China's Ghost Cities
China's Ghost Cities: How Rapid Urban Expansion Ran Ahead of Demand
Over two decades, Chinese local governments built hundreds of new towns and districts ahead of demand, financed by land sales and off-balance-sheet borrowing. Ordos Kangbashi, designed for a million residents, held a fraction of that for years and became the global symbol of the phenomenon. Vacancy figures are genuinely contested — China publishes no official vacancy rate — but independent surveys and satellite positioning data point to tens of millions of unoccupied homes. This investigation examines how fiscal design, not incompetence, produced overbuilding on a continental scale, and what programme leaders should learn about building supply against speculative rather than real demand.
Source: investigation key facts.
| Measure | Value |
|---|---|
| Original budget | No published programme-level budget; Kangbashi New Area's build-out was reported in foreign media at up to US$161bn equivalent (figure unverified and contested) |
| Final cost | Unknown at programme level; losses and stranded capital unquantified — Ordos municipal debt reported around RMB 300bn (~US$47bn) by 2012 |
Contents
What Happened
Researchers have linked China's rapid urban expansion to a fiscal model that encouraged local governments to rely heavily on land sales and LGFV financing. The 1994 tax-sharing reform pushed the bulk of tax revenue to Beijing while leaving local governments with heavy spending obligations. Land filled the gap: municipalities requisitioned rural land cheaply, rezoned it, and sold long-term leases to developers at large mark-ups. Off-balance-sheet local government financing vehicles (LGFVs) borrowed against future land sales to build the roads, utilities and civic monuments of entirely new districts. Glaeser, Huang, Ma and Shleifer calculate that developers built roughly 100 billion square feet of residential space between 2003 and 2014, with prices reaching two to ten times construction cost. Ordos Kangbashi in Inner Mongolia became the emblem. Construction of the new district began around 2003, some 23 kilometres from the existing city, with a master plan ultimately aimed at housing around one million people. Coal wealth and stimulus-era credit paid for museums, an opera house and thousands of apartments. When coal prices fell in 2011–12, the local economy stalled, property prices collapsed, developers defaulted and Ordos was left servicing debts reported at around RMB 300 billion (about US$47 billion) while the new city held only tens of thousands of residents. How empty is China? Honest answer: nobody knows precisely, because no official vacancy statistics are published. The China Household Finance Survey put urban vacancy at about 22.4 per cent in 2013 — roughly 49 million units — with little improvement in later waves. A 2015 study using Baidu mobile positioning data mapped systematically under-occupied districts across dozens of cities. In 2023 a former deputy head of the national statistics bureau remarked that even China's 1.4 billion people could not fill its empty homes. Estimates of total vacant units therefore range from tens of millions to figures an order of magnitude higher; any single dramatic number should be treated with caution. The reckoning arrived through the developers. The pre-sale model — selling flats years before completion and using the cash to fund the next project — unravelled after Beijing's 2020 'three red lines' curbed leverage. Evergrande defaulted in 2021, leaving hundreds of thousands of prepaid buyers waiting for unfinished homes. Some new districts, Kangbashi included, have slowly filled as government offices, universities and hospitals were relocated into them; the ghost city label is now less apt there. But the fiscal machinery that built too much, too early, on borrowed money remains the larger unfinished story.
Why It Matters
China's new town programme is the largest single experiment in supply-led urban development in history, and its failure mode — building ahead of demand with borrowed money — is one every megaproject sponsor should recognise. The incentive structure made overbuilding rational for every individual official: land sales funded budgets, construction inflated GDP, and promotion rewarded visible output, while the debt sat off-balance-sheet. Property came to hold the majority of Chinese household wealth, so the correction is not a local embarrassment but a macroeconomic risk that suppresses consumption and local government spending across the country. For project leaders, the lesson generalises: when the metric you reward (starts, floor space, revenue booked) diverges from the outcome you need (occupied, viable places), the system will deliver the metric and abandon the outcome.
