs. But notice the same pattern. The revenue that sustains R&D runs through Western partners.

Raids, Breakthrough, crisis, and the fragility of partnership

Return to the ASCO moment. Now for the full story.

Legend grew from a small team inside GenScript Biotech, operating as the “Legend Project” starting in late 2014. The bet was specific: target BCMA in multiple myeloma using a novel dual-binding-domain CAR construct, leveraging GenScript’s genetic engineering expertise and China’s access to a large patient population. The 2017 ASCO presentation happened. The $350 million J&J deal followed. FDA approval came in 2022. NMPA approval in August 2024 based on a trial showing 87.9% overall response. Manufacturing at scale in Ghent, Belgium. Over 7,500 patients treated globally.

Then the crisis. In September 2020, three months after Legend’s $424 million Nasdaq IPO, Chinese customs raided GenScript and Legend offices. Founder Frank Zhang was detained and later arrested for suspected smuggling related to human genetic resources regulations. Legend’s stock plummeted more than 15%. The company survived because Ying Huang, a former Wall Street analyst with a Columbia Ph.D., stepped in and held the J&J partnership together.

Legend’s model exploited a division of labor: China for discovery and early clinical work, J&J for the rest. It reached the world faster and with less capital than vertical integration would have required. But it also exposed a kind of risk that Western biotech boards have never had to plan for, the political vulnerability of Chinese founder-led companies working at the intersection of genetic data, cross-border research, and state security.

A Swiss Passport and a New Name

The arc reaches its high point with the company that made the biggest gamble and won, and then had to reckon with what winning meant.

In 2010, John V. Oyler, an American serial entrepreneur, and Xiaodong Wang, a member of the U.S. National Academy of Sciences, set out to build a vertically integrated, globally competitive oncology company from China. Not a licensing shop. A company that would discover molecules, run global trials, build its own salesforce, and sell its own drugs in the United States and Europe. First-in-human study of zanubrutinib in Australia in 2014. Commercial infrastructure built across the U.S., Europe, and Asia, burning through cash for years. R&D at approximately 52% of revenue in 2024.

The moment that changed the narrative was the Phase 3 ALPINE trial, a direct head-to-head against J&J’s ibrutinib in CLL. A loss would have been devastating. Zanubrutinib won. Reduced the risk of progression or death by 75% vs. bendamustine-rituximab in treatment-naïve CLL. The first FDA approval of a new chemical entity discovered in China. From roughly $42 million in 2020 revenue to $2.6 billion in 2024. Approved in 70+ markets. 100,000+ patients. $30 billion market cap.

Then came the identity reckoning. In November 2024, the company announced it would rebrand as BeOne Medicines. It changed its Nasdaq ticker from BGNE to ONC. In May 2025, it redomiciled from the Cayman Islands to Basel, Switzerland, joining Roche and Novartis as a Swiss-domiciled pharma company. John Oyler called it a reflection of “who we are today as a leading global oncology company.”

That is one way to read it. Another: a $30 billion company that was built in China, on Chinese scientific talent, with Chinese regulatory pathways and Chinese patient data, concluded that its Chinese identity had become a liability. The BIOSECURE Act was advancing through Congress. U.S.–China tariff uncertainty was escalating. More than three-quarters of American biotech companies contract out services to Chinese firms, and the political environment around those relationships was shifting fast. BeOne’s redomiciliation was not a corporate address change. It was an acknowledgment that the most successful Chinese-origin pharmaceutical company in history believed its origin story was becoming structurally incompatible with its commercial future.

A system built to produce Chinese pharmaceutical innovation produced a company so successful that it needed to stop being visibly Chinese. That is the contradiction, and it is not incidental. It runs through everything.

The Same Great Wall

Hengrui, HUTCHMED, Legend, BeOne. Four very different strategies. All four drew on the supply-side preconditions that decades of industrial policy built. The reformed regulatory system. The talent pipelines. The research infrastructure. All four produced globally validated molecules: FDA approvals, head-to-head trial wins, tens of billions in licensing payments from the world’s most demanding buyers.

All four share a structural dependency. They need Western revenue. The Chinese domestic market, despite being the catalyst that activated their innovation, compresses the pricing needed to sustain it. That paradox, a market that simultaneously creates and constrains innovation, turns out to be the central mechanism of the story. The sharpest academic evidence available helps explain how it works.

Demand Side Policy is Industrial Policy

The supply-side preconditions grew steadily through the 2010s. More publications, more scientists, more infrastructure, year after year. But the clinical trial surge was not steady. Something happened around 2016 that converted latent capacity into an explosion of activity.

A working paper published this year by the National Bureau of Economic Research, “From Free Rider to Innovator: The Rise of China’s Drug Development,” by Panle Jia Barwick at the University of Wisconsin, Hongyuan Xia at Cornell, and Tianli Xia at the University of Rochester, provides the most comprehensive attempt so far to disentangle the competing explanations. The authors compiled a synchronized database spanning clinical trials, scientific publications, drug sales, and government policies across China, the U.S., and Europe from 2010 to 2024.

A few caveats before the findings. This is a working paper. It has not been peer-reviewed. The identification strategy is credible, an event-study framework with disease-by-country fixed effects, placebo tests, and multiple robustness checks, but the specific magnitudes will be debated, and the decomposition leaves a substantial unexplained residual. What follows should be read as the best available evidence on a specific question, not as settled fact.

The paper zeroes in on a specific policy: the National Reimbursement Drug List reform.

Before 2016, China had nearly universal health coverage, more than 95% of the population, but the national formulary prioritized low-cost generics and excluded innovative drugs. Patients paid full price out of pocket for novel therapies. The innovative drug market totaled just ¥66 billion ($10 billion) in 2015. A market of 1.4 billion people that was, for innovative drugs, effectively tiny.

Starting in 2016, the National Healthcare Security Administration launched centralized “price-for-volume” negotiations. Firms accept average price cuts of 50–60% in exchange for NRDL inclusion, which guarantees access to roughly 1.4 billion insured lives at low coinsurance rates. By 2024, 699 drugs had been included. Cumulative spending reached ¥460 billion ($67 billion) between 2016 and 2024. The inclusion criteria explicitly favor new molecular entities and reward clinical novelty and unmet needs.

The paper documents what happened to individual drugs after NRDL inclusion. Retail prices dropped roughly 50% (66% for oncology). But quantities sold increased 350% on average, and nearly ten-fold for cancer drugs. Net producer revenue increased about 100% on average and roughly 500% for oncology. Given low marginal production costs (the authors estimate about 18% of pre-negotiated price), these revenue gains flowed through to profits.

