Travel Time and Distance in International Perspective: A Comparison between Nanjing (China) and the Randstad (The Netherlands)

Abstract

While Western countries are trying to reduce car dependency on the back of low carbon objectives, the ownership and use of private cars in urban China is increasing dramatically. In this paper, light is shed on both developments through a comparative study of the travel behaviour in two regions with a very different built environment: Nanjing, China, and the Randstad in the Netherlands. Controlled for car ownership, daily travel time and distance are analysed in both regions. The results indicate that, in the case of Nanjing, the suggestion is that the configurations of current land use which support walking and cycling should be preserved as much as possible and that, in the meanwhile, investments should be made in fast public transport to facilitate economic developments. As regards the Randstad, it would seem wise to promote the use of walking and cycling by continuing to encourage compact land use patterns in combination with relatively fast public transport developments.

1. Introduction

Low carbon economies and cities are high on both policy and research agendas world-wide (Worldwatch Institute, 2009). Estimates of 13 per cent greenhouse gas emissions and 23 per cent energy-related CO₂ emissions caused by transport (EEA, 2009) automatically generate attention for sustainable urban transport.

Private car use is largely responsible for the increase in transport-related energy use and emissions. Although future technological developments could lead to a change in emphasis, these days car ownership is still an important indicator of the sustainability of transport systems. In 1985, there were approximately 285 000 private vehicles on the roads in China. This number skyrocketed to 59.4 million in 2010, an increase by a factor of 208 in 25 years. Despite this rapid process of motorisation, the Chinese car ownership ratio is still low in comparison with Western countries. In 2007, the number of cars per 1000 persons was 9 in China, 457 in the Netherlands and 740 in the US (Ng et al., 2010). This low level of motorisation is also reflected in data on daily car use. Only 10 per cent of all trips in Chinese cities are made by car and non-motorised transport and public transport are still the dominant travel modes (Ng et al., 2010). Walking and cycling do account for 60 per cent of all trips in Chinese cities (TSW, 2009).

The rapid expansion of car ownership could dramatically change this situation and potentially threaten the sustainability of urban transport in China. However, not much literature is available on travel behaviour in China. A comparison of travel behaviour in Chinese and Western cities is valuable for theoretical reasons. Muller (1995/2004) introduced a four-stage model of transport in US cities in which he linked the dominant transport mode in an era (like walking, horse-drawn vehicles, electric streetcar and automobile) with the urban form of the metropolitan area. However, from an analytical-empirical perspective, this relationship has not yet been proven, mainly due to the lack of disaggregated data. In the late 1960s, when travel diary data became available, Western cities had already entered the era of car dominance for metropolitan travel patterns. By contrast, there is still a substantial representation of non-motorised transport modes in the modal split in Chinese urban areas, despite the rapid increase in car ownership and use and related changes in metropolitan areas. As far as Nanjing is concerned, the combined share of walking and cycling is still 66 per cent, compared with 38 per cent in the Randstad (2008). Hence, a comparative analysis of travel behaviour in both metropolitan areas provides a unique opportunity to develop a better understanding of interdependencies between transport modes and urban form in different stages of metropolitan development.

This comparison is also interesting for policy reasons. Despite the contextual differences, an analysis of the use of sustainable transport modes in Chinese cities can provide insights which are useful for the organisation of daily life in future low carbon cities. In addition, a comparison with a car-based metropolitan system will provide Chinese policy-makers with a better understanding of the implications of increased motorisation for travel behaviour and urban form in their metropolitan areas.

Taking these aims as a basis, this paper compares the travel behaviour of residents in Nanjing (China) and in the Randstad (Netherlands). We focus our comparison on two characteristics of travel behaviour: individual daily travel time and distance. Daily travel time represents an indicator of the mobility of people pursuing their daily lives. In general, daily travel distance is a reasonable proxy for transport emissions, energy consumption and a number of other environmental impacts of transport, especially for cities with transport systems dominated by car and public transport (Stead, 1999). In order to understand the differences regarding these characteristics of travel behaviour, we ask the following research questions. What are the differences in daily travel time and distance in Nanjing, China, and the Randstad, the Netherlands? Which factors determine these characteristics in Nanjing and the Randstad? What does this suggest about travel behaviour in the future? In view of the expected impact of car ownership on activity and travel behaviour, a distinction is made between car owners and non-car-owners. The data for this comparison come from two different surveys conducted in 2008. These are the Nanjing Residents Travel Survey (NRTS) and the Netherlands Mobility Survey (MON).

The research questions are answered in the five sections following this introduction. Section 2 summarises the academic literature on determinants for travel time and distance. In section 3, we present the characteristics of the datasets and analytical methods, followed by a comparative description of travel time and distance in Nanjing and the Randstad (section 4). Section 5 presents information about determinants for these factors through a number of regression models. Finally, section 6 concludes and discusses main results of the analyses.

2. Literature on Travel Time and Travel Distance

There is an extensive body of literature on factors that determine travel time and distance. In general, the car is identified as an important facilitator for individual and household travel behaviour (Giuliano and Narayan, 2003). Car ownership is determined by many factors that also determine travel behaviour in general (Giuliano and Narayan, 2003; Potoglou and Kanaroglou, 2008): socio-demographics (for example, income, household composition, employment, age, gender), attributes of the transport provision (for example, purchase costs of cars, relative costs of car and transit use, access to transit) and attributes of the built environment (for example, density, diversity and design). In particular, households with a high number of active and senior adults, workers, high incomes and homeownership have higher levels of car ownership, while single-parent households, one-person households, households in high-density or mixed land use areas and households with good transit access tend to own fewer cars. Næss (2005) also observed that car ownership is also determined by residential location relative to the city centre: the greater the distance from the city centre, the higher the car ownership ratio.

