public transport, rail systems, private transport, rail cities, World Transport Policy, Sustainable Transport, median value, passenger transport, urban transport, public transport systems, automobile dependence, statistical significance, SRCs, NRCs, Island Press, transport infrastructure, Felix Laube, international study, bus systems, public transport system, statistically significant, Transport systems, urban rail systems, urban rail, LRT, transport system, urban density, Institute for Science and Technology Policy, light rail, Jeff Kenworthy, Transport Planning, Authors Bionote Jeff Kenworthy, Transportation Research Board, Peter Newman, Laube F., U.S.A., University Press of Colorado, Murdoch University, Washington DC, Laube
An International Review of The Significance of Rail in Developing More Sustainable Urban Transport Systems in Higher Income Cities
Introduction With growing attention being paid to sustainability issues, most cities are making efforts to restrain the growth in automobile dependence. Many avenues are available to cities in the pursuit of this goal. Physical planning policies can aim to make development more compact with mixed land uses, thus building in less auto-dependence at the start (Cervero 1998, Newman and Kenworthy 1999a). Economic policies towards the automobile can seek to minimise car ownership and use through higher prices that perhaps better reflect the car's true social cost, as has happened in Singapore for example (Ang 1990, 1993). Amongst these efforts, there is a general recognition that the role of public transport needs to be enhanced, along with its companion modes, walking and cycling, and the latter for reasons of health, not just transport (Pucher 2002, Pucher and Dijkstra 2003). Within this general recognition that public transport can play a much greater role in most cities, arguments exist about the most appropriate modes to install to achieve enhanced public transport use and other desirable qualities, such as cost-effectiveness, integration with land uses and ability to shift people out of cars. In particular, there is considerable debate about buses versus rail (e.g. Henry 1989, Pickrell 1990). Some argue that rail is very capital intensive and that well-conceived bus systems can achieve the same results at a fraction of the cost (Bonsall 1985, Kain and Liu 1999). This argument is strongly used in lower
income cities where there appears to be less financial capacity to afford the extra capital cost
s of rail systems (Badami 2005). Others argue that rail systems in general have greater intrinsic passenger appeal and that they compete better with cars (Newman and Kenworthy 1991). Hass-Klau et al (2003) have made extensive studies of European cities with and without light rail systems and have concluded strongly that those cities that develop LRT systems consistently outperform, across many criteria, those cities that attempt to run their public transport systems only using buses. Likewise, a report from Litman (2004) of the Victoria Transport Policy Institute called `Rail Transit In America: Comprehensive Evaluation of Benefits' evaluates rail's benefits in terms of transport system performance in 130 U.S. cities. It finds that cities with large, wellestablished rail systems have a wide range of system-wide benefits relative to those that have no urban rail (see later). It is further argued that rail stations are natural sites for dense residential and mixed-use development which can help to reshape the city into a more sustainable urban form (Cervero 1995, Kenworthy 1995, Cervero 1998, Newman and Kenworthy 1999a, Hass-Klau, et al 2004). In order to contribute a more international perspective on the issue of the merits of rail in cities, this paper will explore a wide range of transport, economic and environmental features in
World Transport Policy & Practice___________________________________________________ 21 Volume 14. Number 2. July 2008
60 higher income metropolitan areas
that have strong urban rail systems compared to those that have weak rail systems or no rail systems at all. The term `cities' in relation to data in this paper refers generally to whole metropolitan regions, not the smaller administrative unit at the heart of the region, which often bears the same name (e.g. City of New York etc.). Higher income cities were defined for the purposes of this research as those with annual GDPs per capita of $US10 000 or more (i.e. it embraced those cities that are generally perceived as being part of the `developed world', as opposed to cities that are clearly in developing nations). It will examine the evidence for whether urban rail in a city's public transport system appears to make any observable, statistically significant difference to the broad patterns of transport and related factors at a metropolitan scale.
This paper draws upon the Millennium
Cities Database for Sustainable Transport
developed by Kenworthy and Laube
(2001), which in turn built on and
extended earlier work by Newman and
Kenworthy (1989) and Kenworthy and
Laube (1999). Some details about items
in the Millennium database, including
methodologies behind the research can be
found in Kenworthy and Laube (1999),
Kenworthy and Laube et al (1999) and
Newman and Kenworthy (1999a). More
specific details about other variables in
the Millennium database are available
from the author.
The list of 24 `strong rail', 28 `weak rail' and 8 `no rail' cities involved in the research in this paper, together with their 1995/6 populations, appears in table 1.
Rail in this study is defined as the combined modes of trams, LRT, metro and suburban rail. The strong rail cities (SRCs) have been defined using three criteria: · To be classed as a SRC, cities were required to have more than 50% of their total public transport task (public transport passenger travel measured as passenger kilometres) on rail, the weak rail cities (WRCs) have rail systems that account for less than 50% of their total public transport passenger kilometres and no rail cities (NRCs) have either no rail systems or rail systems that are so negligible in terms of extent and usage as to be tantamount to having no rail. Cities in table 1 that fulfill this last criterion are Tel Aviv
, Denver, Los Angeles
and Taipei where rail usage in 1995 is negligible due to the existence of only very small rail systems. · SRCs also had to have no less than 40% of total public transport boardings by rail modes. · Finally, for classification as a SRC, cities were required to have rail systems that are competitive with the car in speed terms. The overall average speed of all rail modes in each city was calculated, weighted by passenger hours, and expressed as a ratio of the average road traffic speed. Only those cities with an average rail speed that was equal to or greater than 0.90 of the road speed were classed as SRCs. Most SRCs exceeded this criterion, often by a considerable margin.