The timing evidence is clean. Benchmarking China against the U.S. at the disease-country-year level, the paper shows flat pre-trends before 2016 and a sharp divergence starting in 2016–2017. By 2024, China’s trial volume had expanded by 172% relative to the U.S. High-novelty trials increased 123%. And 88% of the post-2015 increase was driven by domestic firms, not multinationals relocating R&D.

The paper’s strongest finding may be the placebo test. Disease categories never exposed to NRDL expansion showed no systematic increase in trial activity after 2015. The diseases with the largest NRDL expansion, including non-small cell lung cancer, liver cancer, and renal cancer, saw the sharpest increases. The contrast is stark.

The decomposition exercise attempts to quantify relative contributions for oncology: 43% attributed to the NRDL reform, 24% to knowledge accumulation and talent flows, less than 1% to other government policies. About 32% remains unexplained.

Those numbers need careful reading. The less-than-1% figure for industrial policy does not mean industrial policy was unimportant. It means that within the paper’s framework, once you account for the NRDL, knowledge stocks, and talent, the additional explanatory power of counting policy documents is small. But the regulatory reforms, biotech parks, talent programs, and capital market infrastructure that industrial policy built over decades are preconditions baked into the base of the model. The NRDL would not have produced a clinical trial surge if no firms had been capable of running trials, no regulatory system capable of processing them, and no scientists capable of designing them.

The honest reading is that both were necessary. Industrial policy built the capacity over decades. The NRDL activated it by changing the expected returns. The paper’s contribution is to show that the demand-side market expansion was a larger factor than most observers had assumed, and that its design features (rewarding novelty, targeting unmet needs) shaped the direction of innovation, not just the volume. That is a genuinely important finding. It does not mean the supply side was optional.

One more result, and it may be the most consequential for anyone thinking about drug pricing. For 57 cancer drugs included in the NRDL, the short-term consumer surplus gain from expanded access is approximately ¥35.6 billion ($5.1 billion) per year. The long-run gain from roughly 60 additional oncology drugs induced by the reform is approximately ¥67.6 billion per year. If those estimates are broadly right (and they are one study’s estimates), the long-run gains from induced innovation are roughly three times the short-run gains from improved access.

The standard framing of drug pricing treats it as zero-sum: lower prices increase access but reduce innovation incentives. The NRDL evidence complicates that. Under specific conditions (massive population coverage, genuine monopsony power, inclusion criteria that reward novelty), a price-volume reform may be able to expand both access and innovation simultaneously. Whether that finding generalizes beyond China’s specific institutional context is a question the paper cannot answer. But the direction is suggestive enough that health system architects in other countries should be paying attention.

Strengths and Fragilities Are the Same Thing

So the system works. The supply-side capacity was built. The demand-side shock activated it. The companies produced globally validated drugs. The licensing revenue is flowing.

But the system’s contradictions are not footnotes. They are structural, and they are getting worse.

Start with the domestic funding crisis. The NRDL expanded the market for innovative drugs and activated the capacity that industrial policy built. It also compressed domestic margins. Average negotiated price cuts of 50–63% mean that Chinese companies cannot generate enough domestic revenue to fund globally competitive R&D from the home market alone. Meanwhile, China’s national priorities have shifted. Government subsidies that once flowed generously to life sciences are being redirected toward semiconductors and computing, leaving local venture capital firms with less support. Venture investment in Chinese biotech peaked at $15.7 billion in 2021 and collapsed to $4.2 billion by 2024. Cash-strapped Chinese biotechs are increasingly out-licensing assets not because they want to, but because they have to, selling molecules to Western buyers to stay solvent and fund their remaining programs. At JPMorgan Healthcare Conference 2025, industry observers reported that Eastern funds and biotechs arrived ready to sell, proactively showcasing portfolios, a reversal from previous years when deals had to be hunted.

This is not a healthy equilibrium. It is a fire sale dressed up as a licensing boom. The numbers look like acceleration, and they are. Deal value hit $135.7 billion in 2025 and is on pace to exceed that in 2026. UBS projects that China's innovative drug market will grow at an annualized pace of 20 percent between 2026 and 2030, with the broader pharmaceutical industry generating revenues of $2.1 trillion by the end of the decade. But those projections sit alongside a domestic venture market that has collapsed by 73 percent since its peak. The innovative drug market is growing. It is not growing fast enough to replace the funding sources that are disappearing. Every quarter of record-breaking outbound licensing is also a quarter in which Chinese biotechs are selling molecules to Western buyers because the domestic capital environment no longer supports carrying them to commercialization alone.

Then there is the geopolitical problem, which is not abstract. On December 18, 2025, President Trump signed the BIOSECURE Act into law as part of the FY2026 National Defense Authorization Act. The law restricts U.S. federal procurement and grants involving biotechnology products or services provided by “biotechnology companies of concern.” Any company on the Department of Defense’s 1260H list of Chinese military companies is automatically designated. A process led by the Office of Management and Budget will name additional companies, with annual updates.

The law’s reach extends deep into the pharmaceutical value chain. Its definition of “biotechnology equipment or services” covers sequencing tools, data-storage platforms, cloud-based bioinformatics, CDMO services, and certain reagents or diagnostic devices. Any U.S. pharma company that conducts federally funded research or manufactures products purchased by government programs would need to validate the provenance of its entire upstream infrastructure. Prominent contract research and manufacturing organizations like WuXi AppTec and WuXi Biologics are not currently on the 1260H list and are not immediately subject to the law’s restrictions. But more than three-quarters of American biotech companies contract out preclinical and clinical services to Chinese companies, and 30% depend on China-linked companies for manufacturing of approved medicines. The most significant consequence may not be the immediate prohibition but the gradual decoupling of federal biomedical programs from globally integrated supply chains.

BeOne’s redomiciliation to Basel. Hengrui’s NewCo structures. Legend’s J&J partnership. These are all corporate architectures designed to navigate decoupling. But they are navigating it from a position of dependency. The pharmaceutical supply chain between the U.S. and China is too integrated for clean separation. Chinese companies were 35% of announced pharma M&A targets in 2025. It is also too politically charged for business-as-usual. The result is an unstable middle ground where every deal, every trial, every licensing transaction now carries a geopolitical risk premium that did not exist five years ago.

