Future Housing Systems

Housing systems is the delivery question: how a dwelling gets from a parcel of land to a household, what each step costs, and which of those costs actually sets the price paid. This brief takes four things in order. The cost decomposition, and which term dominates. The delivery models — site build, panelised, volumetric — and the failure record of the industrialised attempts. Mass timber, the one recent case where an engineered structural system argued its way into the codes on test evidence. And the exotic question at the end: whether housing can ever be put on a manufacturing cost curve at all.

The framing that organises everything here. Delivered price decomposes into five terms — land, hard cost, soft cost, finance, developer margin. Industrialisation, in every form it has taken, attacks hard cost and only hard cost. Whether that is worth doing depends on how large hard cost is against the other four, and on whether a reduction in it reaches the household or is captured upstream. The literature has answers to both questions. Neither answer is the one the popular argument assumes.

Two declared holes, stated at the front rather than buried. The research pack behind this brief could not locate a single source giving the five-term decomposition on a consistent boundary for any market, in any country, over any period. That is its highest-priority gap and this brief does not fill it from memory. The pack's fetch channel also failed partway through the session, and the modular-industry failure record went with it: Katerra is named nowhere in the retrieved literature, and no figure for its capital, its pipeline or its collapse appears anywhere in this brief. Section 14 says what would settle each.

Scope. The machines are Automated Construction Systems, on which this brief depends and from which it inherits a contested premise. Population-scale single structures are Arcologies; the subsurface variant is Underground Cities. Estimation and outturn are Megaproject Governance. What is left here is the house, the parcel, and the buyer.

Established The central quantity in this subject has never been published in comparable form. Everyone in the housing-cost argument reasons from a share — land is a third, hard cost is half, soft cost is the hidden killer — and nobody publishes the five shares on a stated boundary. Established Tsachageas states the problem from inside the field: land costs “play a pivotal role in housing affordability but are often misrepresented in urban research” (Critical Housing Analysis, 2025). Frontier His own contribution is a method rather than a number — GIS interpolation of land-price surfaces for Volos, Greece, showing that land prices can be tracked geostatistically even in small and imperfect markets. Established The methodological claim matters more than the Greek result: land is simultaneously the largest disputed term in the decomposition and the worst-measured one. Every argument about whether cheaper construction produces cheaper housing is being conducted without the arithmetic that would settle it.

Established Supply constraint is real, measured, and tightening. Baum-Snow and Duranton, reviewing United States housing prices, quantities and regulation since 1980, report descriptive evidence “indicating that supply constraints appear to be increasingly binding” (SSRN, 2025). Frontier The effect size has been estimated outside the United States as well. Huang, extending Saiz's supply-elasticity framework to 272 Chinese cities using land-transaction microdata and satellite-derived geography, finds government land regulation cutting housing supply elasticity from 1.086, which is elastic, to 0.750, which is not — a fall of more than 30 per cent (SSRN, 2021). Frontier Within that, land-use regulation carried a larger marginal effect than either land allocation or land supply regulation. Frontier And the finding the deregulation argument tends not to quote sits in the same abstract: physical geographic constraint had a marginal effect higher still than land-use regulation. Geography beat the planners. That is one estimate in one country, and it should be replicated before it is leaned on.

Frontier The single most consequential result for this brief is a causality test that ran the wrong way. Tsai, testing Taiwan's construction cost index, rental price index and house price index against one another, reports that “the HPI stimulates changes in the CCI and the RPI, although construction costs and rent are viewed as fundamentals in the existing literature”, that the effect is “rather obvious when deviations of the latter two indices from the HPI are greater”, and that the relationships are nonlinear (Asian Economic Journal, 2012). Frontier Read plainly: house prices drive construction costs, not the reverse. Speculative If that generalises, the entire manufacturing-cost-curve programme is aimed at a dependent variable. Reducing hard cost would not reduce the price of housing; it would widen the developer's margin, or be bid into the land, and the household would see none of it. Frontier This is one paper, on one national market, and it is badly under-replicated — which is itself the finding. A result this destructive to the field's central assumption should have twenty replications and has approximately none.

Established Two constraints on housing output are private, and no public policy addresses either. Meen and Whitehead identify output as limited not only by land-use planning but by “the behavioural responses and incentives structures facing local authorities, builders and landowners”, and specifically by “the business model of larger speculative builders; the skills base and productivity; and the extent to which demand is there to maintain increased output levels” (Understanding Affordability, ch.10, 2020). Established That last clause is the absorption-rate constraint stated in polite form. Speculative A speculative builder sitting on a land bank maximises profit by releasing units at a rate that does not depress the price it is achieving on the units already released. Under that objective function a technology that builds twice as fast is worth nothing in volume terms, because the builder will not sell twice as fast. Speculative It converts instead into land held for less time — a working-capital gain on the builder's balance sheet, invisible in completions. Frontier The second constraint appears in Pal's study of Toronto's Ward 8, where the most restrictive neighbourhoods showed “the highest growth in house prices, the greatest increase in housing services per unit upon rebuild, and disproportionately low per-square-foot property values in comparison to its sale price” (2022). Speculative If the number of units is capped, the only margin left is quality per unit. Restriction does not merely produce scarcity; it produces luxury, and it predicts that construction savings in restricted markets get spent on better houses rather than cheaper ones.