World Transport Policy & Practice___________________________________________________ 22 Volume 14. Number 2. July 2008
STRONG RAIL CITIES Washington New York Brisbane Sydney Wellington Barcelona Berlin Berne Brussels Frankfurt Hamburg London Madrid Munich Oslo Paris Ruhr Stockholm Stuttgart Vienna Zьrich Osaka Sapporo Tokyo
POPULATION (1995/6) 3,739,330 19,227,361 1,488,883 3,741,290 366,411 2,780,342 3,471,418 295,837 948,122 653,241 1,707,901 7,007,100 5,181,659 1,324,208 917,852 11,004,254 7,356,500 1,725,756 585,604 1,592,596 785,655 16,828,737 1,757,025 32,342,698
WEAK RAIL CITIES
Calgary Atlanta Chicago S. Francisco Montreal San Diego Toronto Vancouver Melbourne Perth Amsterdam Athens Copenhagen Dusseldorf Graz Helsinki Lyon Marseille Nantes Rome Geneva Glasgow Newcastle Manchester Milan Hong Kong Singapore Seoul
767,059 2,897,178 7,523,328 3,837,896 3,224,130 2,626,714 4,628,883 1,898,687 3,138,147 1,244,320 831,499 3,464,866 1,739,458 571,064 240,066 891,056 1,152,259 798,430 534,000 2,654,187 399,081 2,177,400 1,131,000 2,578,300 2,460,000 6,311,000 2,986,500 20,576,272
NO RAIL CITIES Ottawa Denver Houston L. Angeles Phoenix Bologna Taipei Tel Aviv
POPULATION (1995/6) 972,456 1,984,578 3,918,061 9,077,853 2,526,113 448,744 5,960,673 2,458,155
Table 1: Strong rail, weak rail and no rail cities in the study
The Millennium Cities Database contains complete data for 84 metropolitan areas worldwide, of which 24 can be considered as lower income (i.e. with a GDP per capita of less than $US10 000 per annum). All of these cities, apart from those in Eastern Europe
and South Africa
are clearly located in `developing nations'. However, Eastern European cities such as Prague in 1995 had low GDPs per capita but cannot be considered as `developing cities', whilst South African
cities present a starkly mixed picture whose GDPs per capita are low because of the huge
World Transport Policy & Practice___________________________________________________ 23 Volume 14. Number 2. July 2008
majority poorer populations. Attempts were made to conduct the analysis of the role of urban rail in all these lower income cities as well, but by the criteria just described, only three of these 24 cities could be considered as having strong rail systems. A larger sample of lower income cities worldwide for which comprehensive and reliable data were available would yield more SRCs so that the analysis could be meaningfully conducted, but this was not possible for this paper. The focus of this paper is therefore on cities in the `developed world', as shown in table 1 whose GDPs per capita range from $US10 305 up to $US54 692 per annum. Tables 2 to 7 systematically examine how the strong rail, weak rail and no rail cities perform on a wide range of factors using 1995/6 data. The values for each variable in the tables are the medians for the three groups of cities, since the data in each case are generally skewed distributions where the median value is a better representation than the mean. In order to test the statistical significance
of the difference amongst the medians, the nonparametric Kruskal-Wallis test from SPSS was used. The Kruskal-Wallis test is used for simultaneously testing multiple cases and eliminates the increased probability of significant results that occurs where, in this case, three separate pair-wise tests could have been undertaken for each variable. Since the samples are relatively small and the asymptotic significance value is not accurate enough, the Monte Carlo simulation of the Kruskal-Wallis test was employed using 100 000 iterations, which gives a 99% confidence level
for the pvalue (significance of the difference in the medians for each variable). P-values of 0.05 or less (95% confidence level) were considered statistically significant and
these are shown in the last column of each table, with significant results marked with an asterisk*.
Urban form and GDP Table 2 shows the differences in urban form between the groups of cities, as reflected by density and centralisation of jobs, as well as economic differences in the cities expressed through the GDP per capita of the urban regions.
systematically higher in the cities with rail
and lowest in the no rail cities, the result
is not statistically significant. Since
density is a powerful determinant of
transport patterns, especially private car
use (e.g. Kenworthy and Laube et al
1999, Newman and Kenworthy 1999), it
is useful for the purpose of this research
that differences in densities between the
three groups of cities are not significant.
On the other hand centralisation of the
city, as measured by the proportion of
metropolitan jobs in the CBD, is clearly
highest in the SRCs (18.2%) and lowest
in the NRCs (10.2%) and the differences
are statistically significant. This might be
expected, given the link between radial
urban rail systems and the development
of strong city centres, through rail's
capacity to deliver large numbers of
people into small areas (Thomson 1978).