-
policy
1994
Tax-sharing reform centralises revenue in Beijing while leaving spending duties with local governments, making land sales their decisive income source.
-
policy
1998
Housing reform ends welfare allocation of homes and creates a private residential property market.
-
construction
2003
Construction of Ordos Kangbashi New Area begins, 23 km from the existing city of Dongsheng.
-
policy
2008-11
A RMB 4 trillion stimulus and relaxed credit trigger a surge in LGFV borrowing and new town approvals nationwide.
-
warning
2009-10
Al Jazeera and TIME document the near-empty streets of Kangbashi; 'ghost city' enters the global vocabulary.
-
failure
2011-12
Falling coal prices hit Ordos; property values collapse, developers default and the city is left servicing debts reported near RMB 300 billion.
-
inquiry
2013
The China Household Finance Survey reports a 22.4 per cent urban housing vacancy rate — roughly 49 million empty units.
-
policy
2014
The National New-type Urbanization Plan (2014-2020) targets 60 per cent urbanisation while urging better-quality, demand-led development.
-
inquiry
2015
Baidu's Big Data Lab publishes a positioning-data study mapping systematically under-occupied districts across dozens of Chinese cities.
-
policy
2016-12
Beijing adopts the line 'houses are for living in, not for speculation', signalling a policy turn against investment-driven demand.
-
policy
2020-08
The 'three red lines' policy caps developer leverage, cutting off the refinancing model that sustained pre-sale construction.
-
failure
2021-12
Evergrande formally defaults, exposing the pre-sale funding chain and leaving prepaid buyers of unfinished flats across the country.
-
warning
2023-09
Former statistics official He Keng says even China's 1.4 billion people could not fill its empty homes, an unusually blunt official-adjacent admission.
-
milestone
2023-24
Kangbashi's population passes 100,000 after government offices, schools and hospitals are relocated there — still far below its planned capacity.
Root cause through the Project Failure Pyramid™ lens
The visible indicators: delays, cost overruns, quality defects, team attrition
Districts that filled — usually those attached to existing job centres and transit — were demand-led; the failures were forecast-led. Phased release tied to occupancy triggers would have stranded far less capital. The same staging logic applies to any capacity programme, from railways to server farms.
Every actor in China's new town boom — mayor, LGFV, developer, bank — behaved rationally under the incentives they faced. If your programme's governance rewards starts and land sales rather than occupancy and cash flow, you will get starts and land sales. Audit the incentive map before you audit the plan.
Chinese households bought empty flats as stores of value because capital controls and thin capital markets left few alternatives. Sales figures therefore overstated end-user need by a wide margin. Any business case built on absorption rates must distinguish occupancy demand from investment demand — the second can vanish overnight.
Project Failure Pyramid™
The 4-Level Diagnostic for Cascading Failures — explore the framework →
Leadership Blind Spot Matrix™
Identifying What Leaders Cannot See — explore the framework →
The Resource-Constrained Growth Model™
Scaling Infrastructure When Capital Is Limited — explore the framework →
Continue Investigating
Apple Park: How Design Ambition Shaped a Multi-Billion-Dollar Headquarters
HS2: How Cost, Scope and Schedule Changed Britain's High-Speed Rail Programme
The Weekly Brief — one investigation, one executive lesson and one project warning sign each week.
Subscribe freeLessons for Leaders
Follow the Incentives, Not the Master Plan
Every actor in China's new town boom — mayor, LGFV, developer, bank — behaved rationally under the incentives they faced. If your programme's governance rewards starts and land sales rather than occupancy and cash flow, you will get starts and land sales. Audit the incentive map before you audit the plan.
Speculative Demand Is Not Demand
Chinese households bought empty flats as stores of value because capital controls and thin capital markets left few alternatives. Sales figures therefore overstated end-user need by a wide margin. Any business case built on absorption rates must distinguish occupancy demand from investment demand — the second can vanish overnight.
