Chapter 5. Mobility in cities

This chapter presents long-term scenarios on the development of passenger mobility in cities and the related emissions up to 2050. The results, based on the new model for mobility in cities of the International Transport Forum (ITF), comprise modal shares, mobility levels and emissions of both CO2 and local pollutants. The first section looks at the development of the modelling framework and analyses the impacts of different transport, environment and technology substitution measures on mobility. The long-term implications of the three policy scenarios in terms of accessibility are then analysed, using a new methodology to compute accessibility in cities. The chapter concludes with a case study on certain cities in Asia, applying the same policy scenarios on a subset of cities from China, India and Southeast Asia.

  

By 2050, there will be 2.4 billion additional urban dwellers compared to 2015, when 4 billion persons lived in urban areas. This rapid urbanisation process will create substantial new demand for mobility in cities, making the provision of efficient, sustainable and equitable transport even more of a challenge. The combined effects of rapid urbanisation, income growth and rising private vehicle ownership will result in a surge in emissions, congestion and public health issues. Under a business-as-usual scenario where no additional policies are implemented, CO2 emissions are projected to grow by more than 26% between 2015 and 2050. This creates pressure to pursue energy savings and vehicular emission reductions, especially in developing countries, where 94% of the new urban dwellers will live.

At the same time, the increasing rate of urbanisation and the growing size of cities mean urban transport systems are unable to deliver the benefits they are expected to. Cities face strong pressure to maintain and expand transport systems to ensure good access to opportunities for their population, while keeping negative externalities such as congestion and pollution to a minimum. At a time when private cars and two-wheelers still provide the quickest way to move around in most urban areas, policy makers are facing a difficult choice between short-term economic efficiency and the long-term liveability of their cities.

This chapter presents a global snapshot and the outlook for mobility, accessibility and emissions in cities. It introduces three policy scenarios, describing three different pathways for urban mobility, from a baseline scenario where private cars remain the dominant source of mobility in cities to a scenario where all policies, from land-use and transport planning to fiscal instruments, align to deliver a low-carbon future.

Modelling passenger transport demand in cities

Population, urbanisation and economic development are the key drivers of passenger mobility demand, particularly at an aggregated level and in the long run. Population and urbanisation trends indicate that the additional mobility demand will be concentrated in urban agglomerations of developing economies. According to UN projections (United Nations, 2014), by 2050, the world population will reach 9.55 billion, of which 66%, or 6.34 billion, will be urban. Urban areas will have to accommodate 2.4 billion additional inhabitants and 94% of them will be moving to cities in developing regions.

This Outlook considers all cities above 300 000 inhabitants in 2014, for which individual population projections are available (Figure 5.1). The total population of these cities amounts to 2.2 billion in 2015, making up 31% of the total world population and 57% of the world’s urban population. This figure will reach 3.6 billion by 2050, representing more than 37% of the world population and remaining a stable share of 56% of the urban population. The share of population represented by these cities analysed in this chapter is stable across regions.

While amounting to less than a third of the world population, the total Gross Domestic Product (GDP) of all the cities in this study represents more than 50% of world GDP in 2015 (Figure 5.2). This share grows to 54% in 2030 and 56% in 2050. The GDP concentration in urban areas leads to urban populations reaching higher income levels sooner. For instance, in China, the GDP per capita of Beijing is more than three times the national level in 2015. Income will grow more slowly in cities than rural areas in most developing regions, as cities are starting from a higher base. By 2030, the national GDP per capita for China will be around 94% higher than in 2015, but the growth for Beijing will be around 73%.

Figure 5.1. Total population of cities over 300 000 inhabitants
Million inhabitants
picture

United Nations (2014), World Urbanization Prospects: The 2014 Revision.

 StatLink http://dx.doi.org/10.1787/888933442738

Figure 5.2. GDP per capita in cities and countries by region
2005 International USD
picture

 StatLink http://dx.doi.org/10.1787/888933442743

Income growth generates transport demand and has, in particular, a positive impact on the ownership of passenger cars (see car ownership projections in Chapter 2). Income levels will grow highest in developing countries, especially in Asia. In Chinese and Indian cities, the average GDP per capita doubles between 2015 and 2030 and is projected to reach more than three times its 2015 level in 2050. Most changes in transport demand and mobility patterns are expected in these regions, which are the subject of a specific section at the end of this chapter.

Towards a global model for passenger transport demand in cities

Most urban passenger transport models apply at a local level. To explain travel behaviour for the population of a specific region or urban area, these models (e.g. Kitamura et al., 2000; Mandel et al., 1997) rely on highly disaggregated individual data and methods, which are unavailable at the global level. Passenger travel demand forecast in cities (e.g. Bowman and Ben-Akiva, 2001; Jovicic and Hansen, 2003; Vovsha et al., 2002) currently rely on detailed and interconnected modules requiring both large household travel surveys and ad hoc consumer preference surveys. These detailed models have the advantage of better capturing behavioural aspects but their findings are case specific with low transferability to other areas or regions.

Such methodologies are not replicable for a global level analysis. A commonly adopted approach for national or global level estimates of urban passenger travel demand, energy consumption and emissions would be to use vehicle stock to estimate total emissions by assuming average distance travelled per vehicle and fuel economy levels. This approach has been widely implemented in modelling different transport modes separately, especially in the private car sector (Daly and Ó Gallachóir, 2011; Meyer et al., 2012; Yan and Crookes, 2010). Comprehensive multi-modal analysis on a global scale is scarce. The Mobility Model (MoMo) developed by the IEA is one such model that estimates and projects the travel indicators, energy consumption, pollutant emissions and CO2 emissions for all modes and regions of the world up to 2050 (IEA, 2015).

Being less complex and data-intensive, there are a number of caveats related to the validity of such long-term projections for transport demand. They do not take travel behaviour into account explicitly, as the projections are entirely based on vehicle stocks. A similar modelling approach has been implemented by other researchers for national and global regionalised projections (Cai and Xie, 2007; Meyer et al., 2007; Yan and Crookes, 2010). Some researchers have criticised these studies in that they extrapolate vehicle fleets based on growth rates independently for each transport mode, and are thus unable to account for the competition between transportation modes and potential mode shift (Schafer, 2012; Schafer and Victor, 1999).

Also working on a global scale are studies based on the concept of “travel time budget” and “time-money budget”, following an idea first expressed by Zahavi and Talvitie (1980). These studies (Meyer et al., 2012; Schafer, 1998; Schafer and Victor, 2000, 1999; Singh, 2006) work on the assumption that an average individual’s daily travel time and the share of travel expenditure in the individual’s overall budget are constant. As passenger kilometres travelled increase, due to income growth for example, travellers have to shift towards more flexible and faster transport modes to maintain their travel time budget constant. This can model the impact of pricing policies or of policies attempting to regulate vehicle use. However, it does not explore the use of non-motorised modes and only works at a very aggregate level. While travel expenditure appears to have some stability at the aggregate level, it gives widely different results at different times and locations (Mokhtarian and Chen, 2004).

The ITF model is a new tool developed to evaluate the impacts of transport, environment and technology substitution policies through projections of travel demand, CO2 emissions and accessibility in cities up to 2050. Annex 5.B details the methodology of the model (see also Chen and Kauppila, 2017).

The model differs from existing models in two main areas. Firstly, it has a global perspective, extending the geographical reach of the 2015 edition of the ITF Transport Outlook. It considers each city above 300 000 inhabitants in 2014, and combines data from various sources to form one of the most extensive databases on mobility in cities (Box 5.1). It analyses five transport modes: private cars, public transport, motorcycles, walking and cycling. Secondly, it represents travel behaviour explicitly, modelling the aggregate behaviour for a segment of travellers as a function of the characteristics of the alternative modes and the socio-demographic attributes of the group (Koppelman and Bhat, 2006). The mode share module describes the interactions between the different modes, in a way that existing models, which analyse the evolution of each mode separately, are unable to do.

Box 5.1. City Mobility database

The database used for the modelling contains the 1 692 cities listed in the United Nations (2014) report, World Urbanization Prospects: The 2014 Revision. After merging cities belonging to the same urban agglomeration (for instance Pretoria and Johannesburg), 1 557 entries remain. The urban boundary for each selected city is provided by the Global Built-up Reference Layer (BUREF, 2010; Pesaresi and Carneiro Freire Sergio, 2014), complemented by the space-based land remote sensing data LANDSAT for the year 2010. Other GIS data sources, such as road and public transport supply, come from the intersection of this global urban boundary layer with the open-source OpenStreetMap layer.

The dataset contains the main socio-economic indicator for each city, such as GDP, population or area size. GDP at city level is estimated by redistributing the national GDP volume into the urban areas according to the GDP distribution map obtained from LANDSAT 2010, which provides GDP information for each cell of a grid with one square kilometre resolution. Future GDP in cities come from the application of an S-shaped curve to the growth of the national GDP projections, using the estimated relation between the concentration of population and the concentration of GDP shown by urban agglomerations in each country. When urban agglomerations are small, the elasticity between GDP and population concentration is low, which will then rise as population grows. Finally, when agglomerations become very large, the marginal benefit of increasing the concentration of population begins to decrease.

To enable the analysis of transport demand, demand related indicators are added for a large group of cities: transit fare, parking cost, average vehicle occupancy, mode share, average travel distance, trip rates, and so on. This information is collected through the analysis of multiple data sources, including individual city household surveys where available (see Table 5.A1.1 in Annex 5.A). The resulting dataset is an integrated cross-sectional dataset from multiple sources for the 1 557 urban agglomerations for the year 2010.

Transport policy scenarios

This Outlook assesses the impacts of combinations of policy measures with three alternative scenarios for future urban passenger transport: a baseline scenario, the Robust Governance scenario (ROG), and the Integrated Land Use and Transport Planning scenario (LUT). The measures concern all areas of urban life and include land-use planning, public transport development, economic instruments and governmental regulations. Exogenous drivers, such as urbanisation, population and income growth, do not change between scenarios.

Baseline

In the baseline scenario, no additional measure aiming at influencing travel demand and reducing CO2 emissions is implemented during the 2015-50 period. This scenario constitutes a business-as-usual reference for travel demand and CO2 emissions in the urban transport sector against which to measure the efficiency of additional policies and compare alternative scenarios. It assumes that the future trends of car ownership, road supply, public transport supply, pricing structure and urban area growth will follow the trajectories of the past, as calibrated in each of the sub-models. For instance, public transport provision continues to grow with population and GDP per capita as observed in the historical data. For more details on these relationships, see Annex 5.B.

Advanced vehicle technology and alternative fuels penetrate the market at a relatively low rate, as in the latest 4°C Scenario (4DS) of the Mobility Model developed by the International Energy Agency (IEA). The 4°C Scenario (4DS) takes into account recent pledges by countries to limit emissions and improve energy efficiency, which help limit the long-term temperature increase to 4°C. In many respects the 4DS is already an ambitious scenario, requiring significant changes in policy and technologies. For example, this corresponds to a global average on-road fuel efficiency of passenger cars of 6.4 litres gasoline equivalent per 100 kilometres in 2050 compared to 10.3 litres gasoline equivalent per 100 kilometres in 2015.

ROG Scenario

The Robust Governance (ROG) scenario assumes that local governments play an active role and adopt pricing and regulatory policies to slow down the ownership and use of personal vehicles from 2020 onwards. Existing literature has proved the effectiveness of rigorous pricing strategies. For example, Meyer (1999) studied the effectiveness of various transportation demand management actions, and concludes that the actions which tend to increase the generalised cost of travel for personal vehicle use are most effective. A cross-country study by Greening (2004) indicates that fuel prices and strong government policies related to vehicle and fuel taxes play a significant role in shifting travel demand from private cars towards modes with lower carbon intensity. Studies on the demand for public transport show that playing on transit fares, parking pricing and car ownership is the most efficient way to encourage transit use (Litman, 2004; Paulley et al., 2006).

Following these findings, every city in the world implements pricing policies on fuel prices, fuel taxes, vehicle taxes and fees, parking fees, and transit fares in the ROG scenario. Each policy measure is specified as follows:

  • The public transport pricing sub-model estimates the elasticity of the price of a single transit ticket with respect to GDP per capita for each country group. In this scenario, the price of a ticket grows according to the lowest regional elasticity.

  • In 2030, fuel prices in each country correspond to a fictive oil price of USD 120 (real USD per barrel, 2005). Such prices can result from higher taxation, high oil prices, or a combination of both. The growth rates of oil prices between 2030 and 2050 are assumed to be the same as that in the baseline.

  • Parking prices are 50% higher than in the baseline.

  • Governments regulate car registration cost, purchase cost, operational cost, and so on at the national level, lowering overall car ownership levels, but without car restriction policies such as those implemented in some Chinese cities (see also the section on Asian cities). The elasticity of car ownership with respect to GDP per capita is lower than in the baseline.

  • The size of urban areas and public transport provision (including mass transit) expands with the population and income as in the baseline.

  • Road supply follows a need-based expansion strategy: more new roads are built to serve the new urban area. However, contrary to the baseline scenario, higher GDP growth levels do not trigger the expansion of the road network, which itself could lead to higher car ownership levels.

  • Vehicle load factors, fuel efficiency standards and the market penetration of advanced vehicles and alternative fuels in this scenario reflect the assumptions made in the latest 2°C Scenario (2DS) of the Mobility Model developed by the International Energy Agency (IEA). The 2DS lays out an energy system deployment pathway and an emissions trajectory consistent with at least a 50% chance of limiting the average global temperature increase to 2°C. The world average on-road fuel efficiency of passenger cars becomes 4.4 litres gasoline equivalent per 100 kilometres in 2050, down from 6.4 litres in the baseline.

LUT Scenario

In addition to the policies introduced in the ROG scenario, the Integrated Land Use and Transport Planning (LUT) scenario assumes stronger prioritisation for sustainable urban transport development and a joint land-use policy. As land use and transport planning decisions interact, it is widely acknowledged that better co-ordination and integration are a prerequisite for sustainable development (Geerlings and Stead, 2003). In contrast to the ROG scenario, the LUT scenario anticipates higher supply of public transport, extensive deployment of mass transit and restrictions on urban sprawl in cities.

Better public transport options combined with more compact urban development are expected to directly contribute to increased public transport use and decreased trip distance. Many studies have found that the application of Transit-Oriented Development (TOD) has positive impacts on the sustainability of the cities. Cervero and Arrington (2008) surveyed several TOD cases in five United States metropolitan areas and found substantially lower trip rates by private cars for the dwellers in those areas. Wang et al. (2016) confirmed that concentrating population has the impact of raising the transit mode share, reducing car mode share, and decreasing the average trip distance.

