Transportation Research Part D 59 (2018) 361–374

Contents lists available at ScienceDirect

Transportation Research Part D

journal homepage: www.elsevier.com/locate/trd

Travel mode choices in small cities of China: A case study of

T
Changting
⁎
Hong Hua, Jiangang Xua, Qing Shenb,c, , Fei Shia, Yangjin Chena
a
Department of City and Regional Planning, Nanjing University, China
b
Department of Urban Design and Planning, University of Washington, United States
c
Department of Urban Planning, Tongji University, China

AR TI CLE I NF O AB S T R A CT

Keywords: The existing literature on urban transportation planning in China focuses primarily on large cities
Mode choice and neglects small cities. This paper aims to fill part of the knowledge gap by examining travel
Small cities mode choice in Changting, a small city that has been experiencing fast spatial expansion and
China growing transportation problems. Using survey data collected from 1470 respondents on week-
Built environment
days and weekends, the study investigates the relationship between mode choice and individuals’
Attitude
socio-economic characteristics, trip characteristics, attitudes, and home and workplace built
environments. While more than 35 percent of survey respondents are car owners, walk, bicycle,
e-bike, and motorcycle still account for over 85 percent of trips made during peak hours. E-bike
and motorcycle are the dominant means of travel on weekdays, but many people shift to walking
and cycling on weekends, making non-motorized and semi-motorized travel especially important
for non-commuting trips. Results of multinomial logistic regression show that: (1) job-housing
balance might exert different effects on mode choice in different types of urban areas; (2) ne-
gative attitude towards e-bike and motorcycle is associated with more walking and cycling; and
(3) land use diversity of workplace is related to commuting mode choice on weekdays, while land
use diversities of both residential and activity places do not significantly affect mode choice on
weekends. Our findings imply that planning and design for small cities needs to differentiate land
use and transportation strategies in various types of areas, and to launch outreach programs to
shift people’s mode choice from motorized travel to walking and cycling.

1. Introduction

Travel behaviors have been widely investigated in large metropolises of China (Wang and Chai, 2009; Zhao et al., 2014), but have
been mostly overlooked in the hundreds of small cities. According to the Annual Report on the Development of Small and Medium-Sized
Cities in China 2015, cities with a population less than 500,000 are classified as small cities (CDRI, 2015). In the recent decades, most
small cities in China have witnessed intensified infill development in central areas and rapid urban expansion at the periphery. The
central areas of these cities are typically characterized by relatively high population density, concentrated amenities, mixed land
uses, and diversified activities, while the suburbs have smaller populations and are often developed as predominantly single land use
areas such as industrial zones or residential zones. Fast paced urban development in small cities also leads to accelerated motorization
and severe traffic congestions, which have been commonly observed in large cities (Pan et al., 2009; Ye et al., 2013). In 2012, the
Chinese government began to implement the “new urbanization” strategy that encourages the in-situ urbanization (ben di cheng zhen

⁎
Corresponding author at: Department of Urban Design and Planning, University of Washington, Office: 410E Gould Hall, Seattle, WA 98195-5740, United States.
E-mail address: qs@uw.edu (Q. Shen).

https://doi.org/10.1016/j.trd.2018.01.013

1361-9209/ © 2018 Elsevier Ltd. All rights reserved.

H. Hu et al. Transportation Research Part D 59 (2018) 361–374

hua) and economic development of small cities (Wang and Wang, 2015). This national policy will not only promote further growth of
small cities, but also stimulate the demand for motorized travel in these cities. China’s urban planners need to develop effective land
use and transportation strategies to maintain the traditional “green transportation” (e.g. walking, cycling, and bus riding) in small
cities by encouraging the choice of non-motorized travel modes.
The contextual factors of China’s small cities, such as economic development, urban form, infrastructure construction, and
household socio-economic characteristics are different from those in large cities. Consequently, travel behaviors of residents in small
cities are different from what we have learned in large cities. For instance, in response to the challenge of reducing traffic congestion,
air pollution, and greenhouse gas emission, China’s large cities have intensely promoted public transportation by constructing rail
transit systems, adding new bus lines, and improving service. The mode share for public transportation remains quite high in cities
such as Shanghai where bus and rail transit serve almost one third of trips in the central area (Shen et al., 2016). In contrast, public
transportation shows a low modal split (less than 10 percent) in most small cities of China due to limited transit networks, poor
service characterized by frequent delays in departure and arrival, and small fleets of buses (Wan et al., 2013).
In general, we know relatively little about transportation problems in small cities. The existing knowledge of travel behaviors and
their underlying causes in these cities is inadequate for guiding the practice of urban planning and policy making.
This paper aims to answer two research questions: (1) what are the key characteristics of residents’ travel behaviors in small cities
of China? and (2) what factors and to what extent these factors influence residents’ travel mode choices in these cities? We ap-
proached these questions through a case study of Changting, which is located in Fujian Province. Using data collected through a
travel survey, we estimated multinomial logistic regression models to investigate the effects of built environment and attitudinal
factors on mode choice while controlling for individuals’ socio-economic and trip characteristics. The next section briefly reviews
previous research on travel mode choice and highlights urban transportation challenges facing small cities in China, followed by a
description of the study area, data collection, and modeling approach in the third section. Then, descriptive statistics are presented
and results of the regression models are interpreted. Our research findings allow several important conclusions to be drawn in the last
section of the paper.

2. Literature review

2.1. Factors affecting travel mode choice

2.1.1. Built environment factors

Although there is no consensus about factors affecting travel mode choice, many studies have found that built environment
characteristics including density, land use mix, and street design influence people’s mode choice (Cervero, 2002; Handy et al., 2005;
Senbil et al., 2009; Bergman et al., 2011; Haybatollahi et al., 2015; Shen et al., 2016). To be specific, high density and mixed land use
are shown to make people prefer non-motorized travel modes. Some examine the impacts of built environment at both trip origin and
destination, and find that workplace environment creates greater effects on mode choice than residential environment (Chen et al.,
2008; Kwoka et al., 2015). In addition, a dense public transportation network helps to keep users, while a less developed public
transportation system can cause transit riders to switch to cars (Susilo and Maat, 2007). Local topography like hilly roads in mountain
cities help make motorcycle a popular mode choice (Weinert et al., 2008).
Studies conducted in Chinese cities find similar relationships between built environment characteristics and mode choice (e.g. Ye
et al., 2014; Tana et al., 2015). Compact and clustered developments with high job and population densities and balanced job-
housing opportunities are shown to help reduce automobile use and increase non-motorized travel (Wang and Zhou, 2016). Higher
levels of mixed land use, transportation connectivity, and accessibility to commerce and jobs are also associated with lower shares for
motorized personal travel modes (Zhang et al., 2013). Pan et al. (2009) choose four neighborhoods from inner and outer-city areas of
Shanghai to study the impact of neighborhood types on mode choice, and find that pedestrian/cyclist-friendly urban form can help
lower the level of motor vehicle ownership. Naess (2010) presents findings from Hangzhou indicating that living close to the primary
center of the metropolitan area is associated with higher proportions of trips by bicycle and on foot than living close to sub-centers. A
recent study focusing on aggregated travel behavior in 161 Chinese cities identify positive effects of city size, but negative effects of
density, on car ownership and usage (Sun et al., 2015). However, relatively few studies actually evaluate the magnitude of the
impacts of statistically significant factors.

2.1.2. Socio-economic factors

Prior studies show that socio-economic indicators of households and individuals strongly influence people’s mode choice
(Cervero, 2002; Van Acker and Witlox, 2010). For instance, men tend to drive more frequently than women (Giuliano, 1983). In the
Chinese context, higher income, higher job status, and car ownership are associated with greater probabilities of car use, while female
and older people prefer walking or cycling (Pan et al., 2009; Shen et al., 2016). The existing literature indicates that income, gender,
age, and car ownership are the most frequently examined variables in the mode choice analysis.

