01 — Housing & Land Use

01 — Housing & Land Use · STAGE 1: EVIDENCE BRIEF

Referee panel: Prof. Olivia Brandt (#57, statistics, C) · Prof. June Takahashi (#17, econometrics, C) · domain specialist Prof. Samuel Birch (#10, urban economics, C)

Ground rules from the panel: every number below comes from a source we opened, with the data year. Where something is a secondary source (Wikipedia, news summary of a paper), we mark it. Today is 2026-09-24; we use the most recent figures available.

---

A. Affordability and burden

1. Renter cost burden is at record breadth. 22.7 million renter households (49%) paid more than 30% of income for rent and utilities; 12.1 million (26%) paid more than 50%. Data year 2024. — JCHS, State of the Nation's Housing 2026 press release (2026)
2. Owner burden. 20.7 million homeowner households (24%) were cost-burdened in 2024. — JCHS SONH 2026 full report (2026)
3. Price-to-income. The 2025 median existing single-family sales price was "nearly five times" median household income, vs. a ~3.2 average in the 1990s. Home prices up 54% since 2020. — JCHS SONH 2026 (2026)
4. Carrying costs beyond the mortgage. Property taxes +31% and homeowners-insurance premiums +72% from 2019 to 2025. — JCHS SONH 2026 press release (2026)
5. The bottom of the rental stock is vanishing. Units renting under $1,400/month fell by 9.3 million from 2014–2024 while units at $1,400+ rose by 11.8 million; ~7 million units under $1,000 disappeared. 11.0 million extremely-low-income (ELI) renter households compete for 3.8 million units affordable to them. — JCHS, America's Rental Housing 2026 (2026) and SONH 2026 (2026)
6. But market rents are soft. National rent growth near zero from mid-2023 through 2025; asking rents −0.6% y/y in Q4 2025; rental vacancy 5.2% in Q4 2025. Cost burdens nonetheless rose in 44 states and 88 of the 100 largest metros over five years. — JCHS America's Rental Housing 2026 (2026)
7. Homeownership rate fell for a second year to 65.2% (2025). Existing-home sales 4.1 million, a three-decade low. Residential mobility 11.2%, a record low (2024). — JCHS SONH 2026 (2026)

B. Supply, construction, costs

1. Construction pipeline, August 2026 (SAAR): permits 1,394,000 (+3.5% y/y); starts 1,275,000 (−1.2% y/y); single-family starts 918,000; completions 1,128,000, −27.1% y/y (from 1,548,000). — Census, New Residential Construction, Aug 2026 (Sept 2026)
2. Multifamily wave is receding: 416,000 multifamily starts in 2025; 686,000 units under construction vs a record 996,000 in 2023. — JCHS America's Rental Housing 2026 (2026)
3. Input costs: combined residential construction inputs +40% since January 2020 (steel mill products +84%, gypsum +47%). JCHS flags that foreign-born workers are a disproportionate share of construction labor and that deportations "threaten to push up costs even further" (no percentage given). — JCHS SONH 2026 (2026)
4. Tariffs: Section 232 duties effective Oct 14, 2025: 10% on softwood lumber/timber, 25% on kitchen cabinets and vanities (scheduled to rise to 50% on Jan 1, 2026); combined with existing duties, Canadian lumber faces ~45%. The U.S. imports about one-third of lumber consumed; Canada supplies ~85% of imports; domestic sawmills ran at ~64% capacity. — NAHB (2025; industry source)
5. Demand side shifting: household growth fell to 1.1 million in 2025 from a 2.0 million average in 2021; net international migration fell ~50% in 2025 with a further large drop expected in 2026. — JCHS SONH 2026 press release (2026)

C. Rates and lock-in

1. Mortgage rates: 30-yr fixed 6.95% (Sept 17, 2026), up from 6.76% the prior week and from 6.26% a year earlier. — Freddie Mac PMMS (2026)
2. Lock-in: each percentage point by which market rates exceed a borrower's locked rate cuts the probability of sale by 18.1%; lock-in prevented ~1.33 million sales mid-2022 to late-2023 and raised prices ~5.7%, more than offsetting the ~3.3% price reduction from higher rates. — FHFA Working Paper 24-03 (Batzer, Coste, Doerner, Seiler) (2024)

D. Homelessness

1. PIT count January 2025: 745,652 people, −3% vs 2024; 266,320 unsheltered. HUD notes 1,456,923 people were either homeless or in HUD-funded homeless housing programs; HUD reports chronic homelessness +81% from 2013 to 2025. — HUD AHAR 2025 release (2026). Referee note: the 2025 release uses a new framing (Housing First as baseline) — the numbers are HUD's; causal readings are not.
2. Cross-metro drivers: Colburn & Aldern (Homelessness Is a Housing Problem, UC Press 2022) find rents and rental vacancy explain more variation in per-capita homelessness than poverty, mental illness, drug use, climate or benefit generosity; e.g., King County ~5 per 1,000 vs Miami-Dade ~1 per 1,000. — Sightline summary (2022; secondary summary)
3. Vouchers and homelessness (RCT): HUD's Family Options Study found permanent housing subsidies reduced returns to shelter at 20 and 37 months, "almost halved" child separations and "more than halved" foster placements, at ~9% higher cost than usual care. — HUD USER, Family Options Study (2015–16)
4. Assistance rationing: only ~1 in 4 eligible households receive federal rental assistance; average wait ~28 months for those who get a voucher (2020 data). Vouchers serve 5.3 million people. — CBPP wait times; CBPP HCV overview (advocacy think tank; HUD data)

E. Institutional investors

1. GAO-26-108675 (Mar 2026): in six metros (Cincinnati, Dallas, Jacksonville, Nashville, Phoenix, Seattle) institutional investors owned <1% to 3% of all single-family homes and 4% (Seattle) to 22% (Jacksonville) of single-family rentals; holdings grew in all six from 2018–2024 (fastest 2021–23, slowing in 2024); in Nashville 35% of investor homes were bought from owner-occupants, 15% newly built. — GAO (2026); GAO blog, ~3% nationally (2026)
2. The law changed. The 21st Century ROAD to Housing Act became law July 11, 2026 (per HousingWire, without the President's signature). It bars investors owning 350+ single-family homes from buying more, with exemptions for build-to-rent, renovate-to-rent, foreclosure/loss-mitigation and others; no divestment. It also creates a $200M/yr competitive grant for localities that demonstrably increase supply (7-yr sunset), drops the manufactured-housing permanent-chassis rule, expands NEPA categorical exclusions, and ties some CDBG to production. — BPC section-by-section; HousingWire (2026)
3. CBO: direct-spending effect of H.R. 6644 ≈ −$6M outlays and revenues −$6M over 2026–2036, net deficit effect ≈ $0; CBO did not estimate discretionary (appropriated) spending. — CBO cost estimate (2026)

F. Subsidy programs

1. LIHTC placed ≥3.65 million units in service 1987–2022; ~$13.5B/yr tax expenditure (2023 est.); GAO (2018, 2011–15 data) median development cost per unit ranged $126k (TX) to $326k (CA); affordability typically 30 years, but the "qualified contract" process allows exit after 15. — Wikipedia summary of GAO/JCT figures (secondary)
2. 2025 reconciliation (OBBBA) LIHTC expansion: permanent 12% increase in 9% allocations and bond-financing test cut from 50% to 25% starting 2026; JCT scored ~$15.7B over 2026–2035; Novogradac projects ~1.22 million additional affordable rentals over 2026–2035 (1.14M from the bond test). — Novogradac (2025; industry projection)

G. Supply, zoning, and prices — the core causal literature

1. Shortage estimates span 0 to 7.3 million. NAHB 1.5M; Freddie Mac 3.7M (3.8M in its 2020 vintage); Zillow 4.5M; NAR 5.5M; NLIHC 7.3M (ELI-affordable units only, a different construct); Urban Institute's own ~2M; McClure & Schwartz: no national shortage. Up for Growth's 2024 figure: 3.5M. — Urban Institute, Bringing the Housing Shortage Into Sharper Focus (2025); JCHS blog; Up for Growth
2. Regulatory tax framework: Glaeser & Gyourko argue that where prices far exceed minimum profitable production cost, the gap mostly reflects regulation, far exceeding plausible externality costs. — NBER w23833 / JEP 2018
3. Hsieh–Moretti (AEJ:Macro 2019) claimed large GDP gains from loosening constraints in NY/SF/San Jose. The authors later acknowledged compounding errors that, if anything, understated their headline (3.7% → ~14% imperfect mobility; 8.9% → ~36% perfect mobility) — Econlib. But Greaney (AEJ:Macro 2026 comment) fails to replicate: the original counterfactual would lower output, results depend on an arbitrary population unit, and a unit-independent fix yields effects "two orders of magnitude smaller." — Greaney 2026
4. Skeptic evidence: Louie, Mondragon & Wieland (NBER w33576, 2025, rev. Feb 2026): income growth predicts the same growth in prices, quantities and population regardless of estimated supply elasticity; "relaxing regulatory housing supply constraints may not affect housing affordability." — FRBSF WP. Companion Economic Letter (Feb 2026): ~85% of metros saw housing units grow faster than population 2000–2020, including LA and SF. — FRBSF EL 2026. Rebuttal: Furth (SSRN 2025) argues total income is co-determined with population and thus housing supply, so their regressions mix equilibrium outcomes with demand. — Furth
5. Local effects of new market-rate buildings: Asquith, Mast & Reed — new large buildings in low-income areas lower nearby rents 5–7% (Zillow listings, ~2-block radius) and increase in-migration from low-income neighborhoods. — Upjohn. Pennington (San Francisco, fire-induced construction 2003–17): rents ~$40 lower nearby, effect fades to zero within ~2 km; displacement to lower-income zip codes down ~20%; affordable (subsidized) buildings show no measurable neighborhood spillover. — Berkeley Matrix
6. Filtering chains: Mast (JUE Insight 2023): a new market-rate building housing 100 people leads 45–70 people to move out of below-median-income neighborhoods, mostly within 3 years. — JUE via RePEc
7. Rent control: Diamond, McQuade & Qian (AER 2019): SF's 1994 expansion cut tenant mobility/displacement ~20% but landlords reduced rental supply 15%, likely raising long-run market rents. — AER

H. Natural experiments in upzoning

1. Auckland (2016 Unitary Plan): a 2025 peer-reviewed review of three quasi-experimental studies finds upzoning added 21,800–43,500 units and reduced rents ~28% relative to counterfactual within six years; ~84% of consents became net additions; it rebuts critiques by Murray & Helm. — *Land Use Policy* 2025
2. Minneapolis: Pew: 2017–22 housing stock +12% vs +4% rest of state; rents +1% vs +14%; ~87% of ~21,000 permitted units in 20+-unit buildings, ~1% duplex/triplex. — Pew 2024. Counter: Minneapolis Fed (2025): the rent divergence begins in summer 2020 across datasets; vacancy spiked to 9.5% (Oct 2020); authors infer a negative demand shock; only 225 units in 87 duplex/triplex buildings 2020–24; permits down 77.6% (2023) and 92.4% (2024) from the 2019 peak; effect of the 2040 Plan "remains unclear." — Minneapolis Fed 2025
3. Houston: minimum lot size cut from 5,000 to 1,400 sq ft (1998 inner city; 2013 citywide); ≥34,000 townhouses built 2007–20; converted townhouse median assessed value $340k vs $545k for other new SF homes; only 0.5% of SF parcels converted. — Pew 2023
4. Austin: −4.2% y/y rent in Q1 2025, lowest of 150 largest metros, after peaking at +14.7% (Q4 2021) and a construction boom. — NMHC (industry source)
5. State preemption results so far: California SB 9 (2021 lot-split/duplex) rarely used in first year per Terner Center; LA approved 38 of 211 applications in 2022 (secondary, Wikipedia summary). Portland's Residential Infill Project: ~1,400 ADU/middle-housing units permitted Aug 2021–Jun 2024 (news report). Montana's 2023 reforms: enjoined then reinstated; unanimous Montana Supreme Court upheld them in March 2026; cities not required to comply until May 2026; no production numbers yet. — APA Planning Feb 2026

