Your hypothesis — that slowing population growth puts the US on Japan's path of slow, decades-long housing deflation — gets the direction of the demographic drag right, but not the scale. On current evidence, a Japan-style national, nominal collapse (50–80% from peak, still below peak 30 years later) is not plausible in the next 5–10 years. What is plausible is quieter: long real (inflation-adjusted) stagnation nationally, and genuinely Japan-like decline in the places already losing people.
A 50–80% national fall spread over 15+ years. The US has none of Japan 1991's preconditions nationally: no burst land bubble, no population decline yet, and a measured housing shortage — not a surplus. That shortage is itself now contested: KU and Zelman research argues it is an affordability shortage, not a unit shortage (§5).
Ageing imposes a drag of roughly 0.5–1 percentage point a year on real prices (Takáts 2012). Partly offset by income growth and shrinking households, that nets to houses appreciating at or below inflation for a long stretch — while nominal prices still creep up.
Depopulating rural counties, parts of the Rust Belt and Midwest, and overbuilt Sun Belt metros are already seeing nominal/real falls, rising vacancy and surplus — while supply-constrained growth metros keep appreciating. Japan sorted the same way.
Confidence in the headline verdict is high for the next 5–10 years: shortage estimates, flat-to-positive 2026 forecasts, mortgage lock-in and Census/CBO growth all point the same way. Confidence in the size of the real demographic drag is only medium — academic estimates agree on sign but differ by a factor of 2–3, and the field's most famous demographic forecast was wrong in sign for 20 years (see §1).
Japan and Italy aged while much of the world still grew. The follow-up research (8–9 Nov 2026) finds the synchronization premise broadly supported for ageing, but staggered for population decline — and the key mechanism has not been directly modelled. No paper found models millions of homes being released in concert across the US, Europe, Japan and China, or quantifies a “synchronization premium” on top of the ~0.5–1%/yr national demographic drag. The foreign-buyer offset that might disappear was always small — $56bn, 2.5% of US dollar volume in Apr 2024–Mar 2025 (NAR), down from $153bn at its 2016–17 peak. On municipal finance, property-tax revenue does not fall one-for-one with values — rates can rise to hold the levy — until caps, delinquency, abandonment and outmigration bind; the likely US path is gradual fiscal strain, with discrete crises only where decline meets a large pension and debt overhang. See §7 · See §8
A dedicated research pass (9 Nov 2026) tested the hypothesis that second / vacation-home ownership is primarily Boomers. It holds as a lifecycle and wealth claim — ownership rates peak at ages 55–74 and Boomers hold ~41% of all US real estate value — but no source publishes a generational share of the second-home stock, and in 2025's financed purchases Gen X was the plurality (~46% estimated, vs Boomers ~27–29% estimated), with cash buying (57% of vacation buyers) skewing the all-buyer picture older than the mortgage data show. The stock is 6.2m homes, 4.3% of US housing (NAHB / ACS, 2024). See §9
Italy has already run a decade of the experiment: population peaked in 2014 at 60.35m (Istat) and has fallen ~2.3% since — held flat in 2025 only because net migration (+296,000) exactly offset the natural decrease (−296,000) — yet nominal house prices are only ~3% below their 2011 peak (Eurostat) after an ~18% credit-crisis slide: a long real erosion, not a nominal collapse. Locally, though, demography is priced: a 2026 Bank of Italy study finds a 10% municipal population decline comes with ~7% lower house prices, and the losses are asymmetric. Milan appreciates; shrinking towns deflate. It points the same way as the verdict: demographics set the slope; credit and supply set the breaks. See §3
A further research pass (9 Nov 2026) tested a different mechanism — flat prices while the cost of owning keeps climbing (insurance +72% and property taxes +31% since 2019, JCHS), making renting the rational choice, collapsing transactions, and starving local revenues and the mortgage / title / brokerage industries. The cost and freeze halves of the premise are established: buying a median home now costs 108% more per month than renting (CBRE, Q2 2025), and existing-home sales have sat at a 30-year low (4.06m) for two straight years. But no Fed, academic or think-tank paper found names a “doom loop” for this mechanism — the one named housing doom loop (Apollo's) is a different loop — and the evidence points to gradual stagnation with severe localised pain, not a national self-reinforcing collapse: levies adjust, owners hold $34tn of equity, and rate lock-in removed sellers faster than buyers. The loop is observable in one segment — Florida condos — and completed in one city, Detroit (§8). See §10
The link is real, but it runs through households and age structure — not raw headcount — and it is weaker, and easier to get wrong, than the headline model suggested.
Mankiw & Weil (1989) built a demographic demand index from Census cross-sections: housing demand rises sharply at ages 20–30, then falls about 1% per year after age 40. Demand growth peaked at 1.66% per year in 1980 as boomers flooded their twenties. Their forecasting equation implied real house prices should fall ~3% per year — 47% by 2007 — as boomers aged out of high-demand years.
What happened instead: real US house prices doubled between 1987 and 2007 (Green & Lee 2016, on the S&P/Case-Shiller index). Mankiw & Weil themselves replied (1992) that the 47% came from a simple equation with a time trend; their structural model produced swings no larger than 22%, and they were “agnostic… regarding exactly how large a fall” would occur.
Green & Hendershott (1996) showed older people in the 1980 Census consumed less housing not because people trade down as they age, but because older cohorts were less educated and earned less over their lifetimes. Controlling for education and income, housing demand does not decline with age. Eichholtz & Lindenthal (2014) found the same for England: demand is driven by human capital, and total demand can still rise as a population ages. Engelhardt & Poterba (1991) applied the same method to Canada and found no significant relationship at all.
Across countries, population size and the working-age share push prices up; a rising old-age dependency ratio (65+ relative to 20–64) pushes them down. The effects are detectable but modest:
| Study | Sample | Price effect of +1% population | Effect of +1% old-age dependency |
|---|---|---|---|
| StudyTakáts (2012) | Sample22 developed nations, 1970–2009 | +1% population+1.05% | +1% dependency−0.68% |
| StudySaita et al. (2013) — Japan | SampleJapanese prefectures, 1976–2010 | +1% population+2.02% | +1% dependency−1.73% |
| StudySaita et al. (2013) — US | SampleUS states, 1975–2011 | +1% population+1.81% | +1% dependency−0.54% |
| StudyHiller & Lerbs (2016) | Sample87 German cities, 1995–2014 | +1% population+1.38% | +1% dependency−1.83% |
Saita et al. elasticities are quoted via Hiller & Lerbs (2016), ZEW Discussion Paper 15-024 — the original Bank of Japan paper was not opened directly in this research (see caveats). Takáts projects an average ageing drag of ~0.8 pp/yr on real prices; Saita implies −2.4%/yr for Japan, 2012–2040. A 2024 study of 10,000+ German municipalities estimates prices would have been up to 12% higher in 2020 had the age distribution stayed at 2008, and projects −18% to −28% by 2050 on age structure alone.
The crucial fact for your hypothesis: Japan's crash was triggered by credit, not demographics. Prices peaked in 1991; the population peaked in 2008 — 17 years later. Demographics amplified and prolonged the fall; they did not time it.
Commercial land in the six largest cities rose more than 300%; the residential land index roughly tripled and commercial quintupled from 1983. Total Japanese land value at peak was widely put at ~4× all US land. Tokyo commercial land tripled; residential nearly doubled.
The Bank of Japan raised rates to 6.0% (Aug 1990); the Ministry of Finance imposed property-lending curbs (sōryō kisei, 27 Mar 1990). The Nikkei fell over 60% in three years. Peaks were staggered: Kansai in Sep 1990, Tokyo 23 wards plateauing Sep 1990–Sep 1991, some regional towns not falling until 1993–96. Not until Sep 1996 were all 174 tracked cities falling at once.
MLIT: nationwide land prices “continued to decline, except for 2007 and 2008” after 1991. Residential land fell in all but three years between 1992 and 2017. A banking crisis, balance-sheet recession and deflation followed — deflation itself raised real debt burdens and delayed clearing.
Population peaked at ~128m in 2008. A Bank of Japan–affiliated prefecture analysis found the correlation between elderly share and land-price change was strongly positive in 1995 (bubble metros fell most) but negative and increasingly significant by 2010 — by then, high-elderly provincial prefectures were falling fastest. That is demographics taking over from the bubble.
Under Abenomics and low rates, residential land has risen since 2018 and commercial since 2016; by 1 Jan 2025 all categories were up for a fourth consecutive year. But levels remain far below the bubble (chart below). Tokyo is one of only two prefectures (with Okinawa) that grew in 2020–2025; 90.6% of municipalities lost residents. Japan did not deflate uniformly — it sorted.
MLIT Residential Property Price Index, 2010 = 100 — nominal yen
Source: MLIT series compiled at japanhouseprices.com. Condos are the exception: the index is ×2.27 since 2010 (+10.8% in 2025), and the average new condo in Tokyo's 23 wards hit ¥136.13m in 2025 (+21.8% YoY) — approaching, but mostly still below, nominal bubble peaks, on the thinnest new supply since 1973.
Japan had ~9.0 million vacant dwellings — 13.8% of the housing stock — in the 2023 Housing and Land Survey, double the 4.5m of 1993, with ~3.85m in the “no designated use” category. Nomura Research Institute projects roughly 1 in 3 homes vacant by the early/mid-2030s if nothing changes. Tokyo's vacancy is lower (~10%, ~900,000 homes). Population has fallen from 128m (2008) to ~123.0m (2025); the IPSS medium projection is ~107m by 2040 and under 100m by 2056. Fertility is 1.14–1.15; births in 2025 (706,000) were the lowest since 1899.
Two structural features the US does not share: (1) Elastic supply — one national zoning system, 12 categories, as-of-right approval, no local veto; Japan sustained over 1m housing starts a year for 40 straight years (1968–2008), and Tokyo builds ~5× per capita what California builds. (2) Depreciating structures — a wooden house has a statutory useful life of 22 years and loses all economic value within ~15–25 years, so old houses can be worth less than the vacant land once demolition is costed. In 2007, 79% of home sales occurred at a loss. Add near-zero immigration and deflationary expectations, and the fall feeds itself.
The biggest percentage falls were in the bubble metros (−70–80% commercial), because they had risen most. The longest, most irreversible falls were outside them: of 131 cities tracked 1983–1999, 66 ended below their 1983 price, and as late as 2016 some regional cities sat at 10–20% of their March 1983 level on their best commercial site. Rural Japan largely never had a bubble to lose — it lost two-thirds of pre-bubble value anyway. Meanwhile the three major metros and four major regional cities have now risen four consecutive years. The national average conceals a sorting story — and sorting is the most likely US analogue.
Italy is the other advanced economy already living the experiment: its population peaked in 2014 at 60,345,917 (Istat) and had fallen to 58,943,000 by 1 January 2026 (provisional) — down ~1.40m, or 2.3%. Its housing story rhymes with Japan's — but in a milder key, and with the same lesson about triggers: prices peaked in 2011, three years before the population did, felled by the euro-area debt crisis and a credit crunch, not by demography. A deeper research pass on 8–9 November 2026 — Istat, Eurostat, Banca d'Italia and OECD sources — replaced this section's earlier secondary figures; what the deep research could not pin down is flagged as an open question rather than smoothed over.
Eurostat / Istat house price index, annual average, 2015 = 100 — nominal euros (Eurostat series via YCharts; Istat confirms 2018–2025)
The worst year was 2013 (−6.5%), amid the sovereign-debt crisis — before population decline had properly begun. Prices drifted down through 2019 even as the population fall stayed gentle, then rose ~18% from trough to 2025 while the population fell in almost every one of those years — helped by cheap credit, then inflation. The quarterly index reached 121.0 in Q2 2026 (+4.0% YoY; +5.2% in Q1 2026), a high for the post-2010 quarterly series — though quarterly 2011 peaks are not in the sources, so “above the 2011 peak” should not be claimed on that basis alone. In real terms the fall was much deeper: the OECD's database shows real prices down at least 20 index points between 2010 and 2019, and still falling in real terms over 2019–2024; estimates of the real peak-to-trough fall range from roughly one-fifth to one-quarter depending on window and deflator — there is no single canonical figure (see caveats). Eurostat's European comparison is stark: between 2010 and Q4 2024, EU prices rose +55.4% and Italy was the only EU country down (−4%).