Amongst these high-income cities, the SRCs are clearly wealthier than both other groups of cities in a statistically significant way, and as the next section shows, they are also more public transport-oriented. This undermines the idea that cities inevitably become more auto-dependent and move inexorably away from public transport as they become wealthier. In this significant international sample of higher income cities, the reverse would
World Transport Policy & Practice___________________________________________________ 24 Volume 14. Number 2. July 2008
appear to be true. We have argued elsewhere that excessive automobile dependence drains the economy of cities and there is some tacit support for this in the results in table 2 (e.g. see Kenworthy et al 1997).
The additional relevance of some of these data to the arguments made in this paper will become more apparent in later discussions.
Urban form and GDP
Strong Rail Cities
Weak Rail Cities
No Rail Cities
Urban density (persons per ha) Job density (jobs per ha) Proportion of jobs in the CBD (%) Metropolitan GDP per capita (US$1995)
47.6 27.4 18.2% $35,747
36.6 16.1 14.6% $26,151
27.7 13.4 10.2% $27,247
0.453 0.293 0.008* 0.014*
Table 2: Median values and statistical significance for urban form and GDP in strong, weak and no rail cities (1995)
Operational performance of public transport Table 3 examines differences in public transport operational performance (service and use). The first item reveals a key basis for the formation of the groups of cities. It shows how the SRCs clearly rely much more heavily on rail systems to deliver public transport mobility, with a median value of 74% of passenger kilometers on rail modes, compared to 43% and 0.4% respectively for the other two groups of cities. Looking more broadly at the public transport operational measures, table 3 shows that the supply of public transport service rises systematically from NRCs to SRCs for both vehicle and seat kilometres of service per capita. SRCs have over four times higher seat kilometres of service per capita than the NRCs. In usage, there is the same ascending pattern from NRCs to SRCs for boardings, passenger kilometres and the proportion of total motorised passenger kilometres on public transport. Public transport use is some
three to four times higher in the SRCs than in the NRCs, depending on the measurement used. This is especially interesting in the light of the urban density data in table 2, which show that there is no statistically significant difference in the median population and job densities between the three groups of cities. Interestingly, however, despite these big differences in the supply and use of public transport, per capita use of public transport energy is only some 1.6 times higher in the SRCs than in the NRCs, though the difference amongst the medians on this factor is statistically significant. This demonstrates the intrinsically high energy efficiency of public transport systems in providing mobility (i.e. service and use are four times higher in the SRCs compared to the NRCs, while energy use to run the systems is only 1.6 times higher).
World Transport Policy & Practice___________________________________________________ 25 Volume 14. Number 2. July 2008
Public transport operational performance indicators Percentage of pubic transport passenger kms on rail Annual public transport vehicle kilometres of service per capita Annual public transport seat kilometres of service per capita Annual public transport passenger trips per capita Annual public transport passenger kms per capita Percentage of total motorised passenger km on public transport Annual public transport energy use per capita (megajoules: MJ)
Strong Rail Cities 74% 77 4,086 275 1,628 21.8% 1,107
Weak Rail Cities 43% 50 2,704 188 975 12.3% 880
No Rail Cities 0.4% 29 969 77 496 5.3% 675
0.000* 0.000* 0.000* 0.002* 0.000* 0.004* 0.019*
Table 3: Median values and statistical significance for operational performance of public transport in strong, weak and no rail cities
Overall, each of the factors in table 3 varies in a strong, statistically significant way in favour of greater rail-orientation of the city. This suggests that for public transport to maximise its role within the passenger transport systems of cities in the developed world, it would appear necessary to move increasingly towards urban rail as the backbone and mainstay of those systems.transport infrastructure
Table 4 presents a range of public and
parameters for the three groups of cities.
The data on the extent of transport
performance reveal, not unexpectedly,
that the SRCs have very significantly
higher reserved public transport route on
a spatial and per capita basis. The vast
majority of reserved right-of-way (ROW)
in cities is rail; physically segregated
busways are very rare (which can be
inferred from the fact that in the NRCs,
which have either no or negligible
amounts of rail ROW, the quantity of
reserved public transport route in total is
indeed very small).
The SRCs have the lowest total per capita road supply and lowest per capita freeway provision of all three groups of cities and the NRCs have the highest. For example, the NRCs have 71% greater per capita supply of freeways than the SRCs and 65% greater road provision. Although in both cases the differences amongst the median values between the groups are not significant, the consistent direction of the results suggests that higher income cities with more significant rail systems appear to be able to function with fewer roads and freeways. Perhaps not surprisingly, the data show that SRCs have very much reduced parking supply in their CBDs (68% less than the NRCs), as do WRCs (48% less than NRCs). This is due to rail's capacity for effectively delivering high volumes of people into constrained sites such as CBDs and sub-centres, which eliminates the need for the extensive CBD parking areas found in cities that have no rail systems. Thomson (1978) found similar results in his `strong-centre' cities.