And the pipeline itself faces concentration risk. Innovation remains heavily concentrated in oncology, where NRDL coverage went deepest. The PD-1/PD-L1 overcrowding, more than 80 molecules in China, is what demand-pull incentives look like when they cluster around a single validated pathway. China now accounts for 24% of the world’s first-in-class pipeline, and more than 40% of the pipeline is fast-follower or first-in-class. The composition is genuinely shifting. But data integrity remains a credibility issue. A peer-reviewed analysis of FDA inspections from 2016 to 2023 found no significant difference in clean inspection outcomes between Chinese and Western sites in multi-regional trials, which is reassuring. But transparency in disclosing inspection findings still lags, and perception matters for a system that depends entirely on Western regulatory acceptance.

Each of these problems traces back to how the system was designed. The price compression that drives the licensing boom is the flip side of the market expansion that activated innovation. The geopolitical exposure comes from the integration that commercializes the innovation. The concentration in oncology reflects the targeting that directed R&D toward unmet needs. You cannot have the strengths without the fragilities. They are the same thing.

The Factory in Lianyungang

In 1970, the Chinese government opened a pharmaceutical factory in a mid-tier coastal city. It made generic drugs. In 2024, that company overtook AstraZeneca as the most prolific clinical trial sponsor on earth.

Between those two dates, a system was assembled. Decades of industrial policy laid the regulatory infrastructure, the talent pipelines, the research parks, the capital markets. An insurance reform activated that system by changing the expected returns to innovative drug development. Four companies took four different paths through it and each produced globally validated molecules. The best available academic evidence suggests the long-run welfare gains from the induced innovation substantially exceed the short-run gains from expanded drug access.

But the system has a problem it has not solved. It produces world-class pharmaceutical innovation. It cannot pay for it at the prices needed to sustain the investment. Domestic venture funding has collapsed. Government priorities are shifting toward semiconductors. The companies that embody the transformation are licensing their assets to stay afloat, redomiciling to foreign jurisdictions to shed their identities, and navigating a law whose stated purpose is to decouple the supply chains they depend on.

The most successful Chinese-origin pharmaceutical company in history concluded that it needed a Swiss passport and a new name. The licensing boom that the deal tables celebrate is, in part, a funding crisis selling molecules at market. And the geopolitical environment on which this system’s commercial viability depends is tightening, not loosening.

Can the system that created the innovation learn to sustain it? That is the question, for the companies, for Beijing, and for every Western pharmaceutical executive trying to figure out whether the next check they write buys them a molecule or a geopolitical liability.

• Inverted incentives. Prevention is 6-to-14x cheaper, but savings take decades. Budget cycles run annual. Officials are going to defer.

• Remove/Reduce the ability to defer. Cities that implemented automated dispatch means a detected pothole triggers a repair order without a human deciding whether to act.

• Institutions compound, mayors don’t. NYC’s system over the past 20 years got better across four administrations because each inherited what the last built. The machinery survives the election.

On March 15, 2026, a Saturday, more than 80 New York City Department of Transportation crews fanned out across all five boroughs starting at 6 a.m. and filled 7,200 potholes before the day was over. That is roughly a full week’s worth of repairs compressed into a single shift. A second five-borough blitz followed the next weekend. By mid-March, the city had repaired over 66,000 potholes since Mayor Zohran Mamdani took office on January 1, maintaining an average response time of about two days. And they were doing it in the worst pothole season in recent memory: 311 complaints were running roughly 33 percent above the same period in 2025, driven by what Commissioner Flynn described as an extended cold spell that caused potholes to form “all at the same time” once conditions finally thawed.

The numbers are impressive. But the numbers are not the point (in this case). The point is the bet Mamdani is placing; one most American mayors never would (FOR SOME REASON).

The conventional (and cliché) political playbook says mayors earn capital from new things. Bridge openings. Park dedications. Groundbreakings with hard hats and oversized scissors. Economists have a name for this tendency: the “ribbon-cutting bias,” describing the documented preference of political systems for new construction over maintenance. The logic is straightforward. Maintaining existing infrastructure is invisible, unglamorous, and generates no ceremony. Building something new gets your name on a plaque.

Mamdani is running the opposite experiment. His framing, “no problem too big, no task too small, no pothole too deep,” is not subtle, and it does not need to be. Eighty crews at dawn is the ceremony. The Saturday deployment schedule guarantees local news coverage, or at least crew-members’ TikTok videos. The real-time 311 dashboard is that oversized ribbon. The pothole count is a metric every voter understands. He is making maintenance visible, making it fast, and betting that mundane competence generates more durable political loyalty than any groundbreaking ever could.

And if that bet pays off in New York City, the implication for every other American city is uncomfortable. If Mamdani’s DOT can hold a two-day average response time under historically bad conditions, the question for every other mayor is pointed: what’s your excuse? Can you beat the city that can’t get a building permit processed in under six months?

I think the answer, for most cities, is that they could. They just haven’t.

Before we explain why, a few things to note. A 2015 audit by the New York City Comptroller found that DOT’s tracking system contained duplicate and triplicate entries that inflated reported repair counts. Whether those data quality problems have since been corrected is not established in available documentation. Every pothole figure in this article, from every administration, should be read with that finding in mind. Beyond the counting question, more than a quarter of pothole service requests were still listed as open, pending, or in progress at the time of a recent review, and Queens alone accounts for nearly half of the city’s 11,300-plus complaints. The two-day average masks a distribution with real tails. And on March 9, Jaikarran Seenarian died after his electric scooter struck a pothole on Liberty Avenue in Richmond Hill, Queens. Like always, be very wary of aggregate statistics, they do not capture the distribution of failures. A single unaddressed pothole has killed.

His roughly two-day average sits slightly above the Adams administration’s benchmark of 1.8 days, but maintaining that level of performance under a 33 percent demand increase is arguably the harder achievement from an ops standpoint, especially how nasty the winter storms were.

The caveats are real. The improvement is also real. And the question this article sets out to answer is not simply “why are there potholes?” but something harder: why do cities that know potholes are coming, know how to prevent them, and have the technology to fix them still fail, year after year, to get the problem under control? The answer is institutional. And NYC’s trajectory from unmeasured chaos to two-day response is the proof.

But first: how big is the problem we are actually talking about?

How Big Is This (Pot)hole?

Pothole damage costs American drivers more than the GDP of Iceland. A 2022 AAA survey put the figure at $26.5 billion in vehicle repairs in 2021 alone, roughly $600 per affected driver, with one in ten drivers sustaining damage severe enough to require a mechanic. Earlier AAA estimates from a multi-year survey had put the figure at around $3 billion annually. The nearly ninefold jump reflects both worsening road conditions and the rising cost of vehicle repairs, but even the lower estimate understates the problem, because it counts only the direct cost to drivers and ignores the costs borne by municipalities, transit systems, and taxpayers.

