High Speed Transit Networks

Established High-speed rail is the only new intercity passenger mode to have entered service and scaled anywhere in the past sixty years. It is not an exotic technology. Steel wheel on steel rail at 250–350 km/h is engineering that has been in revenue service since 1964, and no serious party disputes that it works. Established What is disputed is the arithmetic around it: at what corridor density, over what distance band, and under what institutions the capital returns more than it consumes. That question has been argued for four decades on three numbers, and this brief's central finding is that all three arrive at the reader damaged.

Established The first is construction cost per route-kilometre, where the band everybody quotes measured a different mode, in a different decade, on a different scope basis. Established The second is the minimum demand threshold for a socially profitable line, which is cited as a floor that projects must clear when the paper's own authors say the true floor is lower. Established The third is ridership forecast error, the empirical backbone of the sceptical literature, whose canonical citation carries a formal retraction dated 21 February 2012 that essentially nobody downstream has noticed.

Frontier Against that, the most consequential recent result in the field is not a construction result at all. Frontier Spanish open-access competition on already-built track is associated with average fare reductions near 30 per cent and total high-speed traffic growth above 75 per cent — a demand shift produced by a regulatory decision with no new civil works, which makes “existing corridor demand” an output of policy rather than an input to it.

Speculative Beyond high-speed rail sit two would-be successors: superconducting maglev, which is under construction in Japan and is not speculative in any technical sense, and evacuated-tube transport, which is. Handwave This brief assesses their physics and their economics separately, because they fail in entirely different places and conflating the two is how the argument is usually lost.

Established The most-quoted capital-cost band in rail is US$50–100 million per route-kilometre, and it is a band for urban rail. It comes from Flyvbjerg, Bruzelius and van Wee in the European Journal of Transport and Infrastructure Research in 2008, whose deposited abstract states that for European projects, excluding outliers, total capital cost per route-kilometre including stations and rolling stock lies mainly between US$50–100 million at 2002 prices. Established Count the qualifiers: urban rail, not intercity high-speed rail; European, not global; outliers excluded, which removes precisely the projects that generate the argument; stations and rolling stock included, where most published high-speed figures are infrastructure-only; and 2002 prices, roughly a quarter-century stale and almost never re-based. Established Five qualifiers, routinely all dropped, and then the residue is set beside a high-speed number built on a narrower scope. Frontier The comparison inflates the apparent gap in a direction nobody controls for, and it is committed as often by advocates comparing a cheap foreign line against an expensive domestic one as by opponents doing the reverse.

Established Two hard scope facts show why any per-kilometre figure without a terrain class is not evidence. Established Taiwan's high-speed line runs 345 km, of which 252 km is continuous viaduct — roughly 73 per cent of the alignment elevated, a consequence of seismicity and alluvial geology, not of procurement. Established And construction method alone moves schedule by nearly a factor of two on identical scope: a comparative study of a 5 km, 125-span viaduct finds full-span launching completing in 732 days against 1,350 days for span-by-span. Established Scope choices compound this. HS2's Phase One design carries 6.6 km of green tunnel selected after an options assessment — a negotiated mitigation feature, not a topographic necessity — and the choice between slab and ballasted track trades capital cost against maintenance frequency in a way that whole-life comparisons need and headline per-kilometre figures never carry. Frontier There is a published argument, from the engineering side, that substantial documented-but-underused civil-engineering practice would cut high-speed infrastructure cost; it is the standard claim of every cost-reduction paper and it is unquantified, but it is the correct counterweight to treating cost per kilometre as a physical constant.

Established The demand-threshold literature is narrower than its reputation. The founding statement is de Rus and Nombela, Is Investment in High Speed Rail Socially Profitable?, in the Journal of Transport Economics and Policy in 2007. Established Its complete deposited abstract says that the case for a high-speed project is highly dependent on the existing volume of demand in the affected corridor; that the authors use real construction, maintenance and rolling-stock costs of European lines then in operation, potential time savings, standard values of time and expected demand growth to estimate a minimum level of demand from which the investment could be considered socially profitable; and that other benefits, such as providing long-term capacity where overcrowding is expected, could reduce the minimum demand thresholds reported in this paper. Established The abstract contains no number. The famous figure lives in the body, which this brief did not obtain and therefore does not print. Established Two things follow that matter more than the figure. The estimate is explicitly conditional on European costs including rolling stock, on the specific mid-2000s lines, on standard values of time and on an assumed growth path — transplanting it to another cost environment or another decade is extrapolation, not citation. Established And the authors state the threshold is an upper bound on the requirement, not a lower one: capacity value pushes the true floor down. The paper is routinely deployed as a hurdle a project must clear, which is a stronger conclusion than it draws.

