5 Mobility: Transport Is a Real Estate Issue—The Design of Urban Roads and Transport Systems
The Need for Mobility
Cities are primarily large labor and consumer markets. These markets work best when the possibility of contact increases between workers and firms, among firms themselves, and between consumers and commercial and cultural amenities. The term “mobility,” in the context of this book, defines the ability to multiply these contacts with a minimum of time and friction.
A worker’s ability to choose among many jobs and a firm’s ability to select the most qualified workers depends on mobility. Mobility is not defined by the ability to get to one’s current job quickly, but by the ability to choose among all jobs and amenities offered in a metropolitan area while spending less than 1 hour commuting. Mobility increases when the number of jobs and amenities that can be reached within a specific amount of time increases. Because of the impact of mobility on the welfare of a city, it is important to measure it and monitor its variations—up or down—as a city’s population increases, its land use changes, and its transport system improves or deteriorates. I will propose ways to measure and compare mobility in different cities in the section “Mobility and Transport Modes” later in this chapter.
The objective of an urban transport strategy should be to minimize the time required to reach the largest possible number of people, jobs, and amenities. Unfortunately, many strategies, such as “compact cities,” only aim to minimize the distance traveled by inhabitants. These strategies reduce the income of the poor, for whom employment opportunities are reduced to jobs located within a narrow radius of their homes.
Cities thrive on changes, possibilities, and innovations. Therefore, an urban transport system that would solely minimize travel time between home and current jobs for all workers would result in poor mobility, as in the future, workers might not be able to reach many alternative jobs that would improve their job satisfaction or salary.
Mobility and Recent Immigrants
During a recent visit to the Tenement Museum in New York, a docent told us that in the 1850s, immigrants who were “fresh off the boat” would typically stay only a few months in a tenement; they would then keep moving as their employment and financial circumstances changed. A typical length of stay in the same tenement would be about 6–8 months. My wife and I then looked at each other, remembering that this was exactly what we did when—in January 1968—we were also “fresh off the boat” in New York. We changed apartments three times in 30 months. We moved from a flophouse on the Upper East Side that was soon going to be demolished, to a studio apartment in an “old law tenement” on the Upper East Side, and then to an entire floor in a townhouse in Brooklyn Heights. I also changed job three times. Each time, I changed for a more interesting job and a higher salary. This is the type of mobility that we will discuss in this chapter: the ability to move from job to job and from dwelling to dwelling made possible by a transport infrastructure that gives access to millions of potential jobs in less than 1 hour of commuting time.
This mobility was made possible by a buoyant housing and job market, ensuring a low transaction cost of changing jobs and location. By contrast, in Paris (where we came from), housing mobility was hampered by 2-year leases that could not be broken without penalties. Additionally, job mobility was frowned on as a sign of instability—changing jobs three times in 30 months would have resulted in a resume that raised a lot of eyebrows.
When—after just 6 months with my first employer in New York—I found a job that was a better fit with my long-term interests, I was terribly embarrassed by the prospect of telling my employer that I was quitting. My colleagues at work reassured me that this was done all the time in New York, and that a higher salary was a very honorable reason to change jobs. Indeed, my employer gave me a good luck party when I quit!
This is mobility. A flexible labor market, an open housing market—the flophouse with its low standards but very low rent was essential to getting us started—and a transport system that is fast, affordable, and extensive enough to allow individuals to look for jobs in an entire metropolitan area rather than just in limited locations.
The benefits of urban mobility are not limited to saving on commuting time. Mobility is also necessary to facilitate random face-to-face encounters between individuals of different cultures and fields of knowledge. These serendipitous encounters increase cities’ creativity and productivity. The multiplicity of easily accessible meeting places available outside the work setting increases the possibilities of chance encounters and therefore increases the spillover effects found in large cities. The agora of ancient Greek cities or the forum of Roman cities were precisely fulfilling these needs. Agoras and forums were places where people assembled to conduct business, to meet friends, to attend religious ceremonies and political meetings, to receive justice, and to frequent public baths. Modern cities have many of these functions in separate locations. Unfortunately, rigid zoning regulations often constrain the existence and location of these multifunction places.
When transport systems provide adequate mobility, then the large concentration of people in metropolitan areas increases productivity and stimulates creativity. Empirical data confirm the link between large human concentrations and productivity. Physicists from the Santa Fe Institute have shown that, on average, when the population of a city doubles, its economic productivity per capita increases by 15 percent.1
The interesting findings of the Santa Fe Institute’s scientists should be qualified, though. Their database included 360 US metropolitan areas with, by world standards, a very good transport infrastructure network that ensures mobility together with spatial concentration. In a way, these scientists’ use of the word “cities” assumes the availability of transportation. It would be wrong to interpret their work as demonstrating that human concentration alone increases productivity.
Mobility explains the link between city size and productivity. Human concentration alone does not increase productivity. Some rural areas in Asia have gross densities that are higher than the density of some North American cities like Atlanta or Houston, for instance. However, in these rural areas, mobility is poor to nonexistent between villages. In absence of mobility, there is no increase in productivity despite the high density. The productivity of cities therefore requires both concentration of people and high mobility.
When the time and cost required to move across a city increase, mobility decreases. When this happens, workers have fewer choices among the potential jobs available in a city, and firms have fewer choices when recruiting workers. In these conditions, metropolitan labor markets tend to fragment into smaller, less productive ones; salaries tend to decrease, while consumer prices increase because of lack of competition. In practical terms, labor market fragmentation means that a worker might not find the job for which she is qualified, because she cannot commute in less than 1 hour to the firm who could employ her. Conversely, the firm looking for a worker with specialized knowledge cannot find him because he cannot reach the firm in less than a 1-hour commute. Workers having to commute for more than 1 hour each way are penalized by a social cost that progressively destroys their personal life. Poor mobility may also result in high transport overhead cost for firms having to exchange goods and services in an urban area. Increasing mobility in urban metropolitan areas is therefore indispensable to the welfare of urban households as well as to the creativity and prosperity of firms.
In a city, a worker’s mobility often depends on his income. In some large Indian cities, for instance, the poorest workers can only afford to walk to work. Even a very long walk of 90 minutes would give them access to a very small number of possible jobs, decreasing their potential earnings. Planners should measure separately the mobility of different income groups, accounting for the modes of transport that each group can afford.
Mobility generates not only benefits but also costs, including congestion, pollution, noise, and accidents. To reduce those nuisances, many urban planners advocate limiting or at least discouraging mobility. They dream of creating cleverly planned land use arrangements that would require only short trips easily covered by walking or bicycling, even in megacities. These utopian land use arrangements usually rely on complex land use regulations2 that would enable planners to match employers’ locations with employees’ residences.
Mobility is an urban necessity that must be encouraged, not curtailed. Poor mobility keeps much of the economic potential of existing large cities from being realized. Unfortunately, in many cities, the lowest-income households suffer the most from poor mobility. They would enormously benefit if their mobility increased, so that they could look for jobs as well as cultural and commercial amenities throughout an entire metropolitan area. Instead, they are limited to the small area around their homes, circumscribed by their limited mobility.
As urban metropolitan areas increase in size and population, their potential large labor markets may fragment into smaller markets because of the lack of mobility. It is therefore necessary to differentiate between the potential and actual size of the labor market. The potential size is equal to the number of workers and jobs in a city. Its real size equals the average number of jobs that a worker can reach in a 1-hour commute.
Commuting Trips and Other Trips
Throughout this chapter, we will examine mobility in the context of commuting trips (trips from home to work and back), even though commuting trips are only a fraction of urban trips. In the United States in 2013, commuting trips represented only 20 percent of weekday urban trips, 28 percent of vehicle kilometers traveled, and 39 percent of public transport passenger-kilometers traveled.3
Households and firms generate many types of trips that have different purposes (e.g., trips to work, to school, to visit friends, to shop). Many trips have multiple purposes. Transport engineers call such excursions chained trips or linked trips. On a chained trip, a person might drop a child off at school, go to work, and shop, for instance. Chained trips are both convenient for the commuter and efficient in terms of transportation, as they save time and reduce the distance traveled compared with the same trips done separately. For the United States, the Pisarski and Polzin study indicates that 19 percent of all women’s trips are chained trips compared to 14 percent for men. While chained trips are transport efficient, they are nearly incompatible with public transport and carpooling.
Despite commuting trips representing only a fraction of all trips, I will continue to use them to measure mobility. For the economic viability of a city, the most important trips are those to and from work—the commuting trips—as the labor market generates the wealth that makes the other trips possible. In addition, the timing of commuting trips is usually not chosen by the traveler. Instead, they often occur at peak hours, and they are the ones causing most congestion and pollution. Therefore, the transport infrastructure capacity needs to be calibrated on the demand during peak hours, largely determined by commuting trips.
Some elective trips, like holiday shopping or leisure trips on summer weekends, may also cause heavy congestion, but they are seasonal and therefore do not have such high annualized costs as the daily congestion caused by commuting trips.
Improving Mobility Is Not As Simple As Making Cities Denser
Ideally, the closer people and firms are to each other, the shorter the trips required to meet and transact business would be. In an urban area with a given population, people, firms, and amenities are closer to one another when population and job densities are higher. It may seem, therefore, that for a given population, mobility simply increases when density also increases due to the shorter distances between households and firms. Similarly, it would seem that mobility would decrease as the distance between firms and employees increases.
Unfortunately, things are not that simple. Let us consider the average distance d of commuting trips between random points A and B selected in a city built-up area. For a given population, the distance d will indeed be shorter if the city’s density is higher. However, mobility increases when the time t needed to cover the trip distance d from A to B decreases, and not necessarily when d alone decreases. Therefore, mobility increases not only when the trip distance d decreases but also when the trip travel speed v increases (t = d/v). The trip’s speed v depends on the mode of transport and the area devoted to roads. Therefore, increasing densities may decrease the average distance d between people and jobs, but it may also increase congestion and therefore decrease travel speed v.
Let’s make this come alive through an example. Nineteen-century London, with its sweatshops and slums, was extremely compact. In 1830, according to Shlomo Angel and colleagues,4 London’s population density had reached a very high density of 325 people per hectare. By 2005, however, the density of London had decreased to only 44 people per hectare. The large decrease in London’s density since the Industrial Revolution has not caused a corresponding decrease in mobility. On the contrary, transport modes in London in 1830, largely walking and horse carriages, were much slower than those available in 2015 London with its choice of various motorized transport modes. In 2015, commuters can reach the city center from the suburbs as far as 26 kilometers away in less than 1 hour by public transport. In contrast, in 1830, commutes from the edge of London, about 7 kilometers from its center, would have taken about 1.5 hours. In this case, a seven fold decrease in density did not result in a decrease in mobility; on the contrary, an improvement in transport technology generated an increase in mobility despite the sharp fall in population density.
The goal is to reduce the time spent traveling and the cost of transport, not necessarily to reduce the distance between trip origin and destination. The mode of transport and the design of the transport network will have a much more impact on mobility than distance traveled.
How to Increase Mobility as Cities Expand
The population of successful cities is constantly increasing because of the economic advantages provided by large labor markets. To maintain mobility as city populations increase, urban transport systems must adapt to the new size of cities. In relatively small cities—around 200,000 people, like Oxford or Aix-en-Provence—a combination of transport modes like walking, bicycles, and city buses provide adequate means of transport in the downtown area, while individual cars and motorcycles are used for trips in the periphery. However, when a city’s population increases above 1 million, these means of transport become inadequate, and new means of faster transport must be built. Because the land in the center of large cities becomes more expensive, new transport must not only be faster but also should use less of the expensive urban land, hence the necessity to develop underground or elevated transport systems.
Transport systems that may be adequate for a given city size soon become deficient in a larger city. Transport systems cannot just be scaled up but need to be entirely redesigned when cities grow larger. It is futile to use Amsterdam’s or Copenhagen’s transport system as a model for much larger cities like Mumbai or Shanghai.
As the size of cities increases and traditional land use patterns keep changing, it is imperative to keep monitoring mobility, the direct cost borne by commuters, and the negative impact it creates on the city environment.
Mobility Creates Friction
Urban mobility creates friction. The larger the city, the more severe the frictions caused by mobility will become. These frictions include the time and cost required to go from one part of a city to another and the congestion and pollution created by doing so.
The frictions caused by urban transport are not new. Urban congestion did not begin with the advent of cars. A golden age when cities were congestion free never existed. The Latin poet Juvenal, in his Satire III, mentioned the difficulties of moving around ancient Rome in the first century. Traffic congestion in the Roman Empire is even the subject of a recently published book!5 In the seventeenth century, the poet Boileau wrote a satirical poem about les embarras de Paris (the gridlocks of Paris).
At the end of the nineteenth century, pollution due to transport was such a concern that some saw it as a limiting factor in the growth of cities. London was at the time the largest city in the world, with 6 million people. The pollution so reviled then was the enormous quantity of horse manure produced daily by horse-drawn buses and cabs. Indeed, it was a major health concern. Acknowledging London’s future increase in population, scientists projected that the quantity of horse manure produced by transport would soon bury the city like a modern Pompeii! With hindsight, we know that the introduction of the automobile saved London from submersion in manure, but congestion and pollution remain a major constraint in urban transport today.
Various frictions caused by urban transport have been a constant concern since the dawn of urbanization. They cannot be eliminated, at least with current technology, but they can be decreased. These frictions will be discussed separately below: they include direct travel cost, time spent traveling, congestion, pollution, and other indirect costs. Any city that can significantly decrease frictions due to transport will see a corresponding increase in productivity and the welfare of its citizens because more time will be left for work and leisure.
A primary task of city managers should be to minimize the frictions caused by urban transport. This job is never done—as a city expands, the distance covered by commuting trips becomes longer. A city structure and its transport system must adapt continuously to its changing scale. As a city expands from 1 million to 10 million (e.g., as Seoul Municipality did between 1950 and 2015), the original transport system cannot simply be expanded; it must change in nature and technology to reflect the new scale of the labor market being served. The goal is to maintain mobility such that the majority of commuting trips stay below 1 hour, in spite of the much longer distances involved.
The unfortunate tendency of many current traffic managers is to restrict trips to avoid congestion. Instead they should better manage the road space available or adopt new technology to allow even more and faster trips.
The Death of Distance Has Been Greatly Exaggerated
In the “Star Trek” television series, the words “beam me up” were all that was needed to transport people and goods anywhere instantly through the teleportation machine. This imaginary technology allowed universal frictionless mobility. Unfortunately, it was fictional.
If frictionless mobility were possible, the dense concentration of people in cities would not be required. I could start the morning in a small town in New Jersey, a few minutes later have coffee and a croissant in a café in Paris, and a few seconds after finishing my coffee, I could start working in an office in Mumbai, or anywhere else in the world. In a world allowing frictionless mobility, location would no longer matter. Most of us have already replaced some physical trips by virtual ones. For instance, I used to visit bookstores on a regular basis; I now buy my books online, and they are delivered to me electronically. The visit to the bookstore was a face-to-face encounter that has been replaced by a “beam me up” operation, except that it is the book that is being beamed up, not a person.
While the teleportation machine from Star Trek is likely to remain fiction, would communication technology—in particular, increasingly realistic teleconferencing—make location obsolete, providing a substitute for frictionless mobility? Or, in simpler terms, could communication technology replace the face-to-face contacts that generate most of our commuting trips? Indeed, it is much cheaper to move data than to move people. This is precisely the main argument developed by Frances Cairncross in her book The Death of Distance (2001). Cairncross suggests that the Internet and the global spread of wireless technology are increasingly making distance irrelevant. Communication technology would make face-to-face contact obsolete, and, in this sense, we would be getting closer to a Star Trek–like frictionless mobility, replacing the mobility of individuals by that of data. Virtual reality encounters would replace the necessity of “in the flesh” face-to-face encounters.
Work-at-Home Individuals
The increasing share of people working at home but constantly connected to a front office seems to confirm Cairncross’ prediction. Among eight of the ten US cities with the largest number of work-at-home individuals (figure 5.1), the share of people working at home is larger than the share of workers using public transport. In all nine cities, the increase of work-at-home individuals has been larger than the increase in public transport users. However, with the exception of San Francisco, all the cities shown on figure 5.1 have rather low densities by world standards. This may explain the low proportion of growth of urban transport compared to the growth of work at home. In addition, many workers commute only a few days a week and are only working at home part-time. If this trend continues, could the home become the main place of work, rendering commuting obsolete and resulting in trips mostly for leisure or personal reasons?
Mode share work at home versus public transport in some US cities. Source: Wendell Cox, NewGeography, May 30, 2015.
Working from home is not new, of course. Up to the early twentieth century, artisans and service workers were often working from home, delivering their finished work to their employers weekly. These included washerwomen and lace makers, but also Swiss farmers making watch mechanical parts. What is new, however, is that clerical and technical workers, who traditionally worked in large office pools, have replaced these manual workers working from home. But is it likely that a very large part of the workforce—say, more than 25 percent—will start working from home full time, significantly reducing peak hour traffic? So far, this possibility seems limited.