2.1.3. Attitudinal factors

In addition to observable socio-economic and built environmental characteristics, less tangible factors such as attitudes, motives,
and preferences affect travel mode choice as well (Ben-Akiva et al., 1999; Collins and Chambers, 2005; Elias and Shiftan, 2012).
Individuals in similar situations and with comparable socio-economic characteristics often make different mode choices because of
various attitudes toward traveling (Li et al., 2013; Haybatollahi et al., 2015). The enthusiasm for driving cars or the pursuit of privacy

362
H. Hu et al. Transportation Research Part D 59 (2018) 361–374

and comfort would likely lead to car usage, while strong environmental awareness might lead to the choice of a non-motorized mode
(Chang and Wu, 2008). Special preferences or attitudes play a mediating role between built environment and mode choice, which
lead people to select themselves to particular living environments (Heinen et al., 2011; Klöckner and Friedrichsmeier, 2011; Gim,
2016). For instance, people with a strong environmental concern could prefer to live in downtown areas to reduce their car use
(Schwanen and Mokhtarian, 2005). However, environmental concern may not be a highly influential factor of self-selection in
developing countries like China (Wang and Lin, 2014). Such self-selection effects are often cached by indicators of attitudes toward
transportation alternatives (Cao, 2009; Ewing and Cervero, 2010; Hong et al., 2014).

2.1.4. Trip chain

In the recent decades, the complexities of trip chaining including frequency, number of activity purposes, location, time of day,
and sequence of activities have been found to influence people’s mode choice (Bhat, 1999; Ben-Akiva and Bowman, 2000; Bhat and
Singh, 2000; Hensher and Reyes, 2000; Ye, 2004; Krygsman et al., 2007; Yang et al., 2007; Primerano et al., 2008; Islam and Habib,
2012). A trip chain is usually defined in the literature as travel from home to one or more activity locations and back home again
(Brunow and Gründer, 2013). Some researchers have used Beijing as an example and found that trip chain pattern affects mode
choice on weekdays, whereas the causal relationship is inverted on national holidays (Yang et al., 2016). Workers with high income,
households with children, and women tend to take more stops within one tour (Ma et al., 2014). In one of the few published studies
on transportation problems in small Chinese cities, the researchers have found that commuters using flexible modes such as walk,
bicycle, motorcycle and car are more likely to make complex trip chains, whereas those using a nonflexible mode like public transit
tend to make simple trips (Yang et al., 2007).

2.2. Urban development and transportation challenges in small cities of China

Small cities in China are usually monocentric, with concentration of public facilities in the central area. Since the economic
reform in the early 1980s, suburban developments in many Chinese cities have taken place in the urban fringe in the forms of
development zones, industrial zones, and high-tech zones (Wei, 2015). Historically, low incomes, high density, and good job-housing
balance have made small cities suitable places for walking and cycling. However, the expansion of urbanized area into suburban
locations leads to longer trips, and thereafter people tend to shift from traditional walking and cycling to various motorized modes
including car, motorcycle, e-bike, and public transportation (Cherry, 2007; Feng and Chen, 2010). “E-bikes” are usually defined as
electric two-wheelers, which have become a substantial component of the urban transportation system in China (Ling et al., 2015).
Public transportation has not been able to meet rising travel demands in many small cities due to a lack of efficient management
and operation (Wang et al., 2011; Wan et al., 2013; Ling et al., 2015). Consequently, motorized personal modes such as motorcycle
and e-bike have been heavily used. These options are very appealing because of they are inexpensive, ease to ride, and highly flexible
(Chen and Chen, 2011). However, since there is no separate lane for these modes, they are often mixed with cars or bicycles in the
traffic lanes, and thus worsen congestion, increase accidents, and aggravates environmental impacts (Cherry et al., 2009; Lan et al.,
2013). Although local governments have implemented various policies to regulate the usage of motorcycles and e-bikes, they still
have high mode shares in small cities (Cherry, 2007; Chen and Lai, 2011).
Existing literature on travel mode choices in China focuses primarily on large cities, such as Beijing, Shanghai, Guangzhou, and
Hangzhou. Only a small number of publications are devoted on medium sized cities such as Huizhou, Yangzhou, and Shijiazhuang
(e.g. Yang et al., 2007; Weinert et al., 2008; Naess, 2010; Liu et al., 2016). The hundreds of small cities have drawn little attention
from urban transportation researchers.
Based on the existing literature and also taking into account different characteristics of small cities in China, we conceptualized
four categories of factors influencing residents’ travel mode choice in small cities: individual socio-economic characteristics, trip
characteristics, attitudinal factors, and built environment features at both trip origin and destination. We then applied these four
categories of explanatory variables in the case study of Changting described below.

3. Data and methodology

3.1. Study area

The city of Changting is located in the mountainous western part of Fujian Province, China. It was initially built in the Han
Dynasty, and has been designated as a national historic city since 1994. In 2014, there were a total of 400,000 residents living in the
city and its surrounding rural areas. Reflecting the rapid economic growth and urban development, the urban built up area increased
from 6.4 square kilometers in 2005 to 11.1 square kilometers in 2014. Changting is quite hilly, and its streets and roads have
traditionally been narrow, which make e-bike and motorcycle highly popular as travel modes.
Changting is a typical monocentric city. The main administrative, educational, commercial, and service land uses concentrate in
the historical central districts. Residential areas stretch from the central districts to suburbs. An industrial zone has been developed in
the urban fringe since 2000, which includes textile and clothing, food processing, and machine making industries. In recent years,
Changting has been making efforts to transform its economy from traditional manufactures into modern logistics and E-commerce.
Therefore, the traditional industrial zone has declined, while job opportunities have increasingly been created in the newly developed
suburban areas. Based on the land use characteristics of Changting, we classified the study area of 21 districts into four types of areas:
the central urban area (four districts), the urban suburban transition area (four districts), the industrial area (three districts), and the

363
H. Hu et al. Transportation Research Part D 59 (2018) 361–374

Fig. 1. Location and land use of the study area.

new suburban area (ten districts) (Fig. 1).

Although Changting is sprawling quickly, the average travel distances of both commuting and leisure trips are moderate due to
the relatively small city scale. However, the narrow roads and concentrated land use function in the historical central districts make
e-bike and motorcycle highly mixed with cars, walking, and cycling in the central urban area. In recent years, the frequency of traffic
jams and injuries is rising particularly in the commuting peak hours. Public transportation has developed slowly since 2004. There
are seven bus lines in total in the study area and the routes are duplicated to a large extent. The current bus lines mainly concentrate
in the central urban area or connect the central area to suburban areas. The bus service quality is low and there are not enough bus
stops. The travel efficiency and safety problems have become a great concern for the public and local planners. These transportation
problems faced by Changting are not unique; they are rather common in small cities in China.

364
H. Hu et al. Transportation Research Part D 59 (2018) 361–374

3.2. Data collection and processing

To understand factors influencing travel mode choice in Changting, we conducted a survey of local residents in January 2016.
Since we did not have the resources to do strictly scientific sampling, we cooperated with the local government to do the survey. The
local government assisted us to ask school teachers to distribute our surveys to their students, who then returned the completed
questionnaires randomly because only some parents responded. This is the common process to conduct surveys in a small city in
China. All classes in all schools located in the study area were selected. We sent out about 30,000 questionnaires at the beginning,
which covered all students in the study area. In our survey, we asked an adult in the household to report trips on a typical day (could
be either workday or weekend). At the conclusion of our survey, we collected 1470 valid responses, including 1049 responses for
weekdays and 421 responses for weekends. The response rate was 29.4%. Each household only filled out one questionnaire and
reported three categories of information: (1) socio-economic characteristics of the respondent, including age, gender, family income,
and car ownership; (2) travel characteristics of the respondent, including all trips made on a typical workday or a weekend day, origin
and destination districts of each trip, travel mode, trip purpose, starting and ending time of each trip, and trip distance; (3) the
respondent’s attitudes toward different travel modes and built environments. The built environment characteristics were derived
from the land use map, road network map, and census provided by the local government, including population density, road density,
intersection density, and land use mixture at both origin and destination locations.
Based on collected data on travel characteristics, the activity duration was estimated and trip chains were constructed for each
respondent. Activity duration was estimated as the difference between the reported ending time of one trip and the starting time of
the following one. A trip chain was defined as a sequence of out of home activities that start and end at home (Ye et al., 2007). We
distinguished between simple and complex chains for commuting-based trips. A simple chain contains only the working activity. The
complex chain contains working and leisure activities such as shopping, entertaining (e.g. going to restaurants, cinemas, sports, etc.),
visiting friends, and others (any activities other than shopping, entertaining, visiting friends, and working). If a complex trip chain
involved more than one travel mode, we used the commuting mode as the main travel mode. Trips on workday mainly focused on
commuting trips. Those trips that included only non-work activities during weekdays accounted for just a small percentage of the
total, and we did not include them in our analysis. We did not include trip chains for leisure activities on weekends due to the small
sample size.
There were seven travel modes identified in our survey: walk, bicycle, e-bike, motorcycle, bus, taxi, and car. Respondents who
used taxi for one trip accounted for less than 1% of all respondents, and hence we dropped these respondents from our analysis.
Since traffic congestion in Changting appeared to be caused primarily by e-bikes and motorcycles, we asked each respondent
whether or not these modes should be restricted in the central urban area of the city (1 for yes, 0 for no). The answers allowed us to
measure attitudinal variables that are useful for understanding travel mode choice. We also requested each respondent to rate his/her
satisfaction levels with regard to, respectively, the working environment, air quality, green space, and leisure environment (1 for
dissatisfied, 2 for neutral, and 3 for satisfied).
We used the entropy index (ENT) to measure land use mixture in each district (Song et al., 2013). We used four land use types
related to working activities on weekdays: public service, commercial, industrial, and residential. We chose three land use types
related to leisure activities on weekends: commercial, green, and residential. ENT was calculated as follows.
k
⎡ ⎤
ENT = −⎢ ∑ Pj ln(Pj )⎥/ln(k )
⎣ j = 1 ⎦ (1)
where k is the number of land use types (k ≥ 2); Pj is the percentage of each land use type j in the district. ENT varies from 0
(homogeneous land use) to 1 (most mixed land use).