World Transport Policy & Practice___________________________________________________ 26 Volume 14. Number 2. July 2008
Transport infrastructure and performance indicators Total length of reserved public trans. routes per 1000 persons Total length of reserved public transport routes per urban ha Length of road per capita (metres) Length of freeway per capita (metres) Parking spaces per 1000 CBD jobs Total private and collective passenger VKT per km of road Overall public transport system speed (km/h) Ratio of public transport system speed to road traffic speed
Strong Rail Cities 172 9.0 3.0 0.070 186 2,026,433 31.3 0.86
Weak Rail Cities 78 3.0 4.1 0.098 303 1,461,402 23.8 0.70
No Rail Cities 7 0.4 5.8 0.120 585 1,615,749 22.6 0.49
pvalue 0.000* 0.000* 0.398 0.282 0.002* 0.708 0.000* 0.000*
Table 4: Median values and statistical significance for transport infrastructure and infrastructure performance in strong, weak and no rail cities
Finally, the data in table 4 show that in the high-income cities, the intensity of road usage or congestion, as measured by the total private and collective passenger VKT per kilometre of road, is highest in the SRCs, but the differences in the medians are statistically very insignificant. The more important point here, however, is not so much the level of congestion as the competitiveness between private and public
transport. In this respect it is very clear that the more rail-oriented the city, the higher the overall average public transport speed for all modes (39% higher in SRCs compared to NRCs) and the higher the ratio between the overall speed of the public transport system and the speed of general road traffic. The median value of this ratio for SRCs is 0.86, while for the NRCs it is only 0.49, which suggests that in speed terms public transport will generally struggle against the car in wealthier cities with no rail systems, while in cities with strong rail systems, public transport speed competitiveness will be much better. The results for both the overall speed of public transport and the speed ratio between public transport and general road traffic are statistically very
significant with p-values of 0.000 in each case. It has been suggested elsewhere that it is this relative speed between public and private transport that is a critical factor in giving public transport a competitive edge over private transport (Laube 1998, Newman and Kenworthy 1999a, b). Overall, it can be suggested that rail systems help in minimising the amount of road, freeway and parking infrastructure required in cities and are a central ingredient in developing public transport systems that can successfully compete with cars in the critical area of travel speed. Private transport patterns Table 5 provides a core set of data related to patterns of private transport and broader modal split in the three groups of cities. The data reveal that in terms of modal split, there is a systematic pattern in these high-income cities of enhanced use of both non-motorised modes and public transport and reduced use of private modes the more rail-oriented are the cities, and the results have a high level of
World Transport Policy & Practice___________________________________________________ 27 Volume 14. Number 2. July 2008
statistical significance. For example, in the SRCs, the median value for the percentage of total daily trips by private transport is 47%, whilst in the NRCs, it is 84%. The WRCs also have only 56% of daily trips by private transport. Likewise, the median value for non-motorised mode use is almost three times greater in the SRCs than the NRCs, while public transport use for daily trip making is some four times higher. Despite this modal split pattern, table 5 reveals that there is very little difference
between the car ownership and actual car travel (VKT and passenger kms per capita in cars and motor cycles) in SRCs and WRCs. However, there is a considerable difference between these more railoriented cities and the cities with no rail, though overall the differences amongst the medians are not statistically significant. Despite this lack of overall statistical significance amongst the medians, the NRCs do have about 70% higher median car use than both the SRCs and WRCs.
Private transport indicators Total cars and motor cycles per 1000 people Private passenger vehicle VKT per capita (cars + mc) Private vehicle passenger kilometres per capita (cars + mc) Percentage of all trips by non-motorised modes Percentage of all trips by public transport Percentage of all trips by private transport
Strong Rail Cities 463 5,133 6,981 31.2% 19.3% 47.5%
Weak Rail Cities 476 5,151 7,014 20.8% 13.8% 56.3%
No Rail Cities 544 8,732 11,736 11.3% 4.7% 83.8%
pvalue 0.256 0.276 0.252 0.001* 0.007* 0.000*
Table 5: Median values and statistical significance for private transport indicators in strong, weak and no rail cities
What is quite interesting about this pattern of private transport use is its relationship to the density and GDP data presented earlier. First, there is a very strong and statistically significant negative relationship found between urban density and private transport use per capita in the higher income cities in this study (R2 of 0.8392); it is virtually the strongest correlation found between all the variables in the entire database. As such it could be expected that the NRCs, with a lower median value of urban density (27.7 per ha) than the SRCs (47.6 per ha), would tend to have higher car use per capita, just based purely on their more sprawling land use patterns. Based on the equation of the regression curve
between urban density and car passenger kilometres per capita, the NRCs could be expected to have approximately 2 700 more car passenger km per capita than the SRCs. In fact, the difference in Table 5 is 4 700, perhaps suggesting that without the superior public transport systems of the SRCs, the NRCs struggle to substitute car use with public transport use. There is some support for this suggestion in the literature in what is known as the `transit leverage effect' where one passenger km of public transport travel replaces multiple kilometres of travel in cars (Neff 1996, Newman and Kenworthy 1999).
