This is probably the most widely believed myth about public transport, and therefore the most dangerous. It’s an old story that, like current transport policy, originates in the first major American freeway study:
The conditions of land use and density….are the major determinants of the travel market. If demand is constrained by these factors, it is unlikely that changes in supply will have any great impact on the number of users.
—Chicago Area Transportation Study 1956
There was no alternative to freeways in Chicago, the road planners said, because the city was too spread out and low-density. The road lobby and its supporters have been using the ‘spread-out city’ as an excuse for freeway building ever since. The story has been repeated so often in Melbourne that many urban planners, commentators, and even some environmentalists believe it.
Gus Braidotti is rightly impressed with the quality, cost and frequency of public transit in Tokyo. But Tokyo has about nine times more residents per square kilometre than Melbourne. This is one of several fundamental reasons why Tokyo has such a high quality transit system and why Melbourne, with its highly dispersed and fragmented population, struggles to sustain the system we have. Until there is a wider community appreciation here of these fundamentals of how a metropolis functions, it seems Melburnians will continue to build a very low density city and as a corollary get a low quality, high cost, urban transit system.
—David Mayes, Australian Institute of Urban Studies, The Age, 20 February 2006
Let’s be frank about public transport – the outer suburbs of Melbourne are poorly served, and, with low population densities, are unlikely to be well served at any reasonable price.
—Sinclair Davidson, RMIT School of Economics, The Age, 11 July 2006
Sprawl means people have to travel more for almost everything…. and they have to travel by car, since low densities make public transport unviable…. Green architecture, therefore, is not made of mud-brick or adobe. Nor is it Federation bungalows surrounded by trees and birds and gardens – much as this might look nature-friendly. Green architecture is made of concrete and glass and high-speed lifts…. The key to the green city, the essence of it, is density.
—Elizabeth Farrelly, Blubberland: the Dangers of Happiness, 2007
Melbourne is one of the world’s most spread-out cities, small islands of high-density housing amid a vast sea of low-density, mostly single-storey homes set in gardens. It is inevitable that a city like that will rely on cars for its transport needs.
—Tim Colebatch, The Age, 9 December 2008
No public transport system can efficiently cope with low density. All the good systems in the world belong to dense cities; and none of the sparse cities has a good system…. The only way we can ethically justify this wasteful ecological damage is to invoke heritage values and deplore the living conditions in other parts of the world…. But some aspects of heritage deserve to be disowned. Most of inner Melbourne is unfit for the purpose of a town of 3.6 million people. Our heritage is largely dysfunctional.
—Robert Nelson, The Age, 15 November 2010
Melbourne is a spread-out, low-density city in which most traffic is driving from one suburb to another. We need freeways to improve inter-suburban access, just as we need a Metro to improve access to the city and inner suburbs.
—Tim Colebatch, The Age, 8 May 2013
In the 2000s the density myth became the centrepiece of the Bracks/Brumby Government’s Melbourne 2030 planning strategy. Apparently, in order to encourage public transport, vast tracts of inner Melbourne would have to be rebuilt at higher densities. As The Age put it:
The deal implicit in urban consolidation is that people forgo private space, backyards and cars for a more compact lifestyle…. [instead] Melburnians are opting for ever bigger, more energy-consuming homes. They need to spare a thought for the environment in which their children will be brought up…. The compact city vision is also under pressure from knee-jerk resident groups and councils in established suburbs…. Unreasonable opposition to higher-density housing in existing streets only adds to the pressure for car-dependent fringe estates.
—The Age (editorial), 5 January 2007
Note that this was a decade before the housing supply crisis of the 2020s brought new weight behind the need to densify the city. Back then, the implication was that we had to give open slather to developers to build high-rise towers throughout the suburbs because, we were told, this is the only way to achieve higher rates of public transport use.
The problem, of course (apart from the fact that public transport use doubled of its own accord between 2005 and 2010, with very few high-rise towers to be seen), is that although many Melburnians are open to the idea of forgoing at least some car use, the idea of forgoing private space and backyards is much less popular.
Pretexts and Fabrications
Many of the contemporary supporters of higher densities are perfectly well-meaning, and speak from genuine concern with the sustainability of Melbourne’s urban form. But not everyone’s motives are so benign. Over the decades many Australian transport planners and economists, and their allies in the media, have wanted to convince us that Melbourne is the most decentralised, low-density city in the world because they have an implacable ideological hostility to public transport (especially rail) and a love affair with roads (especially motorways). No-one has summed up this attitude better than the neo-liberal Institute of Public Affairs, who to their credit have held to this position consistently for over three decades:
In spite of public transport benefiting from massive subsidies, the coverage of its ability to carry people to their destinations quickly is highly restricted…. It can only operate effectively in urban conditions and only really effectively in urban areas with high densities and concentrated origin and destination points. A rule of thumb is that, to be commercially viable, rail-based systems require [400 people per hectare] and express bus systems [250]. Melbourne has an average density of [15]….
For the main part….cities should adapt to the car and the truck. Road systems are far and away more important than fixed track systems, and buses can make good use of them…. It is therefore vital that the road system be upgraded to keep pace with the demand for car transportation.
