Housing Market Predictions: A Calibrated Reading of the Post-2020 Break

The post-2020 run broke affordability, not the banking system. Where 2008 analogies fail, where they quietly hold, and the insurance channel the registry seals at 62% — with ranges, not price targets.

Insurance is how climate risk becomes a price. When the price becomes unavailable, the mortgage market finishes the sentence.

The run, and where it broke

The numbers of the post-2020 run are public and stark. Between 2019 and 2024, the median US home price rose 31% — from $321,500 to $420,300 — while median household income rose 22%; by 2025 the national price-to-income ratio stood at 5.08, against a 2022 peak of 5.83 and a 1980 reading of 3.65. In seven large metros the ratio now exceeds eight; in San Jose it exceeds twelve, a record. Harvard's Joint Center, which compiles these figures, describes them without adjectives, and none are needed: relative to what its buyers earn, US housing has essentially never been this expensive.

What broke was not prices — it was the bridge to them. With the 30-year fixed averaging about 6.5% in June 2026, the arithmetic on a median home implies a monthly payment roughly $680 higher than the same house financed at the locked-in rates half the country still holds, more than $8,000 a year on principal and interest alone. The result is a market frozen rather than falling: owners will not sell out of 3% mortgages, buyers cannot reach 6.5% payments at 2026 prices, and transactions — not values — took the first hit. Affordability did not bend after 2022. It snapped, and the market has spent four years routing around the break.

The current national print reads accordingly: the Case-Shiller National index rose just 0.8% year over year in April 2026 (May's preliminary reading, 1.4%) — a stall, not a crash — with existing-home supply at 4.6 months, the textbook edge of a balanced market, while new-home supply sits at 10.3 months, a genuine builder glut concentrated where builders actually built. A single national housing number is now close to meaningless, because underneath it the country has split into two opposite markets — the subject of the regional section below. First, the comparison everyone reaches for, and mostly shouldn't.

Why the 2008 analogy mostly fails

Start with credit, because 2008 was a credit event wearing a housing costume. The 2004–07 boom ran on subprime and Alt-A originations, teaser-rate ARMs with built-in payment shocks, and documentation standards that collapsed under their own name (“stated income”). The 2020s cycle ran on the opposite: roughly two-thirds of post-2020 mortgage originations went to borrowers with credit scores above 740, the ARM-reset wall essentially does not exist, and the dominant instrument is the fully-amortizing 30-year fixed — often at rates the borrower will never see again. The 2008 crash required millions of loans engineered to fail. The 2026 stock of loans is engineered to be kept.

Second, the lock-in effect — 2008's structural inverse. As of Q1 2026, 49.9% of all outstanding US mortgages still carry rates below 4%, down from over 65% at the peak in early 2022 but still half the market. A homeowner at 3% holding a 6.5% market rate is not a potential forced seller; they are a rationally immobilized non-seller, and foreclosure pipelines remain thin. 2008's downward spiral needed distressed supply clearing at any price. Today's equivalent owner defends the price floor by simply not moving. This is terrible for transaction volumes, mobility and first-time buyers — and it is powerful crash insurance, which is an uncomfortable pair of facts to hold at once.

Third, supply. The 2006 bust landed on a construction overhang; the 2020s run landed on a deficit. How large is genuinely disputed — NAHB counts roughly 1.2 million units, Goldman Sachs about 3 million, Zillow over 4 million, Brookings around 5 million, McKinsey 8 million, and White House economists put it at 10 million or more in April 2026; the span is wide because “shortage” is a definitional choice. But the sign is consistent across every serious method, and it points the opposite direction from 2006. The honest caveats: builders demonstrably can overshoot locally — the 10.3 months of new-home supply is real, and concentrated in Texas and Florida — and on price-to-income the market is more stretched than 2006 ever was. The analogy fails on credit structure. It does not fail on stretch.

The underrated channel: insurance retreat

The channel most 2026 housing content ignores is the one this registry rates most likely to matter. Insurance is how climate risk becomes a price: as long as a hazard can be underwritten, it is a line item; when insurers cannot charge risk-true premiums under rate regulation, they do not argue — they withdraw. The risk then migrates to state backstops built as last resorts: California's FAIR Plan exposure has swollen past $500 billion, roughly triple its 2020 level, and Florida's Citizens has ranked among that state's largest insurers since the private market thinned. The final link is mechanical — lenders require insurance, so an uninsurable property is an unmortgageable one, and its buyer pool collapses to cash.

The loss record explains the retreat. Insured natural-catastrophe losses reached $137 billion in 2024 and $107 billion in 2025 — the sixth consecutive year above $100 billion — with a record 92% of the 2025 total coming from so-called secondary perils: wildfires, severe convective storms, floods. The January 2025 Los Angeles fires alone cost insurers roughly $40 billion, the largest insured wildfire loss ever recorded. The precedent for what happens next is thirty years old: after Hurricane Andrew bankrupted eleven insurers in 1992, Florida's market never returned to private normalcy. What is new is the geographic breadth — wildfire, wind and flood exposure now touches the collateral under a meaningful share of the largest asset class in the United States.