Off-Balance-Sheet Debt Is Still Debt
LGFVs existed precisely to keep borrowing off municipal books. The result was a debt stock whose true scale even Beijing struggled to measure. Financing structures that place liabilities outside conventional municipal budgets should be treated as a red flag in themselves, not a clever optimisation.
Absence of Data Is a Decision
China has never published an official housing vacancy rate, leaving policy makers, lenders and households guessing at the scale of oversupply. When the sponsor of a programme declines to measure its most embarrassing variable, outsiders should price in the most unfavourable plausible range.
Build-to-Order Beats Build-to-Forecast
Districts that filled — usually those attached to existing job centres and transit — were demand-led; the failures were forecast-led. Phased release tied to occupancy triggers would have stranded far less capital. The same staging logic applies to any capacity programme, from railways to server farms.
A Slow Recovery Does Not Cancel the Failure
Kangbashi now houses over 100,000 people after government offices, universities, hospitals and other institutions were relocated there.. That was an administrative salvage operation, not market validation. Programme evaluation must discount recoveries that depend on the state moving its own demand into the building.
Executive Recommendations
Tie every capital release to independently verified absorption data — Fund districts, phases or tranches only when the previous phase reaches defined occupancy or utilisation thresholds, measured by a body that does not benefit from continued construction.
Consolidate all programme debt onto one audited balance sheet — Prohibit delivery vehicles whose liabilities are invisible to the sponsor's accounts; the true leverage of a programme must be legible to its board at all times.
Stress-test demand cases against investment-driven buying — Model scenarios in which speculative purchasers exit simultaneously; if the business case only works while everyone believes prices rise, it does not work.
Publish the uncomfortable metric — Mandate public reporting of vacancy, utilisation or ridership — whichever the programme fears most — on a fixed cadence, so bad news arrives while it is still cheap to act on.
Align official tenure with asset performance — Create accountability mechanisms that follow decision-makers beyond their posting, so the cost of overbuilding cannot simply be handed to a successor.
PMOS Intelligence
Preview — illustrative assessment; PMOS is in development“The ghost city story is usually told as incompetence — planners who could not count. That is backwards. The planners counted perfectly well; PIA's interpretation is that the incentive structure rewarded construction and land sales more strongly than long-term occupancy outcomes.. When an official's career, a city's budget and a bank's loan book all depend on the next land sale, 'demand' becomes whatever the spreadsheet needs it to be. Kangbashi is filling now, two decades late, which defenders cite as vindication. It is not: capital that sits idle for twenty years while debt compounds is a failure by any cost-of-capital test, whatever the final occupancy photograph looks like.”Ramesh's Insider Take — opinion
Documentary Evidence
A Real Estate Boom with Chinese Characteristics
Glaeser, Huang, Ma and Shleifer; Journal of Economic Perspectives 31(1) · 2017
Chinese developers built roughly 100 billion square feet of residential real estate between 2003 and 2014, with prices reaching two to ten times construction cost and vacancy rising sharply.
Ghost Cities Analysis Based on Positioning Data in China
Chi et al.; arXiv:1510.08505 · 2015
Baidu mobile positioning data reveals systematically under-occupied new districts across many Chinese cities, the first large-scale empirical mapping of 'ghost city' vacancy.
Wade Shepard (author of 'Ghost Cities of China', Zed Books), Vagabond Journey field report on Kangbashi
Wade Shepard (author of 'Ghost Cities of China'
Ordos accumulated debts of about RMB 300 billion while Kangbashi, planned for hundreds of thousands, held roughly 70,000 residents a decade after construction began; property prices fell precipitously from 2011.
China's Phantom Urbanisation and the Pathology of Ghost Cities
Sorace and Hurst; Journal of Contemporary Asia 46(2) · 2016
China's overbuilding is a 'phantom urbanisation' driven by fiscal and career incentives of local officials rather than by population demand.
