mode of travel transit

What is Modal Shift and How Can it Change the Way We Travel?

Explore a new way of thinking about your mode of transport, and discover how your trips can have an environmental impact.

What’s a Rich Text element?

The rich text element allows you to create and format headings, paragraphs, blockquotes, images, and video all in one place instead of having to add and format them individually. Just double-click and easily create content.

Static and dynamic content editing

A rich text element can be used with static or dynamic content. For static content, just drop it into any page and begin editing. For dynamic content, add a rich text field to any collection and then connect a rich text element to that field in the settings panel. Voila!

How to customize formatting for each rich text

Headings, paragraphs, blockquotes, figures, images, and figure captions can all be styled after a class is added to the rich text element using the "When inside of" nested selector system.

It’s true that as humans, we find solace in being comfortable, and no less can be said about the way we choose to travel. We want predictable and reliable when it comes to our transit modes, and deviating from this path takes us into a world of unknowns. But change offers us the chance to do better—to create sustainable transport beyond the traditional travel models. Here’s how Modal Shift and a change in transit behaviour can make a difference.

What is Modal Shift?

Modal Shift can be deemed as a new way of thinking about the way we travel. It encourages innovation, sparking alternative means of transit that combat the problems incited by previous travel models. At its core, Modal Shift pushes people towards more sustainable transport to benefit society.

Essentially, Modal Shift is the shifting of travel modes that humans go through based on a range of variables; both external and internal influences that impact the way we move. For example, during the COVID-19 pandemic, the way in which people travelled drastically altered, with a decline in public transport and private car usage following the rise of working from home. 

To adopt Modal Shift requires a change in transit behaviour and habits, and it can be difficult to achieve this without intentionality and focus. It will take the cooperation and commitment of governments and communities in order to create equitable and convenient transit access for all.

What are the benefits of Modal Shift?

Modal Shift is an opportunity for Transit Agencies and travellers to experience the advantages of new forms of transport. This presents a number of benefits on both a macro and micro scale.

Environment   

It’s not news to us that our world is suffering environmentally, and our lifestyles, especially transit, have a lot to do with this. According to the United States Environmental Protection Agency , in 2018, transportation accounted for the highest level of US greenhouse gas emissions at 28 per cent, and from this, 59 per cent were attributed to light-duty vehicles.

Further evidence suggests that much of society is choosing solo transit modes over shared, adding to the growing environmental concerns around harmful emissions.

“The United States has one of the highest levels of car ownership in the world with one car for every two people.” – The Geography of Transport Systems .

Adopting Modal Shift into transportation can have a substantial impact on the environment. Reframing transit and reducing our reliance on solo vehicles means that fewer cars frequent our roads, reducing air pollution and damage to our infrastructure.

Active transportation, such as walking or cycling, is one of the most traditional modes of transit, but is often disregarded.

According to census data , the number of commuters in the US choosing walking as their means of getting to work has declined substantially between 1980 and 2012 from 5.6 per cent to 2.8 per cent. Cycling had increased by 0.1 per cent, but remained a low figure at 0.6 per cent. 

A Modal Shift towards active transportation not only alleviates our carbon footprint, but has a number of health benefits, including reducing obesity and preventing chronic conditions such as diabetes and cardiovascular disease.

Time and money

Emerging technologies have opened up new avenues of transport, flipping the lid on both solo travel and public transit models. With the development of new smartphone Apps, Transit Agencies have been able to look towards efficient and dynamic routing that saves both the commuter and driver transit time.

Microtransit is a trend to have come about from such developments, offering a shared transit experience that is dynamic and recalculates based on the commuters registered. A Modal Shift towards this model not only means that travellers are taken on the fastest route to their destination, but they also save on costs as it is cheaper than solo travel.

Modal Shift has the potential to influence our lives and the world we dwell in, improving our carbon footprint, supporting our general health and saving us valuable time and money.

Want to dive deeper?

Introducing the Definitive Guide to Rural DRT.

What are some ways to influence Modal Shift?

Establishing Modal Shift requires commitment and resources from the government, Transit Agencies, and the community.

If Modal Shift were to be directed towards active transportation, an investment in bicycle and pedestrian facilities would be required by Transit Agencies and local governments. Increased bicycle lanes and walking paths, as well as a surplus of bikes for hire, would help support this.

Another option in favour of Modal Shift would be establishing intermodal passenger transport, which involves using multiple means of transit in a given journey. An example of this would be taking a bus to an inner city stop and providing cycling facilities for the commuter to travel the rest of the journey. 

And with society constantly changing as technology evolves, there is a space for Modal Shift to exist through Demand-Responsive Transport (DRT)—a model that deviates from fixed-route transit to more dynamic journey-mapping through the use of smartphone Apps. It’s essentially public transit that you can book on your phone—and with the assistance of technology, this is becoming a viable and affordable option for the future of transportation. By working with existing Public Transport infrastructure, and enhancing it with new technologies, Modal Shift can be easily attained.

There are a number of options for our society to facilitate Modal Shift that would begin to make a difference to the environment as well as to our own wellbeing.

How can Modal Shift impact our future?

Modal Shift is key to us moving forward in the future of transport. It allows a space for Transit Agencies to innovate, and for communities to adapt towards mobility that is healthier for our society.

Re-evaluating how we travel and evolving our current transit modes has the potential to significantly improve our environment. It can lead to a reduction in harmful greenhouse gas emissions while also supporting the infrastructure of our roads. Not only this, but Modal Shift can contribute to our overall health by encouraging active forms of transportation and has the potential to alleviate time and expenses required for our daily commute. 

Modal Shift means that we can move towards a better and more sustainable society, creating transport models that will have a significant and beneficial impact on generations to come.

Looking for more?

Sign up to our newsletter.

About the Author

Cassie Holland

Cassie has forged a career out of her passion for the written word and love of hearing people's stories. She utilises her writing and editing skills to craft pieces that tell a story.

Related Content

Tech-enabled on-demand transport that doesn’t leave passengers behind.

Transportation leaders discuss implementation, rider education and adoption of Demand-Responsive public transport.

Liftango announces the next major product features and enhancements in “Abrazo” product release

Liftango is very pleased to announce the next major product release “Abrazo” for its Carpool, On Demand and Fixed Routes Apps.

Where Should You Deploy DRT?

Liftango data scientist, James Willcox, practically demonstrates where Demand-Responsive Transport deployments are best implemented.

• About The Journalist’s Resource
• Follow us on Facebook
• Follow us on Twitter
• Criminal Justice
• Environment
• Politics & Government
• Race & Gender

Expert Commentary

Transportation safety over time: Cars, planes, trains, walking, cycling

2013 study by Northwestern University on the relative fatality risk of a broad range of motorized and non-motorized transportation modes in the United States.

Republish this article

This work is licensed under a Creative Commons Attribution-NoDerivatives 4.0 International License .

by Leighton Walter Kille, The Journalist's Resource October 5, 2014

This <a target="_blank" href="https://journalistsresource.org/economics/comparing-fatality-risks-united-states-transportation-across-modes-time/">article</a> first appeared on <a target="_blank" href="https://journalistsresource.org">The Journalist's Resource</a> and is republished here under a Creative Commons license.<img src="https://journalistsresource.org/wp-content/uploads/2020/11/cropped-jr-favicon-150x150.png" style="width:1em;height:1em;margin-left:10px;">

In May 2013 the National Highway Traffic Safety Adminstration released estimates of U.S. traffic fatalities for 2012, and the results were troubling: 34,080 people died in motor-vehicles crashes that year, an increase of 5.3% over 2011’s total and a reversal of the long-term downward trends. The meaning of the one-year shift is unclear and there is a great deal of nuance within all the numbers, but the litany of deaths remains sobering — an average of more than 93 every day.

The lowest year in recent history was 2011, when 32,367 people died on U.S. roads and highways. As horrifying as that number is, it actually constitutes progress: In 1994, 40,716 died on the roads — 26% more, and nearly 112 deaths per day. Beyond the absolute numbers, progress has been made in overall mortality rates : 1.1 per 100 million vehicle miles travelled in 2011 (it rose to 1.16 in 2012). Much of this is due to advances in vehicle safety — air bags, anti-lock brakes and increased crashworthiness. But technology can have its drawbacks as well: Since 1980 the average horsepower of U.S. cars more than doubled , and speed limits have risen significantly, greatly increasing the potential for damage, loss of life and injuries. Motorcycling is also on the rise, and fatality rates have increased in lockstep with its popularity and inherent riskiness.

A 2013 study in Research in Transportation Economics , “Comparing the Fatality Risks in United States Transportation Across Modes and Over Time,” looks at the historical trends to paint a fuller picture of where this all stands. The researcher, Ian Savage of Northwestern University, prefaces his findings with an important caveat on measures of “safety”:

The focus on fatalities is primarily motivated by a greater confidence that this measure of safety is reported more consistently and accurately across modes and time. In general, cross-sectional and time-series comparisons in fatalities are also indicative of differences in non-fatal injuries, illnesses, and property damage. Albeit that the correlation is not perfect. In particular, fatalities are a poor measure of some of the environmental risks associated with the transportation of oil products and hazardous materials. In addition many of the advances in safety in recent decades have focused on “crashworthiness” whereby design changes have been made to increase the survivability of crashes and mitigate the severity of injuries. Consequently it is possible that a reduction in fatalities may be partly compensated for by an increase in the number of injuries.

Savage’s analysis involved two datasets: The first involved the relative risk of different travel modes — cars, buses, planes, trains, and more — from 2000 to 2009; the second was a time-series analysis for each mode from 1975 to 2010. The findings of the study include:

• Between 2000 and 2009, on average 43,239 people in the United States died each year in transportation-related incidents. Based on the average number of U.S. residents over that period, the annual risk of dying in a transportation-related accident is 1 in 6,800.
• Transportation-related fatalities constituted just under 2% of the 2.43 million deaths per year from all causes in the United States, or 1 in 56. Transportation was the biggest source of all “unintentional injury deaths” (38%) — those not caused by old age, disease, suicide or homicide.
• Whatever the vehicle, highways are by far the most common place of transportation fatalities in the United States: 94%. If deaths that take place at rail-highway grade crossings are included, the total is even higher, at just over 95%.
• Despite significant fatality rates for highway travel, overall transportation is becoming less dangerous: “The rate in 2010 is just one-third of that in 1975 (1.11 versus 3.35 fatalities per 100 million vehicle miles). The 1980s and early 1990s were the era of the greatest rate of improvement.”

Cars, trucks and SUVs

• Nearly three-quarters of people who died in highway crashes (74%) were occupants of automobiles and light trucks. More than half (55%) occurred in single-vehicle incidents without a prior collision, including roll-overs; vehicles striking fixed objects, animals or debris; or catching fire.
• The proportion of fatal single-vehicle crashes is much higher for light trucks (66%) than it is for automobiles (47%). Light trucks — including minivans, pickups and SUVs — often have a relatively higher center of gravity and thus a greater propensity to roll over.
• Drivers or passengers in cars or light trucks faced a fatality risk of 7.3 per billion passenger-miles: “A person who was in a motor vehicle for 30 miles every day for a year faced a fatality risk of about 1 in 12,500. Relative to mainline trains, buses and commercial aviation the risk was 17, 67, and 112 times greater, respectively.”
• Because private individuals operate the vast majority of motor vehicles, their risk is highly dependent on personal behavior: “Unlike the commercial modes where passengers are victimized randomly, the risk to individual highway users varies considerably depending on age, alcohol consumption and the type of road used.” (See highway risk factors below.)
• “One might argue that transportation equipment, and in particular the motor vehicle, must be the most dangerous machines that we interact with on a daily basis,” the researcher states. “The annual toll in motor vehicle crashes exceeds the deaths resulting from the next most dangerous mechanical device, firearms, by about 40%.”

Motorcycles

• Nearly 10% of all highway fatalities — one in ten — were motorcyclists: “When a motorcycle is involved in a collision with another vehicle, the motorcyclist invariably receives more serious injuries. The ratio of fatalities in two-vehicle collisions was 70 motorcyclist fatalities for each fatal injury sustained by the occupant of the other vehicle.”
• Over the period studied, motorcycles became increasingly popular, with use rising as much as 75%. As a consequence, fatalities have increased proportionally. This trend has been exacerbated by the “general rollback in the number of states requiring motorcycle riders to wear helmets.” ( Earlier research has indicated that when a state repeals or weakens a helmet-use law, motorcyclist fatalities typically rise nearly 40%.)
• Motorcycles had a fatality rate of 212 per billion passenger miles, by far the highest of all modes: “A motorcyclist who traveled 15 miles every day for a year, had an astonishing 1 in 860 chance of dying — 29 times the risk for automobiles and light trucks.”

Highway risk factors

• Urban roads are far safer than those in rural areas: “Based on data from 2009, highways in rural areas have a fatality risk that is 2.7 times greater than that in urban areas. In general the lower average speeds, greater provision of lighting, greater deployment of traffic control devices and fewer curves in urban areas more than compensate for factors such as the greater number of intersections and the presence of pedestrians.”
• Gender and youth play are significant factors in fatality risk: Males are three times more likely to die in a road accident than females, while people between the ages of 18 and 29 are at a 50% to 90% greater risk.
• Seat-belt use is a significant factor: Half of vehicle occupants who die in automobiles and light truck incidents (49%) were not wearing seat belts or using child safety seats.
• Alcohol played a role in approximately a third of all highway fatalities, with at least one of the involved parties having a blood-alcohol level above 0.8 grams per deciliter.
• Related research has shown that drivers using cell phones show greater impairment than drunk drivers, and hands-free systems offered no improvement over handheld devices. Cell-phone conversations have a more profound effect on driver performance than other forms of in-car distraction, including talking to passengers and listening to the radio.
• Mainline railroads claimed an average of 876 lives a year, the majority of which occur during collisions with highway users and pedestrians. The largest number of deaths, 490, involve people and vehicles not at grade crossings, and a significant portion of those deaths, approximately 85 to 110, were possibly suicides.
• The balance of rail-related deaths involve motorists at grade crossings (281), pedestrians at grade crossings (68) employees and contractors working on the tracks (26). Per year on average, only seven passengers traveling on mainline trains die.
• The overall fatality rate for long-haul train service is 0.43 per billion passenger miles. Excluding pedestrians and others not on trains — 64% of total fatalities assigned to railroads — the fatality rate is approximately 0.15 per billion passenger miles.
• Vehicles with a capacity of 10 passengers or more represented just 0.1% of the total fatalities. On average, there were approximately 40 fatalities per year, with drivers and other bus-company employees representing 25% of lives lost.
• Scheduled and charter service accounted for 44% of total bus fatalities . The balance of deaths occurred with school buses (23%), urban transit (11%) and a variety of private shuttles, church buses and other services (22%).
• The fatality rate per billion passenger-miles for buses is relatively low, 0.11. However, this is still 65% greater than that for aviation, and doesn’t include victims of crime. (Also, see statistics on “curbside” bus services. )
• The majority of aviation fatalities that occur each year (85%) involved private aircraft (known as “general aviation”). On average, 549 people die each year in activities such as recreational flying (41% of flight hours), business travel (24%), and instruction (17%).
• Excluding acts of suicide and terrorism, commercial aviation was the safest mode of travel in the United States, with 0.07 fatalities per billion passenger miles: “A person who took a 500-mile flight every single day for a year, would have a fatality risk of 1 in 85,000.” (One variable to note: Takeoffs and landings are where the risk is, not in the number of miles flown, so risk-per-flight calculations are higher.)

Walking, bicycling

• Between 2000 and 2009, on average 6,067 pedestrians and bicyclists died on U.S. highways and in collisions with other modes of transport. Of these, 4,930 died when hit by cars and trucks operated by private users, 545 deaths resulted from collisions with commercial carriers, and 592 from commercial users not on highways.
• In all, fatalities of pedestrians and bicyclists make up nearly 15% of annual average highway fatalities. More than 90% of pedestrian fatalities occurred when the victims were hit by automobiles and light trucks.
• A related study on risk factors for on-road cycling commuters indicated that prior to car-bicycle accidents, 89% of cyclists were traveled in a safe and legal manner. In addition, vehicle drivers were at fault in 87% of the events.

A related 2013 study published in the Annals of Emergency Medicine , “Safety in Numbers: Are Major Cities the Safest Places in the United States?” examined the overall injury risk in urban areas compared with suburban and rural areas. Among the conclusions: Rural counties demonstrated significantly higher death rates than urban counties — 1.22 times greater for the most rural compared with the most urban. The majority of the difference was in unintentional incidents such as accidents, but some increase in suicide risk was also seen in rural areas. Overall, the study finds, “U.S. urban counties were safer than their rural counterparts, and injury death risk increased steadily as counties became more rural.”

Also of interest is a 2013 report by the World Health Organization. “Global Status Report on Road Safety 2013: Supporting a Decade of Action.” The study was based on country-level data and included information on newer risk factors such as cell phone use while driving. Among the findings, the report states that more than 1.24 million people die every year as a result of road traffic injuries, making it the eighth leading cause of death globally, and the leading cause of death for young people aged 15-29. Based on anticipated trends, by 2030 road accidents are projected to be the fifth leading cause of death globally.

Keywords: cars, trains, Amtrak, Northeast Corridor, planes, airlines, pedestrians, bicycling, safety, distracted driving, alcohol , multitasking, driving

About The Author

Leighton Walter Kille

Mode choice

Mode choice model form and structure

Factors influencing mode choice

Traveler characteristics (socio-demographics, attitudes, perceptions, lifestyle)

Modal availability and characteristics

Mode characteristics

Characteristics of the journey

Mathematical formulation

Practical Issues in Mode Choice Modeling

Market segmentation

Transferability of mode choice models

Calibration/Validation

Page categories

Activity Based Models

Needs Review

Topic Circles

More pages in this category:

Mode choice is the process where the means of traveling is determined. The means of travel is referred to the travel mode, which may be by private automobile, public transportation, walking, bicycling, or other means. How desirable a travel mode is usually is expressed using utilities . In most travel models, mode choice is applied to travel that has already been estimated, meaning that mode choice is applied to a trip or tour, or group of trips or tours, where the origin and destination are already known.

Mode choice is an important part of models that are used for analyses such as:

• Major transportation investment projects since they may attract travelers not only from competing facilities but also competing modes;
• Transit service changes, which may encourage or discourage travelers from using transit;
• Long range forecasts, where changes in demographics or in travel conditions (e.g. increased congestion) may alter the relative worths of different modes for some or all travelers;
• Pricing policy analyses, which may discourage travelers from using modes with increased prices; and
• Land use planning analyses, where changes in development patterns may make certain modes more or less attractive relative to others.

If a model is vehicle-trip based (i.e., if vehicle trips rather than person trips are generated and distributed), there is no need for mode choice or vehicle occupancy modeling capabilities.

# Mode choice model form and structure

In a conventional four-step model, mode choice is considered the third step in the process, following trip distribution and preceding network assignment . In these models, the outputs of trip distribution are person trip tables, which are matrices of trips where the rows and columns represent some aggregate geography, usually transportation analysis zones (TAZ). By convention, at this point in the modeling process, the rows correspond to trip productions and the columns to trip attractions . The mode choice process determines the percentage of trips made by each mode for each cell in the matrix (zone pair).

In tour based models , mode choice is typically separated into two stages, the tour level and the trip level. At the tour level, it is assumed that prior steps in the model have generated tours whose origins (usually home) and primary activity locations are known. In a disaggregate activity based model application, for each tour, the probability of using each mode is computed, and the chosen mode is realized through Monte Carlo simulation. At the trip level, it is assumed that prior steps have generated the stops on each tour and their locations and that tour mode choice has been simulated. The trip mode choice component simulates the mode for each trip between every two stops on a tour, including the origin and the primary activity location. Trip mode choice is dependent on tour mode choice in that a chosen tour mode will allow only certain trip modes as valid. For example, on an auto tour, where the vehicle is brought along to all stops, transit may not permitted for trips on that tour.

In both trip based and tour based models, mode choice is determined using probabilities for each mode estimated from the characteristics of the trip, the modes, the traveler, and the environment in which the travel occurs. In an aggregate model, such as a four-step model, mode choice is realized by applying the probabilities as percentages to the trips to which the probabilities apply. In a disaggregate model, such as an activity based model, the probabilities inform a Monte Carlo simulation process that simulates the mode for each tour/trip.

Typically, mode choice is formulated as a discrete choice model with alternatives corresponding to the specific tour or trip modes. The following types of mode alternatives may be found in these models:

• Drive alone
• Shared ride with 2 occupants
• Shared ride with 3 or more occupants

In some models, auto modes may be further stratified by having separate sets of alternatives for potential priced road users and non-priced road users.

• Transit with walk access at the home end

In some models, transit modes may be further stratified by having separate sets of alternatives for by type of transit, such as commuter rail, heavy rail, light rail, express bus, local bus, etc. In some models, school bus may be an alternative for tours or trips with a school purpose.

Some regions might specify more complex travel models than are warranted based on existing travel choices represented in a travel survey. This is often the case for regions that anticipate testing new alternatives in the future. For example, a region that does not have HOV lanes in a base year might specify a model that stratifies the auto mode by group size rather than simply modeling drive alone and carpool trips. This would provide the capability for testing alternative carpool lane treatments. Likewise, an area without fixed guideway transit service may specify mode models that include walk and drive access (to estimate park-and-ride lot usage) and fixed guideway submodes in order to test future alternatives.

# Factors influencing mode choice

# traveler characteristics (socio-demographics, attitudes, perceptions, lifestyle).

The characteristics of individual travelers affects their choice of mode. These may include observed characteristics, such as age, gender, driver’s license status, or worker or student status, and unobserved characteristics, such as awareness of transportation options, consideration of specific modes, attitudes, and personal preferences. Characteristics of a traveler’s household, such as income level and vehicle availability, may also affect mode choice.

# Modal availability and characteristics

Obviously, for a mode to be chosen, it must be available to the traveler. Availability may not be able to always be clearly defined. For example, an auto driver mode is generally unavailable to a young child, but it may also be unavailable to a disabled person, someone without a driver’s license, or someone who does not have a vehicle available. Sometimes these characteristics may not be observable; for example, someone whose household does not own a car may still be able to borrow or rent an auto or use a car sharing service.

In general, there are no restrictions for any traveler to be an auto passenger. Auto driver modes may be limited to those who can be defined as eligible to drive, such as persons over the legal driving age, and perhaps those holding driver’s licenses, if that characteristic is simulated in the model. It is more likely that such characteristics would be known in disaggregate activity based models than in household based aggregate models.

In many models, however, auto drivers are not distinguished from passengers, and the auto mode definitions are based on vehicle occupancy (e.g., drive alone, two person shared ride, etc.). In this case, any restrictions are usually applied only for the drive alone mode.

The availability of the walk mode is in theory unrestricted, unless those with mobility limitations that would prevent the ability to walk are specifically identified in the model. However, as a practical matter, very long trips or tours almost never use the walk mode, and in many models, trips over a certain length are deemed not to have the walk mode available. The same type of restriction is sometimes implemented for bicycle modes as well (thought with a longer maximum trip length.

Availability of public transit modes is generally not restricted; however, there may be no reasonable transit paths for some origin-destination combinations. It is common practice to separate transit modes into those with walk access/egress at the home end and those with auto access/egress. The walk access mode often has restrictions on the maximum walk length, similar to those used for the walk mode, and so there may be origin-destination combinations where there are no valid transit paths (for example, if the origin is not within the maximum walk distance of any transit stop). In theory, there is no limit on auto access distance, but there may be practical limitations imposed in some models. In general, a transit mode is deemed to be available if the transit path building process is able to find a valid path from the origin to the destination.

# Mode characteristics

Mode characteristics are usually measures of the level of service provided by the mode. For motorized (auto and transit) modes, the following characteristics are often used:

• In-vehicle time
• Walk egress time, which may be referred to as “terminal time”
• Parking cost

For auto modes, the auto operating cost is usually used as a characteristic. This is generally defined as the marginal cost of operating the vehicle for the trip or tour, which in large part is the fuel cost but may also include other costs that vary by the distance traveled.

For transit modes, other characteristics that are often part of the mode choice model include:

• Walk access time
• Auto access time
• Initial wait time
• Transfer wait time
• Transfer travel time (e.g. walking between stops)
• Number of transfers

For non-motorized modes, characteristics are usually limited to time and or distance. Because speeds may vary much more among travelers for non-motorized than for motorized modes, it may be difficult to estimate travel times for individual travelers. For bicycle modes, a limited amount of research has been done indicating that bicyclists prefer paths that include dedicated bicycle facilities, fewer turns, less vehicular traffic, fewer traffic controls, smooth pavement, level terrain, and no on-street parking. Some of these characteristics, however, may be difficult to quantify in some modeling environments.

# Characteristics of the journey

Characteristics of the tour or trip may have an effect on the mode choice decision. Some examples include:

Time of day – This may affect not only the level of transportation service (which may also be considered in the mode characteristics variables) but also considerations such as a higher preference for bicycling, walking, or waiting in daylight hours.

Stops on a tour – In a tour based model, the presence, number, and type of stops may have an effect on mode choice. For example, it may be more desirable to use an auto for a tour with many shopping stops, or with stops related to picking up or dropping off passengers.

Certain land use characteristics may favor the use of some modes. In general, more densely developed areas see higher shares of transit and walking modes. The land use characteristics that affect mode choice are often included in the form of variables representing density of development or other features near the origin or destination of the tour or trip. Some examples include:

• Population density
• Employment density, often by type
• Measures of “mixed land use” density that consider how much different types of development (e.g. commercial and residential) occur
• Intersection density, which considers the density of the road network as well as its connectivity

# Mathematical formulation

Most mode choice models are discrete choice models , with logit (multinomial, nested, or cross-nested) being the most comnmon formualtion.

# Practical Issues in Mode Choice Modeling

# market segmentation.

Experience shows that different segments of the population behave differently regarding mode choice. Some examples are straightforward; for example, travelers without a car available are less likely to use auto modes than travelers who always have a car available. Therefore, some segments may be defined as person or household characteristics (gender, household income, number of vehicles or vehicle sufficiency). Other segments may represent geographic or land use characteristics. For example, a model may include a segment for travel to the central business district (CBD) or to the regional airport. While some of the effects of these types of segmentation may be captured in other model variables (e.g., parking prices at the airport or density of CBD land use), other effects may not be easily captures (e.g., the fact that many attractions can be reached easily without an auto in the CBD).

A critical issue regarding segmentation in mode choice modeling is the availability of data for model estimation and validation. Since most tours or trips are made by auto modes, estimation data sets are dominated by auto travel. The data to estimate separate models or parameters by segment may often be insufficient. There are also data sufficiency issues regarding validation and calibration as there may be insufficient data to determine specific values of segment-specific parameters.

# Transferability of mode choice models

Because of the expense of data collection and the relatively low shares of non-auto modes in survey data sets, transferred mode choice models are fairly common. The research on the validity of mode choice model transferring is mixed, as documented by Cambridge Systematics et al., 2012 and Rossi and Bhat, 2014.

# Calibration/Validation

(see Model calibration and validation )

The FHWA Model Validation Manual (Cambridge Systematics, Inc. et al., 2010) includes detailed information about validation of mode choice models in Chapter 7. The main sources of data for validation of mode choice models include the following:

Transit ridership counts – Transit ridership counts have the best information on the total amount of travel by transit, usually at the route level. It is important to recognize, however, that ridership (boarding) counts represent “unlinked trips,” meaning that a person is counted each time he or she boards a new transit vehicle. So a trip that involves transit transfers is counted multiple times. Mode choice models generally consider “linked trips,” where a trip including transfers counts as only one trip. Information on transfer rates is required to convert unlinked trips to linked trips; such information generally is obtained from transit on-board surveys.

Transit rider survey – A transit rider survey (typically an on-board survey) is an invaluable source of information for validation of the transit outputs of mode choice models but may have also been a data source for model estimation. A wealth of information that cannot be obtained from transit counts is available from on-board surveys, including:

Transit trip origin-destination patterns by trip purpose

Access modes;

Transit paths (surveys should ask riders to list all routes used in order for the linked trip);

Transit submodes used (e.g., bus, light rail);

Transit transfer activity; and

Characteristics of the surveyed riders and their households.

It should be noted that transit on-board surveys usually provide data only for individual transit trips, not tours, and so their use in estimating transit travel in tour-based models is limited.

• Household travel/activity survey – If such a survey is available, it may have also been a data source for model estimation although data from other sources such as transit on-board surveys may also have been used in model estimation. The household survey is the best source for information on nontransit travel data since the number of observations for transit travel is usually small. The expanded household survey data can be used to produce observed mode shares for nontransit travel by purpose for a number of geographic and demographic market segments.
• Highway usage data – Data on highway usage such as toll road and high-occupancy vehicle lane use would be helpful validation data for models that include related modal alternatives.

The most basic aggregate checks of mode choice model results are comparisons of modeled trips or tours by mode, or mode shares, to observed data by market segment. Market segments include trip or tour purposes as well as demographic segments, such as income or vehicle availability levels, and geographically defined segments. It is critical to remember, though, that aggregate validation of regional mode shares is insufficient to validate the mode choice model. Disaggregate validation, and validation of market segments of interest, are also needed.

Mode choice models are typically applied using trip tables (or their tour-based equivalents) as inputs. The mode choice model’s results, therefore, are shares of the total trip table for each market segment that use each of the modal alternatives. Validation of the model’s aggregate results involves checking the shares for the model’s base year scenario results against observed mode shares.

A household survey is the only comprehensive data source covering all modes, and there-fore is the only source for mode shares. However, mode shares for modes that are used relatively infrequently – notably transit modes – as well as mode shares for relatively small segments of the population (such as members of zero-vehicle, high income households) cannot be accurately estimated from household surveys due to small sample sizes.

Transit ridership counts provide estimates of total transit trips, not mode shares. To convert these trips to shares, an estimate of the total trip table for each market segment is needed. Assuming good validation of the trip generation and distribution components (or their tour-based equivalents), the trip table outputs from the trip distribution model can provide this information. Basically, the transit trips by submode, access mode, trip purpose, and other segmentation level, segmented using the transit rider survey data, can be subtracted from the total trips represented in the trip table to obtain estimates of “observed” nontransit trips. The nontransit trips can be separated into trips by individual mode (auto and nonmotorized submodes) using information from the household travel survey. While it may be problematic to find an alternate source for some segments or modes (such as bicycle travel), transit trips and shares by segment may be estimated using data sources including ridership counts and transit rider surveys.