Bhat and Guo (2007) argue that household demographics have more dominant effects on car ownership than built environment characteristics. In many analyses, income is considered the most important determinant for the number of cars in a household, with some transport economists arguing that per capita income is the single most important explanatory factor for car ownership and use (Ingram and Liu, 1999). As regards the two main indicators for our own approach (travel time and distance), results show that both increase in conjunction with household car ownership (Dieleman et al., 2002; Schwanen et al., 2002).

Besides car ownership, the literature discerns two main groups of determinants for individual travel behaviour: built environment characteristics and socio-demographic variables (Stead, 2001). People travel distances to participate in various economic and social activities. By determining the distance between the locations at which these activities take place, and by providing various modes of transport, the built environment provides a set of conditions that encourages some types of travel behaviour and discourages others (Næss, 2010). Studies focusing on the effects of the built environment on travel demand have analysed the impact of numerous attributes like the snone of the municipality or region, density and the land use mix (Dieleman et al., 2002; Schwanen et al., 2002; Pan et al., 2009; Ewing and Cervero, 2010). Here, we limit our discussion to variables available in our datasets.

Various authors have studied the influence of residential density on both travel time and travel distance. Residential density is usually regarded as a proxy for local accessibility to goods and services (Giuliano and Narayan, 2003), while higher densities often result in shorter trips and less overall travel distance (Stead, 1999; Næss, 2005). However, the influence on travel time is ambiguous in that shorter distances in areas with high density overlap with lower travel speeds resulting from increasing congestion and higher shares of walking and cycling trips. Levinson and Kumar (1997) suggest a threshold of between 7500 and 10 000 persons per square mile for residential density at which the decrease in distance is overtaken by the congestion effects, resulting in the shortest car commute times.

The distance of place of residence from the urban centre is another built environment characteristic which features frequently in analyses. In general, city centres are concentrations of various facilities served by good public transport links. Closeness to the city centre increases the likelihood of residents travelling shorter distances and making greater use of public transport (Næss, 2010). Gordon et al. (1989) show that travel distance for both work and non-work journeys is generally lower for central-city residents. Likewise, Næss (2005) identifies a direct link between an increase in travel distance and an increase in distance between home and the urban centre. He also finds that the distance between home and the urban centre influences travel times, but that this influence is weaker, perhaps due to lower travel speeds in central areas because of congestion and higher mode shares of walking and cycling trips.

Various studies investigate the relationship between land use mix and travel distance and time. From a theoretical perspective, facilities are distributed relatively close to each other in areas with highly mixed land use. Residents in these areas are expected to travel shorter distances to access these facilities. Cervero (1996) concludes that a balanced job–housing ratio in the San Francisco Bay Area leads to relatively short commuting distances and times. Wang and Chai (2009) find similar results in Beijing. However, Schwanen et al. (2002) show that a quantitative balance does not necessarily result in a qualitative balance between types of jobs and homes in a particular area. Cervero and Kockelman (1997) reveal a negative relationship between land use mix and non-work travel distance, although total travel distance is unaffected by land use mix.

In general, it can be concluded that the relationship between built environment and travel distance and time is highly complex. Different characteristics of the built environment—such as density, land use mix and urban structure—are interwoven and have a composite impact on travel behaviour (Dieleman et al., 2002). In addition, socio-demographic characteristics of individuals are strong determinants for travel distance and time. Many studies have evaluated the impact of socioeconomic characteristics on travel behaviour by focusing on gender, age, income, education, occupation, household car ownership and household structure. Compared with impacts of built environment variables, socioeconomic characteristics might play an even more important role as regards travel patterns (Stead, 2001).

This brief review of literature on the impact of built environment attributes on travel behaviour highlights the potential value of a comparison of travel time and distance in China and the Netherlands. The expectation is that the dominance of walking and cycling infrastructures and compact spatial structures will have a reducing impact on travel time and especially travel distance in Nanjing. These differences will be larger for non-car-owners than for car owners.

3. Research Design

Nanjing is the capital of Jiangsu Province and the second-largest commercial centre in the Yangtze River Delta. It covers an area of 4723 square kilometres and was home to a population of 6.24 million in 2008 (Nanjing Statistics Bureau, 2008). Nanjing is a monocentric metropolitan area characterised by a hierarchical centre structure with a main centre and three sub-centres. Our data on travel behaviour in Nanjing result from the Nanjing Residents Travel Survey (NRTS; in Chinese: Nanjingshi jumin chuxing chouyang diaocha). The NRTS survey was conducted on all the Tuesdays, Wednesdays and Thursdays in June 2008. During this month average temperatures varied between a very warm summer and a cold winter. Our sample contains 3894 cases from 1414 households in 10 urban districts. The sub-sample size of each district is proportional to its population.

The Randstad is the name of a horseshoe-shaped ring of cities in the west of the Netherlands, surrounding the ‘Green Heart’, a sparsely populated green area in the centre (see Figure 1). The region has 6.7 million inhabitants in an area covering 6400 square kilometres (Regio Randstad, 2004). The Randstad is a polycentric urban region with four big cities specialised in different economic sectors. Amsterdam (population 991 000) is a centre of advanced producer services (for example, business and financial services); Utrecht (640 000) is an important commercial centre for the domestic service economy; Rotterdam (1.4 million) is home to the main Dutch seaport; and The Hague (810 000) is the seat of the national government and many international organisations (Regio Randstad, 2004). Our data on travel behaviour in this region are taken from the Netherlands Mobility Survey (in Dutch: Mobiliteitsonderzoek Nederland—MON) performed in 2008. This is a continuous survey of daily travel behaviour of a random sample of residents of the Netherlands. The survey is conducted monthly using individual travel diaries for one day. In 2008, 40 000 respondents participated in the survey. However, for comparability reasons, we carried out data handling procedures. First, we selected residents of the four main Randstad cities from the MON respondents. Secondly, we excluded the winter months of December, January and February and the relatively warm, summer holiday months of July and August, leaving March, April, May, June, September, October and November in the dataset. Thirdly, we excluded data collected during weekends.