---

Contested evidence (where the literature genuinely disagrees)

• Is there a national shortage, and how big? Vacancy-based (NAHB, Freddie Mac) vs construction-trend (NAR) vs ELI-affordability (NLIHC) vs "no shortage, distribution problem" (McClure & Schwartz). Urban's ~2M sits at the low end. With household growth halving (fact 12) and national vacancy at 5.2%, the national number is getting smaller; metro shortfalls remain.
• Do supply constraints explain cross-metro price growth? Glaeser–Gyourko, Saiz-style elasticity work, and Furth vs. Louie–Mondragon–Wieland. This is the live frontier dispute and is not settled.
• Macro costs of zoning. Hsieh–Moretti's large GDP numbers are now seriously undermined by Greaney (2026). The micro evidence that constraints raise local prices is much sturdier than the macro growth-loss number.
• Minneapolis. Pew's supply story vs. Minneapolis Fed's demand-shock story. Both use real data; the Fed paper is more careful about timing.
• Auckland. Peer-reviewed quasi-experiments (Greenaway-McGrevy et al.) vs. informal critiques (Murray & Helm). The weight of published evidence favors a real supply and rent effect; the magnitude depends on counterfactual construction.
• Local new supply and displacement. Asquith–Mast–Reed, Pennington, and Mast agree on sign (lowers nearby rents / reduces displacement). Earlier anti-displacement literature (e.g., neighborhood "signal" or amenity effects) is less causally identified; we found no recent causal study showing net rent increases from market-rate supply, but effects are small (5–7%, ~$40) and local.
• Institutional investors. Ownership shares are small nationally (≤3%) but concentrated (22% of SF rentals in Jacksonville). Causal price/rent effects: we could not verify a primary causal study in this review; treat as open.
• LIHTC crowd-out. A classic literature (Sinai & Waldfogel; Eriksen & Rosenthal; Baum-Snow & Marion) finds partial-to-large crowd-out of private construction in some settings. We did not open the primary papers in this round — labeled unverified.

— Brandt, Takahashi, Birch

STAGE 2: OPENING POSITIONS (Part 1 of 3 — Left bloc)

Prof. Rachel Stein (#73, urban studies, L) — "The market cannot build the bottom of the ladder"

Causal claim. The crisis that matters morally is at the bottom: 11.0 million ELI renter households chasing 3.8 million units they can afford (JCHS 2026). That gap is not a zoning artifact; it is an income problem colliding with a market that has deleted 9.3 million sub-$1,400 units in a decade (JCHS Rental 2026). Meanwhile, luxury construction in gentrifying neighborhoods raises nearby rents and pushes out long-time residents — new glass towers are a signal to capital that a neighborhood is "turning."

Evidence. Facts 1, 5, 6 in the brief. Note especially that national rents were falling in Q4 2025 while cost burdens rose in 88 of 100 metros. If supply were the whole story, soft rents should have eased burdens. They didn't, because the tenants in trouble aren't in the market segment that softened.

Remedy. (a) Deep, permanent subsidy — vouchers as an entitlement for ELI households; (b) public or nonprofit acquisition of existing unsubsidized low-rent stock before it filters up; (c) anti-displacement tools (right to counsel, rent stabilization with a new-construction exemption) in neighborhoods receiving upzoning.

Tradeoff I concede. Rent stabilization without a new-construction exemption reduces supply; I accept Diamond–McQuade–Qian's 15% supply reduction as a real cost of the San Francisco design.

Falsifiable prediction. If I'm right, then in metros with the biggest 2023–25 multifamily completion waves, the count of units renting under $1,000 will keep falling through 2028 even as average rents stay flat. If sub-$1,000 inventory rises in, say, Austin or Phoenix by 2028, filtering is working faster than I think.

---

Prof. Rosa Delgado-Finch (#7, political economy, Soc) — "Housing as an asset class is the disease"

Causal claim. Housing was financialized: private equity and Wall Street landlords bought up a fifth of the single-family homes in the Sun Belt after 2008 and now set rents algorithmically. Housing prices are high because shelter is priced as a yield-bearing asset, and policy (the mortgage interest deduction, GSE guarantees, LIHTC syndication fees) is built to protect asset values.

Evidence. GAO shows institutional holdings grew in every metro studied 2018–2024 and that in Nashville 35% were bought from owner-occupants — that's direct conversion of ownership to tenancy (fact 19). Congress itself just conceded the problem by passing the ROAD Act's 350-home cap (fact 20). LIHTC units exit affordability after as little as 15 years (fact 22): we rent affordability from investors instead of owning it.

Remedy. A federal social-housing developer (permanently affordable, mixed-income, publicly owned, financed at Treasury rates); close the ROAD Act build-to-rent loophole; national rent stabilization; right of first refusal for tenants and nonprofits on sales.

Tradeoff I concede. Public developers have a poor cost record in the U.S.; I concede that a social-housing agency that builds at the $326k/unit end of LIHTC (fact 22) will not scale.

Falsifiable prediction. If financialization drives rents, then in the ROAD Act's first three years, rent growth in single-family rentals in the most investor-heavy metros (Jacksonville, Atlanta) should decelerate relative to low-investor metros (Seattle, Cincinnati). If there's no differential by 2029, I'm overweighting investors.

---

Prof. Patience Mensah (#95, housing design, C-L) — "The Left should be the party of building"

Causal claim. Scarcity is a policy choice made parcel by parcel. The anti-displacement left and the local-control right have converged on the same tool — discretionary review — and it's the tool that delivered California. Legalize the missing middle, kill parking mandates, allow single-stair buildings, and you get cheaper homes and less displacement.

Evidence. Houston's lot-size reform produced ≥34,000 townhouses at a median $340k vs $545k for other new houses (fact 33). Minneapolis froze rents by legalizing triplexes citywide: rents +1% vs +14% statewide (fact 32). Pennington and Asquith–Mast–Reed show new market-rate buildings reduce displacement and nearby rents (fact 28). Mast's chains: 100 new market-rate residents → 45–70 moves out of below-median neighborhoods (fact 29).

Remedy. State by-right approval for 2–6 units on residential lots and mid-rise near transit (the Oregon/Montana/California SB 79 model), with shot clocks and a builder's remedy for non-compliant cities; national building-code reform (single-stair to 6 stories).

Tradeoff I concede. Middle housing is slow: SB 9 barely moved (fact 35). By-right laws are necessary, not sufficient — they don't reach ELI households without subsidy.

Falsifiable prediction. If I'm right, Portland's middle-housing production (≈1,400 units in its first ~3 years) should roughly double in the next three years as financing and designs standardize, and Montana cities that comply after May 2026 should show a measurable increase in 2–4-unit permits by 2028.

STAGE 2: OPENING POSITIONS (Part 2 of 3 — Center bloc)

Prof. Colleen Farrow (#34, local government, C) — "Preemption, but with capacity"

Causal claim. Local land-use decisions impose regional costs on people who can't vote in them (future residents), a classic externality that justifies state override. But preemption without implementation capacity (planners, pre-approved plans, infrastructure finance) produces paper capacity, as SB 9 shows.

Evidence. Auckland's regionwide upzoning cut rents by about a third and added tens of thousands of units (fact 31). The contrast with SB 9's weak uptake (fact 35) is instructive: Auckland rezoned three-quarters of residential land at once and paired it with infrastructure planning; SB 9 allowed lot splits with owner-occupancy strings attached.

Remedy. State housing targets with automatic by-right approval for cities that miss them; state-funded pre-approved plan sets; conditioning state transport and infrastructure money on permits issued (not plans adopted). Federally, scale up the ROAD Act's $200M supply grant and tie it to completions.

Tradeoff I concede. Preemption shifts political conflict to the state capitol, where cities will fight back through fees, impact studies, and slow-walking; the Montana litigation (fact 35) cost two years.

Falsifiable prediction. If state targets with automatic consequences work, California cities subject to builder's-remedy exposure should show higher permitting per capita 2025–2028 than comparable cities that met targets on paper. If there's no difference, enforcement is theater.

---

Prof. Daniel Mbeki (#70, demography, C) — "The shortage number is mostly an artifact; the demand side is moving"

Causal claim. There is no single national shortage number worth defending. The estimates range from 0 to 7.3 million because they answer different questions (fact 24). The biggest near-term housing variable in 2026 is demand: household growth has halved from 2.0M (2021) to 1.1M (2025) and net immigration fell ~50% in 2025, with more decline in 2026 (fact 12). Housing supply constraints do not explain cross-metro price growth nearly as well as income growth does (Louie–Mondragon–Wieland, fact 27).

Evidence. In ~85% of metros, including LA and SF, units grew faster than population 2000–2020 (FRBSF 2026). Minneapolis's celebrated rent divergence starts in 2020 with a demand shock and a 9.5% vacancy spike, not with triplexes (Minneapolis Fed, fact 32). National rents fell in Q4 2025 with vacancy at 5.2%.

Remedy. Stop anchoring federal policy on a national shortage number. Target where price-to-cost gaps are large (Glaeser–Gyourko is the right metric, even if its causal story is contested), and target the income gap for ELI renters directly.

Tradeoff I concede. "Units grew faster than population" can coexist with a shortage if household formation was suppressed (the young living with parents). Freddie Mac's estimate explicitly adds such "missing households."

Falsifiable prediction. If demand is dominant, then with immigration down sharply, national real rents should be flat-to-negative through 2027 even though completions fell 27% y/y (fact 8). If rents re-accelerate above 4% nominal by late 2027 with low immigration, the supply-shortage camp was right about the stock.

---

Prof. Grant Albrecht (#6, antitrust/IO, C) — "It's the cost curve, not just the permit"

Causal claim. Even where zoning allows it, housing is expensive to build and construction productivity has stagnated for decades. Input costs are up 40% since January 2020 (fact 10). The federal government is currently adding to costs: 45% combined duties on Canadian lumber, 25–50% on cabinets (fact 11), and removal of immigrant construction labor. Rates at ~7% (fact 13) raise carrying costs on every project.

Evidence. Completions down 27% y/y (fact 8); multifamily under construction down from 996k to 686k (fact 9). Permits are up 3.5% y/y — so entitlement isn't the binding constraint at the margin right now; financing and cost are.

Remedy. Exempt residential building materials from Section 232 tariffs; a construction-labor visa track; federal preemption of building-code variation for factory-built housing (build on the ROAD Act's chassis-rule repeal); R&D and procurement support for modular.

Tradeoff I concede. Tariff exemptions cost revenue, and modular has a long history of failed scale-ups (e.g., Katerra's collapse). Code preemption shifts safety risk to the federal level.

Falsifiable prediction. If costs bind, then removing the lumber/cabinet tariffs should show up in the PPI for residential construction inputs within 12 months and in single-family starts within 24. If starts don't respond, the constraint is demand or rates.

STAGE 2: OPENING POSITIONS (Part 3 of 3 — Right bloc)

Prof. Walter Pruitt (#8, law & economics, Lib) — "Abolish the permission regime"

Causal claim. Zoning is a cartel run by incumbent owners. The regulatory tax is the gap between price and construction cost (Glaeser–Gyourko, fact 25), and it is enormous in coastal metros. Hsieh and Moretti showed that housing constraints in three cities cut aggregate U.S. growth so much that GDP would be 36% higher without them. This is the largest self-inflicted wound in the American economy.