The mode choice model validation process is tied in with the highway and transit assignment validation processes because these processes are better able to make use of independent aggregate data at the link or route level. It is common practice to compare overall mode choice model results to the observed data for the region. It is important to recognize that this type of regional check is not sufficient to determine that the mode choice model is validated, any more than it would be sufficient to validate a highway assignment model simply by comparing total regional vehicle miles traveled (VMT) to observed VMT. Aggregate validation must also be performed for all relevant market segments for which information can be obtained.

Any calibration of the transit assignment process may lead to model changes that affect mode choice, whether they are network changes, revisions to path building or skimming , or other changes to the model. The mode choice models cannot be considered validated until the transit assignment model has also been validated.

In disaggregate validation, model predictions are compared with observed data to reveal systematic biases. Disaggregate checks are appropriate for estimated models, as opposed to transferred models where the estimation data set would not be available. Logit models are disaggregately estimated (one record per trip/activity), and therefore disaggregate validation should be performed when logit mode choice models are estimated, along with the aggregate checks described above.

Generally, disaggregate validation is performed by applying the model using a data set with known choice results (such as a revealed-preference survey data set) and checking the results by one or more segmentation variables. Examples of segmentation variables include:

• Income level;
• Vehicle availability level;
• Geographic segmentation (e.g., counties, area types); and
• Trip length segments.

Disaggregate validation of a model ideally should be performed using a data set that is independent of the data set used for model estimation. However, most urban area household travel surveys have such small sample sizes that the entire data set is needed for model estimation and so there is no independent model estimation data set available for validation. This is especially true for mode choice models, where the household survey itself is often inadequate for model estimation due to low incidence of transit travel.

Limited disaggregate validation can be performed using the same data set used for model estimation, but reporting the results by market segment. Logit model estimation software has the capability to apply the estimated model to a data set in the same form as the estimation data set. For example, a logit mode choice model could be applied to the data set used for estimation but the results may be reported by vehicle availability or income level. It might be found, for example, that transit with auto access is being chosen too often in the model by households with zero vehicles.

(opens new window) provides examples of mode choice model parameters for level of service variables and relationships among parameters.

Sensitivity testing can be performed for mode choice models by varying model inputs and checking results for reasonableness. Model inputs that can be varied include level of service variables used in the trip distribution model (time/speed and cost) and the demographic or zone-level variables that are used as model inputs. Some example tests include:

• Increasing or decreasing highway or transit travel times by a fixed percentage regionwide;
• Increasing/decreasing parking costs in the CBD by a fixed percentage;
• Increasing/decreasing headways on selected transit routes or submodes by a fixed percentage or amount;
• Increasing/decreasing fares on selected transit submodes by a fixed percentage;
• Changing development patterns for forecast years by moving projected new activity among different parts of the modeled region (e.g., from suburbs to small urban centers or from outlying areas to infill); and
• Reallocating the number of households by income level for a forecast year.

The resultant changes in demand due to changes in a model input variable reflect the sensitivity to the variable; the sensitivity level is determined by the coefficient of the variable in the utility function. Simple “parametric” sensitivity tests can be performed by introducing small changes in the input variable or in the parameter itself and checking the results for reasonableness. The changes in demand for a modal alternative (or group of alternatives) with respect to a change in a particular variable can be expressed as arc elasticities.

# References

(opens new window)

← Destination choice models Network assignment →

This site uses cookies to learn which topics interest our readers.

National Geographic content straight to your inbox—sign up for our popular newsletters here

This Graphic Maps the Greenest Modes of Transportation

What's the most energy efficient way to go from New York City to Toronto? You might be surprised.

There are many ways to go from here to there.

And the mode of transportation we choose can have a big impact. It can affect time, comfort, cost, and other factors. But how we travel can also have a big impact on the planet.

In the U.S., the transportation sector is responsible for about 28 percent of total greenhouse gases, according to the EPA . That's the largest contributor, just edging out the energy sector.

And although cleaner alternatives are coming on board, such as electric cars and biofuels, over 90 percent of the fuel used for transportation in the U.S. remains derived from petroleum.

In this graphic, we break down the most efficient ways to travel on a typical trip:

• Nat Geo Expeditions

Become a subscriber and support our award-winning editorial features, videos, photography, and more—for as little as $2/mo.

Related Topics

• ENERGY EFFICIENCY
• TRANSPORTATION
• FOSSIL FUELS

You May Also Like

Thousands of oil rigs are nearing the end of their life—what will happen to them?

Why manatees often lurk close to Florida's power plants

What fuel made from the sun could do for the planet

2023 was our hottest year yet. What other milestones did we reach?

The essential guide to visiting Dublin

• Environment
• Paid Content

History & Culture

• History & Culture
• Terms of Use
• Privacy Policy
• Your US State Privacy Rights
• Children's Online Privacy Policy
• Interest-Based Ads
• About Nielsen Measurement
• Do Not Sell or Share My Personal Information
• Nat Geo Home
• Attend a Live Event
• Book a Trip
• Inspire Your Kids
• Shop Nat Geo
• Visit the D.C. Museum
• Learn About Our Impact
• Support Our Mission
• Advertise With Us
• Customer Service
• Renew Subscription
• Manage Your Subscription
• Work at Nat Geo
• Sign Up for Our Newsletters
• Contribute to Protect the Planet

Copyright © 1996-2015 National Geographic Society Copyright © 2015-2024 National Geographic Partners, LLC. All rights reserved

Planes, Trains, Cars and Buses: We Do the Math to Find the Cheapest Way to Travel Per Mile

See exactly how much each transportation mode costs for trips under 1,000 miles.

Most major American travel routes will give you multiple options for how to get there.

After the lean years of the early COVID-19 pandemic, travel bounced back in 2022 in a big way. More than 43% of American adults (about 112 million people), are planning to travel by airplane, car, bus or train this holiday season, per a Thanksgiving travel survey from The Vacationer.

If you are traveling a long distance, it's hard to beat air travel for convenience and price, but the considerations get trickier with shorter trips. Time spent in security and boarding could be spent cruising down the road (or rail). The number of people traveling is another big factor -- airplane, train and bus fares are all per person, while car prices generally stay the same when you add more passengers. 

Luckily for you, we did the math for traveling by airplane, car, train or bus for four of the most popular travel routes in America, and the results are enlightening.

Cost of airplanes, cars, trains and buses per mile

* Because of limited train service between Los Angeles and Las Vegas, and Atlanta and Orlando, the media cost per mile for trains was calculated using only the routes from New York to Chicago and Boston to Washington, DC. 

If you're traveling by yourself on a major airline route, it's always fastest to fly, and it's also often the cheapest method. Only bus fares dip below airfare for any of our routes, and the overall median cost per mile for flights easily beats automobile and train transportation too. However, if you've got a traveling companion or two (or if you need a car at your destination), driving becomes a more attractive option financially. And while buses might give a slight discount over airplanes, they'll take at least twice the time.

Read on to see how much each mode of travel costs for each of our four weekend journeys, and how to calculate the best way to travel to save money. For more We Do the Math tips, read how much you can save by shopping at Trader Joe's .

How we did the travel math for transportation costs

I used travel sites SkippedLag , momondo and Travelocity to find the lowest one-way airfares for afternoon travel on Friday, Jan. 13, 2023 (with the caveat that prices may vary greatly depending on time of year and day of the week). I used Amtrak for all one-way train fares, and I searched FlixBus , megabus and Greyhound for the lowest one-way bus fares using the same January date. I calculated per mile rates by dividing by the distance between cities per Google Maps .

For driving costs, I used AAA average gas prices on Nov. 16, 2023 for the states where gas fill-ups would be most likely, then calculated an exact number by simulating an average small American car with a 15-gallon gas tank getting 25.4 miles per gallon . After calculating the gas fees, I added in 9.68 cents times the number of miles for an additional maintenance cost, per AAA .

How much does it cost to fly, drive or take a bus or train from Los Angeles to Las Vegas?

It's about 270 miles from Los Angeles to Las Vegas, and the quick trip is one of the most popular flying routes in the country. The flight generally takes a little more than an hour (or 2.5 hours with airport security and boarding), while driving will take a little less than four hours at an average driving speed of 70 miles per hour. 

Using a departure date and time of Friday, Jan. 13, 2023 in the afternoon, we were able to find several flights for as low as $42. That fare is about half of what it would cost to drive ($84) using a current California gas price of $5.40 a gallon, while also including AAA's standard of 9.68 cents per mile for maintenance costs.

However, driving in a car is the only travel option that gets cheaper per person as more people travel with you. If you've got four people driving with you from LA to Las Vegas, you're suddenly each paying half of what it would cost to fly.

If you've got extra time, a bus can get you from Los Angeles to Las Vegas for half the price of flying.

Train fare from LA to Las Vegas isn't quite fair to compare, since there's no direct service. You can take an Amtrak train to Oxnard, California, and then transfer to a bus, but it will cost five times more than a standard bus. 

Speaking of buses, we found several $20 fares on both Greyhound and FlixBus . If you're looking for the absolutely cheapest way to get from Los Angeles to Las Vegas, it's hard to beat that bus trip, though it will take at least five hours.

How much does it cost to fly, drive or take a bus or train from Atlanta to Orlando, Florida?

The drive from Atlanta down to the home of Disney World is a popular route and a manageable 439 miles straight down I-75 till you veer left at Lake Panasoffkee . It's also the seventh busiest flight route in America, per Jetline Travel .

If you're looking to make that trip from Georgia to Florida, you have plenty of options for flights, trains and buses. By surveying Skiplagged, momondo and Travelocity, we found a few $49 flights for the afternoon of Friday, Jan. 13, 2023.

That's twice as cheap as driving, if you're going solo. Using the current average gas prices in Georgia and Florida, you'll pay $55.44 in gas, plus another $42.50 in maintenance costs. Bring a passenger with you, and your cost per person will be about exactly the same as flying.

Unfortunately, the train isn't much of an option for southern travel either. In order to take an Amtrak train from Atlanta to Orlando, you'd need to take a multi-day trip that goes northeast to Greensboro, North Carolina, and from there down to Orlando, more than doubling the travel distance. You can do it, but it will take two to three days and cost at least $169.

A Greyhound bus costs about the same as airfare: $49. However, while a typical flight from Atlanta to Orlando takes an hour and a half (or about 3.5 hours with security and transportation factored in), a bus will arrive in about eight hours.

How much does it cost to fly, drive or take a bus or train from New York City to Chicago?

Connecting the two biggest financial metropolitan areas of the eastern US, the New York to Chicago route is always among the most popular flights. Even though it's the longest trip in our comparison -- 790 miles -- the train rates aren't much more than flying. You can get a direct train from New York City to Chicago for the afternoon of Friday, Jan. 13, 2023 right now for $90 on Amtrak. The trip takes about 19 hours.

However, even that competitive train price can't beat flying. Flights out of New York airports on that Friday afternoon are currently hanging out at $86, and those flights will get you there in a little more than 2.5 hours (or about five hours considering security, boarding and transportation to and from the airport).

The bus won't help you much here. Current prices for Greyhound trips from New York to Chicago on a Friday afternoon are $110, and your trip will take just as long as the train. 

Driving solo might cost more than flying, but you can set your own pace and use your car at your destination.

And what about driving? It's pretty much a straight shot across six states on I-80, but the longer your trip, the more gas prices will eat at your wallet. Using average gas prices from New York, Ohio and Indiana, we calculate that you'd need to pay $174 in gas and $74 in maintenance costs for $248 total. You'd want at least two other people with you for the road trip to cost about the same as flying.

One of the biggest factors in deciding whether to fly or drive when traveling is whether or not you need a vehicle at your destination. Renting a car for the weekend in Chicago right now will cost you at least $300 and up to $1,000 or more for a large or luxury vehicle. Those prices for rental cars make it a financial no-brainer to drive if you need a car on your trip, and can afford the extra time to get there. (I'm not even gonna get into the calculus of Uber/Lyft/taxis vs. public transit vs. storing and parking your own car.)

Is there any US route where it's cheaper to take the train? What about Boston to Washington, DC?

When you're talking about the 440-mile journey from Boston to Washington, DC, the price of airfare and train fare are nearly identical. You can take the Acela -- Amtrak's fastest , with speeds up to 150 miles per hour -- from Boston to Washington in less than seven hours for a little more than $70.

If you're on the Acela line -- which has 14 stops in Massachusetts, Rhode Island, Connecticut, New York, New Jersey, Pennsylvania, Delaware, Maryland, and Washington, DC -- you'll find competitive train prices down to DC and back up to Boston.

Your flight from Boston to DC would cost about the same for a one-way ticket on a Friday afternoon and take about one hour and 40 minutes (or about four actual hours with airport transportation, security and boarding processes). It's certainly possible that you could find a train ticket for Boston to Washington, DC, that's cheaper than airfare. In fact, I found several for weekend and weekday dates.

Like all of the other routes, driving ends up costing a little more once you factor in the 9-plus cents per mile maintenance costs. Filling up in Massachusetts and then just a little gas in Maryland will get you from Boston to DC at $66 for gas and $43 for maintenance for a total of $109 -- definitely a bargain if you have traveling companions.

Super bargain hunters should keep an eye on FlixBus. The Greyhound competitor was offering a $59 rate from several Boston stations to DC at the time of this article. You'll pay for that $15 in savings, though, since the trip takes 10 to 11 hours.

The bottom line on the cheapest way to travel

When deciding on a mode of transport for a trip, you'll have a lot of personal factors to consider, but some constants emerged from our calculations. If you're traveling alone on a major route, it's hard to beat flying. Buses were only cheaper for two of our routes, and they generally took several hours longer.

Trains might provide a change of pace (and scenery), but they only really make sense if you're on the Acela line in the US Northeast. Taking a train at any longer distances will increase your travel time dramatically for no cost savings.

Driving a car makes the most sense if you have multiple passengers or need a car at your destination, but as your trip gets longer, flying becomes more and more financially advantageous.

To glean more insights into calculating financial decisions, read about how much more it costs to buy organic produce .

Which form of transport has the smallest carbon footprint?

How can individuals reduce their emissions from transport.

This article was first published in 2020. It was updated in 2023 with more recent data.

Transport accounts for around one-quarter of global carbon dioxide (CO 2 ) emissions from energy. 1 In some countries — often richer countries with populations that travel often — transport can be one of the largest segments of an individual’s carbon footprint.

If you need to travel locally or abroad, what is the lowest-carbon way to do so?

In this chart, we see the comparison of travel modes by their carbon footprint. These are measured by the amount of greenhouse gases emitted per person to travel one kilometer .

This data comes from the UK Government’s Department for Energy Security and Net Zero. It’s the emission factors companies use to quantify and report their emissions. While the overall rankings of transport modes will probably be the same, there may be differences across countries based on their electricity mix, vehicle stock, and public transport network.

Greenhouse gases are measured in carbon dioxide equivalents (CO 2 eq), accounting for non-CO 2 greenhouse gases and the increased warming effects of aviation emissions at high altitudes. 2

Walk, bike, or take the train for the lowest footprint

Over short to medium distances, walking or cycling is nearly always the lowest carbon way to travel. While they’re not in the chart, the carbon footprint of cycling one kilometer is usually in the range of 16 to 50 grams CO 2 eq per km depending on how efficiently you cycle and what you eat. 3

Using a bike instead of a car for short trips would reduce travel emissions by around 75%.

Public transport is usually your best option if you can’t walk or cycle. Trains are particularly low-carbon ways to travel. Taking a train instead of a car for medium-length distances would reduce emissions by around 80%. 4 Using a train instead of a domestic flight would reduce your emissions by around 86%. 5

In fact, if you took the Eurostar in France instead of a short-haul flight, you’d cut your journey’s footprint by around 97%. 6

What if you can’t walk or cycle and don’t have access to public transport?

If none of the above are options, what can you do?

Driving an electric vehicle (EV) is your best mode of private transport. It emits less than a petrol or diesel car, even in countries with a fairly high-carbon electricity mix. Of course, powering it from a low-carbon grid offers the greatest benefits.

The chart above only considers emissions of EVs during their use phase — when you’re driving. It doesn’t include emissions from car manufacturing. There have been concerns that when we account for the energy needed to produce the battery, an EV is actually worse for the climate than a petrol car. This is not true — while an EV does have higher emissions during its production, it quickly “pays back” once you start driving it. 7

The next best is a plug-in hybrid car.

Then, where you take a petrol car or fly depends on the distance. Flying has a higher carbon footprint for journeys less than 1000 kilometers than a medium-sized car. For longer journeys, flying would actually have a slightly lower carbon footprint per kilometer than driving alone over the same distance.

Let’s say you were to drive from Edinburgh to London, a distance of around 500 kilometers. You’d emit nearly 85 kilograms CO 2 eq. 8 If you were to fly, this would be 123 kilograms — an increase of almost one-third. 9

Some general takeaways on how you can reduce the carbon footprint of travel:

• Walk, cycle, or run when possible — this comes with many other benefits, such as lower local air pollution and better health;
• Trains are nearly always the winning option over moderate-to-long distances;
• If travelling internationally, going by train or boat is lower-carbon than flying;
• Electric vehicles are nearly always lower-carbon than petrol or diesel cars. The reductions are greatest for countries with a cleaner electricity mix;
• If traveling domestically, driving — even if it’s alone — is usually better than flying;
• Car-sharing will massively reduce your footprint — it also helps to reduce local air pollution and congestion.

Appendix: Why is the carbon footprint per kilometer higher for domestic flights than long-haul flights?

You will notice that domestic flights have higher CO2 emissions per passenger-kilometer than short-haul international flights, and long-haul flights have even slightly lower emissions. Why is this the case?

In its report on the CO 2 Emissions from Commercial Aviation , the International Council on Clean Transportation provides a nice breakdown of how the carbon intensity (grams CO 2 emitted per passenger kilometer) varies depending on flight distance. 10

This chart, with carbon intensity given as the red line, shows that at very short flight distances (less than 1,000 km), the carbon intensity is very high. It falls with distance until around 1,500 to 2,000 km, then levels out and changes very little with increasing distance.

This is because take-off requires much more energy input than a flight's “cruise” phase. So, for very short flights, this extra fuel needed for take-off is large compared to the more efficient cruise phase of the journey. The ICCT also notes that less fuel-efficient planes are often used for the shortest flights.

The IEA  looks at CO 2  emissions  from energy production alone — in 2018, it reported 33.5 billion tonnes of energy-related CO 2  [hence, transport accounted for 8 billion / 33.5 billion = 24% of energy-related emissions.

Aviation creates several complex atmospheric reactions at altitude, such as vapor contrails, creating an enhanced warming effect. In the UK’s Greenhouse gas methodology paper , a “multiplier" of 1.9 is applied to aviation emissions to account for this. This is reflected in the CO 2 eq factors provided in this analysis.

Researchers — David Lee et al. (2020) — estimate that aviation accounts for around 2.5% of global CO 2 emissions but 3.5% of radiative forcing/warming due to these altitude effects.

Lee, D. S., Fahey, D. W., Skowron, A., Allen, M. R., Burkhardt, U., Chen, Q., ... & Gettelman, A. (2020). The contribution of global aviation to anthropogenic climate forcing for 2000 to 2018 .  Atmospheric Environment , 117834.

Finding a figure for the carbon footprint of cycling seems like it should be straightforward, but it can vary quite a lot. It depends on several factors: what size you are (bigger people tend to burn more energy cycling), how fit you are (fitter people are more efficient), the type of bike you’re pedaling, and what you eat (if you eat a primarily plant-based diet, the emissions are likely to be lower than if you get most of your calories from cheeseburgers and milk). People often also raise the question of whether you actually eat more if you cycle to work rather than drive, i.e., whether those calories are actually ‘additional’ to your normal diet.

Estimates on the footprint of cycling, therefore, vary. Based on the average European diet, some estimates put this figure at around 16 grams CO2e per kilometer. In his book “ How bad are bananas: the carbon footprint of everything ”, Mike Berners-Lee estimates the footprint based on specific food types. He estimates 25 grams CO 2 e when powered by bananas, 43 grams CO 2 e from cereal and cow’s milk, 190 grams CO 2 e from bacon, or as high as 310 grams CO 2 e if powered exclusively by cheeseburgers.

National rail emits around 35 grams per kilometer. The average petrol car emits 170 grams. So the footprint of taking the train is around 20% of taking a car: [ 35 / 170 * 100 = 20%].

National rail emits around 35 grams per kilometer. A domestic flight emits 246 grams. So the footprint of taking the train is around 14% of a flight: [ 35 / 246 * 100 = 14%].

Taking the Eurostar emits around 4 grams of CO 2 per passenger kilometer, compared to 154 grams from a short-haul flight. So the footprint of  Eurostar is around 4% of a flight: [ 4 / 154 * 100 = 3%].

The “carbon payback time” for an average driver is around 2 years.

An average petrol car emits 170 grams per kilometer. Multiply this by 500, and we get 85,000 grams (85 kilograms).

A domestic flight emits 246 grams per kilometer. Multiply this by 500, and we get 123,000 grams (123 kilograms).

Graver, B., Zhang, K. & Rutherford, D. (2018). CO2 emissions from commercial aviation, 2018 . International Council on Clean Transportation.

Cite this work

Our articles and data visualizations rely on work from many different people and organizations. When citing this article, please also cite the underlying data sources. This article can be cited as:

BibTeX citation

Reuse this work freely

All visualizations, data, and code produced by Our World in Data are completely open access under the Creative Commons BY license . You have the permission to use, distribute, and reproduce these in any medium, provided the source and authors are credited.

The data produced by third parties and made available by Our World in Data is subject to the license terms from the original third-party authors. We will always indicate the original source of the data in our documentation, so you should always check the license of any such third-party data before use and redistribution.

All of our charts can be embedded in any site.

Our World in Data is free and accessible for everyone.

Help us do this work by making a donation.

• Ways to Give
• Contact an Expert
• Explore WRI Perspectives

Filter Your Site Experience by Topic

Applying the filters below will filter all articles, data, insights and projects by the topic area you select.

• All Topics Remove filter
• Climate filter site by Climate
• Cities filter site by Cities
• Energy filter site by Energy
• Food filter site by Food
• Forests filter site by Forests
• Freshwater filter site by Freshwater
• Ocean filter site by Ocean
• Business filter site by Business
• Economics filter site by Economics
• Finance filter site by Finance
• Equity & Governance filter site by Equity & Governance

Search WRI.org

Not sure where to find something? Search all of the site's content.

5 Shifts to Transform Transportation Systems and Meet Climate Goals

• transportation
• public transit
• climatewatch-pinned
• Urban Mobility
• data visualization

Transportation connects us to one another. It’s how we get to school and work, how we visit our families, and how we access our food and health care. It’s also how we ship goods and deliver services. As economies and populations grow, so does the need for efficient, accessible and sustainable transportation.

The current global transport system accounts for  15% of global greenhouse gas  emissions, which continue to grow. In 2019,  71% of transportation-related greenhouse gas emissions  came from roadways alone (with the rest primarily from maritime and aviation, and a small portion from rail and other sources). Transport has been a laggard for years, with the sector  falling behind others , like power and heating, in its decarbonization rate. To limit global warming to 1.5 degrees Celsius (2.7 degrees Fahrenheit) and prevent some of the worst impacts of climate change, we need to reverse course and drive transportation emissions down as low and as quickly as possible.

Our current global transport system accounts for 15% of global greenhouse gas emissions per year

In 2019, 72% of transport-related greenhouse gas emissions came from road transport alone, with the rest made up of shipping, aviation, and other sources

Shifting to electric vehicles will play an important role but transforming transportation will take real systemic change. What does a systems approach involve? Solutions must bring jobs and services closer to where people live as well as promote public transportation, walking, cycling and other low-carbon and clean-energy transportation alternatives. Finding new solutions to decarbonize shipping and aviation are also crucial.

There are five major shifts identified by  Systems Change Lab  that, if achieved together, can drastically reduce emissions and spark the necessary change for the planet and people to thrive:

1) Guarantee Reliable Access to Safe and Modern Mobility 

Future transportation systems — in addition to being low-carbon — must be safe, modern and center around improving health. For example, expanding the infrastructure around public transportation systems

with dedicated walkways or bike paths will not only combat vehicular congestion and reduce air pollution, but will also encourage more physical activity such as walking, cycling or using scooters.

A safer system will also prevent crashes and death associated with vehicle travel. About  17 out of every 100,000 people  in the world were killed on roadways in 2019, and  nearly half of them were pedestrians or cyclists . In 2021, the United Nations  set a target  to halve injuries and deaths from road traffic crashes by 2030. Achieving this goal requires concerted efforts to protect pedestrians, cyclists and drivers alike, such as improving street designs, creating protected bicycle lanes and enforcing traffic laws.

Transportation systems must also be affordable and accessible to all. This has not always been the case — for example, women face potential danger from  unlit bus stops  and potential harassment on public transit. This kind of consideration is often left out of transportation planning. Fortunately, some cities are already working to tackle these challenges — the city of Peshawar, Pakistan recently  launched a bus rapid transit system  that addresses common issues facing cisgender and transgender women as well as people with disabilities on public transit.

2) Reduce Avoidable Vehicle and Air Travel

Explore System Spotlights

This article is a part of a series profiling the major systems tracked by Systems Change Lab , a collaborative initiative — which includes an open-sourced data platform — that is designed to spur action at the pace and scale needed to limit global warming to 1.5 degrees Celsius, halt biodiversity loss and build a just and equitable economy.

Also in this series:

4 Shifts Needed to Create a Carbon-free Power System 

6 Shifts the Finance System Can Make to Build a Sustainable Future

Convened by WRI and the Bezos Earth Fund, Systems Change Lab supports the UN Climate Change High-Level Champions and works with key partners and funders including Climate Action Tracker (a project of New Climate Institute and Climate Analytics), ClimateWorks Foundation, Global Envirentonm Facility, Just Climate, Mission Possible Partnership, Systemiq, University of Exeter and the University of Tokyo’s Center for Global Commons, among others. Systems Change Lab is a component of the Global Commons Alliance.

The switch from internal combustion to low- and zero-carbon technologies is vital but also unlikely to happen fast enough to decarbonize the entire transportation sector at the necessary speed and scale. Even a complete electrification of cars, buses and trucks would pose challenges due to the increased electricity demand. We need solutions that will also help solve other problems such as traffic congestion, threats to pedestrian safety and inequitable transportation access.

Because of this, in parallel with the electrification process, we must also change the way we move around by limiting the most carbon-intensive forms of transportation, such as cars and airplanes.

Globally, the distance traveled by passenger cars is rising. This is especially noticeable in high-income, car-dependent places like Europe and North America. The  share of kilometers traveled by passenger cars  grew to 44% in 2020, and without intervention, this is expected to reach 50% in 2030. To reverse this trend, we must ensure people around the world have access to high-quality, safe alternatives such as high-speed rail or well-designed local bus systems. Bringing the share down to  34% to 44% by 2030  would help get the transportation system on track to keeping global warming below 1.5 degrees C.

To reduce car dependence, cities can also embrace  higher-density development , devising new ways to increase access to shops, services and leisure opportunities without needing vehicles. Cities can also disincentivize automobile travel through increased parking costs, fuel taxes or congestion-charging schemes.

However, these policies could hurt those who may not have the means to pay and who rely on cars as their only transportation option. Thus, policies to discourage car travel must be paired with sufficient alternatives, such as increasing safe public transportation options.

Strategies such as fuel taxes and frequent flier levies are also helpful to restrict airplane travel. For short-distance or moderate-distance flights, it is possible that as much as  15% of all regional trips  taken by plane could instead be served by high-speed rail. Increasing the quality and affordability of alternatives to plane travel, such as trains and ferries, is essential to boost ridership.

3)  Shift to Public, Shared and Non-Motorized Transport

Currently,  almost three quarters of transport carbon dioxide emissions  come from road travel —largely from cars, vans, buses and trucks. Convincing drivers to shift to more efficient modes will require fundamental cultural change around the car-centric design of many cities. This, alongside holding back car adoption in places where cars are not as prevalent, will be driven by investments that vastly improve other modes of travel.

Across the 50 highest-emitting cities,  rapid transit tracks and infrastructure  increased from about 13 kilometers (roughly 8 miles) per million people in 1990 to about 19 kilometers (or nearly 12 miles) in 2020. At the same time, the length of  high-quality bike lanes  increased by 6.5 kilometers (4 miles) per person from 2015 to 2020.

Growth in these indicators is a step in the right direction, but recent progress will have to accelerate by  6 times for rapid transit  and  more than 10 times  for bicycles by 2030, to help steer the world toward a 1.5 degree C-compatible scenario.

Alongside crucial emissions reductions, these changes will bring far-reaching benefits for public health and quality of life. A reduction in car usage could open more spaces for walking and cycling, public parks or outdoor dining and socializing. As of 2022, there are  11 countries with targets  to shift away from car travel and prioritize public transport — including India and China. As of 2019,  103 countries had plans  for walking and bicycling infrastructure.

4)  Transition to Zero-Carbon Cars, Trucks and Buses

We need to phase out fossil-fuel-powered vehicles as fast as possible. Fortunately, electric vehicles (EVs) provide a similar service without directly emitting carbon dioxide or air pollution.  While  EV uptake numbers are surging , it’s not fast enough.

To stay on track to limit global warming to 1.5 degrees C by 2030,  all new cars sold globally need to be electric .  In 2021,  only 8.7% of new cars  were electric. EV sales are  rising rapidly  thanks to improving economics and government support, and  dozens of countries  plan to end sales of gas and diesel cars by or before 2040. However, recent progress needs to  accelerate by 5 time s to meet global climate goals.

EVs’ upfront costs are falling, largely due to declining battery prices. It’s been expected to reach price parity with fossil-fuel-powered vehicles across Europe  by 2027  (although recent supply chain interruptions have likely  pushed this back  a few years). In some countries, like Germany and the Netherlands, it is already  cheaper to own and operate some types of  EVs than their fossil-fuel-powered equivalents. It is reasonable to assume that hitting the milestone of price parity will represent a  tipping point  that could contribute to accelerated growth in sales.