Established The forecast-error result that underwrites the sceptical case rests on a withdrawn paper. Crossref's record for Flyvbjerg, Skamris Holm and Buhl, Inaccuracy in Traffic Forecasts, Transport Reviews 26(1), 2006, returns the literal title string RETRACTED: Inaccuracy in Traffic Forecasts and carries an updated-by relation of type retraction dated 21 February 2012, with the notice deposited separately as Banister, Retraction, Transport Reviews 32(2):265. Established This is the origin point of the claim that rail passenger forecasts are systematically overestimated while road traffic forecasts are roughly symmetric around zero error — the asymmetry on which a large fraction of high-speed-rail scepticism, and a good deal of journalism, is built. Established A retraction is not a finding of falsity. This brief did not read the notice, could not establish the reason, and will not speculate about it. Established The sibling brief Megaproject Governance carries the full treatment. Frontier The consequence for this subject is specific and awkward: a load-bearing sceptical result in high-speed rail travels on a formally invalid citation, thousands of downstream users have not noticed, and the substantive claim needs rerouting to sources that are not retracted — Skamris and Flyvbjerg's 1996 Danish sample, whose abstract reports construction costs tending to be underestimated and traffic overestimated, with cost overruns of 50 to 100 per cent common and traffic forecasts off by 20 to 60 per cent, and Næss, Flyvbjerg and Buhl's 2006 companion, which poses the road-versus-rail asymmetry as a question in its own title.

Frontier The most important measured result of the last five years came from deregulation, not construction. An ex-post evaluation of open-access competition on Spanish high-speed rail reports that competition is associated with large and persistent fare reductions averaging 30 per cent and substantial demand expansion, with total high-speed traffic increased by more than 75 per cent. Frontier The preprint version quantifies welfare: consumer gains of €450–520 million annually and external benefits of €120–150 million per year. Frontier Both carry a 2026 deposit date that this brief flags rather than relies on, and the two versions have not been reconciled here. Frontier If the demand response to market structure is of that order, the “existing volume of demand in the affected corridor” that de Rus and Nombela treat as the decision variable is partly endogenous to the regulatory regime imposed on the corridor — and nobody has yet redone the threshold calculation with a post-liberalisation demand function.

Established The Chinese evidence shows large accessibility gains and regional divergence at the same time. A study of 40 cities in the Yangtze River Delta over 2009–2018 finds high-speed rail increased a city's rail accessibility by about 50 per cent, producing closer intercity connections, while the difference in the urban intensity index in 2018 is significantly greater than in 2009 — concentration at the hubs, not convergence. Frontier A panel study over 2008–2019 finds the high-speed network significantly increases innovation in local and neighbouring cities, with the effect diminishing beyond roughly 400 kilometres and large cities benefiting most. Established That 400 km is an innovation-spillover distance and not a mode-share distance; the two are routinely conflated. Frontier On the revenue side, Chinese researchers modelling Beijing–Shanghai argue the existing ticket price is not the optimal one and that there is still room for a price increase — which means the world's most heavily used high-speed line is, on its own operators' research, priced below its revenue-optimal level. Established And on competition, a survey of 1,080 Chinese passengers finds frequency, price, perceived behavioural control and attitudes to be the significant determinants of choice between low-cost carriers and high-speed rail, not journey time alone.

Established The cost record this subject most needs is either unpublished or was unreachable in this pass. The exact study required — a three-stage comparison of preliminary estimate, initial contracted cost and final outturn across a national high-speed programme — exists for Italy, by Bruzzone, Cavallaro and Nocera in 2023, and deposits no abstract; its record is verified and its content is unread here. Established The single best-documented cost-escalation case in the world, HS2, is documented in four Institution of Civil Engineers volumes that are the engineering record and not the cost record; the sequence of published estimates, price base years, scope changes and northern-phase cancellations lives in national audit and infrastructure-authority documents that a provenance gate refused on every attempt. Established The same gate closed off Chinese high-speed network length, cost per kilometre by line and terrain class, China State Railway Group liabilities and debt-to-asset ratio, and the Channel Tunnel's outturn against forecast. Established These are declared gaps and this brief does not fill them from memory. What can be said without them is that the comparative cost table everyone wants does not exist in a scope-consistent form, and that the Chinese figures in circulation are generally converted at market exchange rates and generally exclude land, which is transferred administratively rather than purchased.

Established Faster-than-high-speed ground transport already exists as a construction programme, and it is not a hyperloop. Japan's Chuo Shinkansen is under construction using superconducting maglev, with a continuous engineering publication record running from at least 2010 through 2025, down to the semiconductor power-conversion chain. Established Whatever one concludes about its economics, the technology is not speculative, and it is the only real-world control that exists for any claim about ground transport above high-speed-rail speeds. Frontier The honest framing is that very-high-speed ground transport is a largely solved engineering problem and an unsolved economic one — and that the correct test is whether maglev cost per kilometre exceeds high-speed rail's by less than the value of the time saved at the relevant volume, a calculation for which the Chuo programme is the only dataset in existence.