TOD policies apply to neighbourhoods, so it is difficult to define such policies at city level. However, there is evidence that land-use factors, such as density, land use mix, transit access or parking restrictions, have cumulative and synergetic effects on travel behaviour (Litman, 2016; Litman and Burwell, 2006). While it matters which type of density is considered, it has been widely shown that densities are negatively correlated with per capita vehicle travel in cities. Research put forward by Newman and Kenworthy (2011) showed that the relationship between density and car travel is significant in 58 higher-income cities, with moderate increases in density leading to large reductions in vehicle travel. Proximity tends to reduce distances to destinations, and the necessity to use private cars (Banister, 2008). Chattopadhyay and Taylor (2012) found a 10% increase in a city’s residential density, jobs per capita and public transit infrastructure, would lead to a 20% decrease in vehicle miles travelled per household in the urban areas in the United States. Higher densities also make the deployment of large-scale public transport systems more feasible, increase the share of public transport and encourage non-motorised travel (Holz-Rau et al., 2014).

The additional policies in the LUT scenario reflect this body of evidence and are represented in the model as follows:

  • In all regions of the world, the expansion rate of public transport supply with population follows the path found in Europe, which is the highest of all.

  • The thresholds in population density and GDP per capita needed for the development of a mass transit system is 20% lower than in the baseline scenario for every region.

  • Urban area size remains constant from 2020 onwards. While the urban density is constant in the baseline, the 2015-50 growth in urban density in this scenario ranges from 20% to 83% depending on the region. The highest growth happens in Africa, where population growth is also highest.

Table 5.1. Specification of the three policy scenarios for city passenger transport

Variables

Baseline

Robust governance

Integrated land-use and transport planning

GDP

BAU

BAU

BAU

Population

BAU

BAU

BAU

Urbanisation

BAU

BAU

BAU

Car ownership

BAU

Low growth

Low growth

Road supply

BAU

Need-based expansion

Need-based expansion

PT stop supply

BAU

BAU

EU expansion pattern

Mass transit

BAU

BAU

Low thresholds of GDP per capita and population density

Fuel price

Current oil price + IEA-Momo-4DS

Current oil price + High taxation

Current oil price + High taxation

Parking price

BAU

50% higher in all countries

50% higher in all countries

PT ticket price

BAU

Low price elasticity to GDP per capita

Low price elasticity to GDP per capita

Urban sprawl

BAU

BAU

Constant urban area

Load factors

IEA-MoMo-4DS

IEA- MoMo-2DS

IEA-MoMo-2DS

Energy intensity

IEA- MoMo-4DS

IEA-MoMo-2DS

IEA-MoMo-2DS

Carbon intensity

IEA- MoMo-4DS

IEA-MoMo-2DS

IEA-MoMo-2DS

Local pollutant standards

ICCT – baseline

ICCT – baseline

ICCT – baseline

Passenger mobility in cities up to 2050

The alternative scenarios represent different ways to fulfil the mobility demand resulting from increasing population and income. The total passenger-kilometres are of the same order of magnitudes in all three scenarios, the common figure reflecting the need for transport of the population in cities. However, mode shares differ significantly between a baseline scenario where private mobility, especially by car, continues to increase and the LUT scenario where public transport becomes the dominant form of transportation in many regions. The following paragraphs discuss these elements in detail; the following section will examine the consequences of the scenarios in terms of emissions.

Mode shares

In the baseline scenario, the share of private vehicles continues to increase in all developing regions, but slightly decreases in developed economies. The share of public transport, motorcycles and non-motorised modes in developing countries are all projected to decrease until 2050, meaning that more residents of cities in developing regions will be shifting to private car use. At the end of the period, the private car is the dominant form of urban transport in all the regions of the world.

Figure 5.3. Car share in cities by region
As a percentage of all trips, Baseline, Robust Governance (ROG) and Integrated Land-Use and Transport Planning (LUT) scenarios
picture

 StatLink http://dx.doi.org/10.1787/888933442752

Figure 5.4. Mobility by mode of transport, Asia and North America
Billion passenger-kilometres
picture

 StatLink http://dx.doi.org/10.1787/888933442760

Table 5.2. Share of car and public transport by region
As a percentage of all trips

Private car

Public transport

2015

2050

2015

2050

Regions

Baseline

ROG

LUT

Baseline

ROG

LUT

Africa

20.2

27.4

 9.8

 7.4

27.1

25.1

64.3

71.0

Asia

28.3

40.3

19.2

16.2

23.8

20.6

56.3

61.7

EEA + Turkey

56.4

44.4

19.7

18.4

19.4

24.9

49.9

52.7

Latin America

40.5

42.4

24.6

21.7

22.1

21.3

47.4

52.0

Middle East

54.6

56.3

38.8

35.5

15.9

15.9

37.8

42.4

North America

81.2

76.1

61.1

60.5

 7.0

 9.5

20.9

21.6

OECD Pacific

59.6

48.9

24.1

23.4

16.8

22.0

46.2

47.8

Transition

54.4

57.9

26.3

22.9

23.5

21.7

57.5

62.5

The highest growth in car share occurs in Asia where the average share of cars reaches 40%, or 1.5 times the 2015 level. China and India reach the highest value in the region. Following closely are other developing Asian countries, where the average car share grows from 30% in 2015 to 41% in 2050. In developed regions, we observe a natural reduction in car share. For instance, in Europe, the share of private cars decreases by 12 percentage points from 2015 to 2050. In North America, the reduction is 5 percentage points. Behavioural aspects win over purely exogenous growth factors. Former car users shift to public transport or non-motorised modes while the growth for transport demand remains low. This is in line with recent research showing that per capita daily travel demand has uncoupled from income in some high income countries, for instance the United Kingdom (Metz, 2012, 2010). In many large European cities with extensive public transport network, the pressure put on car usage has brought down the mode share of private cars (TfL, 2010, OMNIL, 2012).

In the ROG scenario, car use is lower in all regions because of the lower expansion rate of the road network and more stringent pricing policies increasing the fixed or variable costs of car ownership and car use. Reduction in the share of cars already happens in 2030, except in China and India, where the mode share of cars continues to increase due to rapid income growth. By 2050, public transport becomes the dominant urban transport mode in every region except North America, whose car share is still around 61%.

The share of motorcycles is on a downward trend in the ROG scenario in all developing regions, with travellers shifting to public transport. This is due to two reasons. The first is the result of per capita GDP growth, as high income travellers prefer safer and more convenient modes (Wen et al., 2012). The second reason is the stringent pricing policies in the ROG scenario, which make motorcycle and car use more expensive and public transport more affordable and attractive.

Another element explaining the growth in transit use is the shift of non-motorised trips to public transport, especially in developing regions, with two main underlying reasons. First, there is strong demand for faster mobility due to income growth. In 2015, the average travelled distance in OECD countries is three times the amount of non-OECD economies; it is only twice that amount in 2050. Second, the expansion of cities increases the length of trips for the urban residents, making walking and cycling less viable and encouraging a shift to motorised modes. Walking and cycling become an effective complement to public transport, meeting the requirements of short distance travel.

In the LUT scenario, the additional policies of land-use planning and TOD reinforce the use of public transport and further reduce car use, due to controlled urban sprawl, higher population density, higher transit network coverage and mass transit availability. Developing regions are more sensitive to the LUT policies than developed regions because cities in these regions are less mature. By 2050, the car share of developing Asian countries is 3 percentage point lower than in the ROG scenario. The figure is only 1.2 in North America and 1.3 in the European Economic Area (EEA) and Turkey region.

Table 5.3. Total mobility by world region
Billion passenger-kilometres

2015

2030

2050

Region

Baseline

ROG vs Baseline

LUT vs Baseline

Baseline

ROG vs Baseline

LUT vs Baseline

Africa

989.8

2 016.5

12.1%

0.0%

3 788.1

37.7%

6.5%

Asia

6 476.6

10 785.2

8.9%

3.8%

15 281.5

25.6%

13.9%

EEA + Turkey

1 699.5

2 047.4

5.4%

3.7%

2 484.1

13.2%

8.3%

Latin America

1 875.0

2 513.8

4.3%

-0.1%

3 397.8

17.5%

7.8%

Middle East

431.3

712.4

2.4%

-4.8%

1106.5

9.6%

-7.3%

North America

3 039.3

3 704.7

-1.9%

-1.8%

4 701.9

-1.3%

-1.0%

OECD Pacific

2 975.1

3 152.9

0.3%

0.7%

3 375.8

10.5%

11.8%

Transition

504.6

567.3

5.2%

4.8%

764.7

16.5%

7.4%

World

17 991.3

25 500.1

5.5%

1.7%

34 900.4

19.4%

9.1%

Transport demand

Under the baseline scenario, total motorised mobility (measured in passenger-kilometres) in cities rise by 42% in 2030 and 94% in 2050 compared with 2015, reaching 25 500 billion and 34 900 billion passenger-kilometres, respectively.

The policy measures of the ROG scenario do not impact the total mobility levels significantly. For developing countries, where mobility demand is growing fast, favouring transit use by improving public transport quantity and quality and reducing its cost significantly improves mobility levels. Public transport being more affordable, motorised mobility becomes available to a larger group of people. This is why total mobility levels in these regions are higher in the ROG scenario than in the baseline, where the reliance on cars limits the uptake of motorised mobility. On the contrary, in already highly motorised developed regions, public transport needs to compensate for the increased cost of car use (see Figure 5.4 for an illustration). If restrictions in car use are put in place without significant improvements to the public transport system, overall mobility levels will decrease.

In the LUT scenario, mobility demand is fulfilled through less carbon-intensive mobility options and reduced distance travelled. The overall passenger distance travelled figures are smaller under the LUT scenario than in the ROG scenario because the effective control on urban area size, which leads to lower urban sprawl, higher population density, and more transit-oriented development (TOD) patterns, contributes to a reduction in trip distances.

Emissions from mobility in cities up to 2050

Emissions from transport in cities have received a lot of attention because of the large impact local pollutants can have on health. The quality of the outdoor air is a more immediate concern to the inhabitants of cities than CO2 emissions and has become the subject of much debate and policies. The policies range from direct and indirect restrictions in car usage to efficiency standards for new cars. However, the climate change impact of urban transport cannot be neglected. The total CO2 emissions from all urban agglomerations in this study are 1 639 million tonnes (Mt) in the base year, amounting to slightly more than half of global surface passenger transport CO2 emissions.

CO2 emissions

In the baseline scenario, the level of total CO2 emissions in large cities is 26% (419 Mt) higher in 2050 compared to 2015. Global emissions do not change between 2015 and 2030 due to the large fuel efficiency gains expected during the decade to come, and the low economic growth for the years up to 2020 (see also Chapter 1 for details on the short-term macroeconomic projections). However, emissions grow again from 2030 onwards based on the assumptions related to vehicle technology and fuel efficiency in the 4DS scenario of the IEA’s Mobility Model. The world average fuel efficiency for on-road passenger light duty vehicles improves by 29% from 2015 to 2030 but only 14% from 2030 to 2050. This pace of technology improvement is not enough to offset the growing mobility demand between 2030 and 2050.

Policy interventions, especially rigorous car pricing policies, lower transit fares and higher vehicle technology improvements introduced in the ROG scenario could intensely mitigate CO2 emissions from the urban passenger transportation sector. With solely the policy measures from ROG, the avoided CO2 emission could reach 397 Mt in 2030 and 886 Mt in 2050 compared with the baseline. The additional policies introduced in LUT scenario would further reduce the CO2 emissions by 48 Mt in 2030 and 104 Mt in 2050. Under the most effective policy scenario LUT, the global CO2 emissions level from the urban transport sector would be 26% lower in 2030 and 35% lower in 2050 compared with 2015 levels.

Private cars are the main contributor of CO2 emissions in cities, representing around 82% of all emissions in 2015 and around 75% in 2030 and 2050. With the implementation of the policy measures of the ROG and LUT scenario, the contribution of cars decreases to 40% in 2050.

Bus and motorcycle emissions represent 11% and 7% respectively of all emissions in cities in the base year, going up to 15% and 10% in 2030 and remaining stable until 2050, in the baseline scenario. In the ROG and LUT scenarios, the contribution of buses in total emissions increase both because of the additional public transport supply in these scenarios and because of the lower level of emissions from cars. Buses emit as much as cars in the ROG scenario in 2050; in the LUT scenario, buses even become the main contributor of CO2 emissions in 2050. CO2 emissions from urban rail are null in this model, as only tank-to-wheel CO2 emissions are taken into account and urban rail is assumed to be fully electrical. A life-cycle analysis would increase the CO2 emissions from urban rail, especially in India and Africa where the IEA projects that electricity production will remain carbon intensive through 2050.

Figure 5.5. CO2 emissions in cities by mode of transport
Million tonnes
picture

 StatLink http://dx.doi.org/10.1787/888933442770

The technology aspect of the ROG and LUT scenarios contribute most significantly to the CO2 mitigation potential of these two scenarios. Figure 5.6 presents the avoided CO2 emissions for the two scenarios, broken down by type of measure in 2030 and 2050. Technology improvements alone reduce global CO2 emissions in cities by 15% in 2030 and 22% in 2050, compared to 2015.

Figure 5.6. Mitigation potential by type of measure
Million tonnes of CO2 avoided in cities
picture

 StatLink http://dx.doi.org/10.1787/888933442780

Behavioural changes of the scale discussed in this Outlook have an impact on emissions (e.g. 11% less CO2 emissions in the LUT scenario in 2050 compared to the base year) and they are essential in combatting congestion or health issues. However, emphasising behavioural changes in the fight against climate change does not take into account the surge in mobility which will result from the economic development of lower income countries. Even in the LUT scenario, despite the strict policies in place, car mobility increases in almost all developing regions. A complete decarbonisation of the transport sector in cities would require extreme changes in mobility patterns, on a scale out of proportion with the efforts currently deployed around the world. Such changes could take the form of much higher taxation of car mobility in cities or higher penetration of alternative fuels. The necessary penetration of electric vehicles in urban fleets in 2050 to bring emissions down to half their 2015 levels would be 65%, on top of all other policies already present in the ROG scenario, which is only likely to happen if all policies are aligned in favour of electric cars (see also Box 5.2).

Box 5.2. IEA electric vehicle outlook

The global deployment of electric vehicles of all types is an integral part of the necessary actions to meet sustainability targets, alongside the optimisation of urban structures to reduce trip distances and a modal shift towards public transport. The Electric Vehicle Initiative 20 by 20 target calls for an electric car fleet of 20 million by 2020 globally. The Paris Declaration on Electro-Mobility and Climate Change and Call to Action sets a global deployment target of 100 million electric cars and 400 million electric 2- and 3-wheelers in 2030.