—Alan Moran, “The Public Transport Myth”, Institute of Public Affairs, October 2006
Though the increased popularity of public transport – and growing public discontent with its inadequacy – has made open political support for this idea less common since the 1990s, one politican who has been happy to regurgitate it without shame is former Prime Minister Tony Abbott.
Sydney (along with other Australian cities) is different too. Sydney has four million people, mostly living in detached houses, in an area larger than London with its ten million plus. Short of attaining London-style population densities, it’s fanciful to imagine that Sydney could have a London-style public transport system…. Mostly, there just aren’t enough people wanting to go from a particular place to a particular destination at a particular time to justify any vehicle larger than a car, and cars need roads.
—Tony Abbott, Battlelines, 2009
One cannot always believe the statistics that appear when, again and again over decades, density is used to discredit public transport. In 1992 the Federal Bureau of Transport and Communications Economics calculated that Melbourne’s population density was only 5.03 persons per hectare, less than 20% of Los Angeles’ 27.1! They did this by dividing the population of the entire Melbourne Statistical Division by its gross area, and the population of the City of LA into the area of residential land. The latter calculation is a reasonable one, as we see below, but the former is nonsense: around 75% of the 1992 MSD wasn’t in the urban area at all.
Melbourne 2030, likewise, claimed that Melbourne’s density is only half that of Montreal and Toronto by using inconsistent definitions of the respective urban areas. And as recently as 2015, the managing director of planning at Transport for London committed the same error, exaggerating the difference in density between Melbourne and London:
Currently Melbourne has some of the lowest population density in the world: half the population of London in an area five times bigger, according to Mr de Cani.
—“Car-free housing encouraged in London, says city’s transport planning boss”, The Age, 29 November 2015
For the record, the area of Greater London (bounded by the famous ‘Green Belt’) is 1,572 square kilometres. The Melbourne urban area, as of the 2011 census, was 2,543 square km: so about 60 per cent larger, not “five times”.
Those BTCE figures back in 1992 were concocted in order to scuttle the proposition that Australian cities could support substantial new rail systems. In short, some transport planners keep saying our city is like Los Angeles because they want it to become like Los Angeles. “LA is inevitable” is an easier argument to sell the public than “LA is desirable”.
Since the alternative – to admit that transport planning in Melbourne was for half a century all-but-deliberately skewed toward boosting car use at the expense of public transport – reflects poorly on those responsible, it’s perhaps understandable they would prefer to argue there was somehow no alternative.
Of course, like the most successful urban myths, the ‘density narrative’ has a grain of truth. It’s perfectly obvious that Melbourne is more spread-out than English and European cities. And a century ago, Melbourne was a low-density city by world standards, because its extensive public transport system enabled people to live in the suburbs long before the car came along. In 1883 a visitor to Melbourne observed: “Terraces and attached houses are universally disliked and almost every class of suburban house is detached and stands in its own garden”.
Melbourne is now, however, more densely populated than most US and Canadian cities, which spread outwards more recently under the influence of the car. US Census figures published for the first time in 2008 (see below) show that Melbourne is more dense than Boston, Chicago and Portland – three of the cities with the most successful public transport in the USA (if still modest by world standards). And even the comparison with London is easily exaggerated, as we see above. London is still overall around three times more dense than Melbourne – but that doesn’t mean one can’t easily find parts of Melbourne that match or exceed the average density of London (and not all of these are in the inner city).
Density Done Badly
The US Census figures also reveal a startling paradox about Los Angeles itself. LA, in reality, is overall one of the most dense urban areas in North America. All American cities are surrounded by low-density suburbia, but LA’s suburbs were settled with densities somewhat higher than elsewhere – compensating for other factors such as the huge amount of space given over to roads. LA’s transport problems stem not from low density, it turns out, but from high density combined with car-centric transport policy – precisely where Melbourne is headed if we aren’t careful.
For an object lesson in what happens when planners obsess about density to the exclusion of all other matters, consider the sad case of East Portland in the US. It turns out that Portland, Oregon, combines some of the best and worst urban planning North America has to offer. Portland’s downtown and inner-suburban areas have built up a justifiable reputation as a new-urbanist success story. Meanwhile, to the east lie acres and acres of formerly rural land that, in accordance with a 1996 blueprint, was zoned for development at densities comparable to Melbourne’s highest density suburbs like St Kilda or Fitzroy. Yet this came with no mandate compelling the provision of services or infrastructure, resulting in vast tracts of apartments in a suburban desert without even footpaths, let alone adequate public transport.
City leaders now admit mistakes after years of complaints from residents. East Portland grew too quickly and without the sidewalks, parks and transportation system bestowed on other high-growth areas such as the Pearl District, Portland’s utopian planning playground….
Expectations were lofty. Large lots would be divided into small blocks with cozy streetscapes. Roads would be paved, sidewalks built, trees planted, transit service improved, the entire area cleaner and safer…. Planners blanketed the area with multi-family housing designations along key transit routes, enabling the construction of 22 to 65 units an acre.