Figure 1.

Nanjing in China and the Randstad in the Netherlands.

Although the two areas are comparable in population size and land area, there are substantial differences in spatial structure, social structure and development context. Nanjing is a monocentric metropolitan area, while the Randstad is a polycentric urban region. Nanjing is less advanced economically than the Randstad, as reflected in much lower levels of income and car ownership. With regard to the urban development context, Nanjing is still at the beginning of the process of massive motorisation and public transport is insufficiently developed, with only one metro line in 2008. The Randstad is characterised by massive motorisation and relatively good public transport. In particular, there is an advanced intercity railway system. In addition, the degree of economic specialisation in the Randstad is also higher than that of Nanjing.

Daily travel time and distance per person per day are the dependent variables of our analysis. Figure 2 shows the socio-demographic and built environment variables included in the data. Distance travelled in Nanjing is calculated as the linear distance between the geometric centres of origin and destination traffic analysis zones (TAZs) of trips. For trips within a TAZ, we assume a travelled distance of two-thirds of the diameter of the TAZ. In the Randstad, we calculate distances in a similar way, with the four-digit postal code zone as the basic unit. In view of the expected impact of car ownership on travel behaviour, and to understand the differences in organising daily travel, a distinction is made between car owners and non-car-owners. However, this distinction masks effects of residential location on car ownership and therefore leads to an underestimation of impacts of the built environment on travel.

Figure 2.

Variables used in the analysis.

We identified five transport modes in Nanjing: walking, cycling (including e-bike), public transport (bus and metro), private car and other modes (mainly motorcycle, taxi). With only one metro line in operation in 2008, public transport mainly means local bus travel. Due to the nature of (mostly intracity) trips in Nanjing, the train is not a travel alternative here. By contrast, trains are an important mode of transport for intercity trips in the Randstad. To ensure more or less similar public transport categories in both areas, train trips have been included in the ‘other’ category. For the Randstad, we therefore use walking, cycling, public transport (bus, tram and metro), private car and other (motorcycle, moped, taxi and train) as travel modes. Four trip purposes are identified—namely, commuting, shopping, leisure (social visits, sports, culture and recreation) and ‘other’ (the latter combining work-related travel, trips for studying and picking up family in Nanjing and work-related travel, trips for education and training, picking up family and other trip purposes in the Netherlands).

The socio-demographic variables in our models are gender, age, education, household income, employment status, number of workers in the household and household type. As part-time employment is virtually non-existent in the Nanjing sample, employment status here includes only two categories: non-worker and worker. In the case of the Randstad, we differentiate between non-workers, part-time workers (12–30 hours a week) and full-time workers. The household typology is based on a combination of household size and presence of children (<20 years old). For the Randstad, we use three household types: single, couple and family. In view of the Chinese tradition of co-residence of parents and married children, we distinguish between three additional categories in Nanjing: ‘adult family’ (households without children but more than two adults), ‘extended family’ (three generations or more) and ‘other family’ (households with single parents and/or single grandparents).

The built environment attributes included are population density, distance to the urban centre and land use mix. Population density is calculated as the ratio of population and area of the sub-district in Nanjing and the postal zone in the Randstad. Distance to urban centre in Nanjing is defined as distance between the respondent’s home and the main city centre (‘xin jie kou’) in Nanjing. In the case of the Randstad, we calculate the shortest distance between home and the closest of the four big city centres. Land use mix simultaneously accounts for the variety and prevalence of different functions in the area. Following Cervero and Kockelman (1997), we calculated an entropy index

s = - \sum_{j} \frac{[P_{j k} \times ln (P_{j k})]}{ln (J)}

In this equation, S refers to land use mix (entropy); j is the type of land use (j = 1, 2, …, J); k the TAZ of Nanjing city or the postcode zone in the Randstad (k = 1,2,…, K); P_jk the proportion of land use j within the TAZ or postcode zone.

The entropy ranges from 0 (homogeneity—only one type of land use) to 1 (heterogeneity—shares of uses evenly distributed over all land use categories). We include six land use types with highest relevance for residents’ daily activities: residential, commercial, public, industrial, offices and research sites, and parks and recreational use.

4. Descriptive Comparisons of Travel Time and Travel Distance

4.1 Travel Time in Nanjing and the Randstad

As Table 1 shows, people in Nanjing travel approximately 40 minutes per day. The average travel time in the Randstad is 70 minutes per day, matching the average travel time per day in Western cities (Schafer and Victor, 2000). The relatively small average daily travel time in Nanjing seems to be at odds with the time budget hypothesis. As observed by Schafer and Victor (2000), the travel time budget is constant on average, but large variations exist for cities and small populations. Differences in sizes of cities, congestion levels and socio-demographics might cause large deviations from the average. In this respect, Nanjing is very different from the Randstad. Another reason could be that walking and cycling, which are the dominant transport modes in Nanjing, are easily underestimated in datasets due to the short related travel distances and times. The travel time in the Randstad by public transport is 1.5 times and by ‘other’ modes about 4 times the figure for Nanjing. This could be caused by the inclusion of train trips in the ‘other’ category in the Randstad, with intercity trips as the main contributor to longer travel times because of long distances and additional waiting times for transfers.

Table 1.