3.3. Model specification

We employed the multinomial logistic regression model (MNL), which has been widely used for studying mode choice, for this
research. MNL is a statistical method for examining the relationship between one dependent nominal variable (with more than two
alternatives) and various independent variables. Our dependent variable is mode choices with four alternatives: walk/bicycle, e-bike/
motorcycle, bus, and car, and our independent variables include socio-economic characteristics, trip characteristics, attitudinal in-
dicators, and built environment features. MNL estimates k-1 multiple linear regression functions defined as follows.

P(y = 1) ⎞
logit(y = 1) = log ⎛
⎜ ⎟ = β0 + β1·S + β2·T + β3·A + β4 ·B
⎝ 1−(P (y = 1)) ⎠ (2)

P(y = k−1) ⎞
logit(y = k−1) = log ⎛⎜ ⎟ = β0 + β1·S + β2·T + β3·A + β4 ·B
⎝ 1−(P (y = k−1)) ⎠ (3)
where k is the number of alternatives of the dependent variable, S is a vector of socio-economic variables, T is a vector of trip
variables, A is a vector of attitudinal variables, and B is vector of built environment variables, β is the set of regression coefficients
associated with each explanatory variables.
We attempted alternative specifications of the MNL model by regrouping the choice alternatives in different ways, and we also
tried the nested logit model with groups of motorized and non-motorized modes. The results were not better than the original MNL

365
H. Hu et al. Transportation Research Part D 59 (2018) 361–374

Table 1
Built environment characteristics in the four types of areas.

Variable Central Urban Urban Suburban Transition Industrial Suburban ANOVA

(four districts) (four districts) (three districts) (ten districts) p value

Built Up Area (km2) 2.5 2.2 2.2 4.2 0.203

Land Uses (%)

Residential 37.2 56.9 24.5 53.0 0.024
Public Service/Administration 9.0 8.0 0.7 9.0 0.254
Commercial/recreational 8.2 5.7 2.0 0.6 0.000
Industrial 37.0 7.0 69.2 23.0 0.000
Green 1.7 4.0 1.3 4.8 0.864

Land Use Index (0–1)

For Working Activities 0.67 0.60 0.49 0.51 0.280
For Leisure Activities 0.50 0.30 0.21 0.15 0.014

Road Network
Intersection Density (count/km2) 183 41 29 34 0.000
Road Density (km/km2) 17.5 9.6 8.6 10.1 0.069
Bus Stop Density (count/km2) 11 8 1 3 0.000
Population Density (persons/km2) 21,717 12,945 4851 7582 0.015

model with those four choice alternatives. In small cities of China, e-bike/motorcycle and bus are typically competing travel modes
with very different characteristics. Therefore, it may not be suitable to group them together as motorized modes within a nested
modeling framework.

4. Results

4.1. Built environment and socio-economic characteristics

The four types of areas identified in this paper have different built environment characteristics. We did the analysis of variance
(ANOVA) to test the null hypothesis that data from the four areas are drawn from populations with identical means. The p value
represents the probability that the null hypothesis is true, and if p value is less than 0.05, it indicates that the null hypothesis is not
supported by the data. Table 1 shows that built environments in the four types of areas are quite different from each other. Land use is
more mixed in the central urban area than in other areas, and that population density decreases from the central urban area to the
suburban and industrial areas. The urban suburban transition area and the suburban area have relatively large portions for residential
land use (more than 50%), whereas the industrial area is predominantly industrial land (69.2%). Commercial land accounts for
considerable portions of the central urban area (8.2%) and the urban suburban transition area (5.7%), but has much less significant
presence in the industrial area (2.0%) and the suburban area (0.6%). Moreover, the industrial area shows a scarcity of public service
land (only 0.7%), which is quite evenly distributed among the other three types of areas. Since the central urban area consists mostly
of historical neighborhoods, blocks are small and streets are narrow (usually just one lane wide). This area has the highest street
density of 17.5 km per km2, intersection density of 183 per km2, and bus stop density of 11 per km2. The industrial area has the
lowest street density and least public transportation service.
Table 2 shows that the differences in socio-economic and attitudinal characteristics among respondents in the four types of areas
are mostly either not statistically significant or only marginally significant. In general, more males commute on weekdays and more
females make non-commuting trips on weekends. Almost half of the respondents have annual family income of more than 100,000
yuan. Car ownership is reported by more than 35% of respondents, but the percentage is lower for the central urban area. A clear
majority of respondents would like to restrict e-bikes and motorcycles on both weekdays and weekends (58%), and somewhat higher
percentages of respondents in the industrial area and the suburban area show negative attitudes towards e-bikes and motorcycles. The
satisfaction indicators show similar values (more than 2) across the four areas, implying that respondents are generally satisfied with
the living environment.

4.2. Trip characteristics and mode choices

Tables 3 and 4 display trip characteristics of respondents on weekdays and weekends in the four type of areas. On weekdays, with
commuting modes changing from lower to higher level of motorization, commuting distance increases from 1.2 km (walk/bicycle) to
4.3 km (car). The average commuting distance by bus is 2.6 km, and that by e-bike or motorcycle is 2.9 km, indicating that e-bike and
motorcycle play an important role in medium distance commuting. Non-commuting trip distances by each travel mode are slightly
shorter than commuting distances in most areas. This indicates that residents tend to travel shorter distances for leisure activities, but
endure somewhat longer commuting distances. Respondents living in the central urban area walk or cycle relatively shorter distances
to their workplaces than those in other areas. However, this difference is only marginally significant. In contrast, respondents living
in the industrial and suburban areas drive longer distances for leisure activities. These summary statistics suggest that the more mixed

366
 H. Hu et al.

Table 2
Socio-economic and attitudinal characteristics of respondents on weekdays and weekends.

Variable Weekdays Weekends

Total Central Urban Urban Suburban Industrial Suburban ANOVA Total Central Urban Urban Suburban Industrial Suburban ANOVA
(N = 1049) (N = 293) Transition (N = 103) (N = 411) p value (N = 421) (N = 126) Transition (N = 31) (N = 178) p value
(N = 242) (N = 86)

Gender (%) 0.841 0.087

Female 45.4 44.7 46.7 48.5 44.3 56.5 47.6 57.0 58.1 62.4
Male 54.6 55.3 53.3 51.5 55.7 43.5 52.4 43.0 41.9 37.6

Age (%) 0.026 0.020

20–40 Years Old 46.4 53.9 43.8 42.7 43.6 44.2 51.6 39.5 61.3 38.2
≥40 Years Old 53.6 46.1 56.2 57.3 56.4 55.8 48.4 60.5 38.7 61.8

Annual Family Income (%) 0.104 0.246

367
< 50,000 (CNY) 16.4 16.4 13.2 13.6 19.0 19.7 23.0 11.6 25.8 20.2
50,000–100,000 (CNY) 35.5 43.3 34.3 36.9 30.2 29.0 24.6 32.6 35.5 29.2
> 100,000 (CNY) 48.1 40.3 52.5 49.5 50.8 51.3 52.4 55.8 48.7 50.6

Car Ownership 0.137 0.025

Yes 35.5 30.7 36.4 42.7 36.5 37.1 26.2 40.7 45.2 41.6

Should Restrict E-bike/Motorcycle 0.256 0.054

(%)
Yes 58.5 55.6 58.3 67.0 58.6 58.2 52.4 51.2 71.0 63.5

Satisfaction Level (1–3, mean(SD))