Evidence. Houston (fact 33): no zoning, lot-size cut, 34,000 townhouses, entry prices ~$200k below other new homes. Austin (fact 34): build, and rents fall — −4.2% in Q1 2025. Rent control is the opposite lesson: 15% supply reduction in SF (fact 30).

Remedy. Replace use-based zoning with nuisance law and objective, ministerial building standards; abolish minimum lot sizes and parking mandates; repeal the ROAD Act's investor purchase ban — it restricts the demand side of rental supply and protects nobody.

Tradeoff I concede. Nuisance law is litigation-heavy and slower for genuine externalities (industrial uses next to homes). Some neighborhoods will change faster than residents want.

Falsifiable prediction. If I'm right, the five fastest-permitting metros of 2021–24 should have the lowest real rent growth 2024–2028 among metros with similar job growth.

---

Prof. Benjamin Ashford (#31, federalism, R) — "Don't nationalize the zoning map"

Causal claim. Local control is not a cartel; it's self-government over the most important investment most families make. Upzoning mandates from Sacramento or Washington are a centralization of power with thin evidence: SB 9 produced almost nothing, Minneapolis's rent story is a 2020 demand shock, and Montana's "miracle" has no production numbers. And homelessness — the issue most used to justify these mandates — is primarily a problem of addiction and untreated mental illness, which Housing First has failed to address: HUD itself reports chronic homelessness up 81% since 2013.

Evidence. Facts 15, 32 (Minneapolis Fed), 35. Montana's reforms were litigated by homeowners for two years (fact 35).

Remedy. Federal role limited to removing federal barriers (NEPA — as the ROAD Act did; the chassis rule), and voluntary incentives. States should give cities options menus, not mandates. Homelessness: treatment-first programs and enforcement of public-camping laws post-*Grants Pass*.

Tradeoff I concede. Local control will produce under-building in some high-demand places; I accept that some regional price premium is the price of self-government.

Falsifiable prediction. If mandates are unnecessary, cities that voluntarily compete for the ROAD Act supply grants should show production gains comparable to mandated jurisdictions over 2027–2030.

---

Prof. Martin Oyelaran (#12, finance, C-R) — "The rate shock and lock-in, not zoning, explain 2022–2026"

Causal claim. Zoning explains the level differences between Boston and Birmingham; it does not explain why national affordability collapsed after 2021. That was rates and lock-in. FHFA estimates lock-in prevented 1.33 million sales and raised prices 5.7% (fact 14), more than offsetting the rate-driven price decline. With the 30-year at 6.95% (fact 13), a huge share of owners won't sell. Existing sales at a 30-year low (fact 7) are the symptom.

Evidence. Facts 7, 13, 14. Price-to-income near 5 vs ~3.2 in the 1990s (fact 3).

Remedy. GSE-backed mortgage portability and broader assumability; fiscal consolidation to bring down term premia; no rent controls; LIHTC as is (just expanded), and oppose investor bans that add friction.

Tradeoff I concede. Portability disproportionately benefits existing owners (older, wealthier) and imposes interest-rate risk on the GSEs/taxpayer.

Falsifiable prediction. If lock-in dominates, a fall in the 30-year rate to ~5.5% should raise existing-home sales above 5 million within 18 months without a matching rise in new construction. If sales stay near 4 million, lock-in is weaker than I think.

STAGE 3: CROSS-EXAMINATION (Part 1 of 3)

Exchange 1 — Pruitt (Lib) challenges Stein (L). Four rounds.

Pruitt, round 1. Rachel, your claim that new "luxury" towers raise nearby rents and drive out residents is precisely the claim the best causal studies test, and it fails. Asquith–Mast–Reed: large new buildings in low-income areas lower nearby rents 5–7% and pull in movers from low-income neighborhoods (fact 28). Pennington uses fires as an instrument in San Francisco — the most gentrification-anxious city in America — and finds rents ~$40 lower nearby and displacement to poorer zip codes down about 20%. Mast's chains (fact 29): 100 market-rate residents → 45–70 moves out of below-median neighborhoods within three years. What causal study shows the sign you asserted?

Stein, round 1. I'll grant the sign on the building-level studies. But look at the magnitudes. $40 a month against a San Francisco rent is noise to a family paying 60% of income. 5–7% within two blocks. These are real but small, local effects. Meanwhile JCHS shows 9.3 million sub-$1,400 units gone in a decade (fact 5). Filtering is working against a much larger tide of upward filtering — renovations, conversions, rent increases on old stock. And Pennington herself finds affordable buildings show no neighborhood spillover but do house their own residents — she calls market-rate and affordable housing complements. I'm arguing for the complement you'd defund.

Pruitt, round 2. Then withdraw the claim that market-rate construction raises rents. On the "tide": why did 9.3 million units filter up? Because demand for existing stock rose faster than new supply, so old units got bid up. That's the shortage showing up in the bottom of the distribution. Austin (fact 34) is the counterexample you need to explain: after the boom, rents fell 4.2% y/y. The Diamond–McQuade–Qian result (fact 30) is the cautionary tale for your rent stabilization: 20% less displacement for covered tenants, 15% less rental supply.

Stein, round 2. I'll amend on the record: the best causal evidence does not support the claim that new market-rate buildings raise nearby rents; the sign runs the other way, with small magnitudes. I'll keep the rest. Austin is one boom-bust metro; NMHC is an industry source; and the Q1 2025 decline is off a +14.7% peak. What I want to see is sub-$1,000 inventory rising, not average rents falling. On rent stabilization: the SF design had no new-construction exemption and allowed condo conversion. Oregon- and California-style caps (inflation plus a margin, with new construction exempt for 15 years) are a different instrument. Show me the Diamond-quality study of those caps before you generalize.

Pruitt, round 3. Fair that the anti-gouging caps are less studied — but "unstudied" is not evidence of harmlessness. Oregon's cap binds rarely because it's loose; if it's tightened, you get San Francisco. And your prediction — sub-$1,000 inventory in boom metros through 2028 — is a good one. I'll add mine: in the same metros, the share of renters who are cost-burdened should fall by 2028 if supply is the lever.

Stein, round 3. Accepted as a joint test. Note what you just conceded: a loose cap is nearly harmless. That's the politics of the next decade: supply plus guardrails. I'm not caving on vouchers — the Family Options RCT (fact 17) is the strongest causal evidence in this entire thread, and it's about subsidy, not zoning.

---

Exchange 2 — Oyelaran (C-R) challenges Delgado-Finch (Soc). Three rounds.

Oyelaran, round 1. Rosa, "a fifth of the single-family homes in the Sun Belt" is not what the data say. GAO-26-108675 (fact 19): institutional investors own less than 1% to 3% of all single-family homes in the six metros studied, including Jacksonville, Dallas and Phoenix. Nationally, ~3%. The 22% figure is the share of single-family rentals in Jacksonville — a different denominator. And the ROAD Act you cite exempts build-to-rent precisely because 15% of Nashville's investor homes were newly built: that's supply. Also: "algorithmic rent-setting" in single-family rentals — which study?

Delgado-Finch, round 1. On the denominator, you're right and I'll take it to the referees. But concentration matters more than national shares. The ROAD Act's Senate sponsor put it at more than 72,000 homes in metro Atlanta — "more than one in four" single-family rentals there (HousingWire, fact 20). In Nashville, 35% of investor homes were bought from owner-occupants: that is owner-occupancy converted to tenancy at the exact entry-level price points first-time buyers compete for. And the growth rates — Phoenix +177%, Dallas +114% 2018–24 (GAO blog) — are not trivial.

Oyelaran, round 2. Growth off a small base. What's missing is any causal evidence in this brief that investor purchases raised prices or rents metro-wide; the referees explicitly flagged that as open. Lock-in (fact 14) has an FHFA estimate — +5.7% on prices. Your investors don't have one here. Meanwhile, the ROAD Act already capped them. So what is your marginal policy?

Delgado-Finch, round 2. Absence of a verified causal estimate in this brief isn't evidence of no effect; it's a research gap, and the GAO data show investors rarely sell (never more than 8% of holdings per year) — this stock isn't coming back to owner-occupancy. My marginal policy is social housing: a permanently affordable, publicly owned stock that can't be sold to anyone. On algorithmic pricing, I withdraw the single-family claim; the litigation I had in mind concerns multifamily rent-setting software, not SFR, and it's not in our brief.

Oyelaran, round 3. Then we agree the investor issue is (a) concentrated, (b) now legislated, (c) causally unmeasured. I'd support funding the evaluation — which is Proposal material.

---

Exchange 3 — Ashford (R) challenges Mensah (C-L). Three rounds.

Ashford, round 1. Patience, "Minneapolis froze rents by legalizing triplexes" is contradicted by the Minneapolis Fed (fact 32). 225 units in 87 duplex/triplex buildings from 2020–24. The rent divergence begins in summer 2020, alongside a 9.5% vacancy spike — a demand shock. Permitting then collapsed by 77.6% and 92.4% in 2023–24. Even Pew's own numbers say ~1% of the 21,000 permitted units were duplex/triplex. And SB 9 (fact 35): LA approved 38 of 211 applications in year one. Your showcase reforms don't show much.

Mensah, round 1. You're right about triplexes and I'll correct it: Minneapolis's supply came overwhelmingly from 20+-unit buildings enabled by parking-mandate repeal and transit-corridor zoning — ~87% per Pew — not from triplexes. My claim should have been "Minneapolis's multi-pronged reforms," and even then the Fed paper tells me the rent attribution is unclear. But Houston is not ambiguous: ≥34,000 townhouses over 13 years, $340k vs $545k (fact 33), with 0.5% of parcels converted. And Auckland (fact 31) has peer-reviewed quasi-experimental evidence: 21,800–43,500 extra units, ~28% lower rents relative to counterfactual in six years.

Ashford, round 2. Houston is a special case — cheap land, flat terrain, deed restrictions doing what zoning does elsewhere. Auckland is a single-country, single-city natural experiment where the critics (Murray & Helm) are credible economists. And Montana's own Supreme Court fight shows these mandates lack local legitimacy.

Mensah, round 2. The Montana Supreme Court upheld the reforms unanimously in March 2026 (fact 35). That is the legitimacy question answered in the state's own constitutional forum. On Auckland, the critiques were informal (blogs/social media) and the 2025 Land Use Policy review addresses each one: sample selection, consents vs completions (~84% of consents became net additions), and the counterfactual method. I'm not claiming Auckland's 28% transfers to Cleveland — I'm claiming the sign and order of magnitude in a high-demand metro.

Ashford, round 3. I'll concede this much: the Auckland supply response is well identified. I don't concede that the U.S. federal government should be the one to do it. A state court upholding a state law is federalism working. Keep Washington out.

Mensah, round 3. On that we can build a bridge: state preemption, federal incentives. I'll take it.

STAGE 3: CROSS-EXAMINATION (Part 2 of 3)

Exchange 4 — Mensah (C-L) and Hartley (#4, monetary econ, Lib) challenge Mbeki (C). Four rounds — the central empirical fight.

Mensah, round 1. Daniel, you lean hard on Louie–Mondragon–Wieland. But Furth's response (fact 27) identifies the core problem: total income growth in a metro is mostly population growth, and population growth is co-determined with housing supply. Regress prices on an outcome of supply, and of course elasticity "doesn't matter." Your "85% of metros added units faster than population" stat has the same problem — households per unit falls when the young can't form households, which is the shortage.