Average daily travel time and distance by users of transport modes, Nanjing and the Randstad

Variable	Area	Car ownership	Walking	Cycling	Public transport	Cars	Other	Total
Travel time (in minutes)	Nanjing	No car	27.2	29.8	57.9	47.5	28.2	42.6
		Car	24.5	30.8	52.9	38.5	27.7	40.7
		All users	26.9	29.9	57.4	40.7	28.2	42.2
	Randstad	No car	19.4	29.6	80.1	36.6	125.8	81.6
		Car	21.5	21.7	68.5	44.2	90.3	66.4
		All users	20.8	23.1	77.0	44.0	121.6	68.9
Travel distance (in km)	Nanjing	No car	1.9	3.1	6.2	7.1	3.9	4.3
		Car	1.7	3.1	5.8	6.1	4.6	5.1
		All users	1.8	3.1	6.2	6.4	4.0	4.4
	Randstad	No car	1.5	4.1	14.7	16.4	70.6	21.5
		Car	2.1	3.7	9.9	20.5	52.4	25.1
		All users	1.9	3.7	12.5	20.4	64.5	24.5

Note: In the table, the travel times and distances using different transport modes by the same person are assigned to those specific modes, respectively.

Source: own calculations.

Car ownership reduces daily travel times in both regions, but to a much larger extent in the Randstad. This might be due to the limited quality of relevant transport facilities in Nanjing (parking space, road provision and congestion), reducing differences in daily travel time between transport modes and making car ownership less attractive.

4.2 Travel Distance in Nanjing and the Randstad

Table 1 also documents substantial differences in daily travel distance between the two areas. While Nanjing residents travelled an average 4.4 kilometres per day, their counterparts in the Randstad covered 24.5 kilometres. This substantial difference mainly stems from the relative limited use of public transport and cars, which can be attributed to the low car ownership and insufficiently developed public transport system and the related urban structure in Nanjing. Larger travel distances in the Randstad could also be the result of more highly specialised labour and housing markets.

Our calculations with respect to travel purpose show that people in both regions travel longer for commuting than for other purposes. However, while leisure trips in Nanjing cover almost the same distance as commutes, these distances are very different in the Randstad. Shopping trips are the shortest trips in both regions.

Car ownership expands the range of daily travel distances in both areas. The lower degree of attractiveness of car ownership in Nanjing (for example, lack of parking space, bad road conditions and severe congestion) leads to a different role for the car in comparison with the Randstad. Since owning and using a car is, to some extent, considered a luxury in China (Li et al., 2010), the car mainly functions as a status symbol and less for convenience, as for example in the Randstad.

5. Multivariate Analyses

5.1 Models for Travel Time in Nanjing and the Randstad

Table 2 presents two models on travel time per person per day for car-owning and non-car-owning participants. In the case of Nanjing, both models possess a relatively low explanatory power. The model for car owners includes four and the model for non-car-owners eight significant variables. The limited number of significant variables in the car owner model can be explained by the considerable group homogeneity. The relatively small number of car owners in China share some distinct characteristics: their educational level and income are high; their age is between 30 and 50; and their household composition is comparable (Li et al., 2010).

Table 2.

Regression mode of total travel time per person per day, Nanjing and Randstad

Nanjing					Randstad
	Car		No car			Car		No car
	B	Beta	B	Beta		B	Beta	B	Beta
Gender (male = ref.)					Gender (male = ref.)
Female	−0.127**	−0.093	−0.031	−0.022	Female	−0.098	−0.048	0.301*	0.111
Age (20–29 = ref.)					Age (20–29 = ref.)
30–39	0.022	0.015	−0.074	−0.043	30–39	−0.456***	−0.173	−0.014	−0.004
40–49	0.124	0.079	−0.100**	−0.062	40–49	−0.335**	−0.139	0.224	0.058
50–59	−0.113	−0.063	−0.088*	−0.047	50–59	−0.422***	−0.165	−0.308	−0.093
More than 60	−0.240	−0.082	0.001	0.001	More than 60	−0.568***	−0.243	−0.376	−0.140
Education (lower = ref.)					Education (lower = ref.)
Middle	0.137	0.094	0.202***	0.140	Middle	0.125	0.057	0.091	0.029
High	0.168	0.121	0.275***	0.182	High	0.377***	0.175	0.339	0.114
Employment status (non-worker = ref.)					Employment status (non-worker = ref.)
Worker	−0.155	−0.088	0.076	0.048	Part-time worker	0.180	0.062	0.646	0.165
					Full-time worker	0.218*	0.106	0.828*	0.285
Workers in household	0.084	0.087	−0.024	−0.028	Workers in household	−0.136**	−0.125	−0.169	−0.086
Household income (EUR per year; more than 10,000 = ref.)					Household income (EUR per year; more than 50,000 = ref.)
Less than 2000	0.081	0.018	0.009	0.004	Less than 9999	−0.241	−0.024	−0.271	−0.050
2000–5000	−0.227**	−0.120	−0.004	−0.003	10 000–29 999	−0.319**	−0.137	−0.230	−0.082
5000–10 000	−0.082	−0.059	−0.053	−0.032	30 000–39 999	−0.268**	−0.110	0.085	0.026
					40 000–49 999	−0.184**	−0.075	0.113	0.015
Household type (family = ref.)					Household type (family = ref.)
Single	0.141	0.020	0.215**	0.042	Single	0.079	0.027	0.560*	0.209
Couple	0.155	0.075	0.076	0.038	Couple	0.032	0.016	0.168	0.056
Adult family	0.112	0.076	0.116**	0.078
Extended family	0.268**	0.139	0.035	0.015
Other family	−0.066	−0.013	0.011	0.003
Population density	−0.041	−0.074	−0.087***	−0.165	Population density	−0.048	−0.025	0.166	0.070
Distance to city centre	−0.017*	−0.092	−0.041***	−0.196	Distance to city centre	0.000	−0.008	0.000	−0.017
Land use mix	−0.179	−0.029	−0.170	−0.027	Land use mix	−0.439**	−0.066	0.070	0.009
(Constant)	3.614***		3.819***		(Constant)	4.365***		2.885***
R ²	0.035		0.058		R ²		0.072		0.140
N	529		2460		N		1099		223

Notes: * = a ≤ 0.10; ** = a ≤ 0.05; *** = a ≤ 0.001. Dependent variable is the natural logarithm of travel time (in minutes).