Frontier The live structural experiment is market design, not civil engineering. Spanish liberalisation is a single country and a single regime change, but it is the only recent result that moves the threshold arithmetic rather than refining it. Frontier The publishable next step is explicit: re-derive the minimum demand threshold using current construction, maintenance and rolling-stock costs, current values of time, and a demand function estimated after open access rather than before it. Speculative Nobody has done this. If the elasticity holds outside Spain, a class of corridors currently below the published threshold is above a corrected one, and a class of already-built lines widely described as failures is mispriced rather than misplaced.

Frontier The second frontier is functional rather than technical: high-speed rail is being planned as commuter infrastructure. Backbone-network planning for the Beijing–Tianjin–Hebei agglomeration uses density-based clustering to select lines and states the objective as realising commuting between important urban nodes. Frontier Daily commuting multiplies annual trips per user by roughly two orders of magnitude over intercity leisure travel, which transforms the demand-threshold calculation in a way the intercity model never contemplated. Speculative The constraint moves to the fare box: commuter-level pricing cannot support high-speed unit costs without either subsidy or land-value capture, and the value-capture route — formalised as spillover tax revenue in a 2023 Asian Development Bank Institute study of Taipei — is untested at network scale and rests on an attribution step nobody has validated.

Frontier The political-economy question has moved from commentary into the literature. A 2026 ADBI paper titled When Speed Outruns Demand: Political Drivers and Economic Outcomes of High-Speed Rail in Spain names the mechanism in its title; its record is verified, its abstract is not deposited, and no authors appear in the metadata. Speculative The hypothesis it encodes — that every high-speed network eventually extends past its own viability frontier for political reasons, and that this is a structural property of the asset class rather than a national failing — is decision-relevant and untested. Speculative The test is assemblable: show whether the marginal line's benefit–cost ratio declines monotonically with network extent across several countries, using national appraisal archives that already exist.

Frontier Forecasting methodology may not have advanced in a quarter of a century, and this is checkable. The same review of state-of-the-art high-speed ridership forecasting practice appears in a 1998 volume and again in a 2020 one; reading them side by side is itself evidence about how much the field has moved. Speculative Against that, revealed-preference data, mobile-phone traces and post-liberalisation outturn series are genuinely new inputs. Speculative The decisive test is whether forecast error on post-2010 openings is measurably smaller than on pre-2000 openings. It is testable with existing data and nobody has published it.

Frontier The emissions case is not yet separated into substitution and induction. The systematic ex-post review of mode substitution and induced demand following high-speed introduction exists precisely because the two had not been distinguished; only substituted trips subtract emissions, while induced trips add them. Frontier A 2025 optimisation study reports a 56.97 per cent CO2 reduction for high-speed freight — against air freight only, under an assumed carbon-tax regime, in a model. Speculative A corridor-level decomposition of substituted against generated trips, paired with an honest embodied-carbon amortisation over a stated grid trajectory, is the binding piece of work and it is not done.

Frontier At the exotic end, the interesting frontier work on evacuated tubes is about branching, not about speed. Pressure-wave behaviour at tube junctions was still experimental science in 2024, in a serious wind-engineering journal. Speculative That matters more than it sounds: a tube that cannot branch is a point-to-point shuttle rather than a network, and network topology is where the capacity argument is actually decided. Frontier Separately, a national standards body published on hyperloop transport services in 2024, which tells you the standardisation process reached document stage regardless of the commercial state of the developers. Speculative Read it for what the standards community believes is actually specifiable, which is a more informative question than what the developers believe is achievable.

Established The binding constraint is not demand. It is that nobody has ever decomposed cost per kilometre into its sources. The workback target is a high-speed line in a low-density country covering full capital and operating cost from fares and directly attributable revenue with no ongoing subsidy — currently considered impossible. Established The ordered chain runs: (1) a defensible, condition-stated demand threshold, re-derived with current costs and a post-liberalisation demand function; (2) a decomposition of per-kilometre cost into terrain-forced, standard-forced, mitigation-forced and procurement-forced components across at least five countries; (3) an achievable cost floor for a given terrain class under best-practice procurement; (4) the demand ceiling attributable to market structure on a given corridor rather than in aggregate; (5) the trips-per-capita gain available if the line is planned as commuter infrastructure; (6) an honest price for non-fare revenue, net of displacement from the origin region; (7) the arithmetic — does (3) fall below (4) plus (5) plus (6) at the country's real corridor densities; (8) five audited years without subsidy, capital service included.