The bust was in quantities first: residential transactions fell from ~869,000 (2006) to ~403,000–407,000 (2013), −53%, and building permits from 278,602 (2005) to ~47,000 (2014), a post-war low. S&P's contemporary judgement (April 2014) was that the downturn reflected “weakness in the domestic economy and persistent pressures on the credit market rather than a correction of an overvalued housing market.” New and existing homes also diverged: by 2025 the new-dwellings index stood at 131.9 against 112.8 for existing dwellings (2015=100).
Resident population, millions — Istat; projections base 1/1/2024, median scenario
Fertility hit a record low of 1.14 children per woman in 2025 (1.18 in 2024; Sardegna 0.85, Trentino-Alto Adige 1.40). Births in 2025 were 355,000 — the fewest since Italian unification in 1861, down 38% since 2008 — against 652,000 deaths: a natural balance of about −296,000. What held the total flat in 2025 was migration: net international migration of +296,000 almost exactly cancelled the natural decrease. Foreign residents are 5.56m (9.4% of the population); for Italian citizens alone, the natural balance was −333,000. Italy is also the EU's oldest country: at 1 January 2026, 25.1% were 65+ (14.82m), and Eurostat puts its median age at 49.1 years and its old-age dependency ratio at 39.0% — both EU highs. Istat projects the 65+ share to reach 34.6% by 2050. Open question: some secondary and UN-based sources cite a peak near 60.8m rather than Istat's 60.35m; the ~450k gap is a census-revision vintage difference the research could not resolve from primary text.
The household offset. Even as population fell, households numbered 26.6m in 2024–25 — over 4m more than in the early 2000s — because average household size shrank from 2.6 to 2.2 over 20 years and single-person households became the largest type (37.1%, vs 25.9% twenty years ago). That is why housing demand did not fall one-for-one with population. The cushion is finite, though: once single-person formation saturates and the large older cohorts die, it thins — the same arithmetic that applies to Japan and the US.
At national level, Italy's timing breaks the simple population→prices story: prices fell hardest in 2012–2015, while population was still near its peak and fell only ~2% peak-to-2025 — the bust was a financial and sovereign-debt crisis bust (transactions halved, credit collapsed, unemployment hit ~13%), amplified by two decades of flat real incomes. Where demography shows up is place by place — and as of 2026 it has been measured directly:
Casolaro, Luccioletti & Neri, “Demographic trends and housing prices” (Questioni di Economia e Finanza No. 1043), a municipal-level panel covering 2010–2025. In Banca d'Italia's own summary: “a 10 per cent decline in population is accompanied by a roughly 7 per cent decrease in housing prices. This relationship emerges gradually over time, [and] remains robust across [specifications].” That elasticity (~0.7) is a long-run, cross-municipal association — not a year-to-year forecasting rule and not proof of causation: jobs, incomes and amenities move people and prices together. One detail circulating in secondary summaries — that full adjustment takes roughly 19 years — is not in the abstract or newsletter text the research read (the full paper is Italian-only), so it is flagged here as unverified.
Tax-office (OMI) price data tell the same story independently. Grouping 3,658 towns by their own population change over 2015–2025 (Magic Towns' analysis of Agenzia delle Entrate OMI price bands, 17,795 zones — a secondary analysis of official data):
| Town population change, 2015–2025 | Mean price change (nominal) |
|---|---|
| Population changeLosing more than 1% a year | Mean price change−11.0% |
| Population changeLosing 0.5–1% a year | Mean price change−7.6% |
| Population changeLosing 0–0.5% a year | Mean price change−1.7% |
| Population changeGrowing 0–0.5% a year | Mean price change+3.2% |
| Population changeGrowing 0.5–1% a year | Mean price change+6.0% |
| Population changeGrowing more than 1% a year | Mean price change+5.2% |
The analysts' own conclusion: “a point of population lost costs roughly twice what a point gains” — prices respond faster on the way down than up — with their own warning attached: “A relationship in the register, not a cause: jobs and incomes move both people and prices.” The main exception proves the demand point: Lake Garda towns (e.g. Cavaion Veronese, +80.6%) rose despite high vacancy, because the demand is foreign and seasonal and never appears in the residents' register.
Italy is the middle case between Japan and the US: real population decline for a decade, yet no nominal national collapse — an ~18% nominal slide concentrated in a credit-crisis decade, a long real erosion, and sharp regional sorting. As an analogue for the US it is better than Japan on institutions (a market economy, immigration as a safety valve, no 1980s mega-bubble — Italy's pre-bust run-up was +85% nominal over 2000–H1 2008 on Nomisma figures, against +15,000% for land in Japan's six largest cities over 1955–89) but worse on mechanism (its price fall was a crisis bust, not a slow demographic deflation, and its vacancy is mostly second homes, not abandoned akiya). The Italian lesson for the US: expect geographic divergence, not national deflation — shrinking places lose real value, with downside asymmetry, while productive, immigrant-receiving, amenity-rich metros can rise even as the national population ages, provided migration and household formation continue. If US immigration were cut sharply and household formation stalled, Italy's local elasticities — ~0.7 long-run, losses biting about twice as hard as gains — are the best available guide to the downside.
Sources: Istat, Demographic Indicators – Year 2025 (31 Mar 2026) and Population and household projections, base 1/1/2024; Istat, House prices, Q4 2025 and Q2 2026; Eurostat, House prices and rents, Q4 2024 (8 Apr 2025) and Population structure and ageing; Banca d'Italia QEF 1043 via RePEc abstract and Research Newsletter No. 83 (July 2026); OECD Affordable Housing Database HM1.2; OMI-zone and town analysis: Magic Towns (secondary analysis of Agenzia delle Entrate OMI and Istat data). Deep research 8–9 Nov 2026; hedges and open questions are carried in the text and the caveats below.
The US is ageing and its growth is decelerating fast — but from growth, not decline. The swing factor is immigration policy, and the housing-relevant variable — household formation — turns down decades before the population does.
The Census Bureau's 2023 projections were the first main series ever to project an eventual decline — population peaks at ~370m in 2080, easing to 366m by 2100 (+9.7% from 2022). But the spread across immigration assumptions is 209 million people. Pick a scenario:
Natural decrease (deaths exceeding births) begins in 2038; by 2100, 29.1% are 65+ vs 16.4% under 18. Assumes net migration of 853,000–976,000 a year — now above 2025–26 actuals, so treat the 2080 peak as an upper bound under the old assumption.
Source: U.S. Census Bureau, “U.S. Population Projected to Begin Declining in Second Half of Century,” 9 Nov 2023. The CBO's January 2026 outlook is faster: population 349m (2026) → 364m (2056), growth stopping in 2056, deaths exceeding births from 2030 — CBO Demographic Outlook 2026–2056.
Population: 341.8m on 1 July 2025; growth slowed from 0.96% (2023–24) to 0.52% (2024–25) — among the smallest in US history outside the 0.16% pandemic year. Even so, immigration was 71% of 2024–25 growth; natural increase was only 519,000. Fertility hit a record low of ~1.60 in 2024 (CDC 1.599), and the 65+ population reached 61.2m (18.0%); median age 39.0.
Harvard JCHS, State of the Nation's Housing 2026: the 2035–45 pace would be the weakest decade in at least 100 years; the low-immigration variant is 690k/yr for 2025–35. Because immigrant households mostly form 1–2 years after arrival, the 2025 migration drop bites later: JCHS estimates recent-immigrant household growth falling from 640k (2024) to ~143k (2027) if migration stays near 0.3m — about 500,000 fewer households a year by 2027.
Smaller households. Average household size fell from 3.67 (1940) to 2.51 (2020) — 2.53 in the 2020–24 ACS — and solo households rose from 27.3% (2009) to 28.7% (2024). More units are needed per person even as population growth slows, so “population flat” ≠ “housing demand flat.” Caveat: the offset is weakening — the fastest declines were 1960–80, and adults-per-home has risen since the financial crisis as young adults doubled up.
Regional divergence is already here. Growth slowed in 48 states in 2024–25; five lost population — Vermont −0.29%, Hawaii −0.15%, West Virginia −0.07%, New Mexico −0.06%, California −0.02% — and 17 states had natural decrease (deaths > births). Over 2010–2025 only three states shrank outright (West Virginia −88k, Illinois −126k, Mississippi −16k), while Utah grew +8.2% in 2020–25 alone. Eleven states — and nearly half of all counties — already have more over-65s than under-18s.
Sources: Census Vintage 2025 estimates via Brookings (11 Feb 2026) and Pew (May 2026); household projections from Harvard JCHS (11 Aug 2026).
Note the asymmetry: the aligning sources mostly predict a drag, a handoff, and local softness — not a Japan-scale nominal crash. No major academic paper forecasts a US repeat of Mankiw–Weil's −47%.
Takáts (2012) across the OECD; Saita et al. imply −2.4%/yr for Japan 2012–40; German city evidence (Hiller & Lerbs 2016) points the same way. At the US ageing pace, a 0.5–1 pp/yr real headwind for decades is the central academic estimate.
JCHS projects 859k/yr (2025–35) then 510k/yr (2035–45) — the weakest decade in a century; 690k/yr in the low-immigration variant. A 2026 MBA white paper warns demographic change, “accelerating due to immigration policies,” will have “profound implications.”
McKeveny & Kaczmarek (Zelman Associates, 23 Jul 2026) argue declining birth rates, ageing and reduced immigration are already leading to slower household-growth projections for the coming decade — an industry research house, not just academics, making the demographic-deceleration case. Their conclusion that the market is broadly balanced cuts the other way on the shortage itself (card below).
If net migration stays near the projected ~0.3m, JCHS estimates immigrant household formation collapses (640k in 2024 → ~143k in 2027). This is the one variable that could pull the US population peak forward by decades — the Census zero-immigration scenario has decline starting in 2024.
Realtor.com (Oct 2026): Boomers + Silent Generation release 13.9m homes — +74% vs the prior decade — rising from 1.27m (2027) to 1.52m (2036), peaking after 2036. Freddie Mac (secondary reporting): ~9.2m fewer Boomer homeowner households by 2035. Oldest Boomers turn 80 in 2026; ownership “holds steady into mid-70s, begins to slip in early 80s, falls steeply in mid-to-late 80s.” HousingWire's coverage of the same study (5 Oct 2026) adds that Boomers overtake the Silent Generation as the larger releasing cohort around 2029, and sets the release against JCHS household growth of ~859,000/yr in the coming decade vs ~1.2m/yr since 2000 — release running ahead of new-household formation.
Of the 13.9m released homes, 9.9m (71.2%) are family homes and 3.6m are large homes — and turnover is fastest in the Rust Belt and Midwest, not expensive coastal markets (NAHB analysis, via HousingWire). Softening concentrates where demand is already weakest: the provincial-Japan pattern.
Florida median −0.39% YoY and Texas −0.81% (Oct 2025); new-home supply at 9.8 months (mid-2025) vs 5–6 balanced; multifamily vacancy a record ~8.5% (end-2025, CoStar); rents falling (TX −2.2%, AZ −2.8%, FL −1.7%); Sarasota vacancy 17.4%; Austin added ~32,500 units in 2024 plus 17,500 in 2025. 2026 forecasts: −4.5% for New Orleans (Zillow), −2% to −5% in select Sun Belt/West metros (J.P. Morgan).
On Francke's (2025) cohort logic, the buyer impulse peaks when a cohort is in its late 20s/30s; millennials (born 1981–96) are now 30–45. Gen Z cohorts are smaller and form households later — 32.5% of 18–34s lived with parents in 2024. This is an argument, not a measurement: JCHS still credits millennial size with supporting aggregate demand.
Most mainstream estimates find the US short of homes (chart below), where Japan has a 9m-home surplus — though KU and Zelman research now disputes the aggregate (card below). Freddie Mac notes 5.8m units were built in 2020–2024 yet the deficit shrank by only 100,000 — latent demand absorbed almost everything. Even the low outlier (J.P. Morgan, 1.2m) is a shortage, not a surplus.