Frontier Regulation's effect on affordability is not monotonic, and past a threshold it stops operating on price at all. Goytia and Pasquini, surveying land-use regulation across Argentine metropolitan municipalities matched to household census data, report that “municipalities that have incorporated more land planning regulatory measures into their legal and regulatory frameworks also face the cost of larger informal land sectors”, with negative effects on formality from higher residential approval costs and from more authorities involved in approvals, and positive effects on formal tenure where inclusionary policies exist (SSRN, 2024). Established The binding outcome there is not what the dwelling cost but what kind of claim the household has on it. Established The same margin runs through the subsidised sector from the other direction: Meen and Whitehead record rental supply subsidy moving from local authorities to housing associations, then toward private finance and the recycling of past subsidy, while “allocation principles have also changed, moving away from accommodating lower-income working households to emphasising provision for vulnerable households of all types” (ch.11, 2020). Speculative A housing model that treats regulation purely as a price effect is missing the tenure margin on both flanks.

Established On delivery models, the measured record is thinner than the rhetoric by a wide margin. Site build remains the default everywhere. Panelised construction ships flat and assembles slowly; volumetric modular assembles fast and ships badly; the two have never been combined because the trade-off is geometric rather than technical. Established Hedayati and colleagues, reviewing industrialised housing delivery, find that “adoption remains uneven due to fragmentation across regulatory, organisational, and technological systems” (Buildings, 2026), and Razak and colleagues report industrialised building systems “still unable to meet the demands of the stakeholders” (Buildings, 2022). Established The one worked cost comparison in this corpus is Syafriandy's study of the Batang III industrial worker flats, where a value-engineered precast method costing Rp 5,650,987,713 delivered 11.96 per cent savings (Jurnal Indonesia Sosial Teknologi, 2023). Established Read the comparator before quoting it: that is value-engineered precast measured against full precast, not against conventional cast-in-place. Established And when Kouhirostami and Chini went looking for the life-cycle evidence base behind modular's standard benefit list, they reported it missing — frameworks for comparing modular against site-built construction are “lacking” (MOC Summit, 2022). Frontier Mossman and Sarhan add the operational reason offsite underperforms its brochure: failure to synchronise factory and site production causes losses and delays for every party, and their remedy is a buffer between fabricator and site (Construction Economics and Building, 2021). Speculative Variability is not eliminated by moving work indoors. It is relocated to the interface and then paid for in inventory, which is the thing lean manufacturing exists to remove.

Established Mass timber is the counterexample, and it is a real one. Stegner and Fotheringham state the position before the change: structural light-frame wood captures “over 90 percent of the market” for United States low-rise residential, and wood-frame construction was at that time “limited to 5 stories and 85 feet in height” (Forest Products Journal, 2022). Established That limit moved, and their account attributes the change to “research and testing”. Established The stock that followed is not a set of one-offs: Victorero and Bustamante document approximately 300 tall timber buildings worldwide as of 2024, led by Paris with over 35 and London with over 17 (Buildings, 2025). Established The best-documented single case is UBC's Brock Commons in Vancouver — 18 storeys, 53 m, at completion in 2016 the world's tallest hybrid wood building, with 17 storeys of mass-timber superstructure carrying all gravity loads on a concrete podium and two concrete cores (Connolly, Loss, Iqbal and Tannert, Buildings, 2018). Established The schedule split is the number to carry: the concrete cores took fourteen weeks; the mass-timber superstructure took ten. The timber went up faster than the thing holding it up. Frontier The same team found laminated-veneer-lumber cores with C-shaped walls to reduce torsion could meet National Building Code of Canada seismic and wind criteria — meaning the concrete core was a convention, not a physical requirement.

Frontier And now the premise this brief inherits, which is contested at the level of measurement. The argument for industrialised housing rests on a stagnation story: construction productivity flat or falling for decades while manufacturing compounded. Huynh, Le-Hoai and Do state the consensus version plainly — “the global construction sector has witnessed a decline in productivity over the past two decades” (Construction Economics and Building, 2025). Frontier Asked of firm-level microdata rather than sector aggregates, the sign flips. Jaffe, Le and Chappell, using New Zealand's Longitudinal Business Database for 2001–2012, found construction labour productivity growing 1.7 per cent a year and multi-factor productivity 0.5 per cent, against 0.5 per cent and 0.1 per cent for firms in the overall measured sector. Frontier Construction outperformed the rest of the economy on both measures, and dispersion was “no greater in construction than in other industries” (SSRN, 2016). Frontier Kulakov's UCL study of UK construction company accounts removes the usual explanation for the deficit: “there is broadly no correlation between enterprise size and productivity” and “no evidence to support the generally held belief that industry fragmentation affects productivity” (UCL, 2025). Frontier The lever that study does find is contractual rather than technological — a more integrated, in-house “make” model is more productive, with less subcontracting correlating with higher productivity. Speculative A large part of the case for industrialising housing therefore rests on a stagnation premise that may be a deflator artefact. That does not make the case wrong. It makes it unproven at the point where it is usually assumed, and it is why this brief refuses to open with the flat-productivity chart.