Working Environment 2.3(0.6) 2.3(0.6) 2.4(0.5) 2.4(0.6) 2.3(0.6) 0.203 2.3(0.6) 2.2(0.6) 2.4(0.6) 2.4(0.6) 2.3(0.6) 0.062
Air Quality 2.4(0.6) 2.3(0.6) 2.4(0.6) 2.4(0.6) 2.3(0.6) 0.379 2.3(0.6) 2.2(0.7) 2.3(0.6) 2.4(0.5) 2.4(0.6) 0.121
Green Space 2.2(0.6) 2.2(0.6) 2.3(0.7) 2.2(0.7) 2.2(0.6) 0.117 2.1(0.6) 2.0(0.6) 2.2(0.7) 2.2(0.7) 2.2(0.6) 0.095
Leisure Environment 2.3(0.6) 2.3(0.6) 2.3(0.6) 2.3(0.7) 2.2(0.7) 0.077 2.2(0.6) 2.1(0.7) 2.3(0.7) 2.4(0.6) 2.3(0.6) 0.075
Transportation Research Part D 59 (2018) 361–374
H. Hu et al. Transportation Research Part D 59 (2018) 361–374

Table 3
Trip characteristics and mode choices on weekdays.

Commuting trips on weekdays Total Central Urban Urban Suburban Transition Industrial Suburban ANOVA
(N = 1883) (N = 521) (N = 440) (N = 180) (N = 742) p value

Commuting Distance (km, mean(SD))

Walk/Bicycle 1.2(1.1) 0.9(0.6) 1.2(0.8) 1.3(0.8) 1.1(1.4) 0.071
E-bike/Motorcycle 2.9(2.3) 3.0(2.4) 2.9(2.0) 3.0(2.4) 2.8(2.3) 0.529
Bus 2.6(3.0) 5.0(6.2) 2.0(0.0) 1.9(1.4) 2.2(1.5) 0.197
Car 4.3(3.8) 4.4(4.2) 4.9(4.2) 3.8(3.9) 3.7(3.2) 0.302

Commuting Peak Hours (%)

7:00 AM – 8:00 AM (41.6%) 0.212
Walk/Bicycle 24.5 26.0 21.2 20.5 26.4
E-bike/Motorcycle 60.7 63.2 59.2 60.3 59.7
Bus 1.8 0.4 1.1 5.1 2.3
Car 13.0 10.3 18.4 14.1 11.6
1:00 PM – 2:00 PM (28.5%) 0.432
Walk/Bicycle 20.5 24.1 21.0 16.0 18.4
E-bike/Motorcycle 66.8 65.8 63.6 66.0 70.3
Bus 0.6 0.0 0.0 2.0 1.1
Car 12.1 10.1 15.4 16.0 10.3
Working Duration (hour, mean(SD)) 4.3(2.0) 4.4(2.0) 4.3(1.9) 4.4(1.9) 4.2(2.0) 0.230
Number of Commuting Trips (per day) 1.9 1.9 1.9 1.9 1.9 0.753

Trip Chain (%) 0.007

Home-Work-Home 96.1 98.1 96.4 92.8 95.3
Home-Work-Other activity-Home 3.9 1.9 3.6 7.2 4.7

Table 4
Trip characteristics and mode choices on weekends.

Leisure trips on weekends Total Central Urban Urban Suburban Transition Industrial Suburban ANOVA
(N = 754) (N = 224) (N = 150) (N = 51) (N = 329) p value

Trip Distance (km, mean(SD))

Walk/Bicycle 0.9(0.7) 1.0(0.7) 0.9(0.7) 0.7(0.4) 1.0(0.8) 0.494
E-bike/Motorcycle 2.5(2.3) 2.8(2.6) 2.3(1.6) 3.0(3.3) 2.3(1.9) 0.145
Bus 3.2(0.8) – 2.5(1.4) 3.2(4.0) 3.5(0.6) 0.409
Car 4.0(4.2) 3.8(3.3) 2.6(1.6) 7.7(6.9) 4.2(4.6) 0.036

Trip Starting Time (%)

Before 12:00 PM (62.6%) 0.582
Walk/Bicycle 47.2 46.5 47.4 33.3 50.0
E-bike/Motorcycle 41.5 43.8 38.9 54.5 39.0
Bus 0.8 0.0 0.0 3.0 1.5
Car 10.4 9.7 13.7 9.1 9.5
After 12:00 PM (37.4%) 0.909
Walk/Bicycle 37.6 40.0 41.8 5.6 38.8
E-bike/Motorcycle 46.8 46.2 43.6 61.1 46.5
Bus 1.4 0.0 3.6 5.6 0.8
Car 14.2 13.8 10.9 27.8 14.0

Leisure Activity Types (%) 0.390

Shopping 29.2 27.2 27.3 31.4 31.0
Entertaining 5.2 5.8 4.7 5.9 4.9
Visiting Friends 7.7 5.4 11.3 7.8 7.6
Others 58.0 61.6 56.7 54.9 56.5

Activity Duration (hour, mean(SD))

Shopping 1.2(1.3) 1.5(1.7) 0.8(0.8) 1.2(1.6) 1.2(1.2) 0.089
Entertaining 1.5(0.9) 1.3(1.0) 1.3(0.6) 1.5(0.2) 1.7(1.1) 0.730
Visiting Friends 2.3(1.9) 2.7(2.2) 1.9(1.3) 2.0(1.2) 2.3(2.1) 0.769
Others 2.1(2.1) 2.2(2.4) 2.2(1.8) 2.0(1.5) 2.1(2.0) 0.925

Number of Leisure Trips (per day) 2.0 2.0 2.0 1.9 2.1 0.589

land use pattern of the central urban area facilitates short distance commuting by non-motorized modes, whereas the relatively
homogeneous land use in the industrial and suburban areas makes people drive longer for leisure activities.
Commuting trips include commuting from home to work and commuting from work to home. We found two distinctive peak hour
periods for each of these two types of commuting trips. Our analysis focused on home-to-work commuting, with both a morning peak
(7:00 AM – 8:00 AM) and an early afternoon peak (1:00 PM – 2:00 PM). The early afternoon peak hour does not typically exist in large

368
H. Hu et al. Transportation Research Part D 59 (2018) 361–374

Table 5
Trips and mode choices for the four types of areas on weekdays and weekends.

Weekdays

Home Area Work Area (%) Walk/Bicycle (%) E-bike/Motorcycle (%) Bus (%) Car (%)

Central Urban Central Urban: 49.3 38.5 52.1 1.6 7.8

(N = 521) Urban Suburban Transition: 13.6 22.5 64.8 0 12.7
Industrial: 5.0 15.4 69.2 0 15.4
Suburban: 32.1 11.4 76.0 1.2 11.4

Urban Suburban Transition Central Urban: 25.5 17.9 70.5 1.8 9.8
(N = 440) Urban Suburban Transition: 28.4 36.0 57.6 0 6.4
Industrial: 3.2 7.1 42.9 0 50.0
Suburban: 43.0 13.8 62.4 0 23.8

Industrial Central Urban: 23.9 14.0 74.4 0.0 11.6

(N = 180) Urban Suburban Transition: 3.9 42.9 57.1 0 0
Industrial: 19.4 34.3 48.6 5.7 11.4
Suburban: 52.8 14.7 64.2 5.3 15.8

Suburban Central Urban: 23.3 22.5 63.6 4.0 9.8

(N = 742) Urban Suburban Transition: 14.4 25.2 62.6 1.9 10.3
Industrial: 8.4 17.7 53.2 3.2 25.8
Suburban: 53.9 29.2 59.8 1.8 9.2

Weekends

Home Area Activity Area (%) Walk/Bicycle (%) E-bike/Motorcycle (%) Bus (%) Car (%)

Central Urban Central Urban: 58.5 42.7 46.6 0 10.7

(N = 224) Urban Suburban Transition: 6.7 60.0 13.3 0 26.7
Industrial: 4.9 27.3 63.6 0 9.1
Suburban: 29.9 46.3 44.8 0 9.0

Urban Suburban Transition Central Urban: 32.7 34.7 49.0 0 13.9

(N = 150) Urban Suburban Transition: 36.0 66.7 20.4 0 13.0
Industrial: 7.3 9.1 81.8 0 9.1
Suburban: 24.0 38.9 47.2 0 13.9