~89,000 local governments set zoning; ~75% of residential land in US cities is zoned exclusively for detached single-family homes, with discretionary review and neighbour veto. Japan has one national system. Tokyo builds ~150k homes a year for ~13m people; California ~80–90k for ~39m. The shortage is partly a policy choice that protects existing values — the opposite of Japan's model.
78.1% of mortgages were below 6% in Q4 2025 (FHFA); ~50% are at ≤4% (Fed Gov. Barr, Sep 2026). FHFA estimates lock-in prevented ~1.72m sales between Q2 2022 and Q2 2024; turnover fell to ~3.8% in 2023 and mobility hit a record-low 11.2% in 2024. Supply nationally: 3.7 months (Jan 2026), 1.60m homes / 4.8 months (Sep 2026) — still tight. Risk cuts both ways: a fall to ~5–5.5% rates could unfreeze a wave of listings at once.
Boomers hold ~$17tn in home equity — about half the US total (Federal Reserve data, via secondary reporting). Owners free and clear: 58% of 65–74s, 69% of 75–84s, 82% of 85+. With no mortgage forcing a sale, owners can age in place, bequeath, or hold — slowing price discovery in both directions.
The median new US home costs $459,826 (NAHB 2025); 74.9% of households can't afford it. Japan itself proves the point: Tokyo's record 2025 condo prices are partly attributed to construction costs up 30–40% since 2021. And unlike Japan, US structures don't depreciate to zero in 15–25 years — they're renovated and treated as stores of value.
2026 national forecasts are all flat-to-positive: J.P. Morgan 0%, Zillow +0.7%, Fannie Mae +2–3%/yr through 2030, NAR +4%. Median existing-home price was $417,500 in Sep 2026 (+1.3% YoY). Connecticut and New Jersey rose +7% YoY in 2025; the corrections are −2% to −5% (Cape Coral −9.6% in some measures), not −30% to −50%.
Japan 1985–91: commercial land +300% in six years, land worth ~4× the entire US. The US 2020–22 run-up was rapid but far smaller, off a post-crisis underbuilding base, with 30-year fixed mortgages, stricter underwriting and no land-collateral credit loop. A US deflation would have to be demographically driven — and that drag is ~1 pp/yr real, not 5–10 pp/yr nominal.
Realtor.com itself stresses the handoff “won't eliminate the nation's housing gap,” will arrive gradually, and that softening will be “real, but gradual” — concentrated in family/large homes and slower-growing markets, not a nationwide drop. HousingWire's coverage of the study (5 Oct 2026) reaches the same conclusion: the release — from 36.7m Boomer + Silent owner-occupied homes in 2026 down to 22.8m by 2036, ~1.39m a year — brings localized softening, not a broad national price correction, and will not erase the ~4m shortage, especially at entry level. The 2026 forecast figures above come from secondary market round-ups; not every forecaster's primary page was opened in this research (see caveats).
Both of these weaken the “shortage as a price floor” argument in the panel above — in different ways. The KU study says the aggregate unit shortage largely isn't there; Zelman says the market has already moved toward balance as demographic demand slows.
McClure & Schwartz (University of Kansas) — an affordability shortage, not a unit shortage. Using Census data for 2000–2020, they find only 4 of 381 metropolitan areas — and only 19 of 526 micropolitan areas — experienced a housing shortage. Across 2000–2020, housing production exceeded household growth by 3.3 million units: the 2000–2010 surplus more than offset the 2010–2020 shortfall. Nearly all metros have sufficient units for owner occupancy; what nearly all lack is rental housing affordable to very-low-income households. Limit: the data end in 2020 — before the 2021–25 period — so the finding does not speak to the most recent five years.
Zelman Associates (McKeveny & Kaczmarek, 23 Jul 2026) — “reasonable balance,” not a historic shortage. Challenging the “severe housing shortage” narrative, they argue declining birth rates, ageing and reduced immigration are leading to slower household-growth projections for the coming decade, and that the market has shifted to reasonable balance overall — modest oversupply in multifamily, modest undersupply in single-family. On their read, a substantial retrenchment in new construction seems unwarranted, but meaningful growth is hard to justify without much better affordability. Ivy Zelman made the same case on CNBC the following day — “Our housing market is balanced,” CNBC video interview, 24 Jul 2026 (video source — cited for the interview and its headline claim, not quoted). Limit: the Zelman page read is a summary; the full report sits behind its link and was not opened.
Sources: University of Kansas news release, “Study finds US does not have housing shortage, but shortage of affordable housing”; Zelman Associates, Zelman Insights, July 2026; CNBC, video interview with Ivy Zelman, 24 Jul 2026.
Estimated US housing shortfall, millions of units
Fed Governor Barr (23 Sep 2026) summarised the range as ~2m to 5.5m — “1% to 4% of ~150m stock… small in percentage terms but consequential because markets need vacancy.” The CEA's 10m uses a pre-2008 construction-pace counterfactual for single-family homes only and is widely questioned. Sceptics put the aggregate shortage near zero on some definitions: McClure & Schwartz (KU) find production outran household growth by 3.3m units over 2000–2020, Zelman Associates (Jul 2026) call the market broadly balanced (card above), and Kevin Erdmann argues adults-per-home rose after the crisis. The direction of post-2009 underbuilding (starts fell from 1.8–2.1m/yr to under 600k in 2009) is not contested — whether the cumulative gap still stands after 2020 is exactly what KU and Zelman dispute.
The national average will be the weighted sum of these three — and will most likely look like stagnation, not deflation. A national +1% can conceal −5% in Austin and +7% in New Jersey, just as Japan's index concealed Tokyo's recovery and rural collapse.
Sun Belt growth corridors and supply-constrained Northeast/Midwest metros. Zoning, costs and the shortage dominate; the demographic drag is a rounding error against constrained supply. Connecticut and New Jersey +7% YoY in 2025 are the current template.
Austin, Phoenix, Sarasota, Cape Coral, New Orleans. Pandemic-era construction created a genuine glut: expect a 2–5% correction, and watch inventory and multifamily vacancy for clearing over 12–18 months as starts fall. Cyclical, not structural — unless vacancy keeps climbing past 9%.
The true US analogue of provincial Japan. Natural decrease, the fastest Boomer releases, family/large homes with no starter-home demand behind them. Here the right comparator is not Tokyo — it's the Japanese regional city at 10–20% of its 1983 commercial value.
Sustained net immigration near ~0.3m/yr instead of ~0.9–1.3m, household growth durably below ~600–700k/yr, months of supply above 6 nationally, a zoning liberalisation that removed the shortage, and a rate fall that unfreezes lock-in — all at once. Japan had every adverse factor together, plus a burst bubble and deflation. Any one factor alone is insufficient.
Your distinction is real: Japan (1990s onward) and Italy (2010s onward) aged while much of the world was still growing, so foreign demand, capital inflows and migration could partly cushion them. The US will age alongside most other rich economies. The follow-up research verdict: the premise is directionally supported, quantitatively unproven, and staggered rather than a simultaneous cliff — and, strikingly, no one has directly modelled the concerted-release case.
UN World Population Prospects 2024 (WPP 2024), via UN DESA Policy Brief 167 (14 Nov 2024, verified live in the research): 63 countries/areas — 28% of world population — had already peaked by 2024 (China, Germany, Japan and Russia among them); a further 48 (10%) peak in 2025–2054; and 126 countries (62%) are still growing through 2054, including India, Indonesia, Nigeria — and the US, Canada and Australia. The peaked group is projected to fall 14% over the next 30 years. World population itself peaks at ~10.3bn in the mid-2080s. Crucially, the UN says the still-growing rich countries “would likely peak much sooner without immigration.”
| Country | Total-population peak | Working-age peak | Population 2024 → 2050 → 2100 |
|---|---|---|---|
| CountryJapan | Total peak2010 (~128m; some vintages say 2008) | Working-age peak1995 | 2024 → 2050 → 2100122.4m → 105.1m → 76.8m (−37%) |
| CountryChina | Total peak2021 (~1.41bn) | Working-age peak2015 | 2024 → 2050 → 21001,419m → 1,215m → 633m (−55%) |
| CountryItaly | Total peak2014 (~60.7m) | Working-age peak1992 | 2024 → 2050 → 210058.9m → 51.9m → 35.4m (−40%) |
| CountryGermany | Total peak2024 (~84.6m) | Working-age peak1997 | 2024 → 2050 → 210083.6m → 78.3m → 70.9m |
| CountrySouth Korea | Total peak2021 (~51.9m) | Working-age peak2017 | 2024 → 2050 → 2100Peaked 2021; dependency ratio overtakes Japan's by 2045 (below) |
| CountrySpain | Total peakListed among the already-peaked in UN summaries; exact year not verified from a primary source (secondary lists give 2023/2024) | Working-age peak— | 2024 → 2050 → 2100— |
| CountryFrance / UK | Total peakGrowing through 2054 (UN group); exact peak year not verified | Working-age peak— | 2024 → 2050 → 2100— |
| CountryUnited States | Total peakNo peak this century (UN medium); Census 2023 main series peaks >369m in 2081, then 366m by 2100 | Working-age peakNo peak before 2100 (UN medium) | 2024 → 2050 → 2100345m → ~381m → 421m (UN) / 366m (Census main) |
| CountryCanada / Australia | Total peakNo peak this century (UN medium); would peak much sooner without immigration | Working-age peakNo peak before 2100 | 2024 → 2050 → 2100Grow >20% 2024–2100, largely via immigration (UN) |
Working-age peaks are a secondary calculation from WPP 2024, not a UN table the research inspected directly — flagged as such. Japan's total peak is given as 2008, 2009 or 2010 in different vintages; the difference is definitional, not substantive. The UN and Census US totals differ (421m vs 366m by 2100) because of different immigration assumptions: Census 2023 puts the US at 226m by 2100 under zero immigration, 319m under low immigration and 435m under high immigration, with main-series net migration assumed at only ~853,000–976,000/yr. The EU as a whole is projected by Eurostat to peak at ~453m in 2026 (secondary reporting; the Eurostat primary was not directly opened).
Old-age dependency ratio, 65+ per 100 people aged 20–64 — OECD Working Paper 1807 (2024), verified live in the research. Do not mix this denominator with the UN's 65+ / 15–64 basis.
| Country | 2000 | 2023 | 2050 (projected) |
|---|---|---|---|
| CountryJapan | 200028% | 202355% | 205079% |
| CountryItaly | 2000— | 202341% | 205077% |
| CountrySouth Korea | 2000— | 202327% | 205080% |
| CountryChina | 2000— | 202322% | 205054% |
The same OECD paper: in some Asian and Eastern/Southern European countries the working-age population may shrink by more than a third by 2050, and further immigration “may help prevent a decline in the working-age population but can only play a moderate role in holding back the strong rise in old-age dependency” — stabilising the ratio would require very high net migration. The EU Ageing Report 2024 (via a cepInput summary, index-level only; denominator not confirmed) puts the EU ratio at 36.1% in 2022 → 59.1% in 2070 — Germany 37.4% → 55.0%, France 38.2% → 57.8%, Italy 40.8% → 65.5%, Spain 33.3% → 64.5%, Poland 31.9% → 63.7%. South Korea, on OECD projections reported in Korean press (Nov 2026, secondary), reaches 103.3 by 2070 on the 65+ / 15–64 basis, overtaking Japan in 2045, with its working-age population down 51.8% by 2070. Open questions: no consistent 2030 ratio table for all countries was found in primary sources — the 2023 baselines above are the nearest verified anchor — and no directly comparable US ratio for 2050/2070 on the OECD basis was verified; Census instead has US 65+ overtaking under-18s around 2029 (71m vs 69m), reaching 24.4% of the population by 2060 and 29.1% by 2100, with working-age adults per older adult falling from ~3.4 today to ~2.4 by 2060.