A  rollout of integrated charging networks  is needed to speed EV adoption. The  number of public EV chargers  around the world reached 1.8 million in 2021. Policies to push electrification are emerging in many countries. As of 2021, there were 18 jurisdictions — including Norway, the United Kingdom and Singapore — with  targets  to phase out the sales of fossil-fuel-powered cars. A complementary policy, a  zero-emission vehicle sales mandate , has shown up in 47 jurisdictions — including California, China, and the United Kingdom — as of 2022.

This zero-carbon transition must also occur for buses and trucks. Government ownership of many bus fleets can make them low hanging fruit in this respect because they can make decisions about large fleets with predictable schedules:  battery electric and hydrogen buses  made up 44% of global bus sales in 2021. To reach our climate goals, this number needs to be 100% by 2030. Progress has been uneven in recent years as most sales have taken place in China, but  increasing sales in other regions  like Europe and North America are likely to continue to push progress forward. Electric buses have surged in the United States over the past few years — a large purchase by a Midwest transit operator brought the  total  from just over 2,000 electric school buses in the third quarter of 2021 to just over 13,000 at the end of 2022.

Medium- and heavy-duty trucks are more difficult to decarbonize.  Just 0.2% of medium- and heavy-duty truck sales  were electric or hydrogen-powered in 2021, and there is an urgent need to bring technologies to commercial maturity. This will require more manufacturers offering  hydrogen  and  electric  options, more countries setting targets for phasing out gas-powered vehicles, and more places building charging infrastructure.

5)  Transition to Zero-Carbon Shipping and Aviation

Both shipping and aviation are seen as hard-to-decarbonize sectors, where zero-carbon technologies are still in infancy. Each is responsible for around  3% of global greenhouse emissions . 

However,  both   sectors  have pathways to a greener future. Decarbonizing shipping and aviation will require a combination of technological solutions such as zero-emission fuels and batteries alongside operational and efficiency improvements.

For shipping,  5% to 17% of fuel needs to be zero-emission by 2030  to stay on track to limit global warming to 1.5 degrees C. By 2050,  87% to 100%   of fuel needs to be zero-emission. Green hydrogen and ammonia, which can be made with renewable electricity, are typically viewed as the most promising fuels to decarbonize shipping. Synthetic fuels made from electricity, hydrogen and captured carbon may also play a part, and batteries could be useful for short-distance trips. Pilot projects will play a key role by helping to prove technical feasibility and demonstrate commercial viability — as of 2022, there were  over 200 pilot projects  underway.

For aviation, zero-emission options are beginning to emerge, including sustainable aviation fuels made from green hydrogen and captured carbon dioxide or sustainably sourced biomass. Biomass-derived sustainable aviation fuels are the only zero-emission solution commercially available in 2022, but new demonstrations of zero-emission planes are increasing. The  share of these fuels  needs to rise from less than 0.1% now to 13% to 18% by 2030 and 78% to 100% by 2050, supported by policy incentives.

It is key that zero-emission fuels are not derived from unsustainable biomass sources, such as food crops, which could  make it harder to feed  people, protect biodiversity and sequester carbon in natural ecosystems. Generally, a massive scaling up of investment and policy is needed in this shift.

Transforming a Global System

Together, these five shifts can transform our global transportation system. They offer a new system where opportunities and services are easily and equitably accessed through clean, safe mobility from walking to electric buses to bike share programs; a system where planes run on clean fuel and road crashes are not the  leading cause of death in children ; a system where high-speed trains are prevalent and popular.

Of course, this is only possible if we rapidly accelerate progress in these shifts to meet climate and equity targets in this decisive decade.

Relevant Work

5 ways to shape a greener, more equitable recovery through transport, decarbonizing freight: how u.s. policies and investments are reducing emissions in the sector, tracking climate action: how the world can still limit warming to 1.5 degrees c, 5 ways to cut oil and gas use through clean transportation, how you can help.

WRI relies on the generosity of donors like you to turn research into action. You can support our work by making a gift today or exploring other ways to give.

Stay Informed

World Resources Institute 10 G Street NE Suite 800 Washington DC 20002 +1 (202) 729-7600

© 2024 World Resources Institute

Envision a world where everyone can enjoy clean air, walkable cities, vibrant landscapes, nutritious food and affordable energy.

An Open Access Journal

• Open access
• Published: 25 September 2023

Factors that make public transport systems attractive: a review of travel preferences and travel mode choices

• Jessica Göransson   ORCID: orcid.org/0000-0002-5405-2872 1 , 2 , 3 &
• Henrik Andersson 4  

European Transport Research Review volume  15 , Article number:  32 ( 2023 ) Cite this article

7338 Accesses

8 Citations

Metrics details

Many regions worldwide are struggling to create a mode shift from private cars to more sustainable transport modes. While there are many reviews regarding travellers’ preferences and travel mode choices, there is a lack of an updated review that provides a comprehensive overview of the factors that make public transport systems attractive.

This review aims to fill the knowledge gap by offering insights into the factors influencing travel behaviour and the demand for public transport. It has two primary objectives: • Summarize general conclusions drawn from international literature reviews. • Present specific insights on the topic pertaining to the Nordic countries. To the best of our knowledge, this is the first review with a Nordic focus regarding public transport preferences and travel mode choices. The special focus on these countries is motivated by their relatively more ambitious policies for reducing emissions in the transport sector compared to many other countries, and their relatively high usage of public transport today.

To achieve these objectives, we conducted a review of existing literature. This review encompassed international literature reviews and included an examination of results from the Nordic countries.

The findings show that reliability and frequency are important factors for creating an attractive public transport supply. However, there is only limited evidence regarding the impact of improvements in these attributes on public transport demand, so this needs more research. This review highlights the importance of understanding the underlying motivations for travel mode choice and provides recommendations on areas for further investigation to understand the attractiveness of public transport supply.

1 Introduction

Transportation accounts for 24% of direct C0 2 emissions from fuel combustion worldwide, with road vehicles accounting for nearly 75% of this total [ 26 ]. To achieve the United Nations 2030 agenda for sustainable development and the Paris Agreement on climate change, actions need to be taken. The European Commission has responded to this need by introducing the European Green Deal to achieve climate neutrality by 2050. One crucial element of achieving this goal is a shift towards more sustainable transportation modes (European Environment [ 16 ]. Although travellers’ preferences and travel mode choices have been studied for decades, many regions around the world are struggling to create a mode shift from private cars to more sustainable travel modes.

Although there exist several reviews regarding travellers’ preferences and travel mode choices, to the best of our knowledge an up-to-date review that provides a broad perspective of what makes public transport (PT) systems attractive is missing. For example, Redman et al. [ 51 ] review what quality attributes of PT that attract car users, Hansson et al. [ 23 ] review preferences in regional PT systems, and Iseki and Taylor [ 29 ] review PT preferences concerning transfers. Therefore, by summarizing findings from existing international literature reviews, we will contribute to the literature by presenting some general conclusions and rules of thumb.

In addition, we will contribute by drawing insights from the Nordic countries. First, the goals for emission reduction in the transport sector in the European Union (EU) and other European countries differ with policies being more ambitious in the Nordic countries compared to the other EU countries [ 9 ]. Second, the Nordic countries, and especially the Scandinavian Footnote 1 ones have some common characteristics regarding PT: They have a relatively high use of PT today, an ambitious plan for the role that PT plays in emission reduction in the transport sector, and PT strategies that are in harmony with broader national and subnational objectives for economic development, land use planning, and social cohesion [ 52 ]. Nordic countries also have similarities in PT supply, population density, climate, norms, and socioeconomic factors, which allows us to treat the countries as one “global region”. Moreover, the Nordic countries outperform the rest of the world regarding The World Economic Forum´s Global Social Mobility Index , Footnote 2 which means that they have more equally shared opportunities compared to other countries [ 30 ]. We have not found any literature review that focuses on factors that make PT systems attractive in the Nordic countries, which is noteworthy as they have a relatively high use of PT today, high ambition levels regarding PT, and many similarities with each other. Hence, by reviewing results from the Nordic countries and by understanding what makes PT systems successful in the Nordic context we: (1) gain insights into the factors that contribute to the attractiveness of PT systems in these countries, and (2) facilitate comparisons and draw conclusions between the experiences and findings in the Nordic and other countries. The aim of this review is to provide insights into the current state of knowledge regarding what affects travel behaviour and travel demand for existing and potential PT users. This knowledge is important to create an attractive PT system, which enables a travel mode shift from car to PT. This review addresses two research questions: (1) What makes the public transport system attractive? and (2) What factors encourage us to travel more with public transport and less with private cars? These questions are addressed from two perspectives: one general and one with a focus on Nordic countries. We believe that this review provides good insight into the research about travellers’ preferences for someone new in the field, and that it will inspire new research in the area.

The structure of the article is as follows: In Sect.  2 , the examined literature and the search method are introduced. Section  3 provides a summary of the international and Nordic results relevant to research question (1) What makes the public transport system attractive? Similarly, Sect.  4 summarizes the international and Nordic results relevant to research question (2) What factors encourage us to travel more with public transport and less with private cars? Sect.  5 presents a discussion, followed by conclusions in Sect.  6 .

2 Literature examined

In this section, the examined literature and the search method are described. This section is divided into two subsections: International reviews and Nordic literature.

2.1 International reviews

A literature search was performed in August 2022 to identify previous reviews in the research area. In Table 1 , the four search strings used in Web of Science, Scopus and Transport Reviews are presented. The search in Web of Science was set as TOPIC, in Scopus as TITLE-ABS-KEY and in Transport Reviews as ALL. None of the searches were limited in timespan. The document type was restricted to “review”.

The search resulted in 190 findings. Titles and abstracts were screened. Articles were selected for further reading if they included travellers’ preferences or attitudes, and articles were excluded if they addressed only tourist travel. After relevant duplicates were removed (21 reviews), the screening process resulted in 28 reviews, and they were read in full. Reviews that did not follow the above criteria were excluded (9 reviews). In total, 19 reviews were found, which are summarized in Table 2 . The reviews consist of 20–130 international studies each. More than half of the reviews (10/19) were published in 2016 or later, and none of the reviews were published before 2000. In 16 of the studies, PT was included, car usage and ownership were included in eleven studies, and active modes (walking and bicycling) were included in ten studies. Four of the reviews included monetary values of attributes, three included travel demand effects, and two included travel mode shift effects.

2.2 Nordic literature

The literature search was performed in August 2022 to identify travel preferences, demand, and customer satisfaction studies in Nordic countries. In Table 3 , the four search strings used in Web of Science and Scopus are displayed. The search in Web of Science was set as TOPIC and in Scopus as TITLE-ABS-KEY. No limitation in timespan was used.

The search resulted in 336 findings. The screening process involved examining the title and abstract of articles, and those that included travel preferences or explaining factors that influence travel behaviour, travel demand, customer satisfaction or travel mode choice in Nordic countries were chosen for further reading. After relevant duplicates were removed (111 articles), 76 articles remained, and they were read in full. Articles were excluded if they only addressed preferences for different car types, active modes or the mode share between car and active modes (41 articles). After the complete screening, the search resulted in 35 articles, which are summarized in Table 4 . All articles contained quantitative data, while only two included in-depth interviews. The methods used to elicit preferences are broadly classified as belonging to two approaches: revealed preferences (RP) and stated preferences (SP). RP studies use individuals’ actual behaviour to elicit preference, whereas SP studies elicit preferences by asking individuals in hypothetical scenarios. The use of SP data and RP data in the articles were quite equal, with eleven SP studies and ten RP studies. Most of the articles with RP data included data from national travel surveys (7/10). Seven customer satisfaction studies were included in this review, a majority of which were from Sweden (5/7). Nine of the studies included monetary values, five included travel demand effects, and six included travel mode shift effects. Seventeen studies included results from Sweden, eleven from Norway, seven from Denmark, one from Finland, one from Iceland and one from Scandinavia as a group.

3 What makes PT systems attractive?

This section summarizes the international and Nordic research findings related to the research question mentioned in the section title. It is divided into five subsections. Each subsection begins with the international findings that reflect general knowledge on the subject. This is then followed by a subsection that provides a summary of insights from Nordic countries on the same topic.

3.1 Preferences regarding waiting time, transfer, and comfort

Studies show that travellers dislike walking, waiting and transferring more than in-vehicle time. A common rule of thumb is that walking and waiting time have twice the burden than in-vehicle time for nonbusiness trips, even if some studies find the burden to be higher and some studies find the burden to be lower. The burden is higher if the wait takes place in an unpleasant or threatening environment [ 29 ]. Studies show that PT users have a greater disutility for unexpected and unpredictable delays compared to expected and predictable waits. Studies also show that the disutility for transfers is higher than the disutility for waiting time [ 29 ]. Iseki and Taylor [ 29 ] and Diab et al. [ 14 ] refer to a review by Reed [ 50 ] that found that the disutility of waiting time ranges from 1.5 to 12 times that of the disutility in-vehicle time.

3.1.1 Nordic findings

Björklund and Swärdh [ 7 ] estimated policy values for comfort (getting a seat) and crowding reduction on board local PT in the three largest urban areas in Sweden, i.e., Stockholm, Gothenburg, and Malmö, by analysing SP data with 2003 participants. No geographical differences were found, which indicates that the same value for comfort and crowding can be used throughout Sweden. The value of the travel time savings (VTTS) multiplier for the worse scenario (standing, high level of crowding) was 2.9 compared to that of the reference scenario (sitting, low level of crowding). Sitting when there was a high level of crowding had a lower VTTS multiplier compared to standing with a low level of crowding. Differences were found depending on age, income, travel purpose and transport mode. For example, the willingness to pay (WTP) for sitting instead of standing was lower for tram passengers than for bus and commuter train passengers.

Pursula and Weurlander [ 48 ] analysed SP and RP data with 294 participants from Helsinki in Finland. The results showed that the disutility of two transfers was more than twice the disutility of one transfer and that the disutility of standing was higher than that of one transfer but less than that of two transfers.

Nielsen et al. [ 43 ] analysed RP data with 4810 observations in the Greater Copenhagen Region in Denmark to investigate whether the transfer penalty depends on transfer attributes, i.e., wayfinding, shopping availability, escalators and shelter. The transfer penalty was found to differ between 3.5 and 30 min of in-vehicle time, depending on the transfer attributes. The transfer penalty for one transfer varied from 5 min of in-vehicle time for the best possible transfer (easy wayfinding, shopping available and two escalators) to 12 min for the worst possible transfer (difficult wayfinding, no shops, and no escalators). Few observations included waiting times over ten minutes per transfer, which indicates that the participants dislike routes with long waiting times.

Vautard et al. [ 58 ] performed an SP study with 554 train passengers in Sweden and found that departure time adjustments were valued less than in-vehicle time. The time multipliers varied from 0.1 to 0.7. Passengers with high multipliers were nonflexible passengers, and passengers with low multipliers preferred a decreased travel time over a favourable departure time. Passengers with high multipliers were morning travellers, business travellers, passengers aged 45 or older, parents and middle-income travellers, whereas passengers with low multipliers were low-income travellers, passengers aged 25–44, females, two-person households and those without car access.

3.2 The impact of trip distance on preferences

Regional and local travellers have similar preferences with some differences. Attributes that are highlighted as important in many studies regarding regional PT are reliability, frequency, comfort, travel time and network coverage. The importance of frequency and reliability seems to decrease to some extent with longer travel distances, and the importance of comfort increases when travel time is longer. For regional PT, comfort is more important than frequency. Studies indicate that station facilities are more important than on-board comfort for regional trips shorter than 25 km [ 23 ].

3.2.1 Nordic findings

Mabit et al. [ 37 ] performed an SP study with 340 international travellers between Scandinavia and central Europe and found that VTTS decreased with trip distance and travellers’ duration of stay at the arrival point. They found that VTTS is not transferable from urban to long-distance international travel contexts. The results showed that VTTS decreased with distance for long-distance international journeys, while the literature on daily travel often shows that VTTS increases with distance. This was found by, e.g., Fröidh and Byström [ 19 ], who revealed that the importance of travel time for interregional journeys in Sweden increased with trip distance. Fröidh and Byström [ 19 ] also found that the importance of comfort and travel time increased with trip distance.

3.3 Preferences related to soft factors

Only Nordic findings were found regarding this topic.

3.3.1 Nordic findings

De Gruyter et al. [ 13 ] performed a meta-analysis, with results from Norway, Sweden, Australia, India, New Zealand and United Kingdom, regarding preferences for soft factors in the PT system. Soft factors included in the meta-analysis were divided into six categories: access (e.g., universal design and access to the station), facilities (e.g., ticket machines), security (e.g., lighting and staff), environment (e.g., noise and temperature control) and conditions (e.g., cleanliness). The results showed that preferences for soft factors in Norway and Sweden were much higher than in the other countries. They state that previous studies also show that Scandinavian countries value soft factors more highly for PT compared to other countries.

Fearnley et al. [ 18 ] analysed SP data from 408 Norwegian participants to estimate values for universal design in local PT. Universal design was defined as factors that make PT accessible to as many passengers as possible, e.g., seating, shelter, and information accessibility. They estimated WTP for five main categories: information at stop/stations, information on board, improved boarding, shelter, cleanliness and ice/snow removal. Each main category had 2–4 subcategories, e.g., shelter with and without a sitting place. The WTP for improved universal design showed a higher value compared to the time value (0.08–0.9 USD/minute compared to 0.07 USD/minute). The WTP for shelter with a sitting place at the bus stop (0.9 USD/minute) and ice/snow removal (0.88 USD/minute) was relatively high compared to the other attributes. The monetary values appear to be higher than those in previous international studies. This might be explained by the high standard of PT in urban areas in Norway and a large share of PT users, which means that these users have a relatively high income, which might lead to a higher WTP.

3.4 PT demand

In PT improvement studies, the attributes of reliability, frequency, travel time, price, comfort, access and convenience are commonly studied. PT demand is highly influenced by individuals’ previous experience of PT, their demographics, and socioeconomic factors [ 51 ]. For instance, high income increases the likelihood of owning a car [ 51 ], car access decreases PT demand [ 30 , 51 ] and the number of dependent children in the household increases car use [ 38 ].

Reliability [ 14 , 23 , 51 ] and frequency [ 23 , 51 ] are attributes with a strong impact on PT demand. If the PT has a low frequency or requires multiple transfers, the PT system fails to attract new users, and a mode shift from PT to private car can occur [ 6 ]. The Mohring effect, coined by the economist Herbert Mohring, posits that an increase in PT users leads to higher frequency, which leads to more passengers. Conversely, a decrease in PT users leads to a lower frequency, which leads to fewer passengers [ 32 ]. Moreover, prices [ 51 ] and travel time [ 40 , 51 ] also affect the PT demand. The effect that price changes have on demand is strongly influenced by other PT attributes such as frequency, travel time and access. Price changes can lead to an initial increase in demand, but the duration of the effect is affected by the quality attributes of PT [ 51 ].

Travellers compare the ticket price with their expectation of a reasonable price for the service they believe is provided [ 51 ]. Travellers’ knowledge and experience of the PT system influence their perception of the PT cost and the travel time [ 29 ], and studies show, e.g., that PT users tend to overestimate their waiting time at bus stops [ 14 , 29 ].

3.4.1 Nordic findings

In 2009, the train supply between Malmö and Gothenburg in Sweden improved due to the deregulation of interregional passenger rail services in Sweden, the ticket prices decreased, and the frequency increased. This led to a significant increase in train demand; the market share for trains compared to cars and airplanes increased from 21 to 28% between 2008 and 2010 on the route [ 19 ]. Fröidh and Byström [ 19 ] conducted an SP study onboard trains with different attributes between Malmö and Gothenburg to evaluate what affects travel mode choice between three different trains on the same route. The results showed that price, followed by travel time, had the highest impact on travel mode choice. Other factors evaluated were train types, train operator, and the quality of food and beverage services. The three trains attracted different travellers: the cheaper and slower train attracted younger passengers to a greater extent, whereas the more expensive trains with a shorter travel time attracted business passengers to a greater extent.

Findings in earlier SP studies show that there is a WTP for more reliable transport. Halse et al. [ 22 ] used RP data to examine whether this was also true in the Oslo metropolitan area in Norway, where there is a high level of competition between trains, cars and express buses. The results showed that train delays had a negative impact on PT demand; a 1% increase in average delay resulted in a 0.04% to 0.1% decrease in demand. The results are in line with previous results in the UK. The demand elasticity calculated in the RP study is lower than that in previous SP studies in Norway, which varied between -0.06 and -0.65. The study concludes that reliability has some effect on PT demand, but an improvement in reliability alone will not lead to a large increase in PT demand.

In 2021, an automated shuttle service was introduced at the Technical University of Denmark between the PT stop and campus as a complement to PT for the first-last mile. During the test period, Thorhauge et al. [ 56 ] conducted an SP study among 249 students and employers at the campus to analyse how improvements in the first-last mile trip affect the overall travel mode choice. The results show that automated shuttles do not have an overall effect on PT market share, but they might shift some existing PT users to use shuttle services.

3.5 Customer satisfaction

Factors that influence customer satisfaction in PT are mainly related to the travel experience [ 34 ]. Reliability [ 14 , 23 , 51 ] and frequency [ 23 , 51 ] are highlighted as factors with a high impact on customer satisfaction. The most frequently mentioned factors that influence customer satisfaction are on-board cleanliness, comfort, safety, behaviour of the personnel [ 23 , 34 ], reliability, frequency [ 34 ], travel time and price [ 23 ]. Factors that influence overall loyalty to the PT system are factors more associated with a trusting relationship between the user and the agency, e.g., the perception of value for money, on-board safety, cleanliness and interaction with personnel [ 34 ].

3.5.1 Nordic findings

Studies show that travel satisfaction is perceived differently by different groups. Börjesson and Rubensson [ 8 ] found that women in Stockholm, Sweden, rated crowding to be more important than men and that passengers over age 30 rated reliability to be more important than younger passengers. Cats et al. [ 11 ] and Abenoza et al. [ 2 ] found that pensioners/passengers older than 64 years were more satisfied than other travellers in Sweden. Ingvardson et al. [ 27 ] found indications that the younger generation in Denmark has a more negative attitude towards PT than the older generation. Cats et al. [ 11 ] found that frequent PT users in Sweden were more satisfied than other travellers and that passengers who travelled longer distances were less satisfied than passengers who travelled shorter distances. Julsrud and Denstadli [ 31 ] found that PT users in Norway had different expectations of using travel time productively depending on their media usage. Passengers who actively used mobile devices had higher expectations of using travel time more productively than those who used mobile devices more passively or not at all. The active mobile user group expressed the lowest customer satisfaction among the PT users.

Abenoza et al. [ 1 ], Cats et al. [ 11 ] and Börjesson and Rubensson [ 8 ] found that reliability and frequency are important attributes affecting customer satisfaction. Abenoza et al. [ 1 ] and Cats et al. [ 11 ] found that frequency, reliability, and travel time were more important attributes in Sweden than suitable PT lines, which indicates a higher preference for direct, punctual and frequent lines over many low-frequency lines that minimize transfers. Börjesson and Rubensson [ 8 ] found that customer satisfaction with crowding and reliability was affected by the actual performance of these attributes. They also found that reliability and frequency were the most important factors affecting customer satisfaction, unless there was a high level of crowding, crowding was the most important attribute. Tanko et al. [ 54 ] found that the factors calmness on the journey, punctuality, cleanliness, access (relative ease of access to boat piers for respondents) and frequency were highlighted as the most important attributes for water PT users in Stockholm County, Sweden. The respondents placed low importance on the factors related to the ability to work on board but were satisfied with how the factor performed.

4 What factors encourage us to travel more by PT and less by private cars?

This section summarizes the international and Nordic research findings related to the research question mentioned in the section title. It is divided into four subsections. Each subsection begins with the international findings that reflect general knowledge on the subject. This is then followed by a subsection that provides a summary of insights from Nordic countries on the same topic.

4.1 The influence that lifestyle, life stage and generation have on preferences

Life events, such as child birth, relocation and retirement, interrupt habits and provide a valuable opportunity to influence travel behaviour and travel mode shifts [ 3 , 38 , 51 ]. Studies show that it is more common for a household to reduce the number of cars when the household size decreases because of a divorce, a child moving out, or the death of one of the partners. A decrease in household income also affects the likelihood of reducing the number of cars, especially due to retirement or residential relocation [ 3 ].

Research on travel behaviour for different generations has increased in recent years. The results show that travel behaviour and travel preferences differ between generations. It is important to remember that generations are not a homogenous group even if some general conclusion can be made [ 30 ].

The car is the preferred mode of transport by many elderly [ 17 , 30 , 36 ]. The main reason for traveling by car is often reported as a lack of valid alternatives [ 17 , 36 ]. The most reported barriers for using PT are unsuitable routes, timetables, and scheduling. Other commonly mentioned factors are low reliability, the risk of having to stand, crowded vehicles, long walking distances, difficulties in understanding timetables, low accessibility, and affordability. PT demand is also affected by bus driver behaviour [ 36 ], and car ownership has a negative effect on PT demand [ 30 ].

The younger generation uses PT and active transport modes more than the older generation. They use multiple transport modes to a larger extent than other generations, e.g., a mix of cycling, driving and PT usage. The difference in travel behaviour between men and women is lesser for the younger generation compared to other generations. An increased income, having car access and holding a driver's licence increase car usage for this group [ 30 ].

Car is the preferred mode of transport by many families with young children. Car usage increases when the PT system has low accessibility, walking and cycling opportunities are low, or the number of dependent children in the household increases. Household income influences transport mode choice; higher income increases car usage, and lower income increases trips made by foot. PT cost is identified as a hindrance to the use of PT when travelling as a family [ 38 ].

Some studies have found a distinction in travel behaviour for adults and children/youths, e.g., a walk-friendly environment can be perceived differently by the two groups. Studies indicate that car access increases the likelihood for parents to drive their children to school. Moreover, the likelihood for a child/youth to be driven to and from school is affected by the parents’ preference for car usage. If parents perceive cars as a convenient and socially acceptable mode of transport, the child is more likely to be driven to and from school. As the distance between school and residence increases, children and youths are less likely to use active modes of transportation. The likelihood for the child/youth to walk and cycle to and from school increases with age and is affected by the perceived traffic safety and “walkability” on the route [ 39 ].

4.1.1 Nordic findings

Prato et al. [ 45 ] used data from the Danish National Travel Survey to evaluate preferences for short trips (< 22 km) in Copenhagen, Denmark. The results showed that lifestyle influenced travel mode choice and the perception of travel modes. Four heterogeneous groups were identified: car-oriented individuals, bicycle-oriented individuals, walking- and PT-oriented individuals, and PT-averse individuals . The groups had different perceptions of travel time; e.g., the car-oriented group evaluated 1 min of car travel as 2.6 min of cycling, whereas the bicycle-oriented group evaluated 1 min of cycling as 2.3 min of car travel. The perception of transfer penalties also differed between the groups, with the PT-averse group having the highest transfer penalty and the walking- and PT-oriented group having the lowest. Socioeconomic factors influenced which group an individual belonged to. For instance, individuals in the car-oriented group were mostly working men living with other adults, with high income and young children, whereas individuals in the walk- and PT-oriented group were mainly young female workers or students without children.

Thorhauge et al. [ 57 ] examined how travel mode choice is affected by trip complexity, activity participation, subjective constraints, and perceived mobility needs by creating a mode choice model based on RP data from Denmark. The results showed that the perceived mobility necessities were influenced by the number of daily activities and how flexible an individual was with arrival time to and from work. Individuals with high perceived mobility necessities (many daily activities and/or low flexibility in activities) were more likely to travel by car and bike and less likely to travel by PT. When using the model to predict the effect of a decrease in travel time for buses and an increase in travel time for cars, it was shown that individuals with high levels of perceived mobility necessities shifted from cars to bikes to a greater extent than those with lower perceived mobility necessities, who mostly shifted from cars to PT.

Rasca and Saeed [ 49 ] found that the probability of commuting by bus to work decreases when the respondents had a person in care, e.g., small children. Two Nordic studies were found that analysed children’s travel behaviour. An analysis of children's (aged between 6–12) travel independence in Norway was conducted by Fyhri and Hjorthol [ 20 ] using data from a survey of 1775 parents and their children. The study revealed variances in travel patterns between school trips and leisure trips, with 17% of school trips using PT and 25% being by car. Conversely, leisure trips predominantly rely on cars, with 66% of sport activity trips being made by car and only 1% by PT. The results showed that the child’s age and the distance to school had the greatest influence on their travel independence, and parental tendencies to frequently use cars raised the probability of driving their child to school. The most stated reasons for driving the child to school were “on the way to the parents’ work” (58%), followed by “dangerous traffic” (21%) and “most convenient” (18%), and the least stated reasons were “the child wants to be chauffeured” (12%), “have much to carry” (12%), “the way to school is unsafe for other reasons” (5%), and “friends are being driven” (2%).

By examining data from a survey of 245 parents in the Värmland region of Sweden, Westman et al. [ 59 ] analysed the choice of parents to transport their children (aged between 10 and 15) to school by car. According to the findings, social convenience, i.e., parents’ desire to spend time with the child (and driving them is perceived as the convenient way to do so), was the primary determining factor in deciding whether to drive them to school. The child's ability to travel independently also played a role in this regard. On the other hand, safety, security, and distance to school were not found to be statistically significant variables to predict whether parents would drive their children to school.

4.2 The impact that distance has on travel mode choices

4.2.1 nordic findings.

Empirical results from Nordic countries show that the distance between residence and workplace or city centre affects travel mode choice. Ahanchian et al. [ 4 ] found that the main competitor to car usage in Denmark differs depending on trip length: For short-distance (up to 25 km) trips metro, cycling and walking were the main competitors for cars, whereas for longer distances, trains and buses were the main competitors.