Mbeki, round 1. Furth's point on total income is fair and I've read it. But LMW also use average income and instruments for demand, and they report rents, not just prices. The broader point stands: the elasticity measures everyone uses (Saiz, Wharton index) do not predict price growth differences across 2000–2020 the way the supply-constraint story says they should. And on household formation, I conceded that in my opening — Freddie Mac's +0.4M missing households is a reasonable adjustment. It doesn't get you to 5.5 or 7.3 million.

Hartley, round 2 (joining). Then let me hand you a within-metro test where elasticity is manipulated by policy, not estimated: Auckland (fact 31). Upzoned vs non-upzoned areas, same city, same income shock. Supply +21,800–43,500, rents ~28% below counterfactual. That's the experiment LMW's cross-section can't run. And Austin (fact 34): the demand shock you want (5.5% population growth, 8% employment growth 2020–22) hit a place that could build, and rents fell after supply arrived.

Mbeki, round 2. Auckland is the strongest evidence on your side and I'll say so. But note two things. First, "28% relative to counterfactual" is not "rents fell by a third" — it's a gap from a synthetic path, and the counterfactual design is what Murray and Helm contested. Second, Austin is exactly my point about timing: rents fell in 2023–25 as demand cooled and a supply wave landed together. You can't separate them from NMHC's descriptive note. Minneapolis (fact 32) shows how easily a demand shock is misattributed to zoning. So my position is: local supply effects are real (Asquith–Mast–Reed, Pennington), metro-level effects of upzoning are plausibly real (Auckland), and national shortage numbers are shaky.

Mensah, round 3. That's a much narrower claim than "the shortage number is mostly an artifact." Your prediction also has a problem: you predict flat real rents through 2027 because immigration is down. But completions are down 27% y/y (fact 8) and multifamily under construction is down 31% from peak (fact 9). If rents re-accelerate in 2027–28, that's the stock story. Will you accept that result?

Mbeki, round 3. Yes — that's why I stated it. If rents re-accelerate above 4% nominal by late 2027 with low immigration, I'll concede the stock-shortage camp was right about the national picture. And I'll amend my opening on the record: "mostly an artifact" overstated it. Better: the national shortage is real but small and shrinking relative to the 3.5–5.5M headline numbers; the binding shortages are metro-specific and at the ELI end. What I won't concede is that LMW is refuted. Furth's critique is a working paper against a working paper; this is open.

Hartley, round 4. Open, fine. But policy doesn't need the national number. It needs the local price-to-cost gap, which Glaeser–Gyourko measure and which you said is the right metric. In the places where that gap is large, by-right upzoning is the only intervention with Auckland-grade evidence.

Mbeki, round 4. Agreed on the metric and the targeting. That's why I'd vote for state preemption aimed at high-gap metros and against another national "3.7 million homes" talking point.

---

Exchange 5 — Imani Walker (#65, African American studies, L) challenges Farrow (C). Three rounds.

Walker, round 1. Colleen, "future residents can't vote" is a clean abstraction. The people who do vote in the neighborhoods targeted for upzoning are often Black and Latino renters who have seen "planning" used against them — urban renewal, highways. State preemption removes one of their few levers. And your Auckland claim is overstated: you said rents fell "by about a third." The review says ~28% relative to a counterfactual, not an absolute fall.

Farrow, round 1. Correction accepted on Auckland: 28% relative to counterfactual, not an observed one-third decline. I'll ask the referees to rate it. On power: the evidence you'd want is Pennington's — in San Francisco, new market-rate construction reduced displacement of nearby residents to poorer zip codes by ~20%. Discretionary review has historically protected owners, who are disproportionately white; renters rarely win at the zoning board.

Walker, round 2. Pennington's effect decays to zero at 2 km, and it's one city. I'd trade preemption for enforceable anti-displacement guarantees — right to return, tenant right-to-counsel, one-for-one replacement of demolished rent-stabilized units.

Farrow, round 2. One-for-one replacement of covered units, yes — California builds that into several state laws. Right-to-counsel, yes; it's cheap relative to shelter costs. A veto, no. The veto is how we got here.

Walker, round 3. Then write it as a package. I'll vote on the package.

---

Exchange 6 — Teresa Quinn (#90, forestry, R) challenges Albrecht (C). Three rounds.

Quinn, round 1. Grant, you treat the lumber tariffs as pure cost. Softwood lumber has been in a subsidy dispute with Canada for 40 years; the duties offset stumpage subsidies. And domestic mills were running at 64% capacity (fact 11) — there's slack to fill. Tariffs rebuild domestic supply capacity, including in rural timber communities.

Albrecht, round 1. Slack at 64% is precisely the problem: if domestic mills could profitably fill the gap at pre-tariff prices, they'd already be running fuller. The U.S. imports a third of its lumber; Canada is 85% of imports. A 45% combined duty is a tax on every stick-framed house. And cabinets at 25% going to 50% has nothing to do with stumpage.

Quinn, round 2. Capacity takes years; tariffs are a signal to invest. And you said inputs are up 40% since 2020 — lumber is a modest share of that, much of which is steel and gypsum.

Albrecht, round 2. Fair on shares: I can't give you a verified per-home tariff cost from this review, and I won't invent one. JCHS says steel mill products +84%, gypsum +47% — those are also tariff-exposed. My claim is directional: in a market where completions just fell 27%, raising input costs is the wrong direction.

Quinn, round 3. Directionally true for builders; I'd still exempt only lumber from allies, not the cabinet duties on China. We'll see that in the vote.

STAGE 3: CROSS-EXAMINATION (Part 3 of 3)

Exchange 7 — Priya Venkataraman (#19, inequality econ, L) challenges Pruitt (Lib). Three rounds.

Venkataraman, round 1. Walter, you said Hsieh–Moretti show GDP "would be 36% higher" without housing constraints. Two problems. First, even in their corrected numbers, 36% is the perfect-mobility counterfactual; the imperfect-mobility figure is ~14%. Second, and decisively, Greaney's 2026 comment in the same journal (fact 26) fails to replicate: their counterfactual would lower output, the results depend on an arbitrary population unit, and a unit-independent version gives effects "two orders of magnitude smaller." You're citing a number the literature has substantially knocked down.

Pruitt, round 1. I'll take the Greaney hit. I should not have used 36%, and I retract it. But note what survives: Glaeser–Gyourko's price-to-cost wedge is a micro measurement, not a growth model. Houston's lot-size reform is a micro natural experiment. Auckland is a micro-to-metro experiment. The case for deregulation never needed a 36% GDP number; it needs the wedge.

Venkataraman, round 2. Then let's look at the wedge's distribution. Price-to-income at ~5 (fact 3) is national, including in places with loose zoning. And your Austin example: rents fell — good — but did burdens? JCHS says burdens rose in 88 of 100 metros even as rents softened (fact 6). Deregulation lowers the price of new stock; it doesn't raise the incomes of the bottom quintile.

Pruitt, round 2. Agreed that it doesn't raise incomes — nothing in land-use policy does. It lowers the rent against which incomes are measured. And "burdens rose in 88 metros" over five years includes the 2021–22 rent spike; the question is the forward path.

Venkataraman, round 3. Then we share a test. I'll add that your "abolish zoning" is also politically inert — Houston has deed restrictions covering much of the city. Minimum-lot-size and parking reform is the achievable core.

Pruitt, round 3. I'll take "achievable core" as a compliment.

---

Exchange 8 — Tamsin Kerr (#30, public policy, C-L) and Grace Holloway (#71, social work, L) challenge Ashford (R). Three rounds.

Kerr, round 1. Ben, you said homelessness is "primarily" addiction and mental illness. Across metros, that's not what the data say. Colburn & Aldern (fact 16) find poverty, mental illness and drug use explain less cross-metro variation than rents and vacancy. West Virginia has severe addiction rates and low homelessness; Seattle's King County runs ~5 per 1,000, Miami-Dade ~1. Individual risk factors decide who loses the game of musical chairs; the housing market decides how many chairs there are.

Ashford, round 1. At the individual level, addiction and mental illness are massively over-represented among the unsheltered. That's not refuted by a cross-metro regression.

Holloway, round 2. Both can be true, and Colburn–Aldern say so ("precipitating events"). But your "Housing First has failed; chronic homelessness up 81% since 2013" misreads the HUD figure. HUD's number is a descriptive count; it doesn't show Housing First caused the rise, and over the same period rents rose sharply in exactly the metros where chronic homelessness rose. The one randomized trial we have on housing subsidy for homeless families — Family Options (fact 17) — found vouchers cut shelter returns, halved child separations and foster placements, and reduced substance use and intimate-partner violence.

Ashford, round 2. Family Options is about families, not chronically homeless single adults with addiction. And the unsheltered numbers in West Coast cities under Housing First regimes are an indictment.

Kerr, round 3. Correct that Family Options is families — and families are a large share of the ELI crisis. On the West Coast: those are also the highest-rent, lowest-vacancy markets. You have to show the policy effect net of rents. You haven't.

Ashford, round 3. I'll amend: homelessness at the metro level tracks housing costs; at the individual level, addiction and mental illness are central risk factors, and the chronically homeless population needs treatment capacity that housing alone won't provide. I withdraw the causal attribution of the 81% rise to Housing First; HUD's figure doesn't establish that.

---

Exchange 9 — Keisha Morrow (#5, economic history, L) challenges Oyelaran (C-R). Three rounds.

Morrow, round 1. Martin, you say lock-in, not zoning, explains 2022–26. But your own source quantifies it: +5.7% on prices (fact 14). Prices are up 54% since 2020 (fact 3). Lock-in is a meaningful slice, not the story. And portability is a subsidy to the owners who already won the 2020–21 rate lottery.

Oyelaran, round 1. The 5.7% is a price-level effect through mid-2023. The larger effect is on transactions and mobility: 1.33 million lost sales, a three-decade-low sales count, record-low mobility (fact 7). That's a labor-market problem — people can't move to jobs. And much of the 54% is the 2020–22 rate collapse plus fiscal stimulus, which is my macro story too: rates, not zoning, drove the national surge.

Morrow, round 2. Rates down and rates up both raised prices? Low rates inflated demand in 2020–21; high rates froze supply of existing homes after 2022. I'll grant both. But zoning determines where that demand turned into prices versus units — that's the Boston vs. Houston point you conceded. On portability: GSEs would bear rate risk; who pays?

Oyelaran, round 2. I'll amend: lock-in explains a meaningful share of post-2022 price resilience (FHFA: ~5.7%) and most of the collapse in existing sales; it does not explain the 2020–22 run-up, and zoning governs how national demand shocks translate into local prices. On portability, there's no official score. It should be piloted with explicit fees priced to the rate risk.

Morrow, round 3. A priced pilot I can live with. A free option to incumbents, no.

STAGE 4: REFEREE FACT-CHECK

Panel: Brandt (#57), Takahashi (#17), Birch (#10). Every empirical claim in Stages 2–3, rated S = Supported · C = Contested (evidence genuinely split, or plausible but not verified by the panel in this review — noted) · U = Unsupported · W = Wrong. "Fact n" refers to the Evidence Brief.