After targeted searches on 8–9 November 2026, the research found no paper that models simultaneous, concerted housing release across the US, Europe, Japan and China, and none that quantifies how much the loss of foreign buyers or capital would add to US price falls beyond the national demographic drag. There is also no comparable cross-country total for simultaneous release (for example, “X million homes released in the EU + Japan + China in 2030–2050”) — the literature quantifies releases nationally, not jointly. Any number for a “synchronization premium” would therefore have to be built — for example by summing national Takáts headwinds and adding a capital-flow reversal — not cited. What follows is the closest established work, with its limits stated.
| Model / paper | Quantitative finding | Relation to synchronization |
|---|---|---|
| ModelTakáts (2010), BIS WP 318; J. Housing Econ. 2012 — panel, 22 advanced economies, 1970–2009 | FindingAgeing lowers real house prices “substantially over the next forty years”; US headwind ~80 basis points/yr, “much stronger in Europe and Japan”; average advanced-economy return ~150bp/yr lower than under neutral demographics (per an MNB summary) | RelationClosest to the hypothesis. Multi-country and forward-looking, with an explicitly global headline — but it estimates a common per-country demographic effect; it does not shock all countries' housing supply at once or model cross-border housing demand |
| ModelNishimura & Takáts (2012), BIS WP 385 — lifecycle model + 22-country panel, 1950–2010 | FindingBoomer saving drove property and money prices up; retirement reverses the process | RelationThe mechanism behind Takáts; no simultaneous-release quantification |
| ModelSun, Chand & Sharpe (2019), MPRA 94418 — panel, 19 advanced economies, 1970–2016 | Finding+1% worker population → +1.62% house prices; +1% retiree population → −0.81% | RelationRefines Takáts by separating fertility from longevity: fewer workers hurt prices, longevity raises them (workers save more in housing), retirees sell. The net depends on which dominates |
| ModelMankiw & Weil (1989) — see §1 | FindingSimple extrapolation → −47% real by 2007; their structural model implied ≤22%, and they disavowed the 47% as a point forecast | RelationThe forecast failed — US real prices rose. In a global context it is the cautionary tale: pure demographic extrapolation misses income, credit, supply, cohort and immigration effects |
| ModelPoterba (2001) vs Abel (2001) — the “asset meltdown” debate | FindingPoterba: retirees do not spend down assets fully (bequests, precautionary saving) → rejects a meltdown. Abel: even accepting that, prices can still fall because the supply of capital adjusts; a bequest motive does not prevent the fall | RelationThe core theoretical dispute. For housing, Poterba's point is stronger — a house is also a consumption good and a bequest — but Abel shows the meltdown logic does not depend on retirees selling everything |
| ModelBörsch-Supan, Ludwig & Winter (2005/06), NBER WP 11850 — multi-country OLG general equilibrium, 7 world regions, 70-year horizon | FindingCapital flows from fast-ageing regions (Germany, Italy, Japan) to the rest of the world are “initially substantial,” then reverse when households decumulate; closed-economy models “miss quantitatively important effects” of capital mobility | RelationClosest multi-country general-equilibrium model of synchronized ageing — but its asset is productive capital, not housing, and it does not model houses being released. It implies the Japan/Italy offset, insofar as it existed, was a capital-flow offset during the saving phase, not a buyer offset during the selling phase — and it is time-limited |
| ModelIMF WEO April 2025, Ch. 2, “Silver Economy” — multi-country heterogeneous-agent OLG | FindingAgeing subtracts ~1.1pp from global output growth through 2050 and ~2pp through 2100 (secondary summary); a healthy-ageing / labour-supply package adds ~0.6pp/yr through 2050 | RelationMulti-country and demographic — but about growth and labour, not house prices |
| ModelDavis & Karim (2026), National Institute Economic Review — literature review | Finding“Demographic effects are detectable for house prices, with the working population tending to drive prices up while retired groups restrain it but only to a limited degree” — the elderly desire bequests and precautionary saving and prefer not to sell or borrow against their property | RelationThe current consensus restated: the effect is real, modest, and dampened by bequest and ageing-in-place behaviour |
Why the gap exists is itself informative: overlapping-generations models with housing consistently produce the same pattern — a large working-age cohort bids up a relatively fixed stock; in retirement and death, demand falls and prices soften unless offset by bequests, longevity-driven saving, immigration or supply contraction — but housing is overwhelmingly a non-traded, local asset, so most modellers treat the market as national or local, with capital flows as the only cross-border channel. Takáts-type panels use time fixed effects that partly absorb common global shocks without identifying a cross-border housing demand channel, and no paper found tests whether foreign housing purchases specifically offset domestic demographic drag.
The release itself is quantified only nationally. In the US: Realtor.com (5 Oct 2026) projects Boomers + Silent Generation falling from ~36.7m homes owned in 2026 to 22.8m by 2036 — 13.9m homes released in 2026–36 (~1.39m/yr, rising from 1.27m in 2027 to 1.52m in 2036), with the pace still rising, so the peak is after 2036; only 0.38m are starter homes and 71% are family-sized. Freddie Mac (Feb 2024 research, via NBER WP 34656, Jan 2026) projects ~9.2m fewer Boomer homeowner households by 2035 and calls it “more of a tide” than a tsunami; Boomers hold ~$17.3tn in housing wealth, about half of US home equity, and are 38% of homeowners. Harvard JCHS 2025 projections (via secondary reporting) put the demand side of the same arithmetic at ~860,000 new households/yr in 2025–35 and ~510,000/yr in 2035–45 — so release (~1.4m/yr) against net new households (~0.5–0.9m/yr) implies a growing national surplus on paper, concentrated in large, suburban, slow-growth markets. Flagged figures: Mortgage Bankers Association numbers circulating in secondary reporting (~250,000 additional units/yr from ageing/mortality as a net figure, and “more than 4m existing homes a year” as a gross flow) appear to measure different things and should not be combined; Nelson's (2020) “Great Senior Short-Sale” scenario of up to 18m more large-lot detached homes than households would desire by 2038 — and a related claim of a glut “as high as 15m by 2040” — was not verified from Nelson's primary paper and is a scenario, not a forecast. Japan's realised version is 9.0m vacant homes, 13.8% of stock, in 2023 — spread over 30 years of decline (§2).
NAR, 2025 International Transactions in U.S. Residential Real Estate (release 14 Jul 2025; period Apr 2024–Mar 2025; verified live in the research)
Foreign buyers bought 78,100 existing homes (+44% YoY, the first rise since 2017) at a record median price of $494,400; 47% paid all-cash (vs 28% of all buyers). That is 2.5% of the $2.2tn in total dollar volume and ~1.9% of existing-home sales (secondary). Composition matters: 56% were resident foreigners or recent immigrants (43,700 homes, $26.9bn), so the truly non-resident “foreign capital” slice is ~$29.1bn — about 1.3% of dollar volume. Top origins: China 15% (11,700 homes, $13.7bn), Canada 14% ($6.2bn), Mexico 8% ($4.4bn), India 6% ($2.2bn), UK 4% ($2bn). Top destinations: Florida 21%, California 15%, Texas 10%, New York 7%, Arizona 5%. Against the 2016–17 peak of ~284,500 homes, foreign buying had already fallen by about two-thirds in dollar terms before any synchronized-ageing effect — driven by prices, rates, capital controls (notably China's) and pandemic disruption.
What that means for the hypothesis. Even a total loss of this flow could not arithmetically absorb a 13.9m-home US release — but equally, its loss alone cannot cause a crash: it is a marginal, high-end, coastal-concentrated flow. The direction still supports you: the largest source country, China, is itself shrinking (working-age peak 2015; population −55% by 2100) and ageing (dependency ratio 22% → 54% by 2050), so its capacity to export both capital and homebuyers falls; Canadian buyers are disproportionately retirees buying in Florida and Arizona — part of the same ageing cohort, not a younger offset. Best read: losing foreign demand removes price support at the top end in a few markets (Florida, California, New York), rather than driving national prices — though a synchronized fall in foreign demand would coincide with a synchronized fall in foreign supply of capital, i.e. potentially higher global rates, which is a separate and larger channel (below).
The people exist: UN WPP 2024 has sub-Saharan Africa growing +79% to 2.2bn by 2054 and +51% to 3.3bn by 2100, with more than half of global growth to 2050 coming from just eight countries (India, Nigeria, Pakistan, DR Congo, Ethiopia, Tanzania, Egypt, Indonesia) plus the US; Latin America and the Caribbean peak around 2058. For the US, immigration is the main driver of growth — the UN has the US growing >20% from 2024 to 2100 “largely as a result of immigration,” and Census has decline starting in 2024 without it. But the age-structure effect is small: Census low- vs high-immigration scenarios differ by only ~0.9pp in the working-age share in 2060 (57.6% vs 58.5%), and the OECD concludes stabilising the dependency ratio through immigration alone would require historically unprecedented sustained net migration.
Illustrative arithmetic (the researcher's own, from the cited figures — not a model): closing the gap between elderly release (~1.4m homes/yr) and net household formation (~0.5–0.9m/yr) via immigration alone would require roughly an additional ~1–2m working-age immigrants a year forming households at typical rates, sustained for decades — well above the Census main-series assumption of ~0.9m net migrants a year in total, and above realised levels except the 2023–24 surge (2.7m, then 1.3m).
| Mechanism | Direction | Evidence / status |
|---|---|---|
| MechanismInterest rates — savings-glut reversal? | DirectionAmbiguous — the most important, and most contested | EvidenceGoodhart & Pradhan, The Great Demographic Reversal (2020): fewer workers → less saving, scarcer labour → higher rates and inflation. Against: Auclert et al. (OECD/CEPR, Jun 2025) — demographics pushed the neutral rate down to date, and multi-country OLG work “soundly reject[s]” the reversal hypothesis; estimates of the demographic drag on the neutral rate range from <100bp (Gagnon et al. 2021) to >300bp (Eggertsson et al. 2019). IMF WP/26/168 (Aug 2026): ageing raises desired saving and reduces investment demand, pushing real rates down, though rising Treasury supply has compressed convenience yields. If the reversal view is right, higher rates amplify the price drag; if Auclert is right, low rates dampen it. No consensus — the key uncertainty |
| MechanismConstruction costs / supply contraction | DirectionDampener | EvidenceIf household growth falls to ~510k/yr (2035–45) and new-unit demand to ~800k/yr (JCHS), builders cut the flow of new homes, partly offsetting the rising stock of released homes. But in shrinking markets builders simply stop building rather than cut prices — so supply contraction helps national prices more than local ones. No quantified offset found |
| MechanismZoning / land-use reform | DirectionDampener, if it happens | EvidenceUpzoning, subdividing large Boomer homes, ADUs and conversion to multi-family could match the type of supply to the type of demand — the main policy lever for the mismatch — but it is local, slow and contested. No national estimate of its price effect in a release scenario was found |
| MechanismRemote work | DirectionAmbiguous / locally amplifying | EvidenceSpreads demand toward cheaper, release-heavy markets (dampening the local glut) but reduces demand for commute-sized suburban homes and makes outmigration from high-tax shrinking cities easier (amplifying the fiscal loop in §8). No quantified study found |
| MechanismClimate migration | DirectionAmbiguous / redistributive | EvidenceMoves away from heat, fire and flood toward the Great Lakes and Northeast could refill some Rust Belt markets while emptying Sun Belt release markets (Florida, Arizona, Texas) — redistributing the glut rather than removing it. No quantified demographic-climate-housing model was found |
| MechanismHealthy ageing / longer working lives | DirectionDampener (delay) | EvidenceIMF WEO 2025: healthy-ageing and labour-supply measures could add ~0.6pp/yr to global growth through 2050, offsetting about three-quarters of the ~1.1pp demographic drag (secondary summary). Longer working lives also delay housing release — but they delay it, not remove it |
| MechanismBequests / ageing in place / stepped-up basis | DirectionDampener (delay) | EvidenceDavis & Karim (2026): bequest and precautionary motives mean retirees “prefer not to sell or borrow against existing property”; US stepped-up basis on inherited homes plus capital-gains lock-in (Parrott / Urban Institute, Sep 2025) reinforce staying put. Result: release is a 30–40-year tide, not a 5-year wave — which is why national models produce a drag (~0.8%/yr), not a crash |
Supported: synchronization of ageing is a fact (dependency ratios rise in every OECD country to 2050/2070; the EU peaks in 2026); the foreign offset was never large and its source countries are ageing fastest, with Börsch-Supan giving a mechanism for it to reverse, not just disappear; and the national drag of ~0.5–1%/yr real is established, with Takáts' ~80bp/yr for the US at its centre. Not supported: any quantified premium on top of that drag — and the US is not Japan or Italy: it still grows on immigration, its working-age population has not peaked, household growth stays positive (~510–860k/yr), and a 13.9m release over a decade (~1.4m/yr) runs against ~4m/yr existing-home sales and ~1.4m/yr new-unit demand — absorbable nationally, painful locally. The binding problem is mismatch — large detached suburban homes vs smaller, starter, rental demand, in slow-growth places — not a uniform glut.