In New York City alone, potholes and road defects cost the city nearly $138 million in legal settlements for pedestrian injuries and vehicle damage over a six-year period, on top of an additional $7.3 million the city allocated to DOT in a single winter to address an extraordinary volume of repairs. Multiply that pattern across thousands of American municipalities and the total dwarfs any individual city’s budget.

The 2025 ASCE Report Card for America’s Infrastructure gave roads a D grade. As the Report Card documents, 39 percent of major U.S. roads remain in poor or mediocre condition, an improvement from 43 percent in 2020 but still a failing grade by any standard that takes driving seriously. Poor road conditions cost the average driver an estimated $1,400 per year in vehicle damage and time lost to delays.

But the backlog is worse than the current snapshot suggests. The Pew Charitable Trusts, in a careful 2025 analysis of Bureau of Economic Analysis data, found that state and local governments have accumulated a deferred maintenance liability of nearly $105 billion on roads and bridges since 2004. That number represents the cumulative gap between what governments spent on preservation and the value of annual depreciation. Even with nominally increasing dollar figures, real, inflation-adjusted road investment actually declined over the first two decades of this century. ASCE estimates that an additional $2.9 trillion is needed across 11 infrastructure sectors to reach a state of good repair, with a $684 billion gap specific to roads over the coming decade.

And the damage is not distributed evenly. Research from Mavis Tire finds that the Northeast experiences 60 percent more pothole-related tire repairs per location than shops in the South and West. The reason is physics: freeze-thaw cycles are the primary mechanism by which potholes form. Water penetrates pavement cracks, freezes, expands, and then thaws, progressively destroying the pavement structure from within. The Northeast and upper Midwest are structurally disadvantaged by their climate, and no technology fully overcomes that.

Most people’s response (especially suburban voters with pavement princesses) to numbers like these is to say the country needs to spend more on roads. That is true, to a point. It’s more about the distribution of the money being spent, rather than spending a total amount.

The Money Problem Is Real. Also Not Real.

The funding argument is correct as far as it goes. The most structurally important fact about American road funding is one that almost never appears in local news coverage of potholes: the federal gasoline tax has been frozen at 18.4 cents per gallon since October 1993, and it is not indexed to inflation. According to the Peter G. Peterson Foundation, that 18.4 cents is now worth roughly 45 percent less in real terms than when it was set. Wikipedia’s fuel tax tracking datashows that inflation has risen more than 122 percent since 1993 with the rate unchanged, and construction costs for roads have risen faster still. The net result, as Transportation for America documents, is that annual gas tax revenue declined by roughly $10 billion per year in real terms from 2010 to 2025. Congress has been patching the Highway Trust Fund with general revenue transfers, $53 billion over a recent five-year period according to Tax Notes, rather than raising the underlying tax.

Why does no one raise it? Before the whole breakdown in the Persian Gulf, increasing a visible per-gallon tax is politically costly in a way that deferred road maintenance is not, at least not until the pothole blows out a tire and the driver calls 311. The result is a revenue system built on a 1993 assumption that has decayed in real terms by more than half, funding a network of roads that costs multiples more to maintain than it did three decades ago.

The Infrastructure Investment and Jobs Act, signed in November 2021, provides approximately $500 billion for roads and bridge programs over five years, a 34 percent increase in surface transportation funding. That is substantial. It is also, given the scale of deferred maintenance, not sufficient to close the gap. And the IIJA is set to expire in September 2026, with reauthorization most likely going to be a no.

All of that is true. But even the money we do have goes to the wrong places.

The Congressional Budget Office has documented this: in 2022, 96 percent of the $52 billion the federal government spent on highways went to capital investment, meaning new construction and major rehabilitation of existing structures, rather than to routine operations and maintenance. That allocation has been typical since the 1950s. The federal government builds. States and localities maintain. And CBO’s reauthorization analysis makes the structural mismatch explicit: federal highway funds are restricted to “federal-aid highways,” the Interstate system and other major roads, leaving most local streets and roads entirely outside the program. The neighborhood streets where potholes cause daily misery are, by design, ineligible for the bulk of federal road dollars. Local roads account for 99 percent of road length in the United States and carry the majority of pothole complaints, yet they are maintained by counties and municipalities from their own general fund revenues, with no systematic federal support.

Even if Congress doubled highway funding tomorrow, most of the increase would flow to state DOTs for state highways. The potholes on your street would remain your city’s problem, funded from your city’s budget.

But spending does not reliably predict outcomes. High-spending cities have persistently terrible pothole records. Lower-spending cities with better-designed systems sometimes outperform them. If money were the binding constraint, that could not happen. Something else is going on.

The way American cities maintain roads is shaped by a set of interlocking incentive problems that make individually rational people choose collectively expensive, irrational outcomes, repeatedly, predictably, and at every level of government. This makes road spending one of the few places where we can cut budgets, and quality can improve. Reducing the amount of roads built, redirecting the existing budgets towards deferred maintenance, or even making roads more manageable by shrinking them.

Why Cities Keep Choosing the Most Expensive Option on the Menu

Consider the math. The Federal Highway Administration has documented that every dollar spent on preventive pavement preservation can save up to six dollars in future reconstruction costs. Michigan DOT research found that proactive maintenance is 14 times cheaper than reactive maintenance, and estimated that the state could save $700 million over eight years through systematic preventive intervention. A widely cited 1996 FHWA report puts preventive treatments at four to five times more cost-effective than allowing pavement to deteriorate to the point of reconstruction. Research summarized by the IMF and FREE Network finds that one dollar of deferred maintenance generates over four dollars in future repair costs, with some industry estimates running as high as seven to one over five years.

The physics compound the economics. Pavement deterioration is non-linear. It begins with cracks. Water (in addition to making new cracks) infiltrates the old cracks, freezes, expands, and destabilizes the road’s base layer. Traffic then removes the compromised asphalt in chunks. Each stage accelerates the next. A crack that costs pennies to seal becomes a pothole that costs dollars to patch and eventually a structural failure that costs thousands to reconstruct.

This isn’t new; most civil engineers will tell you this. That means every transportation agency in America is fully aware. And yet cities systematically choose cold-patch repairs, which last approximately one year, over permanent repairs that require excavating and replacing the base material. The same pothole gets filled three or four times before it is finally reconstructed at five to seven times the total cost of a single proper repair. The math is unambiguous. The behavior is economically irrational. So why does it persist?