A recent Yahoo human resource department’s memo requested employees working from home to resume working at the office, arguing, “Some of the best decisions and insights come from hallway and cafeteria discussions, meeting new people and impromptu team meetings. Speed and quality are often sacrificed when we work from home.” In Silicon Valley, the most successful firms, like Google and Facebook, are building very large headquarters in addition to the large office buildings they recently acquired in downtown San Francisco. These large and costly real estate acquisitions suggest that they do not anticipate that a large part of their workforce will be working full time from home in the future.
The design of the largest Silicon Valley offices, offering their employees an unusual environment with gourmet food cafeterias, gyms, and kindergarten, demonstrates the intention of management to encourage their employees to work in their office and to interact with one another socially as well as professionally. In a way, it seems that Silicon Valley firms are trying to intensify within their offices the knowledge spillovers that are known to happen in large cities. For these reasons, I believe that the number of work-at-home individuals might soon reach a peak and may not affect commuting flow significantly in the future.
If Cairncross had been right in 2001, by 2015 we should have already seen large changes in the price of urban land across the world. The most environmentally attractive but remote rural areas of the world would have higher prices, while the least environmentally attractive, highest-density areas would have lost value. This is not happening. Real estate prices in New York, London, Delhi, and Shanghai are still climbing, proving that the death of distance might have been greatly exaggerated. High real estate prices demonstrate that even in cities where mobility causes severe friction—as in New York, London, or Shanghai—being physically close to a large concentration of people, jobs, and amenities is still worth a very high price.
Measuring a City’s Mobility
A Decrease in Congestion and Pollution Is Not a Measure of Mobility
The objective of urban transport is to increase mobility to maximize the effective size of labor markets. Congestion and pollution are very important constraints on the mobility objective, but they are only constraints. Confusing objectives and constraints when solving problems can lead to false solutions. Urban managers too often try to solve transport problems by focusing exclusively on reducing congestion and pollution without giving much consideration to mobility, as if the objective of urban transport was limited to decreasing the nuisances it causes.
Some policies rely on reducing trip length, others on forcing more commuters into slower transport modes. None of these policies effectively reduces pollution or congestion, but they reduce mobility.
Planners who think that decreasing pollution and congestion is the main objective of urban transport might logically try to fragment a large metropolitan labor market into smaller ones. For instance, some planners suggest that matching the number of jobs with the size of the working population in every neighborhood would significantly decrease trip length to the point where walking and bicycling could provide access to all the jobs in a neighborhood.
Of course, there is nothing wrong with having mixed-use neighborhoods, provided that demand from households and firms drives the land use mix. Even where planners can achieve a perfect match between the number of jobs and housing units, as in satellite towns, experience has shown that workers prefer access to wider labor markets and that there is no decrease in trip length. This has been shown in Seoul’s well-planned satellite towns.6 After an exhaustive survey of land use and trip length in California, the transport economist G. Giuliano concludes that “regulatory policies aimed at improving jobs-housing balance are thus unlikely to have any measurable impact on commuting behavior, and therefore cannot be justified as a traffic mitigation strategy.”7
Understanding why job-housing balance does not reduce trip length is easy. If it did, it would imply that at least one of the following propositions is true:
• All workers within a household only look for jobs within a short distance from their home.
• When workers change jobs, they also change homes, and moving from one home to another has a negligible transaction cost.
• Proximity to work is the only consideration when selecting a home.
Obviously, common sense shows that none of these propositions is true for the majority of households. If any of these propositions were true, then we would observe a fragmentation of labor markets and a decrease in mobility, and therefore a decrease in urban productivity.
The job-housing balance policy is, of course, not implementable in a market economy. This is because the number of jobs and the number of workers is always fluid, and no government, however authoritarian, can force people to live and work in a specific location. Even in the Soviet Union and in pre-reform China, where large state-owned enterprises provided housing for their workers, who often spent their entire career working for the same enterprise, planners could not achieve a spatial match. As I was working on housing issues in China in the 1980s and in Russia in the 1990s, I was surprised to see that, even in command economies, the utopian dream of matching jobs and housing location could not be achieved. Large enterprises had to expand in locations distant from their workers’ housing, and they had to build new workers’ residential estates in areas where they could find the land, which was not necessarily close to their factories. As soon as labor markets opened in both countries, the job-housing balance deteriorated further. A fluid labor market (which is what makes large cities so attractive) and a job-housing balance are incompatible.
However, despite these negative experiences, planners still devise land use regulations aimed at matching people with jobs. For instance, a regulation in Stockholm requires developers to match the number of jobs and the number of dwelling units in new suburban locations. Allowing mixed-use development is a good land use policy, as it allows households and firms to select locations that best meet their needs without the rigidity of arbitrary top-down land use zoning. Requiring a perfect match between population and jobs in each neighborhood in order to reduce trip length is an unattainable utopia.
Other policies also eagerly sacrifice mobility to reduce pollution and congestion. For example, several Latin American cities (e.g., Bogotá, Santiago, and Mexico City) have instituted a vehicle rationing system called “pico y plata” that restricts the circulation of vehicles on 2 days per week, depending on the last number of the vehicle’s license plate. This policy reduces mobility.8 It forces drivers either to switch to public transport or to carpool for 2 days per week. The change of transport mode is likely to require a longer commute time during the days drivers are obliged to switch. If public transport were faster, they would have used it before the restriction on driving was put in place.
Studies show that drivers circumvent “pico y plata” regulations by buying a second car with a different last number on the license plate. Traffic and congestion initially decreases after this regulation is implemented, but then increases again when the second cars join the traffic. The result is more pollution, because there are more cars on the road and because the second car bought is usually an older, more polluting model. Many studies across cities in different countries and incomes have confirmed this result. However, the regulation, which restricts mobility in order to reduce pollution and congestion, is still popular among city managers. This stance is counterproductive.
In some cities, exceptional climatic events may cause extremely dangerous pollution peaks on some days. In this case, restricting individual car use is, of course, legitimate as an emergency measure—as it is legitimate to ask factories to stop operating during the emergency—but it is not efficient to use such restrictions as permanent policy.
Accessibility and Mobility: What Is the Best Way to Measure Urban Mobility?
Transport policy should aim to increase mobility while decreasing congestion and pollution. Often, reports that claim to quantify mobility in fact only measure the cost of car congestion and pollution. For instance, the “mobility report”9 prepared by Texas A&M Transportation Institute (2012) argues that a shift from car to public transport, which indeed obviously reduces road congestion, is considered an improvement in mobility. For some reason, the reduction in the time spent commuting for the drivers who keep driving is considered a benefit, while the longer commuting times for the drivers who have shifted to public transport is not considered a cost. Mobility would increase only if the commuting time of the driving commuters who have changed to public transport becomes shorter because of the shift. However, if commuting by public transport were faster than commuting by car, drivers would have switched modes already.
Congestion clearly decreases mobility, but measuring it is not a substitute for measuring mobility. For instance, imagine a person having to walk 1 hour to work because of poverty but eventually being able to afford a collective taxi to make the same trip in 30 minutes. The collective taxi will contribute to congestion; walking did not. However, the mobility and welfare of this worker shifting from walking to a collective taxi ride would have increased. We should therefore measure and monitor the variations in mobility for different income groups.
While congestion is usually measured for car traffic only, congestion can occur at bus and Bus Rapid Transit (BRT) stops and in metro stations. While attempting to use the BRT in Mexico City in 2014, I saw three buses pass the station where I was waiting without being able to board, the buses being able to take only a few passengers among the more than 100 individuals waiting on the platform. This is also congestion, which planners should measure. Trying to shift commuters from one congested transport mode to another congested mode doesn’t decrease congestion problems. To my knowledge, the Beijing Transport Research Center is the only monitoring institution measuring daily congestion in metro stations. This institution measures the time required to board a train at peak hours.
Consequently, measuring and monitoring mobility for all modes of transport at the metropolitan level is an indispensable step for improving urban transport. A quantitative index measuring mobility improvements or setbacks is necessary to provide substance to urban transport policy. Advocating “mobility” without a way of measuring it will just add a new faddish slogan similar to “sustainability” and “livability.” Both slogans are unmeasurable and are too often used by urban planners to justify whatever policies they favor. Measuring mobility is not easy. I will describe some of the methods currently used and some that are emerging thanks to new data-recording technology.
In chapter 2, I explained why large labor markets are the raison d’être of cities. Large labor markets result in higher productivity than smaller ones. However, the size of a labor market is not necessarily equal to the number of jobs in a city. If inadequate or unaffordable transport prohibits workers from accessing all of a city’s jobs within an hour’s commute, the effective size of the labor market is only a fraction of the total number of jobs in the city. The productivity of a city is proportional to the effective size of its labor market. Mobility allows workers to have access to a measurable number of jobs within a specified travel time and can therefore be measured by the effective size of a city’s labor markets given a specific travel time.
A useful measure of urban mobility would calculate the average number of jobs that workers can commute to within, say, an hour one way. We could calculate such a mobility measurement by aggregating the number of jobs accessible in less than 1 hour from every census tract, weighted by their population. A mobility index, therefore, would have to be calculated in two stages: first, by calculating the number of jobs accessible from every census tract within a selected time limit; second, by calculating the worker-weighted average of the accessibility of all census tracts to form an index reflecting the entire metropolitan area.
Traditionally, transport planners have measured job accessibility from different census tracts in a metropolitan area by measuring the number of jobs accessible from the census tract corrected by coefficients that reflect distance, cost, and elasticity of demand related to distance. The formulas used to measure accessibility of census tracts are usually similar to the ones I highlight in equations 5.1 and 5.2.
Equation 5.1 Job accessibility per census tract
The index of accessibility can be calculated using the equation
where Ai is the index of accessibility of census tract i, Kj is the number of jobs in census tract j, e is the base of natural logarithm, β is an elasticity coefficient, c is the unit cost of traveling the distance dij between census tract i and census tract j. While these formulas provide a way of measuring access to jobs or amenities from a specific location, the measurement they provide is an abstract index dependent on the way distances, costs and speed, and cost elasticity are calculated. Transport planners have a tendency to make accessibility measures more complex by adding more variables reflecting the complexity of commuters’ behavior. Unfortunately, this complexity renders accessibility calculations more difficult to interpret. As a result, their “black box” effect prevents their use in formulating transport policies that non specialists like mayors or city councils must approve. It is therefore indispensable to develop a much simpler accessibility index, based solely on the size of the labor market available to residents of a particular census tract based purely on travel time using existing transport modes. The advances in Geographical Information System (GIS) technology allow interactive use of maps where areas accessible within a given travel time can easily be verified, as shown by the Buenos Aires example below.
Equation 5.2 Number of jobs accessible by census tract within a set travel time
The first step in developing a mobility measurement that reflects the number of jobs accessed within a set travel time would be to change the traditional accessibility formula into the simpler and more explicit:
where Ai is the number of jobs accessible from census tract i within a commuting time lower or equal to maximum travel time T, and v is the average travel speed to cover the distance dij between tract Ai and tract Kj using the network of the mode of transport selected.
The values of v and dij are dependent on the mode of transport: public transport, bicycle, or car. Therefore, we should calculate the different value taken by Ai for each mode of transport. This accessibility index measuring the number of jobs accessed in less than a trip time T would be repeated for all census tracts in the urban area and for the major mode of transport available: public transport, cars, motorcycles, bicycles.
A few years ago, such calculations would have been extremely cumbersome and costly and, if performed, unlikely to be repeated for periodic monitoring. Two factors now allow for easy monitoring of an urban mobility index. First, new GIS-based technology enables the development of interactive tools accessible to any user. Second, the standardization of transport data networks (the General Transit Feed Specification, or GTFS)10 is becoming universal. This allows for the calculation of accessibility based on real transport networks and real travel times—including transfers between stations and walking times to and from stations—instead of crude “as the crow flies” distances between census tracts. In the following paragraphs, I use data extracted from research conducted by Tatiana Quirós and Shomik Mehndiratta11 for Buenos Aires.
Figure 5.2 shows the area accessible within 60 minutes by car (left) and public transport (right) from an arbitrarily selected suburban census tract (marked by a small red circle) in Buenos Aires. By overlaying the job census data with these maps, one can calculate the total number of jobs accessible in less than 60 minutes from the census tract marked by the circle. The number of jobs that can be reached in less than 60 minutes are 5.1 million jobs (95 percent of the total number of jobs in Buenos Aires) for workers commuting by individual cars and 0.7 million jobs (15 percent of the total number of jobs) for workers using public transport.12
Accessibility of a suburban location in Buenos Aires by public transport and by car. Source: Wb.BA.analyst.conveyal.com.
The difference in accessibility between the two different modes of transport is striking. However, it is not true that if every worker in Buenos Aires switched to cars, the average mobility would be increased. The commuting speed of the car users depends in large part on the number of cars on the road. An increase in car users would increase congestion and possibly cause gridlock, decreasing the speed and therefore the mobility of car users. I discuss below the necessary complementarity of various transport mode in improving mobility.
Additionally, there are two caveats for car commuting: first, the speed implied in figure 5.2 is an average speed and is not adjusted for different times of day; second, the availability of parking in different locations is not considered. In some areas, the scarcity or cost of parking might significantly decrease the practicality of commuting by car.
My purpose in showing the Buenos Aires accessibility map here is limited to providing a concrete example of the twin concepts of accessibility and mobility. The interactive map found on the website allows any Buenos Aires citizen to test its accuracy compared to their own experience. This can reduce the black box effect that habitually decreases the impact of sophisticated transport studies on urban policies.
Using this method, we could calculate a mobility index for an entire metropolitan area (equation 5.3). This global measure would be an indicator that planners should monitor regularly as a city develops.
Equation 5.3 City mobility index
After obtaining the job accessibility index of every census tract, we can calculate a city mobility index that represents the average job accessibility of all census tracts weighted by their population. The mobility index M expressed by the formula below shows the number of total jobs reachable within a commuting time T, for a given transport mode, for the average city resident,
where M is the mobility index, Ai is the number of jobs accessible from census tract i in less than T travel time, n is total number of census tracts, Pi is the active population in tract i, and P is the total active metropolitan population.
From an operational point of view, it is necessary to be able to measure mobility by location: how many jobs a worker can reach from a given location within a given time by different modes of transport. This type of data would show the most transport-deficient areas of a city. Various combined factors may explain the high unemployment rate in some urban neighborhoods. Indeed, an adequate transport system that gives easy access to jobs in the metropolitan region is often a prerequisite to decreasing local unemployment.
Measuring the Cost of Mobility
Mobility is a benefit provided by urban transport, but it has a cost. There is no point in advocating for increased mobility without also measuring the marginal cost associated with this increase. However, the economic costs of transport systems are particularly difficult to evaluate. Typically, urban commuters—whether they use public transport or individual vehicles—pay only a small fraction of the real cost of their trips.
Urban transport is different from other consumer products because its users pay only a part of its cost. Car users pay a market price for their car and the gasoline they consume (in most countries), but they are usually not paying for the public road space they use, or for the pollution, congestion, and other costs they have imposed on others. Users of publicly operated transport pay a fare that represents only a small part of the system’s operational and maintenance costs and usually pay nothing for the capital cost of the system. Obviously, car owners and public transport users eventually pay collectively all these costs through their taxes, but the cost they pay is not related to the quantity of the service they use. Because of the lack of real pricing, we can expect urban transport to be overused and undersupplied. Because of our inability to recover the cost of trips, mobility is significantly less than it could be. Therefore urban productivity could be greatly increased if we could price urban trips at their real costs.
Evaluating these costs is difficult, as many subsidies are not transparent. In addition, the cost of what economists call “negative externalities” (i.e., the cost imposed on others, like congestion and pollution) is not easy to evaluate. Since the 1980s, it has become clear that we should add the cost of global warming to the other traditional externalities. A worldwide price for carbon emission should reflect the cost of global warming caused by greenhouse gas (GHG) emissions. However, because of the worldwide failure to price carbon, pricing different modes of transport and comparing their price to their performance is even more difficult.
Faced with the difficulty in calculating the real cost of trips, many transport policy advocates renounce any attempt to make even an approximate calculation and just append the word “sustainable” to the mode of transport they favor. When comparing the cost and benefits of different modes of transportation, I will differentiate the transport costs that have a clear cash value (e.g., the cost of a car or of a subway ticket) from those like pollution or GHG emissions, which I will evaluate in units of gas emitted per vehicle/km or passenger/km without attempting to price it. In the same way, I will not attempt to give a cash value to the time spent commuting, but will just provide the average speed or time traveled. Transport economists attribute a dollar value to the time spent commuting based on the opportunity cost of the time of the person traveling. Using this convention, the cost of an hour of travel by a worker earning the minimum wage is lower than the cost of the same hour of travel by an executive paid a multiple of the minimum wage. Although this type of calculation is legitimate to calculate the aggregate economic cost of urban transport, it does not necessarily reflect how individuals select their choice of travel mode. Besides, the social cost of long commutes for low-income workers might be much higher than the one reflected by their hourly salary.