For simultaneous release to depress US prices materially beyond the national drag, the research concludes all of the following would need to hold together — presented in the report explicitly as synthesis and speculation, not a cited model: (a) a bunched release, if ageing-in-place and bequest behaviour breaks (e.g. removing stepped-up basis, or a health/care shock forcing sales); (b) immigration falling and staying low — net migration materially below the Census main-series ~0.9m/yr, so household formation falls toward or below ~500k/yr while release rises toward 1.5m/yr; (c) rates rising on demographic reversal (Goodhart–Pradhan right, Auclert wrong); (d) foreign capital reversing, with foreign buying — already down two-thirds from peak — falling further and removing top-end support in Florida, California and New York; and (e) local concentration of the release in metros with weak demand and high pension/debt burdens, triggering the fiscal loop in §8. If (a)–(e) do not jointly hold, the expected outcome is the Takáts outcome: a persistent ~0.5–1%/yr real headwind, larger in slow-growth older metros, smaller or zero in growing, supply-constrained, immigration-heavy ones — divergence, not a uniform national fall.
Your second question: if home values fall for years, do US municipalities — funded largely by property taxes on those values — face a financial crisis? The research answer: dependence is real and schools are the most exposed, but revenue does not fall one-for-one with values. The crisis, where it comes, is mostly gradual fiscal strain — with discrete crises reserved for places where decline meets fixed legacy costs and binding tax caps.
| Measure | Figure | Source / date |
|---|---|---|
| MeasureProperty tax share of local general revenue | Figure~30% | SourceCensus via Lincoln Institute, Significant Features of the Property Tax (US, 2020) |
| MeasureProperty tax share of local tax revenue | Figure~72% | SourceCensus of Governments (via secondary summaries) |
| MeasureProperty tax share of local revenue, Northeast | FigureUp to 62% | SourceLincoln Institute (Colorado testimony, citing Census 2019) |
| MeasureProperty tax share of local school revenue (direct levy) | Figure63.31% ($272.1bn of $429.9bn) | SourceCensus Annual Survey of School System Finances, FY2024 |
| MeasureProperty tax share of total school revenue | Figure27.35% ($272.1bn of $994.9bn; down from 29.05% in FY2012) | SourceSame, FY2024 |
| MeasureProperty tax share of state + local revenue | Figure~16.6% | SourceLincoln Institute (Michigan profile, 2020) |
Caveat carried from the research: the school figures count only direct property-tax levies by independent school systems; in dependent-system states (e.g. CT, MD, MA, VA, NC) schools are funded by parent-government appropriations whose property-tax share Census cannot isolate — so 63% / 27% understates schools' true property-tax dependence. Either way, the property tax is the largest single own-source revenue for local government and the dominant funder of K–12 schools; no other local tax comes close (sales tax is ~7% of local general revenue nationally, ~13% of local tax revenue).
The arithmetic is revenue = assessed value × tax rate — but in most US jurisdictions the levy (the total dollars to be raised) is set first by the budget, and the rate is derived as levy ÷ base. So if the tax base falls 10% and the levy is unchanged, the rate rises ~11% to compensate. That is why Detroit's effective rate (~3%, among the highest in the US) coexists with collapsed values: the same levy spread over a smaller base needs a higher rate (University of Michigan Poverty Solutions, on 2023 ACS data; national average 1.38%). The research's net assessment: a 0.5–1%/yr real national price drag does not by itself cause a municipal revenue crisis — rates can absorb it for years. A crisis requires the drag to be locally concentrated, persistent over decades, and combined with a fixed-cost overhang (pensions, debt, infrastructure built for a larger population). That is a statement about the mechanism, not a quantified threshold — no threshold model was found.
Falling values → smaller base → higher rates or lower revenue → worse services and/or higher tax bills → outmigration (especially of younger, mobile, higher-income households) → further value falls → more abandonment and delinquency → higher rates on those who remain → … Each link is documented in at least one case — Detroit (values → rates → outmigration), Prichard (population loss → pension non-payment), Japan (depopulation → per-capita costs → service withdrawal → further outmigration) — but no paper found estimates the loop's gain (e.g. “a 1% value fall causes a further 0.X% fall via fiscal channels”) for the US as a whole. That is a second modelling gap, alongside the missing simultaneous-release model in §7.
Pensions, bonds and schools are the fixed-cost overhang. The Lincoln Institute (Dye & Gordon, 2012) finds the local governments in trouble are disproportionately (a) those losing people and jobs over time — Detroit had twice as many retirees as active workers at filing — and (b) those that rode the housing boom and bust (Stockton, Vallejo, Central Falls); “mobile residents and businesses could flee communities that levy higher taxes to rebuild pension assets rather than to provide basic services.” State and local pension funds' equity holdings fell from $2.3tn (Sep 2007) to $1.2tn (Mar 2009) in the financial crisis — the asset side is exposed to the same headwinds. General-obligation bonds are backed by taxing power: as the base shrinks and rates hit caps, ratings fall, borrowing costs rise and capital spending is deferred, degrading services and values further. Schools are the most exposed service: enrolment falls with the child population but per-pupil fixed costs (buildings, transport, pensions) do not; state per-pupil aid falls with enrolment while local property-tax revenue falls with values — a double squeeze that ends in consolidation and closure, which itself depresses local home values (a second loop). No national quantification of the school channel was found — the research calls it the largest unquantified fiscal risk in this work.
Population fell from 1.85m (1950) to ~1.0m (1990) to 713,000 (2010) — about −1.4%/yr for 60 years — while infrastructure built for 1.85m served 713k. By 2008 there were >67,000 foreclosures and median values <$10,000 in many neighbourhoods; unemployment was 22% and poverty 35%. Detroit filed Chapter 9 on 18 July 2013 — the largest municipal bankruptcy in US history, at ~$18bn in obligations. Flagged as secondary reporting: tax revenue had fallen by >40% over the preceding decade; pensioners took a ~4.5% benefit cut with COLAs eliminated and retiree healthcare cut ~90% for most retirees. After bankruptcy the city still has >30 sq mi of vacant land and ~3% effective tax rates — in the research's phrase, bankruptcy “ratified the collapse” rather than reversing decline.
Youngstown (OH), Cleveland and Pittsburgh follow the Detroit pattern at smaller scale. Prichard lost >45% of its population after 1970, had <23,000 residents by 2010, stopped sending pension checks in September 2009 and declared bankruptcy a month later (Lincoln Institute). Pittsburgh is the counter-example: it remade its economy around health, education and tech and avoided bankruptcy — showing that economic reinvention, not demography alone, determines the outcome.
9.0m vacant homes (13.8% of stock, 2023), with Wakayama and Tokushima at 21.2% vacancy and ~59% of vacant homes inherited (secondary); population projected from ~128m at peak to 107m by 2040 and <100m by 2056 (IPSS 2023 medium). More than half of municipalities are designated depopulated; a 2017 projection (Sakakibara, via STH Press) put 896 towns and villages as non-viable by 2040 — flagged as a secondary, dated projection, not a government forecast. The fiscal mechanism differs from the US: Japanese local tax includes a fixed-asset (property) tax, but municipalities depend heavily on central transfers (the Local Allocation Tax), so depopulation is fiscalised nationally. Costs show up as rising per-capita infrastructure costs; the 2015 Vacant Houses Act (revised 2023) letting municipalities strip the residential-land tax reduction — up to a 6× tax rise — and order demolition (the state raising tax on vacant homes to force action, the opposite of a revenue windfall); and municipal mergers — the Heisei mergers cut municipalities from ~3,200 to ~1,700 — as the main adaptation.
Population ~59m → ~54.7m by 2050 (ISTAT median), with the South −15% (2025–50) against the North ~−1%; births have been below deaths for over a decade and immigration “can no longer fully reverse the trend” (secondary ISTAT summary). In inland Campania, municipalities are projected at −10% to −20% over 2020–30, yet built-up area rose 14.5% in Benevento province over 2010–21 while population fell 6.8% — sprawl with shrinkage, raising per-capita service costs (OECD, Preparing for Demographic Change in Campania, 2025). Responses — €1 homes (Salemi, from 2008), a 7% flat tax for foreign retirees in small southern towns (2019, 10 years), mountain-residency grants — are demand-subsidy policies: no evidence was found that they have reversed depopulation at scale, so they are best read as evidence the fiscal problem is recognised, not solved. Italian municipalities are also heavily transfer-dependent, so the crisis appears as service withdrawal and state-funded incentives, not bankruptcy.
Has anyone documented a municipal fiscal crisis driven by demographic housing decline alone? No clean case was found. Detroit, Prichard, Stockton, Vallejo and Central Falls are all confounded by industrial decline, housing-bust leverage or governance failure. Japan and Italy are the pure-demography cases — and they produce gradual fiscal strain + central transfers + municipal mergers, not bankruptcy, because their intergovernmental systems socialise the loss. The US system — local property-tax dependence + state caps + Chapter 9 — is more likely to produce discrete local crises than Japan's or Italy's, but that is an institutional inference, not a modelled result. Policy dampeners exist: diversifying local revenue away from residential property tax (land-value tax, sales/income tax, state aid) would break the loop but requires state action — Lincoln Institute work on split-rate taxation in Detroit (Anderson & Allen, 2022; index-level only in this research) is the closest applied example — and managed shrinkage (Japan's mergers; Youngstown's “planned shrinkage” and land banks) reduces per-capita costs without restoring values.
Gradual (the modal outcome): rates rise to hold levies, services are trimmed, schools consolidate, maintenance is deferred and real per-capita costs rise — Japan and Italy today, and much of the US Rust Belt and rural counties. A slow crisis, visible in budgets and service quality, not in a bankruptcy filing. Discrete (the tail outcome): a Chapter 9 filing, pension non-payment (Prichard, 2009) or service collapse (Detroit water shut-offs, 2014–20, affecting up to 141,000 residents) occurs when three conditions coincide — (i) sustained population and value decline, (ii) a large fixed legacy burden (pensions, OPEB, debt), and (iii) a binding cap or political inability to raise revenue. Demographic housing release alone is unlikely to trigger (iii); it needs the legacy burden already present. Watch the intersection, not demography alone — the watchlist rows in §11 (property-tax base, delinquency and effective rates in release-heavy metros; pension funded ratios in shrinking jurisdictions; school enrolment in shrinking districts) are the early-warning set.
Your hypothesis — that second / vacation-home ownership is primarily Boomers — is confirmed in its lifecycle and wealth form, but not as a majority claim. Second-home ownership is overwhelmingly an older, richer person's activity, peaking at ages 55–74, and Boomers hold the largest share of all housing wealth. But no primary source found publishes a direct generational share of the second-home stock, and in the most recent financed buying flow (2025), Gen X, not Boomers, is the plurality. In the short-term-rental buyer pool, Millennials lead. The distinction between stock, flow, financed vs cash, and vacation vs investment is decisive — and most popular claims conflate them.
Ownership rate by age: peaks at 55–74 — Boomer + late-Gen X ages; dominance confirmed as an age / lifecycle effect. Share of all housing wealth: Boomers hold ~41% of all US household real estate value (Fed DFA, Q1 2026, via secondary reporting) and ~$19 trillion in housing wealth. Share of all home buyers / sellers (NAR 2026, primary residences only): Boomers 42% of buyers, 55% of sellers — the largest generation, confirmed, but not a second-home share. Share of financed second-home purchases (Redfin 2025): Boomers ~27–29% estimated — not the largest; Gen X ~46% estimated. Share of vacation-rental buyers (Vacasa 2022): Boomers 18% — Millennials lead at 43%. Share of the second-home stock by generation: not published anywhere found — a majority claim cannot be verified.