Because the people who pay for prevention are not the people who benefit from it. Preventive treatments yield savings over a 10-to-20-year horizon. Municipal budget cycles run annual. The official who authorizes a $50,000 crack-sealing program today will not be in office, and may not even be in the same job, when the savings materialize. The official who lets roads deteriorate and then funds a high-visibility pothole-filling blitz gets to take credit immediately, on camera, with a crew and a truck behind them. The incentive structure is inverted. The economically rational choice and the politically rational choice point in opposite directions, and politics wins every time.

The ribbon-cutting bias runs deeper than preference for novelty. The IMF and FREE Network research identifies four compounding reasons that governments systematically underinvest in maintenance: political economy, meaning officials prefer projects that generate visible credit; fiscal, meaning maintenance budgets are cut first in lean years; institutional, meaning capital and operating budgets are managed separately (with FHWA’s own guidance actually drawing a sharp bureaucratic distinction between “preventive” and “corrective” maintenance that reinforces the silo); and informational, meaning asset condition data is often simply unavailable. The Economic Policy Institute adds a further wrinkle: because most states operate under balanced budget requirements, infrastructure maintenance is discretionary spending that gets squeezed during recessions, exactly when potholes proliferate most as neglected roads face winter freeze-thaw cycles and revenue-starved cities defer even routine crack-sealing. Maintenance spending is procyclical. Conditions are countercyclical. The system is designed to fail at the worst possible moment.

Even if these incentive problems were fixed, the underlying math of American development patterns would still make the task harder every year. Since World War II, American municipalities have systematically extended road infrastructure to accommodate new development, accepting near-term tax revenue in exchange for long-term maintenance obligations that the new revenue stream rarely covers. Strong Towns, the nonprofit led by Charles Marohn, has built an influential body of work arguing that this pattern constitutes a fiscal Ponzi scheme: cities use revenues from new growth to service the maintenance obligations of previous growth, requiring ever-more growth to remain solvent. The “Ponzi” framing is contested — Market Urbanism and Better Cities Project note that the empirical evidence is uneven, and City Journal has found the analysis broadly credible while questioning the label. But the core insight, that municipalities routinely assume maintenance liabilities well in excess of the tax revenue new development generates, is well-grounded. The practical consequence: total lane miles grow; maintenance budgets don’t. Consider Tulare, California, which told the state legislature in 2017 that, forced to prioritize arterial and collector roads due to funding shortfalls, its residential street network had declined from a Pavement Condition Index of 61 in 2010 to 53 in 2016, and the cost to restore it was growing exponentially with each year of deferral.

Fragmentation compounds everything else. On any given urban road, responsibility for the pavement may be shared among city, county, state, and private utility companies. The Federal Highway Administration’s manual on pavement utility cuts documents how utility excavations, in Washington, D.C., alone more than 5,000 per year at peak, structurally weaken adjacent pavement and dramatically shorten road lifespan. Utility companies are typically required to patch their cuts temporarily, but permanent restoration is often delayed years due to funding gaps. When the resulting pothole appears, it is frequently unclear which agency is responsible, and each agency has an interest in pointing to the other. Cities like Cleveland explicitly disclaim responsibility for utility-cut repairs, referring drivers back to the contractor. The driver, meanwhile, does not care whose jurisdiction the hole is in. They want the hole filled.

Former New York City DOT deputy commissioner Lou Riccio, speaking to Crain’s New York Business, captured the structural problem with characteristic directness: “New York City streets, in particular, get cut so often it’s like continuous open-heart surgery.” Every utility repair, every underground pipe, every cable installation reopens the street and disrupts freshly laid asphalt. Riccio estimated that the city needs at least 1,000 lane miles of resurfacing per year just to maintain current quality, and more than 1,200 to actually improve conditions.

The result is a system where no one bears the cost of what they defer. Federal policymakers who set gas tax rates do not drive on the local roads that deteriorate when revenues fail. State legislators who cut transportation budgets during recessions are not accountable for the potholes that appear three winters later. City officials who defer crack-sealing this fiscal year have moved on by the time the road requires full reconstruction. Utility companies that cut pavements are not on the hook for the premature road failure that follows. The FHWA and the Pew Charitable Trusts are right that better data and asset management plans would help. But the deeper problem is not informational. It is incentive-structural. Until that changes, the potholes will persist.

So what does it look like when a city breaks this pattern? The answer has less to do with technology than you might expect, and more to do with closing the loop between knowing about a problem and doing something about it.

The Cities That Broke the Pattern

In Chicago (to most people’s surprise), the Department of Transportation filled approximately 143,000 potholes in the first two months of 2024. February alone: over 93,000 potholes, a 63 percent increase over February 2023 and a 100 percent increase over February 2022. But the volume is not what makes Chicago interesting. What makes Chicago interesting is the systems design behind the volume.

The city’s 311 portal feeds directly into the open data platform at data.cityofchicago.org, allowing both internal management tracking and public transparency. Service requests are generally completed within seven days. The design choice that matters most: when a crew arrives to repair a 311-reported pothole, they do not fill that one hole and leave. They fill every pothole on the block. This sounds obvious. It is not. Most municipal repair systems dispatch crews to specific complaints. Chicago’s system dispatches them to streets, which means each complaint triggers a sweep that catches unreported defects. It is a force-multiplier that prevents the whack-a-mole dynamic of returning repeatedly to the same location. And because progress is publicly visible on a real-time dashboard, elected officials have a political incentive to maintain performance, not merely to claim it. Transparency creates accountability.

Pittsburgh illustrates a different facet of the same principle: that data quality itself can be a form of governance. Through the Bloomberg Philanthropies What Works Cities initiative, the city’s Department of Innovation and Performance built “Dashburgh,” a public data dashboard presenting 311 data broken down by neighborhood, providing granular accountability the city had never had before. But the more consequential innovation was internal. Pittsburgh’s Department of Public Works built analytical tools that automatically link multiple requests for the same pothole into a single ticket, flag requests that lack precise location details, and, most importantly, identify street segments that have been patched three or more times in two months. That last flag is the key. A street segment patched three times in eight weeks is not a street that needs patching. It is a street whose underlying pavement needs rehabilitation. The data system recognizes the pattern that human decision-making, under the pressure of daily complaint volume, reliably misses. This is the shift from reactive to systemic maintenance in microcosm: using information to detect when a short-term fix is producing long-run waste.