Frontier Mass timber's open questions are now specific enough to be experiments rather than positions. The most interesting is structural-system inversion. Chapagain and Crovella's study of the 340+ Dixwell affordable-housing project in New Haven uses a honeycomb system in which floors and roofs bear on mass-timber walls rather than post-and-beam, consuming “nearly twice the amount of timber”, and reports A1–A5 embodied carbon 40 per cent below the median for other multi-family residential buildings on the Carbon Leadership Forum benchmark, with biogenic carbon storage per square foot 60 per cent above the average for reference mass timber types, assessed under ISO 14044, 21930 and 21931 with quantities from Revit and the model in One Click LCA (Sustainability, 2025). Frontier Note the mechanism: the carbon advantage came from using more material, not less. Frontier They introduce Timber Use Intensity as a design variable the prior literature had not evaluated alongside life-cycle impact.

Frontier The all-timber lateral system is the frontier that would make tall timber a delivery innovation rather than a material substitution. Every flagship tall timber building in the record keeps a concrete core. The Brock Commons feasibility work showing LVL cores with C-shaped walls meeting NBCC criteria is the demonstration path; nobody has built one at height. Frontier Naumovski, Gams, Pazlar and Azinović modelled a 12-storey platform-type CLT building replacing a seismically inadequate 1960s residential block, calibrating a shear wall against experiments before expanding to the full structure with nonlinear-spring connections, and found “large lateral displacements and local failures of connections” while still judging the typology usable at moderate seismic risk such as Ljubljana (2CroCEE, 2023). Frontier Connections govern, not panels. Hussein argues separately that rigid-bond assumptions in laminated-composite analysis “greatly underestimate the true performance” of these systems (IntechOpen, 2021) — if that survives scrutiny, current design values are conservative in an unquantified direction.

Established The fire evidence is maturing in public, with numbers, and that is exactly why timber got a code path. Chang, Lin, Chuang and Tsai ran 72 full-scale tests on CLT walls with four joint types at three pressure differentials, under both non-fire and fire exposure: leakage was 0.00 m³/h unexposed, and under fire at a 25 Pa differential averaged 8.17 m³/h with a maximum of 8.27 (Fire, 2025). Frontier They then propose a single-joint evaluation method converting leakage into smoke-layer descent time and thence allowable evacuation time — the shape a performance-based approval argument has to take. Frontier At the landscape scale, Hayajneh and Naser modelled a two-storey CLT building under wind-driven Douglas Fir plantation fire in FDS across fuel load, moisture, separation distance, terrain slope and wind, and report that fuel load and terrain slope drove peak heat release rate, that separation distance substantially reduced local exposure, that the highest heat release rate did not always correspond to the highest building temperature, and that convective heating dominated over radiative heating at the monitored locations (Fire, 2026). Speculative If that holds, wildland-urban-interface codes built on radiant-heat assumptions may be assessing the wrong transfer mode.

Frontier Automating the approval itself is blocked on data quality, not on logic. Kincelova, Boton, Blanchet and Dagenais built a BIM-based automated code-checking method against the National Building Code of Canada for tall timber fire safety, and found “the main challenges for rule checking are the modeling standards and the elements' required levels of detail” — geometrical and non-geometrical requirements were implementable, but sprinkler and fire-damper positioning was not (Buildings, 2020). Speculative That is a useful negative: the barrier to machine-checked compliance is the discipline of the model, which is a procurement requirement, not a research problem.

Frontier On the economics side, four frontiers are live. Platform Design for Manufacture and Assembly is being treated as a procurement construct rather than a project technique, which moves industrialised housing from a factory question to a client question. Frontier Reference class forecasting has reached public housing: Chadee, Martin, Gallage and Rathnayake built a 23-project reference class from 82 housing projects matched on design-build procurement and local contractor involvement, and report a 50th-percentile risk acceptance giving a closer estimate to actual cost than the contracted cost did — described as the first practical application in public housing worldwide (Buildings, 2023). Frontier Natarajan layers machine learning onto the method and demonstrates that overrun distributions are “non-normal and fat-tailed” (Project Management Journal, 2022), which should reshape how housing programmes set contingency. Frontier And Zheng and Wang bring the vertical dimension into hedonic pricing, finding a price premium for housing in immediate neighbourhoods with more open mid-rise buildings and low plants (Environment and Planning B, 2023) — the beginning of a method for pricing three-dimensional urban space, which the neighbouring briefs in this cluster all lack.