Current sales are still modest. In 2015, the global stock of electric cars went up to 1.3 million, a near-doubling over the stock of 2014 (IEA, 2016a). Although the share of electric cars in the global vehicle stock is still only 0.1%, this is a marked improvement from historic levels. The increase in sales has also been accompanied by a growth in electric vehicle supply equipment. The recent rise of electric cars has emerged both as a result of continuous technological improvements and because of mounting policy support.

According to the New Policies Scenario, a wider availability of electric car models and charging infrastructure will continue to drive electric car deployment: the stock rises by 50% per year to about 9 million by 2020 and 30 million by 2025; by 2040, the global stock of electric cars reaches more than 150 million, around two-thirds of which are plug-in hybrids (Figure 5.7). Yet, this growth of the global market for electric cars only has a minor impact on fuel consumption. Indeed, despite a significant decline in battery costs, electric cars might not become easily competitive with conventional cars, due to still higher payback times. The payback is quicker for commercial cars with high annual mileages, such as taxis, company fleets or car-shared vehicles. These vehicles having higher mileage, the CO2 savings are also important for these types of vehicles, which should be prioritised by policies.

Figure 5.7. Stock of electric cars by region, 2015 to 2040
IEA New Policy Scenario, million vehicles
picture

IEA (2016), World Energy Outlook.

 StatLink http://dx.doi.org/10.1787/888933442790

Despite the high growth in emissions expected in developing economies, their average CO2 emissions per capita is still only one-third that of OECD countries in 2050. In the baseline scenario, city inhabitants of OECD countries emit on average 1.2 tonne of CO2 in 2050 for their transport activities, against 0.4 tonne for inhabitants of non-OECD economies. The average CO2 intensity (emissions per kilometre travelled) is also lower in non-OECD countries, both in 2015 and in 2050, because of the more common use of non-motorised modes, and in particular walking.

Table 5.4 presents the regional breakdown of total urban CO2 emissions. In the baseline scenario, the most dramatic increase in CO2 emissions occurs in Africa, where emissions in 2050 is projected to be almost three times their 2015 levels. However, the highest growth in absolute value takes place in China and India. The combined emissions of these two countries grow by 297 Mt. CO2 emissions in all regions significantly decrease in the ROG scenario. Regions with the greatest CO2 mitigation potential are North America, because of the widespread use of private cars on the continent, and China and India, due to the high motorisation potential of these two countries.

Table 5.4. Total CO2 emissions in cities by region
Million tonnes

2015

2030

2050

Baseline

ROG

LUT

LUT vs ROG

Baseline

ROG

LUT

LUT vs ROG

Africa

41.9

75.3

60.3

52.8

-12.4%

155.7

76.5

59.9

-21.7%

Asia

323.4

510.1

401.7

376.9

-6.2%

760.3

437.2

385.8

-11.8%

EEA + Turkey

163.8

134.5

109.3

107.0

-2.1%

132.8

96.1

90.9

-5.3%

Latin America

133.5

161.3

133.7

126.9

-5.1%

216.4

141.5

126.6

-10.6%

Middle East

45.2

70.5

59.2

54.5

-7.9%

118.7

72.2

60.2

-16.7%

North America

592.6

469.8

344.1

343.5

-0.2%

457.1

238.1

237.7

-0.2%

OECD Pacific

303.8

202.2

129.2

128.9

-0.2%

168.2

88.5

88.2

-0.4%

Transition

34.9

35.3

24.1

23.5

-2.7%

49.5

22.5

19.7

-12.2%

World

1639.1

1659.0

1261.6

1214.0

-3.8%

2058.5

1172.5

1069.0

-8.8%

The additional urban policies in the LUT scenario have high impacts on the developing regions (e.g. an additional 22% in CO2 reduction for Africa in 2050). In developed economies, the additional effect of the LUT policies is negligible. In Europe, Japan and Korea, the reason lies in the already (comparatively) high level of public transport infrastructure. In North America, Australia or New Zealand, the low CO2 mitigation potential results from the low elasticity of mode choice to changes in pricing policies or public transport supply.

Local pollutants

In addition to its climate change impacts, urban transport is an important contributor to local air pollution, principally through the emission of NOx, SO4 and particulate matters (PM), which can contribute to severe health problems, including cardiovascular and respiratory diseases and numerous cancers. These problems are widespread: the World Health Organisation estimates that more than 90% of the world population lives in area where air pollution is above the limits for healthy living (WHO, 2016).

The effects of urban activity on CO2 and local air pollutants are not always correlated. While emissions of CO2 are strictly proportional to fuel consumption, the quantity of local pollutants per litre of fuel in exhaust fumes can vary greatly. Regulation has historically focused on limits on tailpipe emissions, because it was assumed that consumer pressure would result in fuel efficiency gains, and thus in less CO2 emissions. While there is a controversy regarding differences in the level of emissions of local pollutants between test and on-road conditions (Franco et al., 2014), the strengthening of emission standards means that, in the European Union, new passenger cars in 2014 emit 100 times less PM than new cars in 1996. It is estimated that the most advanced emission controls could effectively eliminate over 99 percent of local air pollutants from engines (Chambliss et al., 2013).

To estimate the emissions of local pollutants resulting from the urban mobility levels of the three scenarios, this Outlook uses emission factors from the Roadmap model of the International Council on Clean Transportation (ICCT, 2014). The Roadmap includes expected improvements in vehicle efficiency standards, and their probable penetration in vehicle fleets until 2030.

Figure 5.8. NOx, SO4 and PM2.5 emissions by region
Thousand tonnes
picture

 StatLink http://dx.doi.org/10.1787/888933442803

Figure 5.8 shows the evolution of the emissions of local pollutants between 2015 and 2030 by region. In the baseline scenario, emissions of SO4 and NOx moderately increase, and those of PM2.5 even decrease. Car mobility, which increases most in this scenario, becomes much cleaner. In the more public transport-oriented scenarios, all local pollutants grow more because of the development of bus travel necessary to replace car mobility (Figure 5.9). Diesel buses, which form most of the bus fleet in many countries, have higher emission factors than cars and will not become much cleaner, especially in developing countries. For instance, diesel buses in China emit on average 7.7 g of NOx per kilometre in 2015 and 5.7 g in 2030 when emissions from gasoline cars go down from 0.1 g per kilometre in 2015 to 0.03 g in 2030.

Figure 5.9. Vehicle activity by mode
Billion vehicle-kilometres
picture

 StatLink http://dx.doi.org/10.1787/888933442815

It is difficult to predict the health impacts of these scenarios, as transport is only one contributor of local pollutants. Several other factors, such as the topography and climate of cities, as well as the presence of industry, also enter the equation. In regions with heavily polluting industries, such as coal plants, its share can be as low as one-third (Beijing Municipal Environmental Protection Bureau, 2014). However, our projections show that most of the increase will take place in developing countries, where many cities are already choking on local pollution. Any additional emission is likely to bring significant health issues. Extra effort will be required to develop clean public transportation systems, especially in medium cities where rail investment is not an option, for instance through the phasing in of buses powered by alternative fuels or through the development of coherent planning policies. In the LUT scenario, the decrease in the average distance travel causes some reduction in the emissions of local pollutants.

Accessibility

Policy packages that are well co-ordinated and address externalities from increasing motorised mobility in cities make a significant difference for sustainable urban futures. The policy scenarios introduced earlier in this chapter (Robust Governance and Land-Use Transport Planning) do not merely influence mobility patterns in cities. They also affect the way in which people, jobs and other urban functions can be accessed in cities.

Accessibility, however, is only to some extent about mobility. Designing urban transport policies has traditionally focused on travel time savings and congestion relief. According to this vision, transport planning aims at maximising the distance that infrastructure users can travel within their time and money budgets. However, there is growing consensus that this is not recognising the actual purpose of transport, which is providing accessibility to opportunities such as employment, goods or services. Taking this shift of emphasis seriously requires rethinking the governance and finance models for connecting transport and land use policy and planning.

Improving accessibility, not merely mobility, will be decisive for sustainable and inclusive cities. In a meta-study reviewing the effects of the built environment on travel behaviour, Ewing and Cervero (2010) highlight the importance of accessibility to valued destinations. Improving the level of accessibility in cities is an important dimension of social inclusion (Viegas and Martinez, 2016).

Measuring accessibility in cities

Ways of measuring accessibility abound and remain subject to wide discussion (Bhat et al., 2005; Geurs and Wee, 2004; Handy and Niemeier, 1997; Murray et al., 1998). There is a broad consensus about understanding accessibility as the ease or potential for reaching valued locations, “where services, goods or opportunities are available” (Paez, 2016; see also Hansen, 1959; Owen and Levinson, 2014). However, research differs according to the dimensions of accessibility emphasised, such as transport, land-use factors, time constraints or individual characteristics (Geurs and Wee, 2004). Similarly, different perspectives of measurement persist, varying from relatively simple infrastructure-based and proximity measures (i.e. the walking distance to a transport stop) to more complex ones, taking into account individual utility functions (ibid.). Location-based metrics, focusing on the accessible mass of opportunities at different locations, strike a good balance between theoretical soundness, data and computing requirements on the one hand, and policy relevance on the other (for a comprehensive discussion of accessibility measures see Geurs and Wee, 2004).

When it comes to comparative studies of accessibility in cities beyond the case study format, the objectives of accessibility measures are usually more modest and often based on the concept of proximity. This has been largely due to limited availability of data and computing power. As a recent example of applied research, the European Commission measured accessibility to public transport in European cities (Poelman and Dijkstra, 2015). The study calculates the share of the population living within walking distance of public transport facilities and assesses the level of service frequency at stops and stations. This metric allows for comparison of the population share covered by the public transport network and the quality of service, as measured through frequency. The advantage of such proximity-based metrics is the relative ease of computation, moderate data requirements and the clarity for policy messaging. However, the policy implications might be limited as the distribution of actually valued destinations, and hence the constraints to reach them are not sufficiently taken into account (Peralta, 2015).

The arrival of new standardised sources and tools for computation and measurement make it possible to go beyond accessibility metrics that are limited to single case studies and proximity-based measures. Innovative research by the Accessibility Observatory located at the University of Minnesota estimated in a series of reports the potential accessibility to jobs in more than 40 US metropolitan areas by car, public transport and walking for 1990, 2000 and 2010 (Owen and Levinson, 2015; 2014; Levinson, 2013). Combining disaggregated census population data, job locations and detailed information of the urban transport network, timetables and travel speeds, this research estimates the number of jobs an average city dweller can reach in ten minute time bands up to one hour. Valuing job locations with shorter travel times higher than with longer ones, this location-based metric adds an element of gravity.

In a similar vein, the World Bank calculates average accessibility to jobs by different modes of transport for a number of cities in Latin America, Africa and Asia. The metric computes the average accessibility to jobs by mode within an assumed maximum commuting threshold of one hour. Working on a case-by-case basis this approach uses detailed and locally specific data, taking into account travel and land use patterns allowing the assessment of different accessibility scenarios, comparison between modes and different points in time (Peralta and Mehndiratta, 2015). The drawback of this approach is that it remains case-study based and dependent on locally available data, notably job locations, which are difficult to obtain or unavailable on a broader scale.

The Outlook accessibility index

This Outlook analyses accessibility in cities for both road and public transport. We define a common accessibility indicator for both modes, following these two principles:

  • Conceptual simplicity. While the state of the art in accessibility research has been moving towards more complex models of accessibility (Geurs et al., 2012), a conceptually simple approach allows overcoming the challenge of extensive data and computing requirements. Contour-based metrics calculating the number of opportunities within a time threshold are best suited to bridge data availability, theoretical complexity and comparison between cities and regions.

  • Global data availability, or of global reach. As the derived value of global metrics is to compare, situate and benchmark cities across the world, data need to be standardised across countries and widely available. As a consequence, only globally standardised datasets are used for this (see Annex 5.A on data sources).

The accessibility index in this Outlook is defined as the average number of inhabitants that can be reached within a 30 minute threshold by private car or public transport. The spatial distribution of the population in cities is used as a proxy for opportunities. While population does not represent actual opportunities, there is some empirical evidence that population density correlates with opportunities such as jobs and services. For instance, Kaufman et al. (2016) showed that the distribution of services, offices and commercial spaces within cities is highly correlated to the one of population. Such a proxy is very useful in the case of a global study; detailed analysis of specific cities should however rely on the actual location of opportunities.

Additionally, to analyse mobility in cities by means of public transport, we also consider the notion of public transport coverage, corresponding to the share of the population which has access to public transport by walk and represents the number of potential users of the public transport network. This metric is similar to the People near Transit (PnT) metrics developed by the Institute for Transportation and Development Policy (ITDP, 2016).

Accessibility by car is calculated for 1 390 cities among the 1 557 urban agglomerations used in the previous section of this chapter; public transport coverage for 1 014 cities. These two indicators are built using OpenStreetMap data. Some cities are excluded because of data quality (see Figure 5.10 for the cities sufficiently covered by OpenStreetMap). Accessibility by public transport is computed for a sample of 23 cities, with GTFS data. The General Transit Feed Specification (GTFS) format is increasingly recognised as a global standard for public transport schedules, but only a small sample of cities was used due to incomplete data coverage.

Figure 5.10. Coverage of cities by OpenStreetMap (OSM)
picture

Accessibility in cities today

Road accessibility

In large and dense cities like Beijing, inhabitants can, on average, reach 3 million inhabitants in a 30 minute car ride, or 13% of the total population. In a sprawled city like Buenos Aires, this number goes down to 0.8 million inhabitants, or 5% of the total population. Figure 5.11 shows the average road accessibility by world region. Clear geographical patterns, correlated with density patterns, appear: Asian and Middle East cities offer the highest level of accessibility, regardless the size of the urban agglomeration considered. Northern American cities as well as cities from the transition economies are characterised by low accessibility albeit for different reasons. Northern American cities are sprawled but provide an efficient and relatively uncongested road network. Although the density observed in transition economies is equivalent to European cities, they lack high capacity trunk roads and thus suffer from low speeds and high congestion.

Figure 5.11. Road accessibility in cities by region and city size
picture

Overall, the differences in road accessibility are explained by three factors: population density, free-flow speed and congestion. Population density impacts the distance people need to travel to reach a given number of inhabitants. Free-flow (uncongested) speeds vary from city to city depending on the provision of fast and high-capacity road infrastructure. The level of congestion, defined as the relative travel time loss at peak hour, is the last major driver of accessibility. Together these three factors explain 75% of the difference between the cities. The elasticity of road accessibility with respect to density is 0.7, according to which increasing population density by 10% would lead to a 7% increase in road accessibility.