Industrial Central Urban: 43.1 22.7 72.7 4.5 0

(N = 51) Urban Suburban Transition: 7.8 0 50.0 0 50.0
Industrial: 21.6 36.4 45.5 0 18.2
Suburban: 27.5 21.4 42.9 7.1 28.6

Suburban Central Urban: 29.8 43.9 41.8 0 14.3

(N = 329) Urban Suburban Transition: 10.0 36.4 45.5 3.0 15.2
Industrial: 10.6 22.9 60.0 8.6 8.6
Suburban: 49.5 53.4 37.4 0 9.2

cities of China. This may be attributed to the relatively short commuting distances and a two-hour lunch break for workers in
Changting, which is commonly practiced in small cities. Many workers return to home to have lunch and then go back to workplace.
During both periods, e-bike/motorcycle is the most frequently used travel mode (more than 60%), followed by walk/bicycle, and
then by car, while bus only has a tiny mode share (less than 2%).
There is no peak hour for leisure activities on weekends. We divided these non-commuting trips into morning trips and afternoon
trips according to trip starting time. About two thirds of respondents choose to start non-commuting trips in the morning. Walking/
cycling is the most popular mode for these morning trips (47.2%), while e-bike/motorcycle has the largest mode share in the
afternoon (46.8%). Again, car accounts for a modest mode share (10.4% in the morning and 14.2% in the afternoon), and bus is used
by only a small number of respondents. In fact, no respondent living in the central urban area reported riding bus on weekends.
We defined a complex commuting trip chain as “Home-Work-Other Activity-Home”. Other activities include shopping, en-
tertaining, visiting friends, and other non-working activities. In general, respondents tend to go back home directly after work.
According to our survey, most commuters make new trips from home to other activities instead of combining other activities with
their commuting trips. The central urban area has only 1.9% of trips taken in the form of complex trip chains, the lowest percentage
among the four areas, while the industrial area has the highest percentage, at 7.2%. This difference can be explained by differences in
location and the pattern of land use mix. For people living in the central urban area that has more diverse land uses, the accessibility
to daily activities is good. These people tend to go directly back to home after work and then make separate trips for other activities,
because their travel distances and costs for making non-commuting trips are relatively low.
Table 5 shows that in both the central urban area and the suburban area, about half of the respondents commute within the area,
which is much higher than the percentages for the urban suburban transition area and the industrial area. A large portion of people
living in urban suburban transition and industrial areas commute to the suburban area, and about a quarter work in the central urban

369
 H. Hu et al.

Table 6
Multinomial logistic regression of travel mode choices on weekdays and weekends.

Variable Weekdays Variable Weekends

Walk/Bicycle Bus Car Walk/Bicycle Bus Car

Odds Ratio Sig. Odds Ratio Sig. Odds Ratio Sig. Odds Ratio Sig. Odds Ratio Sig. Odds Ratio Sig.

SOCIO-ECONOMIC SOCIO-ECONOMIC
Gender (male = 1) 0.676 0.003 1.096 0.802 2.490 0.000 Gender (male = 1) 0.427 0.000 0.114 0.061 2.180 0.012
Family Income (base level: < 50,000) Family Income (base level: < 50,000)
50,000–100,000 RMB 0.639 0.017 0.490 0.222 0.988 0.978 50,000–100,000 RMB 0.780 0.399 – – 0.530 0.246
> 100,000 RMB 0.615 0.008 1.345 0.561 2.329 0.046 > 100,000 RMB 0.603 0.064 – – 0.277 0.011
Car Ownership (yes = 1) 1.248 0.129 1.180 0.681 72.394 0.000 Car Ownership (yes = 1) 1.089 0.723 0.372 0.338 46.022 0.000

TRIP TRIP
Peak Hour (base level: other time) Activity (base level: other activity)
7:00 AM – 8:00 AM 0.810 0.165 0.611 0.200 1.120 0.613 Shopping 2.533 0.000 18.001 0.020 1.053 0.891
1:00 PM – 2:00 PM 0.648 0.011 0.171 0.006 0.973 0.914 Entertaining 4.527 0.002 19.989 0.075 0.716 0.668
Complex Chain (yes = 1) 0.783 0.454 1.818 0.450 3.281 0.018 Visiting Friends 2.070 0.078 8.014 0.175 2.267 0.120
Commuting Distance (km) 0.314 0.000 0.999 0.990 1.265 0.000 Trip Distance (km) 0.244 0.000 1.134 0.512 1.184 0.002

370
ATTITUDE ATTITUDE
Should Restrict E-bike/Motorcycle 1.355 0.024 0.933 0.854 1.306 0.177 Should Restrict E-bike/Motorcycle 1.535 0.039 0.856 0.860 1.276 0.439

BUILT ENVIRONMENT BUILT ENVIRONMENT

Landuse Index in Home District 0.698 0.478 0.391 0.506 0.261 0.065 Landuse Index in Home District 1.585 0.489 6.147 0.479 0.333 0.260
Landuse Index in Working District 1.320 0.632 91.313 0.005 0.036 0.000 Landuse Index in Activity District 0.571 0.438 0.232 0.589 0.212 0.136
Commute within the Same area 0.880 0.579 1.077 0.901 0.356 0.002 Travel within the Same area 1.055 0.882 – – 0.361 0.043
Home in Central Urban 0.692 0.233 0.436 0.354 0.494 0.065 Home in Central Urban 1.712 0.226 – – 1.328 0.665
Home in Urban Suburban Transition 0.843 0.508 0.286 0.126 1.028 0.932 Home in Urban Suburban Transition 0.924 0.826 0.418 0.393 0.779 0.610
Home in Industrial 0.553 0.054 1.532 0.481 0.694 0.354 Home in Industrial 0.280 0.013 1.151 0.899 0.187 0.023
Same area: Central Urban 2.043 0.053 1.825 0.601 5.854 0.001 Same area: Central Urban 0.472 0.244 – – 2.762 0.269
Same area: Urban Suburban Transition 1.642 0.175 0.001 0.867 2.901 0.088 Same area: Urban Suburban Transition 2.654 0.112 0.115 0.000 6.022 0.043
Same area: Industrial 4.087 0.013 3.184 0.278 2.031 0.391 Same area: Industrial 1.087 0.928 0.549 0.000 15.988 0.032

Model Fit Log-likelihood: −1234.27 Model Fit Log-likelihood: −493.73

McFadden R2: 0.323 McFadden R2: 0.361
Likelihood ratio test: chisq = 1177.75 Likelihood ratio test: chisq = 557.36
(p value = .000) (p value = .000)
Transportation Research Part D 59 (2018) 361–374
H. Hu et al. Transportation Research Part D 59 (2018) 361–374

area. This is because many companies in the industrial area are not performing, whereas the central urban suburban areas both have
substantial industrial and public service land uses that provide many job opportunities. It is especially worth noting that walking/
cycling is more frequently selected (29.2–38.5%) for within area commuting than for commuting in general. However, e-bike/
motorcycle still accounts for more than 50% of all within area and most inter area commuting trips.
Leisure trips on weekends show a similar spatial pattern. About half of the respondents travel within the central urban area and
the suburban area, while people living in the urban suburban transition area mainly travel within the same or to the central urban
area. In the industrial area, 43% of people go to the central urban for leisure activities. For within area trips on weekends, e-bike/
motorcycle shows reduced mode shares, while walk/bicycle and car show increased mode shares. Bus is seldom chosen on weekends.