Established The workback plan to housing on a manufacturing cost curve has nine links, and the construction links are not the ones that bind. The target state: dwellings whose delivered price to the household falls with cumulative production volume, at a sustained learning rate, in a real market. The impossible thing is not manufacturing a house cheaply — that has been done repeatedly. It is for the saving to arrive at the household.

Established H1 — the five-term decomposition, published on a common boundary for at least five metropolitan markets across three countries over ten years. Not binding, but it is the gate: without it nobody can say whether the programme is aimed at a term that matters. Cheap, and it should be done first. Frontier H2 — a mix-adjusted and quality-adjusted productivity series. If it returns no deficit, everything downstream is a solution in search of a problem. Established H3 — a repeatable dwelling product with characteristic values, a type approval, and a learning rate of at least 10 per cent per doubling across three doublings in hard cost. Not binding: this is the part the industry already knows how to do.

Frontier H4 — an anchor demand commitment, decoupled from the speculative absorption rate. A public, institutional, build-to-rent or co-operative buyer committing multi-year volume at a fixed specification, whose objective function is units rather than achieved price per unit. This binds for hard cost, and it is the spine of the whole subject. Speculative Industrialised housing has not failed because the factory does not work. It has failed because the factory needs a pipeline it cannot secure, and nobody will commit a pipeline to an unproven factory. A site builder in a downturn shrinks to near-zero cost; a factory keeps paying for the building, the line, the working capital and the retained crew while its pipeline evaporates. Speculative Volumetric modular converts a variable-cost business into a fixed-cost business inside a cyclical demand curve — a financial failure mode in a manufacturing costume. Established H5 — type approval with jurisdictional reciprocity binds jointly with H4: volume you cannot serve from one line is not volume.

Frontier H6 — land supply elastic enough that the saving is not capitalised. This is the binding link for the entire programme. In a constrained market the residual claimant on any efficiency gain is the landowner, because developers bid for sites until margins return to normal. Tsai's causality result is the observable signature; Pal's Toronto wedge between per-square-foot value and sale price is another. Speculative H7 — or, failing H6, land value capture that takes the gain first: betterment levy, land value tax, public land assembly, or ground leases on public land. Politically harder, technically simpler. Frontier H8 — absorption-rate independence: completions must rise when build speed rises. Handwave H9 — delivered price per household falls with cumulative volume, audited, over three doublings. Terminal, and nowhere in sight.

Speculative The structure of that chain is the finding. H3 to H5 are construction problems and are solved or solvable. H6 is a land-economics problem, and it decides whether solving H3 to H5 accomplishes anything for a household at all. Any argument that presents industrialised construction as the answer to affordability without addressing H6 is offering a solution to hard cost as though it were a solution to price.

Established This brief waits on results that are mostly not being produced, and none of them is expensive. The first is inherited from Automated Construction Systems and is the reason the dependency is typed rather than decorative: a hedonic, quality-adjusted construction output index, back-cast several decades in at least two countries, that holds regulated performance attributes constant. Frontier Thermal performance, seismic performance, accessibility, fire compartmentation, acoustic separation and air-tightness have all been added to the product by regulation over fifty years. If quality improvement is booked as price inflation by the deflator, measured real output growth is understated and measured productivity falls mechanically. Frontier Jaffe and colleagues' microdata result is precisely what one would expect if the aggregate deflator, and not the industry, were the defective instrument.

Established Second, an output-mix-adjusted productivity series. Construction's output mix has shifted for decades toward repair, maintenance, retrofit, fit-out and complex urban infill — intrinsically low-productivity, high-variance work — and away from greenfield repetition. Measured sector productivity would fall from mix shift alone with no firm becoming worse at anything. Frontier The complication is that Jaffe and colleagues' decomposition separates continuing-firm change from reallocation and from entry and exit, and reports that in building construction and heavy and civil engineering, productivity was enhanced by net entry and reallocation but reduced by a decline at continuing firms. That cuts against the pure mix-shift story. Something real is happening inside firms, and the mix-adjusted series is the instrument that would separate the two effects. It does not exist.

Established Third, the five-term decomposition, which is this brief's critical gap. Likely sources are national housebuilder association cost models, appraisal texts and central-bank housing reports; none was reachable in the research session behind this brief. Frontier Fourth, supply-elasticity estimates on Huang's method outside China, so that the 30 per cent regulatory cost of elasticity and the finding that geography beat regulation can be tested somewhere else. Speculative Fifth, a replication of Tsai. One nonlinear causality test on one national market cannot carry the weight this brief has to put on it, and the field's failure to replicate it in fourteen years is a fact about the field's incentives rather than about the result.

Frontier The cheapest high-value experiment in this entire subject requires no new technology and nobody has run it. Take an existing housebuilder that has adopted a materially faster method — panelised, volumetric, or simply a better site process — and measure whether their annual completions rose, at constant price. Speculative If completions did not rise, the absorption-rate hypothesis is confirmed, the benefit of the industrialisation programme is a working-capital saving that accrues to the builder, and the public case for subsidising it collapses. If completions did rise, the hypothesis is falsified and the technology case strengthens enormously. Established The data required is annual completions and average achieved price, before and after, from firms that already publish both. This is a fortnight of work and it would move the argument more than another decade of factory demonstrations.