Established Link (2) binds. The argument is always held on demand, and demand is the softer of the two sides: the threshold literature already tells us roughly what volume is needed, conditional on its assumptions. Established What nobody can tell you is what a kilometre has to cost, because the per-kilometre figure has never been separated into the part that is geology and the part that is procurement, standards and negotiated mitigation. Frontier Until it is, “high-speed rail is too expensive for low-density countries” is an observation about institutions dressed as a claim about physics, and the workback plan cannot proceed, because you cannot know whether the floor sits above or below the ceiling. Established Link (2) is also unusually tractable: it needs archival cost data and engineering judgment, not new technology, and it has been started for exactly one country.

Frontier The secondary binding link is (6), non-fare revenue attribution. In every low-density case the fare box alone will not close, so the plan depends on land-value uplift and spillover taxation — and attributing uplift to the line rather than to growth that would have occurred anyway is methodologically unresolved. Speculative Spillover revenue is also substantially a redistribution of location rent; capturing it at the corridor can impoverish the origin region, and the net national figure is the one that matters and the one nobody computes.

Handwave For the evacuated tube the chain is different and the binding links are structural cost per kilometre and vacuum maintenance, not aerodynamics. Speculative A vacuum tube is a pressure vessel hundreds or thousands of kilometres long that must hold precise alignment through thermal cycling and seismic motion; its cost is a structural-engineering question and its leak rate is a manufacturing-tolerance question, and neither has a credible published number.

Established This brief depends on a result produced elsewhere on the site. The retraction status of the canonical traffic-forecast paper, the reference-class-forecasting debate, and the wider question of whether cost overrun is a forecasting failure or a political equilibrium are adjudicated in Megaproject Governance. Established High-speed rail cannot settle its own forecast-error premise, because that premise is a megaproject-wide empirical claim and not a rail-specific one.

Established The comparator problem is inherited from continental network economics. The correct benchmark for any new mode is the improved version of the incumbent, not today's incumbent — Fogel's methodological demand, restated as a design rule in Continental Transportation Systems. Frontier Applied here it produces the strongest anti-high-speed argument available, which is not the cost-per-kilometre argument at all: much of the claimed time saving may be obtainable from conventional-line upgrades at a fraction of the capital, and that case does not depend on any contested number.

Established Three data dependencies are external and currently unmet. National appraisal archives hold the marginal-line benefit–cost series that would test the network-overextension hypothesis, and they are held by the bodies with the least incentive to publish them. Established Infrastructure managers and national audit offices hold the scope-consistent cost histories; this brief could reach none of them, and says so rather than reconstructing them. Speculative And the vacuum-tube question depends on structural cost and leak-rate figures that no developer has published, which is why Intercontinental Rail Systems reaches the same wall from the other direction.

Frontier One dependency runs the other way and is worth naming. This subject is where the scope-corruption failure mode is easiest to demonstrate, because the units are simple, the projects are public, and the qualifiers are printed in the source abstract. Frontier A per-kilometre decomposition done here would transfer directly to urban metro, freight corridors and fixed links, which fail in exactly the same way and have no comparably clean worked example. Speculative That makes high-speed rail the right place to run the experiment even for readers who do not care about trains.

Established Six measurements would settle most of what is contested here, and five of them need archives rather than laboratories.

Established One: the four-way cost decomposition. Take at least five national high-speed programmes and separate per-kilometre outturn into terrain-forced, standard-forced, mitigation-forced and procurement-forced components, on a single scope boundary that states explicitly whether it includes rolling stock, stations, depots, land, electrification and interest during construction, at a single price year. Established The Italian three-stage cost study is the closest existing precedent and covers one country.

Frontier Two: re-derive the demand threshold. Repeat de Rus and Nombela's method with current costs and values of time and a demand function fitted to post-liberalisation Spanish data, and report the result with every conditioning assumption and its price base year attached.

Frontier Three: the forecast-improvement test. Assemble forecast-against-outturn pairs for high-speed openings before 2000 and after 2010 and test whether error has fallen. Established Station-level ex-post work of the kind published on the Hiawatha corridor shows the granularity at which forecast error is actually diagnosable.

Frontier Four: the substitution–induction decomposition on at least three corridors with pre-opening controls, distinguishing trips taken from air and road from trips that would not otherwise have been made. Established This is the only route to a defensible emissions claim, and it must be paired with an embodied-carbon amortisation over a stated grid decarbonisation trajectory rather than a present-day grid mix.

Speculative Five: the marginal-line series. For each network with more than three lines, plot appraised benefit–cost ratio against network extent at the time of approval, and test directly whether the marginal line's appraised return declines monotonically as the network grows. Speculative If it does across several countries, network overextension is a structural property of the asset class and not a national failing, which changes what reform can reasonably be asked to achieve.