4.3. MNL regression results

Table 6 reports the MNL regression results for commuting trips on weekdays and non-commuting trips on weekends of re-
spondents in families with children. In both models, the baseline mode of the dependent variable is e-bike/motorcycle. The alter-
native modes include walk/bicycle, bus, and car. The odds ratio measures the ratio of the probability that people will choose an
alternative travel mode to the probability that they will choose baseline mode for one unit increase in the value of the independent
variable. We tested all variables shown in Tables 1-4, many had insignificant signs. Road density and population density did not show
up as land use index did, indicating that land use diversity could be more influential on mode choice in small cities. Satisfaction with
the environment, working duration, and activity duration were not significant because the values of these variables were more or less
the same among respondents (see ANOVA). We also did collinearity test. The possibility of multi-collinearity problem is low.
In the socio-economic category, gender, family income, and car ownership are significantly associated with mode choice. Because
all odds ratios use e-bike/motorcycle as the reference mode, we can infer that males prefer car to e-bike/motorcycle, and prefer e-
bike/motorcycle to walk/bicycle for travel both on weekdays and weekends. Increased family income is related to lowered prob-
abilities of choosing to walk or cycle and higher probabilities of car usage relative to riding e-bike or motorcycle for commuting trips.
Car ownership shows a strong positive relationship to people’s car use on both weekdays and weekends, indicating an important
impact of motorization on mode choice.
Among the trip characteristic variables, commuting during the afternoon peak hours is associated with decreased likelihood of
walking/cycling relative to riding e-bike/motorcycle. Further, if the commuting trip is a complex chain, the likelihood of driving is
greatly increased, by 3.28 times, relative to choosing the reference mode. In addition, trip distance significantly influences mode
choice on both weekdays and weekends. For every kilometer increase in commuting distance, the likelihood of driving a car increases
by a factor of 1.27, and the likelihood of walking or cycling decreases by a factor of 0.31, in comparison to riding e-bike/motorcycle.
The estimated relative changes in probabilities for choosing these alternatives are 1.18 and 0.24 times, respectively, for non-com-
muting trips on weekends. These results indicate that people generally prefer car for long trips and walk/bicycle for short ones.
However, commuting distance does not show significant influence on the choice of bus as a travel mode.
The attitudinal factors show effects on certain mode choices. For people who support e-bike/motorcycle restriction, the likelihood
to walk or cycle increases by a factor of 1.36 for commuting trips on weekdays, and by a factor of 1.54 for non-commuting trips on
weekends. It means that when people have a negative attitude toward e-bike or motorcycle, their probability of traveling by walk or
bicycle increases significantly.
We tested land use diversity in both origin and destination districts as explanatory variables for mode choice on weekdays and
weekends. The land use index in working district shows significant associations with commuting mode choice. To be specific, if the
land use is more diversified in working district, people prefer bus to e-bike/motorcycle, and prefer e-bike/motorcycle to car for
commuting trips. The land use indices for both home and activity districts do not show statistical significance for non-commuting
trips. It implies that land use diversity in a small city does not influence people’s mode choice for leisure activities as much as in large
cities. Instead, trip distance and trip purpose are the main predictors of mode choice for leisure activities. When people go shopping
and participate in entertaining activities, they prefer walking/cycling to e-bike/motorcycle.
We included interaction terms between “the four types of areas” and “traveling within the same area” to test if traveling within
certain types of areas matters. These interaction terms help to better understand the effects of job-housing balance on mode choice.
The job-housing balance means a reasonable commuting distance/time from residential locations to job locations, which implies a
balanced housing and employment opportunities within a sub-region (Wang and Zhou, 2016). Job-housing imbalance usually results
in higher car usage. According to Table 6, for people who commute within the central urban area, the probability of choosing
walking/cycling compared to e-bike/motorcycle increases by a factor of 2.04, and the probability of driving a car compared to e-
bike/motorcycle increases by a factor of 5.85, in comparison to those commuting from the central urban area to other areas. It
indicates that promoting job-housing balance in the central urban area might stimulate both non-motorized and car commuting. The
results were not consistent with existing findings in large cities. Whether this is a typical small city feature needs additional case
studies. For people who commute within the industrial area, the probability of choosing walking/cycling compared to e-bike/mo-
torcycle increases by a factor of 4.1, in comparison to those commuting from the industrial area to other areas. It implies that job-
housing balance in the industrial area could encourage walking and cycling.
On weekends, for people who travel for leisure activities within the transition area, the probability of choosing bus compared to e-
bike/motorcycle decreases by a factor of 0.12, and the probability of driving compared to e-bike/motorcycle increases by a factor of
6.0, in comparison to those who travel from the transition area to other areas. Leisure trips originated from industrial area show
similar negative effects of within district travel on choosing bus and cross district travel on driving. These findings suggest that
although land use diversity does not show influential impacts on mode choices on weekends, the built environment in general in the

371
H. Hu et al. Transportation Research Part D 59 (2018) 361–374

transition and industrial areas makes residents prefer private motorized travel for within district leisure trips. The central urban area
does not show a significant effect on mode choices on weekends.

5. Conclusions and discussion

China’s rapid urbanization and motorization has captured much attention in the academic world. The large scale urban growth in
both spatial and demographic dimensions results in complicated transportation challenges. In recent decades, many researchers have
focused their work on transportation problems in large cities, and therefore, travel behavior and its connection with the built
environment in small cities remains largely unknown. Small cities have also been undergoing urban transformation since the eco-
nomic reform in the 1980s, although the scale is smaller and the speed is slower. It is important to gain a basic understanding of the
extent to which transportation implications of urban transformation in small cities differ from those in large cities. This paper
presents an effort to fill the knowledge gap by examining travel mode choice on both weekdays and weekends and its connection with
built environment factors in a small city, Changting.
The analysis divided the study area into four types of districts with different built environment characteristics: the central urban
area, the urban-suburban transition area, the industrial area, and the suburban area. Using survey data collected from families with
children and employing statistical methods, including multinomial logistic regression models, we obtained empirical results that can
help address the two research questions raised at the beginning of this paper.
First, as expected, travel behaviors in small cities of China show both similarities and differences in comparison to those observed
in large cities. Average trip distance for leisure activities is shorter than average commuting distance. As travel distance increases,
people shift their travel mode from low to high level of motorization. Due to the small geographic scale, trip distance in Changting is
shorter than it is in large cities. To be specific, average commuting distance is 2.6 km in Changting, while it ranges from 3.7 km to
11.9 km for people living in different types of neighborhoods in Beijing (Wang et al., 2011). Average trip distances for shopping and
entertaining activities are 1.5 km and 1.7 km respectively, compared to 1.98–4.64 km and 1.84–6.04 km respectively in Beijing
(Wang et al., 2011).
The combination of relatively short commuting distances and a relaxing working culture enables most workers in Changting to
take lunch break at home, resulting in an early afternoon peak hour of commuting that does not exist in large cities. Trip chaining is
less common, as chained trips account for only about 3.9% of all commuting trips and less than 1% of leisure trips. In comparison,
Yang et al. found that in Beijing 26.7% made complex commuting trips (home-work-other activities-home), and 71.7% of travelers
made complex trips on holidays.
Short trip distances and the hilly built environment also affect local residents’ choice of more flexible and personal travel modes.
E-bike and motorcycle are much more popular in Changting than in a typical large city, accounting for two thirds of commuting trips,
whereas public transportation plays a very limited role. On weekends, many people shift from e-bike and motorcycle to walk and
bicycle, and mode share for public transportation remains extremely low. It is worth pointing out that the low popularity of public
transportation can be interpreted differently. On one hand, it may indicate that people in small cities have low preferences for bus; on
the other hand, it might be because the low quality of the existing bus service, characterized by inadequate frequency and coverage,
restricts people’s choice.
Secondly, regression results show that attitudinal factors and built environment factors influence mode choice after controlling for
socio-economic and trip characteristics. Negative attitude toward e-bike and motorcycle may cause some people to shift from e-bike
and motorcycle to walk and bicycle. It may indicate that people suffer the most from e-bike/motorcycle choose walk/bicycle over e-
bike/motorcycle, or that people choosing walk/bicycle share values that is the antithesis of what is manifested by e-bike/motorcycle
while people who drive do not. (We took this latter viewpoint from one of the anonymous reviewers.) Planners and policy makers
should consider initiating more outreach programs to educate people about the likely social and environmental consequences of
increased dependence on cars and motorcycles. Such efforts may help reshape people’s attitudes towards, and subsequently choice of,
bus and other private motorized modes. These outreach programs can serve as complementary policy measures to good planning and
design. The function and form of public transportation service need to be carefully reconsidered, perhaps upgrading bus operation on
selected key routes but replacing other bus lines with flexible on-demand service.
In terms of built environment factors, land use diversity of the working district increases bus rides and reduces car driving on
weekdays. However, land use diversity does not show a significant influence on weekend trips. In addition, modal preferences do not
vary across different types of areas in Changting. This is different from observations in large cities, for example, Pan et al. (2009)
found that in Shanghai neighborhood type showed significant influences on mode choice for non-work travel, with walk/cycle as the
most favorite one in more traditional neighborhoods.
As in the large cities, economic development and urban expansion lead to jobs and population moving from urban central area to
industrial zones and suburbs in small cities. Consequently, job-housing relationships have been reshaped in different types of areas.
Our analysis show that better job-housing balance might have various effects on mode choices, depending on the type of the area.
Better job-housing balance might not only increase walking and cycling, but also promoting car usage and e-bike/motorcycle riding
in different areas. It is clear that small cities in China must also promote compact urban form with mixed land use to ensure a
sustainable future. The key issue is the transportation and land use policy needs to be differentiated based on the local context. Small
scale might need refined urban management. Due to the great popularity of e-bike and motorcycle and the increasing level of car
ownership, it is critical for small cities to investigate the pros and cons of e-bike/motorcycle in comparison to car, and to design
appropriate strategies for achieving more desirable outcomes. Given the great importance of walking and cycling as travel modes,
especially for weekends, major efforts should be made to improve pedestrian and bicycling facilities. In the meantime, the potential

372
H. Hu et al. Transportation Research Part D 59 (2018) 361–374

role of public transportation in small cities needs further exploration.

