Tokyo moves 40 million trips a day with 99.8% on-time rate; Houston commutes 25 miles by car. The structural choices that decide which cities work.
The Trips That Cities Move
Why a few networks dwarf everything else on the planet
Greater Tokyo's rail system moves roughly forty million passenger trips on a normal weekday — more than the entire daily population of California — with average delays under sixty seconds and an on-time rate of ✓ Established 99.8% across the Tokyo Metro network [2]. Seoul, Singapore and London each operate networks whose daily ridership exceeds the population of most countries.
The phrase "public transport" hides a planetary spectrum. At one end sit a handful of cities that have built networks capable of moving the equivalent of an entire metropolitan population every working day, repeatedly and on time. At the other end sit cities whose entire transit ridership for a year would barely register as a peak hour in Tokyo. The gap between these two worlds is not technological — every city has access to the same trains, buses and software. It is structural: a function of how streets, zoning, parking, subsidy and time have been allocated over the preceding half-century.
Tokyo's network is the global benchmark for two reasons. The first is scale: Tokyo Metro alone carried 6.84 million passenger trips a day in fiscal year 2024 [2], and once Toei Subway, JR East commuter lines and the dozen private railways are added, the metropolitan rail system handles roughly forty million daily trips. The second is reliability: a ✓ Established mean delay under sixty seconds per train [2] that turns the network into infrastructure rather than transportation — something a commuter plans against in the way a Western worker plans against electricity rather than weather.
Seoul Metro is its closest peer, with approximately eight million daily riders across twenty-three lines and more than six hundred stations under a single integrated T-money fare system [4]. Singapore's MRT moves 3.49 million daily journeys on a network that the Land Transport Authority intends to double to about 460 kilometres of track by 2040, with the explicit policy goal that nine in ten peak-hour journeys be completable in under forty-five minutes door-to-door [3]. London Underground records 1.2 billion passengers a year — roughly 3.8 million daily taps on the rail network and 5 million bus boardings on top [1].
Against these benchmarks, the American picture is structurally different in kind rather than degree. The Federal Transit Administration's 2024 National Transit Summary records national transit ridership at 74.6% of 2019 levels in the spring of 2024, recovering unevenly across systems [13]. Los Angeles reached 81% of its pre-pandemic baseline; Houston 89%. But these recovery percentages obscure the larger fact: at full pre-pandemic operations, US transit ridership concentrated in a handful of legacy systems (New York, Chicago, Washington, Boston, San Francisco) and was already a minority share of commuting nearly everywhere else. ✓ Established Cars account for 93% of all commuter distance travelled in Los Angeles [13] — the second-largest metropolitan area in the United States.
Dubai's pattern is even more pronounced. More than 83% of trips in the emirate are taken by private car [18], and registered vehicles rose from 1.9 million in 2021 to 2.27 million in 2023 despite the Dubai Metro carrying record passenger numbers. The metro is real; it just sits inside a built environment that punishes anyone trying to use it for ordinary daily life. The same paradox describes Phoenix, where Valley Metro Rail carries roughly 9,000 weekday riders on its south Phoenix extension while the surrounding region remains designed almost entirely around the private vehicle.
What separates Tokyo from Phoenix, then, is not the existence of trains. It is the surrounding decisions — about where buildings may be placed, how much land must be allocated to parking, how the costs of motoring are distributed, and how the streets themselves are geometrically organised — that determine whether a transit network can grow into infrastructure or remain a costly amenity used by a fraction of the population. The remaining sections of this report examine those structural choices in turn.
Daily ridership figures conceal as much as they reveal. Tokyo's forty million trips include suburban rail, subway, monorail and tram across multiple operators; London's figure separates Tube from National Rail commuter services that share the same passenger pool. The methodology this report uses — total passenger trips per day across all formal public-transport modes — follows the convention adopted by the McKinsey 24-cities study and the Oliver Wyman Urban Mobility Readiness Index. Comparisons across cities are indicative, not exact.
The Geometry That Decides Everything
Why streets, not preferences, determine whether a city moves
A single lane of urban road can move roughly 2,000 people per hour in cars, 9,000 by bus, and over 20,000 by rail [3]. ◈ Strong Evidence The constraint on every city is geometry: a road network cannot expand fast enough to outpace the demand it itself induces, a relationship now formalised as the fundamental law of road congestion [11].
Public transport is sometimes framed as a question of preference — whether residents like trains, whether they will tolerate a bus, whether suburbanites can be persuaded out of their cars. This framing collapses on inspection. The actual constraint is mathematical. Cars are the least space-efficient mode of urban transport ever designed: at typical urban speeds, a sedan occupies roughly thirty square metres of street for a single occupant, while delivering perhaps one or two passengers per trip. A bus on the same lane moves an order of magnitude more people; a rail line on a dedicated right-of-way moves another order of magnitude on top. No amount of cultural preference can rewrite these ratios.