The registry's seal on this channel is CLI-01: 62% (±10) that insurance retreat triggers a climate repricing of housing in a G7 economy within the 2026–2030 window — a probabilistic simulation, sealed July 2026, whose mechanism is actuarial rather than meteorological. The leading indicators run 12 to 24 months ahead of prices: non-renewal rates in exposed counties, FAIR and Citizens enrollment growth, and January reinsurance renewals, which price the world's climate view a year before retail markets see it. A reader who watches those three series is watching the housing risk most likely to produce the headlines of 2028 — and watching it early.

Regional divergence, honestly

There is no US housing market in 2026; there are at least two, moving in opposite directions. As of June 2026, 77 of the roughly 300 largest metros show year-over-year price declines while 223 still show gains. The fall list is led by Punta Gorda, Florida (−7.9%), Cape Coral (−6.1%) and Austin (−5.7%) — pandemic-boom destinations where builders actually responded to demand and where insurance and property-tax burdens have risen fastest. Measured from the mid-2022 national peak, the West is down 7.3% and the South down 3.5%, while the Northeast stands 12.6% higher and the Midwest 10% higher. The average of those numbers describes nowhere.

The mechanics of the split are instructive because they are the article's earlier sections running regionally. Florida and Texas combined elevated new supply with the insurance channel: Gulf Coast metros carry the country's steepest premium escalation, and their declines have insurance arithmetic in the causal chain — CLI-01 operating at metro scale, years before any G7-wide repricing would qualify the sealed entry. The Northeast and Midwest are the mirror image: almost no pandemic overbuilding, the oldest and tightest housing stock, and inventory recovering fastest exactly where prices are still rising (+8.5% and +7.3% year over year respectively) — while national active inventory remains about 11% below pre-pandemic norms.

The honest regional statement, then, has no single direction in it. Sun Belt corrections are real, orderly so far, and concentrated where supply met softening demand and rising carry costs; a buyer's market by most definitions already exists there. Northeast and Midwest scarcity is equally real and shows no supply response capable of ending it this decade. Neither half generalizes. Any sentence that begins “the housing market will…” and does not immediately specify where has already failed before reaching its verb — which is one reason this platform seals channels and conditions rather than a national price call.

Scenarios as ranges, not points

This platform publishes no 2027 national price target, because a point forecast on housing is a confession of method failure: the outcome is the interaction of a rates path, a labor-market path (FIN-02 seals 55% ±15 on a recession beginning by end-2028), an insurance-loss path (CLI-01, 62% ±10) and dozens of metro-level supply positions. What can be stated honestly are conditional shapes, all of them probabilistic simulation. The base path is the least dramatic and the most likely single shape: national nominal prices oscillating in a low-single-digit band around flat while incomes slowly close the gap — affordability repaired by stagnation, over years, the way the market has already been drifting since 2022.

The tails deserve equal print. The downside shape is compound, not singular: a recession arriving into the existing regional corrections, plus an insurance shock — a mega-catastrophe exhausting a state backstop — could convert orderly Sun Belt declines into the sustained 15%+ metro-level repricing that would resolve CLI-01 as CORRECT; that is what the 62% is pricing. The upside shape is the trap scenario: rate relief unlocks locked-in demand faster than any supply response, prices re-accelerate, and the affordability break worsens while looking like a recovery. Note that the political variable then activates — the median first-time buyer is now 38 years old, against 29 in 1981, and the registry separately seals 60% (±15) on the generational wealth rupture becoming an organized political movement by 2032 (SOC-03). Housing policy is not exogenous on any horizon past 2028.

What would change these shapes, in order of leverage: the January reinsurance renewals and FAIR/Citizens enrollment (the insurance channel's early gauges); the lock-in unwind rate — the below-4% share fell from 65% to 50% in four years, and its slope sets how fast normal mobility returns; months-of-supply by region rather than nationally; and the Sahm-rule gauges that front-run FIN-02. None of this is guidance about buying or selling anything. Housing decisions couple leverage, geography and life circumstance in ways no article can see, and a probability built for calibration scoring is not a basis for a mortgage. The numbers here exist to be graded, and they will be.

The ledger, not the listings

Here is the reading, compressed and exposed. The post-2020 run broke affordability (5.08 times income, mortgages near 6.5%) without breaking credit; the 2008 analogy fails on loan quality, lock-in and supply deficit, and quietly holds on stretch; the national print is a stall masking two opposite regional markets; and the channel most likely to write the next act is insurance — sealed at 62% (±10) as CLI-01, with its failure condition published. Each claim above is either a verified published figure or a sealed simulation carrying its own confession terms. Nothing in this article was written to be safe from grading.