Speculative Six, for the exotic end: publish a tube cost and leak-rate figure. Build an instrumented evacuated tube of at least a few kilometres with realistic joint counts, and report structural cost per kilometre, pumping energy, and leak rate over a year of thermal cycling. Handwave Until that exists, every hyperloop economic case is a capsule cost multiplied by an assumption.

Established The engineering variance that dominates cost is structural, and it is legible in advance. Viaduct and tunnel fraction is the first-order term: Taiwan's 252 km of continuous viaduct on a 345 km line is a seismic and geotechnical requirement, and comparing that alignment's unit cost to a line built at grade across a plain is not a comparison. Established Construction method is second-order but large — 732 against 1,350 days on identical viaduct scope, depending on whether spans are launched whole or built in place.

Established Track form sets the capital-versus-maintenance split. Slab track offers greater dimensional stability; ballasted track requires more frequent assessment and maintenance. Whole-life cost turns on that choice and per-kilometre headlines never carry it.

Established Tunnel cross-section is governed by aerodynamics, not by clearance. HS2's tunnel design is set by pressure-transient and micro-pressure-wave behaviour against line speed — the same constraint, in a different regime, that governs evacuated-tube design. Frontier In a partly evacuated tube the problem sharpens into local supersonic zones and shock formation at transonic capsule speed: the choked-flow regime, which forces either an onboard compressor, a much larger tube cross-section, or a harder vacuum. Each costs money or reliability. Speculative The same paper reports a surface-temperature range spanning more than an order of magnitude, presented with a range separator that is a known source of transcription corruption; this brief declines to print it and recommends that anyone who needs it read the load cases separately from the source.

Established Vehicle capacity is the hard engineering fact at the exotic end. Published hyperloop capsule designs seat 28, 40 or 50 passengers, with total annual cost modelled at €5.5–6.2 million depending on configuration. Frontier Seats per vehicle are held low by pressure-vessel mass, so tube capacity depends almost entirely on headway, and headway in a sealed tube is bounded by emergency braking and evacuation. Established A published safety framework for hyperloop draws its requirements from NASA human-spaceflight practice, which tells you the certification regime is not a rail regime.

Established Superconducting maglev is the nearest adjacent technology and the only one under construction. The Chuo Shinkansen programme has a continuous engineering literature and mature vehicle development, and functions as the empirical control for every claim about ground transport faster than high-speed rail. Frontier Its cost and schedule history, once available, answers by direct measurement what hyperloop advocates answer by assertion.

Established Aviation is the binding competitor and it does not compete only on time. Low-cost carrier competition against high-speed rail in China turns on frequency, price, perceived behavioural control and attitudes across a 1,080-passenger survey. Established Mode share is therefore contestable by scheduling and pricing decisions, not only by journey-time thresholds — which is the same lesson the Spanish open-access result teaches from the rail side.

Frontier The conventional-line upgrade is the adjacent option that most often wins and is least often assessed. Higher line speed on an existing alignment consumes no new corridor and a fraction of the capital, at the cost of disruption-years and a capacity ceiling. Whether it dominates depends on whether the upgrade path is capacity-feasible on the existing alignment, and that is a corridor-specific engineering question with a knowable answer.

Speculative Freight in a vacuum tube is the adjacent form of the exotic proposal that avoids its hardest constraints. A numerical analysis of a high-speed freight vehicle in a vacuum tube exists in preprint; with no human occupants there is no evacuation constraint and the safety case is not a spaceflight case. Handwave If evacuated-tube transport has a viable form, this is the likelier door, and the passenger framing has probably been an error throughout.

Frontier Value capture is an adjacent financial technology rather than a physical one. Spillover tax revenue has been formalised as a cost–benefit component; it is the only candidate for closing the fare-box gap in low-density cases and its attribution step is unvalidated.

Established Land acquisition regime is frequently the largest single line item and it differs categorically between jurisdictions. Where land is purchased at market price through contested compulsory-purchase procedures, cost and schedule both move; where it is transferred administratively, the resource cost does not vanish, it simply leaves the published figure. Established This is the single largest reason that international per-kilometre league tables are not comparable, and it is an institutional fact, not an engineering one.

Frontier Open access requires three institutional conditions and Spain is the only network known to satisfy all three: spare capacity on the built network, a workable slot-allocation regime, and rolling-stock market access. Frontier Producing the short list of other networks that meet them is answerable work and would identify where a near-zero-capital intervention delivers hundreds of millions of euros a year in consumer surplus.

Frontier Reference class forecasting is the standard prescribed cure for the optimism the sceptical literature documents, and it is not settled. One study of 107 major projects reports average cost overrun falling from 38 per cent to 5 per cent after the United Kingdom's 2003 mandate, with the UK exceeding its targeted probability of on-budget completion by 12 per cent and the US underperforming by 17 per cent. Frontier A longitudinal case study of a 23.6-billion-kroner Danish megaproject concludes the opposite, stating that reference class forecasting does not lead to more accurate cost estimates in a project that failed to meet its objectives despite applying it. Established Both are peer-reviewed and both stand. Any appraisal regime that treats the instrument as solved is out of date.