These three factors provide a categorisation of the policies to enhance accessibility. Improving the road network to increase free-flow speeds and alleviating road congestion are the first two levers. Historically it has been the preferred option as accessibility is more sensitive to speed and congestion than density. The elasticity with respect to actual speed is indeed higher, with a value at 1.6.

Policy instruments for increasing density, such as land-use planning and zoning, tend to be more efficient. Accessibility is globally higher in dense cities, even those with low-quality road infrastructure, because the variability of density between cities is much higher than that of speed. The ratio between Northern American cities and African cities ranges between 1:4 and 1:8 for density, depending on city size, when it is only 2:1 for speed.

Increasing road provision has little effect on free flow speeds and congestion. According to our model of congestion (Box 5.5), increasing the density of trunk roads (in km of roads by km2) by 10% induces a 1% increase in free flow speeds. The relationship between congestion and road supply is more complex. A large supply of trunk roads implies high capacities and thus low congestion. However, vehicle-kilometres increase at a similar pace than road capacity: higher speeds lead to a higher demand for road transport which, in turn, calls for more infrastructure building. Statistical analysis (Box 5.5) shows that road congestion first decreases sharply with GDP per capita but quickly stabilises at around 40%. This result extends the well-known rule that, in congested cities, providing more capacity usually does not resolve the problem of congestion.

Box 5.5. A global model of congestion in cities

The TomTom congestion index (TomTom, 2016) estimates the time lost by drivers of a city during the morning peak hour: a congestion index of 50% indicates that, on average, the morning congestion increases driving times by 50% on average. To analyse the influence of some characteristics of cities on congestion levels, the following model is built:

picture

VKM is the total number of vehicle-kilometres in the city, Trunk the length of fast and high capacity roads, and Capacity the average unit capacity (in VKM/hour). The quantities VKM and Trunk are known from the urban passenger model. Road capacity is difficult to assess globally as it varies according to the number of lanes, road geometries, intersection design, speed limitations, etc. It is assumed that Capacity can be written as picture.

The results of the estimation are in Table 5.5. Congestion first sharply decreases with GDP before stabilising at an index between 40% and 50%. The stabilising reflects the inter-relation between road capacity and vehicle use. From a certain level of GDP onwards, vehicle-kilometres and the total length of trunk roads grow at the same pace, leaving the ratio between the two, congestion, unchanged. This result recalls, on a global level, the classical work by Downs (1962, 2004) sometimes known as “fundamental law of peak hour congestion”. It states that on urban commuter expressways, road will be as congested as before after any new investment in road capacity.

Table 5.5. Results of the estimation of the congestion model

Variable

Elasticity

log(VKM/Trunk)

0.12

log(GDP per capita

-0.22

Figure 5.12. Congestion in cities as a function of GDP per capita
picture

 StatLink http://dx.doi.org/10.1787/888933442821

Public transport coverage

Before looking at the accessibility index for public transport, we first focus on public transport coverage. Indeed, accessibility in cities depends on the number of people who live within walking distance from transit stops. Public transport coverage, or People near Transit (PnT), measures the number of residents in a city who live within a walking distance from public transport stops. PnT can also be considered as a proxy for the integration of transport and land-use in cities (ITDP, 2016).

Based on the available data on public transport stops, public transport coverage is calculated for 1 014 cities. The proximity to stops is evaluated for bus and for mass transit, with a maximum walking distance of one kilometre, which is equivalent to approximately 12-15 minute walk. Public transport coverage is calculated as the percentage of residents of a city living within that distance from at least one public transport stop. Figure 5.13 illustrates this indicator on a set of cities.

Figure 5.13. Public transport coverage in selected cities
picture

When coverage is computed, the values are likely to be low estimates of the public transport coverage, because of possible OSM data incompleteness. However, it is likely that the cities for which no data was available are also the cities with the lowest public transport supply, thus skewing the results in the other direction.

On average, around 53% of the residents of the considered cities live in proximity to a public transport stop, with 28% of all the residents covered by mass transit. Cities from Europe and transition economies have the best coverage, with an average of 85% and 80% of the population correspondingly and relatively high shares of mass transit: 51% for Europe and 47% for transition countries (Figure 5.14). OECD Pacific, North America and Asia have lower public transport coverage rates but fairly large shares of mass transit. The rest of the world regions tend to have lower coverage with bus prevailing over mass transit modes.

Figure 5.14. Public transport coverage in cities by region
picture

Vast coverage by public transport is a prerequisite for good accessibility. However, public transport coverage does not provide sufficient information for policy implications. Even if residents live in proximity to public transport stops, variations of the quality of service of the available public transport modes may affect the mobility patterns significantly. Moreover, access to a public transport stop does not imply easy access to the desired opportunities. The next section describes the accessibility index, which solves some of these issues.

Public transport accessibility for a sample of 23 cities

Computing the accessibility index for public transport was only possible for a limited set of city due to limited data availability on public transport services. Figure 5.15 illustrates the computed indicator for the 23 cities in our dataset.

Accessibility by public transport varies greatly by city, with European cities generally offering higher accessibility than developing cities. Around 12% of the 10 million inhabitants of Paris urban area can be reached in 30 minutes by public transport against less than 4% of Cairo’s 17 million inhabitants. Although Cairo is nearly twice as dense as Paris, its public transport system is only half as fast, with less extensive coverage, leading to lower accessibility. Here, the speed of the public transport network needs to be understood as the average speed over all trips, assuming that the geographical distribution of trip ends is similar to that of density; it is not the average commercial speed of the lines composing the network.

Figure 5.15. Accessibility by public transport in 23 cities
30 and 60-minute isochrones
picture

Compared to public transport coverage, the accessibility index provides more meaningful insights on public transport efficiency. Figure 5.16 depicts that no relationship exists between the two indicators. In well-covered cities, accessibility varies from a few per cent to 30%. Conversely, in cities with low accessibility, coverage varies from 100% to 30%. Low frequencies and station density, as well as inadequate networks, can result in very low average speed even when the coverage is very good.

As in the case of roads, public transport accessibility is driven by population density and speed. Those two variables explain 80% of the variability between cities. Accessibility is much more sensitive to speed than density: the elasticities are 2.9 and 1, respectively. Because public transport tends to connect dense areas, a small increase in speed has a great effect on the number of inhabitants people can reach in 30 minutes. This highlights the interest of mass transit as a lever for enhancing accessibility, especially when coordinated with land use.

The accessibility gap in favour of European cities relates to an unequal distribution of public transport speeds. The average speed for cities ranges from 5 to 15 km. Most of the cities with no or little mass transit achieve a speed between 5 and 8 km/hour, one notable exception being the transport network of Nairobi which, with nearly no rail transport, offers a speed of 8.8 km/h thanks to a large and rather efficient informal transport system. The so-called Matatus are mini-coaches operating on a network of 2 000 stops with a peak frequency of nearly 30 buses per hour. The average speed can go up to 15 km/h for cities with significant mass transit provision.

Reducing the accessibility gap between cities requires significant investment in infrastructure and improved services. The high variation in average speed has to do with the unequal provision of public transport services. Yet, quantifying the quality of a transport network is challenging. Common indicators, such as station density or the total length of public transport lines, focus only on the spatial extent of the network and are more easily related to public transport coverage. They fail to grab much of the differences in accessibility between cities of similar coverage (Figure 5.16).

Figure 5.16. Public transport accessibility in cities
Public transport coverage and average accessibility by public transport
picture

 StatLink http://dx.doi.org/10.1787/888933442836

Table 5.C2 in Annex 5.C presents two simple measures that overcome this difficulty for the 23 cities of our sample. These are the number of buses and mass transit stations, multiplied by the average frequency at peak hour, measuring the total number of vehicle calls at stations per hour. These two indicators explain more than 55% of the speed differences between public transport networks and thus appear to be good measures of network quality. In particular, it appears that increasing mass transit provision of 1% increases the average speed by 0.14%.

Policy implications

Accessibility, as measured in this Outlook, is higher for cars than for public transport, except for some large western metropolises, because of the flexibility offered by this mode. Contrary to public transport, which does not cover all inhabitants of a city, cars can, in theory go between any two places. However, promoting accessibility by car raises many issues, not least because of congestion (see also the Outlook section below). The road infrastructure required to accommodate a car-oriented accessibility can be very difficult to build and maintain, especially in dense cities.

Public transport, on the other hand, can deliver accessibility to the greatest number. Investment in transport infrastructure needs to be coupled with wider policy packages, including travel demand management, land-use planning and promotion of active modes. Innovative mobility solutions, such as car-sharing or demand-responsive buses, can also be promoted to form part of the transport options for travellers. Indeed, a recent study by ITF (2016) shows that these two mobility options, if implemented together with adequate mass transit, can provide high levels of accessibility at a reasonable cost and bring additional benefits, such as reduced public parking space and emission reductions (Box 5.6).

Currently, accessibility by public transport is especially low in developing cities, where the motorisation rate is also the lowest. Many inhabitants are thus excluded from fast access to opportunities in these cities. This raises a major equity issue that also applies, to a lesser extent, to North American cities. As investing in public transport infrastructure is costly, there is also a significant policy challenge here: how to develop a mass transit network while maintaining affordability? Indeed, in many developing cities, low-income urban residents are already too poor to afford public transport. For instance, in Sao Paolo, Mexico City and Manila public transport is beyond the reach of the 20% at the bottom of the income pyramid (Carruthers et al., 2005). Financing an inclusive, yet efficient, transport system requires innovative mechanisms, such as taxes to capture land value increases in areas served by public transport systems or contributions of private vehicle users through road infrastructure and parking charges.

Box 5.6. The ITF Shared Mobility Model

The advent of shared mobility on a large scale could change travel patterns in cities and improve accessibility significantly. The ITF Shared Mobility simulation model (ITF 2016), developed for Lisbon, Portugal, shows that the car fleet needed for daily commuting can be reduced to 3% of today’s fleet if all trips are made using a comprehensive shared mobility platform.

The Shared Mobility model simulates the current situation as a baseline and assesses the possible large-scale deployment of a shared vehicle fleet that provides on-demand transport in different scenarios, while keeping the same level of mobility. The shared alternatives in the model are designed to provide high level of acceptance by current car drivers and include on-demand door-to-door rides by “Shared Taxis” and pre‐booked transfer-free rides at pop-up stops by “Taxi-Buses”. While rail and subway services continue to operate as today in the shared-mobility scenarios, all other motorised modes, including taxis and buses, are replaced with shared alternatives.

The main findings of the study, besides the drastic car fleet reduction, include a decrease in congestion, emissions reduced by one-third, and 95% less space needed for public parking. The total vehicle-kilometres travelled is 37% less during the peak hours compared to today, while each car is running almost ten times more kilometres than currently. The induced shorter vehicle lifetimes lead to more frequent fleet renewal, which enables a quicker fleet turnover and a faster penetration of fuel-efficient or alternative fuel technologies. The shared mobility scenarios also result in cheaper trips due to more efficient use of capacity. Other benefits include the decrease in the number of transfers and better accessibility. Figure 5.16 shows the comparison of accessibility to jobs in the baseline and the shared mobility scenarios. With shared mobility, the majority of grid cells have at least 75% of the jobs in the city reachable within 30 minutes.

The challenges for policy makers lie in the creation of the right market conditions and operational frameworks. While a sudden change to a complete shared mobility system is not conceivable, a gradual installation is plausible and may already yield large benefits. In one of the scenarios, individual cars are allowed to drive in the city two working days per week, while other days the shared mobility system must be used. This results in significant reductions of congestion and emissions, and could be an opportunity for car owners to shift to the shared mobility service gradually.

Figure 5.17. Accessibility to jobs in Lisbon before (left) and after (right) the introduction of shared mobility solutions
picture

ITF (2016), Shared Mobility: Innovations for Liveable Cities.

Outlook for accessibility

This section analyses the consequences, in terms of accessibility of the different policy scenarios used for our transport demand and CO2 emission projections.

Despite their slow transport networks, cities from the developing world offer reasonable accessibility by car and public transport because of a high population density. This could change if these cities follow the same urbanisation pathway as in Europe and North America. During the post-war years, large shares of the population began to move outwards, from the city centres to suburbs, in developed countries (e.g., Mills, 1972). This resulted in urban sprawling and in a decrease of population density. This trend is still ongoing. In Europe, between 1990 and 2000, the urbanised area increased by 18.4%, while population density fell by 9% (Oueslati et al., 2014).

The determinants of urban sprawl are known. The monocentric city model (Alonso 1964; Mills, 1981) identifies income and commuting costs as essential drivers of sprawl. When inhabitants of a city have access to cheap and efficient transport means, they tend to relocate in the periphery of cities to increase the size of their dwellings. The empirical relevance of the monocentric model has been tested several times in developed countries (Brueckner and Fansler (1983), McGrath (2005)). In a developing country context, Shanzi et al. (2009) investigated the determinants of the spatial scale of Chinese cities and demonstrated the crucial role that income growth has played in China’s urban expansion. 

The analysis of the ITF city database shows that the two main drivers of urban extent are population and GDP per capita. Urban extent increases at a slower pace than population, thus bigger cities tend to be denser. However, richer cities tend to be more sprawled. For this reason, the average density of cities decreases in the baseline scenario. This is true for all regions of the world, although to different degrees. The density decrease is particularly sharp in Asia, where the GDP per capita will be a significant driver of urban expansion.

The growth in infrastructure in the baseline scenario is not sufficient to maintain accessibility levels. As density decreases, maintaining accessibility constant requires an increase in the speed of transport networks, both road and public. Although trunk road length increases with population and wealth, a wider city also requires a larger network to serve suburban dwellers. Moreover, the combined effects of urban extension, population and income growth will result in a surge in road traffic, and thus call for more road capacity to limit congestion.

The situation is particularly dramatic for Asian cities. The drop in density is sharp, -19% between 2010 and 2050, while road traffic rises up to +532%. Although the trunk road length is projected to grow by 137%, this will not be enough to alleviate congestion. Maintaining road accessibility constant would require multiplying the trunk road network by six, a growth that is not financially and environmentally sustainable. To a lower extent, similar trends are observed in transition economies and Latin America. Road accessibility is thus expected to deteriorate between 2010 and 2050 in most cities of the developed world, unless strict policy packages are put in place.

The loss in road accessibility can be compensated by infrastructure investment in the public transport system. In the baseline scenario, the number of cities with mass transit more than triples in developing countries. In cities with mass transit, public transport accessibility will not drop on average. However, significant shares of the urban population will have to rely on car for travel purposes, as providing efficient public transport services for low density suburbs is difficult.