In Memphis, the breakthrough came from an unlikely place: the city’s transit buses. Working with Google Cloud partner SpringML, Memphis equipped its buses, which were already running cameras for traffic monitoring, with an AI system that analyzes road footage to detect and classify potholes. The model was trained on existing images, refined with data from higher-quality cameras, and then integrated directly with the city’s 311 work-order system so that confirmed potholes automatically generate a repair ticket. No human triage step. No 311 call required. The results: a 75 percent increase in potholes detected compared to prior manual methods, an estimated cost saving of approximately $20,000 per year, and 63,000 potholes repaired in one year. The lesson from Memphis is simple: use what you already have. The city did not buy a specialized detection fleet or hire a team of data scientists. It leveraged buses already running existing routes, cameras already in place, and cloud tools developed for other purposes. The marginal cost of detection approached zero.

Tarrant County, Texas pushed the detection-to-dispatch connection even further. The county integrated reports from Waze, the navigation application with over 140 million users that already allows drivers to flag road hazards, directly into its work-order system. A Waze pothole report auto-generates a maintenance dispatch with no human triage step. No 311 operator, no supervisor review, no scheduling queue. The bottleneck that plagues most municipal repair systems, the chain of handoffs between someone knowing about a pothole and someone doing something about it, is eliminated entirely.

And in Kansas City, Missouri, a pilot program tried to go one step further: not just closing the loop between detection and repair, but getting ahead of the problem entirely. The program combines traffic camera data, weather data, traffic volume, and pavement condition assessments to predict where potholes will form before they appear, with the goal of applying preventive sealant while the road is still intact. Chief Innovation Officer Bob Bennett reported early successes, with fewer potholes reported the following spring, though the sample was small and the causal mechanism difficult to isolate from weather variation. The program comes closest to addressing the root cause: spending money today where data suggest it is needed, rather than where the public happened to complain yesterday.

Bennett also happens to have the best one-liner in the pothole policy literature. Telling Governing magazine: “If we fail to fill the potholes or pick up the trash, we’re going to hear about it. Potholes are one of those things people kvetch about.” He is right that potholes generate immediate, visceral political pressure. The problem is that the political energy they generate goes to reactive speed, not systemic prevention. People kvetch about the pothole in front of their house. They do not kvetch about the crack-sealing program that was cut from the budget eighteen months ago.

What connects Chicago, Pittsburgh, Memphis, Tarrant County, and Kansas City? It is not any particular technology. Chicago uses spray injection trucks. Memphis uses machine learning. Tarrant County uses Waze. Pittsburgh uses internal analytics. Kansas City uses predictive modeling. The technologies are different. The underlying design principle is the same: each city closed a feedback loop connecting detection to dispatch to performance measurement to public visibility. Each removed the discretion and the delay that normally sit between knowing there is a problem and acting on it.

The absence is equally telling. Pennsylvania DOT’s AI road survey program, which partnered with RoadBotics to survey over 2,500 miles using computer vision, found that standardized defect scoring improved regional coordination and reduced inspection costs. But translating condition data into repair schedules required sustained institutional commitment that not every participating municipality provided. Better cameras produced better-documented backlogs unless someone rewired the institution between the camera and the crew.

Technology transfers. Institutions do not.

When did New York start to do something right?

Considering how people feel about the city, this is a bit of a shock. Yet, no city proves the compounding power of institutional investment more completely than New York (in this case), and no city’s story makes clearer that the improvement belongs to the institution, not to any single occupant of the mayor’s office. What Mamdani’s crews did on that Saturday in March 2026 was possible because of decisions made in 2007, 2010, 2014, and 2022 by people who could not have known they were building a system together.

Start at the floor. A 2002 audit by the New York City Comptroller found that DOT lacked any useful operational standard for pothole repair. The agency’s informal target was merely to close 65 percent of repair orders within 30 days. The auditors noted that the 30-day threshold was set arbitrarily, with no operational rationale. DOT could not explain why 30 days rather than 15, the number that actually mattered, since NYC Administrative Code Section 7-201(c)(2)shields the city from civil liability only if potholes are repaired within 15 days of complaint. The tracking system was unreliable. Repair counts were inflated by duplicates and triplicates. There were no written policies or procedures governing the work. This was the institutional floor: roughly 100,000 potholes filled per year, by a count that could not be trusted, on a timeline the agency could not reliably measure.

Michael Bloomberg’s contribution to pothole management was primarily infrastructural rather than operational, and it is the kind of contribution that gets no credit at the time and turns out to be indispensable later. Three investments from his tenure shape everything that followed. First, the city had long operated one municipal asphalt plant at Hamilton Avenue in Brooklyn. Bloomberg opened a second plant in Queens in 2010, giving DOT the materials supply, and the independence from outside vendors, that makes large-scale blitz operations financially and logistically viable. NYC DOT is one of the rare urban agencies in the country that produces its own asphalt, and that structural asset remains central to every subsequent administration’s pothole story, including Mamdani’s. Second, in 2007 and 2008, Bloomberg launched the Street Conditions Observation Unit, or SCOUT, a team of inspectors deployed on three-wheeled scooters to traverse every one of the city’s 6,000 miles of streets at least once per month, reporting conditions via GPS-enabled devices linked directly to 311. As Gothamist reported at the time, Bloomberg’s framing was explicitly about shifting from reactive to proactive: “It’s government’s responsibility to find the problems and fix them, not to sit there and say ‘Duh we didn’t know.’” Third, Bloomberg’s DOT under Commissioner Janette Sadik-Khan launched “The Daily Pothole”, a public-facing blog tracking repair efforts in real time with maps and crew counts. Trivial by today’s standards. Significant at the time. It created a transparency norm around pothole performance that each subsequent administration has had to honor, escalate, or outrun.

And yet, for all of this institutional investment, the operational numbers barely moved during the Bloomberg years. A data science analysis by researchers at Lehman College examining 311 records found a mean repair time of approximately 4.7 days with a standard deviation of seven days, meaning the distribution had a very long right tail. The Adams administration’s own retrospective benchmarks Bloomberg at approximately 4.4 days average. By fiscal year 2015, a second Comptroller audit found the average at 5.6 days on local streets, with nearly 10 percent of complaints unresolved past the 15-day legal liability threshold. The infrastructure was better. The operations had not yet caught up.