[But] the level of community involvement wasn’t sufficient, [community housing provider Nick Sauvie] said…. New zoning in east Portland ushered a massive influx of homes and people. [But] new services to support the growth never materialized.
On 122nd Avenue….city planners justified zoning for as many as 65 units an acre because TriMet’s No.71 bus line was nearby. But frequent bus service hasn’t arrived. To the contrary, the 71 rumbles north and south 109 times each weekday, down from 121 in 1996. [The basic frequency is every 20 minutes, with a brief window of 15 minute services in peak hour on weekdays, and half-hourly service in the evenings and on Sundays.]
The tract’s eastern border is 136th Avenue, a two-lane road where city officials increased zoning to as many as 32 units an acre but never installed a sidewalk. One will be built next year following the death of 5-year-old Morgan Maynard-Cook, hit by a vehicle while crossing the street in February.
“Personally,” said Mark White, the former Powellhurst-Gilbert neighbourhood association president, “I can’t imagine an urban planner at any time in the city’s history thinking, ‘You know, I think it’s a really, really good idea to cram tens of thousands of people into an area with no place to shop, no place to work and no infrastructure.’”
—“East Portland’s housing explosion tied to city plan without basic services“, The Oregonian, 20 December 2013
There are two obvious morals to this story for transport planners. First, good public transport is not something that just materialises by magic if only the urban density is high enough: it has to be explicitly planned and funded by governments. Second, where public transport is of poor quality and unattractive, building up the urban density does not make it any more attractive – or any more financially self-supporting.
The same article notes that similar high-density development plans were abandoned in Portland’s relatively affluent southwest, highlighting another problem that arises when planners attempt to impose pro-density strategies in a top-down manner, without a high degree of community debate or control. The community’s default reaction in these situations, understandably, is outright opposition – and whether this actually succeeds in overturning the plans depends more on the political connections of powerful individuals and communities, more than on the merits of the plans themselves. So in Portland, densification was directed to the poorly serviced east rather than the established southwestern suburbs. And overall, its density remains a good deal lower than Melbourne’s.
Similar patterns of development are seen throughout the English-speaking world, including in Australia. One significant exception, however, is Vancouver in Canada. Vancouver has the distinction of being the only Canadian city where the average travel time to work decreased between 1990 and 2010 – despite rapid growth in population over that time. As it turns out, Vancouver’s planning strategy has two crucial features that distinguish it from others and can be held responsible for its success. First, Vancouver’s Liveable Region Strategy went to great lengths to involve local communities in key decisions from the outset. Second, the strategy prioritised public transport and other service improvements early on, recognising that changes in urban density and form are slow processes, but that people can take advantage of improved transport immediately.
Contemporary urban planning has learned from successful examples such as Vancouver, as well as from the ill-fated ‘dense sprawl’ of cities like Las Vegas, that the way in which a city is dense matters as much or more as how dense a city is. Crucially, it’s also recognised that one can have too much density. Lloyd Alter of the TreeHugger blog suggests cities should really be aiming for a ‘Goldilocks’ level of urban density, neither too little nor too much.
There is no question that high urban densities are important, but the question is how high, and in what form. There is what I have called the Goldilocks density: dense enough to support vibrant main streets with retail and services for local needs, but not too high that people can’t take the stairs in a pinch. Dense enough to support bike and transit infrastructure, but not so dense to need subways and huge underground parking garages. Dense enough to build a sense of community, but not so dense as to have everyone slip into anonymity.
—Lloyd Alter, The Guardian, 17 April 2014
Alter cites as examples of ‘Goldilocks density’ cities like Copenhagen, Barcelona and Vienna – the last of which famously pipped Melbourne for ‘most liveable city’ status in 2018.
It would be vain to search for a magic number of persons per hectare that achieves this Goldilocks effect. Barcelona and Vienna, for example, both share a sense of ‘European walkable urbanism’ to anyone who strolls through them, and both support excellent public transport systems allowing anywhere-to-anywhere travel. The fact Barcelona is actually twice as dense as Vienna would appear quite secondary. Vienna in turn is only a little over twice as dense as Melbourne, suggesting that perhaps we don’t really have so far to go to achieve similar success. (Were we to just double the overall density over one-tenth of the urban area, nearly half a million more people would be accommodated.)
Melbourne: Density versus Form
Melbourne does have a major geographical advantage that works to public transport’s benefit – and not just through being more densely populated than many American cities. Because our initial suburban growth was based around trains and trams, rather than the car, the form of development is more ‘public transport friendly’, with a strong central business district and most major suburban shopping centres near railway stations. Aside from a handful of postwar shopping malls, most of our suburban district centres date back to the late 19th century, contrary to those in most American cities where suburbs only became established after World War II.
While Melbourne’s public transport has historically been far from perfect, it is worth recalling that in 1950 the average Melburnian made 449 trips by public transport, and three-quarters of all motorised journeys to work were made by public transport. That this was possible in a sprawling city of “Federation bungalows”, that would not see its first skyscraper for another eight years, is a fact often overlooked by today’s commentators.