Established The institutional record is also where this brief's evidence ran out. Cost histories, network statistics and operator accounts live in national audit offices, infrastructure authorities and state railway annual reports, and a provenance gate refused those domains on every attempt in this pass. Established The absence of a discovered scope-corrupted headline number in this cluster is a statement about access, not about the literature. Frontier The highest-probability location for one, on the evidence available, is the international cost-per-kilometre league table, where mixed scope, mixed price years and mixed exchange-rate conventions make a factor-of-two artefact close to unavoidable.

Speculative One institutional fix would do more than any of the above and nobody owns it. A single published scope standard for infrastructure cost reporting — stating whether a figure includes rolling stock, stations, depots, land, electrification and interest during construction, and at what price year — would make comparison possible without any new measurement at all. Frontier No body has the standing to impose it, which is why the field has argued about the same numbers for twenty years.

Frontier The regional-equity argument for high-speed rail is, on current evidence, backwards at least as often as it is forwards. The Yangtze Delta study finds accessibility rising by about 50 per cent while inter-city intensity differences widened significantly over the same decade; the innovation study finds benefits accruing disproportionately to large cities and dying beyond roughly 400 kilometres. Frontier A line sold to a peripheral region as a development instrument may move activity toward the hub it connects to. That is a distributive question argued in technical vocabulary.

Established Who pays and who rides are different populations. High-speed rail is financed from general revenue or state borrowing and used disproportionately by higher-income intercity travellers; where it is priced below its revenue-optimal level, as Chinese researchers argue Beijing–Shanghai is, the transfer is larger than the accounts suggest and its incidence is rarely stated.

Established Mitigation scope is an ethical negotiation with a price tag. HS2's 6.6 km of green tunnel is a choice made after an options assessment to reduce local impact, not a geological requirement. Frontier It is a defensible commitment, it is also a cost driver, and treating it as evidence that a country cannot build confuses a preference with a constraint. Frontier The reverse error is as common: reading another country's lower per-kilometre figure as competence when part of the difference is that it declined to buy the mitigation.

Established Hindsight is the other trap, and there is a documented case. A 1994 cost–benefit analysis of high-speed rail in California reported competitive comprehensive rates of return and potential for self-financing — a citable, favourable appraisal of a programme whose later history became the field's standard cautionary tale. Frontier Two honest uses: as a case study in appraisal drift, asking how much of the change is scope, alignment, land and inflation and how much is that the analysis was wrong; and as a guard against hindsight, because the 1994 conclusion was not unreasonable on 1994 information. Frontier The dishonest use is to treat it as self-evidently foolish, which does to 1994 exactly what 1994 did to the future.

Established And there is a research-integrity dimension specific to this subject. A great deal of policy scepticism about rail investment rests on a paper that has carried a retraction since 2012. Established The ethical obligation is symmetrical: not to treat the retraction as proof the underlying finding is wrong, and not to keep citing a withdrawn work because the conclusion is congenial.

Established High-speed networks are among the few artefacts that fix a country's settlement geometry for a century. An alignment chosen once determines which cities are two hours apart and which are five, and that ordering outlives every government that argued about it. Frontier The measured effects run in both directions — accessibility gains that are real, and concentration effects that run against the stated regional purpose — and the net sign is a property of the specific network, not of the technology.

Frontier The nation-building objective is real and is usually smuggled. When a study of the world's most-extended network is titled around political drivers and economic outcomes, the field is conceding that the objective function was never purely economic. Speculative That is not a debunking. Legitimacy, cohesion and the visible demonstration of state capacity are genuine objectives; the failure is that they are pursued through an appraisal apparatus that prices only travel time, so the political return is bought without ever being costed.

Speculative The honest civilizational statement is that this class of asset is bought partly for what it says. A country that can build 300 km/h rail on time has demonstrated something about itself that no report can demonstrate. Handwave A defensible practice would price that return explicitly and compare it against other ways of buying the same thing — which nobody does, and everybody implicitly does anyway.

Frontier There is also a long-horizon argument that has nothing to do with speed. A dedicated high-speed alignment removes the fastest passenger services from the conventional network, and what it leaves behind is conventional capacity for freight and regional traffic on the old line. Speculative Where that released capacity is actually used, the investment is partly a freight investment wearing a passenger badge, and its appraisal is being run against the wrong benefit stream. Speculative Where it is not used, the claim is a decoration. Which of the two holds is measurable line by line and is rarely measured.

What these horizons track is the arithmetic, not the trains. The engineering is settled; what changes over time is whether anyone can say what a kilometre must cost and what demand a corridor will actually produce.