World Transport Policy & Practice___________________________________________________ 28 Volume 14. Number 2. July 2008
Furthermore, it is clear that the SRCs in this study have significantly higher GDP per capita than either the WRCs or the NRCs (37% and 31% respectively: see table 2). It is thought by some commentators that greater wealth in a city tends inevitably towards higher automobile dependence and therefore that the SRCs would be unlikely to have equal or lower car use than the WRCs and NRCs with their considerably lower GDP per capita (Gomez-Ibaсez 1991, Lave 1992). Again, it would appear that the NRCs are experiencing considerably higher dependence on the car than either their urban form or wealth characteristics would point towards. The data on private transport and overall modal split strongly suggest that rail is a significant factor in minimising automobile dependence in cities in the developed world. Strong rail systems apparently help in developing urban characteristics that together favour less private transport use (though not necessarily statistically significant lower car + motor cycle ownership), and greater capacity to exploit both public transport and nonmotorised modes. Economic factors Table 6 summarises some important indicators of the economic performance of urban systems in relation to transport. Many discussions on the overall effectiveness of urban public transport systems focus on the `subsidy' afforded to public transport, particularly as reflected in the operating cost recovery of the system. Whilst it can be argued that this focus constitutes a very limited view of the significance of public transport systems in keeping a city operating
effectively and minimising environmental impact
s (e.g. none of public transport's benefits to non-users such as congestion minimisation appear on the credit side of the balance sheet), and that the word `subsidy' is something of a misnoma, it is nevertheless important to examine this factor. The data show that it is the SRCs that have the best recovery of operating costs (60%) with WRCs at 51%, while the NRCs recover a much lower figure of 35% and these differences in the medians are statistically significant. Although the differences in average public transport vehicle occupancy in table 6 are not statistically significant, the SRCs do have 16% higher occupancy than the NRCs, which would partly explain the better cost recovery result. Rail cities tend to concentrate public transport services into more focussed corridors with more transit-supportive land uses, which generally deliver higher patronage per unit of service supplied. On the other hand, cities with no rail or those relying solely or almost solely on buses, tend to have public transport systems that have to `chase' fewer patrons through lower density settings, which inevitably detracts from higher rates of cost recovery.
World Transport Policy & Practice___________________________________________________ 29 Volume 14. Number 2. July 2008
Economic indicators Public transport operating cost recovery (%) Overall public transport vehicle occupancy Percentage of metro GDP spent on public transport investment Percentage of metro GDP spent on road investment Total passenger transport cost as percentage of metro GDP
Strong Rail Cities 60% 19.8 0.42% 0.73% 9.03%
Weak Rail Cities 51% 17.8 0.20% 0.72% 9.27%
No Rail pCities value
35% 17.0 0.10% 0.88% 11.78%
0.037* 0.192 0.000* 0.774 0.018*
Table 6: Median values and statistical significance for economic indicators in strong, weak and no rail cities
The other three economic items in table 6 refer to how much of the GDP of the cities is spent on investing in their public transport and road systems and how much of their GDP they spend on passenger transport as a whole (both public and private transport operating and investment costs from all sources). The patterns are quite clear and statistically significant: the more rail-oriented the cities, the greater proportion of their GDP goes back into investment in their public transport systems, and the lower is the overall cost to the society of running the entire passenger transport system (9.0% of metro GDP in SRCs compared to 11.8% in NRCs). The cities with rail also spend less of their GDP on road investment, but the overall differences in the median values between the groups of cities is not statistically significant on this factor because of the virtually identical result between the SRCs and WRCs. In summary, the economic data suggest that in this sample of developed world cities
, those where rail is a strong feature have greater wealth and more costeffective urban transport systems overall. They are also investing more in the quality of their public transport systems. Such cities would appear to be wasting less economic resources on passenger transport functions and on this factor are
therefore likely to be more competitive economically than cities which sink a higher proportion of their wealth into transport functions. Environmental factors Transport systems produce a range of environmental impacts
, taken here to include energy use and deaths attributable to transport accidents. Table 7 highlights the relatively favourable position of the more strongly rail-based cities in minimising these impacts. Per capita use of energy in private passenger transport increases as cities become less rail-oriented, with the NRCs being 144% higher in this factor than the SRCs. Because the SRCs and the WRCs do not vary very much in their median values, the overall differences in the medians are not statistically significant, even though there is this clear difference in private transport energy use between cities that have rail and those that don't (as there was with car use in table 5). Per capita generation of local smog producing emissions from transport (nitrogen oxides, carbon monoxide, sulfur dioxide
and volatile hydrocarbons) is also much higher in the NRCs than in the SRCs (100% higher). The pattern of decreasing per capita transport emissions is quite
World Transport Policy & Practice___________________________________________________ 30 Volume 14. Number 2. July 2008
systematic as the strength of rail increases, though the result falls a fraction short of statistical significance at the 90% confidence level. The spatial intensity of smog emissions also rises slightly the less rail-oriented the cities become, but the results fall far short of any statistical significance (the median value for the NRCs is only 6% higher than the SRCs). Finally, the costs incurred through transport-related accidents in cities are significant, especially the loss of life. The data in table 7 reveal a consistent and
statistically significant pattern of increasing transport deaths as the cities become less rail-oriented and of course less public transport-oriented as a whole. This is true both for per capita transport deaths, which are 129% higher in the NRCs than in the SRCs, and also deaths per billion passenger kilometres, which are 58% higher. It would appear that the more rail-oriented cities become, the less exposure there is to the risk of death from transport causes, even though the use of the riskier non-motorised modes also increases with greater rail orientation.