# · Speaker · Claim (as made) · Rating · Basis
1 · Stein · 11.0M ELI renter households vs 3.8M affordable units · S · JCHS SONH 2026 (fact 5)
2 · Stein · 9.3M sub-$1,400 units lost 2014–24 · S · JCHS Rental 2026
3 · Stein · Rents fell (Q4 2025) while burdens rose in 88/100 metros · S · JCHS Rental 2026. Note: burden figure is a 5-yr window incl. 2021–22 spike; the contrast is real but the timing isn't simultaneous.
4 · Stein · New luxury buildings raise nearby rents / cause displacement · U · Contradicted by best causal evidence: Asquith–Mast–Reed (−5–7%), Pennington (−$40, displacement −20%), Mast 2023
5 · Stein · SF rent control: supply −15%, mobility −20% · S · Diamond–McQuade–Qian AER 2019
6 · Stein · Pennington: affordable buildings no spillover; complements · S · Berkeley Matrix interview
7 · Stein · Austin decline follows +14.7% peak; NMHC is industry · S · NMHC 2025
8 · Stein · OR/CA caps = inflation + margin, new construction exempt 15 yrs · C · Panel did not open statutes this round; consistent with panel's knowledge of OR SB 608 / CA AB 1482
9 · Stein · No Diamond-quality causal study of OR/CA-style caps · C · Not verified (panel did not search)
10 · Delgado-Finch · Investors bought "a fifth of the single-family homes in the Sun Belt" · W · GAO-26-108675: <1%–3% of all SF homes in six metros incl. Jacksonville, Dallas, Phoenix; ~3% nationally. 22% is Jacksonville's share of SF rentals
11 · Delgado-Finch · Nashville: 35% of investor homes bought from owner-occupants · S · GAO 2026
12 · Delgado-Finch · LIHTC units can exit after 15 years · S · Qualified-contract process (Wikipedia summary; secondary)
13 · Delgado-Finch · Algorithmic rent-setting in single-family rentals · U · No source offered; withdrawn in cross-exam
14 · Delgado-Finch · Atlanta: >72,000 investor SFRs, >1 in 4 · C · Senate sponsor's statement via HousingWire; GAO did not study Atlanta; not independently verified
15 · Delgado-Finch · Phoenix +177%, Dallas +114% (2018–24) · S · GAO blog 2026
16 · Delgado-Finch · Investors sell ≤8% of holdings/yr · S · GAO 2026
17 · Delgado-Finch · $326k/unit high-end LIHTC cost · S · GAO 2018 via Wikipedia; dated (2011–15 data)
18 · Mensah · Houston: ≥34,000 townhouses, $340k vs $545k · S · Pew 2023
19 · Mensah · Minneapolis "froze rents by legalizing triplexes" · W · Minneapolis Fed 2025: 225 duplex/triplex units 2020–24; Pew: ~1% of permits; rent divergence starts 2020 with demand shock
20 · Mensah · Minneapolis rents +1% vs +14% rest of state (2017–22) · S · Pew 2024 (descriptive; attribution contested — see #19)
21 · Mensah · New market-rate supply lowers nearby rents and displacement · S · Asquith–Mast–Reed; Pennington
22 · Mensah · 100 new residents → 45–70 moves out of below-median areas · S · Mast JUE 2023
23 · Mensah · Montana Supreme Court upheld reforms unanimously (Mar 2026) · S · APA Planning 2026
24 · Mensah · Auckland: ~84% of consents became net additions · S · Land Use Policy 2025
25 · Mensah · Portland ≈1,400 middle/ADU units in ~3 yrs · S · News report of city progress report (secondary)
26 · Farrow · Auckland rents "fell by about a third" · W · Review: ~28% relative to counterfactual, not an observed decline of one-third
27 · Farrow · Auckland rezoned "three-quarters" of residential land · U · Not in any source the panel opened
28 · Farrow · SB 9 lot splits carry owner-occupancy strings · C · Not verified from statute this round
29 · Farrow · Montana litigation cost ~2 years · S · Dec 2023 suit → Mar 2026 ruling (APA)
30 · Farrow · Pennington: displacement to poorer zips −20% · S · Berkeley Matrix
31 · Farrow · CA builds one-for-one replacement into state laws · C · Not verified this round
32 · Mbeki · Household growth 2.0M → 1.1M; NIM −50% (2025) · S · JCHS 2026
33 · Mbeki · Supply elasticity doesn't predict price growth across metros · C · LMW 2025/26 says so; Furth 2025 disputes identification. Open
34 · Mbeki · ~85% of metros: units grew faster than population 2000–20 · S · FRBSF EL 2026
35 · Mbeki · Minneapolis divergence reflects a 2020 demand shock · S · Minneapolis Fed 2025 (authors' inference, supported by timing + 9.5% vacancy)
36 · Mbeki · National shortage "mostly an artifact" · C · Estimates 0–7.3M (Urban 2025); amended in cross-exam
37 · Mbeki · Freddie Mac adds ~0.4M missing households · S · JCHS blog
38 · Hartley · Auckland: +21,800–43,500 units, ~28% rent effect · S · Land Use Policy 2025
39 · Hartley · Austin pop +5.5%, employment +8.0% (2020–22) · S · NMHC 2025
40 · Albrecht · Construction inputs +40% since Jan 2020 · S · JCHS 2026
41 · Albrecht · ~45% combined duty on Canadian lumber; 1/3 imported; 85% from Canada · S · NAHB 2025 (industry)
42 · Albrecht · Completions −27% y/y; MF under construction 996k→686k · S · Census Aug 2026; JCHS 2026
43 · Albrecht · Rising permits imply entitlement isn't binding at the margin · C · Permits +3.5% y/y is fact; inference untested
44 · Albrecht · Katerra's collapse as modular cautionary tale · C · Unsourced here; panel believes accurate (ceased operations 2021) but opened no source
45 · Quinn · Duties offset Canadian stumpage subsidies · C · Long-running trade dispute; not adjudicated here
46 · Quinn · Domestic mills at ~64% capacity · S · NAHB 2025
47 · Quinn · Lumber is a "modest share" of the 40% input rise · U · JCHS gives no decomposition; no source
48 · Pruitt · Hsieh–Moretti show GDP "36% higher" without constraints · W · 36% is the corrected perfect-mobility figure (imperfect ≈14%); Greaney AEJ:Macro 2026 finds replication failure and effects ~two orders of magnitude smaller
49 · Pruitt · Houston lacks use zoning; lot-size reform result · S · Pew 2023
50 · Pruitt · Oregon's cap "binds rarely because it's loose" · U · No source offered
51 · Venkataraman · Greaney's replication critique · S · AEJ:Macro 2026
52 · Venkataraman · Deed restrictions cover much of Houston · C · Not verified this round
53 · Ashford · SB 9 "produced almost nothing" · C · Year-one Terner finding supports "rarely used"; LA later reported 569 duplexes permitted — "almost nothing" overstated
54 · Ashford · Montana has no production numbers yet · S · APA 2026
55 · Ashford · Chronic homelessness +81% since 2013 · S · HUD AHAR 2025 release (HUD's figure)
56 · Ashford · Housing First caused that rise · U · Descriptive count; no causal design; confounded with rents
57 · Ashford · Homelessness "primarily" addiction/mental illness · U · As an explanation of levels across places, contradicted by Colburn & Aldern; at individual level these are risk factors (not rated)
58 · Kerr · Rents/vacancy explain more cross-metro variation than poverty, SMI, drugs · S · Colburn & Aldern 2022 (Sightline summary)
59 · Kerr · West Virginia example · C · Not in the summary the panel opened
60 · Kerr · King County ~5/1,000 vs Miami-Dade ~1/1,000 · S · Sightline summary
61 · Holloway · Family Options results (shelter, separations, foster care, substance use, IPV) · S · HUD USER
62 · Oyelaran · Lock-in: 1.33M sales prevented, prices +5.7% · S · FHFA WP 24-03
63 · Oyelaran · Existing sales 30-yr low; record-low mobility · S · JCHS 2026
64 · Oyelaran · Price-to-income ~5 vs ~3.2 · S · JCHS 2026
65 · Oyelaran · 2020–22 run-up driven by rate collapse + stimulus · C · Plausible; not tested in this brief
66 · Morrow · Prices +54% since 2020 · S · JCHS 2026
67 · Oyelaran · ROAD Act exempts build-to-rent · S · BPC; HousingWire
68 · Delgado-Finch · ROAD Act cap at 350+ homes · S · BPC; HousingWire

Tally (68 claims): Supported 43 · Contested 14 · Unsupported 7 · Wrong 4.

Panel comment (Takahashi): The "Wrong" ratings share a pattern — denominator and attribution errors: all SF homes vs SF rentals (#10); an observed decline vs a counterfactual gap (#26); a reform component vs a reform package (#19); a retracted-and-disputed macro estimate (#48). The Unsupported ratings are mostly causal leaps from descriptive counts (#56, #57, #4).

---

On-the-record retractions and amendments

• Stein (#4): "Retracted. The best causal evidence shows new market-rate buildings lower nearby rents modestly; my argument is about magnitude and the ELI gap, not sign."
• Delgado-Finch (#10, #13): "Retracted 'a fifth of Sun Belt homes.' Correct statement: ≤3% of all single-family homes in GAO's metros, up to 22% of single-family rentals (Jacksonville). Withdrawn: algorithmic pricing in SFR."
• Mensah (#19): "Amended: Minneapolis's added supply came from larger buildings enabled by parking and corridor reforms; the triplex provision produced ~225 units; rent attribution to the 2040 Plan is unclear per the Minneapolis Fed."
• Farrow (#26, #27): "Amended: ~28% below counterfactual. Withdrawn: 'three-quarters of land' — I can't source it."
• Pruitt (#48, #50): "Retracted the 36% GDP figure. Withdrawn the Oregon claim pending evidence."
• Ashford (#56, #57): "Withdrawn the Housing First causal attribution. Amended: metro-level homelessness tracks housing costs; addiction and mental illness are central individual risk factors."
• Quinn (#47): "Withdrawn; I have no decomposition."

STAGE 5: STEELMAN

Each bloc writes the strongest version of an opposing case; the opposing bloc responds.

---

Left steelmans the Right (written by Stein #73 and Venkataraman #19)

"The strongest right-of-center case is not 'markets good.' It is this: the price of housing in high-demand places is mostly a regulatory artifact, and every other intervention is fighting the symptoms. Glaeser–Gyourko's wedge between price and construction cost is large exactly where discretionary review is strongest. The cleanest natural experiments we have — Houston's lot-size reform, Auckland's upzoning — show that when you legalize supply, supply comes and prices of new entry-level product fall ($340k vs $545k; ~28% relative rent effect). The causal filtering literature, including studies from San Francisco, shows market-rate construction lowers nearby rents and reduces displacement. Rent control, by contrast, has a clean natural experiment showing a 15% supply reduction. Subsidy programs are rationed (1 in 4 eligible) precisely because each subsidized unit costs $126k–$326k to build; every dollar of subsidy goes further when land and approval costs fall. And on federalism: preemption by states is legitimate, but federal micromanagement of zoning risks one-size-fits-all rules in a country where Houston and Boston face opposite problems. Finally, on macro: the collapse in existing-home turnover since 2022 is a rates-and-lock-in story (FHFA's 1.33M lost sales), not a 'greed' story, and investor bans restrict one of the few channels that added single-family rental supply, including build-to-rent."

Right bloc reply (Pruitt #8, Ashford #31, Oyelaran #12): Accepted as fair, with one correction from Ashford: "Our federalism case isn't just about one-size-fits-all rules. It's that local self-government is a good in itself, not merely an instrument. I'll pay some price premium for it; I'd like the Left to acknowledge that's a values trade-off, not an error." Pruitt adds: "And we no longer need Hsieh–Moretti to make the case — we've dropped it."

---

Right steelmans the Left (written by Oyelaran #12 and Ashford #31)

"The strongest left case: the housing crisis for the bottom third is an income-rent gap that no plausible supply response closes on a relevant timescale. 11 million ELI households, 3.8 million units they can afford. Rents fell nationally in 2025 and burdens still rose. Filtering is real but slow — Mast's chains take three years to move 45–70 people per 100 new residents, and the effects per building are 5–7% within two blocks. Meanwhile the low-rent stock is being deleted faster (9.3 million units in a decade) than filtering replenishes it. The only randomized evidence on ending family homelessness — Family Options — shows vouchers work, on housing and on child welfare, for ~9% more than usual care. The rationing of vouchers (1 in 4) is a budget choice, not a market fact. Across metros, homelessness tracks rents and vacancies, not addiction or poverty rates, so treatment-only approaches will fail in expensive metros. And guardrails matter politically: upzoning without anti-displacement protections loses at the ballot, and 'loose' rent caps with new-construction exemptions are a different instrument from 1994 San Francisco."