Pritchard and Frøyen [ 46 ] analysed data from a survey that asked 195 workers at a large company in Norway about their commuting behaviour before and after the company relocated from a location 10 km outside the city centre to the city centre. The results showed that relocation led to a decrease in commuting trips made by car/motorcycle (from 72 to 25%) and an increase in commuting by PT (from 12 to 32%) and by active modes (from 16 to 43%). The likelihood of travelling by car and PT increased when the distance between workplace and residence was greater than 7.5 km. Three similar case studies in Norway found that commuting by car decreased when companies relocated to a more central location. Naess et al. [ 42 ] analysed data from 1148 respondents from the Reykjavik capital region in Iceland and found that the probability of being a regular car commuter increased when the distance between residence and city centre increased. Nordfjærn et al. [ 44 ] performed a cross-sectional survey among 441 students on the two largest university campuses in Trondheim, Norway; one of the campuses is 2 km from the city centre and the other 6.5 km from the city centre. The results showed that a longer distance between the respondents’ residence and the university was correlated with more PT usage and less usage of active transportation modes. Isacsson et al. [ 28 ] created a mode choice model for Swedish men travelling to and from work by analysing RP data and employee-establishment data. They found that the likelihood for Swedish men to commute by PT compared to by car, motorcycle and active modes increased with an increased distance to work. Rasca and Saeed [ 49 ] analysed what affects the use of PT for employees in the region of Adger, Norway, by analysing data from a regional travel survey that consisted of 1849 respondents. The results show that the probability of commuting by PT increased with an increased distance between residence and workplace and that the probability of using PT increased when living five minutes or less from a PT stop with a frequency of at least 20 min between departures. They found that respondents with children were less willing to change from cars to PT.

4.3 Psychological factors and intention to use

4.3.1 nordic findings.

In 2018, a trial operation for a first-/last-mile automated bus service took place in Stockholm, Sweden. The automated buses were free of charge and operated on a 750-m route with flexible timetables from 6 a.m. to 6 p.m. every day. During the trial period, a survey was conducted with 574 passengers who lived, worked, or studied near the trial operation area to determine factors that influenced the intention to use the service. The results showed that frequency had the greatest impact on the intention to try the service, and comfort had the greatest impact on the intention to keep using the service [ 12 ].

Eriksson and Forward [ 15 ] examined how well an expanded version of the theory of planned behaviour predicts the intention to use cars, buses, and bicycles. They analysed data from a survey with 620 participants from Falun, Sweden. The results showed that attitude, subjective norms, and perceived behavioural control explained 48% of the intention to use a car, 41% of the intention to use a bus and 38% of the intention to use a bicycle. When car access was included in the model, the model better predicted the intention to use the different modes. Car access had a negative influence on the intention to use buses and bicycles. Car drivers were less willing to use other transport modes than bus and bicycle users.

Lind et al. [ 35 ] examined how travel mode choices are affected by the relative importance of situational factors and personal norms by analysing data from a survey with 1043 participants in urban areas in Norway. The results showed that socioeconomic factors, personal norms, values and beliefs affected travel mode choice and that values and beliefs explained 58% of the variance in personal norms. For example, the participants who stated a strong feeling of moral obligation to sustainable travel modes more often used PT, walked, or cycled, whereas those who stated a low feeling of this moral obligation more often used cars.

Mobility as a Service (MaaS) is an approach to attract car users to more sustainable travel habits. Strömberg et al. [ 53 ] analysed data based on 151 local travellers who participated in a MaaS trial for six months in Gothenburg, Sweden. The participants completed three online questionnaires (before, during and after the trial). The results showed that 42% of the participants reported a behavioural change regarding travel mode choice, and 36% of the participants did not report a behavioural change. Four subgroups were identified that differed depending on socioeconomic factors, motivation to join the trial and expectations of the trial. Participants who used cars before the trial reduced their car usage and increased the usage of more sustainable travel modes. The results showed that participants had difficulties predicting how their own behaviour would change before the trial since they had little knowledge about their preconditions, travel needs and behaviour when joining the trial. The fact that individuals have difficulties in predicting their own behavioural changes indicates the complexity of predicting behaviours.

Andersson et al. [ 5 ] conducted an SP study in Sweden to evaluate how marketing messages motivate a mode shift from car to sustainable transport. The results showed that environment and health messages motivated more than economic and status-related messages and that messages focusing on collective efficacy elicited higher motivations than messages focusing on self-efficacy. The marketing messages had different effectiveness in different groups, which highlights the importance of adapting the message to fit the preferences and behaviour of the selected target group. The results suggested that individuals’ preferences and current behaviour affect their responses to the messages. In line with previous studies, campaigns did not seem to affect devoted drivers and should therefore focus on other target groups who more open to changes [ 5 ].

4.4 Car users

There is no simple solution to attract car users to PT since they are not a homogenous group. To attract them to PT, it is important to understand the underlying motivation for car use for that specific target group [ 3 , 23 , 38 , 51 ]. An example of a target group can be middle-class families with young children in a specific suburb who work in the closest large city. Several of the reviews directly or indirectly state that more research is needed to understand how to attract car users to the  PT system [ 3 , 36 , 38 , 41 , 47 , 51 ].

As previously mentioned, PT demand is strongly influenced by reliability and frequency. Since car users already have these qualities in their current travel mode, a PT system with high punctuality and high frequency is not enough to create a mode shift. To attract car users to the PT system, it must provide a cost-competitive alternative to the car with basic levels of accessibility and reliability together with attributes viewed as important by the target group. It is important to show car users the benefits they can obtain by travelling by PT. Habit-interrupting transport policies and reduced PT prices can lead to an initial mode shift. However, the duration of the effect is affected by how the PT system is perceived [ 51 ].

4.4.1 Nordic findings

Isacsson et al. [ 28 ], Eriksson and Forward [ 15 ], Pritchard and Frøyen [ 46 ], Nordfjærn et al. [ 44 ], Rasca and Saeed [ 49 ], Thogersen et al. [ 55 ] and Hjorthol et al. [ 25 ] found evidence suggesting that car access has a negative effect on PT usage and/or a positive effect on car usage. Rasca and Saeed [ 49 ] found that if the respondent had difficulty finding a parking spot, the likelihood of using PT to commute to work is higher. Thogersen et al. [ 55 ] analysed data from 2607 commuters in Norway to investigate why they drive conventional cars and not more climate-friendly alternatives to and from work/school. Their findings showed that a greater PT frequency reduces the likelihood of travelling by car, while an increased requirement for transfers in PT increases the likelihood of travelling by car.

Using data from the Danish National Transport Survey (2010–2015; 29 089 journeys) and information from the Danish National Transport Model, Ahanchian et al. [ 4 ] created a model to predict the modal shift effect in Denmark. Different scenarios were tested in the model to analyse how three different transport policies would affect travel demand compared to the reference scenario in 2050. The results showed that an increase in the cost of travelling by car had the highest effect on reduced car use (− 30%), followed by reduced costs for sustainable travel modes (− 19%) and expansion of PT infrastructure (− 7%). The greatest effect was found when all three policies were adopted (− 49%). The results also showed that the easiest group to influence was the low-income group.

5 Discussion

The aim of this review was to provide insights into the current state of knowledge regarding what affects travel behaviour and travel demand for existing and potential PT users. This was done by reviewing existing international reviews and reviewing results from Nordic counties. The restriction to published international reviews is considered to capture the general knowledge and rules of thumb in a comprehensive way. By limiting our review to high-quality, English-language articles that target an international audience, we can ensure that we are summarizing the most relevant and valuable information available. However, some may argue that we risk missing out on relevant articles written in other languages.

No major contradictions in the results were found between the international and Nordic studies. However, certain empirical evidence from Nordic studies was not analysed in international studies. Only the Nordic studies examined preferences related to soft factors, the impact that distance has on travel mode choices, psychological factors and intention to use, and crowding. However, only the international review examined travel preferences for elderly individuals. Most studies included in this review focused on specific contexts such as a municipality, a region, or a country. Only one Nordic study examined how stable preferences were between contexts: Björklund and Swärdh [ 7 ] analysed how stable preferences for comfort and crowding were between the three largest cities in Sweden. It is noteworthy that Sweden's data constitute half of the Nordic findings, whereas Finland and Iceland contributed only one study each.

The results show that travellers dislike walking, waiting and transferring more than in-vehicle time. The burden is higher if the wait takes place in an unpleasant or threatening environment. From an international perspective, Hansson et al. [ 23 ] found indications that station facilities were more important than on-board comfort for shorter regional trips. In the Danish context, Nielsen et al. [ 43 ] found that transfer penalties depend on transfer attributes, with better station standards leading to less disutility for transfers. In the Finnish context, Pursula and Weurlander [ 48 ] found that the disutility of two transfers was more than twice the disutility of one transfer and that the disutility of standing was higher than one transfer but less than two transfers. No Nordic study was found that compared improvements in PT stops/stations versus more classical PT improvements, such as lower travel times or higher reliability.

Travellers have a greater disutility for unexpected and unpredictable delays compared to expected and predictable waits. From an international perspective, Diab et al. [ 14 ], Hansson et al. [ 23 ] and Redman et al. [ 51 ] concluded that reliability has a strong impact on PT demand. Only one Nordic study was found that evaluated the impact reliability has on PT demand: Halse et al. [ 22 ] concluded that reliability has some effect on PT demand in Norway, but an improvement in reliability alone will not lead to a large increase in PT demand. The demand elasticity calculated in the RP study is lower than in previous SP studies in Norway. Hansson et al. [ 23 ] and Redman et al. [ 51 ] similarly found that frequency has a strong impact on PT demand. In Norway, Thogersen et al. [ 55 ] found that a greater PT frequency reduces the likelihood of travelling by car. Empirical findings from both international [ 23 , 51 ] and Nordic [ 1 , 8 , 11 ] studies show that reliability and frequency are important factors affecting customer satisfaction.

Evidence from both the international and Nordic perspectives shows that preferences are heterogeneous, and travellers can be classified into distinct subgroups based on their characteristics and preferences. Therefore, improvements and campaigns will have different efficiencies in different subgroups, and the answers to the research questions depend on which subgroup is of interest. While there may not be an easy answer for the two research questions, we believe that it is valuable to gain knowledge from previous research to create a deeper understanding of the factors that influence travel behaviour.

6 Conclusions

The empirical findings show that reliability and frequency are important attributes for creating an attractive PT supply. However, the extent to which improvements in these attributes affect PT demand remains uncertain. Notably, potential PT users have high levels of reliability and flexibility in their current travel modes. Not degrading the current level of reliability and frequency is important to keep existing users, and a high level of reliability and frequency is crucial to make PT a reasonable travel mode for potential users. Car users are not a homogenous group, and to attract them to PT, it is important to understand the underlying motivation for their current travel mode choice. Life events interrupt habits and provide a valuable opportunity to influence travel behaviour and travel mode shifts. Habit-interrupting transport policies and reduced PT prices can lead to an initial mode shift effect. However, the duration of the effect is affected by how the PT system is perceived. To attract car users to the PT system, it is important to show them the benefit they can receive by travelling by PT.

Only one study was found that examined the stability of preferences across different contexts. If preferences remain stable across contexts, then preferences found in one context can be applied to other contexts. Assuming identical preferences across regions or locations can result in suboptimal PT planning that does not reflect the preferences and needs of the local populations. From a policy perspective, it is therefore recommended to further examine how stable preferences are between different contexts since findings from one study/context are often used for other contexts, especially within countries.

In line with many of the previous reviews, we also acknowledge the need to create a deeper understanding of the underlying motivations for travel mode choice for potential PT users. Additionally, it is recommended to investigate the extent to which PT stops or stations contribute to the attractiveness of the PT supply. From a policy perspective, it would be valuable to understand when investing in PT stops or stations is more beneficial than investing in traditional PT improvements such as reduced travel time or improved reliability. On a similar note, more research is recommended to explore the impact of reliability and frequency on PT demand. We believe these recommendations would improve input parameters for PT planning, thus enabling planners to invest in the most effective PT improvements to increase the attractiveness of PT.

Availability of data and materials

All reviewed articles are available in Web of Science, Scopus or Transport Reviews.

The Scandinavian countries, i.e., Denmark, Norway, and Sweden, are part of the Nordic countries. The Nordic countries include Finland and Iceland as well.

Economies with greater social mobility provide more equally shared opportunities that are independent regarding socioeconomic background, geographic location, gender, or origin. The index is based on the performance of five dimensions: health, education, technology, work, protection and institutions.

Abenoza, R. F., Cats, O., & Susilo, Y. O. (2017). Travel satisfaction with public transport: Determinants, user classes, regional disparities and their evolution. Transportation Research Part A: Policy and Practice, 95 , 64–84. https://doi.org/10.1016/j.tra.2016.11.011

Article   Google Scholar  

Abenoza, R. F., Liu, C., Cats, O., & Susilo, Y. O. (2019). What is the role of weather, built-environment and accessibility geographical characteristics in influencing travelers’ experience? Transportation Research Part A: Policy and Practice, 122 , 34–50. https://doi.org/10.1016/j.tra.2019.01.026

Aguilera, A., & Cacciari, J. (2020). Living with fewer cars: Review and challenges on household demotorization. Transport Reviews, 40 , 796–809. https://doi.org/10.1080/01441647.2020.1772405

Ahanchian, M., Gregg, J. S., Tattini, J., & Karlsson, K. B. (2019). Analyzing effects of transport policies on travelers’ rational behaviour for modal shift in Denmark. Case Studies on Transport Policy, 7 , 849–861. https://doi.org/10.1016/j.cstp.2019.07.010

Andersson, A., Winslott Hiselius, L., & Adell, E. (2020). The effect of marketing messages on the motivation to reduce private car use in different segments. Transport Policy, 90 , 22–30. https://doi.org/10.1016/j.tranpol.2020.02.006

Ben-Akiva, M., & Morikawa, T. (2002). Comparing ridership attraction of rail and bus. Transport Policy, 9 , 107–116. https://doi.org/10.1016/S0967-070X(02)00009-4

Björklund, G., & Swärdh, J.-E. (2017). Estimating policy values for in-vehicle comfort and crowding reduction in local public transport ☆ . Transportation Research Part A: Policy and Practice, 106 , 453–472. https://doi.org/10.1016/j.tra.2017.10.016

Börjesson, M., & Rubensson, I. (2019). Satisfaction with crowding and other attributes in public transport. Transport Policy, 79 , 213–222. https://doi.org/10.1016/j.tranpol.2019.05.010

Calmfors, L., Hassler, J., Nasiritousi, N., Bäckstrand, K., Silbye, F., Birch Sørensen, P., et al. (2019). Nordic Economic Policy Review 2019: Climate Policies in the Nordics . Nordic Council of Ministers.

Carlsson, F. (2003). The demand for intercity public transport: The case of business passengers. Applied Economics, 35 , 41–50. https://doi.org/10.1080/00036840210158921

Cats, O., Abenoza, R. F., Liu, C., & Susilo, Y. O. (2015). Evolution of satisfaction with public transport and its determinants in Sweden: Identifying priority areas. Transportation Research Record, 2538 , 86–95. https://doi.org/10.3141/2538-10

Chee, P. N. E., Susilo, Y. O., & Wong, Y. D. (2020). Determinants of intention-to-use first-/last-mile automated bus service. Transportation Research Part A: Policy and Practice, 139 , 350–375. https://doi.org/10.1016/j.tra.2020.06.001

De Gruyter, C., Currie, G., Truong, L. T., & Naznin, F. (2019). A meta-analysis and synthesis of public transport customer amenity valuation research. Transport Reviews, 39 , 261–283. https://doi.org/10.1080/01441647.2018.1461708

Diab, E. I., Badami, M. G., & El-Geneidy, A. M. (2015). Bus transit service reliability and improvement strategies: Integrating the perspectives of passengers and transit agencies in North America. Transport Reviews, 35 , 292–328. https://doi.org/10.1080/01441647.2015.1005034

Eriksson, L., & Forward, S. E. (2011). Is the intention to travel in a pro-environmental manner and the intention to use the car determined by different factors? Transportation Research Part D: Transport and Environment, 16 , 372–376. https://doi.org/10.1016/j.trd.2011.02.003

European Environment Agency. (2021). Transport and environment report 2020: train or plane? Publications Office, LU.

Fatima, K., Moridpour, S., De Gruyter, C., & Saghapour, T. (2020). Elderly sustainable mobility: Scientific paper review. Sustainability, 12 , 7319. https://doi.org/10.3390/su12187319

Fearnley, N., Flügel, S., & Ramjerdi, F. (2011). Passengers’ valuations of universal design measures in public transport. Research in Transportation Business & Management, Accessibility in passenger transport: Policy and management, 2 , 83–91. https://doi.org/10.1016/j.rtbm.2011.07.004

Fröidh, O., & Byström, C. (2013). Competition on the tracks—Passengers’ response to deregulation of interregional rail services. Transportation Research Part A: Policy and Practice, 56 , 1–10. https://doi.org/10.1016/j.tra.2013.09.001

Fyhri, A., & Hjorthol, R. (2009). Children’s independent mobility to school, friends and leisure activities. Journal of Transport Geography, 17 , 377–384. https://doi.org/10.1016/j.jtrangeo.2008.10.010

Gunther, M., Jacobsen, B., Rehme, M., Goetze, U., & Krems, J. F. (2020). Understanding user attitudes and economic aspects in a corporate multimodal mobility system: Results from a field study in Germany. European Transport Research Review, 12 , 64. https://doi.org/10.1186/s12544-020-00456-0

Halse, A. H., Østli, V., & Killi, M. (2020). Revealed and stated preferences for reliable commuter rail in Norway. Transportation Letters, 12 , 288–292. https://doi.org/10.1080/19427867.2019.1586088

Hansson, J., Pettersson, F., Svensson, H., & Wretstrand, A. (2019). Preferences in regional public transport: A literature review. European Transport Research Review, 11 , 38. https://doi.org/10.1186/s12544-019-0374-4

Heinen, E., & Bohte, W. (2014). Multimodal commuting to work by public transport and bicycle: Attitudes toward mode choice. Transportation Research Record, 2468 , 111–122. https://doi.org/10.3141/2468-13

Hjorthol, R. J., Levin, L., & Sirén, A. (2010). Mobility in different generations of older persons: The development of daily travel in different cohorts in Denmark, Norway and Sweden. Journal of Transport Geography, Special issue: The mobility of older people, 18 , 624–633. https://doi.org/10.1016/j.jtrangeo.2010.03.011

IEA. (2020). Tracking Transport 2020, IEA, Paris https://www.iea.org/reports/tracking-transport-2020 . Accessed 2021 May 30.

Ingvardson, J. B., Kaplan, S., de Abreu e Silva, J., di Ciommo, F., Shiftan, Y., & Nielsen, O. A. (2020). Existence, relatedness and growth needs as mediators between mode choice and travel satisfaction: evidence from Denmark. Transportation, 47 , 337–358. https://doi.org/10.1007/s11116-018-9886-3

Isacsson, G., Karlström, A., & Swärdh, J.-E. (2013). The value of commuting time in an empirical on-the-job search model—An application based on moments from two samples. Applied Economics, 45 , 2827–2837. https://doi.org/10.1080/00036846.2012.678981

Iseki, H., & Taylor, B. D. (2009). Not all transfers are created equal: Towards a framework relating transfer connectivity to travel behaviour. Transport Reviews, 29 , 777–800. https://doi.org/10.1080/01441640902811304

Jamal, S., & Newbold, K. B. (2020). Factors associated with travel behavior of millennials and older adults: A scoping review. Sustainability, 12 , 8236. https://doi.org/10.3390/su12198236

Julsrud, T. E., & Denstadli, J. M. (2017). Smartphones, travel time-use, and attitudes to public transport services. Insights from an explorative study of urban dwellers in two Norwegian cities. International Journal of Sustainable Transportation, 11 , 602–610. https://doi.org/10.1080/15568318.2017.1292373

Li, X., & Yang, H. (2016). Dynamics of modal choice of heterogeneous travelers with responsive transit services. Transportation Research Part C: Emerging Technologies, 68 , 333–349. https://doi.org/10.1016/j.trc.2016.04.014

Li, Z., & Hensher, D. A. (2012). Congestion charging and car use: A review of stated preference and opinion studies and market monitoring evidence. Transport Policy, Urban Transport Initiatives, 20 , 47–61. https://doi.org/10.1016/j.tranpol.2011.12.004

van Lierop, D., Badami, M. G., & El-Geneidy, A. M. (2018). What influences satisfaction and loyalty in public transport? A review of the literature. Transport Reviews, 38 , 52–72. https://doi.org/10.1080/01441647.2017.1298683

Lind, H. B., Nordfjærn, T., Jørgensen, S. H., & Rundmo, T. (2015). The value-belief-norm theory, personal norms and sustainable travel mode choice in urban areas. Journal of Environmental Psychology, 44 , 119–125. https://doi.org/10.1016/j.jenvp.2015.06.001

Luiu, C., Tight, M., & Burrow, M. (2018). Factors preventing the use of alternative transport modes to the car in later life. Sustainability, 10 , 1982. https://doi.org/10.3390/su10061982

Mabit, S. L., Rich, J., Burge, P., & Potoglou, D. (2013). Valuation of travel time for international long-distance travel—Results from the Fehmarn Belt stated choice experiment. Journal of Transport Geography, 33 , 153–161. https://doi.org/10.1016/j.jtrangeo.2013.10.003

McCarthy, L., Delbosc, A., Currie, G., & Molloy, A. (2017). Factors influencing travel mode choice among families with young children (aged 0–4): A review of the literature. Transport Reviews, 37 , 767–781. https://doi.org/10.1080/01441647.2017.1354942

Mitra, R. (2013). Independent mobility and mode choice for school transportation: A review and framework for future research. Transport Reviews, 33 , 21–43. https://doi.org/10.1080/01441647.2012.743490

Molander, S., Fellesson, M., Friman, M., & Skalen, P. (2012). Market orientation in public transport research—A review. Transport Reviews, 32 , 155–180. https://doi.org/10.1080/01441647.2011.633248

Muñoz, B., Monzon, A., & Daziano, R. A. (2016). The increasing role of latent variables in modelling bicycle mode choice. Transport Reviews, 36 , 737–771. https://doi.org/10.1080/01441647.2016.1162874

Naess, P., Stefansdottir, H., Peters, S., Czepkiewicz, M., & Heinonen, J. (2021). Residential location and travel in the Reykjavik Capital Region. Sustainability, 13 , 6714. https://doi.org/10.3390/su13126714

Nielsen, O. A., Eltved, M., Anderson, M. K., & Prato, C. G. (2021). Relevance of detailed transfer attributes in large-scale multimodal route choice models for metropolitan public transport passengers. Transportation Research Part A: Policy and Practice, 147 , 76–92. https://doi.org/10.1016/j.tra.2021.02.010

Nordfjærn, T., Egset, K. S., & Mehdizadeh, M. (2019). “Winter is coming”: Psychological and situational factors affecting transportation mode use among university students. Transport Policy, 81 , 45–53. https://doi.org/10.1016/j.tranpol.2019.06.002

Prato, C. G., Halldórsdóttir, K., & Nielsen, O. A. (2017). Latent lifestyle and mode choice decisions when travelling short distances. Transportation, 44 , 1343–1363. https://doi.org/10.1007/s11116-016-9703-9

Pritchard, R., & Frøyen, Y. (2019). Location, location, relocation: How the relocation of offices from suburbs to the inner city impacts commuting on foot and by bike. European Transport Research Review, 11 , 14. https://doi.org/10.1186/s12544-019-0348-6

Pronello, C., & Gaborieau, J.-B. (2018). Engaging in pro-environment travel behaviour research from a psycho-social perspective: A review of behavioural variables and theories. Sustainability, 10 , 2412. https://doi.org/10.3390/su10072412

Pursula, M., & Weurlander, M. (1999). Modeling level-of-service factors in public transportation route choice. Transportation Research Record, 1669 , 30–37. https://doi.org/10.3141/1669-04

Rasca, S., & Saeed, N. (2022). Exploring the factors influencing the use of public transport by commuters living in networks of small cities and towns. Travel Behav. Soc., 28 , 249–263. https://doi.org/10.1016/j.tbs.2022.03.007

Reed, T. B. (1995). Reduction in the burden of waiting for public transit due to real-time schedule information: A conjoint analysis study . Paper presented at the 6th Vehicle Navigation and Information Systems Conference (1995, Seattle, WA) at Piscataway, NJ. Refereed in Iseki and Taylor (2009) and Diab et al. (2015).

Redman, L., Friman, M., Garling, T., & Hartig, T. (2013). Quality attributes of public transport that attract car users: A research review. Transport Policy, 25 , 119–127. https://doi.org/10.1016/j.tranpol.2012.11.005

Rye, T., Velde, D., Wretstrand, A., Nordheim, B., Sørensen, C., & dellenbag, J. (2017). The Scandinavian way to better public transport . Available at: https://www.researchgate.net/publication/343181207_The_Scandinavian_way_to_better_public_transport

Strömberg, H., Karlsson, I. C. M., & Sochor, J. (2018). Inviting travelers to the smorgasbord of sustainable urban transport: Evidence from a MaaS field trial. Transportation, 45 , 1655–1670. https://doi.org/10.1007/s11116-018-9946-8

Tanko, M., Cheemarkurthy, H., Hall Kihl, S., & Garme, K. (2019). Water transit passenger perceptions and planning factors: A Swedish perspective. Travel Behaviour and Society, 16 , 23–30. https://doi.org/10.1016/j.tbs.2019.02.002

Thogersen, J., Vatn, A., Aasen, M., Dunlap, R. E., Fisher, D. R., Hellevik, O., & Stern, P. (2021). Why do people continue driving conventional cars despite climate change? Social-psychological and institutional insights from a survey of Norwegian commuters. Energy Research & Social Science, 79 , 102168. https://doi.org/10.1016/j.erss.2021.102168

Thorhauge, M., Fjendbo Jensen, A., & Rich, J. (2022). Effects of autonomous first- and last mile transport in the transport chain. Transportation Research Interdisciplinary Perspectives, 15 , 100623. https://doi.org/10.1016/j.trip.2022.100623

Thorhauge, M., Kassahun, H. T., Cherchi, E., & Haustein, S. (2020). Mobility needs, activity patterns and activity flexibility: How subjective and objective constraints influence mode choice. Transportation Research Part A: Policy and Practice, 139 , 255–272. https://doi.org/10.1016/j.tra.2020.06.016

Vautard, F., Liu, C., Fröidh, O., & Byström, C. (2021). Estimation of interregional rail passengers’ valuations for their desired departure times. Transport Policy, 103 , 183–196. https://doi.org/10.1016/j.tranpol.2021.02.005

Westman, J., Friman, M., & Olsson, L. E. (2017). What drives them to drive?-Parents’ reasons for choosing the carto take their children to school. Frontiers in Psychology, 8 , 1970. https://doi.org/10.3389/fpsyg.2017.01970

World Economic Forum. (2020). The Global Social Mobility Report 2020 . World Economic Forum.

Google Scholar  

Download references

Acknowledgements

Financial support from the Swedish Transport Administration is gratefully acknowledged. Henrik Andersson acknowledges funding from VTI and the French National Research Agency (ANR) under grant ANR-17-EURE-0010 (Le Programme d’investissements d’avenir (PIA)). We would also like to thank the reviewers and the editor for their comments that helped us improve the manuscript. The usual disclaimers apply.

This study was funded by the Swedish Transport Administration (grant number TRV 2020/26424).

Author information

Authors and affiliations.

VTI - Swedish National Road and Transport Research Institute, Box 55685, 102 15, Stockholm, Sweden

Jessica Göransson

K2 – The Swedish Knowledge Centre for Public Transport, Bruksgatan 8, 222 36, Lund, Sweden

Department of Technology and Society, Lund University, Box 118, 221 00, Lund, Sweden

Toulouse School of Economics, University of Toulouse Capitole, 31015, Cedex 6, Toulouse, France

Henrik Andersson

You can also search for this author in PubMed   Google Scholar

Contributions

JG and HA together developed the aim and objectives of the study. JG performed the literature review and prepared the original draft. HA supervised the process. JG and HA together edited and finalized the manuscript. Both authors have approved the manuscript for submission.

Corresponding author

Correspondence to Jessica Göransson .

Ethics declarations

Competing interests.

The authors declare that they have no competing interest.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Göransson, J., Andersson, H. Factors that make public transport systems attractive: a review of travel preferences and travel mode choices. Eur. Transp. Res. Rev. 15 , 32 (2023). https://doi.org/10.1186/s12544-023-00609-x

Download citation

Received : 01 September 2022

Accepted : 14 September 2023

Published : 25 September 2023

DOI : https://doi.org/10.1186/s12544-023-00609-x

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Travel preference
• Public transport
• Travel mode choice
• Travel demand
• Travel behaviour
• Systematic review
• Nordic countries

The Geography of Transport Systems

The spatial organization of transportation and mobility

8.3 – Urban Mobility

Author: dr. jean-paul rodrigue.

Urban mobility involves three broad categories of collective, individual, and freight transportation. While the mobility of passengers is the outcome of individual decisions based on different rationales, freight mobility is decided in tandem between the cargo owners and transportation service providers.

1. Urban Mobility and its Evolution

Urban areas are the most complex settings in which the mobility of passengers and freight is taking place. Typical urban attributes such as density, diverse economic, cultural, political, and social functions, and land scarcity, jointly generate mobility demands and constraints. In several instances, the mobility of passengers and freight is complementary as they may be using separate routes. Still, both are competing for the usage of scarce land and transport infrastructures:

• Collective transportation (public transit) . The purpose of collective transportation is to provide publicly accessible mobility over specific parts of a city. The systems are usually owned and operated by an agency, and access is open to all as long as a fare is paid; the reason why they are called public transit. The efficiency of public transit systems is based upon transporting large numbers of people and achieving economies of scale. It mainly includes tramways, buses, trains, subways, and ferries.
• Individual transportation . Includes any mode where mobility results from a personal choice and means, such as the automobile, walking, cycling, or motorcycling. Most people walk to satisfy their basic mobility, but this number varies according to the urban context. Some forms of individual mobility could be favored, while others could be impaired. For instance, walking accounts for 88% of all movements within Tokyo’s central area, while this figure is only 3% for Los Angeles. The density and design of the former are more accommodating to the mobility of pedestrians than the latter.
• Freight transportation . Since cities are dominant centers of production and consumption, urban activities are accompanied by large freight movements. These movements are characterized mainly by delivery trucks moving between industries, distribution centers, warehouses, and retail activities, including major terminals such as ports, railyards, distribution centers, and airports. The growth of e-commerce has been associated with increased home deliveries of parcels. The mobility of freight within cities is part of an emerging field related to  city logistics .