Melbourne’s population density declined a little in the years after 1950, but in recent times the decline has actually stopped (despite propaganda to the contrary). According to official figures, Melbourne’s overall density reached a minimum in 1981 and has since been increasing:
| Year | Overall urban density (people/ha) | Data source |
|---|---|---|
| 1951 | 23.4 | Melb. Metro. Planning Scheme 1954, p. 23 |
| 1961 | 21.4 | Australian Bureau of Statistics |
| 1971 | 18.1 | A.B.S. |
| 1976 | 16.75 | Melbourne Social Atlas, 1976 (A.B.S.) |
| 1981 | 15.9 | Social Atlas, 1981 |
| 1986 | 16.05 | Soc. Atlas/”Supermap” Census Data, 1986 |
| 1991 | 16.8 | Social Atlas/Supermap, 1991 |
| 1996 | 17.9 | Department of Infrastructure, 1998 |
More detailed figures from the Census confirm the general trend. In the early 1990s, the proportion of Melbourne’s population growth occuring within 20km of the GPO was 24% – a quite respectable proportion considering this was almost all within established suburbs. But in the three years to 2007, this proportion had jumped to 41%. For a while it had been evident that the population of inner Melbourne was booming; but now, established ‘middle’ suburbs whose populations were static or declining in the 1990s are also showing strong population growth.
Even in the new urban growth areas outside the 20km ring, analysis by transport planner Chris Loader shows that new growth suburbs in outer Melbourne are on average being developed with higher densities than similar existing suburbs (in population weighted terms). These suburbs also have slightly lower car ownership rates and slightly higher non-car mode shares for journeys to work than comparable existing suburbs, though both understandably remain high.
It’s become pretty clear in recent years that – whether in established inner and middle suburbs or in outer growth areas – there is more than enough population both to congest local roads and to fill local public transport services. Arguments about density ignore a real-world reality of overcrowded trains and buses, even in the suburbs.
As for the tripe about low density and public transport – have you ever travelled the Pakenham line? Those supposed low-density outer suburbs see trains full by the time they’ve stopped at five stations (ie. before we even reach Dandenong) at 7am.
—Comment to The Age, November 2010
A ‘back of the envelope’ calculation using the Census data shows that any location in the Melbourne urban area can support – at the minimum – a bus service running every ten minutes throughout the day, assuming just 20 per cent of people make just one return journey by public transport per day (see Appendix).
Perhaps the remarkable thing about the postwar growth of Melbourne is just how little it has altered the basic form of the city laid down in the 1880s land boom. The postwar brick-veneer suburbs are not fundamentally different from their ‘Hawthorn brick’ and ‘California bungalow’ counterparts of the 1880s and 1920s, and are even built to similar densities:
| Suburb | Period of settlement | Gross residential density (per ha) |
|---|---|---|
| Yarraville | 1890s-1920s | 25.4 |
| Balwyn | 1910s-1940s | 24.9 |
| Gardenvale | 1880s-1920s | 21.7 |
| Keysborough | 1960s-1980s | 36.7 |
| Wheelers Hill | 1970s-1990s | 26.0 |
| Bayswater | 1960s-1970s | 22.8 |
The appearance of Americanising tendencies in the 1960s now appears more like a ‘hiccup’ in a consistent long-term pattern. Trends toward ‘gentrification’ of the inner and middle suburbs reinforce this conclusion. Yet many commentators, mesmerised by the US experience and the false leads of the 1960s, seem to be ignoring the real nature of urban development in Melbourne. It often takes overseas visitors to observe the reality:
What you really have here is a European type of development at lower density.
—Matthew Quinn, a British transport analyst visiting Melbourne, 1992
I am embarrassed as an urbanist to have taken this long to see #Melbourne. Could this be the most walkable modern city outside Europe?
—Jeff Speck, author of Walkable City, via Twitter (@JeffSpeckAICP), August 2016
Although some recent development has been poorly planned, Melbourne is not a formless sprawl like modern American ‘anti-cities’ which grew up entirely around the car. We do not share the American pattern of ring freeways linking scattered suburban office parks impossible to serve by public transport. This is precisely why never-ending plans for ring roads and ‘missing links’ are so dangerous to Melbourne‘s future – and why we should be suspicious of phoney ‘agglomeration benefits’ used to sell these plans. Big roads merely promote more sprawl, the exact opposite of densification.
Density, shmensity: it’s all about service
Despite not sharing in Melbourne‘s extensive rail infrastructure and supportive urban form, even low-density North American cities have spent the past few decades developing viable alternatives to the car. In Toronto, for example, where in 1990 the average citizen made more than twice as many public transport trips as in Melbourne, the official transport plan has long aimed to
enhance the attractiveness of travel by transit in the Greater Toronto Area for a variety of trip purposes including, but not limited to, journey to work, and decrease reliance on the private automobile.
—Transit 2020, Toronto, 1993
Vancouver, where public transport use per capita is also higher than in Melbourne, was easily on target to triple patronage by 2021 from its 1991 level, as originally planned, with only the COVID pandemic temporarily getting in the way. As in Toronto, this was done by providing fast, frequent, integrated, safe and cheap public transport.