Source: own calculations.

As stated, there are four significant variables in the car owner model in Nanjing. The first of these is gender: women in car-owning households spend less time travelling than men. Car ownership per household is often limited to one car which is primarily used by the male adult family member, thereby causing increasing travel times for men. Secondly, one income category (2000–5000 Euro per year) significantly influences travel time. The highest income segment was chosen as a reference category, so the negative coefficient for this variable documents a decrease in travel time for the lower-income groups, which is in line with existing results (Wang and Chai, 2009). Thirdly, living in an extended family significantly increases travel times compared with traditional family. Apparently, the presence of an older family member is linked to longer travel times. The final significant variable in the model is distance to the city centre, which displays a reducing effect on travel time and contrasts with existing research in Western cities (for example, Næss, 2005). This difference might be explained by a decreasing number of trips made by people living further from the city centre, resulting from less developed and lower-quality public transport.

The non-car-owner model in Nanjing has slightly more explanatory power. Age is a significant variable with older people spending less time travelling (except for the oldest group). A higher level of education increases travel time and matches findings from existing research. Respondents of single and adult family households have significantly longer travel times than members of traditional families with children, potentially resulting from additional time constraints because of childcare.

Built environment factors have a substantial influence on travel time. Population density has a negative impact on travel time and leads to shorter travel times in areas with higher densities. As in the car owner model, increasing distance to the city centre reduces daily travel times. In general, both models demonstrate that factors influencing daily travel time of car-owning and non-car-owning respondents in Nanjing differ considerably, with car ownership clearly reducing the importance of built environment factors.

A comparison with the travel time models in Nanjing shows that the two Randstad models for travel times are reversed. Here, more significant variables are detected in the car-owner model than in the non-car-owner model. Explanatory reasons could be similar: non-car-owning households in the Randstad are a small group with rather homogeneous characteristics in terms of income, education, age and built environment factors, so only three significant variables were found. Full-time workers spend considerably more time travelling than non-workers. Respondents from single households also travel considerably more than family members and women in non-car households travel for longer on a daily basis than men.

The model for car owners in the Randstad shows expected outcomes. Younger respondents, respondents with a higher level of education or a higher income, full-time workers and members of households with a higher number of workers spend significantly more time on daily travel. Furthermore, land use mix has a negative impact on travel time. This is presumably caused by the larger variety of opportunities within short distances in areas with high levels of land use balance.

5.2 Models for Travel Distance in Nanjing and the Randstad

Table 3 presents the models on travel distance per person per day. As far as Nanjing is concerned the results for car owners and non-car-owners show clear similarities in significance of socio-demographic factors. In both sub-samples, women travel shorter distances than men, older people travel less than young people and respondents with a higher level of education tend to travel further than people with a lower level of education. In the case of household income and household type, the results differ. One category of household income is significant in the model for non-car-owners, where a higher income is linked to longer distances travelled. This could result from specialised jobs for people with higher incomes that cannot be found close to the home (Dieleman et al., 2002). In the case of respondents in car-owning households, income does not significantly impact travel distance. This can be explained by the generally high level of income that these households have with little variance between respondents.

Table 3.

Regression mode of travel distance per person per day, Nanjing and Randstad

Nanjing					Randstad
	Car		No car			Car		No car
	B	Beta	B	Beta		B	Beta	B	Beta
Gender (male = ref.)					Gender (male = ref.)
Female	−0.218**	−0.136	−0.120***	−0.075	Female	−0.271**	−0.101	0.239	0.076
Age (20–29 = ref.)					Age (20–29 = ref.)
30–39	−0.129	−0.076	−0.099	−0.052	30–39	−0.424**	−0.122	0.126	0.030
40–49	−0.037	−0.020	−0.186***	−0.102	40–49	−0.304**	−0.096	0.082	0.018
50–59	−0.343**	−0.163	−0.148**	−0.071	50–59	−0.581***	−0.173	−0.493	−0.127
More than 60	−0.622***	−0.182	−0.284***	−0.127	More than 60	−0.773***	−0.252	−0.549	−0.176
Education (lower = ref.)					Education (lower = ref.)
Middle	0.318**	0.186	0.238***	0.148	Middle	0.206**	0.072	−0.218	−0.060
High	0.381**	0.234	0.326***	0.192	High	0.512***	0.181	0.465	0.134
Employment status (non-worker = ref.)					Employment status (non-worker = ref.)
Worker	0.021	0.010	0.182***	0.102	Part-time worker	0.376**	0.100	0.800	0.176
					Full-time worker	0.473***	0.176	1.302**	0.386
Workers in household	0.077	0.068	−0.056**	−0.059	Workers in household	−0.193**	−0.135	−0.269	−0.118
Household income (EUR per year, More than 10 000 = ref.)					Household income (EUR per year, More than 50 000 = ref.)
Less than 2000	−0.158	−0.030	−0.072	−0.033	Less than 9999	−0.687*	−0.053	−0.200	−0.032
2000–5000	−0.198*	−0.089	−0.006	−0.004	10 000–29 999	−0.441**	−0.144	−0.277	−0.085
5000–10 000	−0.116	−0.071	0.042	0.023	30 000–39 999	−0.254**	−0.080	−0.040	−0.011
					40 000–49 999	−0.185	−0.057	0.006	0.001
Household type (family = ref.)					Household type (family = ref.)
Single	0.864**	0.104	0.192	0.033	Single	0.097	0.025	0.746**	0.239
Couple	0.241**	0.100	0.077	0.034	Couple	0.049	0.018	0.290	0.083
Adult family	0.054	0.031	0.084*	0.050
Extended family	0.317**	0.140	0.012	0.004
Other family	−0.324	−0.055	−0.003	−0.001
Population density	−0.048	−0.074	−0.145***	−0.247	Population density	−0.194**	−0.078	−0.158	−0.057
Distance to city centre	0.007	0.031	−0.024***	−0.102	Distance to city centre	0.000	0.039	0.000	0.051
Land use mix	−0.534*	−0.074	−0.435**	−0.061	Land use mix	−0.348	−0.040	0.143	0.016
(Constant)	8.38***		8.638***		(Constant)	9.856***		7.964***
R ²		0.115		0.111	R ²		0.124		0.195
N		529		2460	N		1099		223