Rapid urban development occurring across much of the globe increased the mobility of passengers and freight within urban areas in absolute and relative terms. There are more urban movements and also more movements per urban resident. Urban mobility also tends to involve longer distances, but evidence suggests that commuting times have remained relatively similar over the last hundred years; approximately 1 to 1.2 hours per day is spent on average commuting. This means that commuting has gradually shifted to faster transport modes, and consequently, greater distances could be traveled using the same amount of time. This underlines the convergence among mobility, the deployment of transport infrastructure, and the diffusion of transportation modes.

Each form of urban mobility, be it walking, the automobile, or urban transit, has a  level of suitability to fill mobility needs. Different transport technologies and infrastructures have been implemented, resulting in a wide variety of urban transport systems around the world. In developed economies, there have been four general eras of urban development, each associated with a different form of urban mobility , with a fifth phase unfolding.

a. The Walking-Horsecar Era (1800s – 1890s)

Even during the Industrial Revolution, the dominant means of getting around was on foot. Walking cities were typically less than 5 kilometers in diameter, making it possible to walk from the downtown to the city edge in about 30 minutes. Land use was mixed, and density was high (e.g. 100 to 200 people per hectare). The city was compact and more-or-less concentric depending on the local landscape. Still, the industrial revolution brought additional populations through rural to urban migrations, improved construction techniques allowing for higher densities and new forms and employment locations. The development of the first public transit systems in the form of omnibus service extended the diameter of the city but did not change the overall urban structure.

The railroad facilitated the first real change in urban morphology. New developments, often called trackside suburbs , emerged as small nodes physically separated from the city itself and one another. The nodes coincided with the location of rail stations and stretched out a considerable distance from the city center, usually up to a half-hour train ride. Within the city proper, rail lines were also laid down, and horsecars introduced mass transit. The realm of urban mobility was expanded.

b. The Electric Streetcar or Transit Era (1890s – 1920s)

The invention of the electric traction motor created a revolution in urban travel. The first electric trolley line opened in the late 19th century, and the technology was quickly adopted in other cities . The operating speed of the electric trolley was three times faster than that of horse-drawn vehicles and did not generate waste on the streets. The streetcar city was able to spread outward 20 to 30 kilometers along the streetcar lines, creating an irregular, star-shaped pattern. Urban fringes became areas of rapid residential development, with trolley corridors as commercial strips that would come to characterize commercial areas of the era. The city core was further entrenched as a mixed-use, high-density zone, gradually losing its residential function. Land use patterns reflected social stratification where outer suburban areas were typically middle class, while the working class concentrated around the central city.

As street congestion increased in the first half of the 20th century due to the diffusion of the automobile, the efficiency of streetcar systems deteriorated as cars infringed on their right of way. Further, many cities had ordinances that prevented fare increases, implying that many streetcar systems became unprofitable, leading to a lack of maintenance and investment in additional services. These factors contributed to the demise of many streetcar systems in the later part of the 20th century.

c. The Automobile Era (1930s – 1950s)

The automobile was introduced in European and North American cities in the 1890s, but only the wealthy could afford this innovation. So no impacts on urban land use and mobility were initially observed. From the 1920s, ownership rates increased dramatically, with lower prices made possible by assembly-line production techniques. As automobiles became more common, land development patterns changed. Developers were attracted to green-field areas located between the suburban rail corridors, and the public was attracted to these single-use zones, thus avoiding many inconveniences associated with the city, mainly pollution, noise, crowding, and lack of space. Still, this phase usually represented the peak share of public transit in urban mobility as suburban developments did not yet account for a large share of the urban landscape, and cities were still high-density and transit-dependent.

d. The Freeway Era (1950s – 2010s)

In the second half of the 20th century, the massive diffusion of the automobile, as well as the construction of highway networks, had substantial impacts on urban mobility. Highways were built to connect the central business district to outlying areas, and, in many cases, complete or partial ring roads were built. The personal mobility offered by the automobile represented a paradigm shift in terms of lifestyle, consumption patterns, and residential locations. The automobile considerably reduced the friction distance, leading to urban sprawl . The emergence of the suburb created a new landscape in which public transit did not fit well, with few services being offered to these new residential areas. Transit ridership fell, and transit companies ran into financial difficulties. Eventually, transit services throughout North America and Europe became subsidized, publicly-owned enterprises. Some tramway systems were being dismantled, and the surviving transit lines were separated from road circulation, namely subway systems.

New light rail systems were introduced, which could generate ridership if large parking lots were provided at suburban stations. Commercial activities also began to suburbanize, creating additional passenger and freight mobility systems that did not rely much on public transit. Within a short period, the automobile became the dominant mode of travel in all cities of North America and, from the 1970s, in a growing number of developed and developing economies. Since the 2000s, a similar process has occurred on a massive scale in China, creating motorized high-density cities. Wherever incomes rise, the growth of automobile use tends to increase accordingly. Motorization and the diffusion of personal mobility have been trends linked with the ongoing and substantial declines in the share of public transit in urban mobility in the second half of the 20th century.

e. The Integrated Mobility Era (2010s onward)

Throughout their evolution, urban transportation modes remained rather disconnected, particularly since they are owned and operated by separate entities such as transit agencies, automobile owners, or trucking companies with limited interaction. The diffusion of information and communication technologies is changing this relationship. Emerging urban mobility systems are gaining from a higher level of integration and collaboration, resulting in better asset utilization levels and the creation of new mobility markets. An early example concerns on-demand vehicle services pooling individual drivers and matching their mobility supply with the consumer demand through a platform accessible through a mobile device. In several high-density markets, the outcome of this convenience was a surge in demand for for-hire vehicles. A further development concerns self-driving vehicles that could expand mobility options and better utilization of automobile assets.

This era is also associated with the diffusion of e-commerce and its associated home deliveries, underlining the issue of city logistics and last-mile freight distribution. Trucks and delivery vans have become more prevalent in urban mobility. Information technologies have also allowed the pooling of resources in the more conventional food delivery market, replacing business-specific deliveries with fleets of on-demand vehicles. An emerging form of urban mobility concerns micromobility , with early forms, such as the bicycle, being developed in the late 19th century and widely used by the early 20th century. In the early 21st century, a new array of electrically assisted conveyances and leasing/sharing systems became available, particularly electric bikes (e-bikes) and scooters. Such systems can be effective in high-density areas and for short trips. However, e-bikes are at least five times more expensive than bicycles. Further, users are concerned about parking e-bikes in public areas and finding locations to recharge them. They are also effective for last-mile parcel deliveries and fast deliveries such as restaurant orders.

In many areas where urbanization is more recent, the above synthetic phases did not occur. Fast urban growth led to a scramble to provide transport infrastructure inadequately, leading to rather chaotic conditions supporting urban mobility. Enduring congestion tends to characterize cities in developing economies.

2. A Taxonomy of Urban Mobilities

Mobility is linked to specific urban activities and land use, with each type involving generating and attracting an array of movements. This complex relationship is linked to factors such as recurrence,  income , urban form, density, level of development, and technology. Urban mobility is either obligatory when linked to scheduled activities (such as home-to-work trips) or voluntary when those generating it are free to decide on the scheduling (such as leisure) and even the mode. The most common types of urban mobility include:

• Pendulum movements . These are obligatory movements involving commuting between locations of residence and workplaces. They are highly cyclical since they are predictable and recurring, most of the time on a daily basis, thus the term pendulum. The historical stability of these movements allowed the planning of transportation infrastructure and services.
• Professional movements . These are movements linked to professional, work-based activities such as meetings, repair, maintenance, and customer services, dominantly taking place during work hours.
• Personal movements . These are voluntary movements linked to the location of commercial activities, which include shopping and recreation.
• Touristic movements . These are important for cities having historical and recreational features. They involve interactions between landmarks and amenities such as hotels and restaurants and tend to be seasonal or occur at specific moments during the day. Major sports events are important generators of urban movements during their occurrence.
• Distribution movements . These are concerned with freight distribution to satisfy consumption and manufacturing requirements. They are mostly linked to transport terminals, distribution centers, and retail outlets. However, the growth of online transactions involves more freight movements being carried to residential areas through home deliveries.

The consideration of urban mobility, both for passengers and freight, involves the consideration of the factors behind their generation, the modes and routes used, and their destination:

• Trip generation . On average, an urban resident undertakes between 3 and 4 daily trips. Mobility in an urban area is usually done to satisfy a purpose such as employment, leisure, or access to goods and services. The activity space of an individual is an important trip generation factor since it indicates the travel that needs to be undertaken. Temporal variations in the number of trips by purpose are observed on a daily and weekly basis, with commuting as the most prevalent pattern. Similar temporal variations are observed for freight mobility, with most of this mobility occurring in the morning when goods are delivered to retail outlets. This often leads to conflicts with the mobility of passengers since vehicles share the same road infrastructure, including parking space, which is the object of capacity constraints in urban areas.
• Modal split . This implies using a series of transportation modes for urban trips, which is the outcome of a modal choice . This choice depends on factors such as cost, technology, availability, preference, travel time (distance), and income. Therefore, walking, cycling, public transit, the automobile, or even teleworking, will be used either as a choice or as a constraint (lack of choice). For instance, locations within five minutes of walking are readily accessible to pedestrians. There is thus a wide variety of modal split across metropolitan areas . Urban freight distribution can also use a variety of modes, but the van and the truck tend to dominate as they allow maximum accessibility to urban locations.
• Trip assignment (routing) . It involves which routes will be used for trips within the city. Passenger trips usually have stable routing. For instance, a commuter driving a car usually has a fixed route between the residence and the place of work. This route may be modified if congestion or another activity (such as shopping) is linked with that trip, a practice often known as trip chaining . The routing of freight distribution is dependent on the types of deliveries involved. Direct deliveries are the norm for large retail outlets, while vehicles will accommodate flexible routing for smaller stores and parcel deliveries. Several factors influence trip assignment, the most crucial being transport costs, time, and congestion levels. The diffusion of information technologies, particularly global positioning systems, allows each vehicle to select a path minimizing distance or time in a dynamically evolving situation. The benefits of such technologies tend not to be fully acknowledged as large-scale routing optimization of individual vehicles significantly reduces total travel time and energy consumption.
• Trip destination . Changes in the spatial distribution of economic activities in urban areas have caused important modifications to trip destinations, notably those related to work. Activity-based considerations are essential since each economic activity tends to be associated with a level of trip attraction. Retail, public administration, entertainment, and restoration are the activities that attract the most movements per person employed. For freight movements, manufacturing, transport terminals, and retail are the activities attracting the most movements. The central city used to be a major destination for trips , particularly passengers, but its share has substantially declined in most areas, and suburbs now account for the bulk of urban trips.

Mobility is also a social issue. The share of the automobile in urban trips varies in relation to location, social status, income, quality of public transit, and parking availability. Mass transit is often affordable, but several social groups, such as students, the elderly, and the poor, are a captive market . There are important variations in mobility according to age, income, gender, and disability, with policies aiming at promoting the accessibility and mobility of groups perceived as disadvantaged. The gender gap in mobility is the outcome of socio-economic differences, as access to individual transportation is dominantly a matter of income. Within households, differences in role and income are related to the respective activity range and mobility of its members. Consequently, in some instances, modal choice is more of a modal constraint linked to economic opportunities.

Central locations generally have the most urban mobility options because private and public transport facilities are present. However, this does not mean mobility is easier since central areas are congested. In locations outside the central core, a share of the population not having access to the automobile faces a level of isolation or at least more limited access to amenities and employment opportunities. Limited public transit and high automobile ownership costs have created a group of spatially constrained (mobility-deprived) people. In a context where mobility is car-dependent, there is a strong incentive to own an automobile irrespective of income level.

3. Urban Transit

Transit is almost exclusively an urban transportation mode, particularly in large urban agglomerations . The urban environment is particularly suitable for transit because it provides conditions fundamental to its efficiency, namely high density and significant short distance mobility demands . Since transit is a shared service , it potentially benefits from economies of agglomeration related to high densities and economies of scale related to high mobility demands. One key advantage of public transit is the higher the demand, the more effective public transit services can be offered. Lower densities are linked with lower demand and a greater likelihood of public transit systems operating at a loss and requiring subsidies. Most public transit systems are not financially sound and must be subsidized , even if several of their core segments are profitable.

Transit systems are comprised of many types of services , each suitable to a specific market and spatial context . Different modes provide complementary services within the transit system and, in some cases, between the transit system and other transport systems.

• Bus transit . One of the most common forms of urban transit includes vehicles of various sizes (from small vans to articulated buses) offering seating and standing capacity along scheduled routes and services. They usually share roadways with other modes and are susceptible to congestion. Bus rapid transit systems offer a permanent or temporary right of way and have the advantage of unencumbered circulation. However, this footprint can come at the expense of other uses.
• Rail transit . Vehicles of fixed guideways usually have their right of way.  Light rail systems are composed of streetcars that can share the right of way, particularly in central areas. Heavy rail systems are commonly called subways or metro since many operate underground. Another type of rail transit concerns commuter rail systems, usually servicing central business districts and peripheral areas along specific rail corridors.
• Taxi systems . Usually, private for-hire vehicles such as automobiles, jitneys, or rickshaws offer point-to-point services. Recent technological developments have enabled car-sharing services and expanded the availability of on-demand transit.
• Alternative transit . Refer to transit systems developed to cope with specific conditions (or niche markets) using alternative modes. Ferries are the most common form of alternative transit as they serve cities with waterways separating different urban districts. Funiculars are also prevalent in locations with steep inclines and enough traffic to justify construction. Aerial lifts are also used in some settings to connect locations that are difficult to access.

Contemporary transit systems tend to be publicly owned , implying that many decisions related to their development and operation are politically motivated. This is a sharp contrast to what took place in the past, as most transit systems were private and profit-driven initiatives. With the fast diffusion of the automobile in the 1950s, many transit companies faced financial difficulties, and the quality of their service declined; in a declining market, there were limited incentives to invest. Gradually, they were purchased by public interests and incorporated into large agencies, mainly to continue providing mobility. As such, public transit often serves more as a social function of public service and a tool of social equity than having an economic role. Transit has become dependent on government subsidies, with little competition permitted as wages and fares are regulated. As a result, they tend to be disconnected from market forces, and subsidies are required to keep a level of service. With suburbanization, transit systems tend to have even fewer relationships with economic activities and the latest dynamism of cities.

Government-owned public transit systems are facing financial difficulties for three main reasons. First, they are often designed to serve taxpayers, not necessarily potential customers. Because of the funding base, transit systems may be spread into neighborhoods that do not provide a significant customer base. The second is that transit unions were able to extract substantial advantages in terms of wages and social benefits, increasing labor costs. This makes public transit highly expensive to operate. The third concerns a technology fixation that incites investment in high-cost transit (e.g. light rail transit) while low-cost solutions (buses) would have been sufficient for many transit systems, particularly in lower-density areas.

Reliance on urban transit as a mode of urban transportation tends to be high in Asia, intermediate in Europe and Latin America, and low in North America. Since their inception in the early 19th century, comprehensive urban transit systems significantly impacted the urban form and spatial structure, but this influence is receding. Three major classes of cities can be found in terms of the relationships they have with their transit systems:

• Adaptive cities . Represent transit-oriented cities where urban form and land use developments are coordinated with transit developments. While a metro system adequately services central areas and is pedestrian-friendly, peripheral areas are oriented along transit rail lines.
• Adaptive transit . Represent cities where transit plays a marginal and residual role, and the automobile accounts for the dominant share of movements. The urban form is decentralized and of low density.
• Hybrids . Represent cities that have sought a balance between transit development and automobile dependency. While central areas have an adequate level of service, peripheral areas are automobile-oriented.

Contemporary land development tends to precede the introduction of urban transit services instead of concurrent developments in earlier phases of urban growth. Thus, new services are established once the demand is deemed sufficient, often after being subject to public pressure. Transit authorities operate under a service warrant and usually run a recurring deficit as services become more expensive. This has led to considerations aimed at higher transit integration in the urban planning process, particularly in cities where such a tradition is not well established.

From a transportation perspective, the potential benefits of better integration between transit and local land uses are reduced trip frequency and increased use of alternative modes of travel (i.e. walking, biking, and transit). Evidence often fails to support such expectations since the relative share of public transit ridership is declining across the board. There is usually a reciprocal relationship between automobile ownership and the use of public transit. Good accessibility to public transit is often associated with lower automobile use. In contrast, areas of high automobile use may impair the development of public transit systems since the automobile is already dominant. Exceptions tend to be cities having very high-density levels.

Community and land use design can have a significant influence on travel patterns. Local land use impacts can be categorized into  three dimensions in terms of accessibility, the convergence of mobility, and the land use integration they provide. Land use initiatives are trying to be coordinated with other planning and policy initiatives to cope with automobile dependence. However, there is a strong bias against transit in the general population because of negative perceptions, especially in North America, but increasingly globally. As personal mobility symbolizes status and economic success, public transit users can be perceived as less successful segments of the population. This bias may undermine the image of transit use within the general population but can be subject to change with the evolution of social norms and values.

The COVID-19 pandemic had complex impacts on public transit systems. In the initial phase of the lockdowns in 2020, most transit systems experienced a decline in ridership in the range of 75%. The benefits of public transit, the massification of trips, became a disadvantage as users became concerned about the transmission risks during a public transit trip. Many transit systems, particularly in advanced economies, did not experience recovery to pre-pandemic levels. A factor is the growth of teleworking, which offered a substitution for transit trips, but a shift to car use is the most important. People actively using cars increased their car use, while active transit users also increased their car use. As a fundamental support to urban mobility, public transit systems remain challenged by revenue generation, rising infrastructure costs, and the willingness of users to shift to other modes.

Related Topics

• 8.1 – Transportation and the Urban Form
• 8.2 – Urban Land Use and Transportation
• 8.4 – Urban Transport Challenges
• City Logistics External link
• 2.4 – Information Technologies and Mobility

Bibliography

• Barry, M. (1991) Through the Cities: The Revolution in Light Rail Transit. Frankfort Press: Dublin.
• Cervero, R. (1998)  The Transit Metropolis: A Global Inquiry , Washington D.C.: Island Press.
• Circella, G. and P.L. Mokhtarian (2017) “Impacts of Information and Communication Technology”, in G. Giuliano and S. Hanson (eds) The Geography of Urban Transportation, 4th Edition, New York: The Guilford Press.
• Cresswell, T. (2006) On the move: mobility in the modern Western world, New York: Routledge.
• Diao, M., Kong, H. and J. Zhao (2021) “Impacts of transportation network companies on urban mobility”. Nature Sustainability 4, 494-500. https://doi.org/10.1038/s41893-020-00678-z.
• Ellder, E. (2014) “Residential Location and Daily Travel Distances: The Influence of Trip Purpose”. Journal of Transport Geography, Vol. 34, pp. 121-130.
• Kwan, M.P., and J. Weber (2008) “Scale and Accessibility: Implications for the Analysis of Land Use-Travel Interaction”. Applied Geography, Vol. 28 (2), pp. 110-123.
• Leinbach, T.R. (2004) “City Interactions: The dynamics of passenger and freight flows” in S. Hanson and G. Giuliano (eds) The Geography of Urban Transportation, 3rd ed. New York: The Guilford Press, pp. 30-58.
• Levinson, D.M. (1998) “Accessibility and the Journey to Work”, Journal of Transport Geography, Vol. 6 (1), pp. 11-21.
• Levinson, D. and D.A. King (2019) A Political Economy of Access: Infrastructure, Networks, Cities, and Institutions, Network Design Lab.
• Marchetti, C. (1994) “Anthropological Invariants in Travel Behaviour”, Technical Forecasting and Social Change, Vol. 47, No. 1, pp. 75-78.
• McKinsey & Co (2021) Reimagining transit in a Post-Covid world, Global Infrastructure Initiative.
• Mokhtarian, P.L. and G. Tal (2013) “Impacts of ICT on Travel Behavior: A Tapestry of Relationships”, in J-P Rodrigue, T. Notteboom and J. Shaw (eds) The Sage Handbook of Transport Studies, London: Sage.
• Muller, P.O. (2017) “Transportation and Urban Form: Stages in the Spatial Evolution of the American Metropolis”, in G. Giuliano and S. Hanson (eds) The Geography of Urban Transportation, 4th Edition, New York: Guilford, pp. 57-85.
• Qi, Y., J. Liu, T. Tao and Q. Zhao (2023) “Impacts of COVID-19 on public transit ridership”, International Journal of Transportation Science and Technology, Vol. 12, No. 1, pp. 34-45. doi.org/10.1016/j.ijtst.2021.11.003.
• Rietveld, P. (2000) “Nonmotorized Modes in Transport Systems: A Multimodal Chain Perspective for The Netherlands”, Transportation Research D, Vol. 5, No. 1, pp. 31-36.
• Schafer, A. (2000) Regularities in Travel Demand: An International Perspective, Journal of Transport Statistics, Vol. 3, No. 3.
• Thomson, J. M. (1977) Great Cities and Their Traffic. London: Victor Gollancz Ltd.
• World Bank (2002) Cities on the Move: A World Bank Urban Transport Strategy Review. Washington: The International Bank for Reconstruction and Development/World Bank.

Share this:

National Association of City Transportation Officials

Programs & initiatives.

• Guides & Publications

Conferences & Events

About nacto, nacto is an association of 100 major north american cities and transit agencies formed to exchange transportation ideas, insights, and practices and cooperatively approach national transportation issues..

• Our Mission
• Member Cities

A micromobility record: 157 million trips on bike share and scooter share in 2023

• Reflections on #NACTO2024: Moving the Movement Forward

NACTO Announces Appointment of Ryan Russo as Executive Director

Stay up to date, nacto email newsletter, check out the latest updates, statements and press releases from nacto., latest news, thousand-page document governing nearly every street in the u.s. gets a refresh.

• A micromobility record: 157 million trips on bike and scooter share in 2023

Statement: State Lawmakers and the MTA Board Must Go Ahead With Congestion Pricing

Featured stories, read more stories, through our initiatives, nacto empowers cities to realize their goals for stronger, safer, and fairer transportation and communities., program areas.

• Designing Better Streets
• Building Capacity
• Shaping Transportation Policy
• Strengthening Peer Networks

All Programs

Featured programs, state and federal policy, cities for cycling, american cities climate challenge, bike share and shared micromobility initiative, guides and publications, nacto’s publications provide a vital resource for practitioners, policy-makers, academics, and advocates., latest publications.

• Shared Micromobility in 2023
• Material Success
• Complete Connections
• Designing for Small Things with Wheels
• Moving Together

All Publications

Featured design guidance, urban street design guide, transit street design guide, urban bikeway design guide, global street design guide, all design guides, explore our conferences, events, trainings and workshops.

• Designing Cities 2025: Washington, D.C.
• About the Conference
• Trainings & Workshops
• Upcoming Events

Featured Events

Nacto meet the cities: midwest – virtual summer series, nacto meet the cities: canada – virtual summer series, nacto meet the cities: southeast – virtual summer series.

• Introduction
• Transit Streets
• Stations & Stops
• Station & Stop Elements
• Transit Lanes & Transitways
• Intersections
• Transit System Strategies
• Project Team
• Endorsement Campaign
• Using the Guide

Designing to Move People

Transit streets are designed to move people, and should be evaluated in part by their ability to do so. Whether in dense urban cores, on conventional arterials, or along neighborhood spines, transit is the most spatially efficient mode.

Traditional volume measures fail to account for the entirety of functions taking place on urban streets, as well as the social, cultural, and economic activities served by transit, walking, and bicycling. Shifting trips to more efficient travel modes is essential to upgrading the performance of limited street space.

Using person throughput as a primary measure relates the design of a transit street to broader mode shift goals.

The capacity of a single 10-foot lane (or equivalent width) by mode at peak conditions with normal operations.

60 people per meter per minute, allowing 1.5 sq. meters per pedestrian and a 75 meter per minute walking speed.

“Ch 6, Bus Transit Capacity.”  TCQSM, 3rd Ed . (2013).

Zhou et al. measured 2,500 bikes per meter per hour on cycle tracks in downtown Hangzhou.

Dan Zhou, et al. Estimating Capacity of Bicycle Path on Urban Roads in Hangzhou, China  (2014).

While street performance is conventionally measured based on vehicle traffic throughput and speed, measuring the number of people moved on a street—its person throughput and capacity—presents a more complete picture of how a city’s residents and visitors get around. Whether making daily commutes or discretionary trips, city residents will choose the mode that is reliable, convenient, and comfortable.

Transit has the highest capacity for moving people in a constrained space. Where a single travel lane of private vehicle traffic on an urban street might move 600 to 1,600 people per hour (assuming one to two passengers per vehicle and 600 to 800 vehicles per hour), a dedicated bus lane can carry up to 8,000 passengers per hour. A transitway lane can serve up to 25,000 people per hour per travel direction. Read More+

A reasonable planning-level capacity for a dedicated transit lane is 80 buses per hour; assuming 100 riders per vehicle (a comfortable articulated bus capacity), 8,000 riders per hour can be moved through a single transit lane. At two-minute headways (or 30 buses per hour), a standard 40-foot bus, assuming 60 passengers, moves 1,800 passengers per hour.

High-capacity LRVs, running four cars per train with a capacity of 125 riders, have a capacity of 15,000 passengers per hour.

Transportation Energy Data Book, edition 34   (2015).

Streetcar and Light Rail Streetcar and Light Rail Characteristics . Regional Transportation District (2012).

References for Designing to Move People : 5 found.

• Ryus, Paul (PI), Alan Danaher, Mark Walker, Foster Nichols, William Carter, Elizabeth Ellis, and Anthony Bruzzone. "Ch. 6: Bus Transit Capacity." Transit Capacity and Quality of Service Manual, Third Edition, TCRP Report 165 , Transportation Research Board , Washington .
• Litman, Todd. "A New Transit Safety Narrative." Journal of Public Transportation 17(4) , National Center for Transit Research , Tampa .
• Dan Zhou, Cheng Xu, Dian-Hai Wang, and Sheng Jin. "Estimating Capacity of Bicycle Path on Urban Roads in Hangzhou, China." Zhejiang University , Submitted to the 94th Annual Meeting of the Transportation Research Board , Washington .
• Regional Transportation District. "Streetcar and Light Rail Streetcar and Light Rail Characteristics." RTD FasTracks , Denver .
• Oak Ridge National Laboratory. "Chapter 8: Household Vehicles and Characteristics." Transportation Energy Data Book, 34th Edition , US Department of Energy , Washington .

No results - please refine your search and try again.

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 27 May 2021

Urban access across the globe: an international comparison of different transport modes

• Hao Wu   ORCID: orcid.org/0000-0002-5526-8827 1 ,
• Paolo Avner 2 ,
• Genevieve Boisjoly 3 ,
• Carlos K. V. Braga 4 ,
• Ahmed El-Geneidy 5 ,
• Jie Huang 6 ,
• Tamara Kerzhner 7 ,
• Brendan Murphy 8 ,
• Michał A. Niedzielski   ORCID: orcid.org/0000-0001-6639-1057 9 ,
• Rafael H. M. Pereira   ORCID: orcid.org/0000-0003-2125-7465 4 ,
• John P. Pritchard   ORCID: orcid.org/0000-0001-8546-4872 10 ,
• Anson Stewart 11 ,
• Jiaoe Wang 6 &
• David Levinson   ORCID: orcid.org/0000-0002-4563-2963 1  

npj Urban Sustainability volume  1 , Article number:  16 ( 2021 ) Cite this article

16k Accesses

19 Citations

430 Altmetric

Metrics details

• Development studies

An Author Correction to this article was published on 11 June 2021

This article has been updated

Access (the ease of reaching valued destinations) is underpinned by land use and transport infrastructure. The importance of access in transport, sustainability, and urban economics is increasingly recognized. In particular, access provides a universal unit of measurement to examine cities for the efficiency of transport and land-use systems. This paper examines the relationship between population-weighted access and metropolitan population in global metropolitan areas (cities) using 30-min cumulative access to jobs for 4 different modes of transport; 117 cities from 16 countries and 6 continents are included. Sprawling development with the intensive road network in American cities produces modest automobile access relative to their sizes, but American cities lag behind globally in transit and walking access; Australian and Canadian cities have lower automobile access, but better transit access than American cities; combining compact development with an intensive network produces the highest access in Chinese and European cities for their sizes. Hence density and mobility co-produce better access. This paper finds access to jobs increases with populations sublinearly, so doubling the metropolitan population results in less than double access to jobs. The relationship between population and access characterizes regions, countries, and cities, and significant similarities exist between cities from the same country.

Similar content being viewed by others

How transit scaling shapes cities

Access to mass rapid transit in OECD urban areas

A suite of global accessibility indicators

Introduction.

Cities exist to enable people to easily reach other people, goods, and services. This is achieved through transport networks, which move people across space faster, and land-use patterns, which distribute people, goods, and services across space. Access (or “accessibility”) is the ease of reaching those valued destinations, and thus is a critical measure of urban efficiency. Higher land use density and travel speeds correspond with greater access 1 . Urban population size has been associated with greater productivity and creativity, a centerpiece of the urban economies of agglomeration literature 2 , 3 . This paper measures how access varies with metropolitan population size, whether larger cities also enjoy increasing accessibility, which we expect will affect how well population produces economies of agglomeration.

To measure access across the globe, we use the cumulative number of job opportunities reachable under a predefined travel time threshold. Since jobs are places of interaction, that provide service either directly or indirectly to customers, jobs are a key indicator of “urban opportunities” 4 serving both as employment opportunities, and as urban amenities.

The role of access in transport, sustainability, and urban economics is increasingly recognized. The positive correlation between access and land value originates from the trade-off between time (transport cost) and space (the price of land) 5 , 6 , 7 , and this positive correlation has been shown with various hedonic models in different contexts 8 , 9 , 10 . Access affects firm location choice 11 , development probability of vacant land 12 , commute mode choice 13 , 14 , and transport emissions 15 . The expansion and rapid growth of urban areas call for a meaningful measure of geographical connectivity, to which a measure of access to job opportunities can be a useful tool 16 .