Frontier The second experiment is the pass-through test, and it is the one that decides the programme. Deliver an identical industrialised product into a high-elasticity and a low-elasticity market, matched on specification and boundary, and measure what fraction of the hard-cost reduction appears in the sale price versus the land price. Speculative The output is a single pass-through coefficient. If it is near zero in constrained markets and near one in elastic markets, the field's entire affordability claim is jurisdiction-specific and should be stated that way. Frontier The third is H1 itself: a published five-term decomposition with the boundary stated explicitly, which is the discipline missing from every construction cost claim in this literature.

Frontier The fourth experiment reframes a statistical dispute as a governance one. Park reports average cost overrun falling from 38 per cent to 5 per cent across 107 major projects after the introduction of reference class forecasting, with the United Kingdom exceeding its targeted within-budget probability by 12 percentage points and the United States underperforming by 17 (EJTIR, 2021). Themsen, studying a 23.6-billion-kroner Danish public megaproject longitudinally, concludes the method “does not lead to more accurate cost estimates” because estimates are “always a relational network effect” of actors' interested efforts to establish them (Financial Accountability & Management, 2019). Speculative Both cannot be right in general. The reconciling hypothesis is that reference class forecasting works by changing the institution's incentives — making a low estimate embarrassing — rather than by improving the estimate, which would explain why it works in a mandated national regime and fails in a single project whose estimate is still produced by interested parties. Speculative The test is whether the forecast is binding on the forecaster. That is a governance variable, it is observable, and as far as this corpus shows it has never been the design variable in a study.

Established The engineering requirements for industrialised housing divide into three unrelated problems, and the field routinely treats them as one. The first is the product: a repeatable dwelling with published characteristic values, tolerances that survive transport and craneage, and a type approval that is a document rather than a per-project engineering judgement. Frontier The second is the interface, which is where offsite construction actually fails — Mossman and Sarhan's finding that synchronising factory and site requires a buffer means the requirement is inventory capacity and float, not precision. Speculative The third is the catchment. Transport economics cap a volumetric factory's radius at a few hundred kilometres, because beyond that you are shipping air; panelised systems ship well and assemble slowly. Nobody has both, and the trade-off is geometric.

Frontier For mass timber specifically the engineering agenda is unusually well defined. Connections govern seismic response in tall platform CLT, not panels, so the deliverable is a connection family with characteristic values and a ductility class rather than a better panel. Frontier Replacing the concrete core with an LVL or CLT core is the single change that would convert tall timber from a material substitution into a delivery change, and the feasibility case exists under NBCC criteria without a built demonstration at height. Established Fire performance needs the leakage-to-evacuation-time chain carried through to a code-usable form. Frontier And Timber Use Intensity has to be traded against structural efficiency inside one objective function, because the carbon case rewards using more timber while the concrete case rewards using less — a mixed-material building faces two contradictory optimisation rules and no design method in this literature resolves them jointly.

Speculative The requirement nobody costs is the model discipline. Automated code checking works for geometry and fails at sprinkler and damper positioning because the building information model does not carry the level of detail the rule needs. That is a specification a client can write into a contract tomorrow, and it is the cheapest available step toward approvals that scale with volume rather than with reviewer hours.

Established The nearest neighbour is Automated Construction Systems, and the relationship is not the one the two literatures assume. That brief carries the machines — extrusion printing, robotic assembly, offsite fabrication — and terminates on an anchor buyer committing multi-year volume sufficient to load one production system. This brief arrives at the same link from housing economics, through the speculative builder's absorption rate. Speculative Two independent chains, one from manufacturing and one from demand, terminate on who is buying and on what terms. That convergence is the strongest structural result in this cluster.

Frontier Arcologies is the same constraint wearing a third costume. A megastructure has no value until nearly complete and no financing instrument tolerates that, so an arcology requires a single committed owner for the whole object — which is the anchor-buyer problem with the volume compressed into one transaction instead of spread across a decade. Speculative Underground Cities inherits the pricing half of the problem: there is no established price for three-dimensional urban space, which is why Zheng and Wang's vertical hedonic work matters well outside its own subject.

Established Megaproject Governance supplies the estimation machinery this brief uses for public housing programmes — reference classes, optimism bias, and the dispute over whether the standard remedy works. Frontier Future Labour Markets carries the other side of the productivity question: if the binding constraint on housing output is trade skill supply rather than technology, industrialisation is a response to a shortage rather than a cause of productivity, and its uptake will track wages rather than technical maturity. Speculative Smart Cities and Circular Infrastructure Systems are adjacent on the operational and end-of-life margins respectively, and the second is where the permanence assumption behind timber's biogenic carbon credit actually gets tested: a CLT panel counts as storage only for as long as it stays out of a fire or a landfill, and that is a demolition-policy question rather than a materials one.