This study has several limitations. Our respondents were from families with children, and consequently the observed mode
choices do not represent the preferences of the whole population. A more inclusive sample to include all types of families in future
research would improve the understanding of travel behavior in small cities. Our questionnaire did not collect enough information
about education, employment status, and more comprehensive attitudinal variables, which might be useful explanatory variables of
mode choice. Moreover, it adopted a cross-sectional research design to test the associations between mode choice and explanatory
variables. The results provide little information on how changes in the built environment can contribute to people’s mode choice in a
long run. Panel data needs to be collected in the future to probe causality in travel behavior. Meanwhile, the single case study of
Changting, while informative, is also quite limited because it is only representative of some, but not all, small cities in China. Indeed,
several empirical results obtained from this case study are difficult to interpret, possibly due to unique local conditions. More studies
with data collected in a larger number of small cities will be essential to advance scholarship in this topic area.

Acknowledgments

This research was supported by the National Natural Science Foundation of China (Grant No. 41501169 and Grant No. 51278239)
and the Fundamental Research Funds for the Central Universities (Grant No. 090214380002).

Appendix A. Supplementary material

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.trd.2018.01.
013.

References

Ben-akiva, M.E., Bowman, J.L., 2000. Activity-based disaggregate travel demand model system with activity schedules. Transp. Res. Part A 1–28.
Ben-Akiva, M., McFadden, D., Gärling, T., Gopinath, D., Walker, J., Bolduc, D., Rao, V., 1999. Extended framework for modeling choice behavior. Marketing Lett. 10
(3), 187–203.
Bergman, Å., Gliebe, J., Strathman, J., 2011. Modeling access mode choice for inter-suburban commuter rail. J. Public Transport. 14 (4), 23–42.
Bhat, C., 1999. An analysis of evening commute stop-making behavior using repeated choice observations from a multi-day survey. Transport. Res. Part B: Methodol.
33 (7), 495–510. http://dx.doi.org/10.1016/S0191-2615(99)00003-X.
Bhat, C.R., Singh, S.K., 2000. A comprehensive daily activity-travel generation model system for workers. Retrieved from https://repositories.lib.utexas.edu/handle/
2152/23978.
Brunow, S., Gründer, M., 2013. The impact of activity chaining on the duration of daily activities. Transportation 40 (5), 981–1001. http://dx.doi.org/10.1007/
s11116-012-9441-6.
CDRI, 2015. Annual Report on Development of Small and Medium-sized Cities in China. By Small and Medium-sized Cities Development Research Institute (China
Small and Medium Cities Development Report).
Cao, X., 2009. Disentangling the influence of neighborhood type and self-selection on driving behavior: an application of sample selection model. Transportation 36
(2), 207–222.
Cervero, R., 2002. Built environments and mode choice: Toward a normative framework. Transport. Res. Part D: Trans. Environ. 7 (4), 265–284. http://dx.doi.org/10.
1016/S1361-9209(01)00024-4.
Chang, H.-L., Wu, S.-C., 2008. Exploring the vehicle dependence behind mode choice: evidence of motorcycle dependence in Taipei. Transport. Res. Part A: Policy
Practice 42 (2), 307–320. http://dx.doi.org/10.1016/j.tra.2007.10.005.
Chen, C., Gong, H., Paaswell, R., 2008. Role of the built environment on mode choice decisions: additional evidence on the impact of density. Transportation 35 (3),
285–299.
Chen, C.-F., Chen, C.-W., 2011. Speeding for fun? Exploring the speeding behavior of riders of heavy motorcycles using the theory of planned behavior and psy-
chological flow theory. Accid. Anal. Prev. 43 (3), 983–990. http://dx.doi.org/10.1016/j.aap.2010.11.025.
Chen, C.-F., Lai, W.-T., 2011. The effects of rational and habitual factors on mode choice behaviors in a motorcycle-dependent region: evidence from Taiwan. Transp.
Policy 18 (5), 711–718. http://dx.doi.org/10.1016/j.tranpol.2011.01.006.
Cherry, C., 2007. Electric Bike Use in China and Their Impacts on the Environment, Safety, Mobility and Accessibility. UC Berkeley Center for Future Urban Transport:
A Volvo Center of Excellence. Retrieved from http://escholarship.org/uc/item/8bn7v9jm.
Cherry, C.R., Weinert, J.X., Xinmiao, Y., 2009. Comparative environmental impacts of electric bikes in China. Transport. Res. Part D: Trans. Environ. 14 (5), 281–290.
http://dx.doi.org/10.1016/j.trd.2008.11.003.
Collins, C.M., Chambers, S.M., 2005. Psychological and situational influences on commuter-transport-mode choice. Environ. Behav. 37 (5), 640–661. http://dx.doi.
org/10.1177/0013916504265440.
Elias, W., Shiftan, Y., 2012. The influence of individual’s risk perception and attitudes on travel behavior. Transport. Res. Part A: Policy Practice 46 (8), 1241–1251.
http://dx.doi.org/10.1016/j.tra.2012.05.013.
Ewing, R., Cervero, R., 2010. Travel and the built environment. J. Am. Plann. Assoc. 76 (3), 265–294.
Feng, J., Chen, Y., 2010. Spatiotemporal evolution of urban form and land-use structure in Hangzhou, China: evidence from fractals. Environ. Plann. B: Plann. Des. 37
(5), 838–856. http://dx.doi.org/10.1068/b35078.
Gim, T.-H.T., 2016. Testing the reciprocal relationship between attitudes and land use in relation to trip frequencies: a nonrecursive model. Int. Regional Sci. Rev. 39
(2), 203–227. http://dx.doi.org/10.1177/0160017613512651.
Giuliano, G., 1983. Getting there: women and transportation. In: Zimmermann, J. (Ed.), The Technological Woman. Interfacing with Tomorrow, Praeger, New York,
pp. 102–112.
Handy, S., Cao, X., Mokhtarian, P., 2005. Correlation or causality between the built environment and travel behavior? Evidence from Northern California. Transp. Res.
Part D 10, 427–444.
Haybatollahi, M., Czepkiewicz, M., Laatikainen, T., Kyttä, M., 2015. Neighbourhood preferences, active travel behaviour, and built environment: an exploratory study.
Transport. Res. Part F: Traffic Psychol. Behav. 29, 57–69. http://dx.doi.org/10.1016/j.trf.2015.01.001.
Heinen, E., Maat, K., van Wee, B., 2011. The role of attitudes toward characteristics of bicycle commuting on the choice to cycle to work over various distances.
Transport. Res. Part D: Trans. Environ. 16 (2), 102–109. http://dx.doi.org/10.1016/j.trd.2010.08.010.
Hensher, D., Reyes, A., 2000. Trip chaining as a barrier to the propensity to use public transport. Transportation 27 (4), 341–361.
Hong, J., Shen, Q., Zhang, L., 2014. How do built-environment factors affect travel behavior? A spatial analysis at different geographic scales. Transportation 41 (3),
419–440.