Environmental indicators Private passenger transport energy use per capita (MJ) Total transport emissions per capita (NOx, CO, SO2, VHC: kg) Total transport emissions per urban hectare (kg) Total transport deaths per 100,000 people Total transport deaths per billion passenger kms
Strong Rail Cities 16,381 96 3,538 5.8 6.4
Weak Rail Cities 17,197 114 3,663 7.8 8.0
No Rail Cities 39,951 195 3,753 13.3 10.1
pvalue 0.317 0.105 0.692 0.000* 0.017*
Table 7: Median values and statistical significance for environmental indicators in strong, weak and no rail cities
In summary rail systems, through their capacity to reduce car use and enhance public transport and non-motorised mode use, are associated with cities that use lower energy for passenger transport and generate lower local emission loads and transport deaths, both on a per capita and per passenger kilometre basis. Discussion The findings in this study are in line with extensive and detailed work by Hass-Klau et al (2003), Hass-Klau et al (2004) and Hass-Klau and Crampton (2002), which has demonstrated the many system-wide benefits in European cities of having Light Rail Transit (LRT) systems compared to only having bus systems, including
busways. These benefits include higher public transport patronage, which was also found in this international study
, but also a wide range of benefits in other factors, which were not examined in this study, but which help perhaps to understand the favourable results found for rail modes in this international comparison. Even though their work refers specifically to LRT systems, some of the findings are likely to be extendable to rail systems in general. Some of their key findings were: LRT requires the least width of corridors busways require most width. LRT normally transports more
World Transport Policy & Practice___________________________________________________ 31 Volume 14. Number 2. July 2008
passengers per hour than
standard buses. Noise and pollution are lowest
Running comfort is best with LRT LRT is better in overall urban
LRT and busways are very similar
LRT vehicles cost much more but
have the longest life expectancy LRT is slightly cheaper than
buses, on a whole-life basis for
similar levels of service. Complementary measures are
critical to the success of public
transport (parking cost and
availability, land use policies,
pedestrianisation, urban design)
complementary measures in order
to reach their maximum potential.
Complementary measures are
easier to implement with LRT and
important to do in all transport
projects to maximise the benefits
of the investment.
Political and psychological factors
related to different transport
considerations e.g. successful
pedestrianisation schemes are
strongly linked to implementation
of LRT systems.
Under equal conditions people
prefer to use LRT than to use
There are a higher percentage of
higher income groups using light
rail than buses (e.g. in Calgary,
Canada). LRT has a strong potential
following among car users, even
in cities with no recent
experiences of LRT or trams.
The study by Litman (2004) comparing 130 US cities with and without rail concluded that those with significant rail systems have: Lower per-capita traffic congestion costs. Lower per-capita traffic fatalities. Lower per capita consumer transport expenditures. Higher per capita public transport ridership. Higher public transport commute mode split. Lower public transport operating costs per passenger-mile. Higher public transport service cost recovery. Of the above factors that were examined in this international study, the results were similar. The Litman study found that residents in cities with large, wellestablished rail systems enjoy about half the per capita traffic congestion delay as people who live in comparable size cities that lack rail. The reason for this is in line with the findings in this international study that people in cities with rail systems enjoy lower per capita annual vehicle kilometres whilst also having an effective alternative when travelling on the most congested corridors. Litman (2004) also found that US cities with large rail systems have about a third lower per capita traffic fatality rates. Residents of the strong rail cities also save approximately $US450 annually per capita in transport costs compared with residents of cities that have no rail systems. The study concluded that rail system service costs are repaid several times over by reduced congestion, road and parking facility costs, reduced traffic accident costs, and consumer cost savings. Such findings are in line with the
World Transport Policy & Practice___________________________________________________ 32 Volume 14. Number 2. July 2008
observed comparative differences in this sample of high-income cities around the world that have rail systems (e.g. lower CBD parking, lower transport deaths, a lower proportion of metropolitan GDP being spent on passenger transport, better cost recovery for public transport, higher public transport use). Rail also has important impacts on urban form in terms of its capacity to increase densities and consolidate both residential and mixed use development around centres or nodes or along corridors. The positive land use impacts of urban rail and their transport flow-on effects are partly responsible for the urban system benefits outlined in this paper. Nodes of development are easier to service with public transport (including bus systems), walking and cycling are more viable for more trips and a polycentric city based around rail stations can help to minimise urban sprawl. These aspects of urban rail and its city-shaping capacity are discussed in detail in other works (Vuchic 1981, Bernick and Cervero 1997, Cervero 1998, Laube, Kenworthy and Zeibots 1999, Newman and Kenworthy 1999a). Conclusions Any developed city wishing to build a better public transport system, to curb or reduce its automobile dependence and to become more environmentally and economically sustainable, should not ignore the potential benefits of building a strong rail backbone as the mainstay of the city's public transport system. The data in this paper point strongly to the idea that public transport systems based on buses alone cannot achieve the same positive urban system results across a wide range of factors as when rail systems assume a more significant role within the public transport system.