Frontier The institutional requirement that dominates everything else is a buyer whose objective function is units. A speculative housebuilder cannot be that buyer, for the structural reason given above: its profit is maximised by controlling release rate. The candidates are public housing bodies, institutional build-to-rent, co-operatives, defence and healthcare estates, and student accommodation providers — clients that need a stock rather than a margin. Speculative What such a client must sign is a programme rather than a project: committed units per year against declared line capacity, over enough years to survive a credit cycle. No amount of capital substitutes for it.

Established Second: type approval with reciprocity. A manufactured-building certification accepted across the anchor's whole catchment turns approval from a per-project cost into a per-product cost, which is the only mechanism by which regulatory overhead can fall with volume. Frontier Mass timber's history shows this is achievable — a code limit of five storeys and 85 feet moved on the strength of a research and testing programme. Speculative But note why it worked and what it does not generalise to. You can fire-test a panel. You cannot fire-test a procurement model. Material innovations have a testable failure mode a committee can ballot on; delivery innovations do not. That asymmetry explains the lopsided history of construction innovation, and it predicts that no amount of evidence will get a delivery change into a code the way timber got a material change in.

Speculative Third, the gate that leaves no published trace: insurance. A code provision permits a system; an underwriter prices it. A novel structural system has neither good nor bad loss experience, and ignorance is priced as risk. A search of the academic literature returns regulatory barriers in abundance and never returns the sentence “the underwriter declined”, which is a fact about what gets written down rather than about what binds.

Established Fourth, and this brief owes the reader one honest paragraph about its own evidence base. Elsevier, Wiley, Taylor & Francis and SSRN deposit no abstracts to the Crossref index that the research behind this brief was able to reach. Their papers appear in the record as verified bibliographic entries whose text was never read. The consequence is structural: the readable corpus for this subject skews toward MDPI, Springer Open, IOP, conference proceedings and preprint servers, while the record-only items come on average from more selective venues. Speculative In this literature, readability is inversely correlated with authority. That is not a complaint about open access, which is doing the work here; it is a warning about what a machine-assembled evidence base over-weights, and it is why the numbers in this brief cluster in Buildings, Fire and Sustainability rather than in the journals a specialist would name first.

Frontier The distributional question in housing technology is not who gets the house. It is who gets the saving. If the pass-through coefficient in constrained markets is near zero, then public subsidy for industrialised construction is a transfer to landowners routed through a factory, and the households it is justified by never see it. Speculative That is not an argument against the technology. It is an argument that the technology must be paired with a capture mechanism, and that a programme funded without one should be described accurately to the people paying for it.

Established The tenure margin is where regulation does its cruellest work. Goytia and Pasquini's Argentine result shows heavier land-planning regulation associated with larger informal land sectors, with higher approval costs and more approving authorities pushing against formality. Established The households that fall off that margin do not experience a higher housing bill; they experience a weaker claim on a dwelling — no security of tenure, no collateral, no standing against eviction. Speculative A cost-only analysis of housing regulation cannot see this outcome at all, which means the standard welfare arithmetic of planning reform is measuring the wrong variable for the households at greatest risk.

Frontier The luxury mechanism raises a subtler equity problem. If restriction converts a unit cap into higher quality per unit, then construction savings in restricted markets get spent on larger and better-equipped houses for the households already inside the market, and the excluded households are unaffected. Speculative Every efficiency gain in a constrained market is therefore regressive by default unless something forces it into unit count. Established And the subsidised sector has drifted the same way from the other side: allocation principles moving away from lower-income working households toward vulnerable households of all types is a defensible policy choice that nonetheless leaves a large group with no route through either channel.

Frontier Two safety questions deserve to be stated without either slogan. Timber's fire performance is measured, imperfect and improving — joints that leak nothing unexposed and measurably under fire, wildland exposure that may be convection-dominated rather than radiation-dominated. Speculative The honest position is that a maturing evidence base is a third thing from “safe” and from “dangerous”, and that residents of tall timber housing are entitled to the numbers rather than to either adjective.

Speculative Housing is the largest capital stock any society owns, and it is the one asset class that has never been put on a learning curve. Wright's 1936 observation that unit cost falls predictably with cumulative output reshaped aircraft, then ships, then semiconductors, then photovoltaics. It has never applied to dwellings, and the reason is not that houses are hard to make. It is that the cumulative-unit count for any given design is typically one.

Speculative If that never changes, the consequences compound rather than plateau. Housing costs rise relative to tradeable goods indefinitely, absorbing a growing share of income; spatial sorting intensifies as access to high-productivity cities is rationed by price; and the intergenerational transfer of housing wealth becomes the dominant determinant of life outcomes in rich countries, which is a political condition before it is an economic one. Speculative Every one of those effects is driven by the land term, not the construction term.

Handwave The opposite future is coherent and is rarely stated. Suppose hard cost fell far enough, and durably enough, that the set of buildable sites changed — parcels previously infeasible becoming feasible, marginal geographies opening, the supply curve shifting rather than the split along it changing. In that world the saving is not capitalised because supply expands to meet it, and industrialised housing does reach households. Handwave It requires a cost fall of a magnitude nobody has demonstrated and a policy environment that lets new sites become buildable, which is the same land-policy variable arriving from the other direction. Speculative The honest terminal position is that the technology question and the land question are not separable, and that a civilisation which solves one without the other has solved nothing that a household can live in.