373
H. Hu et al. Transportation Research Part D 59 (2018) 361–374

Islam, M.T., Habib, K.M.N., 2012. Unraveling the relationship between trip chaining and mode choice: Evidence from a multi-week travel diary. Transport. Plann.
Technol. 35(4), Retrieved from /view.aspx?id=1143664.
Kwoka, G.J., Boschmann, E.E., Goetz, A.R., 2015. The impact of transit station areas on the travel behaviors of workers in Denver, Colorado. Transport. Res. Part A:
Policy Pract. 80, 277–287.
Klöckner, C.A., Friedrichsmeier, T., 2011. A multi-level approach to travel mode choice – How person characteristics and situation specific aspects determine car use in
a student sample. Transport. Res. Part F: Traffic Psychol. Behav. 14 (4), 261–277. http://dx.doi.org/10.1016/j.trf.2011.01.006.
Krygsman, S., Arentze, T., Timmermans, H., 2007. Capturing tour mode and activity choice interdependencies: A co-evolutionary logit modelling approach. Transport.
Res. Part A: Policy Practice 41 (10), 913–933. http://dx.doi.org/10.1016/j.tra.2006.03.006.
Lan, T.T.N., Liem, N.Q., Binh, N.T.T., 2013. Personal exposure to benzene of selected population groups and impact of commuting modes in Ho Chi Minh, Vietnam.
Environ. Pollut. 175, 56–63. http://dx.doi.org/10.1016/j.envpol.2012.12.017.
Li, Z., Wang, W., Yang, C., Ragland, D.R., 2013. Bicycle commuting market analysis using attitudinal market segmentation approach. Transport. Res. Part A: Policy
Practice 47, 56–68. http://dx.doi.org/10.1016/j.tra.2012.10.017.
Ling, Z., Cherry, C.R., Yang, H., Jones, L.R., 2015. From e-bike to car: A study on factors influencing motorization of e-bike users across China. Transport. Res. Part D:
Trans. Environ. 41, 50–63. http://dx.doi.org/10.1016/j.trd.2015.09.012.
Liu, F., Yun, M., Liu, G., 2016. Will motorcyclist shift to bus or car in commuting: a case study of Zhongshan, China. Presented at the Transportation Research Board
95th Annual Meeting. Retrieved from https://trid.trb.org/view.aspx?id=1393958.
Ma, J., Mitchell, G., Heppenstall, A., 2014. Daily travel behaviour in Beijing, China: an analysis of workers’ trip chains, and the role of socio-demographics and urban
form. Habitat Int. 43, 263–273. http://dx.doi.org/10.1016/j.habitatint.2014.04.008.
Naess, P., 2010. Residential location, travel, and energy use in the Hangzhou Metropolitan Area. J. Trans. Land Use 3 (3). http://dx.doi.org/10.5198/jtlu.v3i3.98.
Pan, H., Shen, Q., Zhang, M., 2009. Influence of urban form on travel behaviour in four neighbourhoods of Shanghai. Urban Stud. 46 (2), 275–294. http://dx.doi.org/
10.1177/0042098008099355.
Primerano, F., Taylor, M.A.P., Pitaksringkarn, L., Tisato, P., 2008. Defining and understanding trip chaining behaviour. Transportation 35 (1), 55–72. http://dx.doi.
org/10.1007/s11116-007-9134-8.
Schwanen, T., Mokhtarian, P.L., 2005. What if you live in the wrong neighborhood? The impact of residential neighborhood type dissonance on distance traveled.
Transp. Res. Part D 10, 127–151. http://dx.doi.org/10.1016/j.trd.2004.11.002.
Senbil, M., Kitamura, R., Mohamad, J., 2009. Residential location, vehicle ownership and travel in Asia: a comparative analysis of Kei-Han-Shin and Kuala Lumpur
metropolitan areas. Transportation 36 (3), 325–350.
Shen, Q., Chen, P., Pan, H., 2016. Factors affecting car ownership and mode choice in rail transit-supported suburbs of a large Chinese city. Transport. Res. Part A:
Policy Practice 94, 31–44. http://dx.doi.org/10.1016/j.tra.2016.08.027.
Song, Y., Merlin, L., Rodriguez, D., 2013. Comparing measures of urban land use mix. Comput. Environ. Urban Syst. 42, 1–13. http://dx.doi.org/10.1016/j.
compenvurbsys.2013.08.001.
Sun, B., Zhang, T., He, Z., Wang, R., 2015. Urban spatial structure and motorization in China. n/a-n/a. J. Regional Sci. http://dx.doi.org/10.1111/jors.12237.
Susilo, Y.O., Maat, K., 2007. The influence of built environment to the trends in commuting journeys in the Netherlands. Transportation 34 (5), 589–609. http://dx.
doi.org/10.1007/s11116-007-9129-5.
Tana, Kwan, M.-P., Chai, Y., 2015. Urban form, car ownership and activity space in inner suburbs: A comparison between Beijing (China) and Chicago (United States).
Urban Stud. http://dx.doi.org/10.1177/0042098015581123. 42098015581123.
Van Acker, V., Witlox, F., 2010. Car ownership as a mediating variable in car travel behaviour research using a structural equation modelling approach to identify its
dual relationship. J. Transport Geogr. 18 (1), 65–74.
Wan, Z., Wang, X., Sperling, D., 2013. Policy and politics behind the public transportation systems of China’s medium-sized cities: Evidence from the Huizhou reform.
Utilities Policy 27, 1–8. http://dx.doi.org/10.1016/j.jup.2013.07.002.
Wang, D., Chai, Y., 2009. The jobs–housing relationship and commuting in Beijing, China: the legacy of Danwei. J. Transp. Geogr. 17 (1), 30–38. http://dx.doi.org/10.
1016/j.jtrangeo.2008.04.005.
Wang, D., Chai, Y., Li, F., 2011. Built environment diversities and activity–travel behaviour variations in Beijing, China. J. Trans. Geogr. 19 (6), 1173–1186. http://dx.
doi.org/10.1016/j.jtrangeo.2011.03.008.
Wang, D., Lin, T., 2014. Residential self-selection, built environment, and travel behavior in the Chinese context. J. Transp. Land Use 7 (3), 5–14.
Wang, D., Zhou, M., 2016. The built environment and travel behavior in urban China: a literature review. Transport. Res. Part D: Trans. Environ. https://doi.org/10.
1016/j.trd.2016.10.031.
Wang, J., Wang, X., 2015. New urbanization: A new vision of China's urban–rural development and planning. Front. Arch. Res. 4 (2), 166–168. http://dx.doi.org/10.
1016/j.foar.2015.03.003.
Wei, Y.D., 2015. Zone fever, project fever: Development policy, economic transition, and urban expansion in China. Geogr. Rev. 105 (2), 156–177. http://dx.doi.org/
10.1111/j.1931-0846.2014.12063.x.
Weinert, J., Ogden, J., Sperling, D., Burke, A., 2008. The future of electric two-wheelers and electric vehicles in China. Energy Policy 36 (7), 2544–2555. http://dx.doi.
org/10.1016/j.enpol.2008.03.008.
Yang, L., Shen, Q., Li, Z., 2016. Comparing travel mode and trip chain choices between holidays and weekdays. Transport. Res. Part A: Policy Practice 91, 273–285.
http://dx.doi.org/10.1016/j.tra.2016.07.001.
Yang, M., Wang, W., Chen, X., Wan, T., Xu, R., 2007. Empirical analysis of commute trip chaining: case study of Shangyu, China. Transport. Res. Record: J. Transport.
Res. Board 2038, 139–147. http://dx.doi.org/10.3141/2038-18.
Ye, M., Yu, M., Guo, X., Liu, Y., Li, Z., Ye, M., Li, Z., et al., 2013. Analysis on residents travel activity pattern in historic urban areas: A case study of historic urban area
of Yangzhou, China. Math. Problems Eng., Math. Problems Eng. 2013, e620973. http://dx.doi.org/10.1155/2013/620973, DOI: 10.1155/2013/620973.
Ye, M., Yu, M., Li, Z., Yin, F., Hu, Q., 2014. Modeling the commuting travel activities within historic districts in Chinese Cities. Comput. Intell. Neurosci., Comput.
Intell. Neurosci. 2014, e253289. http://dx.doi.org/10.1155/2014/253289, DOI: 10.1155/2014/253289.
Ye, X., 2004. An exploration of the relationship between mode choice and complexity of trip chaining patterns. Graduate Theses and Dissertations. Retrieved from
http://scholarcommons.usf.edu/etd/1312.
Ye, X., Pendyala, R.M., Gottardi, G., 2007. An exploration of the relationship between mode choice and complexity of trip chaining patterns. Transp. Res. Part B 1 (41),
96–113. http://dx.doi.org/10.1016/j.trb.2006.03.004.
Zhang, Y., Li, Y., Yang, X., Liu, Q., Li, C., 2013. Built environment and household electric bike ownership: Insights from Zhongshan Metropolitan Area, China.
Transport. Res. Record: J. Transport. Res. Board 2387, 102–111. http://dx.doi.org/10.3141/2387-12.
Zhao, J., Xiao, L., Tang, L., Shi, L., Su, X., Wang, H., Shao, G., et al., 2014. Effects of spatial form on urban commute for major cities in China. Int. J. Sustain. Dev. World
Ecol. 21 (4), 361–368. http://dx.doi.org/10.1080/13504509.2014.922132.

374