The consequence is what Gilles Duranton and Matthew Turner formalised in 2011 as the fundamental law of road congestion: across decades of US metropolitan data, ◈ Strong Evidence a one-percent increase in interstate or major-road capacity generates approximately a one-percent increase in vehicle-kilometres travelled in the same area [11]. The relationship is so close to unitary that the authors describe it as a law rather than a tendency. A 2019 replication using two decades of European city data reproduced the finding almost exactly [11]. The implication is brutal: no amount of road building can outpace the demand it generates. Within a few years of opening, an expanded highway carries more cars than the old one, at roughly the same congested speed, having absorbed billions in capital expenditure to deliver no time savings.
Duranton and Turner's 2011 NBER paper established a unit elasticity between road capacity and vehicle-kilometres travelled in US metropolitan areas; a 2019 replication on European cities found virtually the same coefficient [11]. The only interventions that durably reduce metropolitan traffic are pricing (tolls and congestion charges) and modal substitution (transit, walking, cycling). Cities that nonetheless continue to widen roads do so for political rather than empirical reasons.
The corollary is symmetric. If new capacity induces new traffic, removed capacity should reduce traffic — a hypothesis tested by Seoul when it demolished the Cheonggyecheon elevated freeway in 2003-2005. The expected gridlock did not materialise. ✓ Established Bus ridership rose 15.1% and subway ridership 3.3% between 2003 and 2008, and the corridor now draws 64,000 visitors a day [10]. Land values within fifty metres of the restored stream rose 30-50%, roughly double the rate of the rest of Seoul [10]. The freeway, far from being load-bearing, had been carrying traffic that simply chose other modes or other times once the option vanished.
The same pattern appears wherever it has been carefully studied. Paris closed both banks of the Seine to through traffic in 2016. New York removed cars from 14th Street in Manhattan in 2019. London restricted Oxford Street access to buses, taxis and cyclists. In every case, predicted traffic apocalypses failed to occur; ridership on parallel transit rose; and the reclaimed street became one of the most economically productive pieces of public land in the city. These are not idiosyncratic outcomes. They are what the geometry predicts.
Geometry also explains why dense networks scale and sparse ones do not. A subway station is most useful when it sits within walking distance of another subway station, which is most useful when it sits within walking distance of another. The value of a transit line is approximately proportional to the square of the network's density, in the same way that telephone networks were once described by Metcalfe's Law. This is why Tokyo's Yamanote loop, Seoul's interlocking lines, the Paris Métro's mesh and London's radial-and-circle pattern produce extraordinary ridership while a single light-rail line through a sun-belt suburb does not. The line is not the system; the system is the system.
The geometric argument is the one most often missing from American transit debates, which tend to oscillate between cultural fatalism ("Americans love their cars") and individual moralism ("commuters should choose differently"). Neither framing engages with the underlying physical constraint: at the densities of population and employment found in any successful global city, there is no possible street configuration that allows a majority of trips to be made by private car. The choice is not whether to provide alternatives. It is whether to do so deliberately or to allow congestion, parking scarcity and accident rates to ration motoring instead.
A bus stuck in traffic is not a failure of buses; it is a failure of bus priority. A nearly empty light-rail line is not a failure of rail; it is a failure of the land-use pattern around it. The mode rarely fails. The geometry around the mode often does.
— Jarrett Walker, transit planner, Human Transit (2024 edition)The Hidden Subsidy in Every American Street
The cost of "free" parking and the politics of who pays
Every mode of urban transport is subsidised — the only meaningful question is which one and how visibly. The UCLA economist Donald Shoup estimated the annual American subsidy to "free" parking at approximately $127 billion [12], a figure that dwarfs the federal transit operating budget by an order of magnitude. ✓ Established A single structured parking spot costs $25,000 to $65,000 to build [12], costs that are embedded in rents, retail prices and tax bills rather than appearing on the driver's bill.
The most persistent myth in American transport policy is that private motoring is unsubsidised and that public transit relies on public money. The truth is the opposite. Every American who buys a sandwich at a downtown deli, rents an apartment in a building with parking, or pays property tax in a city with off-street parking minimums is subsidising automobile storage. The subsidy is so embedded in the price structure of ordinary life that it is invisible — exactly the reason it has been so politically durable.
Donald Shoup's The High Cost of Free Parking, drawing on more than a decade of empirical research, estimated the annual American subsidy to free parking at roughly $127 billion in 2002 dollars [12]. The number rises in real terms as land values rise. Beyond Shoup's headline figure, the tax-free treatment of employer-provided commuter parking — a US tax-code provision dating to the 1960s — costs federal and state governments approximately $7.3 billion in foregone revenue every year [12]. Neither figure includes the road space allocated to free on-street parking, which represents one of the most valuable land assets in any American downtown.