Cities have been ranked and recognized by economic and demographic statistics that describe their sizes and productivity 17 , 18 , 19 , 20 . The efficiency of transport infrastructure and land use in linking people with opportunities is also a vital measure for cities 21 , 22 , 23 , and its significance is on par with economic and demographic measures. However, there has been no previous large-scale, multi-modal comparison of access to jobs for cities across the globe. Cities differ in the levels of transport infrastructure and development patterns. On the one hand, US cities prioritize mobility over density; the density of US cities are relatively low 24 due to sprawling development patterns. Road length per vehicle is higher in the US than in Europe, Canada, and Oceania 25 . On the other hand, European cities are compact and have a denser road network than US, Canadian, and Oceania cities 25 . These differences among cities and global regions would have accessibility implications, that affect the quality of the transport system within each city. This paper examines the type of development pattern most conducive to accessibility, and whether mobility and density can co-exist for better accessibility.

Understanding global urbanization and transport development are needed 26 . The quantification of access using cumulative opportunities provides a universal unit of measurement for transport and land use, and the opportunity for defining cities by transport and land use with a uniform benchmark. Measuring accessibility sheds light on the basic structure of cities underpinning economic efficiency, and reveals where a city (i.e., metropolitan area) stands relative to its overall size and other attributes, and relative to other cities.

Access has been measured and compared domestically in the United States 27 , Australia 28 , Brazil 29 , Canada 30 , New Zealand 31 , and in European countries 32 using the cumulative opportunities measure. One notable advantage of the cumulative opportunities measure is being in absolute and comparable units 33 , and having clarity of meaning, and being an easily understood and interpretable concept 34 . This paper uses a travel time threshold of 30 min that is consistent with many estimates of one-way travel time budgets to work, to measure access to jobs in all cities 35 , 36 , 37 , 38 , 39 .

This paper compares access globally and covers access by automobile, transit, walking, and cycling where data were available. We examine patterns in the relationship between access and populations, and whether cities cluster by their global region. This work also examines whether the disparity in access to jobs between modes of transport is linked with population size. Access to jobs data in this global comparison are collected by different researchers and organizations. The population provides a consistent measure of metropolitan area size that is comparable across global regions. This paper details the data collected from multiple agencies in the supplementary information, commenting on differences in data sources. Access measures for each model are tabulated and cities ranked by the total population in the supplementary information. The paper examines access in global cities across modes, across countries, and across cities.

Comparison across countries: scaling city-level access with population

In order to compare accessibility across cities worldwide, we first compare the relationship between the population size and accessibility in different countries. Such comparison can help clarify the differences in the relationship, and in the returns to scale of metropolitan population size on accessibility between different countries, i.e., is the increase in access proportional to the increase in population in each country, and how different countries compare.

We use scaling functions to quantitatively measure the relationship between population and the level of access to jobs. The formulation of the scaling function is shown in the methods sections. The scaling coefficient ( β 1 ) signifies the returns to scale, where β 1  > 1 means doubling the metropolitan population will more than double the level of access. We find the city-level access to jobs increases with population. Table 1 shows the model fit and coefficients.

Although the relationship between access and population, in general, is positive, with larger cities exhibiting higher accessibility by all modes than their smaller counterparts within the same counties (the β 1 coefficients are positive), we generally see diminishing returns in access with respect to population (the coefficient is less than one), so we have sublinear scaling, meaning population rises faster than access. Notably, transit in Chinese cities is the only exception to the sublinear scaling: the increasing population in Chinese cities confers proportionally more accessible jobs.

Comparison across modes

We compare city-level access to jobs by four modes of transport: automobile, transit, walking, and cycling. The population and access by different modes on a logarithm scale are plotted in Fig. 1 through Fig. 4 ; scaling functions are fitted as trend lines. Figures showing the accessibility by different modes of transport for each country is provided in the Supplementary Information. Cities above the trend lines are overperforming based on their mode and country category; cities beneath the trend lines are underperforming. Our comparison corroborates that differences in structure exist both within, and between, global regions. Although cities of the same global region tend to share similar trends in access, cities from different global regions can cluster on access attributes.

Walking access to jobs and metropolitan population.

Walking access to jobs is shown in Fig. 1 . Walking is part of every other mode of transport. The walking access alone represents the spatial distribution of population relative to urban opportunities. Urban density, the proximity between residential and employment centers 40 , and mixed land use increase walking access.

For any given population, Chinese and European cities have markedly higher walking access to jobs than cities in other countries. American cities, with their lower densities and auto-orientation, as well as the functional separation between residential and employment districts, have the lowest walking access to jobs globally. Among American cities, New York and San Francisco-Oakland (excluding San Jose) are more similar to European cities than to other American cities. In Oceania, Wellington also clusters amongst European cities with significant walking access to jobs, and Sydney comes close to European access levels.

Cycling access is shown in Fig. 2 . Cycling access in major cities tends to be below automobile, but higher than transit. Chinese and European cities generally have greater cycling access than the US cities with similar population sizes. Cycling access in Oceania cities is comparable to the best American and Brazilian cities, but lower than Chinese and European cities.

Cycling access to jobs and metropolitan population.

Cycling provides better access to jobs than transit in every city where access data of the two modes are available. Cycling has no waiting or transfers time penalties, and cycling routes to and from destinations are less circuitous than transit. Cycling provides better access than automobiles in the city of Shanghai, where congestion reduces access by automobile.

Transit access is plotted against the population in Fig. 3 . Transit service provision is linked to patronage in a positive feedback system 41 , which affects transit performance, so more populous cities with a greater base for transit patronage tend to have better transit quality of service, with higher frequency, shorter access distances, and more direct service, resulting in higher access to jobs 42 .

Transit access to jobs and metropolitan population.

Chinese and European cities have higher transit accessibility than the others. London and Paris have a longer history than North American cities and were well developed before the advent of the automobile. It is expected that these European cities would better support transit and walking. Australian and Canadian cities are similar in terms of transit access, with the exception of Quebec City, which, despite its age, more resembles American cities. Brazil’s largest metropolises, São Paulo and Rio de Janeiro have much lower accessibility levels than what would be expected for their population. This is in part because these two cities have large territories with low population densities coupled with a high concentration of jobs in the city center, and both cities have relatively poor transport conditions with some of the highest average commute times among global cities 43 .

The majority of American cities lag behind in transit access for their population size. New York and San Francisco-Oakland are exceptions for the American cities, with high transit access relative to size. Metropolitan Washington, Boston, Philadelphia, Chicago, and Seattle bear more resemblance to Australian and Canadian cities than to other American cities, although transit access for these cities is still lower than their Australian and Canadian counterparts. The two African cities, Douala and Nairobi both have better walking and transit access than average American cities.

Automobile access is plotted in Fig. 4 . Large cities tend to have well-developed road networks and have employment opportunities proportional to the population sizes, which, generally translates to greater automobile access. Conversely, heavy traffic flows can cause delays in large cities, especially during peak hours when automobile travel time data for this study are measured.

Automobile access to jobs and metropolitan population.

Both globally and within each nation, automobile access increases with population. Chinese and European cities follow a distinct trend from other global cities and have the highest automobile access for each population level. American cities have greater automobile access than Australian and Canadian cities at each level of population. Historically the United States has placed heavy emphasis on auto-mobility, and the Interstate highway system created by the Federal-Aid Highway Act of 1956 44 greatly facilitates movement by automobile both within and between metropolises. New York has high automobile access, but its level did not grow out of proportion to make New York an outlier; New York is an outlier among US cities for high walking and transit access. Oceania and Canadian cities are comparable to American cities that are in the lower quantile of automobile access; Perth more resembles mid-tier American cities at its size.

Comparison across cities: access with populations

Across cities, we find population correlates with access to job opportunities. People in more populous cities generally have better accessibility. Population reflects the need for transport. On the one hand, it is hypothesized that larger metropolitan areas enjoy economies of agglomeration and tend to have better public transport systems and higher residential and employment densities, thus better access. On the other hand, congestion in large cities reduces access, and differences in urban networks and spatial configuration can result in varying levels of access for cities with similar sizes. Similarly, as the population increases, urban densities rise slower than the population, as some of the additional growth expands urban territory, and thus distances required 45 . In the Supplementary Information, we tabulate global cities by the access of different transport modes (where data were available).

The correlation between population and accessibility of different modes is strongest with the automobile (0.69), followed by cycling (0.55), walking (0.48), and transit (0.44). An automobile can generally reach more territory than other modes within a 30-min threshold, so the automobile accessibility relates more strongly with the metropolitan population than other modes. The weaker correlation for the transit mode is largely due to transit being the most susceptible mode to variations in service provision, and urban land use. Small cities with good transit infrastructure can provide similar levels of access to jobs within 30 min as many large cities (which may have more jobs available for longer time thresholds, which are less valuable to residents than nearer jobs). Examples of such small but compact cities with good transit services include most European cities, and Wellington and Christchurch in New Zealand, which have relatively good transit accessibility for their sizes. Large and densely populated metropolitan areas such as New York, London, and Paris have good transit accessibility.

Modal comparison: transit as a benchmark

We explicitly examine the disparity in access provided by different modes of transport from different cities, using transit as the benchmark. The existence and extent of such disparity vary by city, and by the country of the cities 46 . The difference in access by different modes is compared using the ratio of cumulative access so that the difference between modes is scalable. Since the access data from each city is collected by the same source and over the identical geographical extent, the ratio of access by mode provides a reliable gauge for comparing the within-city modal difference. Figure 5 shows the ratio of 30-min access to jobs by automobile and by cycling, relative to the transit mode.

Boxes span 25th to 75th percentile, whiskers extend to max/min of the values, excluding outliers (>1.5 box heights away from box edges, which are shown as circles).

The automobile provides better access than transit in all cities we compared, except in Shanghai, China, where automobile reaches about 90% of the jobs reachable by transit at 30 min. The disparity between transit and automobile access is greatest in American cities; Oceania and Canadian cities have more comparable levels of access between the two modes, although the gap remains significant. Among the Oceania cities, Perth and Wellington are the two extremes (outliers) for respectively having the largest and smallest gaps between transit and automobile access to jobs.

Transit and automobile often have the highest commute mode share in major cities. Income 47 and physical ability affect mode choice, so the relative level of access to jobs provided by automobile and transit has equity implications 48 . The ratio of the automobile to transit access correlates weakly with population ( R  = 0.10), so larger populations reduce the gap between transit and automobile, but generally do not guard against inequitable transit access to jobs. US cities with better transit access tend also to have lower auto to transit access ratio; examples include New York, San Francisco-Oakland, Washington, and Boston.

The automobile provides better access than transit in almost all cities we compared. The disparity between transit and automobile access is greatest in American cities Fig. 5 ; Australian and Canadian cities have more comparable levels of access between the two modes. The ratio of the automobile to transit access does not seem to be affected by the population.

All cities we examined have cycling access higher than that of transit. Oceania and European cities have stable ratios of cycling to transit access, where cycling can reach about twice as many jobs as transit. The gap between cycling and transit access to jobs is larger in the US and Chinese cities. The ratio of cycling to transit accessibility has a weak correlation with populations ( R  = 0.17), so the gaps between transit and cycling are smaller with larger populations.

This research conducts a systematic, multi-modal, international comparison of access to jobs, which is the core variable connecting transport networks and land use, and the central factor in characterizing cities and explaining why cities exist 49 . This paper compares the performance of the transport and land-use system across cities.

One notable finding of this paper is the national difference in the relationship between access and population. While outliers exist, there are remarkable similarities for cities in the same global region. We find mobility and density can co-exist and collectively co-produce greater access to jobs for a metropolitan area. Sprawling development accompanied by an intensive road network, as is common in American cities, results in modest automobile access, but low access for transit and active modes of transport. Oceania and Canadian cities, with US-style land, uses but without US-level freeway networks have relatively low automobile access and are generally situated between American and European cities in transit and walking. This is consistent with previous findings 23 . Chinese and European cities are compact, and well supported by road networks, resulting in the highest accessibility in all modes of transport.

While it is unsurprising that this paper finds city-level access increases with population, the relationship is neither linear nor constant across countries. Access does not increase proportionally with a population (for all modes in all countries where data are available, except transit in China), and presents diminishing returns to scale, so the doubling of city population will likely less than double access to jobs.

More populous cities present more available urban opportunities, usually at higher densities, with higher levels of traffic congestion and better public transport infrastructure. When appropriately matched with transport infrastructure, compact urban development generally improves access to jobs, despite increased congestion. In terms of the disparity between modes of transport, we find larger cities tend to narrow the gap between transit and automobile access to jobs. The disparity between transit and automobile is most significant in the US, mostly as a result of sprawling development, which increases automobile speeds and makes transit service more difficult.

Two major caveats are identified: the demarcation of city boundaries, and the jobs and population data source. The modifiable areal unit problem is present in defining the city boundary for analysis, for example, excluding lower-density outlying (exurban) areas likely inflates the access measure. Although there is no consistent standard for defining city boundaries across nations, the geographical area was chosen for measuring access generally reflects what would be considered the built-up, urbanized area in each city that envelops commute ties to the urban core. This city boundary issue is further alleviated by using the population-weighted access measure, where the population data were available. The second caveat involves the census data on jobs and population numbers collected (or not) by governments of different nations, that vary in accuracy and coverage. We believe that while more consistent standardization of city boundary measurement and employment definition would affect specific numbers, they would not substantially change the general findings and conclusions from this study. By definition in some sense, “informal economy” jobs are excluded in all countries (which underestimates access in some metropolitan areas much more than others), however, the magnitude of this is unclear by its very nature.

This work provides a cross-sectional comparison of cities, focusing on how cities of similar scales compare, in terms of the coupling between transport infrastructure and land use, measured by access. Future research can link access with other city-level characteristics, including income, GDP per capita, transport emissions, and commute duration, to shed light on the interplay between these elements. In the future, it will also be possible to explore time-series data on access across the globe, to observe the longitudinal trend in the co-evolution of access with other urban elements.

Method for measuring accessibility

Access is calculated as the cumulative number of jobs reachable under a 30-min travel time threshold. Equation ( 1 ) specifies the cumulative access measure, for a location of interest. The concept of cumulative opportunities can be conceived graphically as the geographical area covered within a travel time threshold, and the number of opportunities contained in that area.

Selection of travel time threshold affects the access measure 50 , 51 ; although distance decay functions 52 (time-weighted cumulative opportunities 53 ) can be used to avoid the choice of thresholds, the difference in data sources and variations across geographies, as well as different preferences for travel between people in different cities, prevent the use of a consistent decay function. We use a 30-minute threshold without distance decay (Eq. ( 2 )) for consistent comparison across modes.

where, A i , m is access measure for zone i , by mode m ; O j number of jobs at zone j ; f ( C i j , m )is the travel time between zone i and j for mode m ; t is the travel time threshold = 30 min.

Access measures for individual zones are aggregated to produce city-level averages. Person-weighted city averages are used where the population data were available, and the population of each subdivision is used as a weight. For the US, Australian, Chinese, and Polish cities, the working population is used as a weight; the total population is used as a weight for Brazilian, Canadian cities, Paris, and London. City-level access of African and Dutch cities are arithmetic averages of subdivisions and not weighted by population, and are thus expected to be lower than population-weighted measures for the same area. Equation ( 3 ) shows the formulation of population-weighted city-level access. This population-based weighting scheme reflects average access as experienced by the entire population.

where, \({A}_{I,m}^{\,}\) is the city-level access for mode m in the city I ; p j is the population within zone j ; and J is the number of zones within the city.

Accessibility data

Accessibility to jobs in each city is calculated based on the subdivision of the city into zones, the travel time between zones, and the number of jobs within each zone. Traffic 54 and transit schedules can cause temporal variations in access 55 , 56 , so the level of access differs by hours of the day. Automobile accessibility is based on historical traffic data, and includes the effects of congestion; recurrent congestion is included for transit through the digitized transit schedule data; walking and cycling does not consider congestion and use all links where walking and cycling are legal, not only where they are pleasant. We measure access based on the morning peak travel time. A detailed description of data sources is provided in the Supplementary Information.

Pedestrians encounter both signalized intersections and non-signalized street crossings that add additional travel time, which is partially but incompletely accounted for by adjusting pedestrian walking speed, so walking access tends to be overestimated. Cycling access is calculated using all roads, although cyclists selectively use roads depending on vehicular speeds and levels of traffic for safety reasons, so actual access by cycling tends to be lower than estimates using all roads 57 , 58 . Digitized transit schedule information in general transit feed specification format (and a similar system in China) is used for calculating transit access. Transit travel time estimates cover transit station access, egress, waiting, and transfer time, and assumes perfect schedule adherence (on the theory that schedules have been calibrated to improve reported “on-time performance”, but perfect adherence may result in overestimation compared with actual access). The automobile travel time includes the effect of congestion, but not searching for parking.

Scaling access to population

To quantitatively measure the proportionality between population and the level of access to jobs, we group cities by country and fit scaling models 59 (Eq. ( 4 )) to cities of the same countries. The scaling coefficient ( β 1 ) signifies the returns to scale.

where, β 0 , β 1 is the coefficients of the scaling model.

Data availability

The data generated and analyzed during this study are described in the following data record: https://doi.org/10.6084/m9.figshare.13476867 60 . The access data are openly available as part of the figshare metadata record in the file “AllCities.csv”. A list of all cities included in the study, along with the sources of the data used for each city, is available in the file “SI for Urban Access Across the Globe—An International Comparison of Different Transport Modes.pdf”. Both of these files are also available in PDF format via the Supplementary Information of this article.

Change history

11 june 2021.

A Correction to this paper has been published: https://doi.org/10.1038/s42949-021-00035-9

Levine, J., Grengs, J., Shen, Q. & Shen, Q. Does accessibility require density or speed? A comparison of fast versus close in getting where you want to go in US metropolitan regions. J. Am. Plan. Assoc. 78 , 157–172 (2012).

Article   Google Scholar  

Graham, D. J. Agglomeration, productivity and transport investment. J. Transp. Econ. Policy 41 , 317–343 (2007).

Google Scholar  

Glaeser, E. L. et al. Cities, Agglomeration, and Spatial Equilibrium (Oxford University Press, 2008).

Merlin, L. A. A portrait of accessibility change for four US metropolitan areas. J. Transp. Land Use 10 , 309–336 (2017).

Smith, A. An Inquiry into the Nature and Causes of the Wealth of Nations . MetaLibri Digital Library 2 (1776).

Alonso, W. Location and Land Use (Harvard University Press, 1964).

Vickerman, R. et al. Transit investment and economic development. Res. Transp. Econ 23 , 107–115 (2008).

Osland, L. & Pryce, G. Housing prices and multiple employment nodes: is the relationship nonmonotonic? Hous. Stud. 27 , 1182–1208 (2012).

Ahlfeldt, G. If Alonso was right: modeling accessibility and explaining the residential land gradient. J. Reg. Sci. 51 , 318–338 (2011).

Mayor, K., Lyons, S., Duffy, D. & Tol, R. S. A hedonic analysis of the value of rail transport in the Greater Dublin area. J. Transp. Econ. Policy 46 , 239–261 (2012).

Ingram, D. R. The concept of accessibility: a search for an operational form. Reg. Stud. 5 , 101–107 (1971).

Hansen, W., G. How accessibility shapes land use. J. Am. Inst. Plan. 25 , 73–76 (1959).

Owen, A. & Levinson, D. M. Modeling the commute mode share of transit using continuous accessibility to jobs. Transp. Res. A 74 , 110–122 (2015).

Wu, H., Levinson, D. & Owen, A. Commute mode share and access to jobs across US metropolitan areas. Env. Plan. B . https://doi.org/10.1177/2399808319887394 (2019).

Stokes, E. C. & Seto, K. C. Tradeoffs in environmental and equity gains from job accessibility. Proc. Natl Acad. Sci. 115 , E9773–E9781 (2018).

Article   CAS   Google Scholar  

Biazzo, I., Monechi, B. & Loreto, V. General scores for accessibility and inequality measures in urban areas. R. Soc. Open Sci. 6 , 190979 (2019).

Levinson, D. Network structure and city size. PloS ONE 7 , e29721 (2012).

Arcaute, E. et al. Constructing cities, deconstructing scaling laws. J. R. Soc. Interface 12 , 20140745 (2015).

M, B. The size, scale, and shape of cities. Science 319 , 769–771 (2008).

Bettencourt, L. M., Lobo, J., Helbing, D., Kühnert, C. & West, G. B. Growth, innovation, scaling, and the pace of life in cities. Proc. Natl. Acad. Sci. USA 104 , 7301–7306 (2007).

Derrible, S. & Kennedy, C. Characterizing metro networks: state, form, and structure. Transportation 37 , 275–297 (2010).

Roth, C., Kang, S. M., Batty, M. & Barthelemy, M. A long-time limit for world subway networks. J. R. Soc. Interface 9 , 2540–2550 (2012).

Newman, P. G. & Kenworthy, J. R Cities and Automobile Dependence: An International Sourcebook (1989).

Marshall, W. E. Understanding international road safety disparities: why is Australia so much safer than the United States? Accid. Anal. Prev. 111 , 251–265 (2018).

Organisation for Economic Cooperation and Development, Environment at a Glance: OECD Indicators. (Organisation for Economic Co-operation and Development, 2015).

Angel, S. et al. Atlas of Urban Expansion (Lincoln Institute of Land Policy, Cambridge, MA, 2012).

Accessibility Observatory, Access Across America. University of Minnesota, http://access.umn.edu/research/america/ (2017).

Wu, H. & Levinson, D. Access Across Australia. TransportLab, http://hdl.handle.net/2123/20509 (2019).

Pereira, R., Braga, C. K., Serra, B. & Nadalin, V. Desigualdades socioespaciais de acesso a oportunidades nas cidades brasileiras, 2019. Texto para Discussão IPEA 2535, http://repositorio.ipea.gov.br/handle/11058/9586 (2019).

Allen, J. & Farber, S. A measure of competitive access to destinations for comparing across multiple study regions. Geogr. Analysis 52 , 69–86 (2020).

Wu, H. & Levinson, D. Access Across New Zealand. TransportLab . https://hdl.handle.net/2123/21853 (2020).

Dimitrios, P. & Nicolas, W. Benchmarking accessibility in cities: measuring the impact of proximity and transport performance. Int. Transp. Forum Policy Pap. 68 , 8–374 (2019).

Batty, M. Accessibility: in Search of a Unified Theory (2009).

O’Sullivan, D., Morrison, A. & Shearer, J. Using desktop GIS for the investigation of accessibility by public transport: an isochrone approach. Int. J. Geogr. Inf. Sci. 14 , 85–104 (2000).

Levinson, D. & Kumar, A. The rational locator: why travel times have remained stable. J. Am. Plan. Assoc. 60 , 319–332 (1994).

Levinson, D. & Wu, Y. The rational locator reexamined: are travel times still stable? Transportation 32 , 187–202 (2005).

Mokhtarian, P. L. & Chen, C. TTB or not TTB, that is the question: a review and analysis of the empirical literature on travel time (and money) budgets. Transp. Res. A 38 , 643–675 (2004).

Marchetti, C. Anthropological invariants in travel behavior. Technol. Forecast. Soc. Change 47 , 75–88 (1994).

Zahavi, Y. & Ryan, J. The stability of travel components over time. Traffic Engineering and Control (1978).

Sarkar, S., Wu, H. & Levinson, D. Measuring polycentricity via network flows, spatial interaction and percolation. Urb. Stud. https://doi.org/10.1177/0042098019832517 (2019).

Mohring, H. Optimization and scale economies in urban bus transportation. Am. Econ. Rev. 62 , 591–604 (1972).

Verbich, D., Badami, M. G. & El-Geneidy, A. M. Bang for the buck: toward a rapid assessment of urban public transit from multiple perspectives in North America. Transp. Policy 55 , 51–61 (2017).

Pereira, R. & Schwanen, T. Commute Time in Brazil (1992-2009): differences between metropolitan areas, by income levels and gender. Instituto de Pesquisa Econômica Aplicada (Ipea). http://repositorio.ipea.gov.br/handle/11058/964 (2013).

Weingroff, R. F. Federal-aid highway act of 1956: creating the Interstate system. Public Roads 60 (1996).

Marshall, J. D. Urban land area and population growth: a new scaling relationship for metropolitan expansion. Urb. Stud. 44 , 1889–1904 (2007).

Niedzielski, M. & Kucharski, R. Impact of commuting, time budgets, and activity durations on modal disparity in accessibility to supermarkets. Transp. Res. D 75 , 106–120 (2019).

Jara-Díaz, S. R. & Videla, J. Detection of income effect in mode choice: theory and application. Transp. Res. B 23 , 393–400 (1989).

El-Geneidy, A. et al. The cost of equity: assessing transit accessibility and social disparity using total travel cost. Transp. Res. A 91 , 302–316 (2016).

Levinson, D. & Wu, H. Towards a general theory of access. J. Transp. Land Use 13 , 129–158 (2020).

Xi, Y., Miller, E. & Saxe, S. Exploring the impact of different cut-off times on isochrone measurements of accessibility. Transp. Res. Rec. 2672 , 113–124 (2018).

Pereira, R. H. Future accessibility impacts of transport policy scenarios: equity and sensitivity to travel time thresholds for Bus Rapid Transit expansion in Rio de Janeiro. J. Transp. Geogr. 74 , 321–332 (2019).

Paez, A. et al. Accessibility to health care facilities in Montreal Island: an application of relative accessibility indicators from the perspective of senior and non-senior residents. Int. J. Health Geogr. 9 , 52 (2010).

Wu, H. & Levinson, D. Unifying access. Transp. Res. D 83 , 102355 (2020).

Moya-Gómez, B. & García-Palomares, J. C. Working with the daily variation in infrastructure performance on territorial accessibility. The cases of Madrid and Barcelona. Eur. Transp. Res. Rev. 7 , 20 (2015).

Farber, S., Morang, M. Z. & Widener, M. J. Temporal variability in transit-based accessibility to supermarkets. Appl. Geogr. 53 , 149–159 (2014).

Fransen, K. et al. Identifying public transport gaps using time-dependent accessibility levels. J. Transp. Geogr. 48 , 176–187 (2015).

Imani, A. F., Miller, E. J. & Saxe, S. Cycle Accessibility and Level of Traffic Stress: A Case Study of Toronto. Technical Report . (Transportation Research Board of the National Academies, 2018).

Murphy, B. & Owen, A. Implementing low-stress bicycle routing in national accessibility evaluation. Transp. Res. Rec. https://doi.org/10.1177/0361198119837179 (2019).

Wu, H., Levinson, D. & Sarker, S. How transit scaling shapes cities. Nat. Sustainabil. 2 , 1142–1148 (2019).

Wu, H. et al. Metadata record for the manuscript: urban access across the globe: an international comparison of different transport modes. figshare . https://doi.org/10.6084/m9.figshare.13476867 (2020).

Download references

Acknowledgements

The data collection for Chinese cities was funded by the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant no. XDA19040402). Data collection and accessibility estimates for Brazilian cities were funded by the Access to Opportunities Project at the Institute for Applied Economic Research (Ipea). Data collection and accessibility estimates for cities in the Netherlands were funded by the ASTRID—accessibility, social justice and transport emission Impacts of a transit-oriented Development project (NWO project number: 485-14-038). The Canadian travel times and data collection was funded by the Social Sciences Research Council of Canada (SSHRC). Andrew Byrd of Conveyal helped prepare the input data and interpret results for Paris. Christian Quest of OpenStreetMap France processed the 2019 SIRENE database used for Paris job numbers. We knowledge the support of the Polish National Science Centre, within the POLONEZ program, allocated on the basis of decision No. UMO-2015/19/P/HS4/04067 based on the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No. 665778. US Data are from the Accessibility Observatory, funded by the National Accessibility Evaluation. Data collection and analyses for Australia and New Zealand were supported by TransportLab. Transit GTFS data for Douala, Cameroon, as well as the initial job distribution estimation for the African cities, were contributed by the World Bank and financially supported by UKAID/DfID through the Multi-donor Trust Fund on Sustainable Urbanization. We thank the above institutes and individuals for their generous contribution of data and funding.

Author information

Authors and affiliations.

School of Civil Engineering, University of Sydney, Sydney, Australia

Hao Wu & David Levinson

The World Bank, Washington D.C., USA

Paolo Avner

Polytechnique Montreal, Montreal, QC, Canada

Genevieve Boisjoly

Institute for Applied Economic Research (Ipea), Brasília, Brazil

Carlos K. V. Braga & Rafael H. M. Pereira

McGill University, Montreal, QC, Canada

Ahmed El-Geneidy

Key Laboratory of Regional Sustainable Development Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China

Jie Huang & Jiaoe Wang

University of California, Berkeley, CA, USA

Tamara Kerzhner

University of Minnesota, Minneapolis, MN, USA

Brendan Murphy

Institute of Geography and Spatial Organization, Polish Academy of Sciences, Warsaw, Poland

Michał A. Niedzielski

Faculty of Architecture and Town Planning, Technion—Israel Institute of Technology, Haifa, Israel

John P. Pritchard

Massachusetts Institute of Technology, Boston, MA, USA

Anson Stewart

You can also search for this author in PubMed   Google Scholar

Contributions

Study conception and design: D.L.; analysis and interpretation of results: H.W.; draft paper preparation: H.W.; accessibility calculation for respective nations/cities: H.W., P.A., G.B., C.B., A.E., J.H., T.K., B.M., M.N., R.P., J.P., A.S., and J.W. There is no difference in contribution between authors providing accessibility calculation for different nations/cities. All authors reviewed the paper.

Corresponding author

Correspondence to Hao Wu .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary information, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Wu, H., Avner, P., Boisjoly, G. et al. Urban access across the globe: an international comparison of different transport modes. npj Urban Sustain 1 , 16 (2021). https://doi.org/10.1038/s42949-021-00020-2

Download citation

Received : 11 June 2020

Accepted : 26 January 2021

Published : 27 May 2021

DOI : https://doi.org/10.1038/s42949-021-00020-2

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Energy and environmental impacts of shared autonomous vehicles under different pricing strategies.