The register of what would prove this page wrong, in one place: CLI-01 dies if no major metro shows sustained insurance-attributable 15%+ declines through 2030 while the state backstops shrink. FIN-02 dies if NBER declares no recession before end-2028 with unemployment never rising a point off its low. The regional reading dies — visibly, in the same public data that built it — if the Sun Belt reaccelerates while the Northeast stalls. This observatory's Calibration Ledger, sealed 3 July 2026, begins public Brier scoring in July 2027, and the registry entries quoted here resolve on their own dates behind it, hits and misses at equal size.

The housing question deserves a better genre than it gets. The crash-content version sells a date; the industry version sells perpetual sunshine; both are unfalsifiable and both are working. The calibrated version is what this page has attempted: verified figures for the present, sealed probabilities for the channels, ranges instead of points, regions instead of a nation, and a scoring appointment instead of a rhetorical one. Come back when the entries resolve and read the grades — that is the entire proposition, here as everywhere on this platform. Nothing here is advice. All of it is checkable.

Frequently asked

Will the housing market crash in 2026?

The measured record says a national 2008-style crash requires forced sellers, and the 2026 market is structurally short of them: roughly half of outstanding mortgages still carry rates below 4%, post-2020 credit quality is the inverse of 2004–07, and most shortage estimates run in the millions of units. National prices are stalling (+0.8% YoY, April 2026), not collapsing — while 77 major metros, led by Florida's Gulf Coast and Austin, are already correcting. The sealed risk to watch is the insurance channel: CLI-01 at 62% (±10) for 2026–2030, a probabilistic simulation, not advice.

Are we in a housing bubble like 2008?

On price stretch, yes and worse: price-to-income reached 5.83 in 2022 and stood at 5.08 in 2025, above anything 2006 recorded. On structure, no: 2008 ran on subprime credit, ARM resets and a construction overhang; 2026 runs on two-thirds-prime originations, fixed rates half the market will not surrender, and a supply deficit. Stretched prices with sound credit historically resolve through long stagnation rather than cascade — the caveat being that a repriced hidden input (then credit, now insurance) is exactly what converts stall into fall.

Will mortgage rates come down?

This registry seals no interest-rate prediction, because rate point-forecasts have among the worst documented track records in all of finance. The verifiable state: the 30-year fixed averaged about 6.5% in June 2026, and the lock-in math means each downward step releases both buyers and sellers — the below-4% share of outstanding mortgages has already unwound from over 65% (2022) to just under 50% (Q1 2026). Note the asymmetry: meaningfully lower rates would likely lift prices before they lift affordability. Nothing here is advice.

Which housing markets are most at risk?

On the mid-2026 record, risk concentrates where elevated new supply meets rising carry costs: Florida's Gulf Coast (Punta Gorda −7.9% YoY, Cape Coral −6.1%) and pandemic-boom Texas metros (Austin −5.7%), with insurance and property-tax escalation in the causal chain — the CLI-01 mechanism operating locally. Wildland-interface and coastal zips with FAIR/Citizens dependence carry the repricing tail. The Northeast and Midwest sit at the opposite pole: scarce stock, prices +12.6% and +10% above the 2022 peak, and no supply response in sight.

Sources

Harvard Joint Center for Housing Studies, The State of the Nation's Housing 2026 — price-to-income 5.08 (2025), peak 5.83 (2022), 3.65 (1980); prices +31% vs incomes +22%, 2019–2024; seven metros above 8x, San Jose above 12x
S&P Cotality Case-Shiller National Home Price Index (April 2026) — +0.8% year over year; May 2026 preliminary +1.4%
Freddie Mac PMMS / Bankrate (June 2026) — 30-year fixed averaging ≈6.5%
National Association of Realtors (June 2026) — 4.6 months of existing-home supply; US Census Bureau — ≈10.3 months of new-home supply
ResiClub / Fortune analyses (June 2026) — 77 of ~300 major metros with falling YoY prices vs 223 rising; Punta Gorda −7.9%, Cape Coral −6.1%, Austin −5.7%; West −7.3% and South −3.5% vs the mid-2022 peak; Northeast +12.6%, Midwest +10%
eMBS / mortgage-market data via Wolf Street (Q1 2026) — 49.9% of outstanding mortgages below 4%, down from over 65% in Q1 2022; ≈$682/month payment gap on a median home
Fannie Mae statistical summaries / Calculated Risk — roughly two-thirds of originations to credit scores above 740
Housing-shortage estimates — NAHB ≈1.2M; Goldman Sachs ≈3M; Zillow >4M; Brookings ≈5M; McKinsey ≈8M; White House CEA ≥10M (April 2026)
Swiss Re Institute, sigma — insured natural-catastrophe losses $137B (2024) and $107B (2025), a record 92% of 2025 losses from secondary perils; January 2025 LA wildfires ≈$40B insured, the largest wildfire loss on record
California FAIR Plan (2025) — exposure past $500B, roughly triple its 2020 level; Hurricane Andrew (1992) — 11 insurer insolvencies
National Association of Realtors (2024) — median first-time homebuyer age 38, up from 29 in 1981