The mechanism by which this subsidy is enforced is municipal zoning. Most American cities require that new construction provide a fixed minimum number of off-street parking spaces per dwelling unit, per square foot of retail, per restaurant seat. These parking minimums originated in 1950s suburban planning and have proved extraordinarily difficult to dislodge, even though the urban planning consensus against them is now overwhelming. ✓ Established Each structured parking spot costs $25,000 to $65,000 to build, and the cost is shifted by developers to tenants in the form of higher rents and to consumers in the form of higher retail prices [12].
Buffalo, NY became the first major US city to abolish off-street parking minimums in 2017. Dozens of cities have since followed, including Minneapolis (2021), Austin (2023), Anchorage and Raleigh. ✓ Established In Minneapolis, residential rents declined 4% between 2019 and 2024 against a national rent increase of 22% over the same period [14]. Washington State in 2024 passed the strongest statewide rollback yet, capping most residential parking at 0.5 spaces per unit and exempting many projects entirely [14].
Transit, by contrast, is subsidised on the public ledger — and visibly so. The Federal Transit Administration's 2024 National Transit Summary records an average US farebox recovery ratio of 16% [13], meaning fares cover only a sixth of operating costs and the remaining 84% comes from federal, state and local subsidy. The 16% figure is widely cited as evidence of transit's failure. It is more accurately read as a measure of the choice not to invest. European systems with comparable density routinely recover 35-55% of costs at the farebox not because Europeans love trains more, but because they have built denser networks serving denser cities with higher service frequencies, making transit the rational default rather than the marginal alternative [13].
The asymmetry is what generates the political deadlock. Drivers experience their subsidies as natural conditions — free roads, free parking, low fuel taxes — and experience transit subsidies as a transfer to other people. Transit riders experience the same fact in mirror image. The single most important rhetorical move available to a transport reformer is to make the parking subsidy visible: to unbundle it from rents, to price kerb space at market rates, to publish what an employer is paying per employee for "free" parking. ✓ Established The annual US parking subsidy is roughly five times the federal transit programme [12].
Highway spending and road-maintenance budgets are large, routinised and invisible to drivers as a "subsidy." Transit operating subsidies are smaller, line-itemed and visible at every farebox. The political contest is therefore conducted on asymmetric terms: one mode's subsidy is treated as infrastructure, the other's as a handout. Unbundling the costs of motoring — kerb pricing, parking cash-out, congestion charging, scrapping the tax break on commuter parking — is the most powerful policy lever available, and the most politically difficult.
Vienna offers a counter-example of what visibility can achieve at the other end of the price scale. The Austrian capital introduced its annual public-transport pass in 2012 at €365 — one euro per day — and saw holders rise from 360,000 to 850,000 by 2020 [15]. The city now reports 34% of all trips on public transport against 25% by car [15]. The €1-a-day price was a visible political statement as much as a fare; in September 2025 the city announced it would raise the price to €467 in 2026 due to budget pressure, a politically painful reversal that demonstrates how much the original price had become symbolic of the city's transport contract with its residents [15].
The Land-Use Loop
Why zoning determines transit before the first train runs
◈ Strong Evidence Land-use decisions made decades before a single track is laid determine whether a transit system can succeed [14]. The American single-family-zoned suburb is, by construction, a built environment in which transit cannot recover its operating cost — not because residents would refuse it, but because the density does not generate the trips.
Transit policy treats land use as a constraint; in practice, the relationship is reversed. The geographic distribution of jobs, housing, schools and retail decides what kinds of trips a population will make, and only certain distributions can be served at acceptable cost by collective transport. A city of detached single-family houses on quarter-acre lots, with employment dispersed across office parks reachable only by limited-access highways, cannot support a transit network capable of moving a meaningful share of trips. The arithmetic does not work: at four households per acre, a bus route serving one mile of street has perhaps 1,200 households within walking distance of any stop, a fraction of whom commute at compatible hours to compatible destinations. The numerator of ridership is structurally bounded.
The opposite arithmetic produces Tokyo. At twelve thousand people per square kilometre across the metropolitan core, with employment concentrated in dense sub-centres organised around rail stations, a single high-frequency line carries the trips of an American mid-sized city. The Yamanote Loop in central Tokyo is so dense that the surrounding land-use planning effectively required the railway: removing the line would create traffic that no road network could absorb. This is the virtuous loop that produces extraordinary ridership. Density justifies the line; the line justifies further density.