Left bloc reply (Stein #73, Delgado-Finch #7, Holloway #71): Accepted as fair. Delgado-Finch adds a correction: "You left out ownership. Our strongest claim is that affordability rented from private owners for 15–30 years (LIHTC) expires, and permanently affordable, non-market stock is the only thing that doesn't. That's the socialist wing's core argument, even if the U.S. cost record is poor."

---

Center steelmans the anti-supply skeptics (written by Birch #10 — as specialist, not referee — and Farrow #34)

"The strongest skeptic case (Louie–Mondragon–Wieland; the Minneapolis Fed): most 'upzoning worked' stories confuse demand shocks with supply effects. Minneapolis's rent divergence starts in 2020 with a 9.5% vacancy spike; its triplex reform produced 225 units. Austin's rent decline coincided with a demand cooldown. In ~85% of metros, units outgrew population. If income growth — especially top-end income — drives prices one-for-one regardless of elasticity, then upzoning mostly changes who lives in a place and the quality of housing, not its price level. Macro claims like Hsieh–Moretti have not survived replication. So the honest position is: local, building-level effects are real and small; metro-level effects are demonstrated in one strong case (Auckland); national effects are unproven."

Mbeki (#70) reply: Accepted as fair, with a correction: "I'd add that the national household-growth slowdown (2.0M → 1.1M) and the immigration drop make 2026–28 an unusually informative test. If rents stay flat with completions down 27%, that's evidence for our side."

---

Center steelmans the YIMBY left (Albrecht #6)

"The YIMBY-left case: discretionary review is the mechanism that converts neighborhood preferences into regional scarcity, and its costs fall on renters and the young. Pennington shows new supply reduces displacement even in San Francisco; the anti-displacement movement's instincts have been pointed at the wrong target."

Mensah (#95) reply: Accepted as fair. "I'd add: we want subsidy too. The YIMBY left is not the libertarian right."

STAGE 6: REVISED PROPOSALS

Eight proposals emerged from the cross-examination. For each: mechanism · cost/scoring · precedent · key implementation risk. "No official score" means we found no CBO/JCT estimate.

---

P1. State by-right "missing middle + transit" package with anti-displacement guardrails

Mechanism. State law (model act, encouraged but not mandated federally): by-right ministerial approval for 2–4 units on any residential lot; mid-rise (to ~6 stories) within walking distance of major transit; abolish parking minimums near transit; minimum lot size ≤ 1,400–2,000 sq ft; single-stair permitted to 6 stories; shot clocks; builder's-remedy style consequences for non-compliant cities. Guardrails (Walker–Farrow bargain): one-for-one replacement of demolished rent-regulated units, tenant right to counsel in eviction.
Cost. No official score (state policy). Right-to-counsel carries modest state costs.
Precedent. Auckland 2016 (+21,800–43,500 units, ~28% rent effect vs counterfactual, Land Use Policy 2025); Houston lot-size reform (≥34,000 townhouses, Pew 2023); Oregon HB 2001/Portland (~1,400 units in ~3 yrs); California SB 79 (2025, transit upzoning; too new to evaluate); Montana 2023/2025 (upheld 2026; no data yet). Cautionary: SB 9's low uptake.
Risk. Paper capacity: legal permission without financing, infrastructure, or feasible pro formas (SB 9). Local resistance via fees and delays.

P2. Scale and harden the ROAD Act supply-incentive grant

Mechanism. Raise the $200M/yr competitive grant to a larger pool; award on completions per capita relative to a baseline, not plans; extend the CDBG production bonus.
Cost. CBO scored the ROAD Act's direct spending at ~$0 net and did not estimate discretionary spending (CBO 2026). Expansion: no official score.
Precedent. ROAD Act (2026) itself; earlier federal "PRO Housing" grants (not evaluated here).
Risk. Grants too small to change local politics; rewards places already building (Sun Belt), windfall rather than behavior change.

P3. Phase in housing vouchers as an entitlement for ELI families with children and households exiting homelessness

Mechanism. Guarantee a Housing Choice Voucher to ELI families with children, phased over 10 years; pair with mobility counseling; small-area FMRs.
Cost. No official score verified in this review; given 5.3M people served now and only ~1 in 4 eligible households assisted (CBPP), the cost would be large (tens of billions annually at full phase-in — order of magnitude only, unverified).
Precedent. Family Options RCT: vouchers cut shelter returns, roughly halved child separations and foster placements, at ~9% more than usual care (HUD).
Risk. In tight markets, vouchers without supply bid up rents; voucher discrimination by landlords; fiscal cost.

P4. LIHTC permanence and conversion protections

Mechanism. Building on the 2025 expansion (12% allocation boost, 25% bond test; JCT ~$15.7B/10 yrs; Novogradac projects ~1.22M additional units), close the 15-year qualified-contract exit and give nonprofits/tenants right of first refusal at year 15.
Cost. No official score for the closure; likely small revenue effect, but it lowers the credit's value to investors (raising cost per unit).
Precedent. Many states already require extended-use waivers of the qualified contract in their QAPs (not verified here in detail).
Risk. Reduced investor pricing of credits → fewer units per dollar; crowd-out concerns (literature unverified this round).

P5. Exempt residential building materials from Section 232 tariffs

Mechanism. Exempt softwood lumber and gypsum from allies, and residential cabinets/vanities, from Section 232 duties; keep AD/CVD on Canadian lumber subject to negotiation.
Cost. No official score; revenue loss.
Precedent. The 2025 duties themselves (NAHB); prior lumber disputes. No clean evaluation verified.
Risk. Trade retaliation politics; domestic mill investment signals weakened (Quinn); savings may be captured as margin rather than lower prices.

P6. Priced mortgage portability/assumability pilot through the GSEs

Mechanism. FHFA directs Fannie/Freddie to pilot portability of existing low-rate mortgages to a new home for a fee priced to the rate risk; streamline assumability.
Cost. No official score; contingent GSE/taxpayer risk.
Precedent. Portable mortgages exist in Canada and the UK (not verified here in detail). FHFA WP 24-03 quantifies lock-in (−18.1% sale probability per point; 1.33M sales prevented).
Risk. Regressive (benefits incumbents who locked in low rates); adverse selection; mispricing.

P7. Evaluate-then-sunset the ROAD Act institutional-investor purchase ban

Mechanism. Require GAO/HUD to evaluate metro-level price, rent and homeownership effects of the 350-home cap within three years (using GAO's parcel data); ban sunsets in year five unless the evaluation finds measurable owner-occupancy or price benefits.
Cost. Minimal (evaluation).
Precedent. GAO-26-108675 provides the baseline data. No prior U.S. federal ban exists to evaluate.
Risk. Evaluation designs are hard (investor share is 1–3% of SF homes); a null result may reflect low power, not no effect.

P8. Anti-gouging rent cap with new-construction exemption (state level, CPI + ~5–7%, 15-year new-construction exemption)

Mechanism. Oregon/California-style annual cap, not SF-1994-style strict control; vacancy decontrol retained.
Cost. No official score.
Precedent. Oregon (2019), California (2019) — design details not verified this round; no Diamond-quality evaluation found by the panel; Diamond–McQuade–Qian (AER 2019) shows supply losses from strict SF-style control.
Risk. Ratchet toward tighter caps (Pruitt); cap becomes a de facto floor for increases; maintenance deferral.

STAGE 7: RECORDED VOTE

Blocs: Left 34 · Center 33 · Right 32 (moderator #100 abstains). Bridge rule: CONSENSUS only if ≥60% in each bloc (Left ≥21, Center ≥20, Right ≥20).

Proposal · Left yes · Center yes · Right yes · Consensus?
P1. State by-right missing-middle + transit package w/ anti-displacement guardrails · 25/34 (73.5%) · 29/33 (87.9%) · 20/32 (62.5%) · YES
P2. Scale ROAD Act supply grant, pay on completions · 28/34 (82.4%) · 25/33 (75.8%) · 13/32 (40.6%) · No
P3. Voucher entitlement for ELI families w/ children + exiting homelessness (10-yr phase-in) · 33/34 (97.1%) · 21/33 (63.6%) · 7/32 (21.9%) · No
P4. LIHTC permanence: close 15-yr qualified-contract exit; ROFR · 31/34 (91.2%) · 22/33 (66.7%) · 12/32 (37.5%) · No
P5. Exempt residential building materials from Sec. 232 tariffs · 26/34 (76.5%) · 30/33 (90.9%) · 21/32 (65.6%) · YES
P6. Priced GSE mortgage portability pilot · 11/34 (32.4%) · 20/33 (60.6%) · 25/32 (78.1%) · No
P7. Evaluate-then-sunset the investor purchase ban · 10/34 (29.4%) · 24/33 (72.7%) · 26/32 (81.3%) · No
P8. State anti-gouging rent cap w/ 15-yr new-construction exemption · 30/34 (88.2%) · 12/33 (36.4%) · 3/32 (9.4%) · No

Consensus reforms: P1 (narrowly — Right at exactly 20) and P5.

Why the Right vote on P1 is so thin: the 8 libertarians and most of the center-right voted yes; most of the traditional-right federalists voted no on state preemption of cities, or voted yes only because the proposal is state-level, not federal. Why the Left vote on P1 isn't higher: the socialist wing and part of the anti-displacement left voted no despite the guardrails.

---

Vote explanations

Prof. Benjamin Ashford (#31, R) — P1 Yes (reluctantly), P2 No, P7 Yes.
"I voted yes on P1 because it's the state acting on its own creatures — cities — and Montana's Supreme Court affirmed that. I voted no on P2 because federal zoning money grows into federal zoning conditions. What would change my mind on P1: if by 2029 Montana and Oregon show no measurable increase in 2–4-unit production after compliance, I'll conclude preemption is costly symbolism and vote to repeal."

Prof. Rosa Delgado-Finch (#7, Soc) — P1 No, P3 Yes, P4 Yes, P8 Yes, P7 No.
"I voted no on P1 because 'by-right' hands the land-value gain from upzoning to private owners and developers with no public capture. What would change my mind: a version that captures a share of the upzoning uplift for permanently affordable units — or evidence that sub-$1,000 inventory is rising in Auckland-style reformed metros by 2028. I'll own the retraction on investor shares; my vote against P7 stands because three years is too short to detect effects on a 1–3% ownership share."

Prof. Walter Pruitt (#8, Lib) — P1 Yes, P5 Yes, P6 Yes, P7 Yes, P3 No, P8 No.
"P1 is the achievable core of what I actually believe. I voted no on P3 because in supply-constrained metros vouchers mostly bid up rents. What would change my mind on P3: a demonstration that voucher expansion in a supply-elastic metro doesn't raise rents for non-voucher renters at the bottom — and I'd accept the Family Options evidence as showing large benefits to recipients; my objection is incidence, not efficacy."

Prof. Daniel Mbeki (#70, C) — P1 Yes, P3 Yes, P5 Yes, P6 No, P8 No.
"Yes on P1 despite my skepticism about national shortage numbers: the local and Auckland evidence is good enough, and the downside is small. Yes on P3 because the ELI gap is an income problem, and that's where the only RCT points. No on P6: it's a transfer to incumbents to fix a transitory problem that falling rates would fix. What would change my mind on the supply question as a whole: rents re-accelerating above 4% by late 2027 with low immigration — then the stock story wins."