Mobility and Transport Modes
Classification of Urban Transport Modes
Until the middle of the Industrial Revolution—about 1860—walking was the dominant mode of urban transport. The area that workers could reach by walking less than an hour severely limited the expansion of cities. Because of the limitation on the speed of transport, urban labor markets grew primarily through the densification of the existing built-up area. Metropolitan Paris in 1800, before the Industrial Revolution,13 had a density evaluated by Angel at about 500 people per hectare compared to about 55 today. Since then, many mechanized modes of urban transport have allowed cities to grow geographically, densities to decrease, and labor markets to expand. These larger labor markets, in turn, allowed more labor specialization, which has increased the productivity of cities. Faster and better-performing urban transport modes are therefore a crucial element in the growth and prosperity of cities. In addition to increasing the size of the labor market, faster and more flexible transport modes allow urban land supply to expand and respond quickly to growing demand for new and better housing and new commercial areas.
Since the Industrial Revolution, many mechanical urban transport modes have been added to walking, among them cars, bicycles, motorcycles, buses, subways, tramways, and BRTs. Governments had an important role in allowing or funding the different modes of transport that were becoming available as technology changed.
The modes of urban mechanized transport that were already available at the beginning of the twentieth century have not changed much since then. The efficiency in using energy and the speed of cars, buses, and metros have certainly improved, but no new mode of urban transport has emerged. The invention of the BRT system in Curitiba, Brazil, in 1974 is only the application to buses of a technology applied to tramways at the end of the nineteenth century. However, it is quite possible that during the next 20 years, we will see the emergence of completely new modes of transport. The possibilities presented by the combination of vehicle sharing and autonomous vehicles could completely revolutionize urban transport as we know it today.
While no new mode of urban transport has emerged during the past 100 years, the dominant mode is often changing rapidly in emerging economies. The changes in mode reflect changes in income, city size, and the geographic coverage of public transport systems. Changes in dominant transport modes are instructive, as they reflect users’ choices and the way they adapt to the performance—speed, cost, and spatial coverage—of the various modes of transport available.
Individual Modes of Transport versus Public Transport
In a typical medium- or high-income city, commuters choose between a number of transport modes: walking, bicycling, driving, riding in a taxi, or using public transport. They select the mode of transport—or combination of modes—that is the most convenient for their trip, taking into account time of travel, direct cost, comfort, and whether their trip has to be chained with several activities (e.g., working, picking up children at school, shopping). When selecting their means of transport, travelers do not take into account the cost of the negative externalities they create—pollution, global warming, noise, and congestion.
Modes of transport are highly diverse, but they can be conveniently divided into three categories: individual transport, shared individual transport, and collective transport or public transport (figure 5.3). Individual transport and shared individual transport give access to the entire road network, while the various public transport modes are restricted to a network, which by necessity is a fraction of the entire road network. Because individual transport modes use the entire road network, they provide door-to-door travel without the need to change modes of transport on the way. Additionally, individual transport provides continuous 24/7 service, while public transport services are restricted to preset schedules with low frequencies outside peak hours.
Modes of urban transport.
Individual motorized modes of urban transport present many advantages over public transport, especially on-demand door-to-door service. Given these advantages over public transport, why did private firms and then the government provide public transport services?
Complementarity of Various Modes of Transport
In many cities, most modes of transport listed in figure 5.3 coexist. Some modes are heavily dominant, like the motorcycle in Hanoi, which represents 80 percent of commuting trips, or the car in US metropolitan areas (86 percent of all trips). However, in most cities, several transport modes coexist, and their relative share of total commuting trips varies with time. These variations reflect consumers’ choices, which respond to changing conditions in household income, urban structure, or transport mode performance.
Dominant Modes of Urban Transport May Change Rapidly
The shift in dominant modes of transport reflects an increase in population and household income. The share of passengers by transport mode reflects commuters’ preferences but also government action. This supply and demand tends to change rapidly in cities whose economies are growing fast; less so in cities where population and income are more stable. Figures 5.4 and 5.5 illustrate the rapid evolution of dominant modes of transport in cities like Beijing, Hanoi, and Mexico City compared to the relative stability found in Paris.
Changes in the dominant transport mode, Beijing (left) and Hanoi (right). Source: Beijing Transport Research Center, 2015.
Changes in the dominant transport mode, Mexico City (left) and Paris (right). Sources: Mexico City: “Gradual Takeover of Public Mass-Transit by Colectivos, 1986–2000,” Secretaria de Transito y Viabilidad (SETRAVI) Embarq—World Resources Institute; Paris: Syndicat des Transports d’Île-de-France website, www
Beijing’s transport mode underwent a radical change between 1986 and 2014. The bicycle was the main mode of transport in 1986, although the city population was already above 5 million. Many bicycle and public transport users shifted to private cars in the late 1990s. The shift from bicycle and public transport to car trips between 1994 and 2000 corresponds with the rapid increase in household income during this period of about 47 percent. It was certainly not driven by government policy. The massive investment in public transport—the tenfold increase in the distance covered by subway lines from 53 kilometers in 1990 to 527 kilometers in 2014—reversed the decline in public transport’s share of commuting trips. Car traffic congestion combined with a quota system for buying new cars stabilized the growth of car trips. Meanwhile, the share of bicycle users kept decreasing.
Hanoi’s transport transformation has been even more dramatic than Beijing’s. From 1995 to 2008, the share of bicycle trips dropped from 75 percent to barely 4 percent! But unlike Beijing, the motorcycle became the only dominant mode of transport in Hanoi, accounting for 80 percent of all trips (cars and public transport together account for only about 15 percent). As in Beijing, Hanoi’s commuters reacted to changing local conditions. Increased income allowed them to replace bicycles with motorcycles, significantly lowering commuting time—the average commuting time in Hanoi was 18 minutes in 2010. Large areas of Hanoi are accessible through narrow, winding roads that are nearly inaccessible to cars and even less accessible to buses, which were the only means of public transport in 2014. Motorcycles also gave easy access to residents of suburban former villages with only rural unpaved road access, expanding the supply of housing affordable to low-income migrants.
In Mexico City between 1986 and 2007, commuters have dramatically reduced their use of public transport in favor of minibuses and private cars despite strong municipal governmental policies to discourage these private alternatives. The change in dominant mode reflects rising incomes but also a change in the city structure of Mexico. Jobs have dispersed to suburban areas, in part due to government land use restrictions in the Federal District, and traditional public transport networks are less efficient for commuting from suburb to suburb. When jobs are dispersed, minibuses and cars become more convenient. However, the congestion created by cars and minibuses considerably slows down traffic in a city as dense as Mexico City (average density is about 100 people per hectare in the metropolitan area).
In contrast with these three cities, metropolitan Paris between 1976 and 2010 (shown on the right if figure 5.5) does not show any large shift in transport mode. Paris’s population and household income have been much more stable than those of Beijing, Hanoi, or Mexico City. The relative share of car and public transport trips reflects the structure of the city: a very dense core of about 2 million people and suburbs of 8 million. Commuters use public transport for most trips within and toward the core, but they use cars for the roughly 70 percent of commuting trips that originate and end in suburbs (reflecting the same share of job distribution). The extension of fast trains in the far suburbs of Paris has somewhat increased the share of public transport since the mid-1990s. However, car travel remains the dominant mode, reflecting the spatial structure of the city with a majority of population and jobs located in suburbs and, as a consequence, trips originating and ending in suburbs.
The change in transport mode in Beijing and Paris shows that increasing the size of the public transport network impacts commuters’ transport mode preferences. However, household income and a city’s spatial structure are the main determinants of commuter choice. For instance, in Beijing, multiplying the length of subway lines tenfold between 1990 and 2014 has only increased the share of public transport trips by 12 percent. And while Hanoi is building a new subway system that could eventually increase the very low share of public transport, subway trips are unlikely to compete with the speed and spatial coverage provided by the motorcycle.
The existence of various modes of transport reflects the choice of commuters. Commuters choose transport modes based on where they live, where they work, what time they go to work, what time they return home, and what share of their income they are willing to allocate to transport. No transport mode is perfect. Unsurprisingly, residents are often dissatisfied by urban transport. Car commuters complain about congestion and pollution, while public transport users complain about crowding, schedule irregularity, and lack of geographic coverage. In the following sections, I will analyze the pros and cons of the various transport modes accounting for their speed and the various negative externalities they create: congestion, pollution, and GHG emissions. However, we must remember that in the end, the primary objective is to increase mobility while decreasing the negative externalities imposed by that mobility.
Travel Time, Speed, and Travel Mode
The Measure of Travel Time to Work (Commuting)
As already mentioned, average commuting travel time is a common proxy used for measuring mobility. Average travel time becomes a meaningful proxy for mobility if it only includes trips to work and excludes other types of trips when calculating the average. Obviously, an average between travel time to work and travel time to go shopping or to the barbershop would have no meaning as a proxy measure for mobility.
The measurement of commuting travel time should be “door-to-door.” Travel time should include the time of travel from the moment the commuter leaves home to the moment she reaches her workplace. In addition, commuting time should be disaggregated by transport mode.
The examples of average public transport commuting time in the municipality of Paris and in Beijing’s metropolitan area illustrate the importance of door-to-door time measurement when assessing mobility (figure 5.6). The average door-to-door commuting time for trips using the subway in Paris municipality is 31 minutes, but the actual time spent on the train is only 15 minutes. The time required to go to the station and board the train, and then to walk from the station to the workplace, represents 52 percent of the door-to-door commuting time. For the longer trips in Beijing’s metropolitan area, the proportion of “access time” is lower and represents 36 percent of total commuting time.
Average door-to-door public transport travel time for commuters in the Paris municipality and Beijing metropolitan area. Sources: Data for Paris: “Etude sur les deplacements,” Regie Autonome des Transports Parisiens, 2014; Beijing: “Beijing, the 4th Comprehensive Transport Survey Summary Report,” Beijing Transportation Research Center (BTRC), Beijing Municipal Commission of Transport, Beijing, China, 2012.
We should do the same door-to-door calculation for car commuting trips, which typically start from one’s driveway in a suburban home but may end in a parking lot or underground garage, involving a sizable amount of walking time to get to the work place. I could not find statistics disaggregating travel time for car travel that include walking to and from a parking place. My own weekly car commuting from Glen Rock, New Jersey, to New York University in Greenwich Village in Manhattan takes on average 55 minutes of driving time and requires an additional 7.5 minutes walking from an underground parking garage to the university (figure 5.7). Access time is then only 12 percent of total commuting time.
Door-to-door commuting time from suburb to downtown New York (case study, no statistical significance).
Because the access time to transport is usually high, the speed of various modes of transport is a poor indicator of door-to-door commuting time. When trying to increase mobility, reducing access time to various modes of transport is as important as increasing the speed of the motorized part of transport. Table 5.1 shows the ratio between door-to-door speed and vehicle speed for Paris, Beijing, and the New York case study. As a city size increases, public transport networks become more complex and less dense, and they often require transfers between modes (e.g., buses to suburban trains). The increasing distance from home to stations and the necessity of transfers tend to increase access time. Car trips are less vulnerable to long access times, if an allocated parking lot exists at the destination. Trips by car from suburb to suburb have very little access time, because usually parking is available very close to the trip origin and destination. The lower value of suburban land explains why this availability is taken for granted.
|
Ratio between door-to-door speed and transport vehicle speed. |
||||||
|
Commuting mode |
Paris |
Beijing |
New York |
|||
|
Subway |
Bus and subway |
Car |
||||
|
Total average commuting distance (kilometers) |
9 |
19 |
38 |
|||
|
Door-to-door trip time (minutes) |
31 |
66 |
63 |
|||
|
Transport vehicle speed (km/h)a |
33 |
25 |
40 |
|||
|
Door-to-door speed (km/h) |
17 |
17 |
36 |
|||
|
Ratio of door-to-door speed to vehicle speed (percent) |
53 |
68 |
89 |
|||
|
a. For Beijing, this is an average speed for bus and subway. |
||||||
In the examples above, commuters in Paris and Beijing were walking to access the main motorized transport mode, but, of course, many other means of transport could be combined in a single commuting trip. A commuting case study in Gauteng, South Africa, describes one of the most complex and long commuting trips I have ever heard of.14 A single mother of four children commutes every weekday from her home in Tembisa, a township in Gauteng’s metropolitan area (which includes Johannesburg and Pretoria), to Brummeria, a business district of Pretoria, where she cleans offices. She leaves home at 5:00 A.M. to be at the office at 7:30 A.M. She starts her commute with a 2-kilometer walk to a collective taxi stand, where a taxi takes her to a train station. The train brings her to Pretoria, where she takes another collective taxi to a stop in Brummeria, from which she walks to her workplace (figure 5.8). The entire commute one way takes 2.5 hours, including walking and waiting for taxis and the train. Her commuting distance is 47 kilometers. Her average commuting speed is about 18 km/h, although most of the distance she covers is on a commuter train going at an average speed of 46 km/h. Because of the need to connect to the rail network to avoid the higher cost of the collective taxi, the distance she travels (47 kilometers) is much longer than the shorter road distance of 29 kilometers between her home and her workplace. If she had access to a motorcycle or even to a moped, she could commute in about 1 hour, instead of 2.5 hours. Access to a moped would allow her to gain 3 hours a day of disposable time!
Extreme commuting in Gauteng (South Africa) case study. Source: “National Development Plan Vision 2030,” President’s National Planning Commission, South Africa, 2011.
For a given home and job location, commuting time may show large variations depending on the main mode of transport, the number of transfers, and access time. In the case described here, a moped with a speed of 30 km/h would result in much higher mobility than using a suburban train with an average speed of 46 km/h.
Average Commuting Time by Transport Mode
Commuting by public transport takes longer on average than commuting by individual car. Given the amount of urban congestion plaguing most large cities in the world, this seems surprising. Urban congestion affects public buses as much as it does individual cars, but we would expect that public transport trips would be shorter in cities where many commuters are using underground public transport and dedicated bus lanes. Unfortunately, this is not the case. In a comprehensive and authoritative book,15 Robert Cervero, a fervent advocate of urban public transport, admits that faster travel time by car, even in public transport-based European and Japanese cities, is the main challenge in increasing the share of public transport over car trips all over the world. Let us try to understand why that is the case by looking at a sample of specific cities.
Commuting time in five large cities—Dallas–Fort Worth, Hong Kong, New York, Paris, and Singapore (figure 5.9)—confirms Cervero’s observation: commuting travel time by car is significantly shorter than by public transport in all these cities. The increase in travel time between public transport and car commuting ranges from 53 percent for New York to 100 percent in Singapore.
Average commuting travel time by transport mode, Singapore, Hong Kong, Dallas–Fort Worth, Paris, and New York. Sources: Data for United States: “Commuting in America 2013,” US DOT Census Transportation Planning Products Program, Washington, DC, 2013; Paris: “Les deplacements des Franciliens en 2001–2002,” Direction régionale de l’équipement d’Île-de-France, Paris, 2004; Hong Kong: “Travel Characteristics Survey-Final Report 2011,” Transport Department, Government of Hong Kong Special Administrative Region, Hong Kong, 2011; Singapore: “Singapore Land Transport Statistics in Brief 2010,” Land Transport Authority, Singapore Government, Singapore, 2010.
These differences measure average travel times for trips that have many different origins and destinations. The averages may mask a number of trips where the ratio between public transport and car travel time is reversed (i.e., trips that are shorter using public transport than using cars). For instance, some trips in Manhattan or in Paris Municipality are most certainly faster by public transport than by car. Suburban trips for people who live very close to a station and whose workplace is also close to a station might also be shorter using public transport than using a car. We can be confident that when this is the case, commuters choose the faster means of transport. However, average commuting time shows that in all these cities, commuters using cars spend less time commuting than those using public transport. Let us try to find out why it is so.
I did not randomly select the five cities in figure 5.9. Their characteristics are shown in table 5.2. Four of the selected cities have a significant share of public transport use, ranging from 26 percent for New York to 88 percent for Hong Kong. The fifth city, Dallas–Fort Worth, is an outlier with less than 2 percent of commuting trips using public transport. Hong Kong’s and Singapore’s public transport systems are relatively recent, and they benefit from their modernity and are known for their efficiency. The five cities selected show a large variety of densities. Hong Kong and Singapore have high densities, while New York and Paris have medium densities but high job and population densities in their core area, which favors public transport use and makes car use more difficult. Dallas–Fort Worth is the only city in the sample with a very low density (12 people per hectare) but with a population of 6.2 million, about equivalent to Hong Kong’s 6.8 million (2011). Given the very low density of Dallas–Fort Worth, car usage is predictably very high at 98 percent of commuting trips.
|
Density and share of public transport trips in five sample cities. |
||||||||
City |
Population (millions) |
Transit share of commuting trips (percent) |
Density (people per hectare) |
Built-up area (square kilometers) |
||||
|
Dallas–Fort Worth |
6.20 |
2 |
12 |
5,167 |
||||
|
New York Metropolitan Statistical Area |
20.30 |
26 |
18 |
11,278 |
||||
|
Paris (Ile De France) |
11.80 |
34 |
41 |
2,878 |
||||
|
Singapore |
5.60 |
52 |
109 |
514 |
||||
|
Hong Kong |
6.80 |
88 |
264 |
258 |
||||
|
Sources: Population: Census 2010. Density and built-up area: author’s measurements. Transit: Dallas–Fort Worth and New York, Summary of Travel Trends, 2009, National Household Travel Survey, U.S. Department of Transportation, Federal Highway Administration, Washington, DC; Paris, E 2008 Enquête Nationale Transports et Déplacements, Table 5.1. Commissariat général au Développement durable, Paris 2008; Singapore, Land Transport Authority Singapore Land Transport, Statistics in Brief 2010, Singapore Government, 2010; Hong Kong, “Travel Characteristics Survey-Final Report 2011,” Transport Department, Government of Hong Kong Special Administrative Region, Hong Kong, 2011. |
||||||||
However, one should not conclude that because current car trips are usually faster than public transport trips, a shift of public transport trip toward cars would decrease the average commuting time and therefore increase mobility. In the four cities with medium and high densities mentioned above, the current speed of cars depends on the proportion of commuters using public transport. Indeed, in Singapore, the government periodically adjusts the cost of using a car to decrease demand for car trips with the explicit goal of maintaining a minimum speed of car travel for those who can afford it. Most cities where public transport is an important mode of transport also try to control demand for car use, although in a manner less explicit and muscular than in Singapore. Reducing demand for car trips takes many forms. For instance, New York increases tolls for cars in bridges and tunnels, Paris reduces the number of car lanes, and Hong Kong increases taxes on car purchases. Beijing establishes yearly quotas for the purchase of new cars and uses a lottery to determine who may purchase a car. Stockholm, London, and Rome have a special charge to discourage car traffic in the core city. All cities heavily subsidize public transport operation cost to convince commuters to shift from cars to public transport because of the cost difference.