This is also why the chicken-and-egg framing of American transit debates ("we cannot build ridership without service; we cannot fund service without ridership") is, in a precise sense, correct. The land-use pattern is the constraint that breaks the loop. ✓ Established Singapore's Land Transport Master Plan 2040 makes the integration explicit: the policy goal of a "45-minute city, 20-minute towns" is defined as nine-in-ten peak-hour journeys completable in under 45 minutes door-to-door [3], which only works because residential and commercial densities are coordinated with line and station planning by the same agency. The transit network is not built around the land use; the two are designed together.
An apartment building required to provide one parking space per unit cannot achieve the floor-area ratios needed to support frequent transit. The parking podium consumes the ground floor, the construction cost rises by $25,000-$65,000 per unit [12], and the resulting building must be set back from the street to accommodate vehicle access. The combination produces an urban form indistinguishable from suburban arterial development even when the underlying transit infrastructure exists. ✓ Established Parking reform — not zoning bonuses or density caps — is the single most powerful land-use lever available to a city seeking to make transit viable [14].
The American post-war suburb is the largest, longest and most heavily subsidised urban experiment in human history. Beginning with the 1956 Federal-Aid Highway Act, which authorised $25 billion (about $250 billion in 2024 dollars) for the construction of the interstate system, federal policy systematically subsidised dispersal [12]. The interstate system was ostensibly inter-city, but in metropolitan areas it became the spine of a land-use pattern that decoupled residence from employment, employment from retail, and all of them from anything reachable on foot. The current US transit problem is, in significant part, the inheritance of this seventy-year subsidy programme.
Reversing it is difficult because the resulting land use is itself path-dependent. Once a region has committed to single-family zoning over decades, the housing stock, the school catchments, the commercial footprints and the political coalitions all align around the existing pattern. The recent wave of US zoning reforms — California's SB 9 (2021) ending exclusive single-family zoning state-wide, Minneapolis's 2018 abolition of the same, Washington's 2023 statewide middle-housing legalisation — represents a serious counter-current, but the resulting changes will take decades to ripple through the built environment. ◈ Strong Evidence Land-use change is generational where transit operating decisions are annual [14].
The simpler observation is that any city building transit without simultaneously reforming its land use is building a system that will struggle to attract riders. Phoenix's Valley Metro Rail, Houston's METRORail, and dozens of other US sun-belt lines were built without commensurate zoning reform along their corridors. The result, predictably, is per-mile ridership that lags dense systems by an order of magnitude — not because the rail itself is poorly designed, but because the buildings around the stations are not. ⚖ Contested The question of how much rail investment alone can shift modal share in low-density geographies remains genuinely contested among transport scholars [13].
A transit-viable city requires three reforms together: minimum-parking-mandate elimination, single-family-only zoning elimination, and dedicated street space for buses and rail. Each reform on its own is incremental; the three combined are transformative. The cities that have made the largest demonstrable gains — Minneapolis, Seattle's denser neighbourhoods, the new Singapore TOD nodes — have pursued all three, while cities that have pursued only one (such as the wave of light-rail openings without zoning reform) have generally seen disappointing ridership.
What Happened When Cities Removed the Cars
The empirical record from Pontevedra, Oslo, Ghent, Barcelona and Seoul
✓ Established Pontevedra has recorded zero pedestrian traffic fatalities in the past decade, down from 69 in 1998, after pedestrianising 1.3 million m² of city centre [5]. Oslo recorded zero pedestrian and cyclist deaths in 2019 [8]. Ghent's 2017 circulation plan raised cycling modal share from 22% to 34% in just two years [9].
The argument against reducing car access in cities has historically been that the loss of traffic capacity would produce displaced congestion, harm retail, and prove politically catastrophic. The empirical record from a growing number of European cities is unambiguous: none of these effects materialise at scale. The displaced traffic largely disappears (a symmetric version of the induced-demand argument), retail recovers and frequently improves, and the political reversals are rare and limited.
Pontevedra, a Galician city of roughly 85,000 people, has become the longest-running natural experiment in car-free city design. Beginning in 1999, the newly elected mayor Miguel Anxo Fernández Lores pedestrianised the historic core within his first month in office. Over the subsequent quarter-century, the pedestrianised area has grown to 1.3 million square metres [5], on-street parking has been almost entirely removed from the centre, and motorised vehicle journeys have fallen from approximately 52,000 in 1997 to 17,000 today. The results, documented in the municipality's own published programme and confirmed by independent reviews, are striking: ✓ Established CO₂ emissions associated with cars down 66%, fuel consumption down 67%, pedestrian traffic deaths from 69 in 1998 to zero in the past decade [5].