Prof. Grace Holloway (#71, L) — P1 Yes, P3 Yes, P4 Yes, P8 Yes, P6 No.
"I broke with part of my bloc on P1: Colburn and Aldern's data tell me homelessness follows rents and vacancy, and P1 is the only proposal that moves vacancy at scale. What would change my mind: if cities that adopt P1-type reforms don't see rental vacancy rise and per-capita homelessness fall relative to non-adopters within five years."

Prof. Teresa Quinn (#90, R) — P5 No, P1 No, P7 Yes.
"No on P5 as drafted because it drops the cabinet duties on China with the allied-lumber relief. I withdrew my unsourced claim about lumber's share of cost inflation, so I won't lean on it. What would change my mind: evidence that domestic mill capacity utilization did not rise after the 2025 duties — then the tariffs are all cost and no capacity signal."

STAGE 8: VERDICT — Prof. Adelaide Wainwright (#100, moderator)

Established (evidence strong, across ideological lines)

1. Affordability is at or near modern lows by the standard measures. Price-to-income near 5 vs ~3.2 in the 1990s; 22.7M cost-burdened renters (49%) and 20.7M cost-burdened owners; homeownership down to 65.2% (JCHS 2026).
2. The low-rent stock is shrinking fast (−9.3M sub-$1,400 units in a decade), and the ELI gap (11.0M households vs 3.8M units) is not closed by soft average rents.
3. New market-rate housing lowers nearby rents and reduces displacement, modestly. Asquith–Mast–Reed, Pennington and Mast agree on the sign. The claim that such buildings raise nearby rents does not survive, and its proponent withdrew it.
4. Legalizing supply produces supply where demand exists: Houston's lot-size reform and Auckland's upzoning are the strongest cases, and Auckland also shows a meaningful rent effect relative to its counterfactual.
5. Strict rent control reduces rental supply (SF 1994: −15%) even as it protects sitting tenants.
6. Vouchers work for their recipients. The Family Options RCT is the strongest causal evidence in this thread. Rationing (~1 in 4 eligible served) is a budget choice.
7. Lock-in is real and large for transactions and mobility (FHFA: −18.1% sale probability per point; 1.33M lost sales), and moderate for prices (+5.7%).
8. Institutional investors are a small national share (~3% of SF homes) but concentrated (up to 22% of SF rentals in Jacksonville). Congress has now capped them anyway.

Contested

• The size of the national shortage (0 to 7.3M; Urban's ~2M at the low end), and whether "national" is even the right unit. The number is shrinking as household growth halves.
• Whether supply constraints explain cross-metro price growth (Glaeser–Gyourko / Furth vs Louie–Mondragon–Wieland). This is the live frontier dispute.
• Minneapolis's rent story (Pew supply vs Minneapolis Fed demand shock). The careful reading: attribution to the 2040 Plan is unproven, and the triplex provision itself did little.
• Whether tariffs and immigration enforcement are now a first-order cost driver. The direction is clear, but there is no verified per-home magnitude.
• Whether loose anti-gouging caps harm supply. No causal evaluation was found.

Unknown

• Causal effects of institutional investors on metro prices, rents and homeownership.
• Whether state preemption laws (Montana, Oregon, California SB 79) deliver production at scale. Most are too new, and SB 9 is a warning.
• Incidence of a large voucher expansion on non-recipient rents in supply-constrained metros.

Research agenda

1. A pre-registered evaluation of the ROAD Act investor cap using GAO parcel data, with a power analysis before anyone claims a null result.
2. Montana/Oregon/California SB 79 as staggered-adoption natural experiments: 2–4-unit and mid-rise permits and completions, plus rents, 2026–2030.
3. Sub-$1,000 rental inventory in post-boom metros (Austin, Phoenix) through 2028: the joint Stein–Pruitt test.
4. Resolve the Louie–Mondragon–Wieland vs Furth identification dispute using policy-induced elasticity changes rather than estimated elasticities.
5. Evaluate Oregon/California rent caps with Diamond-style designs.
6. Mbeki's test: do national real rents stay flat through 2027 with completions down 27% and immigration down sharply?

Consensus reforms

• P1. State by-right missing-middle + transit reform with anti-displacement guardrails. It passed narrowly; the Right cleared the bar at exactly 20.
• P5. Exempt residential building materials from Section 232 tariffs.

Notably, nothing federal and nothing fiscal reached consensus. Vouchers, LIHTC permanence and larger federal incentives split Left/Center from Right. Rent caps split Left from everyone else. Portability and the investor-ban sunset split Right/Center from Left.

Where Symposium 1 was wrong or too glib (as flagged by this panel)

To the extent the first round leaned on these common talking points, this round corrects them:

• "The U.S. is short X million homes" as a single settled number. It isn't settled. Estimates differ by construct, and the national figure matters less than metro price-to-cost gaps and the ELI affordability gap.
• "Zoning costs the U.S. economy a third of GDP" (Hsieh–Moretti). This should not be cited without Greaney's 2026 replication critique. The micro case for reform does not need it.
• "Minneapolis proves upzoning cuts rents." The Minneapolis Fed's demand-shock analysis makes this unproven, and the triplex provision produced ~225 units.
• "Wall Street owns the Sun Belt." It owns ≤3% of single-family homes in GAO's metros, though concentration in rentals is real.
• "Homelessness is a drug problem" / "Housing First failed." Metro-level variation tracks rents and vacancy. HUD's 81% chronic figure is descriptive, not causal.
• "Supply alone solves it." No evidence here shows market filtering closing an 11M-vs-3.8M ELI gap on a policy-relevant timescale. Subsidy is a complement, as the San Francisco filtering evidence itself suggests.

The honest bottom line is that both "build more" and "subsidize the bottom" are supported by strong evidence, each for a different part of the problem. The ideological fight is mostly over who pays, and over which level of government decides.

— A. Wainwright (#100)

Independent Audit

Audit Panel — Prof. Laura Pennington (#97, operations research, C) & Prof. Olivia Brandt (#57, statistics, C). Brandt sat on this thread's referee panel, so she recuses from any judgment on the Stage 4 ratings. Pennington leads and signs every finding below about referee rulings. (Pennington is no relation to Kate Pennington, the Berkeley author cited in fact 28.) Corrections are appended here. Nothing above has been deleted or edited.

1. Source check (17 claims, fetched 2026-09-24)

# · Claim · Cited figure · Verified figure · Status · URL
1 · Renter cost burden (2024) · 22.7M (49%) >30%; 12.1M (26%) >50% · Same · Confirmed · jchs.harvard.edu …SONH_2026_Press_Release.pdf
2 · Owner burden; homeownership; mobility · 20.7M (24%); 65.2%; 4.1M sales; 11.2% · Same (SONH full report) · Confirmed · jchs.harvard.edu …State_of_the_Nations_Housing_2026_0.pdf
3 · Price-to-income; prices since 2020; taxes/insurance · "nearly five times"; +54%; +31% / +72% · Same · Confirmed · JCHS press release
4 · Low-rent stock; ELI gap · −9.3M <$1,400; +11.8M ≥$1,400; ~7M <$1,000 lost; 11.0M ELI vs 3.8M units · Same. The ELI and <$1,000 figures are in SONH 2026, not the Rental press release the brief cites for them · Confirmed (attribution note) · jchs.harvard.edu/press-releases/new-report-finds-cooling-rental-markets…
5 · Rental market softness · −0.6% y/y asking rents; 5.2% vacancy; burdens up in 44 states / 88 of 100 metros; MF under construction 996k→686k · Same · Confirmed · same
6 · Census, August 2026 · Completions 1,128k, −27.1% y/y (from 1,548k); permits +3.5%; starts −1.2% · Same · Confirmed · census.gov/construction/nrc/pdf/newresconst.pdf
7 · Freddie Mac PMMS · 6.95% (9/17/26); 6.76% prior week; 6.26% a year earlier · Same · Confirmed · globenewswire.com …mortgage-rates-average-6-95
8 · FHFA WP 24-03 lock-in · −18.1% per point; 1.33M sales; +5.7% vs −3.3% · Same (2022Q2–2023Q4) · Confirmed · fhfa.gov/research/papers/wp2403
9 · HUD AHAR 2025 · 745,652; −3%; 266,320 unsheltered; 1,456,923; chronic +81% since 2013 · Same. HUD attributes the −3% to "sanctuary cities," which supports the referees' framing caveat · Confirmed · hud.gov/news/hud-no-26-037
10 · GAO-26-108675 · <1–3% of all SF homes; 4% (Seattle)–22% (Jacksonville) of SF rentals; Nashville 35% / 15%; ≤8% sold per year · Same · Confirmed · gao.gov/products/gao-26-108675
11 · ROAD Act (BPC) · Enacted 7/11/26; 350+ home cap; BTR exempt; $200M/yr, 7-yr sunset; chassis; NEPA; CDBG · Same. "Without the President's signature" (HousingWire) was not checked · Confirmed · bipartisanpolicy.org/issue-brief/inside-the-deal…
12 · CBO, H.R. 6644 · Outlays −$6M, revenues −$6M, net ≈$0; discretionary not estimated · Same · Confirmed · cbo.gov/system/files/2026-07/hr6644.pdf
13 · Greaney (AEJ:Macro 2026) · Counterfactual would lower output; unit dependence; "two orders of magnitude smaller" · Verbatim in abstract · Confirmed · ideas.repec.org/a/aea/aejmac/v18y2026i2p409-28.html
14 · Auckland (Land Use Policy 2025) · "Added 21,800–43,500 units"; ~28% rent effect; 84% · 21,808 (5 yrs) and ~43,500 (6 yrs) are additional building consents, not units; ~84% of consents become net additions; 28% is vs synthetic counterfactual · Minor discrepancy · sciencedirect.com/science/article/pii/S0264837725000316
15 · Minneapolis Fed 2025 · 225 units in 87 duplex/triplex bldgs; 9.5% vacancy (Oct 2020); permits −77.6% / −92.4% · Same numbers, but the 87 buildings are "duplex, triplex, and fourplex" · Minor discrepancy · minneapolisfed.org/article/2025/unpacking-supply-and-demand…
16 · Pew, Houston · Min lot 5,000 → 1,400 sq ft (1998); ≥34k townhouses; $340k vs $545k; 0.5% · 1998 cut: 5,000 → 3,500 sq ft standard, 1,400 only with open-space conditions. Other figures match · Minor discrepancy · pew.org/…/lot-size-reform-unlocks-affordable-homeownership-in-houston
17 · Family Options (HUD USER) · Shelter ↓; child separations "almost halved"; foster placements "more than halved"; +9% cost; (Holloway) substance use and IPV ↓ · Same, including substance use and IPV · Confirmed · huduser.gov/portal/family_options_study.html

Also checked: Upjohn (5–7%, ~2-block radius) and APA Planning (unanimous ruling 3/17/26; suit filed Dec 2023; cities need not comply until May 2026). Both Confirmed.

Tally: Confirmed 14 · Minor discrepancy 3 · Not supported 0 · Could not access 0.