Bill de Blasio came into office in a brutal pothole winter. By his own DOT’s count, 113,131 potholes had been patched by early 2014 before his comprehensive plan was even announced. But the move that mattered was not reactive. It was strategic. De Blasio shifted the city’s fundamental logic from patching potholes after they appeared to preventing them from forming in the first place. The centerpiece was a $1.6 billion commitment over ten years to resurface New York City’s streets. The theory of change was explicit: a freshly resurfaced street is less likely to develop potholes, meaning fewer complaints, fewer reactive repairs, and lower long-run costs. This is exactly the prevention-over-reaction economics that FHWA had documented for decades, applied at city scale for the first time.

It worked. Between 2014 and 2018, the de Blasio administration repaved 5,000 lane miles and drove pothole counts down by 44 percent. By the end of his tenure, DOT had resurfaced 8,356 lane miles total and filled more than 2 million potholes. He also invested in field operations technology: crews received tablets for one-touch, real-time tracking of work, replacing paper-based reporting with digital logistics that could route crews more efficiently. DOT introduced a warm-weather asphalt mix requiring less heat, reducing energy use and extending the paving season closer to year-round. Average response time fell from Bloomberg’s 4.4 days to approximately 3.4 days (though official 2021 releasesreport 3.83 days, suggesting the improvement concentrated in later years). But the 44 percent reduction in pothole formation arguably mattered more than the response-time improvement. Fewer holes means less demand on the repair system even at constant operational capacity.

Eric Adams inherited the prevention framework and operationalized it. By January 2025, his administration claimed an average response time of approximately 1.8 days, nearly a full day faster than de Blasio and more than twice as fast as Bloomberg. But the more revealing number was demand: 88,466 pothole complaints via 311 over the first three years of his administration, compared to 147,640 at the same point under de Blasio, a 40 percent reduction. The administration resurfaced an average of nearly 1,200 lane miles per year, right at the expert-estimated threshold Riccio identified, and consistently deployed 25 to 75 crews during pothole season. When Adams framed his 500,000th pothole milestone as filled “in half the time it used to take” because “proactive paving of 1,200 miles of lanes” cut pothole formation, that was the correct analytical framing. Improvement came from two compounding forces: faster dispatch and lower demand. Both were improving simultaneously.

Mamdani inherited a better system under harder conditions. Approximately two-day average response under 33 percent elevated demand. Over 66,000 potholes through mid-March. The blitz format itself, 80-plus crews in coordinated Saturday deployments, is not new; DOT’s own 2021 press releases from the de Blasio era describe the identical model. What Mamdani adds is not operational novelty. It is the political insight that maintenance, made visible and fast, can do the work that ribbon-cuttings do. Executive attention flows resources and accountability to operations that might otherwise drift. Whether that attention converts into structural improvement, specifically the 1,150 lane-miles resurfacing commitment for 2026 that represents the preventive investment rather than the reactive repair, is the unanswered question. And whether the response time can reach same-day is the unanswered dare.

Most of what drives these numbers is (again) institutional, not mayoral. Asphalt plants, crew systems, 311 integration, and resurfacing programs were built over decades and persist across administrations regardless of who occupies City Hall. The most important long-run metric is not response time but pothole formation rate, which is a function of resurfacing investment, utility excavation coordination, and pavement quality. Riccio’s description of “continuous open-heart surgery” captures a structural vulnerability that no blitz weekend can permanently fix. And all of this data is self-reported by agencies with strong incentives to report improvement. The one time an independent auditor looked carefully, in 2015, the numbers were inflated. We do not know if that problem has been corrected. The evidentiary gap we would most like to close is independent verification of DOT pothole counts, along with borough-level and complaint-source breakdowns that would let us assess the distribution of the response-time average, not just its mean.

What Works, What Is Coming, and the Gap That Is Not Technical

Every success story so far shares a principle: institutional design matters more than technology choice. But the technology is changing, some of it is promising, and understanding what is deployable now versus what remains speculative matters. The persistent question is whether any of it changes the institutional calculus or merely produces more polished documentation of the same dysfunction.

The most commercially mature technology is computer-vision road scanning. The basic model is straightforward: a smartphone or dash-mounted camera captures imagery at roughly 10-foot intervals as normal fleet vehicles run their routes. Images are GPS-tagged, uploaded to a cloud platform, and processed by neural networks that classify defects across a spectrum from hairline cracks to full-depth potholes, producing a color-coded condition map that feeds directly into maintenance management software. Vialytics, a German-American platform, now supports over 600 municipalities and recognizes 15 damage classes; case studies from municipalities including North Tonawanda and Metuchen document crews shifting from reactive patching to planned maintenance after a single survey drive. RoadBotics, now a Michelin company, partnered with the Pennsylvania Department of Transportation to survey more than 2,500 miles using AI-powered analysis, allowing 15-plus municipalities to adopt standardized defect scoring and compare conditions on a common scale. The program was expanded following strong results.

On the repair side, two methods significantly outperform the traditional throw-and-roll approach that remains standard in most cities. Infrared pothole repair uses infrared panels to heat existing asphalt to a workable state without burning it, allowing new hot-mix material to blend seamlessly with the old surface. The result is a seamless patch with no cold joints, eliminating the primary failure mode of conventional patching: water infiltration through seam edges. Repairs take under 20 minutes, and the recycling of existing asphalt reduces both material costs and environmental impact. Ohio DOT research comparing three patching methods found that infrared patches had significantly longer expected lifespans than throw-and-roll or spray injection, and proved more cost-effective for winter patching specifically. The limitation is not technical but organizational: the equipment requires upfront investment, the method demands trained operators, and it cannot address full-depth failures. It is most powerful as a preventive strategy, applied to developing cracks before they become potholes. Spray injection patching, which blasts debris with compressed air, injects asphalt emulsion, and covers with aggregate, is the workhorse for high-volume rapid response.

Better detection is valuable. Better repair methods are valuable. But both face the same institutional bottleneck. Does better detection actually produce faster repair, or does it produce better-documented backlogs? Memphis shows the connection working when detection is wired end-to-end into work-order dispatch. But Pennsylvania’s experience is the cautionary case: condition data improved coordination, but without sustained institutional commitment to translating data into schedules, the information sat unused in many participating municipalities. The technology is only as good as the institution it feeds into.