Oslo's approach has been more incremental but no less successful. From 2017 the city removed approximately 700 parking spaces in the central square mile and replaced them with sixty kilometres of bike lanes and pocket parks; the city-centre ban on private through traffic took effect in early 2019. ✓ Established The city recorded zero pedestrian and cyclist fatalities in 2019 [8], the first year for which this could be documented since automobile-era record-keeping began. The risk per trip of fatal or serious injury fell 47% for cyclists, 41% for pedestrians and 32% for drivers between 2014 and 2018 — outcomes consistent with Vision Zero hypotheses but unprecedented in their magnitude [8].
Ghent's 2017 circulation plan provides perhaps the most data-rich record of a single intervention. The city divided its centre into six sectors that cannot be traversed by car without leaving the centre via the outer ring road, and reserved core streets for pedestrians during business hours. The results in the first two years: ✓ Established cycling modal share up from 22% to 34%, car share down from 55% to 27%, traffic accidents in the centre down 25% [9]. The city's 2030 cycling target was achieved by 2019, eleven years ahead of schedule [9].
Barcelona's superblocks programme remains the most ambitious modelled exercise in deliberate urban redesign. The original 2019 study by ISGlobal estimated that if all 503 proposed superblocks were implemented, ◈ Strong Evidence 667 premature deaths could be prevented annually, life expectancy would rise by approximately 198 days per resident, and the city would save €1.7 billion per year in avoided morbidity and mortality [6]. The measured results from the first three implemented superblocks, published in 2024-2025, confirm the direction: NO₂ within the Sant Antoni superblock fell 25%, PM10 fell 17%, and residents reported gains in well-being, tranquility, sleep quality and social interaction [7].
The traffic does not disappear; it evaporates. When you remove a road, the trips it carried mostly do not appear elsewhere. Some are made on foot, some by bicycle, some by transit, some shifted to different times, some not made at all. Cities have spent a century assuming that traffic is a fluid that conserves volume. It is not.
— Sally Cairns, Carmen Hass-Klau & Phil Goodwin, summarising decades of road-removal evidence (TfL/UCL, updated 2023)Seoul's Cheonggyecheon project is the largest urban highway removal recorded to date. The 5.8-kilometre elevated expressway carried roughly 168,000 vehicles per day in 2003 and was widely considered impossible to remove without metropolitan collapse. The demolition was completed in 2005. The expected traffic chaos did not occur. ✓ Established Bus ridership rose 15.1% and subway 3.3% over the next five years, the restored stream now draws 64,000 daily visitors, and land values within 50 metres of the corridor rose 30-50%, double the rate for the rest of Seoul [10]. The project paid back its construction cost in increased land-tax revenue within a decade.
The American Pattern
Los Angeles, Houston, Phoenix, Dubai — the cost of the car-only choice
✓ Established In Los Angeles, cars carry 93% of commuter distance [13]. In Dubai, 83% of trips are by private car [18]. In Houston, the average commuter travels 25 miles round-trip. These are not natural conditions but cumulative outcomes of seventy years of land-use, road-funding and zoning choices that the cities themselves can reverse if they elect to.
The American sun-belt city is the global archetype of car-dependent urban form. Built largely after the 1956 Federal-Aid Highway Act, expanded under decades of low-cost petrol, and shaped by single-family zoning that excluded apartments from most residential land, these cities arrived at the present century with built environments that effectively preclude meaningful transit ridership. The largest among them — Los Angeles, Houston, Phoenix — have spent the past two decades attempting to retrofit transit onto land-use patterns that resist it. Dubai, despite a metro and a deliberate Vision 2030, is travelling the same trajectory at higher speed.
Los Angeles is the case study most often cited. The city operates the second-largest transit system in the United States by passenger trips, has invested tens of billions in light-rail and subway expansion since the 1990s, and has hosted dozens of internal planning exercises aimed at transit-oriented development. Yet ✓ Established cars still account for 93% of commuter distance travelled in Los Angeles [13], ridership reached only 81% of pre-pandemic levels in 2024, and the corridors served by new rail remain dominated by surface parking and single-family zoning. The lesson is not that Los Angeles failed to invest in rail. It is that rail without parking reform, density reform and street-space reform produces minor share-of-trip gains rather than the network effects observed in Tokyo or Seoul.
Houston offers a sharper version of the same pattern. The city is, in conventional terms, the global capital of automobile geometry: the Katy Freeway (I-10) was widened in 2008 to as many as twenty-six lanes in places, the widest stretch of urban motorway in the world. The expansion delivered the predicted induced-demand result. Within five years of completion, peak-period travel times on the corridor had returned to or exceeded the pre-expansion baseline. The city's transit system, METRO, carries roughly 250,000 weekday boardings — comparable per-capita to a city one-fifth Houston's size — and the average Houstonian commutes 25 miles round-trip by car. These figures are not failures of effort; they are the predictable outcome of a built environment that allocates the overwhelming majority of street space to private vehicles.