These horizons track institutional and land-policy change, not construction technology, which is the term least likely to be limiting.

• 10 yr: Frontier Mass timber continues to compound: more tall buildings, more jurisdictions with explicit code paths, and probably the first tall building with an all-timber lateral system, since the feasibility case exists and only the demonstration is missing. Speculative The five-term cost decomposition and the mix-adjusted productivity series are both achievable in this window by anyone who decides to fund them; neither requires new science. Frontier At least one anchor-buyer programme of real scale is likely somewhere in the OECD, and its published completions data will be the most informative thing to happen to this subject in a generation. Frontier Volumetric modular firms continue to fail at roughly the historical rate, for financial rather than manufacturing reasons.
• 25 yr: Speculative If a pass-through coefficient has been measured by then, the field bifurcates honestly: industrialised delivery marketed as an affordability instrument in elastic markets and as a schedule-and-quality instrument in constrained ones. Speculative Type approval with cross-jurisdictional reciprocity is plausible within one regulatory union and unlikely across several. Speculative Timber's biogenic carbon credit faces its first serious permanence audit as the earliest tall timber stock approaches end-of-life decisions, and the accounting could move either way. Handwave A demonstrated learning rate on delivered price, rather than hard cost, remains the thing nobody has shown.
• 50 yr: Speculative Whether housing is on a cost curve by then is a question about land institutions rather than about factories: some jurisdiction will have run a serious land value capture regime alongside an industrialised supply programme, and the result will be the single most valuable data point in the subject's history. Speculative The construction side of the problem is likely to look solved and unimportant, in the way that the manufacturing of solar cells now looks solved and unimportant next to the interconnection queue. Handwave A genuinely novel delivery model — something that is not site build, panelised or volumetric — is possible and nothing in the current record predicts what it would be.
• 100 / 250+ yr: Handwave On the timescale where the question becomes civilisational, housing either follows the learning curve that every other manufactured good followed, in which case the twenty-first century's housing crises will read as an institutional failure rather than a technical one, or it does not, in which case shelter remains permanently positional because its scarce input is location and location cannot be manufactured. Handwave The second future is the one the evidence currently favours, and it is worth saying that it is not a dystopia — it is simply a world in which the interesting policy instrument is always the land tax and never the factory.

• Depends on This brief depends on Automated Construction Systems, and the dependency carries a contested premise rather than a settled result. Three things transfer. First, the productivity question: the case for industrialising housing rests on a stagnation story that firm-level microdata contradicts in at least one national dataset, where construction labour productivity grew 1.7 per cent a year against 0.5 per cent economy-wide, and where the fragmentation mechanism usually offered to explain the deficit fails testing. Until a quality-adjusted output index exists, this brief cannot say whether there is a productivity problem to solve, and it declines to assume one. Second, the anchor-buyer result: that brief's workback plan terminates on a client committing multi-year volume sufficient to load one production line, and this brief's terminates on the same link approached from the demand side. Third, the boundary discipline: every cost claim in the automation literature fails for want of a stated boundary, and the same failure is what makes the five-term housing decomposition impossible to assemble from published sources.
• Enables Delivers to the land-economics side of every settlement and urban-form brief on this map. The transferable results are four. The pass-through framing — that a construction saving in a supply-constrained market accrues to the landowner rather than to the occupant — applies unchanged to any proposal that claims to make habitation cheaper by making structure cheaper, which is most of them. The absorption-rate constraint, that a private developer will not build faster than it can sell without moving its own price, applies to any speculative delivery model regardless of technology. The testability asymmetry — a panel can be fire-tested and a procurement model cannot — predicts which innovations will reach codes. And the tenure margin, that regulation past a threshold pushes households out of the formal market rather than merely pricing them up, is the variable a cost-only analysis cannot see.
• Adjacent Adjacent to Arcologies and Underground Cities, which are the same financing and pricing problems rotated: the arcology needs one committed owner for an object with no value until nearly complete, and the subsurface city needs a price for three-dimensional space that no market currently quotes. Adjacent to Megaproject Governance for the estimation machinery, including the unresolved dispute over whether reference class forecasting improves estimates or merely embarrasses the estimator, and the fat-tailed overrun distributions that make mean-based contingency systematically inadequate for a programme of many housing projects. Adjacent to Future Labour Markets, because if the binding constraint is trade skill supply then industrialisation is a response to scarcity rather than a source of productivity. Adjacent to Smart Cities on the operational margin and to Circular Infrastructure Systems on the end-of-life margin, where timber's biogenic carbon credit is finally either honoured or written off.

Handwave “If we could build houses like cars, houses would cost what cars cost.” Cars do not come with land. The analogy silently assumes hard cost is the whole price, and the decomposition that would test it has never been published on a consistent boundary — which is why the workback plan here opens with assembling one.