Table 5.6. Changes in some city characteristics between 2015 and 2050 in the baseline scenario

Density %

VKM %

Trunk road provision %

Trunk road need %

Cities with mass transit(1) %

Africa

 -8

325

180

158

460

(5 to 28)

Asia

-19

532

137

295

295

(37 to 146)

EEA + Turkey

 -7

 40

 46

 39

  6

(77 to 82)

Latin America

 -8

152

 49

 92

 51

(37 to 56)

Middle East

 -2

228

 99

 98

175

(4 to 11)

North America

 -1

 68

 39

 36

  9

(35 to 38)

OECD Pacific

 -8

 12

 11

 24

 16

(19 to 22)

Transition

-15

147

 57

120

347

(17 to 76)

In the LUT scenario, where urban extent is assumed to be fixed, density grows at the same rate as population. This implies a major increase in density for regions where rapid urbanisation is still ongoing. In Africa, population density in cities will triple to an average of 24 000 inhabitants per km2. Although high, these numbers are still plausible: in the two densest cities in the world, Dhaka (Bangladesh) and Hyderabad (Pakistan), there are 40 thousand inhabitants per km2. Vehicle-kilometres also rise at a much lower pace and more cities implement mass transit systems, reducing both congestion and improving the share of the population which has access to public transport.

Table 5.7. Changes in some city characteristics between 2015 and 2050 in the LUT scenario

Density %

VKM %

Cities with mass transit %

Africa

195

48

920

(5 to 51)

Asia

140

181

627

(37 to 269)

EEA + Turkey

 41

-30

 12

(77 to 86)

Latin America

 72

27

 81

(37 to 67)

Middle East

127

177

325

(4 to 17)

North America

 54

68

 11

(35 to 39)

OECD Pacific

 24

-37

 37

(19 to 26)

Transition

 61

-27

124

(17 to 38)

This emphasises the crucial importance of land use policies in maintaining accessibility to opportunities in the developing countries. Without strict land use control, urban sprawl will increase the need for infrastructure to a point that is not sustainable. On the contrary, density decreases the need for private car, reduces distances to destinations and makes the implementation of mass transit systems more feasible.

Passenger transport in Asian cities

According to the results of the ITF model for mobility in cities, 43% of the world’s transport demand in passenger-kilometres will be in Asia by 2050. This is a region that is projected to grow significantly and rapidly in population, economic development, urbanisation rate, and motorisation level. Although increases in motorisation will bring positive benefits and contribute to economic growth, high levels of congestion, energy consumption, local air pollution, and CO2 emissions will often follow. This section focuses on urban transport trends and projections in cities in China, India and Southeast Asia.

Motorisation trends in Asian cities

In 2010, China surpassed the United States to be the largest automobile market in the world (CAAM, 2016). More than 20 million vehicles were sold in China in 2014, resulting in a total number of 92 million cars in the same year (CAAM, 2010, 2014). Despite its low vehicle ownership rate, at 58 vehicles per 1 000 persons, compared to 804 vehicles per 1 000 persons in the United States (Wang et al., 2011), it is already the world’s largest CO2 emitter and has recently become the top global crude oil importer. The level of CO2 emissions from the transport sector in China has more than doubled from 2000 to 2010 (CAIT, 2015); China’s demand for oil and subsequent CO2 emissions will only increase as its transport sector grows, unless a comprehensive range of policies and measures is implemented to alter the course of development.

In India, raising automobile sales, household income, urban population, and urbanisation are all leading to higher transport demand. Urban population has increased from 60 million in 1951 to 410 million in 2015 (United Nations, 2014). The number and size of cities have also increased, which will ultimately increase travel demand and total distance travelled. The number of cars has grown from 16 million in 1990 to almost 40 million in 2000, reaching a total number of 131 million in 2014 (IEA, 2015). Most of these vehicles are motorcycles and only 21 million of them are passenger cars. Compared to China, India has a much smaller car market but the transport sector is still a leading contributor to Indian cities’ CO2 emissions mainly because of motorcycles. India’s passenger vehicle market is heavily dominated by motorcycles, which are the fastest growing type of vehicle in India, with an average annual growth rate that is higher than all other types of motor vehicles. Hence, in addition to CO2 emissions, local air pollution is a major problem in Indian cities.

Local air pollution from transport activities, along with traffic congestion, are pressing challenges in many Southeast Asian countries. Transport demand in Southeast Asia has been steadily increasing over the past three decades and has not shown any signs of slowing down. Recent statistics from the IEA (2015) show that the number of motorcycles has grown by 177% in Vietnam between 2000 and 2013. The increase in cars was 600% in the same period. In Malaysia, the growth in cars is 148%, while the Philippines saw a relatively lower percentage at 44% over the same period of time. Indonesia has also experienced high vehicle growth rates for both motorcycles and cars, with an over 600% and 280% increase respectively. Most cars have also become single occupancy vehicles most of the time, leading to greater time delays in urban cities. It is not surprising that due to its affordability and practicality, motorcycles are the leading transport mode choice in many Southeast Asian cities. In Vietnam, 95% of all vehicles are motorcycles. These increases in motorcycles could be a reflection of the lack of an adequate public transport system and the general appeal of personal mobility over public transport. Transport challenges exist in smaller Southeast Asian countries too. Traffic congestion costs Cambodia around USD 6 million per month as a result of lower economic efficiency and the loss of working time and fuel (Sotheary and Kunthear, 2015). In Kuala Lumpur, the World Bank estimated the costs of traffic congestion to be 1.1% to 1.2% of the national GDP in 2014 (Sander et al., 2015).

Although the transport priorities in most Asian cities are traffic congestion and local air pollution, their transport sectors have increasingly become a bigger contributor of CO2 emissions. The high motorisation rate in Vietnam led to a 190% increase in its CO2 emissions from the transport sector between 2000 and 2010, which is higher than China (160%) and India (100%) over the same period (CAIT, 2015). Indonesia, Cambodia and Malaysia also recorded significant increases of transport-induced emissions. The Philippines was the only country selected in this study not to witness a growth in its transport emissions. As carbon emissions become a pressing concern in the region, there is now a greater sense of urgency for Asian cities to adopt more sustainable transport development policies and measures.

In order to successfully reduce carbon emissions, it is important to first identify measures that will also provide climate co-benefits through the development of energy efficient and low carbon transport systems in Asia. Policies and measures that support the adoption of advanced vehicle technology and alternative fuel will reduce CO2 emission by improving energy intensity. However, as a result of the increasing number of vehicles in Asia, advancing vehicle and fuel technology will not be enough to reduce CO2 emissions, as the growth of vehicle ownership and use surpass technological improvements. A range of policies and measures that include land use planning, public transport development, economic instruments, and governmental regulations need to be considered.

The development of public transport in areas with high urban population density and user demand is a way to reduce congestion and emissions. In general and especially so in developing cities, public transport provides services at a lower cost to the user than driving but travel time can be longer for transit than driving and accessibility can also decrease (see section on Accessibility). Increase in public transport efficiency and ridership can also lead to economic benefits through high economic rates of return in dense cities (Cambridge Systematics Inc. and Apogee Research, 1996).

Selected cities

The selected cities for this chapter include five Chinese, five Indian and five Southeast Asian cities that reflect a wide range of population size, motorisation rates and existing transport policies and services (Table 5.8). Some cities offer more transit services than others. For example, out of the 15 cities selected in this study, only four cities offer bus rapid transit (BRT) systems, regular bus and metro rail transit services. Many of the Southeast Asian cities also offer several informal public transport options, either in the form of three-wheelers or mini buses. The 15 cities exhibit different levels of motorisation rates, both for cars and motorcycles. On the other hand, the selected cities are all experiencing high levels of local air pollution and congestion.

Table 5.8. Transport characteristics of selected Asian cities in 2010

City

Population (Million)

Cars (Million)

Two-wheelers (Million)

Car ownership rate

Two-wheeler ownership rate

Vehicle Restriction (Year)

Transit Services

China

Beijing

15

4.81

0.35

321

 23

Yes (2011)

Bus, BRT, Metro

Shanghai

19.55

1.46

1.29

 75

 66

Yes (1994)

Bus, Metro

Guangzhou

10.49

1.36

0.54

130

 51

Yes (2012)

Bus, BRT, Metro

Tianjin

8.54

1.38

0.15

161

 18

Yes (2014)

Bus, Metro

Xi’an

4.85

0.74

0.26

153

 53

N/A

Bus, Metro

India

Mumbai

19.42

0.43

0.77

 22

 40

N/A

Bus, Metro

Delhi

21.94

1.61

3.25

 74

148

N/A

Bus, Metro

Bangalore

8.28

0.51

1.95

 62

235

N/A

Bus, Metro

Ahmedabad

6.21

0.21

1.05

 33

169

N/A

Bus, BRT

Jaipur

3.02

0.19

0.92

 62

304

N/A

Bus, BRT, Metro

Southeast Asia

Manila

11.89

1.13

6.67

 95

561

N/A

Bus, Metro

Kuala Lumpur

5.81

2.86

1.34

493

232

N/A

Bus, BRT, Metro

Jakarta

9.63

2.00

8.76

207

910

N/A

Bus, BRT

Phnom Penh

1.51

0.18

0.73

123

486

N/A

Bus

Hanoi

2.81

0.80

2.20

284

782

N/A

Bus

Ownership rates are per 1 000 inhabitants.

Chinese Cities Statistical Yearbooks for population, cars and motorcycles data in Chinese cities. TERI data for population, cars and motorcycles data in Indian cities. Southeast Asian cities data obtained from the Philippines Ministry of Transport, Malaysian Ministry of Transport, Regional Statistics of DKI of Jakarta Province, JICA, Vietnam Department of Transport, and DKI Transport statistic.

Policy scenarios

This study applies the same set of policy scenarios as in the global urban model, presented earlier in this chapter, tailoring each measure or assumption to individual cities (see Tables in Annex 5.D). Although each scenario contains a different set of policy assumptions, the framework for estimating transport demand and CO2 emissions is similar across all scenarios. The scenario outcomes are not predictions, but different possible futures based on the assumptions applied in each scenario.

The projection of transport demand in Chinese and Indian cities were estimated differently from the Southeast Asian cities due to data availability, which is a wider challenge in the region (Box 5.7).

Box 5.7. Transport data in Asia and the Pacific: Challenges and opportunities

“If you cannot measure it, you cannot improve it”, goes the quote popularly attributed to Lord Kelvin (William Thomson). The need for reliable, robust data sets is felt particularly by those working towards sustainable transport in the developing world. Better data provides direction for informed action: for example, the Cities Air, Climate, Transport database (www.citiesact.org) covering nearly 500 cities across Asia shows that 97% of these cities are not meeting air pollution targets. Identifying what modes of transport generate the air pollution would be a critical point of action.

The Asian Development Bank, together with its partners, is currently undertaking a regional project to improve the quality of transportation data in Asia and the Pacific, and widen the access to this data. The project – “Better Transport Data for Sustainable Transport Policies and Investment Planning” – intends to collect, collate and generate insights from currently available transport data in forty (40) of ADB’s developing member countries (DMCs). The data will be shared through a publicly accessible portal, and used as inputs to a transportation model to assess the potential impacts of future transport scenarios in these countries.

A key challenge towards benchmarking across countries in Asia is the non-availability of standardized definitions. A notable example relates to road vehicles. Each country is using its own set of vehicle categories, most often derived from the usage of the vehicles, which does not follow internationally-recognized classifications. Simple steps can be taken to harmonise, such as using publications like the “Illustrated Glossary for Transport Statistics” (ITF et al., 2009); these guidelines could be further enriched by feedback from users in Asia and the Pacific.

The non-collection of key transport data is also a challenge in the region. Average vehicle-kilometres driven by different vehicle sub-segments, for example, are not commonly collected by developing and middle-income countries. The non-availability of such information prevents generation of other important indicators for transport. Moreover, access to readily-available disaggregated data is often limited. Solutions need not be complicated: collection of data for additional indicators, such as vehicle-kilometres driven, can easily be incorporated into the vehicle registration and renewal processes.

City-level transport data in the region most often comes from ad hoc initiatives, supported by external parties, and is limited to the main cities in developing and middle-income countries. Inconsistencies can sometimes be found between datasets for the same parameter generated by different government agencies from the same country. Increasing the vertical integration between relevant stakeholders through discussions and workshops on transport data collection can accelerate the generation of quality data in the region. Local governments can be better capacitated in transport data collection and tapped to contribute to an integrated database.

The “Better Transport Data” initiative underlines the importance of accounting for local contexts, priorities, resources, and capacities when looking into transport data. At the same time, countries in the region can benefit from mutual sharing of experiences and methods, as well as in having standard guidelines for defining, and collecting transport data. The availability of openly accessible and curated data – through a web portal and possible crowd-sourcing tools – is also envisioned to generate interest, both from state and non-state actors, in improving transport data and research in the region.

Travel demand projections for the Chinese and Indian cities were developed using the ASIF (Activity, Structure, Intensity, Fuel type) approach (Schipper et al., 2000), where their motorisation rates were modelled based on South Korea and Japan’s historical motorisation rates for cars and motorcycles, while projections for bus and passenger rail transportation demand follow population density growth (ITF, 2015). For the five Southeast Asian cities, household travel survey data that included individual trip data such as mode choice, travel distance, travel time, and socioeconomic variables were obtained from the Japan International Cooperation Agency (JICA). These data supported the construction of five mode choice models (multinomial logit), which enabled the projections of 2050 travel demand and CO2 emissions using parameters derived from the choice models.

Baseline

In the baseline scenario, no major new public transport development, economic instruments, or governmental regulations will be implemented in the selected cities, apart from those that have already been introduced. For example, existing restrictions on vehicle registration in Chinese cities will continue but not spread to other cities that currently do not have such regulations. Motor vehicle ownership and use will continue to increase for cities without any restrictions on vehicle growth. These cities include all five Indian cities and one Chinese city, i.e. Xi’an. There is no substantial effort to improve public transport services, in terms of reducing bus and rail travel time or developing BRT systems in cities where such systems are currently unavailable. Similar assumptions were made for Southeast Asian cities. Advanced vehicle technology and alternative vehicle fuel use will continue to penetrate the market but at a relatively low rate, especially for Indian cities. There will also be no significant improvement in fuel efficiency standards, which coincides with the IEA 4DS (IEA, 2015).