Cities That Move
~40 million daily rail trips; 99.8% on-time; mean delay under 60 seconds; integrated bus, rail and private-railway network [2].
~8 million daily metro riders across 23 lines and 650+ stations; integrated T-money fare; Cheonggyecheon highway removal [4].
3.49 million daily MRT journeys; rail to double by 2040; "45-minute city" planning goal; LTA controls land use [3].
1.2 billion annual Tube passengers; 3.6 billion total TfL journeys; integrated bus, rail and cycling; congestion charging since 2003 [1].
34% transit modal share; €1-a-day annual pass since 2012; 850,000 pass holders by 2020; consistent multi-decade investment [15].
Cities That Idle
93% of commuter distance by car; transit at 81% of pre-pandemic baseline; light-rail without commensurate land-use reform [13].
26-lane Katy Freeway widening in 2008 returned to pre-expansion congestion within five years; 25-mile average round commute.
Valley Metro Rail ~135,800 weekday riders system-wide; per-mile ridership a fraction of dense systems; sprawl-driven employment dispersal.
83% of trips by private car; registered vehicles 1.9M (2021) → 2.27M (2023); metro carries record passengers within hostile street geometry [18].
Continued freeway expansion into the 2020s; metro-area transit shares below 5%; per-capita VMT among the highest in the OECD.
Phoenix is the city most often held up as evidence that even sun-belt geography can support new transit. Valley Metro Rail carries roughly 135,800 weekday boardings across its 38.5-mile network, with special events such as the 2024 NCAA Final Four producing peaks of 65,300 single-day light-rail riders. These numbers are real and rising. They are also two orders of magnitude smaller than what comparable mileage of rail moves in Tokyo or Seoul. The constraint is not the rail itself but the built environment around the stations — surface parking, single-family neighbourhoods, employment dispersed across business parks reached only by car. Whether sun-belt rail can generate dense-system ridership without prior land-use reform remains ⚖ Contested a genuinely contested question among transport scholars [13].
Dubai offers an instructive non-American example. The emirate has invested heavily in a driverless metro since 2009 and operates one of the most advanced transit systems in the Gulf. Yet ✓ Established 83% of all trips in Dubai remain private-car [18] and registered vehicles grew from 1.9 million in 2021 to 2.27 million in 2023 [18]. The Dubai-Sharjah commute is among the most congested in the Middle East. The metro is excellent; the surrounding street and land-use environment is hostile to it, and the result is exactly what the geometry predicts. The Dubai case demonstrates that even very large transit investments cannot, on their own, overcome land-use patterns that have committed to the private car.
The deeper observation is that these patterns are not natural. They were chosen, deliberately or implicitly, through decades of road-funding decisions, zoning ordinances, parking mandates and federal subsidies. The same choices can be unchosen. Minneapolis ended single-family-only zoning in 2018 and removed parking minimums shortly thereafter, with measurable rent effects within five years [14]. Washington State in 2024 passed the strongest statewide parking-mandate rollback in the country [14]. The US Department of Transportation has begun, through the Reconnecting Communities programme, to allocate $3 billion to begin undoing urban-highway construction that displaced over a million Americans in the mid-twentieth century [16]. Path dependence is real but not absolute.
A sun-belt metropolitan region that adds light rail without simultaneously reforming zoning, abolishing parking minimums and reallocating street space to buses and bikes will spend billions for ridership outcomes that look like failure. The line itself is not the failure. The decision to build the line without changing the conditions around it is. The Atlanta, Dallas, Phoenix and Houston experiences over the past two decades each illustrate the trap; the few sun-belt corridors that have generated meaningful ridership — Phoenix's Tempe-Mesa segment, the Houston Main Street Line — are precisely those where local zoning has been reformed to permit dense development.
Who Gets Transit, Who Gets Highways
The equity dimension that runs underneath every mode-share statistic
Mid-twentieth-century US highway construction displaced over one million people, disproportionately concentrated in Black and Latino neighbourhoods that had already faced disinvestment [16]. The current $3 billion Reconnecting Communities programme is a partial federal acknowledgement that the highway map of every major American city encodes a specific distribution of who gets transit and who is required to drive [16].
Every transport map is also a map of priorities. The decision to route an urban interstate through one neighbourhood rather than another, to fund a commuter rail line to one suburb rather than another, to maintain a bus route or to cut it — each of these decisions distributes mobility unevenly across a population. The American experience of mid-twentieth-century highway construction provides the clearest historical example: the routes chosen for I-95 in Miami, I-81 in Syracuse, I-787 in Albany, I-94 in St Paul-Minneapolis, I-5 in Seattle, and dozens of others were systematically driven through low-income and minority neighbourhoods that lacked the political capital to resist. ✓ Established Over a million Americans were displaced by these projects [16], and the resulting highways imposed lasting costs in air pollution, noise, accident risk and physical severance on the communities that remained.