Further out, there is autonomous road repair. The ARRES PREVENT robot, developed by Robotiz3d in collaboration with the University of Liverpool and Hertfordshire County Council, passed its first live public road trial in Potters Bar, England, on March 6, 2024. The robot uses AI-powered imaging to detect cracks and road defects, then autonomously fills them with sealing material before they develop into potholes. Robotiz3d claims the technology could deliver a 90 percent cost reduction, a 70 percent increase in speed, and three times fewer CO2 emissions. These claims require heavy skepticism. They come from the developer, not from independent assessment. The trial was a single short demonstration, not a controlled evaluation. UK government funding amounted to just over 30,000 pounds. That said, the underlying concept, autonomous proactive crack sealing before pothole formation, directly addresses the root institutional failure we have been describing: the inability of reactive systems to intervene at the optimal moment in the pavement lifecycle. If the technology matures and costs fall, it could change the maintenance calculus. For planning purposes, it is not there yet.

A different approach comes from Professor Erik Schlangen’s laboratory at Delft University of Technology in the Netherlands: self-healing asphalt. The mix combines standard paving material with tiny steel wool fibers that make the asphalt conductive. When an induction machine passes over the road, the fibers heat up, melting the bitumen binder and allowing microscopic cracks to close and re-bond before they can propagate. The technology has been field-tested on a 400-meter stretch of the Dutch A58 motorway since 2010, and on 12 other Dutch roads by contractor Heijmans, including high-stress locations such as roundabouts and industrial estates. Laboratory research suggests road lifespan can be doubled if the treatment is applied approximately every four years. The material costs about 25 percent more upfront but is estimated to save the Netherlands up to 90 million euros per year if adopted nationwide. Delft’s team is also researching epoxy asphalt, which mixes epoxy resin with bitumen to create a surface less sensitive to temperature extremes, particularly relevant as climate change increases both peak summer temperatures and freeze-thaw frequency.

The limitations of self-healing asphalt matter, though. The process is not autonomous: the induction machine must be driven over the road on a schedule, which presupposes a systematic maintenance system capable of tracking when each segment was last treated. And the technology is most powerful on long-duration highway sections with infrequent disruption; urban roads that are excavated every few years for utility work gain little benefit from materials designed for 40-to-80-year lifespans. It has been in field trials for fifteen years and is still not standard practice even in the Netherlands. We would not bank on it for near-term planning.

Same Asphalt, Different Institutions

If American institutions are the problem, a natural question arises: do other countries actually do this better, or is road maintenance just inherently hard everywhere? Two countries offer useful answers.

Over 90 percent of Japanese road pavements are asphalt, the same material used in the United States. Japanese roads are notably smoother and better maintained, but not because of some exotic technology. Repair work is simply performed far more frequently, ensuring the road surface remains intact before deterioration can compound. Asphalt has a service life of roughly 10 years, but maintenance is applied on a rolling basis so that roads rarely reach the deteriorated state at which potholes form. Japan’s Ministry of Land, Infrastructure, Transport and Tourism mandates systematic periodic inspection, bridges and tunnels once every five years by close visual inspection, and applies the same preventive maintenance philosophy to pavement. Nighttime paving is common, allowing crews to work without traffic disruption and producing better compaction. Procurement standards enforce lifecycle cost thinking rather than lowest upfront bid. These practices are not secret. They are well-understood by American transportation engineers. What prevents their adoption is not knowledge but institutional structure: the fragmented governance of American local roads, the annual budget cycle, and the absence of performance-based procurement at the local level. MLIT documents that 30 percent of Japanese towns and 60 percent of villages have no civil engineering technicians in their workforce, and Japan still outperforms the United States on road maintenance, because the institutional system compensates for local capacity gaps in ways the American system does not.

The Netherlands tells a complementary story. Rijkswaterstaat, the national water and road authority, applies rigorous lifecycle cost analysis to pavement decisions and maintains strong data infrastructure for road condition monitoring. Dutch porous asphalt is designed to absorb both sound and water, reducing noise pollution and improving drainage, but its open structure makes it more vulnerable to raveling, which is why the Delft self-healing research is particularly relevant to Dutch conditions. And the Dutch case demonstrates the lesson most directly: even self-healing asphalt requires institutional infrastructure. The induction treatment every four years presupposes a systematic maintenance scheduling system capable of tracking when each road segment was last treated. Innovation without maintenance institutions behind it does not stick.

Neither country is doing anything technically exotic. They are just more institutionally disciplined: systematic preventive maintenance, funded predictably, implemented through performance-based procurement, and measured transparently. The United States has most of the technical tools needed to match their outcomes. What it lacks is the institutional structure to deploy those tools at scale and sustain them across election cycles.

The Feedback Loop Is the Fix

Which brings us back to Mamdani, and to what his administration actually proves, which is less than the press coverage suggests and more than his critics allow.

The Saturday blitzes are the reactive half of the equation; the half that matters less. Eighty crews filling 7,200 potholes in a day is operationally impressive and politically savvy, but reactive speed is not what separates good road systems from bad ones. Prevention is. The real test of the Mamdani administration is not whether it can fill potholes fast on camera. It is whether the 1,150 lane-miles resurfacing commitment for 2026 actually gets funded and executed; whether the preventive investment survives budget season, whether it compounds across the next administration the way de Blasio’s and Adams’s did. We do not know the answer yet. A man died from a Queens pothole during the blitz. A quarter of service requests were still open. The data has historically been inflated. The hypothesis that visible maintenance can generate political capital is interesting. We will see whenever Mamdani can prove it or not.

But if the famously dysfunctional government of New York City can build a two-day-average system over twenty years, the question for every other American city is a little uncomfortable and specific: what’s stopping you?

The honest answer, for most cities, is not that they lack the knowledge or the technology. That every incentive in the system (annual budgets, political time horizons, fragmented governance, frozen federal revenue) pushes against doing the thing that works. There are proven institutional models. The political economy is designed to prevent its adoption. Pretending otherwise is how you get another decade of cold-patching the same hole four times.

The cities that broke the pattern did not do it by wishing the incentives or the political economy away. They did it by building systems that made the right choice easier than the wrong one: automated dispatch that removes the discretion to delay, public dashboards that make inaction visible, data flags that catch repeat failures before they become structural collapse. They did not fix the politics. They routed around them.

That is the only honest prescription this article can offer. Not that your mayor should try harder (well, that mayor should; you still need political will at the end of the day). Not that the ribbon-cutting bias is a failure of imagination. It is a structural feature of American municipal governance, and it will outlast every mayor who tries to overcome it through sheer will. What does not have to outlast them is the institutional machinery (the asphalt plants, the 311 integration, the resurfacing commitments, the transparency norms) that makes the next mayor’s job easier whether they care about potholes or not.