2. Internal consistency

• Vote math: all 24 yes-count ↔ % pairs recomputed and correct (P7 Right 26/32 = 81.25%, shown as 81.3%). Consensus labels are correct. P1 clears with Right at exactly 20/32, and P5 at 21/32. Every other proposal fails at least one bloc. No mislabels.
• Fact-check tally: 68 rows = S 43 · C 14 · U 7 · W 4. This matches the stated tally.
• Verdict vs fact-check: no claim rated Wrong comes back as established. Verdict item 4 ("Legalizing supply produces supply… Auckland") should read "additional consents, ~84% of which become net additions" (see row 14).
• Roster: all 19 speakers match the roster on name, number, field and lean. "The 8 libertarians" in the vote note matches the roster (#4, 8, 16, 37, 58, 75, 94, 99).
• Corrections appended:

1. Brief fact 33 and P1's "≤1,400 sq ft": the 1998 Houston reform cut the standard minimum to 3,500 sq ft. The 1,400 sq ft figure applied only to lots meeting open-space standards.
2. Fact-check row 53: "LA later reported 569 duplexes permitted" is a referee-supplied number with no source in the thread. Treat it as unverified.
3. Procedure: Birch was a referee, and he also co-wrote a Stage 5 steelman. The post discloses this ("as specialist, not referee"), but the dual role should be avoided in future rounds.
4. Rating convention: this thread rates "plausible but unverified" claims C (#8, 9, 28, 31, 44, 52, 59). Thread 34 rates the same kind of claim U. That inflates Contested here relative to other domains, so readers comparing tallies across threads should adjust.

3. Balance

Non-Supported ratings fall across all blocs. Wrong ratings: one each for Soc (Delgado-Finch), C-L (Mensah), C (Farrow) and Lib (Pruitt). Unsupported: two Left, one Center, four Right. Every Right U is earned. Ashford's "primarily addiction" and "Housing First caused +81%" are causal leaps from a descriptive count. The Quinn and Pruitt ratings cover claims they withdrew. The unverified-claim convention (point 4 above) was applied to speakers in every bloc; it helped Left, Center and Right speakers alike (Kerr, Venkataraman, Albrecht, Quinn's stumpage point). Farrow's W on "fell by about a third" is the harshest call: 33% vs 28% is close in magnitude, and the real error is observed decline vs counterfactual gap. Row 4 is worth noting on the other side: Stein's claim is "contradicted" by the evidence, yet it is rated U rather than W. Row 57 gets the same treatment on the Right, so the two roughly offset. The verdict follows the evidence ("build more" and "subsidize the bottom"). No bloc is systematically disadvantaged.

4. Grade: A−

Every number we fetched checks out except three small descriptive slips (consents vs units, fourplexes, the Houston lot size). Vote and tally arithmetic are clean. The only procedural blemishes are the referee's dual role and the "C for unverified" convention.

— Pennington (lead); Brandt (recused on Stage 4 rulings)

Verdict v2 (audit-corrected)

Prof. Adelaide Wainwright (#100, moderator). This post is appended to the record. Nothing above it has been edited. It does three things: (1) resolves every claim the thread left unverified, plus the audit's "minor discrepancy" items; (2) moves the Stage 4 tally onto the standard scale (Supported / Contested / Unsupported / Wrong, where Unsupported includes "could not verify"); (3) restates the verdict with the corrected figures. Sources were fetched on 2026-09-24.

1. Resolution of unverified and audit-flagged items

Item · Claim · Result · Figure and source
Brief 33 / P1 (audit) · Houston cut minimum lots to 1,400 sq ft (1998) · Corrected · The standard minimum was cut from 5,000 to 3,500 sq ft. 1,400 sq ft applied only to lots that meet open-space conditions. Pew, Lot-size reform unlocks affordable homeownership in Houston (URL as cited in the audit, row 16)
Brief 31 / row 38 / Verdict 4 (audit) · Auckland "added 21,800–43,500 units" · Corrected · These are additional building consents (21,808 over 5 yrs; ~43,500 over 6 yrs). About 84% of consents become net additions. The ~28% rent effect is measured against a synthetic counterfactual. *Land Use Policy* 2025
Row 19 / Brief (audit) · Minneapolis: 225 units in 87 duplex/triplex buildings · Corrected · The 87 buildings are duplex, triplex and fourplex. The 225 units and the 9.5% vacancy figure stand. Minneapolis Fed 2025, Unpacking supply and demand… (URL as cited in the audit, row 15)
Row 8 · OR/CA caps = inflation + margin; 15-yr new-construction exemption · Now verified · Oregon ORS 90.324: the lesser of 7% + CPI or 10%. ORS 90.323: exempt if first certificate of occupancy is <15 yrs old. California Civ. Code 1947.12: the lesser of 5% + CPI or 10%; exempt if certificate of occupancy was issued in the prior 15 yrs; sunsets 1/1/2030. ORS 90.324, ORS 90.323, Civ. 1947.12
Row 9 · No Diamond-quality causal study of OR/CA-style caps · Still unverifiable · This is a claim that something doesn't exist. We did not run a literature search within budget
Row 14 · Atlanta: >72,000 investor-owned SFRs, >1 in 4 · Still unverifiable at a primary source · Secondary corroboration only: CBS Atlanta repeats the figure and attributes it to WABE reporting. We did not open the underlying dataset, and GAO did not study Atlanta
Row 28 · SB 9 lot splits carry owner-occupancy strings · Now verified (secondary) · The applicant must sign an affidavit of intent to occupy one unit as principal residence for ≥3 years. Wikipedia summary of the statute. The leginfo text was blocked to our fetcher
Row 31 · CA builds one-for-one replacement into state law · Now verified · Gov. Code 66300(d) (Housing Crisis Act/SB 330): a project must "replace all existing or demolished protected units" and include at least as many units as existed in the prior 5 years. Gov. Code 66300
Row 44 · Katerra's collapse · Now verified · Katerra filed for Chapter 11 on 2021-06-07. The panel had said it "ceased operations in 2021"; the filing is what we confirmed. ENR
Row 52 · Deed restrictions cover "much of" Houston · Corrected / still weak · The only estimates are dated: Siegan (1972) put coverage at "no more than 25 percent" of the city, and Kapur (2002) gave a similar figure. No GIS inventory exists. Market Urbanism summary. The City's FAQ gives no share. "Much of" is best read as "roughly a quarter, on old estimates"
Row 53 (audit) · LA "569 duplexes permitted" under SB 9 · Now verified (secondary) · "569 duplexes in single-family zones and 4 residential lot splits" permitted (LA 2024 report), per the Wikipedia summary. The rating stays C: "rarely used" in year one, but "almost nothing" is overstated later
Row 59 · West Virginia: severe addiction, low homelessness · Now verified · Colburn: West Virginia is "the home of the opioid epidemic… yet, homelessness is not one of the consequences." UCLA Lewis Center interview
Row 43 · Rising permits ⇒ entitlement not binding at the margin · Still unverifiable · An inference. No test was offered
Row 65 · 2020–22 run-up driven by rate collapse + stimulus · Still unverifiable · Plausible, but no decomposition source was opened
Brief (open q.) · LIHTC crowd-out literature · Still unverifiable · We did not open the primary papers. The P4 risk stays flagged
Brief (open q.) · Causal effect of institutional investors on prices and rents · Still unverifiable · We found no primary causal study. It stays on the Unknown list
P3 cost · Voucher entitlement ≈ "tens of billions/yr" · Still unverifiable · There is no official score. The figure is an order-of-magnitude guess only
P4 precedent · State QAPs already waive the qualified contract · Still unverifiable · Not checked
P6 precedent · Portable mortgages in Canada/UK · Still unverifiable · Not checked
P8 precedent · OR/CA design details · Now verified · See row 8

2. Rating normalization

This thread used C both for genuinely split evidence and for "plausible but not verified." Under the standard scale, the second kind is Unsupported. Ten rows are reclassified C → U: 8, 9, 14, 28, 31, 43, 44, 52, 59, 65. Rows 33, 36, 45 and 53 stay C, because the evidence on them is genuinely split.

· Supported · Contested · Unsupported · Wrong
Stage 4 as published · 43 · 14 · 7 · 4
Normalized scale · 43 · 4 · 17 · 4
After this post's resolutions (rows 8, 28, 31, 44, 59 → S) · 48 · 4 · 12 · 4

Old → new: S43 · C14 · U7 · W4 → S48 · C4 · U12 · W4 (68 claims). No Wrong rating changes. Row 52 stays U: the "much of the city" claim is weaker than stated.

3. Corrected verdict

Established

1. Affordability is at or near modern lows. Price-to-income is nearly 5, against ~3.2 in the 1990s. 22.7M renters (49%) and 20.7M owners are cost-burdened. Homeownership is 65.2% (JCHS 2026).
2. The low-rent stock is shrinking fast: −9.3M sub-$1,400 units in a decade. The ELI gap (11.0M households vs 3.8M units) is not closed by soft average rents.
3. New market-rate housing lowers nearby rents and reduces displacement, modestly (Asquith–Mast–Reed 5–7% within ~2 blocks; Pennington; Mast).
4. Legalizing supply produces supply where demand exists. Corrected figures: Houston's 1998 reform cut the standard minimum lot to 3,500 sq ft (1,400 only with open space) and yielded ≥34,000 townhouses. Auckland's upzoning generated ~21,800–43,500 additional building consents (not completed units), about 84% of which become net additions, with a ~28% rent effect relative to a synthetic counterfactual. The direction of the conclusion is unchanged. The magnitude claim is now stated in consents.
5. Strict rent control reduces rental supply (SF 1994: −15%) while protecting sitting tenants.
6. Vouchers work for their recipients (Family Options RCT). Rationing to ~1 in 4 of those eligible is a budget choice.
7. Lock-in is large for transactions (−18.1% sale probability per point; 1.33M lost sales) and moderate for prices (+5.7%).
8. Institutional investors are ~3% of SF homes nationally, but concentrated (up to 22% of SF rentals in Jacksonville). The Atlanta ">1 in 4" figure remains secondary-sourced.
9. (New, from resolutions) The "loose cap" designs are as described: OR is the lesser of 7% + CPI or 10%, CA is the lesser of 5% + CPI or 10%, both with 15-year new-construction exemptions. California law already requires one-for-one replacement of protected units (Gov. Code 66300). The latter means P1's guardrail extends existing California practice. It is not an untested idea.

Contested

• Size of the national shortage (0–7.3M) and whether "national" is the right unit.
• Whether supply constraints explain cross-metro price growth (Glaeser–Gyourko/Furth vs Louie–Mondragon–Wieland).
• Minneapolis's rent story. Corrected: the 2040-Plan small-building provision produced 225 units in 87 duplex, triplex and fourplex buildings. Attribution of the rent divergence to the Plan is unproven (Minneapolis Fed demand-shock analysis).
• Whether tariffs and immigration enforcement are first-order cost drivers. The direction is clear; no per-home magnitude is verified.
• SB 9's yield: rarely used at first. LA later permitted 569 duplexes (secondary source). Small, but not "nothing."

Unknown

• Causal effects of institutional investors on metro prices, rents and homeownership (no primary causal study found).
• Whether state preemption (Montana, Oregon, CA SB 79) delivers production at scale.
• Incidence of a large voucher expansion on non-recipient rents in constrained metros.
• Whether loose anti-gouging caps harm supply. This moves from Contested to Unknown. The designs are now verified, but no causal evaluation exists, so this is a gap in the evidence, not a split in it.
• (Added) LIHTC crowd-out magnitude under current market conditions (primary papers not opened).

Conclusions that changed.
(a) Verdict item 4 now states Auckland in consents and Houston at 3,500 sq ft. The conclusion that supply follows legalization stands, but the headline magnitudes are smaller and more precisely defined.
(b) Ashford's "deed restrictions do what zoning does" and Venkataraman's "deed restrictions cover much of Houston" are weakened. Dated estimates put coverage at about a quarter of the city.
(c) The rent-cap question is reclassified from Contested to Unknown.
(d) Proposal P1's lot-size spec (≤1,400–2,000 sq ft) goes further than the Houston precedent's general standard (3,500 sq ft). Readers should not treat Houston as direct evidence for a 1,400 sq ft citywide minimum.

The consensus list in the original verdict is left as recorded. The bottom line is unchanged: "build more" and "subsidize the bottom" are both supported, for different parts of the problem.

— A. Wainwright (#100)