Both Toronto and (especially) Vancouver started out as spread-out cities in the early 1990s, but did not use that fact as an excuse for car-dominated transport policies.
| Melbourne | Toronto | Vancouver | |
|---|---|---|---|
| Population density in 1991 (per hectare) | 16.8 | 24.1 | 14.0 |
| Share of total jobs in Central area | 25% | 23% | 21% |
| Share of office space in Central area | 78% | 47% | 63% |
| Share of retail sales in Central area | 11% | 10% | n/a |
| Annual public transport trips (per capita) | 94 | 240 | 129 |
Vancouver also gives the lie to the assertion that public transport in spread-out cities comes only at high cost. Even in the early years of this century its entire budget for roads and public transport corresponded to just $180 per resident, compared with $430 per resident in Melbourne.
Unfortunately, many transport planners have completely failed to make the link between quality of service and patronage. The connection is obvious to anyone who checks the statistics, and holds true in low-density cities as much as in high-density cities. As a result, low patronage on public transport is too frequently excused as being residents’ fault for wanting big backyards, rather than a fairly obvious result of lousy service provision. Recall how it was assumed those high-density East Portland neighbourhoods could be adequately served with a bus every 20 to 30 minutes, and how that assumption failed utterly. Likewise, when Portland was forced to cut some other low-patronage services in the 2008-09 recession, one prominent (and usually much wiser) transport planner wrote:
Four routes are to be eliminated completely, and three of these (27, 154, and 157) are outer-suburban feeders…. All serve relatively low density areas but not especially affluent ones, a reminder that density determines ridership much more than wealth does. There’s not much of an alternative for residents of the areas served…. but if good transit service were really important to you, you wouldn’t live there.
—Jarrett Walker, Human Transit blog, 11 February 2010
As Walker well knows, residents of these Portland suburbs do have an alternative – private cars – and their decision to use them in preference to the cited bus routes was entirely rational. Of the three routes he named, two were ‘commuter’ routes that ran five trips a day on weekdays, while the third operated once an hour with the last bus at about 6:30pm. None of them ran on weekends. In short, they were equivalent to some of the very worst Melbourne suburban bus routes. Blaming density merely serves to excuse poor design and false economies, and lets the planners off the hook when poor service fails.
If density were the key to use of sustainable transport, then of course you wouldn’t live in Portland at all – you’d live in high-density New York, which rates highest in the US for public transport use. But you might just as well consider living in Los Angeles, Miami or Las Vegas: all cities with much higher urban density than Portland. Los Angeles even has a higher density than New York when entire urban areas are compared – a fact not widely known or believed until very recently. The problem is these cities, all ‘high density’ as they are, all have lower mode shares for public transport than even Portland does!
The table below gives the overall urban density and the public transport mode share for journeys to work in a selection of US and Australian cities. There is some relationship evident between density and public transport use, but it is weak and unconvincing, to say the least. Some other factor must be at work to explain why Brisbane, for example, has three times the overall rate of public transport use as LA despite being just one-third the overall density. That factor is good-quality service, which is present in Brisbane (at least in peak hour) but virtually absent in LA. Although we haven’t included Canadian cities, they do even better: Ottawa with 17.2 people per urban hectare has almost the same density as Miami, but differs from Miami in having one of the highest-quality bus systems in the world. Its 21.2% of journeys to work by public transport exceeds that in Miami more than fivefold!
| City | Overall density (/ha) | Travel to work by public transport (%) |
|---|---|---|
| Los Angeles | 27.3 | 4.7 |
| New York | 20.5 | 24.8 |
| Las Vegas | 17.7 | 4.1 |
| Miami | 17.0 | 3.9 |
| Denver | 15.4 | 4.4 |
| Chicago | 15.1 | 11.5 |
| Sacramento | 14.6 | 2.7 |
| Phoenix | 14.0 | 1.9 |
| San Diego | 13.2 | 3.4 |
| Washington DC | 13.1 | 9.4 |
| Portland | 12.9 | 6.0 |
| Boston | 8.9 | 9.0 |
| Sydney | 20.4 | 21.2 |
| Melbourne | 15.7 | 13.9 |
| Adelaide | 13.8 | 9.9 |
| Perth | 12.1 | 10.4 |
| Canberra | 10.6 | 7.9 |
| Hobart | 10.3 | 6.4 |
| Brisbane | 9.2 | 13.8 |
Source: Extracted from Mees, Transport For Suburbia (2010), Table 4.1
The Age also perpetuates the myth that “Melbourne is one of the Western world’s lowest density cities”. No it isn’t. I agree it’s not in the same league as super-dense Western (OECD) cities like Seoul, Tel Aviv, Palermo and Madrid, but it’s as dense as Lyon and Antwerp, and denser than Nice and Toulouse. More relevant to Australian history is that only five of the 40 largest cities in the US are denser than Melbourne. Some of the US cities that are less dense include Washington DC, Portland (Or), Philadelphia, Seattle and Boston.
—Alan Davies, The Urbanist blog, October 2013
Transport for Everyone
In the end, one can’t get away from the fact the 1956 Chicago Transportation Study still retains a tenacious legacy half a century on, particularly among planners and economists in the English-speaking world. The tendency remains strong to assume the demand for public transport depends entirely on factors external to the system (land use, population density) and to forget – or just downplay the evidence – that internal factors such as service frequency, speed and comfort are just as crucial.