Shares are not directly comparable across sources, because the generation boundaries move:
| Source | Silent | Boomers | Gen X | Millennials | Gen Z |
|---|---|---|---|---|---|
| SourceNAR 2026 Generational Trends | Silent1925–1945 | Boomers1946–1964 | Gen X1965–1979 | Millennials1980–1998 | Gen Z1999–2011 |
| SourcePew / Fed DFA (common usage) | SilentBefore 1946 | Boomers1946–1964 | Gen X1965–1980 | Millennials1981–1996 | Gen Z1997–2012 |
| SourceFed DFA generational wealth | Silent“Silent and earlier,” before 1946 | Boomers1946–1964 | Gen X1965–1980 | Millennials + Gen Z1981 or later — Millennials and Gen Z lumped together | |
Consequence: any figure labelled “Millennials” from the Fed DFA includes Gen Z. Redfin's HMDA work reports only age bands (under 35, 35–44, 45–54, 55–64, 65–74, 75+), never generations — the generational splits for Redfin below are estimates from those age bands, assuming uniform distribution within a band, and are labelled as such wherever they appear.
There is no single official count — the answer moves by 2–3 million depending on definition:
NAR publishes no stock count at all — only sales flow: vacation-home sales rose +16.4% in 2020 and were 5.5% of all existing-home sales in 2020, rising to 6.7% in Jan–Apr 2021 (NAR 2021 Vacation Home Counties Report). Its monthly Confidence Index put sales for non-primary use (vacation or rental) at 16% of sales in Dec 2024, of which vacation use was 4% — use annual figures, not single volatile months.
What exists for the stock is ownership rates by age band — how likely a family of a given age is to own a second / other property:
Redfin analysis of HMDA data, published 28 Jul 2026 — mortgaged second-home purchases only, 2025
Ages 45–64 combined took 58.7% of second-home mortgages in 2025. The 2024 shape was the same (55–64: 30.2%; 45–54: 28.4%; 35–44: 20%; 65–74: 12.5%). The financed-buyer distribution implies a median in the 55–64 band — cumulative share reaches 51.6% below age 55 — and cash buyers would push the all-buyer median higher.
| Generation (2025 ages) | Estimated share of financed second-home purchases, 2025 |
|---|---|
| GenerationSilent (80+) | Estimated share~1–2% |
| GenerationBaby Boomers (61–79) | Estimated share~27–29% |
| GenerationGen X (45–60) | Estimated share~46% |
| GenerationMillennials (29–44) | Estimated share~21% |
| GenerationGen Z (18–28) | Estimated share~2–3% |
These generational splits are the researcher's estimates from Redfin's age bands (Pew / NAR boundaries, uniform distribution within 55–64) — not a primary-source figure. They cover mortgaged purchases in one year only and exclude cash purchases — which were 57% of vacation-home buyers in Jan–Oct 2025 (NAR Confidence Index via Realtor.com) and 53% in Jan–Apr 2021 (NAR). Cash buying skews much older: NAR 2026 reports 39% of younger Boomers and 46% of older Boomers buying a home paid all cash, and repeat cash buyers' median age is 68 vs 58 for financed repeat buyers. The true Boomer share of all second-home purchases is therefore materially higher than ~27–29%, and the true Gen X share lower than ~46% — but by how much cannot be computed, because NAR does not publish a generation × vacation-use cross-tab.
NAR's 2026 Generational Trends report (purchases Jul 2024–Jun 2025) has Boomers at 42% of all buyers and 55% of all sellers — but that survey covers primary residences only. It proves Boomers dominate housing transactions overall; it says nothing directly about second homes, and it is routinely misquoted as if it did. First-time buyers were 21% of all buyers in that report — the lowest since NAR began tracking in 1981 — with a record median first-time buyer age of 40.
One clean test cuts the other way: in 2024, Boomers were the only generation whose financed second-home mortgages rose (+4.5% at 65–74; +8.6% at 75+, Redfin) while Gen X and younger cohorts fell — consistent with cash-rich older buyers being rate-insensitive. There is no published “Boomers net bought vs net sold second homes” figure: NAR does not publish second-home sales by generation of seller, and the NAHB stock count does not split by owner age. Flagged as unverified. Intentions point to holding, not selling: in a Clever Real Estate survey of 1,000 Boomers (Jul 2025), 61% said they “never” plan to sell their home (up from 54% in 2024) and only 10% expected to sell in the next five years — for the primary home, but the same lock-in logic applies more strongly to second homes. AARP (2024): 75% of adults 50+ want to stay in their current home as they age.
No published share of second homes owned outright (stock) was found. The proxies:
The vacation-rental / short-term-rental buyer pool skews far younger — and it is a different population:
2020–22: a younger, more investment-motivated surge — in flow, not stock. Vacation-home sales rose +16.4% in 2020; Redfin's HMDA second-home share of mortgages peaked at 5.1% in 2021 (5.0% in 2020), driven by sub-3% rates, remote work and affluent households decamping to vacation destinations. Millennial / Gen X shares clearly rose in the buyer flow — most measurably in the STR / investor segment (Vacasa: 44% → 51% Gen Z + Millennials, 2021 → 2022; Zillow: Millennials 31% → 40% of vacation buyers, 2019 → 2021). Open question: Redfin HMDA age shares by generation for 2021 were not published in the sources retrieved, so no whole-market generational share for 2020–22 second-home purchases could be verified — the Vacasa / Zillow figures are survey / platform-specific and skew young and investor. The pandemic surge was disproportionately younger than the traditional market, on a stock still dominated by older owners; both facts coexist.
After 2022: collapse, then a small rebound. Second-home mortgages fell −42% in 2022, −40% in 2023 and −5% in 2024 — to 86,604, the lowest since Redfin's records began in 2018, at 2.6% of all mortgages — then rose +4.1% in 2025 to ~90,413 (2.7% of mortgages). Even after the rebound, financed buying is still ~65% below the 2021 peak, and 2024 purchases were about half pre-pandemic levels. NAR's all-sales share for non-primary use fell from a 22% peak (Jan 2022) to 16% by Aug 2024 — still above the 14% average of late 2015 to mid-2020. In the hotspots the volume fall was far steeper: Hilton Head −83% and Lake Havasu City −87% from early 2019 to early 2024 (Optimal Blue via Reuters). Drivers: higher rates, the 2022 FHFA rise in second-home loan fees, record prices, return to office, and cooler rental markets.
Release: the 13.9m figure is all homes, not second homes. Realtor.com (Oct 2026, §5) projects 13.9m homes released from Boomer + Silent owner-occupancy in 2026–36 — but no published estimate splits those releases into primary vs second homes, or into sold vs inherited vs rented. Any second-home-specific turnover number would be a new calculation, not a citable forecast. Flagged as a gap. “Released” also ≠ listed for sale.
Inheritance is already at a record — and still small. 340,000 homes changed hands through inheritance in the year to Aug 2025 — 7.4% of all transfers, the highest share on record (Cotality, via Realtor.com) — against ~30m homes owned by householders 65+. Inheritance, not listing, is the mechanism by which second homes will mostly change generation: Cerulli Associates (via Realtor.com / Coldwell Banker) puts $124tn transferring through 2048 (~$105tn to heirs), ~$25tn of it tied to real estate, with Gen X + Millennials to inherit $4.6tn in global real estate wealth over the next decade, $2.4tn of it in the US.
What heirs do — with an intention gap. Trust & Will's 2026 survey (1,000 Americans; 500 past heirs, 500 future heirs; all inherited real estate, 72% of it primary residences): 56% of past heirs sold (26% within 12 months, 30% after), 33% kept the home as a primary residence, 17% kept and rented. Future heirs intend differently — only 35% plan to sell, 36% plan to rent, 35% plan to keep as a primary residence (Millennials most likely to plan to rent, 43%; Gen Z most likely to plan to live in it, 48%). The 19-point intention-to-rent gap (36% plan vs 17% did) warns against taking intentions at face value. No source quantifies the sold / inherited / converted split for second homes: a vacation home is more likely than a primary home to become a listing (the heir usually already has a primary home elsewhere) and more likely to be kept as a shared family asset for sentimental reasons — with carrying costs of roughly $700–$1,500 a month even on a paid-off house (Inherited Property Match 2026 survey of 1,350 heirs) and disagreements in 76% of multi-decision-maker cases pushing toward sale.
Expect a trickle, then a gradual swell — not a cliff. Vacation and second-home markets will feel Boomer release first and hardest, but unevenly: they are discretionary, rate- and insurance-sensitive, and already correcting. The released mix is wrong for the shortage — of Realtor.com's 13.9m all-home releases, only 0.38m are starter homes, against 9.9m family homes and 3.6m large homes — so the likely pattern is more listings in vacation markets than the national average, gradual softening in family / large homes and slower-growing markets, and little relief at entry level. And the second-home stock itself may keep shrinking before it grows: NAHB's count already fell from 7.15m (2020) to 6.2m (2024), and if heirs sell or convert inherited second homes to primary / rental use faster than new ones are bought — financed buying is still ~65% below its 2021 peak — the stock falls even as total Boomer releases rise.
| Country | Second homes | Share of stock | Age pattern |
|---|---|---|---|
| CountryUnited States | Second homes6.2m (2024, NAHB / ACS) | Share of stock4.3% | Age patternOwnership rate peaks at 55–64 (SCF); financed buyers aged 45–64 took 58.7% (2025) |
| CountryFrance | Second homes3.7m (2025, INSEE via Le Monde) | Share of stock9.8% | Age patternTwo-thirds owned by people over 60; 15.1% of 80–89-year-olds own one vs 1.7% of 30–39-year-olds |
| CountryItaly | Second homesNo verified second-home count; over 9.5m vacant dwellings | Share of stock~27% vacant — not second homes | Age patternNo age breakdown verified |
| CountryJapan | Second homesNo verified second-home count; 9.0m vacant dwellings (2023) | Share of stock13.8% vacant | Age patternAkiya (abandoned homes) concentrated in ageing rural areas; heirs renounce inheritance |
France is the best comparison (Le Monde, 17 Feb 2026, on INSEE / SDES data): 3.7m second homes, 9.8% of housing, stable for ~10 years, ~40% coastal and 16% mountain. The generational skew is, if anything, stronger than in the US — but the mechanism differs: only 34% are owned by top-decile affluent households, and many were inherited from parents / grandparents who bought in the 1960s–70s, rather than market-purchased. Younger French owners are more rental / investment-oriented, and very elderly owners often neither sell nor rent, leaving homes empty. Japan's 9.0m vacant homes are a different phenomenon — abandonment, not second-home ownership (§2); of the 2023 total, 3,856,000 are “other vacant” — not for rent, not for sale, not second use. Italy's ~27% vacancy likewise includes second and holiday homes and is not directly comparable to Japan's akiya (§3). The wider European evidence points the same lifecycle way: the Survey of Health, Ageing and Retirement in Europe (SHARE, wave 7, 2017) found 23.9% of people 50+ in 12 European countries owned “other real estate,” concentrated among older households — but there is no Eurostat EU-wide second-home ownership share, and no comparable INSEE-style age breakdown was found for the UK, Germany, Spain or Italy in this research.
Bottom line: older dominance of second homes is a general feature of rich, ageing societies — France confirms it independently. The US is distinctive in that second homes are a smaller share of stock (4.3% vs France's 9.8%), more market-purchased and income-selected, and more entangled with the STR / investment market. And the cohort now buying with a mortgage is Gen X (45–60) — Boomers' successors ageing into the peak second-home ages, exactly what a lifecycle model predicts as Boomers age out of buying and into holding and selling.
Sources: NAHB Eye on Housing, “Count of Second Homes Declines in 2024” (Apr 2026); Census Bureau, “Most U.S. Vacant Housing Is Seasonal Housing” (May 2023); Redfin, “Vacation-Home Mortgages Tick Up For First Time Since Pandemic Boom” (28 Jul 2026) and 2024 second-home mortgages (8 May 2025); NAR, 2026 Generational Trends release (15 Apr 2026); Federal Reserve, Changes in U.S. Family Finances from 2019 to 2022 (SCF); JCHS Belsky, Di & McCue (2006), W06-5; Vacasa 2022 via Inman (6 Jul 2022); Trust & Will, Real Estate Inheritance report (2026); Clever Real Estate (Jul 2025); Le Monde, “Who are France's second-home owners?” (17 Feb 2026). Deep research 9 Nov 2026; every hedge, estimate label and open question from that research is carried in the text and in the caveats below.