ROG Scenario

In the ROG scenario, city governments play a larger role in regulating vehicle use and ownership. Governmental regulations and standards refer to non-market based policies, such as restrictions on annual vehicle ownership growth and fuel economy standards. A cap on vehicle growth rate through tight vehicle quota control regulations will ensure reductions in congestion, local air pollution and global carbon emissions. Auctions or lottery schemes that distribute the limited number of license plates available will accompany such regulations, as currently implemented in cities such as Singapore, Shanghai and Beijing. In this scenario, restrictions on vehicle registration will exist in all Chinese cities and their quotas will gradually decrease between 2030 and 2050. Fuel efficiency standards in this scenario follow the assumptions made in the IEA 2DS (IEA, 2015), which are more stringent than in the baseline scenario. Such standards are usually implemented on a national level but complement local measures.

In addition to changes in vehicle registration and fuel efficiency standards by 2050, other economic instruments will also be implemented. In this scenario, fuel taxes, road tolls and parking pricing will be widely implemented in all Chinese, Indian and Southeast Asian cities, leading to a general increase of 64% in the cost of driving in 2030 and 99% in 2050. There will also be an increase in bus and rail subsidies, which will lower the cost of public transport to users by 30% in 2030 and 50% in 2050.

LUT Scenario

In the LUT scenario, there is a greater urge for sustainable urban transport development in the cities. Therefore, all the measures of the ROG scenario will be implemented together with appropriate urban planning measures that will reduce travel distance and urban sprawl through an increase in population density. Population density will be 15% and 25% higher in 2030 and 2050 respectively in this scenario compared to the Baseline and ROG scenarios for the Chinese and Indian cities, and distance travelled will not increase beyond 2010 levels for Southeast Asian cities. This could be achieved by the development of more mixed use transit corridors. In this scenario, public transport will also be greatly improved through the decrease of bus and rail travel time by 30% in 2030 and 60% in 2050. BRT services are available in all cities in this scenario, which will further decrease transit travel time, yet increase accessibility.

Scenario results

Emission trends vary by city (Figures 5.18). The significant decreases in emissions in Chinese cities are due to the vehicle ownership regulations that will become more stringent over time, as well as reductions in vehicle use due to the assumptions made in the ROG and LUT scenarios. Xi’an, which is the only Chinese city in this study that does not have a limitation on vehicle ownership, will continue to have increasing emission levels in the Baseline and ROG scenarios. CO2 emissions will only decrease in Xi’an when a wider range of policies and measures are implemented as assumed in the LUT scenario.

Figure 5.18. Total CO2 emissions in the selected Asian cities
Million tonnes
picture

The base year is 2010 for the Chinese and Indian cities, 2015 for the other Southeast Asian cities.

 StatLink http://dx.doi.org/10.1787/888933442845

Emission projections in Indian cities differ from those in the five selected Chinese cities and will follow a much more linear trend over the next few decades. Nevertheless, the rate of growth could decrease, as shown in the ROG and LUT scenarios, if cities implement appropriate transport policies to change travel demand and behaviour, and if the pace of technological progress is quicker.

As for the Southeast Asian cities, the policy scenarios trigger the largest decrease in CO2 emissions from the baseline to the LUT scenario in 2050 for Hanoi (90%) and Kuala Lumpur (88%). These decreases largely result from the shift from cars to public transport options. The difference in CO2 emission reduction between the LUT and ROG scenarios is the smallest for Kuala Lumpur (18%) and largest for Hanoi (124%), which shows that the impact of the same combination of policy measures will vary depending on a city’s existing transport mode choice, preferences, alternatives, and policies.

Policy options for sustainable transport in Asian cities

Achieving low-carbon mobility in Asian cities requires targeted policies, which differ according to the varying transport preferences, constraints and needs found in different cities. The same set of policies and measures can trigger different outcomes. Cities are diverse in terms of the modes available and existing transport services, which leads to different policy impacts, even within the same country or region. This calls for special attention to be paid to the local context when examining the policy options of a city. The following paragraphs conclude by bringing forward some common elements for sustainable transport in Asia cities.

Government regulations are complementary to economic instruments in reducing car use and CO2 emissions. Xi’an, the only Chinese city selected in this study without an existing regulation on vehicle registration restriction, had the lowest number of cars in 2010 in China. However, without governmental regulation, its total number of vehicles increases by 82% in 2050 (baseline scenario). In the five Indian cities selected for this study, the increase in cars can be as high as 96% for the same time horizon, as is the case in Mumbai. These outcomes can be avoided by a mix of government regulation and economic instruments, in the form of fuel taxes, road pricing, parking fee, or transit subsidies, which provide incentives to drive less and use public transport more.

To be fully efficient, economic instruments need be coupled with integrated land-use and transport planning. Land-use and transport planning strategies can change the density of urban cities and diversity of activities in neighbourhoods. It reduces the average distance travelled, transport demand and CO2 emissions subsequently. Given their high population density, Chinese and Indian cities are well suited for the development of efficient public transport systems. The effect of the policies in the LUT scenario is particularly high in reducing CO2. Emphasising public transport services using high capacity vehicles also reduces congestion and emissions.

Since the share of motorised two-wheelers is high in most developing Asian cities, especially in Southeast Asia, economic instruments need also apply to them. These include road tolls and parking policies that will serve as transport demand management tools for both types of vehicle. In addition, the high shares of motorcycles often hide an issue with public transport provision. The regulation of motorcycles can only go hand in hand with an improvement of public transport, or cities face the risk of significant loss of accessibility.

The low modal shares for public transport in some Southeast Asian cities can also be explained by the existence of informal public transport services. Such services provide more flexible route and often at a lower cost than structured public transport. Since the cost of public transport is relatively low in Chinese, Indian and Southeast Asian cities, the provision of further transit subsidies may not be effective in encouraging greater public transport use. Another way to increase its appeal is to improve the quality of the services, for example, by decreasing travel time. This implies the development of faster travel modes. Speed can be further enhanced by using dedicated roadways or introducing BRT systems, and enabling smoother transfers between vehicles and other transport modes.

The LUT scenario reflects the improvement of bus and passenger rail travel time and the availability of BRT systems in all 15 cities. Compared to the baseline scenario, bus ridership will increase by 28 to 117% in Chinese cities and by 36 to 138% in Indian cities in 2030. Bus ridership in the ROG scenario, which only includes transit subsidies as a policy to increase ridership, increases on a much smaller scale. A similar trend is observed for passenger rail ridership. The improvement of the quality of the public transport network appears much more efficient in encouraging mode shift than pure economic instruments.

Finally, the results highlight the importance of timing in policy implementation. Cities should act now to reduce transport CO2 emissions. In 2010, Beijing had the highest passenger light duty vehicle demand in China, most certainly because it only started regulating the growth in private cars in 2011, whereas Shanghai proposed its vehicle ownership policy 25 years earlier in 1986 and implemented it in 1994. Despite having a larger population and higher GDP level, Shanghai has managed to keep its vehicle ownership rate low, not just because it implemented a vehicle restriction policy, but also because it started early while its vehicle ownership rate was still relatively low.

Since most of the rapidly growing cities are still at the beginning of their motorisation growth projections, cities have to act now to avoid greater levels of traffic congestion, local air pollution and CO2 emissions. The rate of vehicle ownership need not follow population and GDP increases as proven by cities such as Shanghai, Hong Kong and Singapore. At the same time, higher travel demand, measured in passenger kilometre, need not imply higher emission levels, as more energy efficient transport modes could be chosen for the same distance travelled. The integration of land-use and transport planning policies will continue to maintain a desirable level of accessibility and prevent significant urban sprawl, which will then reduce total distance travelled over time. Together with adequate infrastructure investment, robust pricing policies, improved transport services, and higher market penetration rates of energy efficient vehicles and fuel, cities will be able to achieve sustainable transport.

References

Alonso, W. (1964), Location and Land Use. Towards a Theory of Land Rent, Harvard University Press, Cambridge (Massachusetts).

Banister, D. (2008), “The sustainable mobility paradigm”, Transport Policy, New Developments in Urban Transportation Planning 15, 73-80, http://dx.doi.org/10.1016/j.tranpol.2007.10.005.

Beijing Environmental Protection Bureau (2014), The sources of local pollution in Beijing.

Bhat, C.R. et al. (2005), Measuring Access to Public Transportation Services: Review of Customer Oriented Transit Performance Measures and Methods of Transit Submarket Identification, Centre for Transportation Research University of Texas, www.utexas.edu/research/ctr/pdf_reports/0_5178_1.pdf.

BMCC and BTRC (2004), The Outline of Beijing Transport Development, Beijing Municipal Committee of Communication and Beijing Transport Research Center, Beijing.

Bowman, J.L., Ben-Akiva and M.E. (2001), Activity-based disaggregate travel demand model system with activity schedules. Transportation Research Part A: Policy and Practice 35, 1-28, http://dx.doi.org/10.1016/S0965-8564(99)00043-9.

Brueckner and Fansler (1983), “The economics of urban sprawl: Theory and evidence on the spatial sizes of cities”, Review of Economics and Statistics, Vol. 65, No. 3, pp. 479-482.

CAAM (2010), “Automotives Statistics”, 2010 China Vehicle Sales, China Association of Automobile Manufacturers, Beijing, www.caam.org.cn, assessed March 12, 2016.

CAAM (2014), “Automotives Statistics”, 2014 China Vehicle Sales, China Association of Automobile Manufacturers, Beijing, www.caam.org.cn, assessed March 12, 2016.

Cai, H. and S. Xie (2007), “Estimation of vehicular emission inventories in China from 1980 to 2005”, Atmospheric Environment, Vol. 41, pp. 8963-8979.

CAIT (2015), CAIT Climate Data Explorer, World Resources Institute, Washington, DC, www.cait.wri.org, assessed July 3, 2016.

Cambridge Systematics Inc. and Apogee Research (1996), “Measuring and Valuing Transit Benefits and Disbenefits: Summary,” Transit Cooperative Research Program Report 20, Transport Research Board, National Academy Press.

Carruthers, R. et al. (2005), “Affordability of Public Transport in Developing Countries”, Transport Papers, TP-3, January 2005, The World Bank Group, Washington, DC.

Cervero, R. and G. Arrington (2008), “Vehicle Trip Reduction Impacts of Transit-Oriented Housing”, Journal of Public Transportation, Vol. 11.

Chambliss, S. et al. (2013), The Impact of Stringent Fuel and Vehicle Standards on Premature Mortality and Emissions, The International Council on Clean Transportation. www.theicct.org/global-health-roadmap.

Chattopadhyay, S. and E. Taylor (2012), “Do Smart Growth Strategies Have a Role in Curbing Vehicle Miles Traveled? A Further Assessment Using Household Level Survey Data”, The B.E. Journal of Economic Analysis and Policy, Vol. 12.

Chen, G. and J. Kauppila (2017), “Global Urban Passenger Travel Demand and CO2 Emissions to 2050: A New Model”, presented at the 96th TRB Annual Meeting.

Daly, H. and B.P. Ó Gallachóir (2011), “Modelling private car energy demand using a technological car stock model”, Transportation Research Part D: Transport and Environment, Vol. 16, pp. 93-101.

Downs, A. (2004), Stuck in Traffic. Coping with Peak-Hour Traffic Congestion, Brookings Institution Press, Washington, DC.

Downs, A. (1962), “The Law of Peak-Hour Expressway Congestion”, Traffic Quarterly, Vol. 16/3, pp. 393-409.

Ewing, R. and R. Cervero (2010), “Travel and the built environment: a meta-analysis”, Journal of American Planning, Vol. 76/3, pp. 265-294.

Franco et al. (2014), “Real-world exhaust emissions from modern diesel cars, a meta-analysis of PEMs emissions data from EU (EURO 6) and US (TIER 2 BIN 5/ULEV II) diesel passenger cars”, ICCT White Paper.

Geerlings, H. and D. Stead (2003), “The integration of land use planning, transport and environment in European policy and research”, Transport Policy, Urban Transport Policy Instruments, Vol. 10, pp. 187-196.

Geurs, K.T. et al. (2012), Accessibility Analysis and Transport Planning. Challenges for Europe and North America, Edward Elgar, Cheltenham, www.elgaronline.com/view/9781781000106.xml.

Geurs, K.T and B. van Wee (2004), “Accessibility evaluation of land-use and transport strategies: Review and research directions”, Journal of Transport Geography, Vol. 12, 2004, pp. 127-140.

Greening, L.A. (2004), “Effects of human behavior on aggregate carbon intensity of personal transportation: comparison of 10 OECD countries for the period 1970-1993”, Energy Economics, Vol. 26, pp. 1-30.

Handy, S.L. and D.A. Niemeier (1997), “Measuring Accessibility: An Exploration into Issues and Alternatives Environment and Planning, Vol. 29/7, pp. 1175-1194.

Hansen, W.G. (1959), “How accessibility shapes land use”, Journal of the American Institute of Planners, Vol. 25, pp. 73-76.

He, D., H. Liu, K. He, F. Meng, Y. Jiang, M. Wang, J. Zhou, P. Calthorpe, J. Guo, Z. Yao, and Q. Wang (2013), “Energy Use of, and CO2 Emissions from China’s Urban Passenger Transport Sector – Carbon Mitigation Scenarios Upon the Transport Mode Choices”, Transport Research Part A, Vol. 53, pp. 53-67.

Holz-Rau, C., J. Scheiner and K. Sicks (2014), “Travel Distances in Daily Travel and Long-Distance Travel: What Role is Played by Urban Form?” Environment and Planning A, Vol. 46, pp. 488-507.

IEA (2015), Energy Technology Perspectives 2015. OECD Publishing, Paris, http://dx.doi.org/10.1787/energy_ tech-2015-en.

IEA (2016), World Energy Outlook 2016. IEA, Paris, http://dx.doi.org/10.1787/weo-2016-en.

ICCT (2014), Global Transportation Roadmap Model, www.theicct.org/global-transportation-roadmap-model.

ITDP (2016), People Near Transit: Improving Accessibility and Rapid Transit Coverage in Large Cities, New York.

ITF (2016), “Shared Mobility: Innovation for Liveable Cities”, International Transport Forum Policy Papers, No. 21, OECD Publishing, Paris, http://dx.doi.org/10.1787/5jlwvz8bd4mx-en.

ITF (2015), Low-Carbon Mobility for Mega Cities: What Different Policies Mean for Urban Transport Emissions in China and India. OECD, International Transport Forum, www.itf-oecd.org/sites/default/files/docs/2016-01-19_report_china_india_bd.pdf.

ITF, Eurostat and UNECE (2009), Illustrated Glossary for Transport Statistics.

Jovicic, G. and C.O. Hansen (2003), “A passenger travel demand model for Copenhagen”, Transportation Research Part A: Policy and Practice, Vol. 37, pp. 333-349.

Kaufmann, T., L. Radaelli and E. Shmueli (2016), “Quantitative Land Use Planning: Deploying Data-Driven Methods in The Practice of City Planning”, submitted for publication to the Proceedings of the National Academy of Sciences of the United States of America.

