Nobody Told Them They Were the Bank

Florida did it first and cheapest. In 1926, the land boom collapsed mid-transaction, leaving platted subdivisions where developers had already installed the streetlights along roads to houses nobody ever built. You can still find some of the grids in aerial photography. Lamp posts in the scrub, standing in perfect rows, illuminating nothing.

China's version is taller. That is the only genuinely new thing about it.

The easy version of that observation is wrong, though, and worth killing early. These four events do not share a mechanism. Spain and Ireland built completed-stock overhangs financed through banks — the developer borrowed, the bank lent, the buyer arrived afterward or never arrived at all. China built something structurally different, and the difference is the whole subject here. What repeats across all of them is not the machine. It is the output: where the losses finally come to rest, and who has the least ability to refuse delivery.

Florida is the closest ancestor and the one nobody cites. The binder boys of 1925 traded binders — small deposits securing an option on a lot — passing the paper hand to hand before a single deed changed hands, on subdivisions that existed as survey lines in a drawer. They were trading the future of an object. That is presale by another name, and it is the thing China industrialized.

i · the household was the construction loan

Most coverage treats what follows as a background detail. It is the whole story.

In China, the dominant way to buy a new apartment is 期房 — forward housing. You buy the unit before it is built. You put down the deposit, you sign the mortgage, and you begin making monthly payments on a structure that at that moment may be a hole in the ground. This is not a niche instrument. In 2021, completed units were 10.4% of new homes sold by floor space. Roughly nine in ten Chinese apartment purchases that year were purchases of a promise.

Read that as a financing structure rather than a purchase and it resolves immediately. The household is not a customer. The household is a lender.

Not an uncompensated one — that overstates it, and the overstatement is worth correcting, because the accurate version is worse. The buyer's return was real; it was simply never paid in cash. Forward units price below completed comparables, and the expected appreciation was the coupon. The household was lending at a rate payable entirely in the currency of a rising market. If prices rose, they were paid handsomely. If prices stopped, the loan paid nothing, and it still had no seniority, no covenants, no acceleration clause, and no seat at the restructuring table.

Nor was the loan unsecured on paper. This is the part that makes the whole system legible, and most accounts skip it. Chinese law required presale proceeds to sit in supervised accounts — usable only for construction of the specific project the buyer had bought into, not for debt service, not for anything else. The safeguard existed. It was written down. A cautious buyer who asked the obvious question got pointed at it.

Enforcement was weak, and it was weak in a specific direction. Developers routinely drew the money back out, and what they most often spent it on was land: acquiring the next parcel to launch the next presale, because the model runs on velocity. Presales on tower four don't finish tower three so much as they buy the site for tower five.

Follow that money one step further than the coverage usually does. The land was sold by local governments. The escrow accounts were supervised by local housing bureaus. Same municipality, both sides of the transaction. The office charged with keeping the buyer's money inside the buyer's building was an arm of the entity that received that money the moment it left.

This is why presale by itself explains nothing. The UK, Australia, and the Gulf all sell housing forward and none of them produce fifty-story skeletons in the hundreds. Presale is not the cause. Captured supervision is. The safeguard was real, and the supervisor had a balance-sheet reason to look away.

That is not a bug in the Chinese property system. That is the product.

So when Evergrande stopped paying — defaulting on its offshore dollar bonds in December 2021, carrying liabilities that the Congressional Research Service put north of $300 billion — the loss did not land where losses are supposed to land. It landed on the least sophisticated creditor in the stack, who was also the only creditor legally obligated to keep paying while receiving nothing.

The households understood this faster than the analysts did. By summer 2022, buyers across dozens of cities began simply refusing to service mortgages on unfinished units. It started with a few hundred residents in Jiangxi whose Evergrande towers had been stalled over a year. Within weeks it was tracked across hundreds of developments in most of the country's provinces, and suppliers to the developers began defaulting on their own bank loans in sympathy and self-preservation. It was one of the largest coordinated refusals-to-pay in modern financial history, and it was organized around a demand so modest it barely qualifies as political: finish the building I already bought.

Spain's version of this cruelty was legal rather than structural. Spanish mortgages carried recourse — hand back the keys and you still owed the balance. The asset could vanish; the debt was immortal. Different machine entirely. Same output: the obligation outlives the thing it was attached to, and the person holding it has no exit.

ii · everyone was long the same trade

The second layer down is where the pattern stops being about developers.

Chinese local governments do not have a broad property tax. What they have — what funded the roads, the metro lines, the schools, the salaries — is land. Municipalities sell long-term use rights to developers, and in 2021 those land transfer fees came to 8.7 trillion yuan, about 30% of total local government revenue. Add land-related taxes and it reaches 37%. Which means the local government's budget was a leveraged long position on rising land prices, held by the same entity responsible for regulating the market that set them — and, as established above, for guarding the escrow accounts that fed it.