A different hypothesis from the demographic one in §§1–9: if house prices stagnate for a long period while the cost of owning keeps climbing — property taxes, homeowners insurance, HOA fees, maintenance — buying becomes unattractive next to renting; fewer transactions then squeeze local-government revenues (transfer and recording taxes first, the property-tax base eventually) and destroy jobs across mortgage origination, title and escrow, real-estate law, banks and brokerage teams, feeding back into weaker demand. A dedicated research pass across think tanks, universities, policy institutions and Federal Reserve papers (9 Nov 2026) finds: the premise is largely true on costs and on the transaction freeze — but no established model calls this a “housing doom loop,” and the evidence to date points to gradual stagnation with severe localised pain, not a national self-reinforcing collapse, unless several breakers fail at once.
Extensive searching (“housing doom loop,” “housing vicious cycle,” “housing feedback loop,” plus Fed / think-tank / academic variants) found no paper, Fed working paper or think-tank report that uses “doom loop” — or an equivalent named model — for: flat prices + rising carrying costs → fewer buyers → fewer transactions → lower local revenue + job losses → lower prices and demand. The one named housing doom loop found is a different loop: Apollo chief economist Torsten Slok's “higher rates, higher rent doom loop” (client note reported by Fortune, 5 Oct 2026) — high rates → less building → higher rents → higher inflation → rates stay high. (A counter-view in the same coverage, from Realtor.com's Joel Berner: “It's really only a ‘doom loop’ if there are no forces acting against it” — rents were falling at the time, not rising.) The Apollo primary note was not opened in this research. “Doom loop” is also used for downtown commercial-real-estate / transit fiscal loops — again a different mechanism. The closest established models for your mechanism are Poterba's user-cost framework, the Fed / FHFA mortgage lock-in literature, and the shrinking-city fiscal-feedback literature (§8).
The standard framework is Poterba's user cost (1984, Quarterly Journal of Economics): in equilibrium, rent = user cost × price, where user cost ≈ cost of capital + property tax + insurance + maintenance / depreciation + risk premium − expected capital gain. If expected appreciation falls to ~zero (flat nominal, falling real) while tax, insurance, HOA and maintenance rise, user cost rises mechanically even if the mortgage rate never moves — and where renting is cheaper than user cost, demand for ownership falls until prices adjust. In the model that is a level adjustment, not by itself a loop: a loop needs a second feedback (falling prices → eroded tax base → worse services → lower demand, or falling turnover → job and income loss → lower demand). The components have moved exactly as the hypothesis requires:
CBRE, Q2 2025 — average monthly cost of buying the median home vs average apartment rent
A 108% buying premium — down from 128% at year-end 2023, but far above the 68% pre-pandemic average; in Orange County the premium is 303% (vs 160% in 2019). Only 12.7% of renter households could afford the median home, down from 17.0% in 2019 — 1.8m households priced out in five and a half years — and a 20% down payment equals four years of average apartment rent. JCHS tells the same story from the payment side: total monthly cost on the median home was $3,120 at end-2025 (mortgage $2,420 plus tax and insurance), against $1,240 mortgage-only at end-2020; the income needed is $120,800. Renting is now cheaper on monthly cash flow in most markets — but renters are also historically cost-burdened (49% of renters, 22.7m households, in 2024 — JCHS), so the “motivation to rent” is a lesser-of-two-evils choice, not relief. The buy-vs-rent decision now hinges on horizon, expected price growth and carrying-cost growth — exactly the user-cost terms above.
Existing-home sales, millions per year — NAR via secondary summaries (the NAR primary release was not opened in this research; the figures are consistent across multiple sources)
The seasonally adjusted pace fell from 6.43m in Jan 2022 to 3.91m in Jan 2026 — a ~40% volume collapse with no national price crash. The mechanism is lock-in, and it is measured: each 1pp gap between an owner's mortgage rate and the market rate cuts the odds of selling by 18.1%, and lock-in prevented an estimated 1.33m sales from mid-2022 to end-2023 while adding ~5.7% to prices (FHFA WP 24-03, via secondary summary — primary not opened). A Federal Reserve paper (FEDS 2024-088r1, Apr 2025) attributes 44% of the 2021–22 mobility drop to lock-in, and finds the shock raised prices ~8% by tightening the market. Overall residential mobility hit a record-low 11.2% in 2024, homeowner mobility just 5.1% (JCHS); months' supply was 3.3 months in Dec 2025. Lock-in therefore cuts against the price half of the loop: higher rates removed sellers faster than buyers, so volume — not price — took the adjustment. A stagnant-price loop needs the supply side to unfreeze (forced sales from insurance / HOA / tax distress, or the Boomer release in §5) while demand stays weak. No clean elasticity of sales to carrying costs (tax / insurance / HOA) was found in the primary literature: the Dallas Fed links insurance to relocation (+0.54pp per $1,000 premium rise) and delinquency, not to sales volume directly.
Property tax is ~28.4% of local funding nationally (2024) and ~70% of local tax collections (Lincoln Institute) — but it is a levy system (§8): budgets set the dollars first and the rate is derived, so when values stall, rates rise to hold the levy, with a lag. The first-round effect of stagnation is therefore higher bills and political revolt, not revenue collapse — and reassessment lags (1–4-year cycles) can even raise bills during stagnation as assessments catch up to peak values, fuelling appeals and delinquency. What actually binds: assessment caps (18 states; California's Prop 13 is the extreme, §8), delinquency and abandonment in already-distressed places, and commercial-value collapse shifting the burden onto homes — Chicago's median residential bill rose ~16.7% in tax year 2024 as ~$469.4m of burden shifted from commercial to residential.
Transaction-based revenue — transfer taxes, mortgage recording taxes, recording fees — is where falling volume bites immediately, but nationally it is much smaller than the property tax and concentrated in specific cities and states. New York's MTA real-estate-related taxes fell from $1,281m (2022) to $696m (2024) and, per the State Comptroller (May 2025), have “never fully rebounded to levels prior to [the] Great Recession”; Philadelphia's realty transfer tax ran ~26% under projection, a ~$110m shortfall in the year to June 2024 (secondary). No national total or share for transfer-tax / recording-fee revenue was verified from a primary source — do not quote one. Exposure varies: Texas (no state income tax; ~1.36–1.40% effective property-tax rate; $125.05bn levied in 2024, up from $49bn a decade earlier) is more exposed to value stagnation but also more able to raise rates — the state has spent $51bn over two years on relief (homestead exemption now $140,000), so the constraint is political, not mechanical. Florida local revenue faces the condo channel below; no Florida local-revenue collapse was quantified in the sources found — flagged as a gap.
| Segment | Count / change |
|---|---|
| SegmentMortgage industry employment (BLS, via MBA) | Count / change418,200 peak (2021) → −36% by Q2 2025 |
| SegmentLicensed loan originators (NMLS, via MBA) | Count / change~125,000 (2021) → ~82,500 (2025), −34% |
| SegmentProducing loan officers (RETR / NMLS) | Count / change287,728 (2021) → 226,155 (2025) — secondary |
| SegmentNAR members (REALTORS®) | Count / change1,580,971 peak (end-2022) → 1,438,569 (Jun 2026), −9% — secondary |
| SegmentBrokers & sales agents (BLS jobs, 2025) | Count / change530,600 — via secondary |
| SegmentTitle / escrow / real-estate law / appraisers / inspectors / movers | Count / changeNo consolidated national count found — gap |
MBA production staffing per company fell 43% over four years, and big lenders cut hard (Wells Fargo cut 500+ mortgage bankers in Feb 2023 as originations fell 57% YoY) — but the MBA (Sep 2025) reports industry employment has stabilised over the past six quarters, with originations forecast to rise through 2027. For scale, 2008–2010 was far worse: construction lost 1.5m jobs from Dec 2007 to Jun 2009 (−19.8%, the largest percentage decline of any sector; BLS), and NAR membership fell 26% from 2006 to 2012 — but that required a ~30% national price crash, a credit freeze and a foreclosure wave (2.8m+ filings at peak vs ~360–400k/yr now). No credible estimate of jobs at risk in a prolonged stagnation (as opposed to a crash) was found: the MBA's 2022 implication of −25% to −30% mortgage employment peak-to-trough has largely already happened, and any number beyond the observed −36% mortgage / −9% NAR would be speculation. Title, escrow, law, appraisal, inspection and moving revenue is transaction-linked by construction — direction down, magnitude unquantified in the sources found.
The one segment where the full loop is observable today. Master insurance policies more than doubled in 2022–2024; post-Surfside law mandates milestone inspections and fully funded structural reserves; condo assessments rose 96% in four years (Vantaca); special assessments of $50,000+ per unit land on owners — many of them older, on fixed incomes — who cannot pay; and financing has been withdrawn, with Fannie Mae's condo blacklist covering over 1,400 Florida buildings. Condo pending sales fell 21% YoY in Jun 2025 while inventory rose 37%, to 13.2 months' supply; Florida prices are falling — FHFA's Florida index −2.3% YoY in Q3 2025 against +2.2% for the US, Tampa among the weakest metros nationally. The loop in this segment:
Carrying costs ↑ → owners who can't pay are forced to sell → supply ↑ → prices ↓ → the building's fixed costs (insurance, reserves, maintenance) spread over fewer solvent owners → assessments per remaining owner ↑ → more forced sales. Regulatory and financing tightening is the amplifier that turns a cost squeeze into forced sales — the most plausible tipping mechanism in the whole hypothesis. A second amplifier is dated: Fannie Mae reserve requirements tighten to 15% in Jan 2027.
| Policy / force | Evidence in the research | Assessment |
|---|---|---|
| PolicyProperty-tax relief / caps | Evidence34 states passed 115 property-tax-cut laws since 2020 (NACo); Texas spent $51bn on relief in two years. Florida / Ohio abolition proposals would require roughly doubling the sales tax to replace the revenue (Tax Foundation) | AssessmentCaps protect incumbents but lock in turnover, create inequities and shift burden to newcomers and renters; ITEP (Jul 2025): limits “led to reduced local services, instability in local finance, increased reliance on regressive taxes.” Income-based circuit breakers are the more targeted alternative cited by Lincoln / ITEP — no outcome evaluation found |
| PolicyInsurance reform / FAIR plans | EvidenceFlorida's SB 2-A (Dec 2022) ended one-way attorney fees and assignment of benefits; litigation fell 25% in H1 2025. Citizens / FAIR Plan growth shows the state backstop is the de facto absorber | AssessmentEarly signs of litigation falling in Florida, but premiums remain structurally high; FAIR plans cost more and cover less — they socialise risk rather than reduce it. No evidence reforms have restored affordability at scale |
| PolicyZoning / supply | EvidenceUrban Institute review: construction rises long-term in upzoned areas; Minneapolis 2040 saw a multifamily response and rent declines; Austin added ~120,000 homes 2015–24 (+30% of stock) with median rents −4% vs +10% nationally (Pew, 2025) | AssessmentStrongest evidence base for long-run affordability, weakest for short-run loop-breaking: building takes years |
| PolicyMortgage-rate relief | EvidenceLock-in is rate-gap driven (FHFA / Fed above), so lower rates would unfreeze turnover fastest; a $10,000 seller tax credit (Biden-era proposal) is modelled in the Fonseca et al. lock-in work, effect not quantified in the summary retrieved | AssessmentRates are set for inflation, not housing — and Apollo's loop (§10, top) warns that if scarce supply re-accelerates rents, rate relief is self-defeating |
| PolicyFirst-time buyer assistance | EvidenceNo specific evaluation retrieved | AssessmentGap — do not claim effectiveness either way |
| PolicyMunicipal revenue diversification | EvidenceLincoln / Tax Foundation: replacing the property tax requires more than doubling the sales tax (or tripling the income tax, in the Ohio example); sales taxes are more volatile and regressive | AssessmentFiscally hard; no US city has replaced the property tax at scale |
| PolicyLand banks / managed shrinkage | EvidenceDetroit Rehabbed & Ready: sale prices +11.5%/yr and mortgage purchases +5.6%/yr in programme neighbourhoods (2021); Cleveland demolition reduced foreclosure and raised nearby equity in functioning submarkets; Detroit has demolished 20,000+ buildings since 2014 and still has ~175,000 vacant properties | AssessmentWorks locally and slowly; does not restore a tax base without jobs and population (§8) |
Amplifiers to watch: insurance non-renewals (+96–216% already); HOA / reserve forced assessments and the Fannie Mae condo blacklist; commercial reassessment shifting tax burden to homes (Chicago, live); and above all a jobs shock — the ingredient whose absence in 2022–2025 is the main reason volume collapsed but prices did not.