Kitamura, R., C. Chen, R.M. Pendyala and R. Narayanan (2000), “Micro-simulation of daily activity-travel patterns for travel demand forecasting”, Transportation, Vol. 27, pp. 25-51.

Koppelman, F.S. and C. Bhat (2006), A Self Instructing Course in Mode Choice Modeling: Multinomial and Nested Logit Models, US Department of Transportation, Federal Transit Administration 31.

Levinson, D. (2013), “Access Across America”, Center for Transportation Studies, University of Minnesota, http://access.umn.edu/research/america/auto/2013/.

Litman, T. (2016), Land-Use Impacts on Transport How Land Use Factors Affect Travel Behavior, Victoria Transport Policy Institute, Canada.

Litman, T. (2004), “Transit price elasticities and cross-elasticities”, Journal of Public Transportation, Vol. 7/3.

Litman, T. and D. Burwell (2006), “Issues in sustainable transportation”, International Journal of Global Environmental Issues, Vol. 6, pp. 331-347.

Mandel, B., M. Gaudry and W. Rothengatter (1997), “A disaggregate Box-Cox Logit mode choice model of intercity passenger travel in Germany and its implications for high-speed rail demand forecasts”, Ann Reg Sci, Vol. 31, pp. 99-120.

McGrath, D.T. (2005), “More evidence on the spatial scale of cities”, Journal of Urban Economics, Vol. 58/1, pp. 1-10.

Metz, D. (2012), “Demographic determinants of daily travel demand”, Transport Policy, Vol. 21, pp. 20-25.

Metz, D. (2010), “Saturation of Demand for Daily Travel”, Transport Reviews, Vol. 30, pp. 659-674.

Meyer, I., S. Kaniovski and J. Scheffran (2012), “Scenarios for regional passenger car fleets and their CO2 emissions”, Energy Policy, Modeling Transport (Energy) Demand and Policies, Vol. 41, pp. 66-74.

Meyer, I., M. Leimbach and C.C. Jaeger (2007), “International passenger transport and climate change: A sector analysis in car demand and associated emissions from 2000 to 2050”, Energy Policy, Vol. 35, pp. 6332-6345.

Meyer, M.D. (1999), “Demand management as an element of transportation policy: using carrots and sticks to influence travel behavior”, Transportation Research Part A: Policy and Practice, Vol. 33, pp. 575-599.

Mills, D.E. (1981), “Growth, speculation and sprawl in a monocentric city”, Journal of Urban Economics, Vol. 10/2, pp. 201-226.

Mokhtarian, P.L. and C. Chen (2004), “TTB or not TTB, that is the question: a review and analysis of the empirical literature on travel time (and money) budgets”, Transportation Research Part A: Policy and Practice, Vol. 38, pp. 643-675.

Murray, A.T. et al. (1998), “Public Transportation Access”, Transport Research Part D: Transport and Environment, Vol. 3/5, pp. 319-328.

Newman, P. and J. Kenworthy (2011), “Peak Car Use: Understanding the Demise of Automobile Dependence”, Journal of World Transport Policy and Practice, Vol. 127, pp. 31-42.

OMNIL (2012), Enquête Globale Transport, La mobilité en Ile-de-France, Cahier Numéro 1.

Oueslati, W. et al. (2014), “Determinants of urban sprawl in European cities”, Urban Studies, Vol. 52/9, pp. 1594-1614.

Owen, A. and D. Levinson (2015), “Access Across America: Walking 2014”, Center for Transportation Studies, University of Minnesota, http://access.umn.edu/research/america/walking/2014/documents/CTS15-03.pdf.

Owen, A. and D. Levinson (2014), “Access Across America: Transit 2014”, Center for Transportation Studies, University of Minnesota, http://access.umn.edu/research/america/transit/2014/.

Paez, A. (2016), “Access and social complexity: identifying and managing access requirements across social groups and across the world”, in Sclar, C. et al., Improving urban access. New approaches to funding transport investment, Earthscan, London and New York, pp. 190-217.

Paulley, N., R. Balcombe, R. Mackett, H. Titheridge, J. Preston, M. Wardman, J. Shires and P. White (2006), “The demand for public transport: The effects of fares, quality of service, income and car ownership”, Transport Policy, Innovation and Integration in Urban Transport Policy, Vol. 13, pp. 295-306.

Peralta, T. (2015), “Mobility for all: Getting the right urban indicator. Shifting from the proximity of transport to the accessibility of opportunities”, Connections, Note 25, November 2015, The World Bank, www.worldbank.org/en/topic/transport/brief/connections-note-25.

Peralta, T. and S.R. Mehndiratta (2015), “Accessibility analysis of growth patterns in Buenos Aires. Density, employment and spatial form”, Transportation Research Record: Journal of the Transportation Research Board, Vol. 2512.

Pesaresi, M. and M. Carneiro Freire Sergio (2014), Buref – producing a global reference layer of built-up by integrating population and remote sensing data.

Poelman, H. and L. Dijkstra (2015), “Measuring access to public transport in European cities”, Regional Working Paper, 01/2015, European Commission, http://ec.europa.eu/regional_policy/sources/docgener/work/2015_01_publ_transp.pdf.

Sander, G. et al. (2015), Malaysia economic monitor: transforming urban transport, Washington, DC, World Bank Group.

Shanzi, K. et al. (2009), “Determinants of urban spatial scale: Chinese cities in transition”, Urban Studies, Vol. 46/13, pp. 1-19.

Schafer, A. (2012), “Introducing behavioral change in transportation into energy/economy/environment models”, Policy Research Working Paper Series, No. 6234, The World Bank.

Schafer, A. (1998) “The global demand for motorized mobility”, Transportation Research Part A: Policy and Practice, Vol. 32, pp. 455-477.

Schafer, A. and D.G. Victor (2000), “The future mobility of the world population”, Transportation Research Part A: Policy and Practice, Vol. 34, pp. 171-205.

Schafer, A. and D.G. Victor (1999), “Global passenger travel: implications for carbon dioxide emissions”, Energy, Vol. 24, pp. 657-679.

Schipper, L, C. Marie-Lilliu and R. Gorham, (2000), Flexing the Link between Transport and Greenhouse Gas Emissions: A Path for the World Bank, Washington, DC, World Bank.

Singh, S.K. (2006), “Future mobility in India: Implications for energy demand and CO2 emission”, Transport Policy, Vol. 13, pp. 398-412.

Sotheary, P. and M. Kunthear (2015), “Congestion costing $6 million a month”, The Phnom Penh Post, www.phnompenhpost.com/national/congestion-costing-6-million-month, accessed 19 October 2016.

TERI (2015), Unpublished Transport Data for Indian Cities, The Energy and Resources Institute, New Delhi.

TfL (2010), Travel in London Report 2, Transport for London.

TomTom (2016), TomTom Traffic Index. Measuring congestion worldwide, www.tomtom.com/fr_fr/trafficindex/.

United Nations (2014), “World urbanization prospects: The 2014 Revision”, United Nations Publications, New York.

Viegas, J. and L. Martinez (2016), “Practical approaches to measuring access and social inclusion. Lessons from Lisbon”, in Sclar, C. et al., Improving urban access. New approaches to funding transport investment, Earthscan, London and New York, pp. 170-189.

Vovsha, P., E. Petersen and R. Donnelly (2002), “Microsimulation in Travel Demand Modeling: Lessons Learned from the New York Best Practice Model”, Transportation Research Record: Journal of the Transportation Research Board, Vol. 1805, pp. 68-77.

Wang, Y., T.F. Welch, B. Wu, X. Ye and F.W. Ducca (2016), “Impact of transit-oriented development policy scenarios on travel demand measures of mode share, trip distance and highway usage in Maryland”, KSCE J Civ Eng, Vol. 20, pp. 1006-1016.

Wang, Y., J. Teter, and D. Sperling (2011), “China’s Soaring Vehicle Population: Even Greater Than Forecasted?” Energy Policy, Vol. 39, No. 6, pp. 3296-3306.

Wen, C.-H., Y.-C. Chiou and W.-L. Huang (2012), “A dynamic analysis of motorcycle ownership and usage: A panel data modeling approach”, Accident Analysis & Prevention, PTW + Cognitive impairment and Driving Safety, Vol. 49, pp. 193-202.

WHO (2016), Ambient air pollution: A global assessment of exposure and burden of disease, World Health Organisation, Geneva.

Yan, X. and R.J. Crookes (2010), “Energy demand and emissions from road transportation vehicles in China”, Progress in Energy and Combustion Science, Vol. 36, pp. 651-676.

Zahavi, Y. and A. Talvitie (1980), “Regularities in Travel Time and Money Expenditures”, Transportation Research Record: Journal of the Transportation Research Board, pp.13-19.

Annex 5.A1. Data sources
Table 5.A1.1. Data Sources

Name

Description

Source

City List

Full list of cities with population above 300k by 2014

UN Habitat, WUP2014

Mode Shares

Percentage of trips (all purposes) by different type of modes

Various sources

Main Source

The EPOMM Modal Split Tool - www.epomm.eu/tems/result_cities.phtml?more=1

Other miscellaneous sources

National Household Travel Survey

Statistic year books

Reports from local transport authorities

Reports from different research institutes and organizations

Union Internationale des Transports Publics (UITP), Mobility in Cities Database

Transport Supply

Global road network

OpenStreetMap, www.openstreetmap.org/

Global public transport network

OpenStreetMap, www.openstreetmap.org/

Mobility in Cities Database

UITP

World metro database

http://mic-ro.com/metro/table.html

Rapid transit database

ITDP

Public transport network and timetables

Various public transport operators and agencies based on the General Transit Feed Specification format (GTFS), www.transitwiki.org/TransitWiki/index.php?title=General_Transit_Feed_Specification

Travel speeds

TomTom Traffic Index, www.tomtom.com/en_gb/trafficindex/

Urban Built-up Areas

BUREF – Global Built-up Reference Layer (BUREF2010) is a spatial raster dataset containing an estimation of the distribution and density of built-up areas using publicly available global spatial data related to the year 2010

European Commission, Joint Research Centre, http://publications.jrc.ec.europa.eu/repository/handle/JRC90459

LANDSAT – Landsat represents the world’s longest continuously acquired collection of space-based moderate-resolution land remote sensing data.

A joint initiative between the U.S. Geological Survey (USGS) and NASA, http://landsat.usgs.gov//about_project_descriptions.php

Population

Total population, urban population by country, cities with population above 300K

UN Habitat, WUP2014

Worldpop population raster grid at continental scales for Africa, Asia, and Latin America and the Caribbean (spatial resolution of approx. 1km; year 2010).

Worldpop, www.worldpop.org.uk/

National population raster grid for Australia (spatial resolution of approx. 1km; year 2011)

Australian Bureau of Statistics, www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/1270.0.55.0072011?OpenDocument

Geostat vector population grid for Europe (spatial resolution of approx. 1km; year 2011)

Eurostat and EFGS, http://ec.europa.eu/eurostat/web/gisco/geodata/reference-data/population-distribution-demography

National population vector grid for New Zealand (spatial resolution of approx. 1km; year 2011)

LINZ and Statistics New Zealand, https://koordinates.com/layer/8707-nz-1km-pop-grid/

American Community Survey population layer at Census Block Group level for the United States of America (Census Block group level; year 2014)

American Community Survey, www.census.gov/programs-surveys/acs/

Gridded Population of the World dataset (GPW version 4) to complement the above mentioned data sources for the following countries (Armenia, Azerbaijan, Bahrain, Canada, Cuba, Georgia, Iran, Iraq, Israel, Jordan, Kazakhstan, Kuweit, Kyrgizstan, Lebanon, Oman, Qatar, Republic of Moldova, Russia, Saudi Arabia, Serbia, State of Palestine, Syrian Arab Republic, Tajikistan, Turkey, Turkmenistan, Ukraine, United Arab Emirates, Uzbekistan, White Russia, Yemen).

GPW version 4, SEDAC, http://sedac.ciesin.columbia.edu/data/set/gpw-v4-population-count

GDP

GDP, GDP per capita projection by country

OECD ECO department

GDP by cell grid in 2010

LANDSAT

Car Ownership

Passenger Cars per 1000 inhabitant by country

IRF World Road Statistics 50th Anniversary (Data 2000-11)

Transport Prices

Transportation prices by city, e.g. gasoline per litre, monthly pass, one-way transit ticket, taxi per hour etc.

NUMBEO, open source, www.numbeo.com/cost-of-living/prices_by_city.jsp

Annex 5.A2. Methodology for the global urban passenger model

The scope of this study is all the urban agglomerations with population above 300 thousand, following the definition of UN World Population Prospects (2014 Revision). The full city list contains 1692 cities. The general model structure comprises six sub-models. The transport system is composed by three highly interrelated sub-models: Travel Demand, Transport Supply and Vehicle Fleet. The dynamics among the sub-models play a fundamental role in the quantitative analysis. The interaction between land-use and transport system is represented as Land-use sub-model influencing the mode choice and vehicle ownership and in turn being affected by the transport supply level. The Exogenous Sub-model contains the inputs of population, economy and vehicle technology providing exogenous drivers to the transport system. The outcomes of Vehicle Fleet sub-model feed into the Environment sub-model to compute the CO2 emissions. The sub-models are structured as the following:

  • Replicating the UN Habitat approach to project the urban population from 2030 to 2050;

  • A sigmoid curve to forecast the GDP growth rates for the cities. The relationship between the national share of urban population concentration and the national share of urban GDP concentration follows an S-shaped curve;

  • Regression models for urban transport supply, including road provision and public transport supply;

  • Discrete choice model to estimate the modal split of each city;

  • A sigmoid curve to forecast the passenger car ownership and assumptions to infer the share of other type of vehicles;

  • CO2 intensities and technological pathways by mode for converting vehicle activities into CO2 emission (IEA, Mobility Model).

Figure 5.A2.1. Modelling framework for the ITF global model for mobility in cities
picture

The urban boundary for each selected city is provided by the Global Built-up Reference Layer (BUREF2010) (Pesaresi and Carneiro Freire Sergio, 2014) and complimented by the space-based land remote sensing data LANDSAT of year 2010. This global urban boundary layer is used to intersect other GIS-based transport data layers to get the transport supply indicators for each urban area, such as the road and public transport supply within each urban boundary. The GDP at city level in the base year is estimated by redistributing the national GDP volume from the OECD Environment Directorate into the urban areas according to the GDP distribution map obtained from LANDSAT 2010, which provides GDP raster that measures the GDP density for each cell grid (1 square km resolution).