Notes: *a≤0.10; **a≤0.05; ***a≤0.001. Dependent variable is natural logarithm of travel distance (in km).

Source: own calculations.

Employment only plays a significant role for people without a car. Workers in this category cover noticeably greater distances than non-workers. By contrast, an increase in the number of workers in the household leads to a decrease in daily travel distance per person. While the first variable directly displays an individual characteristic and documents longer overall trip distances for employed participants, the presence of more than one employed person in the household leads to a lower per capita travel distance due to the sharing of household tasks, such as shopping and chauffeuring children to/from school (Dieleman et al., 2002). In the non-car model, there is a slight increase in daily travel distance for respondents in adult families compared with traditional families. In car-owning households, the presence of children (in the reference category) might stimulate their parents to use local services, such as school and local shopping and leisure facilities, which would then lead to shorter travel distances in comparison with single person and couple households (Naess, 2005; Schwanen et al., 2002). Although there are also children present in extended families, sharing childcare tasks with more adult household members reduces constraints and enables larger travel distances. The effects of children in the family are much more explicit in households with cars than in non-car households. Car ownership therefore seems to amplify differences in travel distance.

Compared with relatively weak effects on travel time, built environment characteristics have much stronger impacts on travel distance. In the non-car model, all built environment variables are significant. Population density impacts travel distances negatively, perhaps as a result of there being more opportunities in close proximity in higher-density areas. With regard to the residential location, it seems counter-intuitive that people living far away from the city centre travel shorter distances than people living close to the city centre (for example, Næss, 2010). However, our analysis here only covers non-car-owners. Due to relatively poor public transport provision in residential areas situated at large distances from the city centre, people in this group have to rely more on local facilities than people living near the city centre. In both sub-samples, land use mix also has a significant negative relationship with travel distance. Since this is the only significant built environment characteristic in the car owner model, it confirms the marginalising effect of car ownership on the relevance of built environment characteristics for travel distances.

Finally, we conducted regression analyses for travel distances in the Randstad. Here, the model for non-car-ownership contains only two significant variables. In non-car households, full-time workers travel significantly further than non-workers. Likewise, people in single households travel further than family members. The model for car owners offers considerably more significant variables. Socio-demographic determinants resemble the ones for travel time: educational level and employment status increase travel distance, while age and numbers of workers in the household decrease it. Furthermore, men travel further than women and respondents with higher incomes travel much further than people with lower incomes. In addition, respondents in densely populated areas travel shorter distances than in less dense areas. This is in line with the existing results (for example, Giuliano and Narayan, 2003; Næss, 2005). However, in contrast to Næss’s (2005) findings for Copenhagen, where distance to city centre is positively related to travel distance and time, this variable shows no significant impact in our study. One reason might be the different forms of infrastructure and related facility distributions: the radial infrastructure in Copenhagen potentially leads to weak connections between axes, thereby emphasising the importance of the city centre. In contrast, the Randstad polycentric form fosters tangential distributions of infrastructure and reduces the importance of inner cities (Geurs and van Wee, 2006).

6. Conclusions and Discussion

In order to capture differences in travel time and distance for inhabitants of Nanjing and the Randstad, and to answer our research questions, we made a distinction between car owners and non-car-owners in our analyses. All models for daily travel distance explain the observed variance better than for daily travel time, as a result of the far more complex nature of travel time. It is influenced by travel mode (and its speed), type of road network (for example, motorways or roads with speed bumps in residential areas) and driving behaviour (for example, sportive or relaxed).

