1 Introduction
Worldwide, 55% of the global population lives in urban areas and the present urban population is projected to increase from today’s 4 billion people to 6 billion by 2050 [
48]. Mainly as a result of migration from rural areas, cities are growing in terms of inhabitants and urban area and form new residential areas outside or further away from the city core. However, the speed of urbanization presents challenges such as meeting the growing demand for transport infrastructure and affordable housing. Urban zones take different forms and characterizations and urban growth patterns differ amongst regions as a result of socio- economic, cultural, historical and environmental differences. As an example, in the US, people tend to live in low-density, single-family houses and commute by car to work. In Japan by contrast, high-rise residential buildings dominate and workers commute by public transportation (mostly rail-based) [
2]. In order to identify the most promising city development policy, it is of primary interest to assess the relations between network infrastructure, socio-economic indicators and the transport system performance based on experiences from existing cities; the understanding how cities are shaped by setting the appropriate transport priorities can help to achieve terms of sustainable mobility objectives [
36].
The relationship between transport infrastructure expansion and population growth, spatial expansion and land-use change has been highlighted in many works [
1,
5,
46]. A tight relationship between transport and urban development has been shown in earlier works [
34,
37]. The imbalance between travel demand and transport infrastructure supply as reason for the increase in congestion has been studied by Aljoufie et al. [
1]. High congestion levels cause significant costs to society; it has been estimated that exposure to traffic congestion reduces welfare in the US by $557 million per year [
17] and the estimation of congestion cost to UK economy is approximately £13 billion per year, in a forecast through 2030 increasing to £21 billion per year [
44]. Congestions impede the proper functioning of more sustainable transport modes such as bus services or cycling; as a consequence, existing bus services could neither meet the growing transport demand, nor meet the demand of the cities’ economic development [
45]. Due to these negative impacts, congestion levels are a good candidate as transport performance indicator. More specific relations between infrastructure expansion and various transport indicators have been found in the studies cited below.
The expansion of road network generally leads to lower population density in cities: Baum-Snow et al. [
4] have shown that the integrated effects of ring roads and highways in Chinese cities gave rise to move 25% of central inhabitants to surrounding zones. The empirical estimates from Baum-Snow [
3] show further that each highway expansion within an urban center of US metropolises causes an average 18% drop of inhabitants in the city center. An analysis in Wisconsin within 1980–1990 demonstrated that highway expansions caused population increase in suburban areas and booming the urban sprawl [
13]. Similar results have been shown by analyses in California between 1980 and 1994 [
12]. At the contrary, rail network expansion has been shown to increase population density at nearby urban rail stations or tracks in several studies, thereby strengthening compactness of urban areas [
6,
28,
31].
The strong correlation between road infrastructure expansion and growth of vehicle ownership has been determined for 50 countries and 35 cities [
26]. A positive relationship between highway expansion and car usage has been shown between 1982 and 2009 in the US [
32]. A negative correlation between transit ridership and highways length has been found for the Montreal Region [
15]. A sharp rise in car ownership in cities with low railway intensity and on the other hand a relatively slow rise of car ownership in cities with high railway intensity have been shown for six Asian megacities located in China, Japan and Thailand [
27]. US cities with rail lines experienced larger declines in car usage than cities without rail infrastructure between 2000 and 2009 [
25]. Similar modal shifts have been shown to exist in Europe: averaged over 14 LRT systems, approximately 11% of car drivers have changed to rail [
24]. With growing concerns over traffic congestion and pollution from motorized vehicles, Dill and Carr [
18] have indicated a positive correlation between bicycle usage and bicycle infrastructure expansion in 43 US cities based on data from Bureau of the Census. This finding has been confirmed and quantified based on a survey from 13 European cities [
40].
In summary, an extension of the road network tends to decrease urban population density, decrease the effectiveness of road based public transport -- conditions for favoring an increase in car ownership. A consequence of these effects is a further increase in private road transport demand which is often cited as “induced demand” [
30]. Rail and bike networks have been shown to achieve de-congesting effects.
The choice of suitable and relevant indicators for the analysis of transport policies is not obvious. Different definitions of “accessibility” have been used as indicators. Geurs and Van Eck [
22] has described various components of “accessibility”: land-use, temporal, individual and transport. In an extensive review, Geurs and Wee [
23] identified four types of possible accessibility measures: infrastructure-based, location-based, activity-based and utility-based accessibility.
Based on these findings and conditioned by the availability of accessible data, this study will use the length of transport infrastructure per person to quantify the amount of available transport infrastructure. This term is known as infrastructure accessibility [
21]. The transport performance is quantified by congestion levels.
The aim of the present study is to shed more light on relations between transport-socio-economic indicators and transport performance indicators. The used data is thought to be comparable across all selected cities, allowing an absolute global evaluation of the transport performance indicator. With respect to previous studies, the number of comparable cities is larger and more recent. Concrete transport policies are addressed by answering this question: under which conditions do more railways and bicycle infrastructure reduce congestion levels?
The next section motivates the data collection for this work and explains the principle data processing steps. The analysis and results are presented and discussed in
Analysis and results section, while the conclusions in Sec. 4 summarizes the main findings.
4 Conclusions
In the past, the limited availability of comparable data on socio-economics, transport infrastructure and transport performance of cities prevented a holistic analysis with many indicators, due to the lack of variety. These limitations have been overcome by analyzing OSM data, Tomtom data and data from centralized internet databases. To date, no systematic worldwide infrastructure analyses based on OSM data has been performed. Using the Python package called OSMnx, it has been possible to extract different network-types from the OSM data, downloaded from different urban areas of the world. The 151 analyzed cities are distributed over 51 countries. The cities have been analyzed as a whole and within subgroups of cities with distinct population sizes (small cities, mature cities and metropolises). Relationships between socio-economic indicators, infrastructure accessibility and congestion level have been investigated.
Good correlation values between infrastructure accessibility, socio-economic indicators, and congestion levels have been demonstrated with a reasonable goodness of fit. The analyses have shown that cities with higher GDP have built more infrastructure which in turn results in lower congestion levels. The relation between infrastructure accessibility and congestion levels has been quantified using regression models. For cities with low population density (above approximately 1500 Inh. per sq. km), more roads per inhabitant lead to lower congestion levels. Metropolises and mature cities with high population density have in general lower congestion levels where rail infrastructure per person is higher. There is significant evidence that, in case of high density cities, an increase in train infrastructure accessibility is more de-congestionating than an increase in road infrastructure accessibility.
The available data could be further exploited to determine the transport-related energy consumption in cities, updating the worldwide comparison of Newman and Kenworthy [
35]. However, this would require more information on modal split and trip distances, data which is more difficult to retrieve in a consistent manner.