Transport Supply

The main source for road network and public transport network data is OpenStreetMap (www.openstreetmap.org). OpenStreetMap is an open data source and a collaborative project created with crowdsource approach, which encourages the volunteers worldwide to contribute through the collection of geographic data. Due to the nature of crowdsourcing approach, there are discrepancies in data quality across regions, countries and cities. To reduce the risk of data discrepancy, only five levels of main roads are considered, namely motorway, trunk, primary, secondary and tertiary. And some cities with poor public transport supply coverage are eliminated for the regression analysis.

The equations for the estimation of the total road network length and total number of public transport stops (bus, metro, tram, BRT, etc.) are the following,

picture

picture

Where, rdLeni, ptStopsi, popi, areai, gcapi are total road length, total number of public transport stops, urban area size and GDP per capita of city i, respectively. b is the estimated coefficient for each variable.

National car ownership

The historical car ownership (passenger car per 1000 inhabitant by country) data is collected from IRF with a time span from 2000 to 2011 that includes 169 countries of the world. The conceptualisation of the passenger car ownership model follows the study by Dargay et al. (Dargay et al., 2007). We build a passenger car ownership model that explicitly models the car saturation level as a function of observable urbanisation rate of each country. The elasticity of passenger car ownership with respect to per capita GDP follows an S-shaped curve, with car ownership rising slowly with income while income remains low, accelerating while income goes through medium levels, and slowing down again as incomes reach high levels.

picture

Where, i is the country, n is the continent, j is the income group, uRate is the urbanisation rate, gcap is the GDP per capita, bn denotes the constant term of the saturation level, u is the coefficient for urbanisation rate, gr is the growth rate, and m is the midpoint of the curve.

Car ownership is modelled as the dependent variable in the first instance, based on urbanisation rate and per capita income. The predicted car ownership is then treated as an independent variable in the development of mode share models.

Transport costs

Fuel price (gasoline price per litre) at city level is mainly from an open source database Numbeo (www.numbeo.com/), and complemented by the national level data (pump prices of the most widely sold grade of gasoline) collected from World Bank. A city without fuel price data collected is assumed to have the same price as its closest city in the same country which has fuel price observed. If no such city was found, the fuel price level is assumed to be equal to the national average price collected from the World Bank. The fuel price growth rates by different country groups are taken from IEA’s MoMo model to forecast the future fuel prices from 2010 to 2050.

Transit ticket price per trip is also collected from the same data source. A regression model is estimated to predict the transit ticket price in the future. The formulation is,

picture

Where, cj is the constant term of country group j, b7 is the estimated coefficient of GDP per capita.

Parking price is collected from the a parking rate survey carried out by Colliers International in 2011 (Moore, 2011). Daily parking cost parkingi is a function of car density (cars per square kilometre) carDensi and public transport stop density (number of public transport stops per square kilometre) ptDensi.

picture

Mode choice

Existing studies show that the aggregated mode share for each city is a function of urban development status, including urban scale, geography, economy, land use, personal behaviour and public policy (He et al., 2013, 2011; Norley, 2015). It aims to answer the questions on what are the impacts of urban development policies related to socio-economic development, car ownership, urban structure, road supply, public transport provision and pricing indicators on the aggregated modal split of a city. Mode shares are the parameters which are sensitive to the urban development policies. It is an alternative to the usual individual-based or trip-based behavioural logit models used in travel demand modelling. A standard multinomial logit model is applied.

picture

picture

picture

picture

picture

picture

Where, ASC is alternative specific constant, β is estimated coefficients, fPrice is fuel price, pPark is parking cost, ptFare is transit ticket price (single trip), rDens is road density, ptDens is density of public transport stops, mass is the availability of mass transit mode, gcap is GDP per capita, and C, PT, W, B, M are car, public transport, walk, bicycle, motorcycle, respectively.

The data set contained 247 observations, with an average weighted mode share in value of 42% for car, 30% for public transport, 18% for walk, 6% for bike and 3% for motorcycle. The calibrated model has ρ2 = 0.279, showing a satisfactory explanatory power of the mode choice, and all the variables are statistically significant.

The values for the calibrated parameters, such as the preference factors, are plausible and in line with other studies that suggest higher preference for personal car as compared to public transport and bicycle to be the least preferred mode. All the coefficients are statistically significant. The calibrated coefficients indicate that car ownership and road density have positive impacts on the car use. Positive impacts of transit stop density and the availability of mass transit are found on the use of public transport. The pricing variables, namely, fuel price, parking cost and transit fare are found to have negative impacts on the use of corresponding mode. We find urban density to contribute positively to the ridership of public transport and the use of non-motorised modes and the values of the coefficients are higher in for the public transport, followd by walking and cycling. GDP per capita uses as proxy for income level is found to have negative impacts on the use of motorcycle and non-motorised modes. This finding is in line with the existing studies that the increasing income leads to the growing demand for faster and more convenient transport modes.

Trip rates and distances

Average trip rate in this study means number of trips per day per person considering all trip purposes. In trip generation analysis, the approach involves setting up model to represent the relationship between trip rates and the socio-economic characteristics. In this study, we used a simple regression analysis to find the relationship between the observed average trip rates from the household travel surveys and the GDP per capita.

Average travel distance is defined as the single trip distance regardless of trip purpose. We used the observed sample trip distances to establish a relationship between average distance by private vehicle and the urban area size. We also obtained average differences in travel distance between different modes, such as average travel distances by public transport is 45% longer than car, distance of a bike trip is usually around 32% of a car trip distance. Based on available data, we made such a simplified average value for all cities over our study time period. If more and better data are collected in the future, the trip rates and trip distance estimations will be enhanced. The methodology will be further improved by including more explanatory variables on travel distances by mode, such as land-use mix, population density, and possibility evolve over time as well.

Vehicle technology and CO2 emission

The transport scenarios are translated into CO2 emission scenarios by applying transport technology paths. The technology assumptions and emission calculations are taken from the IEA’s MoMo model and the Energy Technologies Perspectives. The scenario used for the baseline is the four degree scenario (4DS) in the World Energy Outlook, which corresponds to a context in which broad policy commitments and plans that have been announced by countries are implemented. Under this scenario fuel economy standards are tightened and there is progressive, moderate uptake of advanced vehicle technologies (IEA, 2013 and Dulac, 2013). The result is a slow but sustained decrease in fuel intensity of travel and carbon intensity of fuel for all vehicles. Such a decrease is in general higher within the OECD region.

Annex 5.A3. Detailed results for transport speed and densities

The following two tables give regional results for the main indicators used to examine the accessibility by car (Table 5.C1) and by public transport (Table 5.C2).

Table 5.A3.1. Road speeds and density in cities

Density (thousand inhab./km2)

Free flow speed (km/h)

Congested speed (km/h)

Speed loss due to congestion %

Cities > 3 millions

Transition

5.8

21.3

14.1

51

North America

1.8

29.3

21.3

37

Africa1

14.7

17.9

10.8

67

OECD Pacific

1.8

33.6

24.3

38

EEA + Turkey

3.4

26.7

19.1

40

Latin America

8.4

19.9

13.2

50

Asia

8.6

24.6

15.8

56

Middle East

5.6

27.8

18.4

50

Cities > 1 million

Transition

2.8

19.7

12.6

57

North America

1.6

26.7

19.7

35

Africa1

7.0

17.2

10.4

65

OECD Pacific

1.8

28.4

20.7

37

EEA + Turkey

2.3

24.1

17.4

38

Latin America

8.5

19.5

12.8

52

Asia

7.5

21.2

13.3

60

Middle East

6.3

24.8

16.3

52

Smaller cities

Transition

2.4

18.6

11.3

64

North America

1.6

24.7

18.4

34

Africa1

5.6

17.6

11.0

60

OECD Pacific

1.4

25.8

19.1

35

EEA + Turkey

2.1

22.5

16.1

39

Latin America

4.5

17.6

11.7

51

Asia

6.3

19.4

12.1

61

Middle East

3.6

21.2

13.7

55

1. Density from African cities have been computed using a methodology that does not guarantee completely accurate and comparable results. It is most likely overestimated.

Table 5.A3.2. Public transport speeds and provision in cities

Urban area

Country

Area (km2)

Density (thous. inhab./km2)

Speed (km/h)

Bus provision (in number of bus calls/hour)

Mass transit provision (in number of vehicle calls/hour)

Baltimore and Washington, D.C.

USA

3 833

1.8

9.0

103 868

3 557

São Paulo

Brazil

2 488

7.9

9.2

759 835

5 467

Manila

Philippines

1 216

9.8

7.2

451 095

810

Ciudad de México (Mexico City)

Mexico

2 219

9.1

10.6

49 150

8 934

Al-Qahirah (Cairo)

Egypt

1 173

14.4

 7.1

130 171

3 252

Toronto

Canada

1 827

3.4

10.4

16 5374

18 196

Vale do Aço and Belo Horizonte

Brazil

  696

8.5

 7.8

265 348

408

Madrid

Spain

3 242

1.8

11.3

321 996

9 021

Paris

France

3 144

3.3

15.2

302 693

184 881

San Jose and San Francisco

USA

1 924

2.6

7.9

28 532

1 279

Sydney

Australia

1 639

2.7

9.6

171 500

3 558

Roma (Rome)

Italy

2 370

1.7

8.4

604 54

5 533

Athínai (Athens)

Greece

  550

5.6

6.8

21 499

2 882

Nairobi

Kenya

  539

6.0

8.8

64 496

0

Berlin

Germany

1 336

2.6

16.4

210 507

62 169

Colorado Springs

USA

  402

1.4

5.7

360

0

Adelaide

Australia

  792

1.5

8.1

50 047

3 211

Austin

USA

  702

2.0

7.8

11 661

236

Budapest

Hungary

1 374

1.3

9.8

25 310

12 308

Toulouse

France

  596

1.5

8.0

16 271

2 003

Grenoble

France

  322

1.5

7.6

4 866

3 039

Nantes

France

  305

1.9

9.1

21 550

5 560

Wroclaw

Poland

  163

3.9

7.7

10 875

3 633

Annex 5.A4. Scenario assumptions for Asian cities

The following three tables provide the declination at the city level of the assumptions of the three policy scenarios of the ITF model for mobility in cities.

Table 5.A4.1. Chinese cities

Scenario

Baseline

Robust Governance (ROG)

Integrated Land Use and Transport Planning (LUT)

2030

2050

2030

2050

2030

2050

Population Density

UN World Population Prospects

UN World Population Prospects

UN World Population Prospects

UN World Population Prospects

15% increase UN World Population Prospects

25% increase UN World Population Prospects

Public Transport Development

Average Travel Time per Commute Trip (min)

53-58

53-58

53-58

53-58

37-41

21-23

BRT availability

Beijing and Guangzhou

Beijing and Guangzhou

Beijing and Guangzhou

Beijing and Guangzhou

All Cities

All Cities

Economic Instruments

Fuel Tax Increase (%)

N/A

N/A

33 (similar to Korea)

63 (similar to Japan)

33 (similar to Korea)

63 (similar to Japan)

Parking Pricing (USD/hr)

0.78-2.35

0.78-2.35

1.40-4.23

2.03-6.11

1.40-4.23

2.03-6.11

Road Tolls

0.78

0.78

1.17

1.56

1.17

1.56

Bus Subsidy Increase (%)

N/A

N/A

30

50

30

50

Rail Subsidy Increase (%)

N/A

N/A

30

50

30

50

Governmental Regulations

Annual Vehicle Registration Restriction

Beijing

211 200

151 200

144 270

48 090

144 270

48 090

Shanghai

100 000

100 000

146 440

73 220

146 440

73 220

Guangzhou

120 000

120 000

136 482

68 241

136 482

68 241

Xi’an

N/A

N/A

100 000

73 904

100 000

73 904

Tianjin

100 000

90 000

90 000

68 820

90 000

68 820

Fuel Economy Standards

All Cities

IEA 4DS

IEA 4DS

IEA 2DS

IEA 2DS

IEA 2DS

IEA 2DS

Table 5.A4.2. Indian Cities

Scenario

Baseline

Robust Governance (ROG)

Integrated Land Use and Transport Planning (LUT)

2030

2050

2030

2050

2030

2050

Population Density

UN World Population Prospects

UN World Population Prospects

UN World Population Prospects

UN World Population Prospects

15% increase UN World Population Prospects

25% increase UN World Population Prospects

Public Transport Development

Average Travel Time per Commute Trip (min)

45-60

45-60

45-60

45-60

32-42

18-24

BRT availability

Delhi, Ahmedabad, Jaipur, and Indore

Delhi, Ahmedabad, Jaipur, and Indore

Delhi, Ahmedabad, Jaipur, and Indore

Delhi, Ahmedabad, Jaipur, and Indore

All Cities

All Cities

Economic Instruments

Fuel Tax Increase (%)

N/A

N/A

63 (similar to Japan)

63 (Similar to Japan)

63 (similar to Japan)

63 (Similar to Japan)

Parking Pricing (USD/hr)

0.60-0.91

0.60-0.91

1.08-1.64

1.56-2.37

1.08-1.64

1.56-2.37

Road Tolls

0.61

0.61

0.92

1.22

0.92

1.22

Bus Subsidy Increase (%)

N/A

N/A

30

50

30

50

Rail Subsidy Increase (%)

N/A

N/A

30

50

30

50

Governmental Regulations

Fuel Economy Standards

All Cities

IEA 4DS

IEA 4DS

IEA 2DS

IEA 2DS

IEA 2DS

IEA 2DS

Table 5.A4.3. Southeast Asian Cities

Scenario

Baseline

Robust Governance (ROG)

Integrated Land Use and Transport Planning (LUT)

2030

2050

2030

2050

2030

2050

Public Transport Development

Average Travel Time1 Decrease (%)

N/A

N/A

N/A

N/A

30

60

Economic Instruments

Fuel Tax Increase (%)

N/A

N/A

33 (similar to Korea)

63 (similar to Japan)

33 (similar to Korea)

63 (similar to Japan)

Parking Pricing Increase (%)

N/A

N/A

80

160

80

160

Road Tolls

N/A

N/A

USD 1.02

USD 1.36

USD 1.02

USD 1.36

Bus Subsidy Increase (%)

N/A

N/A

30

50

30

50

Rail Subsidy Increase (%)

N/A

N/A

30

50

30

50

Governmental Regulations

Fuel Economy Standards

All Cities

IEA 4DS

IEA 4DS

IEA 2DS

IEA 2DS

IEA 2DS

IEA 2DS

1. Based on household travel survey data from each city. In the Baseline and ROG scenarios, the travel time for all modes used in the mode choice and emission models were all from the household travel survey data. In the LUT scenario, travel time for bus and train would decrease by 30% in 2030 and by 60% in 2050.