In line with the literature, models of Nanjing and the Randstad are quite similar as regards socio-demographic and built environment determinants for daily travel time and distance. However, there are some notable differences. Descriptive comparisons show deviations in daily travel time and even larger differences for daily distances travelled between the two areas. While Nanjing participants travel approximately 40 minutes and 4.4 km per day, their Dutch counterparts spend 70 minutes and cover a distance of 24.5 km every day. Multivariate analyses reveal various determinants for these variances. In both areas, socio-demographics have a strong impact on travel times and distances, especially in the two dominant sub-groups of car owners in the Randstad and non-car-owners in Nanjing. ‘Classical’ determinants like income and education play a more prominent role in the Netherlands. However, there is a striking difference with opposite effects for women belonging to carless households in both samples. In Nanjing, female respondents travel less time and distance. In the Randstad, the reverse is true. This can be explained by the interaction between land use patterns and dominant transport modes as well as the danwei system. Muller’s (1995/2004) four-stage model of intrametropolitan transport and urban growth assumes that the dominant transport system in a certain era shapes the spatial structure of the city. The dominance of walking and cycling in Nanjing has developed spatial structures which offer carless women more opportunities to travel short distances than their Dutch counterparts who live in built environments shaped by fast transport modes. Over the past 50–60 years, the urban form of the Randstad has transformed into a car-based spatial configuration to the detriment of non-car-owners, like women and low-income groups (Urry, 2007). The Chinese danwei system also has an impact on travel time and distance. This remnant of the socialist welfare-oriented housing system provided inhabitants with work, housing and facilities for everyday life (shops, education, medical services and recreation) in close proximity (Wang and Chai, 2009). In combination with other structures that offer many decentralised opportunities in close proximity to housing, the danwei system still reduces travel distances and times, especially for women who are largely responsible for household tasks.

Another main difference between Nanjing and the Randstad concerns the impact of the built environment. The reason for this difference might be related to the choice opportunities offered. In Nanjing, for example, the relatively insufficient opportunities to travel by car and public transport mean that people do not have many options when it comes to travelling short distances apart from walking and cycling. However, people in the Randstad who want to travel similar distances can choose between walking, cycling, car and good quality public transport. Characteristics of spatial structure and land use are explicitly more important in Nanjing and, in particular, people who do not have cars are more sensitive to travel time and distances than people in the Randstad. Due to the dominance of walking and cycling in the daily life of Nanjing inhabitants, users are much more dependent on their own physical energy when it comes to travelling. Increases in distance and travel time lead to a greater use of energy and increased fatigue. By contrast, inhabitants of the Randstad rely more often on motorised modes.

As regards future developments, increasing car ownership could have a huge impact on travel behaviour in Nanjing, especially in view of already important socio-demographic determinants in the Randstad. Economic growth in China will result in higher incomes and then in increasing private car ownership and use. This will lead to new trade-offs between location centrality and travel costs (Muller, 1995/2004). Urban development could spread over larger areas, resulting in lower densities—a development comparable with previous processes in Western countries (Dijst and Vázquez, 2007).

However, Nanjing’s public authorities might be able to limit such processes. The state ownership of land means that the Chinese public authorities are in a much stronger position to control land development and urban form than Western authorities (Pan et al., 2009). Enforcement of planning codes and other policy instruments would enable Nanjing authorities to preserve high urban densities and mixed land uses. The considerable investments in the Nanjing metro network (NPB, 2010) may potentially limit land use sprawl by offering space for compact urban developments near metro stations. This could create opportunities for walking, cycling and the use of high-speed public transport, which is not only relevant for sustainability reasons, but also for supporting inhabitants who cannot afford a car. However, due to diverse—and sometimes conflicting—interests (for example, land consolidation, expected economic benefits from growing motorisation), it is unclear whether policy and planning will focus on these issues (Hsing, 2010).

Various studies have argued that scholars should be careful regarding the temporal and spatial transferability of spatial policies (for example, Badoe and Miller, 1995) and that contextual specificity should be taken into account. By doing just that, this paper can deliver some relevant insights for future developments in Western countries. Our paper documents that walking and cycling can only become serious alternatives to the private car if the spatial configuration of the built environment offers the possibility to reduce car mobility. Compactness might lead to shorter travel distances and travel times. Although the literature on fixed time budgets for travelling (for example, Maat et al., 2005) argues that time saved on shorter trips might be invested in additional activities or other longer trips, some empirical evidence indicates that any such rebound effects are considerably smaller than the reduced travel distances resulting from a compact urban structure (for example, Næss, 2006). Urban designs that support non-motorised transport, thereby providing efficient ways to satisfy individual needs without using cars, are the key to less car-oriented development. However, due to spatio-temporal and physical constraints, walking and cycling are not realistic alternatives for longer travel distances. Hence, a compact built environment only serves as a necessary, but insufficient, condition for less car mobility.

In general, history has shown that fast modes of transport are needed to stimulate economic growth. However, from an environment perspective, fast public transport, rather than private car as the dominant mode, is preferred at metropolitan and international levels. Walking and cycling are proposed as major transport modes at the local level (Newman and Kenworthy, 2000; Schwanen et al., 2004). As far as Nanjing is concerned, this would actually mean preserving as much as possible the configurations of current land use which support walking and cycling and, in the meanwhile, investing in fast public transport to facilitate economic developments. In the case of the Randstad, it would seem to be wise to promote the use of walking and cycling by continuing to encourage compact land use patterns in combination with relatively fast public transport developments.

It should be noted that this paper has some limitations. First, the differences in built-up areas of the Randstad (the high functional specialisation of the four big cities situated around a green heart) and Nanjing (one continuous built-up area) might have influenced our results, since a certain proportion of trips in the Randstad are intercity trips. In future research, this influence could be limited by comparing Chinese metropolitan areas with monocentric areas like Paris or London. Secondly, as the built environment also influences car ownership, the distinction between car owners and non-car-owners has masked the effects of these factors on car ownership. Future analyses could correct this problem.

Footnotes

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

References

Badoe

D. A.

Miller

E. J.

(1995) Analysis of the temporal transferability of disaggregate work trip mode choice models, Transportation Research Record, 1493, pp. 1–11.

Bhat

C. R.

Guo

J. Y.

(2007) A comprehensive analysis of built environment characteristics on household residential choice and auto ownership levels, Transportation Research Part B, 41(5), pp. 506–526.

Cervero

(1996) Mixed land-uses and commuting: evidence from the American housing survey, Transportation Research Part A, 30(5), pp. 361–377.