In his influential book Human Transit (first published in 2012 and updated in 2024), Jarrett Walker (quoted in the previous section) provides valuable insight into the various factors that make public transport both successful and essential for city-building. But he nonetheless falls back on density, and takes issue with the comparative density figures for cities cited just above. He points out quite correctly that these are average density figures for entire urban areas, and mask huge variations between dense ‘urban villages’ and less dense suburbs within the same city. Human Transit devotes a central chapter to arguing why the provision of public transport service will be easier to justify in the former than in the latter.
Walker is perfectly correct here – but he goes on to make the same mistakes we have warned about above, setting aside all the good points he makes in the rest of the book about how to compete with car travel, and rejecting the idea that frequent, highly-patronised public transport can exist in any suburb below a certain density threshold (presumed to be higher than that of typical 20th-century North American or Australian suburbs).
Walker brings up the hypothetical suburbs of ‘Sparseville’ and ‘Denseville’ to illustrate low and high density. The density of Sparseville is stated as 7,500 people per square mile, or 29 people per hectare. This roughly matches the density of the eastern half of Coburg in Melbourne, where all points are within 800 metres of a frequent tram service and, according to the 2016 Census, 27% of people went to work by public transport. As another page details, this ‘low’ density characterises many inner suburbs of Melbourne such as Footscray or Clifton Hill, with similarly healthy rates of public transport use.
‘Denseville’ has twice the number of people per hectare and, by simple arithmetic, has twice the number of people within the catchment of any public transport line compared with Sparseville. All true: but the question remains, is Sparseville really all that sparse to begin with?
Walker urges planners instead to treat ‘ridership’ and ‘coverage’ as competing goals that must be traded off against each other, even within contiguously settled urban areas like Sparseville. If the aim is to maximise patronage and cost recovery, frequent service ought to be confined to the highest-density corridors within cities. The broadacre suburbs, meanwhile, may be provided with as much low-frequency so-called ‘coverage’ service as the budget and public sentiment will allow, but planners should not expect these services to ever attract many passengers. Our own planners draw on similar thinking when trying to justify why buses in so many Melbourne suburbs and regional towns only run once an hour or less.
Walker’s capability and experience are not in question. He understands the factors that make public transport attractive and competitive with car travel, and he devotes the other 16 chapters of Human Transit to explaining – in a very engaging manner – these factors and their importance. But when it comes to discussing how a city transit agency should maximise patronage within an available budget, it’s tempting to overlook two crucial facts about competitive public transport even when they’re acknowledged elsewhere. First, that the more one invests in making a service attractive, the more people will choose to use it rather than driving; and second, that higher patronage translates into higher revenue for the transit agency itself. This is particularly the case in low-density corridors, where capacity constraints are rarely an issue, and the main challenge is to find enough people for the bus rather than to find enough buses for the people.
Perhaps the issue is simply one we’ve seen so much before: that in car-dependent suburbs of ‘New World’ cities it is so difficult to imagine public transport improvements having a ‘revenue multiplier’ greater than one (so that the extra revenue raised actually pays for the incremental operating cost). Yet we’ve long had evidence for this in Melbourne. For example, when the off-peak frequency on the Sandringham train line was first boosted from three to four trains per hour in the early 1990s, there was a 33% increase in operating cost but a 40% increase in revenue – helped along by a boost in the proportion of full-fare paying passengers. The suburbs the line runs through had fairly uniform densities typical of established Melbourne suburbs of any age, around 20-25 people per hectare – again, less than Walker’s Sparseville. (And lest we forget, urban densities in typical 1990s Vancouver and Toronto suburbs were no greater.)
Often, then, the real lesson to be learned from lower-density suburbs is the need for planners to work a bit harder than they traditionally have done, making the service attractive enough to boost the share of the local population who use it. As the chief transport planner in Perth put it:
In Perth, the convenience of the motor car has historically been absolutely paramount. If we’re building new public transport, we must design it to compete with the motor car and be better. Otherwise, don’t bother building it at all.
—Peter Martinovich, WA Government transport planner
There are of course cases where this may not be feasible, and we find in this category a lot of North American ‘sprawl’ developments of the last half century, with their curvilinear street layouts, cul-de-sacs, housing subdivisions scattered through semi-rural land, and entire neighbourhoods with a single entry and exit point obliging buses to make lengthy, wasteful deviations.
In [northeast Houston] we were looking at essentially rural infrastructure, with no sidewalks and often not even a safe place to walk or stand by the road. Many homes are isolated in maze-like subdivisions that take a long time for a bus, or pedestrian, to get into and out of. And as the population is falling, the area is becoming more rural every year.
—Jarrett Walker, Human Transit blog, 2015
But while Australian suburbs have imported some of these features, they are not present to the same extent as in North America, and recent urban design standards actively discourage the worst elements. With a few notable exceptions (mostly due to a breakout of laissez-faire land use decisions in the 1990s), Melbourne’s suburbs mostly have the right geometry (as Walker puts it) to support effective public transport networks even if they’re middling to low when it comes to urban density.