The Reconnecting Communities and Neighbourhoods programme, launched in 2022 and expanded to $3 billion in 2024, is the first federal attempt to systematically address this inheritance [16]. The initial round funded projects including the demolition of the I-81 viaduct in Syracuse (replaced by a community grid), the removal of urban segments in Toledo and Rochester, and dozens of smaller bridges, caps and street reconnections. The programme is a start. It is also, relative to the scale of the original damage and the cost of undoing it, modest. The Interstate Highway System cost approximately $500 billion in 2024 dollars to build over forty years [12]; the cost of selectively undoing its urban segments will be a substantial fraction of that figure.
The equity dimension is not only historical. Contemporary transit-investment decisions continue to allocate access unevenly. A city's decision to build a commuter rail line into wealthier suburbs while underfunding bus service in lower-income neighbourhoods produces an immediate distributional consequence: the commuter rail is faster, more comfortable and more heavily subsidised per passenger, while the bus is slower, less reliable and serves a population with less political voice. This pattern repeats across virtually every American metropolitan region with comparable transit assets.
The Harvard Equality of Opportunity Project's 2015 study of geographic variation in upward mobility found commute time to be the single strongest factor correlating with the chances of a child raised in poverty escaping it as an adult, stronger than crime rates, school quality or two-parent family share. The implication for transit is that the network's reach into low-income neighbourhoods is not only a service question but a mobility-of-opportunity question. The Federal Transit Administration's 2024 equity guidance reflects this finding by tying federal funding shares to demonstrated benefits for environmental-justice communities [13].
The opposite pattern — that car dependence is regressive — is no less established. Transportation is the second-largest household expense in the United States after housing, and for low-income households it is often the first. A household required to maintain a car to participate in the labour market faces a fixed cost of $8,000-$12,000 per year for ownership, insurance, fuel and maintenance — a non-trivial fraction of poverty-line income. The savings from car independence are correspondingly large for households that can substitute transit, walking or cycling for ownership. ✓ Established The US Bureau of Labour Statistics consumer-expenditure survey consistently shows transportation as 15-20% of low-income household spending against 9-13% for upper-income households [13].
The European cities that have most aggressively reformed their transport systems have generally combined access expansion with affordability commitments. Vienna's €1-a-day annual pass [15], Luxembourg's elimination of transit fares country-wide in 2020, Germany's €49 monthly Deutschland-Ticket and France's Pass Rail — each represents a deliberate choice to price transit as a near-universal entitlement rather than a marginal commuter cost. The political appeal is that these schemes simultaneously address modal share, household-budget pressure and the visibility-asymmetry problem identified earlier in this report.
The greatest civil-rights issue in the cities of the twenty-first century is access. Without it, every other right — to housing, to education, to employment, to family life — is conditional on owning and operating a private vehicle.
— Transportation Equity Caucus statement, US Congress (Transit Equity Day 2024)The equity dimension also runs in less obvious directions. Air pollution from cars and trucks falls disproportionately on neighbourhoods adjacent to major roads, which are themselves disproportionately low-income and minority. Pedestrian fatality rates per resident are highest in low-income neighbourhoods, in part because the same neighbourhoods receive less street-design attention and bus stops are sited along high-speed arterials. The benefits of ✓ Established Barcelona's superblocks — measured NO₂ reduction of 25% within the perimeter [7] — accrue disproportionately to lower-income residents who do not have the option of escaping to cleaner-air suburbs at weekends.
The Choices That Decide a City
What the evidence tells us, and what the recoverable interventions are
Public transport is not a technology problem. The trains, buses, software and fare systems used by the world's best networks are commercially available to any city that elects to buy them. The difference between a city that moves and a city that idles is structural: a sequence of choices about street geometry, land use, subsidy and street-space allocation that can be made deliberately or allowed to drift. The evidence from the past three decades is consistent enough to permit confident policy guidance.
The empirical record reviewed in this report converges on a small number of structural observations. Cars are geometrically incapable of moving the trips of a dense metropolitan region; this is a mathematical fact unaffected by cultural preference. Adding road capacity generates almost exactly proportional new traffic and produces no durable congestion relief. Removing road capacity, in cities that have tried it, produces less displacement than predicted and often produces measurable health, safety and economic gains. Parking minimums make transit-supportive density structurally impossible; their removal correlates with measurable rent reductions. Cheap or universal transit passes can shift modal share if they accompany — not substitute for — service and street-space reform.