The higher speed of commuting cars in dense and moderately dense cities is due to the high number of trips using public transport. In these cities, the two modes, car and public transport, complement each other. In Dallas, by contrast, the short commuting time is entirely due to low density. As I will show later, low suburban densities provide larger road areas per households than in high-density cities. This large area of road per person allows higher speeds. As I mentioned at the beginning of this chapter, compact dense cities like Hong Kong and Singapore, while decreasing the average commuting distance, are often associated with a much longer commute than very-low-density cities like Dallas–Fort Worth. The potential advantages of shorter commuting trips in high-density cities are entirely offset by the slower speed caused by congestion, including congestion in transit.
Travel Cost
As mentioned above, I drive once a week from Glen Rock in suburban New Jersey to New York University in south Manhattan. The trip length one way is 37 kilometers. The cost of tolls amounts to $14, plus $17 for parking (of which 18.4 percent is a special municipal tax on parking), plus another $5 for about 2 gallons of gasoline, for a total of $36 for a return commuting trip by car (not counting insurance, maintenance, and capital cost). The commuting time door-to-door, one way, is about 63 minutes on average, corresponding to an average speed of 35 km/h.
The same return trip using public transport (bus plus metro) would cost only $14 but would require 102 minutes door-to-door one way, or an average speed of 22 km/h. In addition, outside peak hours, buses to and from Glen Rock leave only every hour. On a two-way commuting trip driving my car, I am spending an additional $22 to gain 78 minutes over public transport travel time, implying an opportunity cost of my time about $17 per hour. This personal case study has no statistical value, but it does explain the way many commuters select their transport mode. The transport costs that I pay for both public transport and car commuting do not reflect the real cost of providing the transport service that I am using—whether car or public transport. The fare of most public transport trips covers only a fraction of operating cost, and usually no capital cost at all. In the same way, tolls and gasoline costs may not reflect all the maintenance cost of the roads and traffic management service I use during my trip, and even less of the negative externalities on the environment and the congestion I impose on others by using my car.
So far, if we look only at the speed and duration of commuting trips, it seems that car trips have an advantage over public transport. Indeed, as jobs tend to disperse into suburbs and household income increases in many large cities of the world, it seems that the ratio of car trips over public transport trips is also increasing, to the alarm of transport planners. The congestion created by cars is a major concern. I alluded to this problem by warning that in denser parts of cities, the shorter commuting time made possible by traveling by car depended on the number of commuters using public transport. The larger the number of commuters using public transport, the higher the speed of commuters using cars will be. This trend explains the popular support for public transport investments in cities like Atlanta, where most commuters are using cars and intend to keep using cars in the future.
Speed, Congestion, and Mode of Transport
Road congestion is a real estate problem. Through regulations, planners or developers allocate portions of urban land to streets when the land is originally developed. Once a neighborhood is fully built, increasing the area allocated to streets is extremely costly financially and socially, as it requires decreasing the land allocated to uses that produce urban rents while increasing the area of street that produces no rents. It also requires the relocation of households and businesses.
In most cases, cars, buses, and trucks do not pay for the street space they consume; they have, therefore, no incentive to reduce their land consumption. The mismatch between the supply of land allocated to streets and the demand for street space creates congestion—too many users for too little street space.
Congestion decreases travel speed and therefore decreases mobility. In our quest to increase mobility, it is important to measure the street area consumed per passenger for each mode of urban transport and eventually to price it so that users who use large road areas would pay a higher price than those who use small road areas. Being able to price congestion in term of real estate rental value would enable us to increase mobility, not so much by increasing supply as by decreasing consumption. The objective remains to increase mobility by pricing congestion, not to select or “encourage” a preferred mode of transport.
In the next sections, I describe how to measure congestion and various attempts to increase road supply to manage demand.
Measuring Congestion
Congestion is the expression of a mismatch between supply and demand for street space. Traffic engineers define a road as congested when the speed of travel is lower than the free flow speed. The free flow speed of vehicles establishes the noncongestion speed, which traffic engineers use as a benchmark to measure congestion.16 Any speed below the free flow speed is indicative of congestion and is measured by the travel time index (TTI), which is the ratio of travel time in peak periods to travel time in free flow conditions. For instance, a car driving at 15 km/h on Fifth Avenue in New York at peak hours would indicate a TTI of 2.8, if we assume that the free flow speed in New York is equal to the maximum regulatory speed limit of 40 km/h. The mobility report published by Texas A&M Transportation Institute in 2012 evaluates the urban average TTI in 498 US urban areas at 1.18. Los Angeles, with 1.37, has the highest TTI among US cities. New York’s TTI is slightly lower at 1.33. The use of TTI allows us to measure the number of additional hours spent driving compared to what they would have been at free flow speed, and by extrapolation, the additional gasoline spent. From TTI, it is then possible to calculate the direct cost of congestion: the opportunity cost of the driver time plus the additional cost of gasoline compared to what it would have been under free flow conditions.
Using TTI to measure congestion is convenient, but is, of course, arbitrary. Starting November 1, 2014, New York City reduced its speed limit from 30 miles per hour (48 km/h) to 25 (40 km/h). The new regulatory limit is bound to reduce the free flow speed. If we take the new regulatory speed of 40 km/h as the free flow speed, then the TTI for a car running at 15 km/h has consequently decreased from 3.2 to 2.8 between October 31 and November 1. The reduction of the New York speed limit, aimed at reducing fatal car accidents involving pedestrians, obviously did not result in a reduction of average commuting time; it has even probably slightly increased it, in spite of the decrease in TTI implying the opposite. In the case of New York, the decrease in TTI in the fall of 2014 will be a false positive!
Using TTI to measure congestion is useful as a relative measure of mobility in a city (providing the benchmark free flow speed has not changed, of course, as it did in New York in 2014). It is also useful to identify streets where traffic management needs to be improved. However, TTI is not a good proxy for mobility when comparing cities. What is important for mobility is the changes in average travel time.
Passengers using motorbuses are also subjected to road congestion, although they are not the main cause of it, as they consume—at least at peak hours, when the bus is full—very little road space per passenger compared to drivers alone in their car, as we will see later. However, in addition to delays due to congestion, public transport users are also delayed when buses and trains are overcrowded and they are unable to board or when the schedule is unpredictable because of mismanagement or poor maintenance.
Public transport overcrowding is a form of congestion internal to the public transport system, as it does not affect commuters using other modes of transport. To my knowledge, the municipality of Beijing is the only one to monitor in real time public transport overcrowding, measured as a percentage of train capacity. Figure 5.10 shows that a significant portion of Beijing’s metro network is severely congested at peak hours. Trains and buses are assumed to exceed capacity when the density inside the vehicle exceeds 6.5 people per square meter! The discomforts caused by congestion are therefore quite different when alone listening to the radio in a passenger car sitting in traffic versus when sharing a square meter with six other persons in a crowded bus or metro car, or stuck in a subway station unable to board overcrowded cars!
Peak hour congestion in the Beijing metro network, 2014. Source: “Beijing the 4th Comprehensive Transport Survey Summary Report,” Beijing Transportation Research Center (BTRC), Beijing Municipal Commission of Transport, Beijing, China, 2015.
However, public transport congestion does not just result in discomfort for passengers; it also increases travel time and therefore reduces mobility. In Beijing, in spite of the spectacular increase in the length of metro lines built since 2000, reaching 523 kilometers in 2015, the congestion is so extreme during the rush hour that public transport employees have to limit the number of passengers who can board each train to prevent the train from becoming dangerously overcrowded. In 2015, about 64 Beijing metro stations (about 20 percent of total number of stations) had restrictions on boarding during rush hour. Beijing’s metro system had an additional 340 kilometers of lines under construction in 2015. It is hoped that these new lines will decrease public transport congestion.
The Supply Side: Increasing the Area Devoted to Transport
Increasing Road Supply in Already Dense Areas Is Nearly Impossible
Roads are the default transportation system in every city of the world. Roads are indispensable for the construction of a city. The road network is the backbone of any public transport system. Government is usually responsible for the design, construction, and maintenance of the primary road system.
Could the government act like an economic market and supply the road space that would match existing demand? The government or developers are routinely supplying new roads as cities expand into the countryside and in low-density suburban areas. However, attempts to increase the road supply in already dense built-up areas (where demand is highest) have had very limited success in the past.
Haussmann obviously managed to cut into dense existing neighborhoods in the Paris of the nineteenth century (see chapter 3). However, if Haussmann’s work is so well known in the history of planning, it is precisely because it has been nearly impossible to replicate it, and it had few precedents. Besides, all the boulevards created by Haussmann are now congested, and it is clearly impossible and even undesirable to widen them further to accommodate current traffic demand.
Demand for more road space usually occurs in areas located in the most attractive and expensive parts of a city. Because of the high value of the real estate bordering Fifth Avenue in New York, Rue de Rivoli in Paris, or Huaihai road in Shanghai, it is unthinkable to widen these streets, although they are extremely congested. The widening of these streets would destroy valuable real estate that makes them attractive, as well as increase the proportion of unpriced land—the roads—at the expense of high-priced land—the shops, offices, and residences bordering the roads. In addition, pedestrians are also major users of street area in central urban districts; increasing traffic flow in downtown areas is usually incompatible with safe and pleasant pedestrian traffic.
As an alternative to road widening, planners have often attempted to increase street area by building elevated highways above existing streets. Elevated highways, while not quite as destructive as widening existing streets, significantly decrease the value and livability of the neighborhoods they cross. In addition, getting in and out of an elevated highway requires the use of ramps, which involves the destruction of additional valuable real estate while obstructing pedestrian flows.
The plan proposed by Robert Moses for the Lower Manhattan Expressway in New York was an attempt to increase street supply in a high-demand area. The popular grassroots movement, led by Jane Jacobs, against the destruction that the expressway would have caused put a stop to the project. Indeed, the negative impact of elevated highways in dense urban areas is not limited to the eventual destruction of existing side buildings; it often extends for several blocks around. Because of the negative impact and high cost of elevated highways, not only has their construction been practically halted around the world, but a reverse movement advocating the demolition of existing ones is spreading.
Supply Side: Using Existing Road Space More Efficiently
Using existing street space more efficiently could increase the area of street available for the circulation of pedestrians, surface public transport, and cars. Many cities that allow extensive on-street parking in their dense core decrease the area available for the movement of pedestrians, bicycles, buses, and cars. On a typical street in Manhattan, parked cars are using 44 percent of the street area available to vehicles (not including sidewalks). Given the scarcity of road space, transferring all on-street parking to privately operated underground garages would greatly increase city mobility, the safety of pedestrians, and the pleasantness of a city in general. The political feasibility of doing so is remote, as many users of free or quasi-free on-street parking consider it as a basic human right. Any mayor attempting to improve mobility for pedestrians, surface public transport, and cars by removing on-street parking would probably not be reelected or might even be impeached.
Clever use of traffic engineering could also improve mobility without the need to increase the area devoted to streets. Samuel Staley and Adrian Moore, in their book aptly named Mobility First,17 devote an entire chapter titled “Seven Steps to Expanding Current Road Capacity” to the various methods that can improve the speed of vehicles in urban areas. These vary from redesigning intersections to introducing “hot lanes.” New technology in traffic light management could also improve mobility of the existing road network. Particularly promising is the swarm technology, which consists of real-time updating of traffic light patterns to respond to shifting traffic volume and unexpected events, like accidents or civil events. However, these measures, when taken, would undoubtedly enhance mobility but would not solve the problem of congestion durably without being associated with a demand side solution.
Supply Side: Tunneling to Increase the Street Area
The first underground railway dedicated to urban transport opened in London in 1863. It had a modest length of 6 kilometers and used steam locomotives. The cost was high, but it was an alternative to Haussmann’s widening of Paris streets, which had started a few years earlier. The more democratic and liberal nature of London’s political system in the middle of the nineteenth century would not have permitted an Haussmann-type operation in London. The decision to create an underground urban transport system in London was justified by the high price of land.
Building a transport system underground is a way of substituting capital for land. While the capital cost might be high, it should be approximately equal to the value of the road area it saves. The economics of building the first subway system in London must have seemed sound, as new subways were soon built in other capitals of Europe and in the main cities of the United States. By 1914, 13 cities18 worldwide had already built urban underground transport networks.
Building tunnels under the dense downtown area of existing cities is expensive. For example, in January 2017, three stations in a tunnel 2.7 kilometers long, part of the projected Second Avenue Subway, opened in midtown Manhattan at a cost of $4.45 billion (2015) or $1.6 billion per kilometer. It seems an astronomical sum for a relatively short addition to the New York subway network. Is it worth it? The object of the tunnel is to substitute capital for land, or in other words, to create new land by spending capital. To determine whether this cost makes sense, we could compare the cost of the land “created” to the price of land in areas adjacent to the tunnel.
A study of land value conducted by the Federal Reserve Bank of New York19 in 2008 valued the prime land in midtown Manhattan at about $5,500 per square foot or about $60,000 per square meter. If we calculate the cost of the new area created by the tunnel, assuming a right of way of 25 meters and three underground stations of 6,000 square meters each, we find that the cost per square meter of new land created by the tunnel is around $50,000. This is similar to the prime land price of $60,000 per square meter assessed in the area by the Fed report in 2008. In the context of Manhattan land prices, the $4.45 billion investment in a subway tunnel appears reasonable. In addition, it is likely that the new subway line, when completed, will increase the value of the land along Second Avenue compared to its 2008 value.
We should therefore always associate transport issues to real estate prices. High land prices indicate to planners that there is a high demand for the area, and therefore, a high volume of commuters will try to have access to it. At the same time, high land prices preclude widening streets to accommodate traffic. The justification for an underground network of transport depends on the ratio between the unit price of land and the unit cost of tunneling. If this ratio is about equal to or greater than 1, the underground network may make economic sense. If the ratio is much below 1, then another solution should be found.
Note that the construction cost of $1.6 billion per kilometer of subway is probably a world record. The cost of tunneling and constructing underground tracks and stations varies with many factors, including depth, width, geology, labor costs, and technology used. A brief survey of recent subway construction costs shows very large variations, from $600 million per kilometer for the latest Singapore MRT line, to $43 million per kilometer for Seoul line 9 built in 2009. These variations show that underground transport may make economic sense even in cities where land prices are much lower than in Manhattan.
Supply Side: How Much Land Is Available Anyway?
We may not be able to increase the area of street in already built-up cities, but do we even know how much is available to make better use of such a scarce resource?
Some cities’ land use statistics provide a percentage of the built-up area devoted to streets. For instance, the percentage of street area in New York City is 26.6 percent, for London it is 20.8 percent. These numbers are difficult to interpret. It would seem that New York City has about a 28 percent larger street area than London. While the two cities (defined as within municipal boundaries and not as metropolitan area) have about the same population of about 8 million, their population density differs significantly (table 5.3). Because of the difference in density, while the share of road is higher in New York, the area of road per capita is lower in New York than in London. If we assume that the need for street space is proportional to the population, then London offers 9 percent more road space per person than does New York, in spite of a significantly lower percentage of road.
|
Street area per person, New York and London. |
||||
New York |
London |
|||
|
Census year |
2010 |
2011 |
||
|
Population |
8,175,133 |
8,173,941 |
||
|
Built-up area (square kilometers) |
666 |
941 |
||
|
Population density (people per hectare) |
123 |
87 |
||
|
Percentage of street area |
26.6 |
20.8 |
||
|
Area of street per person (square meters) |
22 |
24 |
||
I am using this example to show that any normative approach to fix the area of urban road to an optimum to avoid congestion is a mirage. The population densities in New York and London have varied widely long after the road areas had been fixed forever by property lines. Densities vary with time, and road areas are fixed by history; as a result, the area of road per person, which is clearly correlated with congestion, varies over time. Transport systems have to adjust to what road area is available and not the opposite.