The mechanisms for the advantages of urban rail are complex. However, they appear to relate at least in part to the legibility of rail systems and the greater permanence of rail services
, the positive image of rail in the mind of the public and business community and people's willingness to use rail systems over buses for a variety of reasons, including more competitive travel speed and greater reliability and quality of service. None of this, however, diminishes the critical role that buses play in public transport systems. Buses are essential public transport providers to areas that simply cannot be served by rail and there are many such areas in most cities, and buses provide critical feeder systems into major sub-centres and into rail systems. Well-patronised urban rail systems are usually associated with strong and healthy levels of bus use (Kenworthy and Laube 2001). Where network structures are well devised and services well coordinated, rail and bus are highly complementary and are not in competition with each other, but rather form an integrated, multimodal public transport system that provides competition with the car. Finally, the arguments and research put forward in this paper should not be read or construed in terms putting one mode of public transport above another merely for the sake of it. This is clearly not productive since the best public transport systems emerge out of choosing the right mode for the right task for the multitude of situations in any city. Public transport should be seen as a multi-modal system whose chief aim is to compete with and reduce dependence on the car, building a `virtuous circle' rather than a cycle of decline, which has tended to be the story of public transport in so many cities over
World Transport Policy & Practice___________________________________________________ 33 Volume 14. Number 2. July 2008
the last decade (Kaufmann 2000). Rather, what the paper has shown is that urban systems, whether in auto-dependent North America
or Australia, more transitoriented Europe, or the wealthier parts of Asia, do seem to gain multiple benefits from developing public transport systems that are anchored and shaped primarily by fixed-track modes, the vast majority of which are rail systems, in one form or another. This then forms the basis for a superior overall public transport system, utilising rail modes, buses and in some cases ferries, which fills a much greater role in the city's transport system. Finally, it needs to be said that although the analysis in this paper is based on data from 1995 or 1996, the overall conclusions and patterns between the three groups of cities are unlikely to be altered were the analysis to be conducted using later data. In other words, the systematic differences in the various factors found between strong rail, weak rail and no rail cities are not ephemeral observations, but are based on strong structural differences between the cities, which reveal themselves repeatedly over long periods of time. A similar analysis was carried out with 1990 data on a more limited set of cities listed in Newman and Kenworthy (1999a). The same systematic patterns of variation between the rail cities and no rail cities emerged on the same variables. In addition, the author has begun the update of data on some cities, especially in the USA and the completed transit data for 2005 shows that the US cities with no rail, such as Phoenix, continue to stagnate in transit use with only an 11% increase in annual boardings per capita from an extremely low level of 15.1 trips per capita up to 16.8 (virtually the lowest in
the world). Phoenix is building a LRT system at this moment. Likewise Houston declined slightly in transit use over the 10 year period and has finally voted to build an extensive LRT system. Los Angeles in the mean time has been aggressively growing its rail system (light rail, metro and commuter rail) and has achieved the highest growth rate in transit use of all the US cities studied (39%, up from 49.1 boardings per capita in 1995 to 68.3 in 2005). New York, the most rail-oriented of the US cities, was the other big transit winner with a 28% increase in transit use from 131.5 boardings per capita to 167.7 per capita, the bulk of which came from the NY underground. Thus more recent data are tending in the direction of reinforcing the patterns observed in this paper, so that the ageing nature of the data used do not undermine the policy value of the results and conclusions. Acknowledgement The author wishes to acknowledge the very significant contribution of Dr Felix Laube, co-author of the Millennium Cities Database for Sustainable Transport, in collecting and testing the data that lie behind the analyses in this paper. The author also wishes to gratefully acknowledge work of Mrs Monika Brunetti in the compilation of the 2005 US transit use data referred to in this paper. References Ang, B.W., 1990, Reducing traffic congestion and its impact on transport energy use in Singapore. Energy Policy, 18 (9), 871-874. Ang, B.W., 1993, An energy and environmentally sound urban transport system: the case of Singapore. International Journal
of Vehicle Design, 14 (4).
World Transport Policy & Practice___________________________________________________ 34 Volume 14. Number 2. July 2008
Badami, M.G., 2005, The urban transport challenge in India: Considerations, implications and strategies. International Development Planning Review. 27 (2), 169-194. Bernick, M. and Cervero, R., 1997, Transit villages in the 21st century (New York, U.S.A.: McGraw Hill
). Bonsall, J., 1985, A bus for all seasons. Presented to Seminar on The Canadian Experience: Making Transit Work in the Golden Gate Corridor, San Rafael, California, U.S.A., October 3. Cervero, R., 1995, Sustainable new towns: Stockholm's rail served satellites. Cities, 12 (1), 41-51. Cervero, R., 1998, The transit metropolis: A global inquiry (Washington DC
, U.S.A.: Island Press). Gomez-Ibaсez, J. A., 1991, A global view of automobile dependence. Journal of the American Planning Association, 57 (3), 376-379. Hass-Klau, C. and Crampton, G. (2002) Future of urban transport learning from success and weakness: Light rail (Brighton, U.K.: Environmental and Transport Planning). Hass-Klau, C., Crampton, G. and Benjari, R., 2004, Economic impact of light rail: The results of 15 urban areas
in France, Germany, UK and North America (Brighton, U.K.: Environmental and Transport Planning). Hass-Klau, C., Crampton, G., Biereth, C. and Deutsch, V., 2003, Bus or light rail: Making the right choice - A financial, operational and demand comparison of
light rail, guided buses, busways and bus lanes (Brighton, U.K.: Environmental and Transport Planning). Henry, L., 1989, Ridership forecasting considerations in comparisons of light rail and motor bus modes. In: Light Rail Transit: New system successes at affordable prices, Transportation Research
Board, Special Report 221, (Washington DC, U.S.A.: Transportation Research Board
), pp. 163-189.