As we have argued above, transport planners have little control over the population density in an established suburb, but they can influence the rate at which people use public transport. There can be no doubt that a ‘coverage’ service running once or twice an hour in a typical Australian low-density suburb will be mostly empty, because the rate of use will be low. But run the service more frequently, on a more direct route that allows people to get where they’re going quicker, and as part of a frequent network that connects to almost all destinations with at most one or two smooth changes of service, and big increases in the rate of use are possible even when the population density itself remains low. The Appendix below goes into more detail on this point.
And of course it will be true that if you took a 10-minute bus service in a low-density suburb and put it in a high-density suburb instead, the service would attract more passengers and generate more revenue. But this is not a reason to avoid offering frequent service in low-density suburbs. If the experience of Vancouver or Toronto is any guide, it’s quite likely that if that hypothetical 10-minute service were discarded in favour of a 30- or 60-minute ‘coverage’ service, it would still cost the agency about as much to run, due to an almost complete failure to attract revenue from full-fare paying passengers.
Except in the most poorly designed suburban neighbourhoods of North American or Australian cities, the question of ‘high patronage’ versus ‘coverage’ is a false dilemma. Toronto already demonstrated this in the early 1990s by running buses every 10 minutes or better in its low-density suburbs and achieving a mode share of 20% of motorised trips by public transport. Likewise, bus services in Melbourne’s suburbs should generally aspire to provide both comprehensive coverage and high patronage, by knitting together with trains and trams to provide a reasonably frequent, legible, go-anywhere-anytime network.
None of the above means, of course, that encouragement of medium density housing in parallel with service improvement is not worthwhile, particularly in the stressed housing conditions being seen in the 2020s. Vancouver famously introduced large tracts of high and medium-density housing under its Liveable Region strategy – but it’s critical to bear in mind that a lot of what North Americans call ‘medium density’ is not much different to the terraces and townhouses that have been traditional in parts of Melbourne for over a century. Meanwhile, many of our so-called ‘medium density’ developments would be considered ‘high density’ in North America.
So carefully targeted land-use measures will help – but the real challenge lies elsewhere.
Density is not the main barrier to providing public transport; rather, it is a rationalisation for inaction.
—Paul Mees, Transport for Suburbia
The problem is not inadequate road infrastructure….The problem is not that Melbourne is too dispersed….The above are all pseudo-problems posed by specialists of various persuasions to enable them to discover answers which serve and justify their interests and ideologies. Vast amounts of money are today being spent on pseudo-solutions which keep those interests in business. They have nothing to do with economic prosperity, environmental quality or accessibility in Melbourne.
—Nicholas Low, Senior Lecturer in Planning, University of Melbourne, 1995
Technical Appendix: How Much Population is Really Needed to Support a Bus Route?
The following calculation shows that any part of the Melbourne urban area is capable of supporting a 10-minute bus service for 18 hours a day if just 20 per cent of people make just one return trip each by public transport per day. It shows that under this assumption, any such bus route will cover its costs, even on the urban fringe; if we are prepared to accept a small public subsidy, an even greater level of service can be provided. In established suburbs with higher populations, a higher level of service is justified in any case.
We assume that the cost of running a bus is $90 per hour. This is based on estimates from various sources. Back in 2004, Southern Cross Station contractor Leighton publicly quoted $45 per hour as the cost of running tram replacement bus services in Collins Street, and this becomes about $90 after adjusting for CPI growth to the present day. With the economies of scale in a large bus operator like CDC or Kinetic the figure for an individual bus will be less than this, and industry literature generally quotes figures of between $60 and $90 per hour in present terms. We use the upper-end figure to ensure any errors are on the conservative side.
Estimating the amount of revenue generated on average by each additional passenger trip is difficult (particularly with private operators who treat their revenue and operating statistics as trade secrets), but given a Myki two-hour single-zone full fare starts from more than $3 and in Zone 1 is more than $5, and many passengers will only make one trip on this fare, a figure of $3 per full-fare passenger trip is likely to be a conservative estimate (and leaves some room to reduce fares while keeping services viable). Of course, the contribution from a concession passenger will be half this.
To recoup $90 an hour therefore requires 30 boardings by full-fare passengers per hour, or 540 over the course of an 18-hour day. The number of buses required to operate a 10-minute service over a route of length L is equal to the time taken to cover distance 2L, divided by 10 minutes. With an average bus speed of 20kph (inferred from current bus timetables), a bus can travel 3.3km in 10 minutes, so the number of buses required is N = L / 1.67 when L is expressed in kilometres. Accordingly the number of boardings per day required to break even is 540 x N = 324 x L – that is, 324 per kilometre.
Define the catchment area for the route as the strip extending 400m on either side, and suppose that in that catchment area 20% of people make on average one return journey by public transport per day (for two boardings). Suppose 50% of those are full-fare passengers, with the remainder counting half the value. Then each route kilometre provides 80 hectares (0.8 square km) of catchment and 0.3P full-fare equivalent boardings, where P is the population of the catchment area. For 324 boardings, P must be at least 1080, and so the population density must be at least 1080 / 80 = 13.5 per hectare.