As a national loop: not established, and not the base case in the evidence. As a local / segment loop: yes — already observable in Florida condos, and in high-tax / high-insurance / shrinking markets. The national stabilisers: fixed-rate mortgages (~half below 4%, so no payment shock forces sales); $34tn of homeowner equity with ~40% of homes owned free and clear; no distressed-selling wave (foreclosures ~360–400k/yr vs 2.8m+ at the 2008 peak); levies that adjust rather than collapse with values; and a 3.7–4.9m unit shortage that supports prices even as turnover collapses. Detroit (§8) is the completed analogue — but it took deindustrialisation and population loss over decades, not a housing cycle, and Chicago is a live fiscal variant driven by commercial reassessment, not residential stagnation.
Tipping into a discrete national loop would need several of these together: (a) a jobs / income shock that creates distressed sellers — the missing ingredient of 2022–2025; (b) insurance withdrawal at scale, if FAIR / Citizens capacity or pricing breaks or a major hurricane / wildfire season triggers mass non-renewal; (c) the condo / HOA cliff spreading beyond Florida (Fannie reserve rules tighten Jan 2027); (d) commercial reassessment spreading beyond Chicago; (e) the Boomer release (§5 — householders 80+ roughly double by 2045 on JCHS projections) flooding supply while young demand is weak; and (f) rent re-acceleration (Apollo's loop) forcing the Fed to keep rates high while building stays low — completions were already down 27% YoY in Aug 2026 (secondary). Any one of these is localised pain; several together, plus a recession, is the discrete-loop scenario. No source quantifies the joint probability — that is speculation, not a finding. The §11 watchlist carries the loop's early-warning variables: sales and months' supply, the rate gap, insurance premiums and non-renewals, realised (not projected) delinquency, tax bills vs incomes, Florida condo prices and inventory, starts and completions, household formation, and foreclosures.
Sources: Harvard JCHS, The State of the Nation's Housing 2026 — Key Facts; CBRE, “Fewer Renter Households Can Afford Homeownership” (15 Sep 2025); Federal Reserve Bank of Dallas, “Home insurance premiums influence mortgage delinquencies, relocations” (24 Mar 2026); NAIC, first national homeowners-insurance analysis (5 Aug 2026); Federal Reserve Board, FEDS 2024-088r1, “Locked In” (Apr 2025); MBA, Chart of the Week: Mortgage Industry Employment (2 Sep 2025); ATTOM, 2024 property tax analysis; Zillow / Thumbtack, “Hidden costs of homeownership reach $16K per year” (13 Nov 2025); Vantaca 2026 via ConsumerAffairs (Nov 2026); NY State Comptroller, MTA operating revenue trends (May 2025); Lincoln Institute, “Understanding Property Tax Revolts”; BLS Monthly Labor Review, construction employment in the recession (Apr 2011); Apollo / Slok “higher rates, higher rent doom loop” via secondary reporting of Fortune (5 Oct 2026). Deep research 9 Nov 2026; hedges, secondary-source flags and open questions are carried in the text and the caveats below.
Twenty-four variables to track over the next 5–10 years — the original eleven, plus seven added by the synchronized-ageing and municipal-finance follow-up (§§7–8), two added by the second-homes research (§9) and four added by the doom-loop research (§10) — each with the threshold that would support the Japan hypothesis — and the one that would kill it. The early-warning combination that matters most: immigration ~0.3m and household growth <700k and months' supply >6 and real prices negative for two years running — and, for the fiscal channel, a falling tax base and rising effective rates and rising delinquency in the same metro — and, for the stagnation loop in §10, sales stuck near 4m and realised insurance-driven delinquency climbing and Florida condo supply still rising.
| Variable | Source / frequency | Would confirm (Japan-like) | Would refute |
|---|---|---|---|
| VariableNet international migration | SourceCensus Vintage estimates (annual, Jan); CBO outlook | ConfirmsStays ~0.3m/yr for 3+ years; immigrant household growth <200k/yr | RefutesRebounds >1m/yr; immigrant household growth back >500k/yr |
| VariableHousehold growth | SourceCensus Housing Vacancy Survey (quarterly); JCHS | ConfirmsFalls to <700k/yr and stays; 2035–45 pace <600k/yr | RefutesHolds >900k–1.1m/yr; young-adult formation recovers |
| VariableNatural increase | SourceCensus / CDC births–deaths | ConfirmsDeaths > births before 2030 (CBO date), gap widening fast | RefutesNatural increase stays >500k/yr into the 2030s |
| VariableFertility (TFR) | SourceCDC / NCHS (annual) | ConfirmsFalls to <1.5 and stays there | RefutesStabilises ~1.6–1.7 |
| VariableBoomer releases vs listings | SourceRealtor.com Generational Succession (annual); ACS | ConfirmsReleases >1.5m/yr before 2036 and listings surge in the same metros; starter releases rise | RefutesReleases stay ~1.3m/yr, absorbed; ageing-in-place retention rises |
| VariableSecond-home stock & ownership rate by age | SourceNAHB / Census ACS (annual); Fed SCF (triennial) | ConfirmsStock keeps falling from 6.2m / 4.3% (2024) as older owners release faster than Gen X buys; ownership-rate peak shifts into the 65–74 band | RefutesStock stabilises or rises; the 55–64 ownership rate (20.4%, SCF 2025 via secondary summary) holds as Gen X ages in |
| VariableSecond-home buying flow — financed vs cash | SourceRedfin HMDA (annual); NAR Confidence Index via Realtor.com | ConfirmsFinanced purchases stay ~65% below the 2021 peak and the cash share (57% of vacation buyers, Jan–Oct 2025) falls as older buyers exit — vacation-market listings rise | RefutesFinanced purchases recover toward pre-pandemic levels with the 45–64 share (58.7% in 2025) broadening to younger buyers |
| VariableMonths' supply (national) | SourceNAR Existing-Home Sales (monthly) | Confirms>6 months sustained nationally, not just the Sun Belt | RefutesStays <5 months; inventory <1.5m homes |
| VariableSun Belt inventory / vacancy | SourceRealtor.com; CoStar multifamily vacancy (quarterly) | ConfirmsMultifamily vacancy >9%; single-family months >7 in FL/TX/AZ; falls spread beyond the Sun Belt | RefutesVacancy <7%; rents re-accelerate; glut clears in 12–18 months |
| VariableMortgage lock-in unwind | SourceFHFA National Mortgage Database (quarterly); Freddie Mac PMMS (weekly) | Confirms30-yr rate falls to 5–5.5% and listings jump >20% YoY without a matching buyer surge | RefutesRates stay 6.5–7.5%, or fall gradually with demand absorbing listings |
| VariableReal house prices | SourceS&P CoreLogic Case-Shiller (monthly), deflated by CPI | ConfirmsNegative real growth for 3+ consecutive years nationally; nominal flat/down with CPI +2–3% | RefutesReal growth >1%/yr; nominal +3–4%/yr |
| VariableRegional divergence | SourceCensus state/county estimates; Case-Shiller metros | Confirms>10 states in outright decline; >25% of counties with falling prices; Rust Belt joins Sun Belt | RefutesDecline confined to 5 states / rural counties; growth metros keep +3%+ |
| VariableJapan control | SourceMLIT Land Market Value Publication (annual, Mar); Tokyo RPPI | ConfirmsTokyo land/houses resume falling; regional falls accelerate — even metros not immune | RefutesTokyo/Osaka keep rising; only rural Japan falls — confirms “metros resilient” |
| VariableForeign-buyer volume & China share | SourceNAR International Transactions (annual, Jul; Apr–Mar year) | ConfirmsFalls well below $56bn; the China line ($13.7bn) shrinks — capital-flow reversal in real time | RefutesRecovers toward the $153bn / ~284,500-home 2016–17 peak; non-resident share rises |
| VariableGlobal real rates / neutral rate (r*) | SourceFederal Reserve; IMF World Economic Outlook | ConfirmsDemographic reversal: r* and real rates rise as fast-ageing countries dissave (Goodhart–Pradhan) | RefutesAgeing keeps pushing r* down (Auclert et al.) — cheap mortgages keep cushioning prices |
| VariableNet capital flows from Japan / Germany / China | SourceBIS; IMF Balance of Payments | ConfirmsFast-ageing countries stop being net capital exporters — the Börsch-Supan reversal arrives | RefutesThey keep exporting capital at scale, supporting US asset demand |
| VariableEU / China working-age population & dependency ratios | SourceEurostat; UN WPP updates (watch the WPP 2026 revision) | ConfirmsPeaks and ratios revised earlier / steeper; EU peak (2026) confirmed and decline accelerates | RefutesPeaks revised later on higher immigration; dependency rises flatten |
| VariableProperty-tax base, delinquency & effective rates in release-heavy metros | SourceCounty assessors; Lincoln Institute — Tampa, Miami, Orlando, Tucson, Dayton, Cleveland, Detroit, Pittsburgh | ConfirmsBase falling and rates rising and delinquency rising together — the §8 loop engaging | RefutesBase stabilises; rates flat; delinquency normal — levies absorb the drag |
| VariablePension funded ratios & OPEB in shrinking jurisdictions | SourcePublic Plans Database; Lincoln Institute | ConfirmsFunded ratios deteriorate where population falls; retiree-to-active ratios climb toward Detroit's 2:1 | RefutesShrinking places pre-fund or reform; legacy burden shrinks faster than the base |
| VariableSchool enrolment & per-pupil funding in shrinking districts | SourceNCES; state education agencies | ConfirmsEnrolment and values fall together; closure waves in release-heavy districts precede further value falls | RefutesState aid formulas cushion per-pupil funding; consolidations happen without value spirals |
| VariableHomeowners insurance premiums & non-renewals | SourceNAIC MCAS (annual); Dallas Fed; FAIR plan / Citizens enrolment | ConfirmsPremiums keep rising well above inflation (already +~70% 2019–25); non-renewals climb beyond the +96–216% of 2018–24; FAIR / Citizens enrolment surges; the Dallas Fed's projected delinquencies (203,000/yr) start appearing in realised data | RefutesPremium growth flattens toward inflation; private insurers re-enter FL/CA; non-renewal rates fall; realised delinquency stays near the 2022 level (~31,000) |
| VariableFlorida condo prices, inventory & special assessments | SourceFHFA HPI; Florida Realtors; Fannie Mae project eligibility | ConfirmsPrices keep falling (already ~−6% YoY in the segment), supply stays above ~13 months, $50k+ assessments spread and the Fannie blacklist grows past 1,400 buildings — the §10 loop engaging; the pattern spreads to other states' condo stock | RefutesPrices stabilise; inventory clears toward ~6 months; the Jan 2027 reserve deadline passes without a forced-sale wave |
| VariableForeclosures & distressed sales | SourceATTOM / CoreLogic (monthly) | ConfirmsFilings climb past 1m/yr and distressed sellers start setting prices — the missing ingredient of 2022–25 arrives, converting the volume freeze into a price loop | RefutesStays near ~360–400k/yr (vs 2.8m+ at the 2008 peak); $34tn of owner equity keeps absorbing shocks |
| VariableTransaction-ecosystem employment | SourceMBA / BLS (quarterly); NAR membership (monthly) | ConfirmsMortgage employment falls another leg from −36% (2021–25) and NAR membership heads toward the 2006–12 pattern (−26%) while sales sit at ~4m/yr | RefutesEmployment stabilises (MBA: stable for six quarters to Sep 2025) and recovers as origination volume normalises toward 2027 forecasts |