Cervero

Kockelman

(1997) Travel demand and the 3Ds: density, diversity, and design, Transportation Research Part D, 2(3), pp. 199–219.

Dieleman

F. M.

Dijst

Burghouwt

(2002) Urban form and travel behaviour: micro-level household attributes and residential context, Urban Studies, 39(3), pp. 507–527.

Dijst

Vázquez

(2007) Employment deconcentration in European metropolitan areas: a comprehensive comparison and policy implications, in: Razin

Dijst

Vázquez

(Eds) Employment Deconcentration in European Metropolitan Areas, pp. 265–291. Dordrecht: Springer.

EEA (European Environment Agency) (2009) EMEP/EEA air pollutant emission inventory guidebook: technical guidance to prepare national emission inventories. EEA, Copenhagen (www.eea.europa.eu/publications/emep-eea-emission-inventory-guidebook-2009; accessed 30 June 2011).

Ewing

Cervero

(2010) Travel and the built environment: a meta-analysis, Journal of the American Planning Association, 76(3), pp. 265–294.

Geurs

K. T.

Wee

B. van

(2006) Ex-post evaluation of thirty years of compact urban development in the Netherlands, Urban Studies, 43(1), pp. 139–160.

10.

Giuliano

Narayan

(2003) Another look at travel patterns and urban form: the US and Great Britain, Urban Studies, 40(11), pp. 2295–2312.

11.

Gordon

Kumar

Richardson

H. W.

(1989) Congestion, changing metropolitan structure, and city size in the United States, International Regional Science Review, 12(1), pp. 45–56.

12.

Hsing

Y.-T.

(2010) The Great Transformation: Politics of Land and Property in China. Oxford: Oxford University Press.

13.

Ingram

G. K.

Liu

(1999) Determinants of motorization and road provision. World Bank, Washington, DC.

14.

Levinson

D. M.

Kumar

(1997) Density and the journey to work, Growth and Change, 28(2), pp. 147–172.

15.

Walker

J. L.

Srinivasan

Anderson

W. P.

(2010) Modeling private car ownership in China, Transportation Research Record, 2193(1), pp. 76–84.

16.

Maat

Wee

B. van

Stead

(2005) Land use and travel behaviour: expected effects from the perspective of utility theory and activity-based theories, Environment and Planning B, 32(1), pp. 33–46.

17.

Muller

P. O.

(1995/2004) Transportation and urban form: stages in the spatial evolution of the American metropolis, in: Hanson

Guiliano

(Eds) The Geography of Urban Transportation, 3rd edn, pp. 59–85. New York: The Guilford Press.

18.

Næss

(2005) Residential location affects travel behavior: but how and why? The case of Copenhagen metropolitan area, Progress in Planning, 63(2), pp. 167–257.

19.

Næss

(2006) Are short daily trips compensated by higher leisure mobility?, Environment and Planning B, 33(2), pp. 197–220.

20.

Næss

(2010) Residential location, travel, and energy use in the Hangzhou metropolitan area, Journal of Transport and Land Use, 3(3), pp. 27–59.

21.

Nanjing Statistics Bureau (2008) Nanjing shi tong ji nian jian [Nanjing statistical yearbook]. Nanjing: Zhong guo tong ji chu ban she [China Statistical Publishing House].

22.

Newman

Kenworthy

(2000) The ten myths of automobile dependence, World Transport Policy and Practice, 6(1), pp. 15–25.

23.

W. S.

Schipper

Chen

(2010) China motorization trends: new directions for crowded cities, Journal of Transport and Land Use, 3(3), pp. 5–25.

24.

NPB [Nanjing Planning Bureau] (2010) Nanjing cheng shi zong ti gui hua [Master plan of Nanjing City for the period 2007 to 2010]. NPB, Nanjing.

25.

Pan

Shen

Zhang

(2009) Influence of urban form on travel behaviour in four neighbourhoods of Shanghai, Urban Studies, 46(2), pp. 275–294.

26.

Potoglou

Kanaroglou

P. S.

(2008) Modelling car ownership in urban areas: a case study of Hamilton, Canada, Journal of Transport Geography, 16(1), pp. 42–54.

27.

Regio Randstad (2004) Economische strategie Randstad. Regio Randstad, Utrecht.

28.

Schafer

Victor

D. G.

(2000) The future mobility of the world population, Transportation Research Part A, 34(3), pp. 171–205.

29.

Schipper

(2004) Rapid motorization in China: environmental and social challenges. World Resources Institute, Washington, DC.

30.

Schwanen

Dijst

Dieleman

F. M.

(2002) A micro-level analysis of residential context and travel time, Environment and Planning A, 34(8), pp. 1487–1508.

31.

Schwanen

Dijst

Dieleman

F. M.

(2004) Policies for urban form and their impact on travel: the Netherlands experience, Urban Studies, 41(3), pp. 579–603.

32.

Stead

(1999) Relationships between transport emissions and travel patterns in Britain, Transport Policy, 6(4), pp. 247–258.

33.

Stead

(2001) Relationships between land use, socioeconomic factors, and travel patterns in Britain, Environment and Planning B, 28(4), pp. 499–528.

34.

TSW (Traffic Safety Weekly) (2009) Walking and cycling account for more than 60% of urbanities of Chinese (www.bjjtgl.gov.cn/publish/portal0/tab120/info14220.htm) [in Chinese].

35.

Urry

(2007) Mobilities. Cambridge: Polity Press.

36.

Wang

Chai

(2009) The jobs–housing relationship and commuting in Beijing, China: the legacy of Danwei, Journal of Transport Geography, 17(1), pp. 30–38.

37.

Worldwatch Institute (2009) State of the World. Washington, DC: W.W. Norton and Co.