Population and job densities vary enormously from one part of a city to another; so does the percentage of land occupied by streets. The need for street area is related to neighborhood densities. For instance, in midtown Manhattan, the density of jobs reaches an astonishingly high 2,160 jobs per hectare. By contrast, in Glen Rock, a suburb of New York, the residential density is only 19 people per hectare. Clearly, the local demand for road space will be different in these two neighborhoods. If we combine neighborhood densities with percentage of land devoted to streets, we obtain the area of street per person or per job in different neighborhoods, which is already a little more useful than the aggregate at the city level. Figure 5.11 shows the variations of road space per person or per job in a select number of neighborhoods in various large cities of the world. Table 5.4 shows the infinite combination of densities and areas devoted to streets that explains the large variations in road area per person in each neighborhood.
Road area per person or per job in various neighborhoods.
|
Densities, percentage of road area, and road area per person in various neighborhoods. |
||||||
City—Neighborhood |
Density (people per hectare) |
Road area (percent) |
Road area per person (square meters) |
|||
|
Cairo—Al Mounira |
1,566 |
19.0 |
1.2 |
|||
|
New York—Midtown, jobs |
2,158 |
36.0 |
1.7 |
|||
|
Mumbai—Null Bazar |
1,649 |
27.9 |
1.7 |
|||
|
New York—Wall Street, jobs |
1,208 |
23.1 |
1.9 |
|||
|
Ahmedabad, India—Walled City |
588 |
16.0 |
2.7 |
|||
|
Hanoi—Dong Da |
929 |
27.3 |
2.9 |
|||
|
Tianjin—Heping |
271 |
24.0 |
8.9 |
|||
|
Mexico City—Tultepec |
121 |
11.9 |
9.8 |
|||
|
Paris—6 arrondissement |
266 |
28.8 |
10.8 |
|||
|
Ahmedabad, India—Vijaynagar |
492 |
54.0 |
11.0 |
|||
|
Cairo—Zamalek |
178 |
21.8 |
12.2 |
|||
|
Gauteng, South Africa—Sebokeng |
182 |
23.5 |
12.9 |
|||
|
Singapore—Toa Payoh |
186 |
26.9 |
14.4 |
|||
|
London—Bromley |
49 |
17.4 |
35.4 |
|||
|
Atlanta—Dogwood Drive, Fulton Co. |
22 |
12.2 |
56.3 |
|||
|
Los Angeles—Martinez Drive |
35 |
28.8 |
81.5 |
|||
|
New York MSA—Glen Rock, NJ |
19 |
15.7 |
84.8 |
|||
The road space per person can be below 2 square meters in some neighborhoods in Cairo, New York, and Mumbai, and above 50 square meters in some suburbs of Atlanta, Los Angeles, and New York. The road area per person could be an interesting indicator for mode of transport compatibility. In midtown New York, for instance, because of the very low area of street per person, workers would more than entirely fill the street with vehicles if each of them were using an automobile. At the other extreme, within the suburbs of New York, Atlanta, or Los Angeles, the road area per person is large enough to accommodate a large number of individual vehicles.
The numbers in figure 5.11 and table 5.4 demonstrate that any normative approach applied to densities or road design for an entire city in order to provide better transport is bound to fail. Planners should design transport systems that are adapted to the densities and street design of the neighborhood they are supposed to serve. In the following section, I evaluate the area of street that is consumed per passenger for various mode of transport.
Going beyond Supply
History and the experience of existing cities demonstrate that the possibility of increasing the supply of roads in existing built-up areas is extremely limited. Widening streets or doubling the street area by creating an elevated highway destroys the very quality that attracts traffic in the first place. Tunneling is very costly, and while indispensable in the downtown of large dense cities, it cannot be applied in cities where land is less expensive but where congestion is still very real.
Increasing mobility (i.e., decreasing the time required to go from one part of a metropolitan area to another) therefore requires a concerted action on the demand side. The demand side includes calculating how much street space each commuter consumes and what measures could be taken to decrease not demand for trips but demand for street space.
The Demand Side: Measuring Land Consumption per Commuter per Transport Mode
There are only two ways of decreasing the commuters’ demand for street space: reducing the consumption of street space per commuter and reducing the demand for trips at peak hours. Let us first look at the consumption of street space per commuter related to the various modes of transport.
Measuring the Consumption of Street Space per Commuter and Road Capacity in Passengers per Hour
A vehicle moving on a city street uses the area corresponding to the vehicle’s dimension plus the area required to prevent a collision with the preceding vehicle. The safe distance between two moving vehicles is set by the time that would be required for the following vehicle to stop if the front vehicle had to stop suddenly. The slow reaction time of drivers, not the size of the vehicles, is responsible for most of the road area required by moving cars. Therefore, the area of street required by vehicle depends on the speed of the vehicle: the higher the speed, the larger the area required will be.
The formal way to calculate the minimum safe distances between moving vehicles involves a number of parameters, including human reaction time, maximum braking deceleration, the adherence of the road surface, among other things. In reality, drivers are told that they should allow a reaction time of 2 seconds20 between two vehicles.
On congested streets, it is difficult to maintain a 2-second interval, as vehicles change speed continuously and average intervals tend to increase, further reducing the flow capacity of the road. In the rest of this chapter, I assume a standard safe interval of 2 seconds for moving vehicles except for buses moving in exclusive lanes, where the distances between buses will be fixed by the bus schedule.
When a car is moving at 40 km/h, the safety buffer zone required to maintain a 2-second interval represents 82 percent of the total street area consumed by the moving car. This area increases with speed, as shown in figure 5.12. Because cars move in a lane of standard width, and because most of the street area required by a car is dictated by the 2-second interval, smaller cars do not consume significantly less road space, except at very low speeds, as we will see below.
Street area requirements for cars running at different speeds with 2-second reaction time.
There are only two ways to decrease the street area consumed by moving cars: the first would be to decrease the width of vehicles so that two vehicles fit in the width of one lane (e.g., a motorcycle); the second would be to decrease safely the 2-second reaction time by using technology like self-driving cars. We will explore these possibilities later on.
For vehicles moving on a road, the consumption of street area per passenger is therefore dependent on four parameters: the length of the vehicle, the reaction time to ensure a safe distance between vehicles, the speed of the vehicle, and the number of passengers.
For instance, a commuter driving alone on a New York street at the maximum allowed speed of 25 miles per hour (40 km/h) is de facto required to use 84 square meters of street area per vehicle to stay at a safe distance from the preceding vehicle running at the same speed (figure 5.12). If the car’s driver is the only passenger, then the consumption of street space is 84 square meters per commuter. However, in the United States, on average, there are about 1.25 passengers per car in urban areas. An average commuter driving a car in the United States is therefore consuming 67 square meters when in a car running at 40 km/h. The more passengers per vehicle, the less the consumption of street space, as the distance between vehicles stays constant at the same speed.
The number of cars on a segment of road of a given length determines the speed of the vehicles because of the necessity of keeping a safe interval of about 2 seconds between vehicles. For instance, maintaining the maximum speed limit of 40 km/h allowed in New York City would require that there are no more than 38 cars per lane over a distance of 1 kilometer (left graph in Figure 5.13). If more cars enter the lane, the speed of all the cars will have to decrease to maintain the 2-second interval between cars. If the number of cars per kilometer of lane increases to 100, then the speed of all cars in the lane will drop to 11 km/h. This decrease in speed will also reduce the capacity of the lane to carry passengers. At the speed limit of 40 km/h, one lane can carry 1,900 passengers in one direction; at 10 km/h, this capacity is reduced to 1,360 passengers (right graph in figure 5.13). In reality, the capacity will probably be reduced further, because as new cars enter the lane, drivers have trouble adjusting quickly the interval between cars; it then becomes difficult to maintain a constant 2-second interval. Empirical data show that this interval tends to increase as speed adjusts downward, resulting in even lower passenger capacity than the one shown on figure 5.13, which assumes that the 2-second interval between cars remains constant.
Speed versus vehicle density (left) and lane capacity (right) for cars.
Therefore, in New York City for instance, every car added to traffic in a lane with a density of more than 38 cars per kilometer decreases the speed of all the cars in the lane. This additional car reduces travel speed, therefore increases commuting time, and therefore decreases mobility. In addition, it decreases the existing road capacity to bring commuters to their destination.
In spite of these well-documented problems, traffic planners have not yet found any direct way to control the number of cars entering a street in order to maintain a preselected speed. In other areas of the economy, “gatekeepers” are matching demand with existing capacity in the short term. This is what happens, for instance, in cinemas or restaurants where peak demand may often significantly exceed capacity. If more customers than seats available present themselves, gatekeepers will prevent the customers from entering the cinema or the restaurant. In the long run, prices will be adjusted to adjust demand to supply, or eventually extra capacity will be added to match supply to demand. This is unfortunately not possible for roads. If the road users’ demand is higher than the road capacity, in the absence of gatekeepers or price adjustments, all users will have to reduce speed until all demand has been satisfied, even if this demand results into total gridlock!
I discuss below the indirect means currently used to decrease the number of cars entering an urban network; except for adjustable congestion pricing as used in Singapore, most methods are often clumsy and ineffective.
So far, we have only looked at the links between vehicle density, speed, and lane capacity for cars, which could be taxis or private cars. Let us now compare these relations for various types of moving vehicles.
Table 5.5 shows the street area required per passenger at different speeds for motorcycles, ordinary cars, short cars (“smart cars”), and urban buses using equation 5.4.
Equation 5.4 Street area used per passenger
The street area used per passenger in moving vehicles can be expressed as a function of the vehicle speed S and four parameters: R, L, W, and P:
where
A is the area of street per passenger in square meters,
S is the speed of vehicle in km/h,
R is the reaction time required from drivers to prevent two vehicles from crashing into each other, expressed in seconds,
L is the length of the vehicle in meters,
W is the width of a lane in meters, and
P is the number of passengers per vehicle.
Predictably, people riding bicycles and motorcycles consume less street space than cars (I have assumed that three bicycles and two motorcycles are riding in parallel on a 3.2-meter-wide standard street lane). Smaller cars consume only slightly less than regular cars, and the difference decreases as speed increases. However, the surprising information provided by table 5.5 is how little street area passengers in a full city bus are using compared to any other mode of transport. At 40 km/h, a passenger in a car consumes more than 50 times the street area used by his counterparts riding at the same speed in an admittedly crowded bus!21
Street area consumption per person for various modes of transport operating at various speeds.
This is an argument often used by people who advocate eliminating cars from city streets. It is also the subject of a poster produced by the city of Munster, Germany, in 2001, which went viral on the Internet.22 The poster shows three side-by-side views of the same street with the area of street used by cars, bicycles, and buses to carry the same number of passengers. The goal of the poster is to dramatize the difference of road consumption per person between cars, bicycles, and motorbuses. The street area consumption figures in table 5.5 seem to support the poster’s claim that cars as a mean of transport are extremely wasteful compared to buses.
If it were possible to transport passengers in a bus at 40 km/h across a city using only 1.3 square meters of street per passenger, as implied by the numbers in table 5.5, it would probably justify banning all cars from cities. Unfortunately, both bus consumption numbers in the table and the claims of the Munster poster are grossly misleading.
Speed and Road Capacity
All types of vehicles listed in table 5.5 follow each other at a distance required by safety. However, buses cannot follow each other the way cars can. Buses have to keep a larger distance between each other than cars do, as we will see below. This is why the very low street area consumption shown on table 5.5 is theoretically correct but irrelevant for practical purposes.
Bus departures from bus stops have to be spaced at regular time intervals, called headways, which are much longer than the few seconds that are required between two successive cars. Typical headways in city centers range from 1 to 10 minutes, not the 2- to 3-second headways required for car safety. It is not possible to fill a city street entirely with buses running at 2-second intervals. For this reason, the very small street area that buses use per passenger is irrelevant, as buses using the same bus stop have to run at several hundred meters from each other, and the space between successive buses have to be filled with cars or be left empty.
Why do city buses have to be spaced at headways of often several minutes? The buses must stop at regular intervals to let passengers board and alight, which usually takes between 10 seconds to a minute for each stop. The time buses spend at bus stops is called the dwell time. The dwell time depends on the time required by passengers who are boarding and alighting. The more numerous the passengers boarding and alighting at a bus stop, the longer the dwell time will be. Therefore, dwell time is longer at rush hour, and it has an element of unpredictability, as it depends on the passengers’ agility when boarding and on the number of passengers. Bus headways must take into account the possible accidental variations of dwell time.
Imagine a column of buses following each other in a lane between two stops at a 2-second safety interval. The entire column will have to stop, say, 20 seconds to allow the first bus at the head of the column to let its passengers board and alight. The process will repeat itself for every bus in the column. Therefore, the speed of this imaginary column of buses will be, at best, the length of a bus every 20 seconds or about 2.2 km/h, about less than half of a pedestrian walking speed.
The departure of each bus must be timed in such a manner that there will be a sufficient time interval between buses to prevent following buses from bunching up at bus stops if the preceding bus is detained at the stop. To avoid this problem, the timing between buses is tightly scheduled. Setting headways is a major constraint in the operation of city buses in dense areas where bus stops are close to each other.
In New York, for instance, a typical city bus stops every 160 meters; express buses, operating only at rush hour, stop about every 550 meters and alternate with regular buses. The bus route M1 going from Harlem to Greenwich Village (running nearly the full length of Manhattan) has an average headway of 5 minutes during rush hour (between 7:00 and 9:30 A.M.), including both regular and express buses. The M1 route schedule indicates that the average rush hour speed varies between 9 and 12 km/h over the length of the line for both express and regular buses, express bus average speeds being closer to 12 km/h while regular bus speeds are about 9 km/h. The main advantage of running express buses is to be able to decrease headway. As these buses alternate with regular buses but bypass many stops and have longer runs between stops, the risk of their bunching at bus stops decreases.
Let us compare the speed and lane capacity for buses and cars on a segment of Fifth Avenue between 110th Street and 8th Street, at rush hour between 7:00 and 8:30 A.M. I will use the speed and lane capacity of the M1 route alone and that of the three other bus routes that use the same segment of Fifth Avenue as the M1 route. These other bus routes use different bus stops, which allows compressing headways without bunching. While the average headway of M1 is 5 minute at rush hour, the combined headway of the four bus routes using the same road segment is 1 minute and 48 seconds. Using different bus stops for different routes on the same street allows operators to increase headways and therefore the road capacity for public transport passengers.
Let us now look at the performance of buses and cars as shown in table 5.5 in terms of speed and capacity expressed as passengers per hour per lane (figure 5.14). The lane capacity for cars varies with speed, while the capacity for buses is independent of speed (horizontal lines on the graph) but depends only on headway. Obviously, lane capacity for cars also depends on distances between cars that vary with speed, while public transport headways are independent of speed.
Speed versus lane capacity for buses and cars along M1 corridor in Manhattan.
The speed and therefore the lane capacity of ordinary cars depends on the density of cars per kilometer, as we have seen in figure 5.13. If the cars’ density allows a speed of between 20 and 40 km/h, the lane capacity will vary from about 1,650 to 1,900 passengers per hour per lane. If the density is such that the speed drops below 5 km/h, then the capacity will drop rapidly to close to 0, when congestion creates gridlock.
By contrast, the number of passengers per hour carried by buses is independent of bus speed; it depends only on headway, which is designed to be constant (equation 5.5).
Equation 5.5 Lane carrying capacity
where
C is the lane capacity in passenger per hour per direction,
H is the headway expressed in minutes, and
P is the number of passengers per vehicle.
In New York, the M1 buses that depart at 5-minute average headways are therefore able to carry only 1,032 passengers per hour ((60/5) × 86), less than cars when they are running at speeds above 6 km/h (figure 5.14). However, when the four bus routes that are using this segment of Fifth Avenue are taken into account, the combined capacity of the buses reaches 2,800 passengers per hour.
In practice, the buses are practically using the right lane exclusively and the second lane partially to allow express buses to overtake regular buses. The other lanes cannot be used by buses and are therefore used by cars. The total capacity of the three usable lanes on Fifth Avenue when buses are combined with cars reaches about 5,000 passengers per hour per direction. This is more than the capacity that would be obtained if this segment of Fifth Avenue was used exclusively by cars or exclusively by buses. Therefore, as currently designed, cars might be clumsy and inefficient, but they are an indispensable component of urban transport systems. In addition, if the density of cars could be kept below, say, 50 cars per lane-kilometers, then they will provide a faster way of moving around the city than does public transport. However, the complementarity of the two means of transport is important. In dense urban areas like New York, London, or Shanghai, the large number of public transport passengers contributes to keeping the car density low, and therefore allows their passengers a higher mobility.
Let us now look at the comparative performance of all the various transport modes shown on table 5.4. We will look at speed and road capacity expressed as passengers per hour per lane (equation 5.5). The parameters used to draw the curves in figure 5.15, like reaction time, length of vehicle, lane width, and number of passenger per vehicles, are the same as those used in table 5.5.
Speed versus lane capacity for various transport modes along M1 corridor in Manhattan.