The cities that have most successfully built transit networks have, without exception, combined large-scale rail and bus investment with parking-mandate reform, street-space reallocation, and land-use coordination between the transit operator and the planning authority. The cities that have built rail without these reforms — most US sun-belt experiments — have produced disappointing ridership and now face the politically difficult question of whether to make the missing reforms after the fact, or to allow their light-rail systems to remain expensive amenities used by a fraction of the population.
| Reform Lever | Severity / Leverage | Assessment |
|---|---|---|
| Parking minimum elimination | The single most powerful land-use lever available. Minneapolis recorded a 4% rent decline against +22% national average after elimination [14]. Removes the structural barrier to transit-supportive density at near-zero fiscal cost to the municipality. | |
| Dedicated bus and rail right-of-way | A bus stuck in mixed traffic delivers a fraction of the capacity of the same bus on a dedicated lane. Bogotá's TransMilenio, Curitiba's RIT, Mexico City's Metrobús and London's bus lanes demonstrate that paint and concrete are often more cost-effective than rail tunnels. | |
| Single-family-only zoning reform | Long-lead-time intervention; effects unfold over decades. Minneapolis (2018), California (2021), Washington State (2023) provide live tests. Necessary but not sufficient — must be paired with parking reform and transit investment to convert into ridership. | |
| Congestion charging and kerb pricing | London (2003), Stockholm (2007), Singapore (since 1975) and New York (2025) demonstrate immediate measurable traffic reduction. The political difficulty is high in cities that have never priced their road space; once introduced, reversal is rare. | |
| Universal-pass affordability schemes | Vienna €1-a-day, Germany €49 Deutschland-Ticket, Luxembourg fare-free demonstrate visible benefits but contested modal-share effects [15]. Most effective when paired with service expansion rather than substituting for it. |
The honest synthesis is that public-transport reform is unusually amenable to evidence-based policy. Unlike many areas of urban governance where causation is contested and outcomes are slow to materialise, transport interventions produce measurable results on short time horizons — months for street-space changes, two to five years for parking reform, five to ten years for major rail openings. The Pontevedra, Oslo, Ghent, Barcelona and Seoul cases provide a near-complete recipe for what works. The remaining obstacle is political rather than technical: the willingness of a city's leadership to take a measurable but politically visible cost (removed parking, reallocated streets, higher near-term fares for some users) in exchange for measurable but diffuse benefits (cleaner air, fewer deaths, lower household transport costs, denser commercial cores).
For US cities specifically, the report's analysis suggests that the most important reform is the one with the lowest fiscal cost and the highest measurable benefit: the elimination of off-street parking minimums in zoning ordinances. This single change, adopted now by dozens of cities since Buffalo's 2017 first move [14], removes the structural barrier to transit-supportive development without requiring new operating subsidy. It is the lever that compounds: every reformed lot becomes a denser building, which generates more nearby trips, which justifies more frequent transit service, which justifies further density. The other reforms — congestion charging, street-space reallocation, fare reform, rail investment — are each individually powerful but depend, in the long run, on the land-use foundation that only parking and zoning reform can lay.
The cost of failure to make these choices is not abstract. It is paid daily, in hours of commuter time lost to congestion, in the household-budget burden of mandatory car ownership, in pedestrian and cyclist deaths, in air pollution mortality, in the foreclosed economic and social activity of cities organised around storage of private vehicles. ◈ Strong Evidence Barcelona's modelled superblock programme alone estimated €1.7 billion in annual avoided morbidity costs [6], and that figure represents one intervention in one city. Aggregated across the world's car-dependent metropolitan regions, the avoidable cost of present arrangements runs into the trillions.
What every working transit system has in common is a coordinated choice — sustained across political cycles — to allocate street space, land use, and subsidy in favour of collective rather than private modes. What every car-dependent city has in common is the inverse: a sequence of decisions, often individually defensible, that cumulatively foreclose the collective alternative. The choice is not between "natural" outcomes and "engineered" ones. Both are engineered. The question is whose interests the engineering serves and whether the city in question has the political capacity to redirect it. The cities that have done so look, after a generation, qualitatively different from those that have not.
Tokyo's forty million daily trips, Seoul's eight million, Singapore's three and a half million, London's three and a half billion annually, Vienna's 34% modal share, Pontevedra's pedestrianised million and a third square metres, Oslo's zero pedestrian deaths, Ghent's doubled cycling share, Barcelona's measurable health gains, Seoul's reclaimed stream — these are not historical accidents or cultural artefacts. They are the cumulative outputs of choices that any city, faced with the same evidence, can elect to make. The reverse is also true: Los Angeles's 93% car share, Dubai's 83%, Houston's 25-mile commute, Phoenix's struggling rail — these are not natural conditions but cumulative outputs of different choices made over the same time horizon. The evidence is sufficient. The decision is political.