• Shaopeng Zhong

npj Urban Sustainability (2023)

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Advertisement

Understanding Influencing Factors of Travel Mode Choice in Urban-Suburban Travel: A Case Study in Shanghai

• ORIGINAL RESEARCH PAPERS
• Open access
• Published: 26 March 2023
• Volume 9 , pages 127–146, ( 2023 )

Cite this article

You have full access to this open access article

• Jiankun Le 1 , 2 , 3 &
• Jing Teng   ORCID: orcid.org/0000-0003-2130-0208 2 , 3  

3794 Accesses

8 Citations

2 Altmetric

Explore all metrics

After the rapid expansion of the subway system over the past two decades, some cities are preparing to build more suburban railways. The emergence of suburban railways is bound to change the choice of suburban passenger transportation. This paper studies the factors that affect the choice of travel mode at the construction stage of suburban railways, aiming to design a more rational suburban railway network and urban public transport service system. Taking Shanghai as an example, this study first surveyed revealed preference (RP) and stated preference (SP) among urban-suburban travelers. Then, we used discrete choice models (DCM) and machine learning algorithms to build a travel mode choice model based on data collection and analysis. Furthermore, the importance of each factor was analyzed, and the effects were predicted under several traffic demand management schemes. Finally, this study proposed some strategies for increasing the share of public transport. On the one hand, it is suggested that Shanghai should continue to develop suburban railways and maintain low pricing of public transport services. Considering the construction and operation costs, the government needs to provide certain subsidies to stabilize prices. On the other hand, as passengers are very sensitive to the “last mile” trips in their suburban railway travel, transport planners should strengthen the connection from and to the suburban railway stations by developing services such as shared bikes and shuttle buses. In addition, the results indicated that some traffic demand management measures can also contribute to a larger share of public transport.

Similar content being viewed by others

Accounting for systematic heterogeneity across car commuters in response to multiple tdm policies: case study of tehran.

Modelling second-best choices from the choice-based sample: revelation of potential mode-switching behaviour from transit passenger surveys

Interurban Travel Mode Choice Model Which Based on Departures Frequency and Passengers’ Preferences

Explore related subjects.

• Artificial Intelligence
• Automotive Engineering
• Geoengineering

Avoid common mistakes on your manuscript.

1 Introduction

In the big cities of Western Europe and Japan, the well-developed multilevel urban public transport network has become one of the urban residents' main daily travel choices. In contrast, the fast and intensive construction of subway networks is still on the way in China and other developing countries. These cities are likely to enter a suburban railway construction boom in the coming decades [ 1 ]. Thus, it is necessary to study how the newly opened suburban railway will change the urban-suburban travel pattern of the city.

This paper aims to study the choice behavior of passenger travel mode in the suburban passenger corridor under the background of the suburban railway construction boom in China's big cities. This paper focuses on the influencing factors of the choice of suburban passenger travel modes. Taking Shanghai as an example, from the whole process of urban-suburban travel, the discrete choice model and machine learning method are used to construct various travel mode choice models, which are further evaluated and compared. Finally, we propose some strategies to enhance the competitiveness of urban-suburban public transport services based on the analysis.

The innovation of this study is to use both traditional and improved logit models and two machine learning methods. This study combines the results of various models to obtain more findings than those from the traditional discrete choice model. The results of multiple models can corroborate or complement each other.

The significance of this study is to investigate the travel behavior on urban-suburban passenger transportation corridors, taking Shanghai as a case study. Suggestions to enhance sharing rates of public transit are proposed. The constructed models can provide quantitative support for the design of various aspects of speed, fares, and connecting traffic in the planning and operation of suburban railways.

The rest of the paper is organized as follows. Section 2 reviews related literature. Section 3 demonstrates the methods adopted in the study. Section 4 presents the survey design with variables and scenarios. In Section 5 , using discrete choice models and machine learning algorithms, the modeling results are introduced, discussed, and summarized. Section 6 predicts the effects of various traffic demand management schemes based on the models from the two methods and gives suggestions to enhance the competitiveness of urban-suburban public transport; finally, Section 7 concludes the study.

2 Literature Review

2.1 urban-suburban travel mode choice.

Urban-suburban travel refers to moving between the main sections of a city (urban area/main area) and the suburbs of the city (new town). Urban-suburban travel usually covers longer distance and time, and possibly has more travel links than those that start and end in central or suburban areas. The differentiation between urban and suburban areas in comprehensive transportation construction also complicates the travel mode selection process. Travelers need to make decisions based on a variety of factors.

At present, there are few studies on urban-suburban travel choices, especially considering the accessibility of public transport services and the impact of new suburban railways on mode choice.

Espino et al. [ 2 ] used the RP/SP (revealed preference/stated preference) method to investigate travelers on suburban railway channels and analyzed the impact of several policies on the share rate of public transport. Monson and Gonzales [ 3 ] predicted the first year of passenger flow of the N-III passenger corridor in Madrid and found that the sharing rate of various transportation modes on the corridor would change significantly.

In the research on the suburban railway system in China, there needs to be more research on the travel choice behavior of suburban railways. Tiantian and Yaodong [ 4 ] studied the convenience of public transportation for suburban commuters in Beijing in 2021. Using the structural equation model, she found different perceptions of convenience for commuters with different characteristics and put forward suggestions for promoting suburban public transportation.

In some developing countries or regions where the composition of public transit systems differs from that of developed countries, travel mode choices for suburban trips may also differ. Only a few studies on suburban travel behavior are currently being conducted in developing regions. Danapour et al. [ 5 ] studied the impact of high-speed rail on travel demand for Tehran–Isfahan routes and used the SP survey method to obtain passengers' attitudes toward changing to high-speed rail travel. Dahlan and Fraszczyk [ 6 ] conducted a pre-launch study of the Jakarta Mass Rapid Transit (MRT) system through a survey in Jakarta, Indonesia, and three surrounding areas, and found that most of the respondents declared a willingness to a shift to MRT. Fraszczyk et al. [ 7 ] studied people’s willingness to shift to metro through a survey targeting a group of potential metro users located close to a planned metro line in Salaya, 20 km from the center of Bangkok, Thailand. Tiantian and Yaodong [ 4 ] studied passenger convenience during the transfer process from a more micro perspective by using questionnaires for passengers in the suburbs of Beijing and a structural equation model. The empirical results show that passengers' perception of the convenience of transfer in the suburban commute is characterized by comfort, safety, and quickness, with comfort coming first, then safety, and quickness last. The subjects of these studies may be close to urban-suburban travel, but they focused more on metro and less on the suburban railway.

2.2 Disaggregate Travel Survey Methods

The revealed preference (RP) and stated preference (SP) surveys are the more commonly used methods in travel behavior research. In recent years, researchers have used these methods to study various types of travel behavior.

Using the above RP or SP survey method, Cherchi analyzed the future travel-sharing rate of a railway in Italy that was about to be opened. Espino et al. [ 2 ] studied the travel mode choice behavior in two main urban-suburban travel corridors in Gran Canaria, Spain, and found that punishing car travel was more effective in stimulating the public's choice of public transport than improving the service of public transport. Ali Aden et al. [ 8 ] aimed to investigate the intentions and preferences of travelers toward car-sharing services in Djibouti, Africa, and the data were collected through an online stated preference (SP) survey. Shamshiripour et al. [ 9 ] designed an SP-RP survey implemented in Chicago to investigate how and to what extent people's mobility styles and habitual travel behaviors had changed during the COVID-19 pandemic.

2.3 Behavioral Theory of Travel Mode Choice

2.3.1 discrete choice model (dcm).

The discrete choice model (DCM) has been widely used in the study of travel choice. To date, discrete choice model theory has generally been classified into multinomial logit model (MNL), generalized extreme value model (GEV), mixed logit model, and multinomial probit model. Many transportation scholars have studied and applied DCM combing their own research needs, such as the multiple discrete continuous extremum model (MDCEV) [ 10 ], structural equation-based logit model (SEM-Logit) [ 11 ], and integrated choice and latent variable model (ICLV) [ 12 ].

2.3.2 Machine Learning Methods

In addition to the DCM, in recent years, many scholars have begun to use other models to analyze the travel choice problem, among which the supervised learning model in the machine learning type is more common.

Among them, the support vector machine (SVM) algorithm has developed rapidly since 1995 and has given rise to a series of extended algorithms [ 13 ], and has been widely used in data analysis and mining modeling of intelligent transportation systems in recent years. Cheng et al. [ 14 ] proposed a travel mode prediction method based on SVM and analyzed the travel mode choices of low-income commuters in Fushun. In order to verify the hypothesis that the household head's travel behavior will impact the travel frequency of his family members, Yang et al. [ 15 ] applied the MNL model and SVM model, respectively, when taking the travel data of Nanjing residents as a case. They found that the SVM performed slightly better than the overall average accuracy by comparing the prediction accuracy. Qian et al. [ 16 ] proposed a new SVM classification method, aiming to classify the travel mode selection data more accurately so that the model could predict the travel mode more accurately.

The random forest algorithm [ 17 ] is a combinatorial classification method proposed by Leo Breiman in 2001. Due to its ease of use and high prediction accuracy, it has been gradually applied in transportation science. Cheng et al. [ 18 ] proposed using the random forest algorithm to predict travel behavior and conducted a case study with household travel data in Nanjing.

3 Research Gap

In the early stage of suburban railway development in developing countries, there were few published results related to the travel choice behavior of suburban railway or urban-suburban passenger transport corridors.

The purpose of this work is to conduct a case study of Shanghai's urban-suburban passenger transport corridor on travel choice behavior. The study uses the classical discrete choice model and machine learning methods to investigate the characteristics of urban-suburban travel mode choice. Then we put forward related travel demand management policies based on these characteristics and predict the effect by different models.

In contrast to the previous literature, on the one hand, this study uses Shanghai (a city in a developing country where urban railways are being built) as a case study, while the previous literature has paid less attention to urban railways in developing regions, and has also studied less about suburban railways or urban-suburban corridors. On the other hand, this study experimented with machine learning methods and the traditional discrete choice models in its methodology. Additional findings are obtained by evaluating the results of the two methods.

4.1 Discrete Choice Model (DCM)

The theoretical basis of the DCM is that people choose the option with the greatest “utility” when they make a choice (the random utility theory). In the case of transportation, the sample option of “utility” of the same option differs for different travelers. The characteristics of the level of service, the attributes of the traveler, the purpose, and the cost of the trip all impact the traveler's perception of the “utility” of the option.

Random Utility Theory also assumes that “utility” is not a fixed value, but a random variable with a nonrandom and a random component, as Equation (Eq. 1 ) below.

where U represents the utility of a solution for a traveler, V represents the fixed part of it and \(\varepsilon\) represents the unfixed part.

As shown in Eq. ( 2 ) below, according to utility maximization theory, the probability that the traveler will choose one option \(i\) is

where A is the set of all available options. \(0\le {P}_{i}\le 1\) , \(\sum_{i\in A}{P}_{i}\) .

The observable component \({V}_{i}\) is generally considered to be the sum of the product of the variable \({X}_{k}\) , representing the influencing factor, and its weight parameter \({\alpha }_{k}\) , as shown in Eq. 3 , is

The basic DCM builds mathematical models based on the above assumptions, using mathematical methods to identify the variables \({X}_{k}\) that significantly impact travel utility and solve for their impact parameters \({\alpha }_{k}\) .

In general, the main models for discrete selection are the MNL, the GEV, the mixed logit model, and the multinomial probit model. Since 2010, many scholars have developed and applied complex models for different data forms and problem forms based on these common basic models.

Based on the frequency of use of each type of model in the study, the match with the sample data collected so far, and the effectiveness of the parameter estimation demonstration, the following types of DCM are selected for parameter estimation and testing in this study to conduct comparisons between modeling results.

The basic multinomial logit model

The MNL is a simple discrete choice model in which random utilities are set to follow independent extreme value distributions. The choice probability function is \(P ={ e}^{\beta x}/\Sigma { e}^{\beta x}\) , where the explanatory variable \(x\) can be either the individual socioeconomic characteristics of the chooser or the attributes of options.

The MNL model is the basis of the discrete choice model system, and several types of logit models have been improved to fix some of the drawbacks of the MNL model. In practical research, MNL models are also the most widely used due to their ease of use and low error rate.

The willingness-to-pay MNL model

The difference between the willingness-to-pay MNL model and the basic MNL model is that the levels of travel modes (such as time and comfort) can be multiplied by a parameter and added to the cost of travel, thus considering the cost of time and the perceived cost in the model.

In this study, the willingness-to-pay MNL model can provide a clearer, quantitative response to passenger perceptions of factors such as time and comfort under different travel modes, and support the proposal of relevant travel management strategies.

The perceived-time MNL model

The perceived-time MNL model differs from the basic MNL model in that it assumes that crowding, possible delays, and the distance from the location to the station will affect the passenger's perception of time. For example, passengers may perceive a longer journey time in a very crowded and uncomfortable carriage.

The perceived-time MNL model can multiply a number of travel mode attributes by their coefficients, add or multiply them with time, and subsequently obtain the total perceived time of the traveler. This model can be used to analyze the effect of improvement methods such as comfort enhancement on reducing perceived time.

The nested logit model

The MNL models have the inherent flaw of the assumption of IIA (independence of irrelevant options). When the IIA hypothesis is not tested, MNL cannot be used for modeling. The initial focus of academic research in DCM is to solve the IIA property, which leads to the development of the GEV. The GEV model is not a specific model but a generic term, with the most widely used model in practice being the nested logit model (NL).

The NL model places options that share specific characteristics in the same category. The questionnaire in this study classifies urban public transport as “rail transit” and “ground bus,” but there are many similarities between the two in terms of various factors, and several connection modes are common. Therefore, in the NL model, the two can be classified in one category.

4.2 Support Vector Machine (SVM)

Support vector machine (SVM) is a supervised learning algorithm for data classification. Its principle can be briefly summarized as follows. The training samples are separated by representing the points in the feature space with different coordinates, followed by finding the best plane (hyperplane) in the feature space that separates all the sample points. The advantage of SVM is that it can learn from small samples, emphasizes key vectors without being too sensitive to outliers, and has a strong generalization capability.

This study divides the dataset into a training set and a test set, allowing the SVM to learn prediction rules for travel mode choice from the training set data. Then, the model obtained from the training predicts the final chosen travel mode from the samples in the test set.

4.3 Random Forest

Random forest is also a supervised learning algorithm. The principle can be explained as follows. There is a forest made up of a group of decision trees. First, features are randomly selected for the trees, and the importance of each feature in each tree is tested. Then, all the decision trees are iterated toward the more important features. Finally, the prediction of the forest is obtained based on the results of all the decision trees. A significant advantage of this method is that it can be used for both classification and regression problems, and random forests can also measure the importance of each feature and give a ranking.

5 Research Design

5.1 research framework.

As shown in Fig. 1 , this paper proposes a research framework. The framework begins with a three-part data collection followed by preliminary statistical analysis (including demographic and socioeconomic analysis). Then, DCM and machine learning methods are used in the modeling part, and a comparison of the results of the two models and some findings are presented. In the last part of the framework, the three best-performing models are selected to predict the change in travel mode choice under 19 different policy schemes. Finally, we summarize the multi-part research and put forward some suggestions to enhance the competitiveness of public transit.

Research framework

5.2 Study Area: Public Transport and Suburban Railways in Shanghai

Shanghai, a megacity with a permanent population of more than 20 million, already has a metro system with 20 lines and an overground bus system with more than 1000 lines.

Compared with the metro and bus, Shanghai suburban railway planning and construction started later. Shanghai currently operated only one suburban passenger railway, Jinshan Railway. In 2016, Shanghai's suburban railway entered a period of rapid development. Nine suburban railways were under planning and construction, such as the Jiamin Line and the Airport Link Line [ 19 ]. Shanghai will have a suburban railway network of more than 1000 km by 2035.

Figure 2 , referring to the Shanghai Master Plan 2017–2035 [ 20 ], shows the downtown in red and the suburbs in light blue. The map shows the alignment of the three urban railway lines already in operation or due to be completed by 2025, namely the Jinshan Line, Airport Link Line, and Jiamin Line. Each of the lines straddles the urban and suburban areas to support the needs of the public for urban-suburban travel.

Shanghai's downtown and suburban areas, and the directions of suburban railways

In our study, a trip from the downtown to the suburbs (or from the suburbs to the downtown) is considered an urban-suburban trip. Given that this division is unclear to all participants, the questionnaire may use administrative divisions to distinguish between urban and suburban areas.

5.3 RP and SP Survey Questionnaire Design

5.3.1 the structure of the questionnaire.

This study uses quantitative methods where questionnaires serve as data collection tools to conduct data acquisition for suburban travel choice behavior, aiming to study the influence of multiple factors such as traveler attributes, travel attributes, travel mode attributes, and so on. The formal questionnaire design includes three parts: the socioeconomic attributes of travelers, the RP survey of suburban travel choice, and the SP survey of suburban travel choice.

The first part of the questionnaire investigates the socioeconomic attributes of the respondents, asking about seven attributes such as gender, age, occupation, average monthly income of family members, living area, the distance from his or her residence, and the nearest subway station ( D RS ), and family car usage.

The second part of the questionnaire is the RP survey. The questionnaire inquiries about the actual suburban trips that occur most frequently, including the travel purpose of this trip, the area of the destination, the distance between the destination and the nearest subway station ( D DS ), the one-way travel cost, the one-way travel distance, and the way to reach the public transport station. If the respondents could not give accurate answers to some questions, such as travel cost and distance in the questionnaire, they were asked to make simple estimates.

Attributes of the first and second parts of the survey are defined in Table 1 .

The third part of the questionnaire is the SP survey. The questionnaire gives a hypothetical scenario. Each scenario has three travel modes (rail transit, car, bus), in which rail transit includes subway, suburban railway and so on, and the car includes taking taxis and self-driving. Each mode's attributes (time in transit, one-way cost, possible delays, D RS , D DS , and so on) are assumed. Respondents were asked to choose one of the three travel modes in each scenario.

Twenty-four scenarios were generated using an orthogonal design. Six of them were randomly selected for each respondent to answer. Respondents were asked to answer the following five questions with three options (rail transit, car, bus) for each scenario.

In general, what is your chosen travel mode?

If a shuttle bus is opened between the place of residence and the station with greater frequency and appropriate time, the one-way cost will increase by up to 3 yuan. What would be your travel mode in this situation?

If the shared bicycles are placed near the residence and the station, the number is large, easy to park, and the travel cost is almost unchanged. What would be your travel mode in this situation?

If there is a special public parking lot ( P + R parking lot) next to the station with ample parking space available, and you can directly transfer to the subway after parking, the parking fee is 10 yuan/day. What would be your travel mode?

If there is a railway from the city to the suburbs, when choosing “subway or train,” the travel time can be shortened to 60 min, and the single trip cost becomes 15 yuan. What would be your travel mode?

5.3.2 The Attributes and Levels Design of the SP Survey

In addition to the common attributes of travel purpose, transit time, and travel cost, the attributes selected for the SP questionnaire include distance, congestion, number of interchanges, and possible delays. D RS and D DS are considered to investigate the relationship between the convenience of the passenger reaching the station and his or her choice of travel mode.

For the level design of each attribute, this study obtained the average level and range of variation through the official website of Shentong Metro, Jinshan Railway, Gaode Map software, the Didi TNC application, and references. For example, the level of travel time is limited to between 40 min and 120 min because this is the commuting time of the vast majority of Shanghai residents mentioned in The 2022 Commuting Monitoring Report of Major Cities in China [ 21 ]. The cost of public transport, on the other hand, is generally in the range of 2–15 yuan, which is determined by considering the pricing of Shanghai's metro, Jinshan Line, ground buses, and company buses.

Indicators that are difficult for respondents to visualize are described in a verbalized way (e.g., crowdedness, when it is 3, 6, and 10 people/m 2 is described in the questionnaire as “have seats, not crowded,” “no seats, crowded,” and “no seats, very crowded,” respectively [ 22 , 23 , 24 ]).

The attributes and levels of the SP questionnaire are designed as follows. Table 2 also explains what the different levels represent or where they come from.

5.3.3 The Scenario Designs

In designing the scenarios, the orthogonal design function of the statistical analysis software JMP Pro 13 is used to generate 24 choice scenarios, rejecting some combinations of factor levels that are too unattractive (e.g., highest cost but slowest speed). Due to interview length constraints and to improve the validity of the respondents' answers, each respondent answers five questions in six of the 24 scenarios randomly.

5.4 Data Collection and Sample Size Estimation

The questionnaire survey was conducted on residents living in the suburbs of Shanghai for a long time. We informed respondents at the beginning of the questionnaire that the travel behavior surveyed was for a period when COVID-19 outbreaks had not yet occurred (no large-scale COVID-19 outbreaks in Shanghai from March 2020 to March 2022).

Affected by the COVID-19 epidemic in Shanghai in 2022, this questionnaire survey was conducted in the form of an online survey in April 2022, and a professional survey company was commissioned to conduct the survey. After deleting some invalid results, A total of 575 valid questionnaires were collected. The following is a preliminary analysis of the survey data. After processing the data, a total of 15,540 valid travel mode choice data were generated.

Based on the following inequality Eq. 4 proposed by Orme [ 25 ] and Johnson and Orme [ 26 ] applicable to simple random sampling, this inequality can be used to determine sample size in SP surveys, taking into account crossover effects between factors.

where \({L}_{MAX}\) is the maximum value of the product of the number of levels of the two attributes, \(J\) is the number of options, and \(S\) is the number of choice tasks faced by each respondent.

The maximum value of the number of levels for a single attribute in the questionnaire is 4 and consists of two such attributes, so this survey has \({L}_{MAX}=16\) . The respondent has to choose between three modes of travel, so \(J=3\) ; each respondent answers the questions in six scenarios, so \(S=6\) . Therefore, the minimum value of the sample size is \(500\cdot \frac{16}{3\times 6}\approx 445\) . This survey collected 575 valid questionnaires, which met this requirement in terms of quantity.

6 Analysis of Results

6.1 statistical preliminary analysis, 6.1.1 participants’ demographic and socioeconomic characteristics.

The basic personal characteristics of the respondents in this survey were analyzed. As shown in Fig. 3 , from the perspective of gender, male respondents account for 44.35%, female respondents account for 55.65%, and the overall gender distribution is relatively balanced. From the perspective of age, 2.17% are under 18 years old. Respondents aged 18–30 and 31–40 years account for 48.7% and 38.26%, respectively.

Gender (left) and age (right) of respondents for sample collection

Participants are relatively evenly split in terms of gender, and the majority of respondents are between the ages of 18 and 59, which is the most active age group in travel behavior. It may also be related to the limitation that the elderly have fewer travel times, and many online surveys are difficult to cover the elderly.

Monthly per capita household income is a better indicator of overall affordability than personal income, so it is chosen as the income item (referred to as “income” in subsequent research and analysis) for the survey. The results are more evenly distributed across the income bands.

Regarding occupation, general employees, enterprise managers, and students account for a large proportion, reaching 39.13%, 21.3%, and 19.57%, respectively. Public data from the Shanghai government shows the average monthly salary in Shanghai is around 12,000 yuan, and the median is around 6400 yuan, which aligns with the survey results.

As shown in Fig. 4 , 66.52% of the respondents confessed that they or their families own a car; 33.48% of respondents do not own a car. Public data show that the number of cars per 100 households in Shanghai is around 70–80, and the socioeconomic results are consistent with the actual situation.

The income (left) and the ownership of private cars (right) of the respondents' households collected in the sample

From the above basic personal characteristics, it is determined that the respondent group of this survey covers different gender, age, occupation, and income, which is relatively comprehensive.

6.1.2 Preliminary Analysis of RP Survey Results

The basic RP data are treated to analyze the situation of travelers on their most common suburban trips. From the perspective of travel purpose, 60.48% of the travel purpose is commuting, indicating that commuting is the most common purpose of urban-suburban travel.

From the perspective of travel choice, 65.32% of respondents choose the subway or suburban railway, 25.81% choose self-driving, carpooling, or taxi (online car-hailing), and 8.87% choose the ground bus.

From the perspective of travel time, 60.48% of the suburban travel time is more than 60 min, and 13.71% of the suburban travel time is more than 90 min. Regarding travel expenses, 51.61% of the respondents estimate their expenses to be less than 10 yuan. The vast majority (97.6%) of these low-cost-trip participants choose public transport, reflecting the low-cost character of public transport.

Regarding travel distance, 66.94% of the respondents estimate that their suburban travel distance is between 20 and 60 km, which is in line with the distance and time of suburban travel every day in Chinese megacities.

In terms of the way to get to the subway station or suburban railway station, 58.87% of the respondents reach the station by riding, short-distance shuttle, driving, or taking a taxi. The basic information about the data collected by the RP questionnaire is shown in Table 3 .

6.1.3 Crossover Analysis

In the crossover analysis, we study the relationship between travel mode choice and certain attributes, which are shown in Figs. 5 , 6 and 7 . This analysis helped in the subsequent modeling.

Crossover analysis of gender (or “CarOwn”) and travel mode choice

Crossover analysis of “Income” and travel mode choice

Crossover analysis of D RS and travel mode choice

It is observed that the probability of men choosing self-driving or taxis is higher than women, and women are more likely to choose public transportation than men. This situation may be due to the larger group of male drivers. The respondents with a car are significantly more likely to choose to drive themselves than those without a car.

Moreover, the higher the income of the respondents, the greater the proportion of not choosing public transport. In the income bracket of more than 20,000 yuan, about half of the respondents do not choose public transport for suburban travel. For the general respondents, the farther the rail transit station is from the destination or residence, the greater the proportion of people who choose car travel, and the rate of taking the ground bus also increases to a certain extent.

In the analysis of the cross-section of distance and travel mode, the pattern of “the further the distance from the metro or rail station, the less likely you are to choose rail travel” is basically present. In general, the further the rail station is from the destination or residence, the greater the proportion of trips made by car, and the rate of ground transportation increases to a certain extent.

6.2 Modeling of Discrete Choice Model

This study uses a variety of discrete choice models, and finally uses the base MNL model with good modeling effect, the willingness-to-pay MNL model, the perceived-time MNL model, and the NL model. The independent variables in the models are the various factors mentioned above, while the dependent variables are the utility values that affect the probability of travel mode choice.

The results of the model parameter calibration are presented in Tables 4 , 5 , and 6 . If the parameter estimate for this factor is blank, it means that there is no involvement of this factor in this model.

6.2.1 The MNL Model

Model 1: RP data modeling

In modeling RP data, we consider that the utility functions of the three ways have constant terms. On the one hand, whether to own a car will affect the probability of choosing a car for travel; on the other hand, people with higher income levels are more inclined to travel by car. At the same time, because statistics show that male drivers account for the majority, gender differences may also be related to the travel mode choice.

The discrete choice model program package Apollo of R language is used to estimate parameters and test parameters based on RP survey data and the above model, and the parameters with insignificant influence are eliminated. Finally, the parameter calibration results are obtained, as shown in Table 4 .

According to the parameter estimation results, the distance to the station is the most significant variable. The coefficient is negative, indicating that the farther the distance between the destination and the rail transit station is, the fewer travelers are willing to choose rail transit travel. The income level also has a significant impact on the travel mode, which is negative for rail transit and ground bus, and the absolute value of the ground bus coefficient is larger, indicating that when other conditions are unchanged, the higher the income, the less willingness to choose public transportation, especially the ground bus.

The coefficient of owning a car on traveling by car is positive, and the absolute value is large, which reflects that “car owners” are significantly more willing to choose car travel. In terms of gender, according to the preliminary analysis of the data in the previous chapter, it is believed that women are more likely to choose public transportation than men.

Model 2: SP data modeling

Based on the modeling of RP data, SP data is modeled. The coefficients of “car ownership,” “gender,” and “income” only exist in the utility function of the car, and the coefficients of D RS and D DS only exist in the utility function of the rail transit and bus, and the coefficients of “whether to use a certain connection mode” and “crowding degree” only exist in the utility function of the rail transit and public transit. The final parameter calibration results are obtained as shown on the right side of Table 4 , next to the results of Model 1.

Overall, most of the travel attribute factors have a significant impact on the choice of travel mode, and the significance of the travel time is of the highest and negative value, reflecting that the length of the travel time is still one of the most important travel attributes for travelers to choose travel modes.

Income and car-own, two factors with high significance in the RP model, are also very significant in the SP model. A positive value indicates that people with higher incomes and cars are more willing to choose cars for travel.

In terms of the congestion degree, the test value is significant, while the absolute value of the parameter is not large, indicating that the effect of the congestion degree of public transportation is low, but this could also be due to the level of crowding is not communicated well to the respondents.

The impact of possible delay on mode choice is more intense in public transportation, and for cars, the estimated value is small and insignificant, which can be understood as residents who choose to travel by car having a psychological expectation of road congestion and possible delay.

The test values of all parameters are relatively significant in terms of the three connection methods. The most favored connection method of respondents is the shared bicycle, followed by shuttle car, and finally “ P + R .”

6.2.2 The Willingness-to-pay MNL Model and Perceived-time MNL Model

Based on the SP-based MNL model in the previous section, the attributes such as travel time, congestion degree, and possible delay with their respective parameters, plus the travel cost and multiply the sum by the cost coefficient, are in the willingness-to-pay MNL (WTP-MNL) model (Model 3). The calibration results are shown in Table 5 .

The parameter estimation of the WTP-MNL model is more significant than the basic MNL model overall, and the characteristics of the factors are similar to the basic MNL model.

Based on the SP-based MNL model in the previous section, the D RS and D DS coefficient is modified by the product of the connection mode, considering that the connection tool will affect the distance between arrival and departure stations. On the other hand, the sensing time of the crowding degree is related to the traveling time, so the product of the two is multiplied by the crowding coefficient alone. Finally, these “sensing time terms” are added to the traveling time to obtain the total sensing time and then multiplied by the coefficient of the sensing time, which is the sensing time-MNL model (Model 4). The obtained utility function and parameter calibration results are shown in Table 5 , next to the results of Model 3.

6.2.3 NL Model

Based on the basic MNL model of RP and SP data, considering that rail transit and ground public transportation are similar in timeliness, economy, comfort, and other aspects, a two-layer tree structure is constructed, and “rail transit” and “ground bus” are classified as “public transportation service,” and the selection tree is as following Fig. 8 .