We see that bicycles provide a much higher road capacity at speeds below 15 km/h than any other mode of transport. This result is based on the assumption that bicycles are running on an entire lane dedicated to them, as is often still the case in Chinese cities but is not common in European and American cities. However, this performance in most cases remains theoretical, as demand for 5,000 bicyclists an hour along one road is unlikely in most large, dense cities. The more modest narrow bicycle lanes provided in many cities like New York or Paris have very low bicycle densities. However, figure 5.15 shows that in cities where traveling by bicycle is culturally well accepted (i.e., where there is a high demand for bicycle trips), bicycle lanes could contribute significantly to mobility. The low speed of the bicycle, however, prevents it from becoming a significant mode of commuting in cities of more than 1 million people. Because of their higher speed and increased comfort, electric bicycles, where they are authorized (as in Chengdu, China), could meaningfully compete with buses or cars as a means of commuting in larger cities.
Motorcycles also prove to be an interesting alternative to buses in terms of both speed and road capacity (figure 5.15). In countries where they are culturally well accepted, as in the countries of Southeast Asia, motorcycles could become a very efficient means of mass transportation on a metropolitan scale. Motorcycles use both energy and street space more efficiently than cars do. The noise and pollution associated with traditional motorcycles are soon becoming a problem of the past with the proliferation of electric motorcycles. The increased risk of accident associated with two-wheelers is real, but it is compounded by the neglect shown by urban managers in adapting traffic engineering and design to motorcycles in cities like Hanoi, where they are the major mode of transport. The emergence of more stable, fully enclosed tilting electric three-wheelers like the Toyota i-Road23 suggests that the mobility advantage enjoyed by the citizens of Hanoi could be provided in the future without the nuisance, risks, and discomfort caused by the traditional motorcycle.
By contrast, compared to the motorcycle, compact Smart car performance (except for energy use) in terms of road capacity is not much better than that of an ordinary car (figure 5.15). The width of the vehicle, not its length, is the important parameter to consider when trying to reduce street area consumption. A compact Smart car is 166 cm wide, while a Toyota i-Road vehicle is only 87 cm wide, allowing two vehicles to run in parallel on a standard 320-cm wide lane.
Figures 5.14 and 5.15 show us that speed and road capacity are equally important when assessing the performance of different transport modes. We will see below that the exclusive focus on road capacity, while disregarding speed, may lead to decreased mobility while apparently reducing congestion.
Demand Side: Allocating Street Area to a Preferred Mode of Transport
We have seen that the consumption of street area per commuter varies with transport mode. A passenger riding a bus filled at capacity running at a speed of 30 km/h consumes about 1 square meter of street space, while a passenger in a car running at the same speed with an average occupancy of 1.25 passengers per car consumes 55 square meters.
Given that congestion is a major problem in most large cities of the world, it is understandable that urban managers try to prioritize allocation of road space to transport modes that use less street space per passenger in order to increase street capacity. For instance, many municipalities reserve lanes for the exclusive use of buses and taxis. Many suburban highways have special lanes reserved for high-occupancy vehicles (HOVs) in an attempt to decrease the area of road used per commuter by encouraging carpooling. Commuters will use the HOV lanes when these are not congested and, therefore, they will be rewarded by higher speed. However, carpooling has been decreasing in the United States from 19.7 percent of all commuting trips in 1980 to only 9.7 percent in 2010.24 This trend can be explained by the higher commuting times involved in carpooling. In the United States, apparently, the potential reduction of travel time offered by HOV lanes is not a sufficient incentive to compensate for the additional time required to pick up carpooling passengers. Carpoolers will gain time by using HOV lanes only if they can maintain a speed close to free flow, say, 60 km/h. That supposes a distance between vehicles of about 33 meters or a density of about 30 cars per kilometer of lane. If there are more than 30 cars per kilometer of HOV lane, the speed of vehicles will decrease and possibly become equal to non-HOV lanes. If the car density is less than 30 cars per kilometer, then the speed in HOV lanes will be free flow and carpooling would become an attractive alternative to driving alone. However, if the density falls much below 30 cars per kilometer, then the highway capacity will be reduced and the HOV lane would increase congestion for the majority rather than reduce it.
High-occupancy toll (HOT) lanes are likely to replace HOV lanes in the future. HOT lanes are reserved for HOVs and cars that are willing to pay a toll to go faster. The definition of a HOV is more restrictive than in traditional HOV lanes, which ensures that the traffic in HOT lanes is fluid while the number of vehicles per hour is sufficient to ensure full use of the lane capacity.
HOV and HOT lanes are attempts by local governments to allocate scarce road area to the vehicles that use it more efficiently. We will see below how the use of congestion pricing and technology might one day be generalized to ensure an even more efficient use of urban roads.
The recent multiplication of BRT systems constitutes another way of allocating street area to the exclusive use of one mode of transport. Let us explore how successful this new mode of transport had been in increasing mobility.
Allocation of Roads for BRT Systems
The invention of Bus Rapid Transit in Curitiba, Brazil, in 1974 provides a more radical example of road allocation among transport modes. The first BRT was created in Curitiba as an affordable substitute to a subway. The challenge was to show that it was possible to increase the carrying capacity of a street lane to approach the capacity of a subway at a fraction of the cost. An ordinary bus line with a full load of 86 passengers per vehicle and a 3-minute headway is able to transport about 1,720 passengers per hour per direction (PPHPD). Curitiba BRT is able to carry about 10,800 passengers per hour per direction per lane. More recent BRTs, like Bogota TransMilenio, are able to carry up to 33,000 PPHPD by adding additional express lanes and wider rights-of-way than the one used in Curitiba. A typical subway carries from 22,000 (London Victoria line) to 80,000 PPHPD (Hong Kong Metro).
Since Curitiba’s success, many cities around the world have adopted similar BRT systems with some variations in design and performance. Mayors and urban planners have often celebrated the invention of BRT as a silver bullet that could solve the problem of mass transport in large cities without the heavy capital investment required for the construction of underground metro systems. BRT’s objective is not to add road area, as the BRT is usually created on existing roads, but to increase the capacity of existing roads by allocating existing road space to the most efficient users.
A BRT has two main features: an existing road area reserved to the exclusive use of buses and a system of specially designed bus stations where passengers pay their fares before entering the bus and where they can board and alight quickly, reducing dwell time to 20 seconds or less. The BRT vehicles are usually specially designed with many large doors to reduce dwell time and large passenger capacity: often up to 270 passengers in articulated buses. The high PPHPD capacity is made possible by running very large buses at very short intervals with headways of about 90 seconds.
Does BRT use street area more efficiently than ordinary buses?
In Curitiba, an elevated curb physically separates the BRT lanes from the lanes used by cars. Stations from which passengers board and alight are located about every 500 meters. Figure 5.16 shows the layout of a two-block portion of Curitiba’s BRT Eixo Sul line on Avenida Sete Setembre. The BRT lanes and station use 44 percent of the total road space exclusively for the use of BRT vehicles. In addition, to maintain speed and avoid traffic lights, crossings are minimized by interrupting about three cross-streets out of five.
Map of a typical station and right-of-way for the Curitiba BRT (Eixo Sul line).
A bus sharing a lane with cars uses only the street area that it needs to keep a safe distance from the vehicle in front of it. At 30 km/h, the passengers of ordinary buses use only about 1 square meter of road per person (see table 5.4). However, for a BRT, the street area used per commuter is different from that of an ordinary bus. A BRT uses its own road space, which is not available to other vehicles. Equation 5.6 defines the area of street used per passenger for a BRT.
Equation 5.6 Area of street used per passenger for a BRT
where
A is the area of street per passenger in square meters,
H is the headway in seconds,
S is the speed of the bus in km/h,
W is the average width of the BRT’s right-of-way, and
P is the number of passengers per bus.
The area of street consumed per BRT passenger varies with headway (figure 5.17). The larger the headway, the more street area a passenger consumes. To increase capacity, BRT operators have a strong incentive to reduce headway and to increase the number of passengers per bus. Increasing the number of passengers per bus has the added incentive of reducing operating costs, as the salary of drivers usually represents more than 60 percent of bus operating cost. This is why Curitiba’s BRT uses very large articulated buses, each with a capacity of 270 passengers.
BRT use of road space per passenger when headways vary. Dashed horizontal line indicates the road area per passenger for a car moving at 30 km/h with a headway of 2 seconds.
Headways, however, have a lower limit. When the headway gets too short, say, below 1 minute, buses run the risk of catching up with preceding buses, resulting in bunching at bus stations. Some Curitiba reports are mentioning headways of 90 seconds. This is extremely tight for articulated buses containing 270 passengers. Most common headways for BRT at rush hour seem to be about 2 minutes.25 The use of road space per passenger is dependent on headways and bus occupancies. I show in figure 5.17 how the road area used by passenger varies with headway and vehicle occupancy.
For a headway of 90 seconds and full bus occupancy, the street consumption per passenger is an impressively low 8 square meters (figure 5.17), about seven times less than the road space used by a commuter in a car running at 30 km/h. However, BRT street consumption per passenger increases rapidly as the headway increases and the occupancy decreases, as can be seen in figure 5.17. With a 6-minute headway and 60 percent occupancy, the road consumption of a passenger riding a BRT is similar to that of a commuter in a car. A 2004 report26 on Curitiba’s BRT analyzing data from various sources shows that headway on the north-south axis varies from 3 to 7 minutes. Checking the schedule of bus 507 at 7 P.M. in 2015 confirms a headway of 7 minutes. If these are the range of operational headways for BRT, the resulting street area consumption per passenger for the highest headway is 40 square meters at full occupancy and 66 square meters at 60 percent occupancy. This consumption is not any more impressive when compared to the 55 square meters consumed by a passenger in a car running at 30 km/h.
By contrast, at any speed and even at less than 10 percent occupancy, the road consumption per passenger of an ordinary bus remains a small fraction of the consumption of passengers in a car. To be efficient, BRT lines have to be built along an axis where the demand for capacity at peak hours ensures full occupancy and justifies the very short headways. It is often the case that after initial operation, headways are increased. Based on the basic statistics of bus movement, it appears that the actual headway at peak hours on the main BRT axes varies from 3 to 10 minutes.
BRT requires an existing rather wide right-of-way of at least 36 meters and preferably 42 meters for higher performance. However, because of the more efficient use of road space and their relatively low cost, BRT systems have spread rapidly all over the world in the past 20 years and, in particular, in cities that already had a subway system. Do they constitute an alternative breakthrough to the traditional bus and subway systems by providing mobility while avoiding congestion? BRT is an innovative way of using scarce road space, but its application will be limited in the future by three factors: first, the wide right-of-way it requires; second, the monocentric or linear city structure it implies; and third, its limited speed, which makes it inadequate for large cities.
BRT’s ability to reduce significantly the demand for street space is very much dependent on a continuous high operating performance: maintaining very short headways and high occupancy. Because of the typical distance between stations of about 500 meters in a BRT, maintaining headways as short as 90 seconds requires a faultless system. The dwell time spent by a bus at a station for letting passengers board and alight should not be above 20 seconds. Any delay occurring during boarding or alighting will cause buses to queue at the same station, as the distance between stations is too short to make up the dwelling time delay by increasing bus speed.
It is important to realize that the road capacity expressed in PPHPD, whether high or low, does not imply anything about speed. Figure 5.18 shows the PPHPD corresponding to eight BRT systems and eight different cities’ subway lines currently in operation and their respective speeds. These speeds are the average speed of buses and trains from beginning to end of the line; the vehicular speeds are much higher than the actual speeds of passengers when traveling from their origins to their destinations.
Transport system speed and capacity versus speed, BRT and Metro. Sources: BRT data are from Walter Hook, “Bus Rapid Transit: An International View,” Institute for Development & Transport Policy, New York, 2008; “Mass Transit Railway: Business Overview, Operations and Services,” www
While the capacity of the best-performing BRT may overlap with that of the worst-performing subway line, no BRT can match the speed of a subway line. Hong Kong’s subway has a capacity seven times larger than Curitiba’s BRT and a speed about 75 percent higher.
The limited speed of BRT systems (figure 5.18) suggests that there is a city size beyond which BRT is too slow to provide the mobility required in very large cities. For instance, Seoul’s metropolitan area extends across a circle of more than 100 kilometers in diameter. It is obvious that at a speed of 25 km/h, even the fastest BRT would not be able to provide access to the full job market of a large metropolitan area like Seoul. BRT, however, might be useful for providing high-capacity mobility in a restricted area (like a CBD) or for joining two dense job clusters. However, trips in large metropolitan areas would have to be provided by faster means of transport. The argument that high-capacity public transport reduces congestion is not convincing if the resulting trip duration is longer than what it would have been in a vehicle subject to congestion.
In addition, BRTs and subways are usually designed to serve radio-concentric trips, from periphery to CBD with high job concentration. However, in large metropolitan areas, employment is dispersing into suburbs. The high capacity at relatively slow speeds provided by BRT lines are not well adapted to the newly emerged spatial pattern of dispersed employment. The increase in the number of motorcycles and collective taxis in cities as diverse as Mexico and Johannesburg suggests that commuters are selecting the modes of transport that are more adapted to their trip patterns: low capacity on many diverse routes at higher speed. Unfortunately, most city managers do not accept motorcycles and collective taxis as legitimate means of transport and consequently do not provide road design and lane marking that would increase these vehicles’ efficiency and safety.
Decreasing Congestion by Managing Demand through Pricing
We have seen that, through design and regulations, local government may decide to allocate scarce street space to a preferred mode of transport. The allocation of street space to HOV lanes, exclusive bus lanes, and BRT corridors is motivated by a desire to use street area more efficiently. An administrative allocation of land may be justified in some cases, but it might often result in an inefficient use of street area when demand for road space shifts. For instance, HOV lanes are often either not used enough or are congested. The same problem may plague BRT lines that are justified only when the demand for capacity is high, when headways are below 2 minutes, and when vehicle occupancy or load factor is close to 100 percent design capacity. Administrative allocation of street space to a specific mode of transport may result in some positive outcomes in some circumstances, but it also introduces a rigidity when demand fluctuates at different hours of the day or over time. The rigidity introduced by an administrative allocation of street area to a preferred mode of transport may result in a loss of mobility.
In market economies, supply and demand is matched through pricing mechanisms. Would it be possible to have a pricing system that determines the optimum mix of different modes of transport through a pricing mechanism without relying on administrative allocation of street space to an exclusive mode of transport?
As I have suggested earlier, we should approach urban transport as a real estate problem. A municipality is the owner of the streets. The rent charged for using the street should be based on how large an area, how long, when, and where a commuter is using it. A commuter using an urban street should be submitted to the same type of pricing system as a traveler renting a room in a hotel or on an airline flight. The price to pay for a hotel room depends on its location, its size, the date of rental, and how long it is rented. Ideally, matching supply and demand perfectly would require that a similar rental system be applied to vehicles using urban roads. The charge, as it is practiced for hotel rooms, should be adjusted to maintain as close to full road occupancy as possible. In the case of urban roads, the objective of the congestion charge is not to maximize a city’s income but to prevent congestion above a set level. A car should therefore be charged for the increased travel time imposed on all other drivers due to its presence on the road.
Congestion pricing adjusts demand for roads in two ways: it discourages driving during peak hours for trips that could be done at other hours, and it encourages using vehicles more efficiently, either by increasing occupancy, sharing vehicles, or using less road-intensive means of transport (e.g., motorcycles or public transport).
Singapore is probably the only city in the world so far that is progressively getting toward this theoretical pricing ideal, although the practicality of converting this ideal into reality has not been fully achieved yet.
Singapore was the first city to implement congestion pricing, which it did in 1975. Initially, it was just a toll collected to enter the business district. Eventually, as new technology became available, electronic road pricing was introduced in 1998. The objective of road pricing in Singapore is to guarantee a minimum peak-hour speed for cars in the CBD, main arterial roads, and expressways. The toll is adjusted during the day for time and location. The monitoring of vehicle speed is done continuously, and toll rates are adjusted quarterly to maintain the minimum speed. In addition, measures have been taken to limit the number of cars on the island by auctioning periodically the right to buy new cars.
The effectiveness of Singapore’s congestion pricing policy is demonstrated by figure 5.19. Between 2005 and 2014, the average speed at peak hours has only varied between 61 and 64 km/h for expressways and between 27 and 29 km/h in the business district and arterial roads. During the same period, Singapore’s population increased by 31 percent!
Evolution of vehicular speeds at peak hours, Singapore. Source: Data are from “Singapore Land Transport Statistics, 2005–2014,” Land Transport Authority, Singapore, 2015.
The Singapore congestion pricing system is the closest in the world to the theoretical model I suggested above: charging a rent for roads the way one is charged for a hotel room (ideally, for a room rented by the hour of use). The toll is charged through gantry when entering different zones. However, the charge is not adjusted for the time passed in the area of high demand, although the rate is adjusted depending on the time of entry and the type of vehicle. I have no doubt that the Government of Singapore will keep improving its pricing system as new technology becomes available that allows charging by time of road use without high transaction costs.
Some cities like London and Stockholm are charging a fee for entering the downtown area, but this is a fee and not a rent, as it does not reflect the area consumed or the length of time it is consumed.
Charging for Road Maintenance and Capital Cost
The objective of the congestion fee is to prevent congestion, or rather to maintain commuters’ travel at a set speed during peak hours. However, there is another issue facing urban travel: users are not charged for all the costs involved in building and maintaining roads. This implicit subsidy given to road users may result in an overuse of roads and, therefore, an increase in congestion.