Now, the average population density for the Melbourne urban area is at least 18 per hectare, based on census figures. We can get more detail on specific areas by looking at the population density maps that until 2006 were published in the ABS Melbourne Social Atlas. On the 2001 map, patches of urban area were shaded in one of five colours; the lowest-density patches were shaded pale grey (less than 15 per hectare) or dark grey (15 to 30 per hectare). On the 2006 map, five colours are again used, with the lowest-density patches being yellow (less than 10 per hectare), pale orange (10 to 20 per hectare) or medium orange (20 to 40 per hectare).
In order to be considered part of the urban area at all, a region (with some exceptions) has to have a density of at least 2 per hectare. As a rough estimate, and focussing on the 2001 map for the time being, we take the average density of a pale grey patch to be 8.5 per hectare (halfway between 2 and 15). Erring on the side of caution, we estimate the average density of a dark grey region at 20 per hectare, not far off the Melbourne average of 18.
Now consider a putative bus route in 2001 passing exclusively through pale grey regions of 8.5 per hectare and dark grey regions of 20 per hectare. In order to get a mean density of 13.5 per hectare or more, the route can have up to 56 per cent of its length bordered by pale grey regions, and still get full cost recovery. With the exception of the Mornington Peninsula and the Dandenongs, there were very few putative bus routes on the 2001 Social Atlas map bordered with pale grey regions for more than half their length. Whether the route is in Craigieburn, Montrose or Werribee – or indeed in St Kilda – there would appear to be sufficient population along the route to make it viable.
Looking at the 2006 map, a notable difference from 2001 is the clearer contrast between settled urban corridors and ‘green wedge’ areas within the Melbourne boundary, as a result of introducing a 10-per-hectare threshold. What is quite clear from this map is that most public transport routes will be found in the orange-shaded regions and relatively few in yellow regions such as the Dandenong Valley parklands (though a few routes will have to make a relatively quick traversal of these areas). A typical suburban route will therefore combine mainly pale orange and medium orange patches with possibly some yellow. Even if we conservatively assume the low-density yellow patches account for as much as 20 per cent of the route and that the typical density is only 2 per hectare for yellow, 10 per hectare for pale orange and 25 per hectare for medium orange (near the lower end of each band), one finds that the break-even density is reached with as little as 34 per cent of the route bordered by medium orange (and 46 per cent by pale orange). Again, the majority of practical routes one plots on the map are likely to satisfy these criteria.
Of course, the crucial assumption here is that 20 per cent of people make one trip by public transport each day, which roughly corresponds to public transport being used for 20 per cent of all trips (as was once a State Government policy objective). If the mode share is only 3 per cent, as in many Melbourne suburbs at present, then the density required for viable public transport goes up proportionally: by our calculations, to 90 per hectare instead of 13.5 per hectare. And if concession passengers represent 90% of boardings, rather than just 50%, this pushes the figure up again, to a massive 123 per hectare. Very few parts of Melbourne record population densities this high.
This basic observation lies at the root of road lobby calculations (like those by Alan Moran above) purporting to show that viable public transport requires the kind of enormous population densities seen only in crowded cities like New York, Tokyo or Singapore. Basically, if you assume that drastic improvements in service frequency (from typically 40-60 minutes at present to 10 minutes in our scenario) don’t induce more people to use the service, it’s not surprising the improvement turns out hard to justify!
Our estimate of 13.5 people per hectare required to justify a 10 minute bus service compares favourably with other estimates by transport experts who make realistic assumptions about the way people respond to improved services: 12.5 per hectare in the 1965 Brisbane Transportation Study; 12 per hectare by Thompson in Great Cities and Their Traffic (1977); 15 per hectare by Pushkarev and Zupen in Public Transportation and Land Use Policy (1977); and 14 per hectare by Mees in A Very Public Solution (2000). Each of these authors use slightly different assumptions to arrive at their figure, none of which concide precisely with ours; it is thus especially reassuring that they all arrive at a similar result.
It’s not hard to see how varying some of our assumptions affects the final result; since our assumptions were deliberately stacked against the viability of public transport, the effect is usually favourable. For example, if services are not expected to cover all their costs but can instead be subsidised by up to 50%, we can cover areas with a population density right down to 7 per hectare (including the Mornington and Bellarine Peninsulas and the Dandenongs). If bus priority improvements raise the average speed of buses from 20kph to 25kph, or economies of scale mean buses can be run for $80 instead of $90 per hour, then costs go down roughly in proportion. One can then provide a couple of additional services in peak hour, or run night services, or reduce fares, without increasing the subsidy.
(Perhaps some reading this might also ask: what about people making single trips using multiple services? Won’t their $3 revenue contribution have to be split across these several routes? Yes it will – but this is automatically factored into our calculations above. Keep in mind that when we counted the ‘passenger catchment’ along the route assuming a 20% rate of trip-making by public transport, we only counted boardings by people who are either commencing a new trip starting from home, or completing a trip by returning home. People who have changed from other public transport services are not counted. They will add to the numbers boarding at stops, but the local population density does not include them.)
Last modified: 10 August 2024