Speculative “Cheaper construction means cheaper housing.” Only where supply is elastic. In constrained markets the saving is capitalised into land, because the developer bids up for the site until margin returns to normal. Tsai's finding that house prices drive construction costs rather than the reverse is the empirical form of this, and it is the most under-cited result in the popular housing-technology conversation.

Frontier “Modular will solve the housing crisis; it has been proven abroad.” The foreign successes are systems with stable anchor demand and harmonised approvals — institutional conditions, not technologies. Importing the factory without the order book reproduces the failure mode exactly. And the life-cycle evidence base for modular's claimed benefits was reported absent by the researchers who went looking: Kouhirostami and Chini's abstract lists lower cost, lower waste, higher productivity, faster construction and lower environmental impact as background framing, then reports that frameworks for making the comparison are lacking. Established That is a null result cited for the claim it failed to substantiate, and it is worth naming as a pattern: check whether a benefit list is the paper's finding or its opening sentence.

Established “Precast saves 20 to 30 per cent.” The one worked figure in this corpus is 11.96 per cent, and it is value-engineered precast measured against full precast rather than against conventional construction. Always demand the comparator. A saving quoted without its baseline is not a number.

Established “Mass timber is a marketing exercise; you cannot build tall in wood.” Brock Commons: 18 storeys, 53 m, 17 storeys of mass-timber superstructure carrying all gravity loads, timber erected in ten weeks against fourteen for the concrete cores. Roughly 300 tall timber buildings worldwide as of 2024. A United States code limit of five storeys and 85 feet moved on the strength of research and testing. This one is simply wrong.

Frontier “Mass timber is carbon-negative, full stop.” The carbon result is a function of how much timber you use and of assumptions about how long the biogenic carbon stays stored. The 40-per-cent-below-median and 60-per-cent-above-average figures belong to one honeycomb building using roughly twice the timber of a post-and-beam equivalent, assessed under stated ISO standards with a stated tool and against a stated benchmark. Speculative Quote the building, the boundary and the benchmark, or do not quote the number. And note the inversion: timber's carbon case rewards using more material while concrete's rewards using less, so any mixed-material optimisation applying one rule globally will be wrong. Nobody has reconciled the two, and the reconciliation turns on permanence accounting — whether stored carbon counts depends on end-of-life assumptions made decades before they are tested.

Frontier “Cost overruns prove public housing programmes are incompetent.” Overrun distributions are non-normal and fat-tailed, so exceedance is the expected outcome of mean-based contingency rather than evidence of incompetence. Speculative Portfolio diversification does not rescue a national programme either, because the dominant shocks — materials prices, labour shortage, interest rates — are common-mode and hit every project at once. Frontier Whether the standard remedy works is genuinely disputed: 38 per cent to 5 per cent in one 107-project study, no effect at all in a longitudinal study of a 23.6-billion-kroner megaproject, with a third paper observing that optimism bias in institutional settings is “largely understudied”. This brief declares the tie.

Frontier “Deregulate and supply will follow.” Bounded, and bounded by something the argument rarely mentions. The one estimate available here puts the regulatory cost of supply elasticity at about 30 per cent, while finding physical geographic constraint's marginal effect larger still. Speculative And in the Argentine evidence, heavier regulation shifted households into informal tenure rather than simply pricing them up, which means the composition of a deregulation matters as much as its extent: removing approval-cost friction and removing inclusionary requirements are not the same intervention and do not have the same sign.

Frontier “Construction productivity has been flat for seventy years.” This is a statement about a deflated aggregate series, and it is contradicted by firm-level microdata in at least one national dataset where construction outperformed the whole measured economy on both labour and multi-factor productivity. Speculative Anyone quoting the flat line without naming the deflator is quoting a statistic they have not interrogated. The correct position is not that construction is fine; it is that the instrument generating the crisis has never been validated, and a quality-adjusted index would cost less than one demonstration factory.

Handwave And one this brief cannot answer, stated as a hole rather than a hedge. The standard cautionary tale of industrialised housing is the collapse of a heavily capitalised modular firm — Katerra is the name usually invoked, alongside an earlier generation of postwar industrialised-housing programmes. The research behind this brief could not evidence any of it. Its fetch channel failed partway through the session, and the corpus retrieved contains extensive barrier literature but no case study of a large modular firm's collapse at all: Katerra appears in none of the records obtained, and neither do the postwar programmes. Handwave So no figure for capital raised, pipeline, debts or jobs appears anywhere in this brief, and the widely circulated ones are not repeated. What would settle it: bankruptcy filings and court records, dated trade-press reporting, and any post-mortem academic work — none reachable in this pass. Speculative The barrier literature implies the mechanism, and it is the one argued throughout here: a factory converts a variable-cost business into a fixed-cost business inside a cyclical demand curve, and dies in the first downturn it cannot pre-sell through. But implying a mechanism is not evidencing a case, and the difference is the whole of this paragraph.