Typically, governments recover the cost of maintaining roads, if not building them, through a tax levied on gasoline price. However, many governments impose a gasoline tax that is much too low to cover the maintenance of urban roads. For instance, in the United States, the Federal gasoline tax is only 18.4 cents per gallon. The tax was last raised in 1993 and is not indexed to inflation. Each state adds its own tax to the Federal tax, but the amount varies widely even in adjacent states and is clearly more influenced by local political considerations than by a desire to rationalize transport economics.
In any case, the imposition of a tax on gasoline to recover from users the cost of roads is rather clumsy, even when the tax is indexed to inflation. Cars built more recently—including hybrid cars—consume much less gasoline per kilometer than older ones, paying even less tax per kilometer of road and further distorting the real cost of using roads. In addition, cars that are entirely electric will soon replace a large part of the current gasoline car fleet, making the cost of driving even more remote from its real cost. In addition, energy and, in particular, gasoline is often subsidized in many countries, further distorting the real cost of using roads.
Pricing urban transport closer to its real cost, including energy used, negative externalities generated (pollution, global warming, and accidents), and real estate utilized, would greatly increase the economic efficiency of urban transport. New GPS technology and transponders should allow users to pay the real cost of trips by distance traveled, including fixed costs like insurance. With a more transparent cost per kilometer traveled, trip decisions and transport mode choices might be different. This would include the choice between owning or sharing a car, or using public transport. The pricing of trips closer to their real economic cost may also alter urban land use in the long run. With a transparent real price for transport, it is urban users, in all their diversity, that will decide the proper mix of transport mode and the proper densities, rather than being submitted to clumsy top-down urban planning decisions. “Getting the prices right” is not just an economist’s theoretical dream. It could result in a more grassroots-driven urban environment, improved mobility with less congestion and pollution, and a better supply of affordable housing.
Street Space Allocated to Parking
In every city, a large part of the street area is not allocated to moving traffic but to on-street parking. The possibility of stopping curbside for loading or unloading people and goods is indispensable to the operation and maintenance of cities. Unloading goods to be sold in a store, food to be prepared in a restaurant, and materials to build or repair a building are necessary activities that keep a city functioning. However, it is absurd to use scarce street area to permanently park idle cars on congested streets.
On most Manhattan streets, 53 percent of the street area is allocated to vehicle use, of which 44 percent (usually two lanes) is reserved for on-street parking. Only a small portion of this area is subject to metered parking fees. In Washington, DC, residents of many neighborhoods have to pay only $25 annually to park their cars permanently on the street. A parking space in a private parking garage costs from $200 to $350 per month. Because cars parked permanently occupy most of the street curb space, the indispensable loading and unloading function requires double parking and reduces further the area devoted to traffic, causing more congestion.
Why should a municipality allocate so much scarce street space to permanent parking and subsidize its use? Originally, the high transaction cost of recovering parking fees discouraged pricing parking. However, new technology makes it much easier.
The difficulty has always been to differentiate loading/unloading from permanent parking. The solution, as for traffic congestion, is pricing. The parking fee for each neighborhood could be set so that 20 percent or even 75 percent of the curb space is always vacant. It is the same principle as congestion pricing.
New York implements metered pay parking reserved for commercial vehicles to load and unload in a few streets in Manhattan CBD. The parking time is limited to 3 hours per vehicle with a slightly increasing rate starting at $4 an hour for the first hour, increasing to $5 an hour for the third hour. While the rate progressivity is a step in the right direction, the rate is far too low to maintain enough empty space during the day. The high cost of enforcement—a municipal employee having to check time of parking printed on a ticket deposited on the vehicle dashboard—is unlikely to deter delinquency and abuse. High transaction costs to enforce payments based on parking meters prevent the policy from being effective. The free or underpriced use of public street space for parking will really disappear when technology can automatically identify the vehicle parked (as well as how long and at what time it was parked) and automatically debit the vehicle’s owner for the rent accrued—a system similar to the transponders already used on toll roads.
Concluding Thoughts on Pricing as Affected by Supply and Demand
Congestion is due to a mismatch between road space supply and demand.
Because increasing the supply of urban roads is expensive and difficult, the most efficient way to reduce congestion is to address the demand side. Charging commuters for the use of roads is the best way to adjust demand to supply and to reduce congestion. Tolls are increasingly used to reduce demand for urban road space. However, urban tolls are usually a fixed amount irrespective of how long street space is used. Tolls are therefore a clumsy way to charge for the temporary use of a good in short supply. Considering that urban roads are not a public good but are part of the real estate market, municipalities should charge a rent for their use. The rent charged should vary with the time of day, the location, the area, and length of time the road is used. The rent charged for roads should be similar to the fares charged by airlines to passengers or the room rates charged by hotels, except that the rate would not be for a fixed 24 hours but for the number of minutes the roads are actually used.
Up to now, the transaction cost of charging vehicles for this type of road rent would have been prohibitive. However, current technology could easily be used to charge road users a “road rent” that would reflect all the characteristics of real estate rents for short-term users. The effect of road rents on congestion would be immediate. Commuting trip departure would be spread more efficiently throughout the day or night. For instance, truck delivery would be strongly incentivized to be done at night. Smaller-footprint vehicles would be rewarded, thus decreasing congestion without decreasing the mobility of individual commuters. Charging road rents would have also a beneficial effect on land use. Many activities that generate a lot of vehicular traffic would have an immediate strong incentive to locate in areas where demand for road is low compared to their supply (i.e., suburban areas). Carpooling achieved by electronically matching similar itineraries would also reduce demand for road space while not reducing the mobility of commuters.
Mobility, Pollution, and GHG Emissions Due to Transport
Mobility, Pollution, and GHG Emissions
Mobility consumes energy. Since the Industrial Revolution, energy has been cheap, and its source has come mostly from fossil fuels. Consequently, urban mobility has been a major source of pollution and GHGs. Urban planners, alarmed by pollution caused by transport, are advocating for a reduction of the footprint of cities (increasing the price of housing) instead of concentrating on transport technology that would reduce the pollution created by transport. Constraining land use is the wrong strategy; promoting pollution free transport is the right strategy, and it is now achievable. Because of technology changes, increasing mobility by allowing longer and faster trips is not necessarily the equivalent of more pollution and GHG emissions. It is possible to increase mobility by increasing a trip’s speed and length while both reducing the energy used per passenger and reducing pollution and GHGs per unit of energy used, as I show below.
Pollution and GHG Emissions Are Two Different Problems
Concerns for pollution and GHG emissions are often lumped together under the “sustainability” agenda. In reality, pollution and GHG emissions are two very different issues that require different approaches. Pollution due to transport causes more harm when it is concentrated in central urban areas. The same quantity of pollutant creates very little damage when dispersed over a large area, but it can be lethal when concentrated in a densely populated urban area. In addition, some tailpipe pollutants, like carbon monoxide, are not stable in the long run and are soon changed into innocuous gases. By contrast, GHGs—mainly carbon dioxide—are not dangerous at the location of emission, even if concentrated; however, carbon dioxide is extremely stable and accumulates in the atmosphere. The dangers and associated costs posed by transport pollutants and GHGs are therefore completely different and should be addressed separately.
Urban Mobility and Pollution
Urban transport is responsible for a large part of urban pollution emissions. However, the effect of these pollutants on human health varies widely depending on concentration. Concentration is dependent on three factors: the concentration of vehicles in one location, the rate of emissions of individual vehicles, and a city’s topography and climate. Cities like Los Angeles, Delhi, Beijing, and Paris could have surges of very dangerous levels of pollution when wind and temperature combine to prevent pollutant dispersion. Therefore, the impact of urban pollution on health could be very different among cities whose vehicles are meeting the same emission standards. Measures to curb pollution should therefore be adjusted for each city depending on its climate and topography.
Trends in Pollution Due to Gasoline Vehicles over the Past 20 Years
There is a clear consensus on the necessity of curbing pollution emissions from urban transport vehicles. Ideally, charging for pollution measured at the tailpipe would eventually reduce pollution to an acceptable level or even to zero. So far, the technology to do so is not yet available. The heterogeneity of urban transport vehicles using different types of engines and different fuel qualities makes direct as-you-go measurement difficult.
Governments in North America, Europe, and Japan have established mandatory maximum pollution standards that they impose on new cars. In the fall of 2015, the scandal caused by Volkswagen’s deliberate evasion of new car emission testing shows that government standards are not foolproof, and more progress must be made not only in setting standards but in enforcing them. However, the trend in pollution reduction over the past 30 years has been conclusive in the more affluent countries that maintain maximum pollution standards, in spite of the enforcement shortcomings.
In the United States, the Environmental Protection Agency (EPA) describes the evolution of pollution due to cars between 1970 and 2004 as follows:
The Clean Air Act required EPA to issue a series of rules to reduce pollution from vehicle exhaust, refueling emissions and evaporating gasoline. As a result, emissions from a new car purchased today are well over 90 percent cleaner than a new vehicle purchased in 1970. This applies to SUVs and pickup trucks, as well. Beginning in 2004, all new passenger vehicles—including SUVs, minivans, vans and pick-up trucks—must meet more stringent tailpipe emission standards.27
Figure 5.20 shows the changes in pollution in Germany from new gasoline nondiesel cars over several decades. The changes show that government mandates, while imperfect tools, are still effective at triggering the technological changes necessary to reduce pollution that markets, in the absence of pricing mechanisms, have been unable to provide.
Changes in gasoline-car pollutant emissions in Germany, 1990–2013. Sources: Table 2: Fuel-specific IEF for passenger cars, in Fkg/TJ, Ministry of the Environment, Federal Environment Agency (Umweltbundesamt, UBA), Dessau-Roßlau, Germany.
The decrease in pollution emissions shown in figure 5.20 is due to a combination of technological change allowing a decrease in gasoline used per kilometer and changes in engine and exhaust treatment that reduce tailpipe pollutants.
The data presented in the figure concerns only gasoline vehicles. In the past few years, hybrid cars and electric cars have been manufactured commercially, although they still represent a very small portion of the overall urban transport vehicle fleet. In addition, hydrogen fuel cell vehicles have moved beyond their experimental phase and are slowly appearing in selected cities.28
Electric cars and hydrogen fuel cell vehicles, when they eventually become a sizable part of the urban transport fleet, will completely change the environmental quality of cities. The pollution emission will not occur at the multiple tailpipes of vehicles but at the electricity source. This would not necessarily mean an absence of pollution, but at least it would avoid the pollution concentration in urban areas and will liberate cities in unfavorable climates and topographies from the peak of pollution that currently affects them. The enforcement of pollution standards at electricity sources would be much easier than the current system obliging environmental agencies to test periodically millions of vehicles. This would also greatly improve the possibility of emission control in less-affluent cities that cannot afford periodic pollution control of urban vehicles. Finally, the pricing of pollution at the source of electricity generation would have a better chance of being implemented without large transaction costs or political resistance.
Urban Mobility and GHG Emissions
Gas-Powered Vehicles
As we have seen, pollution due to transport is difficult to measure and price at its real cost. The measurement of GHG emissions poses an even greater challenge.
The simplest measure consists of measuring tailpipe carbon emissions, also called tank-to-wheel emissions. This is rather simple, as every liter of gasoline burned in an engine releases about 2.3 kg of CO2. Therefore, the lower the gasoline consumption, the lower the CO2 contribution will be to global warming. The GHG measurement appears to be simpler than for pollution, which depends not only on gasoline consumption but also on the design of the engine and of the exhaust system. However, this simplicity is only apparent. Many argue, with reason, that what we should measure is the well-to-wheel carbon emission: the carbon emitted when extracting, refining, and transporting 1 liter of gasoline to the tank of a car. This well-to-wheel measurement adds about 10 percent more CO2 to the tank-to-wheel emission, increasing GHGs emissions to about 2.73 kilograms per liter of gasoline. Finally, to be more accurate when measuring the GHG emissions due to transport, it might be legitimate to include the GHG emissions produced when manufacturing the vehicle, maintaining and recycling it, building and maintaining roads, and so forth. This “life span” emission calculation might be tempting to tackle, but it is probably counterproductive, because the complexity of the calculation and its many attendant assumptions might require the development of an entire academic field. Should the GHG emitted by the workers commuting from home to the factory manufacturing the vehicle be included in the final calculation of the CO2 emitted by a liter of gasoline burned in an internal combustion engine? I can imagine every country’s department of transportation becoming a gigantic accounting office, similar to the Soviet Union’s Gosplan, preparing gigantic input-output tables to calculate ever more accurate GHG numbers emitted by a single liter of gasoline. In the following paragraphs, I will use the well-to-wheel emission figure for gasoline-powered vehicles.
We should not forget that a market economy avoids the Kafkaesque complexity of the Soviet Union’s Gosplan by simply using prices to transmit information through the entire economy. If our governments could agree to put a price on carbon, there would be no need to calculate a vehicle life cycle emissions of GHGs. Emissions would decrease in all industries in proportion to the cost of reducing them. The carbon price would stimulate the creation of new technology that would reduce emission across industries.
In the absence of carbon pricing, the second-best solution is for the government to impose industrywide standards. In the United States, the National Highway Traffic Safety Administration and the EPA issued jointly a new national program to regulate fuel economy and GHG emissions for cars built between 2012 and 2016. The European Union, Japan, and Korea are also issuing their own annual mandatory CO2-equivalent (CO2-e)29 maximum emission standards for new cars. The changes in emission standards for new cars since 2000 are shown in figure 5.21.
Average mandatory CO2-e emission standards in various countries, 2000–2020. Sources: Data for Europe, Korea, and Japan: “Monitoring CO2 Emissions from New Passenger Cars and Vans in 2015,” EEA Report No 27/2016, European Environment Agency, Copenhagen, Publications Office of the European Union, Luxemburg, 2016; United States: “Light-Duty Vehicle Greenhouse Gas Emission Standards and Corporate Average Fuel Economy Standards; Final Rule,” U.S. Environmental Protection Agency (EPA) and the Department of Transportation’s National Highway Traffic Safety Administration (NHTSA), Federal Register, Vol. 75, No. 88, May 7, 2010, Rules and Regulations.
The emission standards reflect only new cars, not the entire fleet. However, they anticipate the average emissions of the entire national fleet in the near future. The actual GHG emissions from 2000 to 2014 in the EU have decreased by 23 percent, and if the target is met in 2020, it would represent a decrease of 44 percent. It shows that in the absence of more efficient price signals, mandatory standards are effective in decreasing GHGs.
Hybrid, Electric, and Fuel Cell Vehicles
The standards shown in figure 5.21 concern only gasoline, hybrids, and diesel cars. Increasingly, electricity or hydrogen fuel cells are likely to power an increasing share of the vehicles used for commuting in urban areas. Currently, the market share of electric cars is very small. In 2014, San Francisco, with 5.5 percent of all vehicles, had the largest market share of electric and hybrid cars among major US cities. However, given the R&D investment being poured into electric cars and batteries, it is likely that the market share of electric vehicles will eventually dominate the urban fleet.
The drop in the price of oil in 2016, reflecting an oversupply, might be a damper on the development of electric cars. However, electric technology presents many advantages for urban transport, in particular the absence of noise and pollution. This superior technology will eventually prevail on its own in the future. As a Saudi oil minister once remarked at an OPEC meeting, “The end of the Stone Age was not caused by a shortage of stones!”
For electric vehicles, the power plant generators alimenting the electrical grid will then produce the GHGs, not the car engine itself. Concerns for GHG emissions would then shift to the source of electric power generation and away from car manufacturers.
Currently, there is a wide difference in GHGs emission in various electrical grids, depending on the source of energy fueling the generators (figure 5.22). The low emissions from Swedish and French grids are explained by a combination of nuclear and hydroelectric generation, while the high emissions of the Polish and US grids stem from the use of coal as a fuel in some generators. However, the emissions from the California grid are nearly half those of the US average! The regional differences in emissions in the US grid are also explained by the differences in fuels used for electricity generation: California has a high proportion of hydroelectricity and nuclear plants, while in Michigan generation plants the dominant production fuels are coal and crude oil.
Emissions of CO2-e per kilowatt-hour for the electrical grids of selected countries. Sources: United States: United States Environmental Protection Agency, 2018, eGRID Summary Tables 2016, https://
Anybody concerned with GHG emissions should certainly switch to electric cars in Sweden, France, and California, but should use gasoline when driving in Michigan or Poland!
CO2 Emissions by Various Modes of Transport
Let us now compare the GHG emissions from urban transport vehicles in grams per passenger-kilometer (g/pkm), depending on the brand and the technology used (figure 5.23). For gasoline and hybrid cars, the amount of CO2 equivalent is calculated at the exhaust pipe. The real amount of GHG emission per vehicle is in reality higher than the tailpipe emission, as additional GHG-emitting energy is required to produce and transport the volume of gasoline consumed by the vehicle engine itself. For electric cars that get their energy solely from the electric grid, I have used the average GHG emissions in grams by kilowatt-hour of the electric grid available in the country where the car operates. However, the GHG emissions of the electric grid in various countries is calculated from well-to-wheel (i.e., the emission figures are taking into account the GHG emitted during extraction and transport of the source of energy of each national or regional grid). I have selected the Nissan Leaf as a typical example of electric car among the several models currently available on the market. This GHG emission per kilowatt-hour is then multiplied by the average number of kilowatt-hours required to transport one passenger for 1 kilometer. The large variations in GHG emissions shown in figure 5.23 for electric cars with the same kilowatt-hour consumption reflects the different sources of energy used to produce electricity in their respective countries, as was shown in figure 5.22.