Kain, J. and Liu, Z., 1999, Secrets of
success: Assessing the large increases in
transit ridership achieved by Houston and
Transportation Research A 33, 601-624.
Kaufmann, V., 2000, Modal practices: From the rationales behind car and public transport use, to coherent transport policies. Case studies in France and Switzerland. World Transport Policy and Practice, 6 (4), 8-17.
Kenworthy, J.R., 1995, Automobile dependence in Bangkok: An international comparison with implications for planning policies. World Transport Policy and Practice 1 (3), 31-41.
Kenworthy, J.R. and Laube, F.B., 1999, Patterns of automobile dependence in cities: An international overview of key physical and economic dimensions with some implications for urban policy. Transportation Research A 33, 691-723.
Kenworthy J, and Laube F., 2001, The
Millennium Cities Database for
Belgium: International Union of Public
Transport, UITP and Perth, Australia:
Institute for Sustainability and Technology
Policy, ISTP), CD ROM database.
World Transport Policy & Practice___________________________________________________ 35 Volume 14. Number 2. July 2008
Kenworthy, J., Laube, F., Newman, P. and Barter, P.,1997, Indicators of transport efficiency in 37 cities. Report to World Bank
, Institute for science and technology
Policy, Murdoch University, Western Australia.
Kenworthy, J. and Laube, F. [with Newman, P., Barter, P., Raad, T., Poboon, C. and Guia, B. (Jr)], 1999 An international sourcebook of automobile dependence in cities 1960-1990 (Boulder, U.S.A.: University Press of Colorado).
Laube, F. B., 1998, Optimising urban
passenger transport: Developing a least-
cost model based on international
comparison of urban transport cost,
infrastructure, environmental and best
practice in public transport. PhD thesis
Murdoch University, Perth, Western
Laube, F.B., Kenworthy, J.R. and Zeibots,
M.E., 1999, Towards a city science: City
observation and formulation of a city
theory. In: Siedlungsstrukturen, rдumliche
Mobilitдt and Verkehr: Auf dem Weg zur
nachhaltigkeit in Stadtregionen, edited by
M. Hesse (Berlin, Germany: Institut fьr
Strukturplanung), pp. 99-118.
Lave, C., 1992, Cars and demographics. Access, 1, 4-11.
Litman, T.A., 2004, Rail transit in America: Comprehensive evaluation of benefits. http://www.vtpi.org/railben.pdf
Neff, J.W., 1996, Substitution rates between transit and automobile travel. Paper presented at the Association of American Geographers' Annual Meeting, Charlotte, North Carolina
Newman, P. and Kenworthy, J., 1989, Cities and automobile dependence: An international sourcebook (Aldershot, U.K.: Gower). Newman, P. and Kenworthy, J., 1991, Towards a more sustainable Canberra: An assessment of Canberra's transport, energy and land use. Institute for Science and Technology Policy, Murdoch University, Perth. Newman, P. and Kenworthy, J., 1999a, Sustainability and cities: Overcoming automobile dependence (Washington DC, U.S.A.: Island Press). Newman, P. and Kenworthy, J., 1999b, `Relative speed' not `time savings': A new indicator for sustainable transport. Papers of the 23rd Australasian Transport Research Forum, Perth, Western Australia, 29 September 1 October, Volume 23, Part 1. pp. 425-440. Pickrell, D. H., 1990, Urban rail transit projects: Forecasts versus actual ridership costs (Cambridge, U.S.A.: US Department of Transportation). Pucher, J., 2002, Renaissance of public transport in the United States? Transportation Quarterly 56 (1), 33-49. Pucher, J. and Dijkstra, L., 2003, Promoting safe walking and cycling to improve public health: Lessons from The Netherlands and Germany. American Journal of Public Health, 93, (9), 15091516. Thomson, J.M., 1978, Great cities and their traffic (Harmondsworth, U.K.: Penguin).
World Transport Policy & Practice___________________________________________________ 36 Volume 14. Number 2. July 2008
Vuchic, V. R., 1981, Urban public transportation: Systems and technology (Englewood Cliffs, New Jersey, U.S.A.: Prentice-Hall). Author contact information Jeff Kenworthy, Curtin Sustainability Policy Institute (CUSP), Curtin University of Technology, Kent Street, Bentley, WA, 6102. Current Tel No. in Frankfurt: 49 69 1533 2753 Email: [email protected]
Authors Bionote Jeff Kenworthy is Professor in Sustainable Cities in the Curtin University Sustainability Policy (CUSP) Institute at
Curtin University of Technology in Perth,
Western Australia. He is co-author with
Peter Newman of Sustainability and
Dependence (Washington DC: Island
Press, 1999) and principal author with
Felix Laube of An International
Sourcebook of Automobile Dependence in
Cities, 1960-1990 (Boulder: University
Press of Colorado, 1999) and The
Millennium Cities Database for
Sustainable Transport (Brussels and
Perth: UITP and ISTP, 2001). He has
worked in the areas of urban transport
systems and urban planning for 28 years
and is author of some 200 publications in
World Transport Policy & Practice___________________________________________________ 37 Volume 14. Number 2. July 2008