• 10 yr: Frontier Open-access competition spreads to two or three further European networks, and the demand-threshold literature is re-derived with post-liberalisation elasticities rather than pre-liberalisation ones. Frontier The four-way per-kilometre cost decomposition is published for two or three countries, most plausibly starting from the Italian three-stage study. Established The Chuo Shinkansen produces the first scope-consistent cost and schedule record for very-high-speed ground transport, and the maglev-versus-high-speed comparison stops being hypothetical. Speculative No passenger hyperloop carries a fare-paying passenger, and the surviving vacuum-tube work is freight-framed or standards-framed.
• 25 yr: Speculative A scope-consistent international cost table exists and the “country X builds for $Y per km” comparison stops being made by serious parties. Speculative At least one network is planned from the outset as commuter infrastructure with land-value capture in the financing structure, giving the first real test of item (6) in the workback chain. Speculative The substitution-versus-induction decomposition has been done on enough corridors that high-speed rail's carbon claim is either established or quietly retired. Handwave Evacuated-tube transport, if it survives at all, survives as freight in a short and specific corridor.
• 50 yr: Speculative The marginal-line question is settled empirically across several countries, and network extent is planned against a measured viability frontier rather than a political one. Speculative Whether any low-density country has a line covering full capital and operating cost from fares and attributable revenue is decided by then, and the answer is determined mostly by what the cost decomposition found in the 2030s. Handwave A maglev network of more than one line exists somewhere, or the mode is permanently understood as a single-corridor technology.
• 100 / 250+ yr: Handwave If evacuated-tube transport ever operates at network scale, it will have been solved as a structural-tolerance and vacuum-maintenance problem, not as an aerodynamics problem, and the solution will look like a manufacturing achievement rather than a transport one. Handwave The alternative, and on current evidence the more likely one, is that steel wheel on steel rail remains the fastest ground transport anybody actually buys.

• Depends on This brief waits on a result that belongs to megaproject governance rather than to rail. The empirical claim that rail demand forecasts are systematically biased — and that road forecasts are not — is the premise of most high-speed-rail scepticism, and its canonical citation carries a formal retraction dated 21 February 2012, with the notice deposited under a separate DOI. Whether that finding survives on unretracted sources, whether the mode asymmetry holds at all once the road-planner comparison is posed as a question rather than a conclusion, and whether reference class forecasting actually corrects the bias, are all adjudicated in Megaproject Governance. Until that adjudication lands, the sceptical case here cannot be scored, because its strongest single number is travelling on a withdrawn paper and its prescribed remedy has one strong positive result and one strong negative one, both peer-reviewed. Note what this dependency is not: it is not a claim that the sceptical case is wrong. It is a claim that nobody currently knows how well supported it is.
• Enables What this brief enables is narrower than the subject's profile suggests, and it is a measurement rather than a technology. If the four-way decomposition of cost per route-kilometre is ever published — terrain-forced, standard-forced, mitigation-forced, procurement-forced, on one scope boundary at one price year — then every downstream argument about whether a country can afford fast rail becomes answerable, and so do the parallel arguments about urban metro, freight corridors and fixed links, which fail in exactly the same way. The decomposition is the artefact this subject owes the rest of the field. A second enabling result is smaller and closer: the short list of networks that satisfy open access's three institutional preconditions — spare capacity, a workable slot-allocation regime, and rolling-stock market access — identifies where a near-zero-capital regulatory change delivers a demand response measured in tens of per cent.
• Adjacent The adjacent structure is unusual in that the nearest technologies are not the nearest constraints. Superconducting maglev is technically mature and under construction, and it is the empirical control for very-high-speed claims. Evacuated-tube transport is adjacent in ambition and remote in evidence: its physics is largely settled and mostly unfavourable in the details, while its economics is unmeasured, and the two must be argued separately or not at all. The conventional-line upgrade is the adjacent option that most often wins on the numbers and is least often given a fair hearing. And open-access market design sits adjacent to all of them as the only intervention in this subject that has moved demand by tens of per cent at near-zero capital cost. The tech tree here is therefore shallower than the subject's reputation implies: almost nothing on it is waiting for a physical discovery, and almost everything on it is waiting for a measurement, an archive, or a regulatory decision.

Established “Country X builds high-speed rail for $Y million per kilometre, so we should too.” Almost always invalid. The band most often reached for — US$50–100 million per route-kilometre — is urban rail, European projects, outliers excluded, including stations and rolling stock, at 2002 prices. Established Set against an infrastructure-only high-speed figure at an unstated price year it produces a gap that is an artefact of scope. Established Add terrain: 252 of Taiwan's 345 km are viaduct. Add method: 732 against 1,350 days for identical viaduct scope. Add mitigation: 6.6 km of green tunnel on HS2, chosen rather than required. Add land regime, which differs categorically between jurisdictions. Established A per-kilometre number without terrain class, scope definition and price year is not evidence.