Tailpipe emissions of CO2 per passenger-kilometer for various cars and public transport modes, 2015.
Sources: United States: United States Environmental Protection Agency, 2018, eGRID Summary Tables 2016, https://
Figure 5.23 includes data on the emissions in New York, the most widely used public transport system in the United States. Measured in grams of CO2-e by passenger-kilometer, city buses emit about three times more GHGs than do subways. The explanation is simple: subway lines follow the routes with the highest demand. Many buses are feeders to subway stations and often serve suburban areas with less demand. To maintain patronage for public transport, buses have to run nearly empty outside peak hours. In addition, because drivers’ salaries are one of the major operating costs for a bus service, the size of buses has been increased to carry the maximum number of passengers per driver. This is financially efficient during peak hours but is energy inefficient when the demand is low (i.e., when very large buses carry few passengers outside peak hours). Bus stops are typically located every 150 meters, obliging buses to accelerate and brake at full load very often. This contributes to more energy use, and therefore more emissions. By contrast, New York subways are connected to the New York grid, which according to EPA, emits 411 grams of CO2-e per kilowatt-hour—a rather low emission rate compared to the average US grid.
The bottom of figure 5.23 shows the CO2-e in g/pkm emissions for the most common cars in the US fleet. The trip of a commuter riding in an average US car is responsible for more than twice the CO2-e g/pkm emitted by a trip of the same length by a commuter riding a subway. However, this car trip produces significantly less CO2-e than the same trip by a commuter using a city bus. Commuting trips using hybrid cars, still a very small part of the total number of urban trips, produce about the same CO2-e as the same trip riding a subway. Finally, trips from commuters in electric cars in some Western Europe countries and in California emit about half the CO2-e of a passenger in a New York subway, while the CO2-e emissions from electric cars in Sweden are practically insignificant. Actually, the 5 g/pkm CO2 emission of a Nissan Leaf being driven in Sweden will be less than the 6 grams per kilometer exhaled by a person weighing 70 kilograms and walking at 4.7 km/h!30
The point of figure 5.23 is not to advocate that commuters shift from city buses to Nissan Leaf cars but to show that we should not exclude individual vehicles as a potential transport mode in modern cities because of a concern about global warming. As we have seen, mobility in cities would increase if some trips were made by individual vehicles, possibly shared. Because of the rapidly changing technology, the inclusion of redesigned, possibly shared, individual vehicles as a major transport mode might decrease GHG emissions compared to current modes of transport, whether public transport or individual traditional gasoline cars.
Trends in energy use per passenger-kilometer by mode of transport (figure 5.24) in the past few years confirm the results of figure 5.23. The energy used by commuter rail (including subway and suburban rail) has remained roughly constant and is the most energy efficient compared to traditional gasoline cars and buses. The energy efficiency of public transport (buses and rail) depends in great part on the load factor (the number of passenger per vehicle), which can vary a lot as a city’s spatial structure changes and as incomes increase. By contrast, the load factor of cars used for commuting remains roughly constant (about 1.3 passengers per car in urban areas). Technological improvements stimulated by government fuel economy mandates are responsible for the decrease in the energy per passenger-kilometer in individual cars over time. The large increase in energy per passenger used by city buses is probably due to the two factors discussed earlier in this subsection: the extension of bus services into low-density suburbs and the increase in the size of city buses.
Changes in energy use per passenger-kilometer for different modes of transport, 1970–2012. Source: US Department of Energy, Transportation Energy Data Book, 33rd ed., Washington, DC, 2014.
Finally, the newest cars like the Toyota Prius hybrid and the “plug in” all-electric Nissan Leaf are far more energy efficient than all prior modes of transport, whether public transport or gasoline cars. However, these cars are still using scarce road space inefficiently. While they could be a better solution for low-density suburb-to-suburb trips, they would cause as much congestion in dense city core areas as do traditional cars. In addition, at the moment, these cars represent an insignificant part of the urban car fleet.
Mobility in Evolving Metropolitan Urban Structures
The Three Plagues of Current Transport Modes
Current urban transport systems inflict three major plagues on the cities they serve: congestion, heavy pollution concentration, and high GHG emissions. I have discussed these different aspects in the preceding sections of this chapter. Here I briefly summarize the contribution of each transport mode to the “three plagues.”
• Individual cars. Their major problem is congestion. The valuable real estate that a car, whether running or parked, occupies in the densest part of a city is extremely costly and is usually unpriced. The concentration of pollution is still a serious problem but could be solved in the long run by a change in technology that is clearly emerging. Cars still contribute a sizable part of GHG emissions. In spite of their shortcomings, they are the fastest mode of transport for suburb-to-suburb trips and therefore are likely to remain an important mode of urban transport in spite of their shortcomings.
• Motorcycles. Their use of street real estate is efficient, but they are noisy, polluting, and dangerous. Municipalities do not take them seriously as a means of transport, and consequently these vehicles do not benefit from elementary traffic management measures like special lane marking. They are fast and efficient for providing access to suburbs where the municipality has not yet provided proper roads. They are likely to become a major mode of transport in lower-middle-income cities. The shift from gasoline to electricity or fuel cells would eventually solve the noise, pollution, and global warming problems that these vehicles contribute to.
• Collective taxis and rickshaws. They usually are a major source of pollution and congestion. They provide route flexibility and are often the only means of transport affordable to lower income workers, but they are difficult to manage, their drivers often compete violently for the same routes, and the municipalities are usually more interested in suppressing these vehicles than in managing them efficiently. They have the advantage of adapting rapidly to changing demand in newly developed areas.
• City buses. They use scarce road space efficiently, have flexible routes, and are efficient for short distances. However, they are too slow for long distances in large cities, because they have to stop too often, and they are energy inefficient because of the low load factors outside peak hours. They are inefficient for suburb-to-suburb routes.
• Bus rapid transit (BRT). These systems are higher capacity than buses, but too slow for large cities and long commuting distances. They are not easily adaptable to changing routes, because they require the use of already existing wide right-of-ways. They use too much real estate per passenger outside peak hours. BRTs considerably slow down freight and other traffic because of the dedicated lanes they occupy. And they are not useful for low-volume suburb-to-suburb trips.
• Subways and suburban rail. These systems do not contribute to the pollution of cities. Their contribution to GHG emissions depends on the electrical grid efficiency. They are space efficient in the core of dense cities like New York, London, and Seoul. They are efficient modes of transport for trips from suburb to core city but are subject to heavy congestion in very dense cities like Beijing or Shanghai. They are inefficient for suburb-to-suburb commuter trips. High capital costs limit their extension in low-income cities.
Expanding Current Urban Transport Modes into Ever-Larger Cities Will Not Work
Currently, commuters choose between some mode of public transport and individual vehicles. Many public transport users use two modes, rail and feeder buses, resulting in long commutes because of the long transfer times between modes. Few commuters combine cars and public transport because of the high cost or unavailability of parking around stations. Most metropolitan transport policies consist of trying to increase the number of public transport commuters and decrease the number of car commuters, even in cities where public transport is heavily congested, as in Mumbai. However, in most cities, public transport users’ travel times are always longer than that of car users. It seems that the purpose of many public-transport-focused policies is not to decrease overall travel time but to decrease car travel time for those who can still afford it, as observed by David Levinson in his article aptly titled “Who Benefits from Other People’s Transit Use?”31
The objective of an urban transport policy should be to increase for everybody the number of jobs and amenities accessible in less than 1 hour, not to decrease travel time for those who already have the shortest travel times (car users). The new technologies that have emerged in the past 20 years should allow radical changes in urban transport modes, which have not changed in the previous 100 years! Tentatively, let us now explore (1) the changing special structure of large cities, (2) how transport must adapt, and (3) the role of new demand-driven technologies in driving change.
The Spatial Structures of Large Metropolitan Areas Are Changing
The spatial structures of cities are changing in most of the world. Large lower-density suburbs are developing around the traditional urban core and CBDs. In Asia, where urbanization is still low compared to the rest of the world (except Africa), we are seeing the emergence of large urban clusters like Delhi, Mumbai, Beijing-Tianjin, Shanghai-Suzhou, the Pearl River Delta, and Seoul-Incheon.
These cities of more than 20 million people are evolving into much larger clusters with populations expected to exceed 30 million people by 2020. The spatial structures of these cities are demand driven and reflect the modernization of their economies, wherein large supply chains of services and manufacturing have different spatial requirements than the ones encountered in the traditional monocentric city of the past.
As these cities develop different spatial forms, current transport systems—consisting of mass public transport, collective taxis, and individual transport—are becoming inefficient. Transport inefficiency results in the fragmentation of potentially large productive labor markets into smaller, less efficient ones. In addition, transport inefficiency results in increased congestion, pollution, and high GHG emissions.
Transport Systems Have to Adapt to Constantly Evolving City Structures
Alarmed by the poor performance of current urban transport modes, municipalities and planners often try to contain urban expansion in a more compact form that they feel would be easier to serve with traditional public transport modes: public mass transport and city buses. Policies constraining the emerging demand for land required by the new economy result in extremely high land prices and, in many cases, informal developments not served by adequate infrastructure expansion—as is the case in Mexico City, for instance. Advocates of compact cities often ignore the fact that the compact part of their city is already extremely congested and that governments are unable to provide the public transport line intensity that would prevent congestion for commuters using public transport as well as individual cars. The example of the congestion in Beijing’s new subway lines (see figure 5.10) discussed above is a good illustration of this issue.
Municipalities and planners should face the reality of changing urban land use required by the new economy. Instead of fighting the expansion of cities, which is largely demand driven, to preserve an obsolete and congested type of land use, planners should try to create new transport systems that could serve both the traditional high-density CBDs and the more recent dispersed forms of urban clusters. The schematic representation shown in figure 5.25 illustrates the changes in spatial structures occurring in many large cities and the trip patterns that they generate. The left schematic in figure 5.25 is representative of cities like New York or London, where already more than 70 percent of commuting trips are from suburbs to suburbs. The right schematic in the figure represents the emerging structures of megacities like Delhi, Beijing, or Shanghai, where commuting and freight trips suburb-to-suburb are becoming more numerous and are evolving into even more complex routes. The current urban transport systems in these cities, consisting largely of radio-concentric public transport lines fed by city buses and complemented by cars and collective taxis or rickshaws, are ill adapted to serve the complex urban shapes shown on the right in figure 5.25.
Urban trip patterns in a typical metropolitan area compared to that of an urban cluster.
Emerging Demand-Driven Technology Could Allow Labor Markets to Function in Large Clusters
How should transport systems adapt to new urban forms? The emergence of large urban clusters expanding to distances of about 100 kilometers across and including high-density cores surrounded by low-density suburbs where jobs are mixed with residential areas suggests that we will see major changes in urban transport modes. Emerging technological changes would facilitate this transition.
First, cars will have to become more compact to use less road space and less energy, making them look more like a hybrid between cars and motorcycles. Examples of such vehicles already exist and are being manufactured, for instance, the Toyota i-Road.
Second, individual compact vehicles should become available at subway or suburban rail stations on a rental or share basis. This is already happening—the Toyota i-Road is available at some Tokyo suburban stations and at the Grenoble (France) main railway station. Such vehicles will allow commuters to combine the convenience of individual vehicles with the speed of suburban rails on longer distances for large intra-metropolitan trips.
Third, subway and suburban rails should have fewer stations and run at higher speeds to allow commuters to cross an entire urban cluster in less than 1 hour. That would require rail speed of about 150 km/h. The catchment area of stations would be increased to more than 200 square kilometers because of the availability of individual vehicles at stations (compared to the current 2 square kilometers limited by the 800 meter walking distance from stations). Figure 5.26 and table 5.6 compare the catchment area of a traditional subway with stations spaced 1 kilometer apart to that of a suburban rail system with stations spaced every 10 kilometers but accessible by small individual vehicles (ranging from bicycles to the Toyota i-Road) with a range of 8 kilometers. Using the speed assumptions shown in the table, a commuter could travel a distance of up to 66 kilometers in less than 1 hour and have access to a destination area of more than 600 square kilometers.
Comparison of the catchment area of traditional subways and that of high-speed suburban rail associated with individual urban vehicles.
|
Comparison of distances between stations, speeds, and catchment areas for traditional subways and high-speed suburban rail. |
||||||||
Traditional subway and walking |
Fast suburban train and individual vehicle |
|||||||
|
Suburban rail line length (kilometers) |
50 |
50 |
||||||
|
Average train speed (km/h) |
32 |
110 |
||||||
|
Distance between stations (km) |
1 |
10 |
||||||
|
Number of stations |
50 |
5 |
||||||
|
Radius distance to station (km) |
walk |
0.8 |
vehicle |
8 |
||||
|
Speed from trip origin to station (km/h) |
5 |
35 |
||||||
|
Catchment area of one station (square kilometers) |
.01 |
201.06 |
||||||
|
Total catchment area of the line (square kilometers) |
53.79 |
623.76 |
||||||
|
Train riding time for 50 kilometers (minutes) |
94 |
27 |
||||||
|
Walking or individual vehicle riding time (minutes) |
19.2 |
27 |
||||||
|
Total trip time (minutes) |
113 |
55 |
||||||
|
Total trip length (kilometers) |
51.6 |
66.0 |
||||||
|
Total trip time (minutes) |
113 |
55 |
||||||
|
Average speed (km/h) |
27 |
72 |
||||||
Fourth, self-driving minibuses should be able to pick a few passengers running on the same route and drive them to individual final destinations without having to stop on the way to pick up and drop off other passengers.
Local governments should not favor a specific mode of transport but should favor and facilitate a large mix of transport modes, including combining fast heavy rail with individual vehicles for the same trip. Subway and fast suburban train stations should be designed with large areas for loading and unloading passengers for self-driving vehicles (figure 5.27).
The Toyota i-Road personal mobility vehicle (left) and the Beijing three-wheeler (right) already provide on-demand trips from station to door and from door to station.
Calls to Action for the Future of Mobility
1. Maintaining mobility is an essential task for municipalities and urban planners. This is best done by allowing multimode transport systems to reflect consumer demand. Commuters should be able to choose among the various transport modes available those that best fit their commuting needs.
2. Planners should not select densities and urban spatial structures in order to best fit an existing preselected transport system. Instead, new transport systems should adapt to evolving spatial structures.
3. Because pricing of pollution and GHG emissions are currently difficult to apply without large transactions costs, governments should set pollution and GHG emission targets as a substitute for price (until technology is available to directly charge for pollution emitted and GHGs released).
4. The pricing of road space is also an important metropolitan authority task. Flat tolls on roads should be progressively replaced by congestion pricing that constantly adjusts depending on time and location to maintain a set speed on specific road segments. As currently practiced in Singapore, the target set speed is different in the CBD and in the suburban arterial roads. The technology to do this is currently available.
5. Eventually, individual commuter cars will have to be redesigned to reduce their road footprint and their weight. Emerging new personal mobility vehicles, such as the Toyota i-Road, are examples of a possible replacement for the traditional car that would provide more mobility for less road space and less energy, pollution, and GHG emissions per kilometer.
6. Finally, the possibility of sharing small self-driving vehicles on demand could provide a very efficient alternative in the future for many suburb-to-suburb trips. Self-driving cars would have three important advantages over traditional cars. First, they would save street space by being able to run closer to one another without requiring the 2-second reaction time that human drivers require; that would save about 65 percent of road space at speeds of about 60 km/h. Second, they would dramatically reduce accidents and, therefore, the unpredictability of road commuting times. Third, they would not require large parking spaces in the center of cities where real estate is the most expensive.
Many new technologies are emerging that will have a large impact on urban transport. These technologies could reduce pollution to near zero, greatly reduce urban transportation’s contribution to global warming, prevent most transportation accidents, and increase the capacity of existing urban roads without creating congestion.
A little more than 100 years ago, the horse as the only nonpedestrian means of urban transport was replaced by mechanical vehicles. These vehicles completely changed cities by allowing them to expand without densifying into Dickensian slums, while enlarging the potential labor markets that greatly increased the productivity and welfare of urban dwellers. Since that time, the automobile, buses, and subways as major modes of urban transport have not changed much. We might now be on the verge of an urban transport revolution that could be comparable to the replacement of animal traction by mechanical traction and could also greatly enhance the welfare of the very large part of humanity that is likely to live in cities toward the end of the twenty-first century.
The emergence of small footprint, on-demand, shared vehicles (very different from large buses running on fixed schedules and routes) will change the way urban transport is organized. The pattern of roads and arterials may also change to adapt to these new modes of urban transport. Instead of concentric traffic on a few high-capacity highways or arterials, numerous smaller low-capacity roads would allow the flexibility required by trips from dispersed origins to dispersed destinations.
New types of specialized urban vehicles—collective or individual, shared or not shared, self-driving or with drivers—are likely to multiply in the future. The speed, street footprint, and size of these vehicles will be adapted to the types of trips and commuters that they serve. The types of urban vehicles will therefore be different in very large, dense cities like New York and Mumbai from much smaller cities like Amsterdam and Key West.