Schematic of the NL model

Then we modify Model 1 (RP data) and Model 2 (SP data) into Model 5 and Model 6, respectively, using the selection tree above. The results of parameter calibration in the Apollo package are shown in Table 6 .

In the parameter calibration of the NL model based on RP results, its overall significance is not as good as that of the basic MNL model, which may be because RP data is based on actual travel behavior. There are still factors not involved in the survey in the actual travel behavior to participate in the choice of travel modes, and there is no strong similarity within public transportation.

In the parameter estimation based on SP data, the overall significance is stronger than that of the basic MNL model, and the characteristics and disadvantages of the factor are similar to that of the basic MNL model.

6.3 Modeling of Machine Learning Methods

This study selects two common machine learning algorithms for classification problems. Based on the same set of SP survey data (15540 data is divided into 75% training set and 25% test set), the two algorithms are used to build a trip choice model and test its accuracy. The results are compared with those under the discrete choice model.

6.3.1 Support Vector Machine

Support vector machine (SVM) is a supervised learning algorithm for data classification. Its advantages are that it can learn based on smaller samples, emphasize key vectors, is not too sensitive to outliers, and has strong generalization ability.

In this study, the SP training set is modeled using the LIBSVM program package based on R Language, which is commonly used in data science research. After parameter adjustment, when the function type of nonlinear transformation is the Gaussian kernel function, and the penalty function is set to 3 (gamma=0.7), the prediction accuracy of the test set is the highest, reaching 63.84%.

6.3.2 Random Forests

In this study, randomForest , a program package based on R language, is used to model the training set. When the decision tree size is 2500, the prediction accuracy of the training set is the highest, reaching 63.11%, and the overall accuracy of the test set is 62.68%.

In addition, the importance order of each factor in SP and RP data is printed, as shown in the following Figs. 9 and 10 . Each figure has two diagrams. Mean decrease accuracy represents the degree of reduction in the prediction accuracy of the random forest when this option is randomly selected. Mean decrease Gini computes the effect of each variable on the heterogeneity of observations at each node of the classification tree, thus comparing the importance of variables. The higher the value, the more important the variable is.

Importance ranking of factors in RP models by random forest

Importance ranking of factors in SP models by random forest

According to the factor importance ranking of RP data (shown in Fig. 9 ), it is found that work, travel purpose, gender, and total travel distance are low in the two diagrams and D RS and D DS (DOS and DDS in Fig. 9 ), income, and age are high. The variable “whether to own a car” ranks third in the accuracy diagram. The low ranking in the Gini diagram is probably due to its binary nature.

In ranking factor importance based on SP data (shown in Fig. 10 ), factors such as income, age, gender, job, and whether to own a car rank high, and the ranking of travel time is also relatively high. The rank of crowding degree is in the middle. Possible delays and D RS and D DS (dis1 and dis2 in Fig. 10 ) are low on the list.

Additionally, in DCM models the parameters of some factors have to be uniform and cannot be differentiated by travel modes in order to ensure significance (for example, factors such as cost and time have to be uniformly parameterized). However, in a random forest model, researchers can obtain more detailed results. For example, in the ranking of mean decrease accuracy and mean decrease Gini, travel time by rail is more important than time by bus or car, and people are most concerned about the crowdedness and possible delays of buses and cars. In terms of the cost factor, it is the cost of travel by private car that has the greatest impact.

6.4 Major Findings

The following main findings can be obtained by combining the parameter estimation and factor importance ranking of the above models:

Time has more influence on travel choice and may be greater than the impact of cost. In all models, the time parameter continues to have strong significance. In the random forest model, the importance of the time parameter on the road is also relatively high.

Suburban railways show advantages in the time–cost ratio. The timeliness of the rail transit network, including suburban railways, is enhanced (30–60 min can be saved), and the travel cost is not significantly increased (10 yuan for Jinshan Line and assume 10 to 15 for other suburban lines in Section 4 ) compared to the metro ticket. Compared with cars, the advantage of the low price of public transit is retained.

Among the three connection modes, the proportion of travelers influenced by them to use public transit: Shared bicycle>short distance shuttle > P + R . Considering the dense bus network stations, many connection modes may be more valuable in the travel chain dominated by rail transit.

The delay coefficient of car travel is not obvious in the parameter estimation. The absolute value is small, which may be because the passengers who choose car travel have a psychological expectation for the possible delay in highway traffic.

Travelers who own cars are stickier to car travel, which is reflected in the strong significance of the Car-own coefficient. Moreover, because of their high income (reflected by the income parameter), they are less sensitive to direct travel costs than public transport travelers. On the contrary, travelers who do not choose cars or have not yet purchased a car are also stuck to public transport, which means that they generally do not change their public transport travel rules.

7 Prediction of the Effect of Transportation Demand Management Schemes

According to the results in previous sections, considering each model's advantages and disadvantages, the basic MNL-SP model, NL-SP model, and support vector machine model with better overall performance are selected to predict the change in travel mode choice under 19 different policy schemes. Predictions of the changes in travel mode choices are made under these assumptions. The 19 schemes and predicted results are shown in Table 7 .

In Section 6.1 , we will explain why these schemes are proposed and show the predictions of these schemes. Our suggestions are derived from the research in Sections 4 and 5 and are presented in 6.2.

7.1 Schemes and Effect Predictions

7.1.1 improve timeliness of rail transit.

The timeliness of the travel mode is still a key factor for travelers. This study considers the increase in timeliness from two perspectives. Schemes 1 and 2 are designed to assume the disappearance of extremely long-time trips, while Schemes 3 and 4 are aimed at a general reduction in travel time.

The results show that the effect of Scheme 1 is like that of Scheme 3, and the effect of Scheme 2 is like that of Scheme 4. However, the effect of Schemes 2 and 4 is much more significant than Schemes 1 and 3, and are among the most effective among 19 schemes. The results suggest that Shanghai should continue to promote the construction of municipal railways and subways to improve the timeliness of rail transit.

7.1.2 Improve the Portability of Rail Transit

Apart from the cost, time, and connection method, the congestion degree and the distance between the station and the residence or destination are the most significant travel mode factors. Therefore, Scheme 5 is to improve comfort (reduce carriage congestion), and Scheme 6 is to improve convenience (to shorten the distance between rail transit stations and travelers). The prediction results show that these two schemes have a relatively weak influence.

7.1.3 Improve the Accessibility of Rail Transit

In the SP survey, for each scenario, each respondent was provided with three access modes to and from public transport stations: shuttle bus, shared bicycle, and P + R . We examine the impact of providing multiple connection methods simultaneously by Schemes 10 to 13.

According to the table of prediction results, when there are both shuttle cars and shared bicycles (Scheme 10), the impact is the most significant, and schemes 11 and 12 have a small impact. The P + R mode is more expensive and laborious than other connection methods in construction and management, and the development of the P + R mode in China is not mature enough. Choosing the appropriate placement of shared bicycles and the reasonable arrangement of connecting buses is better.

7.1.4 Restrict Car Travel

Although car travelers may be less sensitive to the increase in travel costs, increasing the cost of car travel is a common and practical travel demand management measure. The government usually uses congestion pricing (Schemes 7 and 8), driving restrictions based on license plates (Scheme 9), and other measures to curb the number of private cars.

Schemes 7 and 8 assume that the car travel cost increases by 50% and 100%, respectively. The results show that the impact is significant.

Scheme 9 assumes that people can no longer travel by themselves but choose public transit or taxis (no carpooling), and the cost of car options is increased to the taxi price level (80 yuan). The results show that scheme 9 has a relatively significant impact on the basic MNL model.

7.1.5 Using Multiple Methods

According to the results of schemes 14 to 19, it is found that when time and cost measures are used together, the overall impact is more significant. Multiple measures can be considered in specifying the competitiveness of suburban public transport services, which may be better than implementing one measure alone.

7.2 Suggestions to enhance competitiveness of public transit

Improving the timeliness of the public transport system through efforts to develop suburban railways. The predictions in Schemes 1 to 4 show significant passenger demand for more excellent timeliness in the public transport system. If travel times can be reduced by 30%, or extra-long trips over 60 min can be eliminated, the rail share could be significantly increased.

The Shanghai Metro generally travels at speeds below 40 km/h, with most trains stopping at every station and shorter station spacing (generally 1–5 km). In this case, consideration should be given to improving the timeliness of the rail network through the development of a faster (fastest travel speed of around 100 km/h) suburban railways with fewer stations and greater station spacing (generally 3–10 km).

It is desirable that suburban rail pricing remain largely consistent with metro pricing rates. In the SP experiment we found that if other suburban lines were kept at the same price level of no more than 15 yuan as the Jinshan Railway (10 yuan), they would not only have the advantage of timeliness, but their price advantage would be substantially retained. If it were to work with the metro to offer various combined fares, it might attract more passengers and get them into the habit of traveling by public transport.

Because of the high investment and low pricing, it is difficult for municipal rail operators to profit in the short term. Therefore, subsidies could be provided to the operators to ensure the long-term operation of the transport services, or allow them to profit from land sales and real estate development.

Scientific placement of shared bikes near stations and running shuttle buses. Schemes 10 to 14 demonstrate that shared bikes and shuttle buses are the more significant connection methods in terms of impact. Bicycle sharing has the most significant impact overall, with the added benefit that its investment often comes from private companies and is less costly for the government.

Shuttle buses are preferred by those who cannot cycle or are in bad weather. Priority can be given to smaller, more frequent, demand-responsive forms of public transport to reduce waiting time for transfers. In order to achieve the goal of linking the various modes of transport, consideration could be given to integrating the operators of the public transport system, to provide a more complete transport service.

Using methods such as license restrictions to increase the cost of car travels. Schemes 8 and 9 contribute more significantly to the increase in rail sharing. The prediction results show that for every 10% increase in the cost of car travel, the rail share rate is likely to increase by 1–2%. This reflects the continued sensitivity of travelers to the cost of traveling by car. Considering that license restrictions have been implemented in some Chinese cities, policymakers can continue to consider this policy while looking at other measures, such as reducing subsidies for car purchases and increasing parking fees.

The combination of multiple types of measures is more effective than a single measure. The results of schemes 14 to 19 can prove this.

8 Conclusions

This paper focuses on the passengers' travel choice behavior in urban-suburban passenger transport corridors. Based on the collected RP and SP data, we apply the discrete choice models and machine learning methods to construct the models, which are further compared and comprehensively analyzed. At the same time, this research focuses on the impact level of each factor on the selection decision considering a newly built urban railway or providing connection services to the suburban public transit stations.

The following findings are obtained.

It is found that time and cost play a vital role in travel choice, and the low price of suburban railways (10–15 yuan) compared with car travel still retains the advantage of the low price of public transport, which is attractive to many travelers.

Rail transit passengers in Shanghai prefer shared bicycles and shuttle buses on the “last mile” trips, but the P + R transfer method is not as attractive.

People who do not use public transport are more insensitive to the price and remain more tolerant of predictable congestion and delays. Strong charging policies such as congestion charges, traffic bans, and restrictions on car purchase qualifications may discourage car travel.

Based on the above findings, we make the following suggestions.

Shanghai should continue to promote the planning and construction of the suburban railway and maintain the one-way price level at 10–15 yuan.

Scientific placement of shared bikes or logical operation of feeder buses near stations can improve the accessibility of rail transit.

Considering the construction and operation costs, the government needs to provide certain subsidies to stabilize prices.

Use methods such as license restrictions to increase the cost of cars because participants are sensitive to the cost of car travel.

Implementing multiple measures in traffic demand management is also suggested, which has a significantly better effect than a single measure.

This study uses discrete choice models and machine learning algorithms to construct a travel mode choice model of suburban passenger transport corridors. In Section 5 , the results of the two methods support each other, highlighting the most significant and influential factors, such as cost and time. In Section 6 , based on the findings of the previous section and considering the social and transport development of Shanghai, 19 schemes are proposed. Conclusions and suggestions are obtained by comparing the prediction results of the two types of models.

Additionally, the research and prediction methods that combine actual conditions with different algorithms can provide application data support in speed designs, station spacings, and connecting traffic modes when designing or operating public transport services.

However, the collected RP and SP data in this study are not fused because of the difference in data sizes. Further studies may consider integrating the two approaches used in this paper, in which the role, value, and effect of travel mode choice need to be explored. This paper does not further integrate the two approaches in use. Further studies can continue exploring the value of combining new methods (such as machine learning) with the DCM in travel mode choice, studying the effect it can bring when used with the traditional discrete choice models.

As urbanization enters the era of metropolitan area, railway construction in big cities is booming. CBNWeekly. https://www.yicai.com/news/100043552.html

Espino R, de Dios OJ, Román C (2007) Understanding suburban travel demand: flexible modelling with revealed and stated choice data. Trans Res Part A Policy Pract 41:899–912. https://doi.org/10.1016/j.tra.2007.03.002

Article   Google Scholar  

Monzon A and Gonzales J (2000) Travel demand impacts of a new privately operated suburban rail in the Madrid n-iii corridor. Planning & Management of Public Transport Systems Proceedings of Seminar E.

Tiantian J, Yaodong Z (2020) Research on the development of beijing suburban public transportation based on SEM. J Phys Conf Ser 1549(2):022027

Danapour M et al (2018) Competition between high-speed rail and air transport in Iran: the case of Tehran-Isfahan. Case Stud Transport Policy 6(4):456–461

Dahlan A, Fraszczyk A (2019) Public perceptions of a new MRT service: a pre-launch study in Jakarta. Urban Rail Transit 5:278–288. https://doi.org/10.1007/s40864-019-00116-0

Fraszczyk A, Weerawat W, Kirawanich P (2019) Commuters’ willingness to shift to metro: a case study of Salaya, Thailand. Urban Rail Transit 5:240–253. https://doi.org/10.1007/s40864-019-00113-3

Ali Aden W, Zheng J, Ullah I, Safdar MN (2022) Public preferences towards car sharing service: the case of Djibouti. Front Environ Sci 10:449

Shamshiripour A, Rahimi E, Shabanpour R, Mohammadian A (2020) How is COVID-19 reshaping activity-travel behavior? Evidence from a comprehensive survey in Chicago. Trans Res Interdiscip Perspect 7:100216–100216

Google Scholar  

Calastri C, Pawlak J, Batley R (2022) Participation in online activities while travelling: an application of the MDCEV model in the context of rail travel. Transportation 49:1–27. https://doi.org/10.1007/s11116-021-10166-8

Han Y, Li W, Wei S, Zhang T (2018) Research on passenger’s travel mode choice behavior waiting at bus station based on SEM-logit integration model. Sustainability 10(6):1996. https://doi.org/10.3390/su10061996

Toorzani A, Rassafi A (2022) Pro-environmental attitude and adherence to a travel mode in an integrated choice and latent variable (ICLV) model: results from a revealed preference survey. Int J Civ Eng 21(2):235–249. https://doi.org/10.1007/s40999-022-00757-6

Cortes C, Vapnik V (1995) Support-vector networks. Mach Learn 20:273–297. https://doi.org/10.1007/BF00994018

Article   MATH   Google Scholar  

Cheng L, Chen X, Lam HKW, Yang S and Lei D (2017) Public transit market research of low-income commuters using attitude-based market segmentation approach: Case study of Fushun, China. In Social Economic, Sustainability, and Human Factors in Transit, vol 2671(1). National Research Council, pp. 10–19 https://doi.org/10.3141/2671-02

Yang S, Deng W, Deng Q, Fu P (2016) The research on prediction models for urban family member trip generation. KSCE J Civ Eng 20:2901. https://doi.org/10.1007/s12205-016-0806-9

Qian Y, Aghaabbasi M, Ali M, Alqurashi M, Salah B, Zainol R, Moeinaddini M, Hussein EE (2021) Classification of imbalanced travel mode choice to work data using adjustable SVM model. Appl Sci 11(24):11916. https://doi.org/10.3390/app112411916

Breiman L (2001) Random forests. Mach Learn 45:5–32. https://doi.org/10.1023/A:1010933404324

Cheng L, Lai X, Chen X, De Vos J, Witlox F (2019) Applying a random forest method approach to model travel mode choice behavior. Trwaissavel Behav Soc 14:8415234. https://doi.org/10.1016/j.tbs.2018.09.002

National development and reform commission (2021). Planning of multi-level rail transit system in the Yangtze river delta region. https://www.ndrc.gov.cn/xxgk/zcfb/ghwb/202107/P020210702639387154556.pdf

Shanghai urban planning and land resource administration bureau (2018) SHANGHAI MASTER PLAN 2017–2035. https://www.shanghai.gov.cn/newshanghai/xxgkfj/2035002.pdf

Ministry of housing and urban-rural development, China academy of urban planning and design & Baidu map. (2022). The 2022 commuting monitoring report of major cities in China. https://huiyan.baidu.com/cms/report/2022tongqin/

Zhang Z, Guan H, Zhu H, Zhu J (2019) Analysis of travel mode choice behavior considering the indifference threshold. Sustainability 11(19):5495. https://doi.org/10.3390/su11195495

Hu H, Xu J, Shen Q, Shi F, Chen Y (2018) Travel mode choices in small cities of China: A case study of Changting. Trans Res Part D Transport Environ 59:361–374. https://doi.org/10.1016/j.trd.2018.01.013

Zhu H, Guan H, Han Y, Li W (2019) A study of tourists’ holiday rush-hour avoidance travel behavior considering psychographic segmentation. Sustainability 11(13):3755. https://doi.org/10.3390/su11133755

Orme B (1998) Sample size issues for conjoint analysis studies. Inc S S. Sequim

Johnson R and Orme B (2003) Getting the Most from CBC. Sawtooth software research paper series. http://www.sawtoothsoftware.com/education/techpap.shtml

Download references

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by Research on suburban railway operation management system under the background of integration of Yangtze River Delta (2021F023).

Author information

Authors and affiliations.

Urban Mobility Institute, Tongji University, 4800 Caoan Highway, Shanghai, People’s Republic of China

Key Laboratory of Road and Traffic Engineering of the Ministry of Education, College of Transportation Engineering, Tongji University, 4800 Caoan Highway, Shanghai, People’s Republic of China

Jiankun Le & Jing Teng

Shanghai Collaborative Innovation Research Center for Multi-network and Multi-modal Rail Transit, Tongji University, Shanghai, 201804, People’s Republic of China

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Jing Teng .

Ethics declarations

Conflict of interest.

The authors declare that they have no conflict of interest.

Additional information

Communicated by Baoming Han.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file 1 (DOCX 20 kb)

Supplementary file 2 (docx 24 kb), rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Le, J., Teng, J. Understanding Influencing Factors of Travel Mode Choice in Urban-Suburban Travel: A Case Study in Shanghai. Urban Rail Transit 9 , 127–146 (2023). https://doi.org/10.1007/s40864-023-00190-5

Download citation

Received : 22 September 2022

Revised : 19 December 2022

Accepted : 28 January 2023

Published : 26 March 2023

Issue Date : June 2023

DOI : https://doi.org/10.1007/s40864-023-00190-5

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Travel mode choice
• Urban-suburban transportation corridors
• Discrete choice models
• Influencing factors
• Machine learning algorithm
• RP and SP survey
• Find a journal
• Publish with us
• Track your research

10 Methods Of Transportation, Ranked From The Least To The Most Eco-Friendly

Your changes have been saved

Email is sent

Email has already been sent

Please verify your email address.

You’ve reached your account maximum for followed topics.

America’s Coolest Small Town Is In North Carolina

This charming oregon town is the state's hottest point, 7 state parks just like glacier national park, but better.

• Air travel is convenient and can transport large numbers of passengers, but it has a high environmental impact due to carbon emissions and fuel needs.
• Gas-consuming vehicles provide convenience and flexibility but have high carbon emissions and contribute to traffic.
• Public transit like buses, trains, and subways are the most eco-friendly option due to their high capacity and low per-person emissions. Walking and cycling have no carbon footprint but are limited in practicality and range.

Transportation is one of the most important aspects of planning a vacation. Whether plotting stops on an epic road trip or booking a plane ticket for an island getaway, figuring out how to get there is half the challenge.

When it comes to transportation, not all options are created equal. Different options have a variety of pros and cons, including convenience, accessibility, and cost.

One other less thought-of factor when it comes to transportation choices is its environmental impact. Those looking for more environmentally friendly ways to travel may opt for different means of transportation than they normally would. From emissions per person to fuel type, these popular forms of transportation have been ranked based on their environmental impact.

Wondering how these modes of transportation were selected? Data pulled from sources like National Geographic and the US Environmental Protection Agency (EPA) was cross-referenced to create this ranking. All modes of travel have pros and cons, and finding the right balance of practicality and environmentally-conscience is key to a successful travel experience.

Related: What Is Eco-Travel? Here's Every Type Of Ecotourism You Should Know About

10 Air Travel

There is no denying that air travel is an integral part of the modern travel experience. As explorers realize the beauty of every corner of the globe, planes have quickly become the most popular way to reach those destinations.

Airplanes can transport hundreds of passengers in one trip, which is usually a sign of more eco-friendly transportation. Unfortunately, the gas burned to achieve liftoff is simply too much to overcome, leaving air travel at the bottom of the list of eco-friendly travel options.

So, what are the options when a plane ride is unavoidable? Simplifying flight plans, particularly to destinations like Europe, is a great alternative. While a flight from the US to Europe may be necessary, once on the continent, visitors can skip the short European flights and take advantage of abundant public transportation or book an excursion with one of many eco-friendly tour companies around the world that operate in Europe (some of which specifically plan plane-free itineraries!).

• Pros of Air Travel: Time Saving, Travel Long Distances, High Capacity
• Cons of Air Travel: High Fuel Needs, High Carbon Emissions, Low-Capacity Flights Have High Emission/Person Ratio

9 Personal Combustible Engine (Gas-Consuming) Vehicle

Arguably the most common form of transportation, particularly in the US, personal gas vehicles are one of the least eco-friendly modes of transportation available today. Especially when used by a single person, vehicles with a combustible engine have one of the highest per-person CO² emission levels of any mode of transportation.

Despite its lack of eco-friendly stats, there is a reason cars continue to be a favorite for travelers. Their convenience and flexibility are undeniable, and that comfort is hard to sacrifice, even for the sake of sustainability. Thankfully, more eco-friendly vehicles are becoming available. Even those not interested in purchasing one can look at renting an electric or hybrid vehicle for their next cross-country road trip.

• Pros of Gas-Consuming Vehicles: Convenient, Flexible
• Cons of Gas-Consuming Vehicles: High Fuel Needs, High Carbon Emissions, Traffic Producing

8 Personal Motorcycles And Motorized Scooters

Like traditional personal vehicles, motorcycles and motorized scooters use gas to run, a disadvantage that automatically ranks them lower on the eco-friendly scale. These vehicles benefit slightly from their smaller size, meaning their impact is slightly reduced.

Also, like cars, motorcycles can be swapped for Electric options without an expensive purchase. Rentable electric scooters, in particular, are increasing in popularity in urban areas. These simple swaps make it possible to keep the autonomy that makes these vehicles so popular while also helping out realistically.

• Pros of Personal Motorcycles/Motorized Scooters: Convenient, Flexible, Lower Fuel Needs Than Larger Vehicles
• Cons of Personal Motorcycles/Motorized Scooters: Relatively High Carbon Emissions

7 Plug-In Electric Hybrid Vehicles

In this modern world, cars are one of the best ways to get from place to place efficiently while still having the flexibility to spontaneously pull over and enjoy a detour. While common gas vehicles don't rank high in terms of eco-friendliness, some of their alternatives can offer various degrees of improvement based on how fully they move away from those carbon-producing fuels.

The plug-in electric hybrid vehicles rank a step ahead of the traditional, able to charge up for a short trip without dipping into the gas tank. This alternative is particularly popular with those looking for a budget-friendly way to make their trip more environmentally friendly.

• Pros of Plug-In Electric Hybrid Vehicles: Lower Fuel Consumption, Higher Fuel Efficiency
• Cons of Plug-In Electric Hybrid Vehicles: Still Primarily Fuel-Powered, Relatively High Carbon Emissions

Related: Ecotourism For Dummies: 10 Ways To Get Involved With Sustainable Missions While Traveling Abroad

6 Hybrid Vehicles

The next step up from the plug-in hybrid vehicles are true hybrid vehicles, which still have a gas tank but can go that much further before they use it. Great for trips that cover a slightly larger area but still mostly favor one or two cities, hybrid vehicles can be a realistic compromise for those who need (or just like) the independence of their own vehicle but want to include some eco-friendly measures into their travel.

Hybrid vehicles can also be cost-saving for travelers, as they can get far more miles per gallon of gas than combustion engine vehicles, saving money at the pump during a time when many are looking for budget-friendly family vacations in the US (and around the world).

• Pros Hybrid Vehicles: Lower Carbon Emissions, Cost-Saving, Higher Fuel Efficiency
• Cons Hybrid Vehicles: Large Battery, More Limited Driving Range

5 Electric Vehicles

When it comes to personal vehicles, electric options are widely considered the most eco-friendly option, skipping the gas altogether in favor of a battery-powered vehicle. While electric vehicles may not have the range of traditional fuel-powered vehicles, their in-city flexibility is equal to their counterparts.

Electric vehicles are also an excellent way to practice the “travel slower” approach, one of many things to be aware of when practicing eco-tourism , which means spending more time in one place instead of burning fuel to bounce around a wide range of areas.

• Pros of Electric Vehicles: No Tailpipe Emissions, Cost-Saving
• Cons of Electric Vehicles: Large Battery, More Limited Driving Range

Wondering how to book a hybrid or electric vehicle for that road trip or weekend getaway? Rental companies like Hertz have a Green Travel Collection that offers competitively-priced hybrid vehicles.

4 Carpooling/Ridesharing

As the list transitions from individual to group transportation, the overall sustainability of the transportation increases dramatically. While public transportation may not be available everywhere, rideshare options are becoming increasingly common and are among the greenest transportation methods for those looking to cut their carbon footprint (but this way is not completely free of environmental impact, of course).

Still, ridesharing, or carpooling, helps reduce the carbon footprint of vehicles by reducing the overall number of cars on the road. Simply carpooling within the family can reduce emissions by a metric ton over the course of the year , so those who opt for ridesharing can feel confident that even their single trip is having a positive impact.

Considering that private, personal vehicles are very much a necessity in the world today, these more eco-friendly variations are a practical way to enjoy the best of both worlds.

• Pros of Carpooling/Ridesharing: Reduced Fuel Emissions, Reduces Cars On The Road
• Cons of Carpooling/Ridesharing: Carbon Emission Producing Vehicles Used

Ferries are a unique category of transportation, having a very specific purpose for transporting people across the water. While not as widely utilized as the other modes of transportation on this list, ferries are popular and necessary to reach many unique destinations. In fact, several of the national parks in the US, including Dry Tortugas National Park off the coast of Florida and Isle Royale National Park on Lake Superior, can only be reached by ferry ride.

Like private vehicles and public transit, different Ferries use different forms of fuel, impacting their carbon footprint. Overall, however, their ability to take multiple passengers at once makes them an eco-friendly alternative to driving over bridges in a personal car.

• Pros of Ferry Rides: High Capacity, Low Person/Emission Ratio
• Cons of Ferry Rides: Limited Use, Fuel-Consuming, Water Polluting

2 Public Transit (Buses, Trains, Trams, And Subways)

Easily the most eco-friendly way to travel long distances, public transit, which includes buses, trains, and subways, takes advantage of their high passenger loads to lower their per-person emission levels. Taking Public Transit is also one of many great ways to be eco-friendly on vacation without having to compromise on destination options.

Popular tourist stops like Germany, one of the safest countries to visit in Europe , have a complex network of trains and buses, ensuring that even the most remote corners of the country are accessible without a personal vehicle.

• Pros Public Transit: High Capacity, Low Person/Emission Ratio, Widely Available
• Cons Public Transit: Cannot Access Remote Regions, Limited Time Options

Related: 10 Super Local Train Routes That Are As Scenic As They Get In The Fall

1 Walking And Cycling

Walking and cycling is undoubtedly the most eco-friendly mode of transportation, with no emissions, no batteries, and no carbon footprint. In addition, walking and biking allow access to areas that cars, motorcycles, and buses cannot reach, making it a favorite for those hoping to reach some of the less populated corners of the world.

While walking and cycling are indeed one of the greenest modes of transport, walking and biking both have their limitations, of course. It is neither practical for longer-distance travel nor for agendas that want to hit a large number of stops in a single day. For those hoping to incorporate this eco-friendly mode of transportation into their next vacation, a number of walking tours are available in most cities in lieu of private vehicles or bus tours.

• Pros Walking/Cycling: No Carbon Footprint Created
• Cons Walking/Cycling: Limited Range of Practicality, More Vulnerable Mode of Transportation

Google Maps will let you plan 'mixed modes' of travel for complicated, multipart trips

If your daily journey to work requires that you cycle to the bus stop or train station before taking an Uber to your job, then planning out your route could be getting easier, thanks to Google Maps. 

The navigation app  is launching a tool to help you plan complex, multipart trips in advance. The latest function, called "Mixed Modes," is rolling out under the app's "Transit" tab in the coming weeks, the tech giant announced Tuesday.

The tool will allow you to choose from ride-sharing and cycling options within public transit routes so you don't have to piece together a series of trips yourself. Public systems will be the primary way to get around, and Google will let you choose walking, biking and ride-sharing for "your first and last mile."

Breathe!  Here's why your Apple Watch tells you to inhale all the time

Emergency alerts:  Facebook wants to keep you 'in-the-know' about active shooters 

How does it work?

After you enter the address of your destination in the search box, you’ll get route options that include multiple modes of transportation strung together in one trip.

If you’re taking an Uber or Lyft as part of your journey, "you’ll see helpful information about each leg of your trip: how much your ride will cost, how long the wait is, if there’s traffic on your ride and when your bus or train departs," Google said in a press release.

The latest addition to the app comes after several Google Maps updates over the summer.

Google added a feature in May that makes  ordering food more efficient  and incorporated a tool in June that warns you about how crowded buses and trains are in advance. It also added a  bike-sharing feature in July.

Follow Dalvin Brown on Twitter: @Dalvin_Brown.