Urban China: Toward Efficient, Inclusive and Sustainable Urbanization
World Bank and Development Research Center of the State Council · 2014
Land-based municipal finance and rapid government-led urban expansion created structural pressure to convert and build ahead of demonstrated need.
Everything from this investigation
One investigation. Many outputs. Everything derived from this case, connected in one place.
Read & download
- Executive summaryRead the executive summary
- PDF briefDownload the PDF brief
- TimelineExplore the timeline
- EvidenceReview the documentary evidence
- SourcesSee the source list
- Framework analysisRead the framework analysis
Watch & listen
- DocumentaryIn production
- Episode pagePlanned
- Related documentariesNone yet
- ShortsNot published
- Podcast episodePlanned
Share & follow
- LinkedIn articlePlanned
- Medium articlePlanned
- Substack articlePlanned
- Downloadable templatesPlanned
- NewsletterGet this investigation by email
Connected research
- CompaniesChina Evergrande Group
- CountriesChina
- IndustriesReal Estate & Urban Development
- Related investigationsBoeing 737 MAX · Denver Airport Baggage System · Kansai International Airport
- A Real Estate Boom with Chinese Characteristics (Journal of Economic Perspectives 31(1): 93-116) American Economic Association · 2017 · peer-reviewed-journal
- A Real Estate Boom with Chinese Characteristics (NBER Working Paper 22789, open full text) National Bureau of Economic Research · 2016 · working-paper
- Ghost Cities Analysis Based on Positioning Data in China Baidu Big Data Lab / arXiv · 2015 · research-preprint
- China's Phantom Urbanisation and the Pathology of Ghost Cities (Journal of Contemporary Asia 46(2)) Taylor & Francis · 2016 · peer-reviewed-journal
- Urban China: Toward Efficient, Inclusive and Sustainable Urbanization World Bank and Development Research Center of the State Council, PRC · 2014 · multilateral-institution-report
- New Trend of Urbanization in China: Government-led vs resident-led development modes Lincoln Institute of Land Policy working paper · 2013 · working-paper
- The Story of Ordos Kangbashi, China's Most Famous Ghost City Vagabond Journey (Wade Shepard, author of 'Ghost Cities of China', Zed Books 2015) · 2013 · expert-field-report
Frequently Asked Questions
There is no official count, and the term has no agreed definition. A 2015 Baidu positioning-data study mapped systematically under-occupied districts across dozens of cities, and researchers such as Christian Sorace and William Hurst describe 'phantom urbanisation' as a nationwide pattern rather than a list of specific places. Named examples frequently cited include Ordos Kangbashi, Chenggong (Kunming), Yujiapu (Tianjin) and Zhengdong New District in its early years.
China publishes no official vacancy rate, which is itself telling. The China Household Finance Survey estimated 22.4 per cent urban vacancy in 2013, about 49 million units, with similar levels in later waves. Other estimates, including remarks by a former statistics official in 2023, imply far higher numbers. Treat any single figure — low or dramatic — with caution.
Largely no, but only after an administrative rescue. The district now houses over 100,000 people after government offices, universities and hospitals were relocated there. Critics note that moving state employees into empty flats is not market validation: the capital still sat idle for the better part of two decades while the debts compounded.
Because the model paid everyone involved until it didn't. Local governments needed land sales to fund budgets after the 1994 tax reform; officials were promoted on GDP and visible construction; LGFVs could borrow off-balance-sheet against future land values; and households bought empty flats as savings vehicles because capital controls left few alternatives. Demand measurements were never allowed to interrupt the loop.
The same pre-sale financing model that filled new towns with speculative stock funded the developers. When Beijing's 2020 'three red lines' capped leverage, Evergrande and peers could no longer refinance, defaulted in 2021, and left hundreds of thousands of prepaid buyers with unfinished homes — the demand side of the ghost city story arriving at its balance-sheet reckoning.