Established “de Rus and Nombela proved high-speed rail needs N million passengers.” The deposited abstract contains no number. It states that the estimate is conditional on European construction, maintenance and rolling-stock costs, on the lines then in operation, on standard values of time and on an assumed growth path — and it ends by saying that other benefits, such as long-term capacity where overcrowding is expected, could reduce the minimum demand thresholds reported in this paper. Established The figure is deployed by sceptics as a floor a project must clear. The authors say the true floor is lower. Established This brief does not print the figure, because it did not read the body, and anyone quoting it has either read the body or is copying a secondary source.

Established “Rail forecasts run high and road forecasts are honest.” The canonical citation for that asymmetry is retracted. Crossref returns the title RETRACTED: Inaccuracy in Traffic Forecasts with a retraction relation dated 21 February 2012 and a notice at a separate DOI. Established Retraction is not falsification, the reason could not be established in this pass, and this brief will not speculate about it. Frontier What can be said is that the sceptical literature's strongest empirical support has been formally withdrawn for over a decade without the downstream literature noticing, and that the substantive claim should now be sourced to unretracted work — the 1996 Danish sample, or the 2006 companion that poses the road-versus-rail question in its title.

Established “Ridership on line Z is N million.” Check whether N is an outturn or a forecast. Forecasts survive in advocacy material for years after outturn data exist, because the forecast sits in the business case and the outturn sits in an operator report nobody reads. Established Never quote a ridership number without saying whether it is forecast or realised, and for which year.

Frontier “Chinese high-speed rail is cheap.” Published Chinese per-kilometre figures are generally converted at market exchange rates rather than purchasing-power parity, and land is transferred administratively rather than purchased. Established Neither adjustment is normally applied when the figure is set beside a European cost. Established This brief could not reach the primary Chinese cost and debt sources and does not print a China-versus-Europe ratio; anyone who does should state the conversion basis and the land treatment alongside it.

Established “High-speed rail converges regional economies.” The Yangtze Delta evidence points the other way: accessibility up around 50 per cent, and urban intensity differences significantly wider in 2018 than in 2009. Frontier Innovation spillovers accrue most to large cities and diminish beyond roughly 400 kilometres. Established That 400 km is a spillover distance, not a mode-share distance, and conflating them is a common and consequential error.

Frontier “Fare revenue shows high-speed rail cannot cover its costs.” On the world's busiest line, Chinese researchers argue the existing Beijing–Shanghai fare is not optimal and that there is room for a price increase. Frontier Observed farebox shortfall partly reflects a pricing decision rather than a demand ceiling, and the two are not distinguishable from the accounts alone.

Established “The emissions saving is established.” The systematic ex-post review of mode substitution and induced demand exists because substituted and generated trips had not been separated. Established The 56.97 per cent reduction figure in circulation from a 2025 modelling study is against air freight only, under an assumed carbon-tax policy, in an optimisation model. Quoted without those three qualifiers it becomes a claim the paper does not make.

Established “Maglev is vapourware.” It is not. The Chuo Shinkansen is under construction with superconducting maglev and a continuous engineering publication record from 2010 to 2025. Frontier Whatever one concludes about its economics, the technology is not speculative — which matters most for calibrating vacuum-tube claims, because it establishes what a state-backed, technically mature, faster-than-high-speed programme actually costs in years and money.

Established “Hyperloop will carry high-speed-rail volumes.” Published capsule designs seat 28, 40 or 50 passengers. Established Capacity in a sealed tube is headway-limited, and the safety framework being drawn for hyperloop comes from NASA human-spaceflight practice — which is a statement about certification cost as much as about risk. Frontier Quote a marketed throughput only alongside the seat count and the assumed headway, and say what headway the safety case actually supports.

Speculative “Vacuum-tube physics is unsettled.” The physics is largely settled and mostly unfavourable in the details: choked flow and shock formation in a partly evacuated tube have been computed, and branch-junction pressure waves are now being measured experimentally. Handwave What is unsettled is the cost per kilometre of a vacuum-tight, precisely aligned, thermally cycling tube and its leak rate at real joint counts — the engineering economics, not the aerodynamics. Argue the two separately.

Frontier “A number in a peer-reviewed paper is safe to quote.” One aerothermal hyperloop paper in this pack's own bibliography reports a capsule surface-temperature range spanning more than an order of magnitude, using a range separator that is a well-known source of transcription corruption between conventions. Established The upper end is not physically absurd for a transonic capsule in a partly evacuated tube, which is exactly what makes it dangerous to repeat. Established This brief declines to print it. The same failure mode — a defensible number stripped of its scope — produced the cost-per-kilometre error at the top of this section, and it happens inside the frontier literature as readily as in journalism.