Stack it up. The developer needed prices to rise to roll the debt. The bank needed prices to rise to justify the collateral. The local government needed prices to rise to make payroll. The household needed prices to rise because the apartment was the retirement account, the marriage prerequisite, and the only inflation hedge a capital-controlled saver could reach.

Every participant was on the same side of the same bet — including the referee. In a market with a functioning correction mechanism, someone profits from the fall and therefore has an incentive to price it in early. Here, the party best positioned to call the top was the party whose payroll depended on the top never arriving. So when the correction came, no one was positioned to absorb it. Losses in a market with no adversarial position don't get transferred. They get distributed, downward, to whoever has the least ability to refuse delivery.

iii · who owned the instruments

It is tempting to conclude from this that a system dependent on perpetual growth is structurally incapable of representing its own failure state — that it simply has no organ for that, and so the failure arrives as a surprise.

Tempting, and not quite right, and the accurate version is worse.

The information existed. Vacancy was measurable and was measured. The local housing bureau knew to the yuan what was and was not in the escrow accounts it supervised, because supervising them was the job. Nothing in this story was unrepresentable. It was unpublishable — because at every level, the organ doing the sensing was owned by the party with the most to lose from what it sensed. The bureau auditing presale funds was funded by land sales. The statistical apparatus measuring growth reported to the officials being evaluated on growth.

That is the finding, and it does not stay inside China. A system whose measurement apparatus is owned by the party being measured cannot distinguish coherence from suppression — and neither can anyone standing downstream of its numbers. It is also what turns "nobody told them" from a rhetorical flourish into a structural fact. The buyer could not price developer solvency, not because the information was difficult, but because producing it was somebody's job and that somebody's salary depended on the answer.

The scale of what was never published is, fittingly, not officially known. A former deputy head of the National Bureau of Statistics remarked publicly in 2023 that China's empty homes might house three billion people — and then, apparently reconsidering, said even 1.4 billion probably couldn't fill them. The number is unverifiable. That a retired official felt free to say it out loud, and a serving one would not, is the more informative data point.

Beijing has since deployed the standard toolkit: mortgage rates cut toward 3.5%, whitelists to get bank credit flowing to specific developers, state purchases of unsold stock for conversion to affordable housing, a multi-trillion-yuan stimulus package. Evergrande drew a wind-up order from a Hong Kong court in January 2024. The regulators have also, belatedly, begun moving the market off presale altogether — completed units rose from 10.4% of floor space sold in 2021 to roughly a third by early 2025.

New home prices have nonetheless been falling for roughly two years of monthly readings. The interventions are real, and they are also, structurally, an attempt to restart the growth that produced the problem, executed by the institutions whose revenue depends on it restarting.

iv · what the concrete records

The towers will mostly stay up.

Not from sentiment — from arithmetic. Demolition costs money and produces nothing. Completion costs money and produces units into a market with no buyers. The cheapest available action, for a developer in liquidation or a municipality with a collapsed revenue line, is the null action. Leave it. Post a fence. Let the rebar oxidize. The unfinished tower persists because persistence is free and every alternative has a price.

So the built environment ends up functioning as an involuntary public archive — the one ledger nobody could edit. Every ghost estate in Leinster, every skeleton outside Seseña, every rusting frame in Zhengzhou is a durable physical record of a promise a system made and could not keep, maintained at no cost, visible from the highway, for decades. The statistics were owned. The concrete isn't. Concrete does not compost. It just sits there, being evidence.

And the generation that paid for it has drawn the obvious conclusion, which is the part that should worry Beijing more than the balance sheets.

They were told: bind your savings, your marriage timeline, and your retirement to a single asset class, on the state's implicit assurance that the asset class only goes one direction. A meaningful number of them did exactly that and got a fenced-off frame and a mortgage. The rational response to that experience is to stop betting — to hold cash, defer the marriage, skip the second child, decline the leverage. 躺平, lying flat, is usually written up as a youth attitude problem. It is more accurately described as correct risk assessment by people who watched the last cohort get liquidated.

And that — not the whitelists, not the rate cuts — is the system's failure state finally being represented.

It took a decade. It cost a generation its savings and a chunk of its household formation. And it was performed by the people who absorbed the loss rather than by any of the institutions that caused it, none of which could afford to run the measurement. But it is measurement. A cohort that could not get an honest number out of any bureau in the country derived one from the wreckage and acted on it at population scale. The sensing organ the system lacked got built out of the injury, by the injured, after the fact.

Which is also why it doesn't resolve. Precautionary saving is deflationary. Beijing now needs that generation to bet again, and it won't, and the refusal is itself the next phase of the contraction. The immune response and the disease are the same motion.

Spain took roughly a decade to clear its overhang and permanently lost a chunk of a cohort to emigration. Ireland socialized the bank losses onto a population that had not made the loans. Florida's 1926 lots sat unsold through the Depression and were quietly re-platted a generation later by people who'd never heard of the binder boys.

The template is stable. What varies is how many people were standing underneath when it executed — and who ends up doing the measuring afterward.

It has never once been the institutions holding the instruments.