Spatial and Social Characteristics
of Urban Transportation in Beijing
Jiawen Yang
This paper aims to improve knowledge of urban transportation in China
by analyzing the spatial characteristics of urban transportation in Beijing,
China’s capital. Neighborhood-level indicators—such as automobile ownership per household, commuting time, commuting mode percentages,
and household income—were all extracted from a household survey conducted in 2006. Urban transportation was then analyzed spatially with a
reference to Beijing’s geometric center, job centers, and urban rail stations. The analysis led to several findings. First, the variation in commuting time by gender and household composition was similar to what has
been observed in American cities. Second, automobile ownership and
commuting modes exhibited no systematic, spatial variation, unlike their
American counterparts. These differences might be explained by a third
finding: in China, relatively high-income households preferred centrally
located housing.
STUDY CASE
Study Region
Beijing lies on the northeastern margin of the North China Plain.
The spatial organization of this city is characterized by the continuous addition of ring roads. Today Beijing has five rings completed
for public use. The first ring circles the Forbidden City (the palace
of the emperors). Because the area has no dense residential areas and
few employees, the ring does not carry heavy regional traffic, and
has remained local. All other ring roads were added to either mitigate existing traffic congestion or to open suburban land to urban
development. Construction of the second ring began in the 1950s
and then completely changed and became an urban expressway in
1992. The third ring road was completed in 1994. The fourth and
fifth ring roads were added in 2001 and 2003, respectively. Figure 1
presents the locations of the five ring roads.
Today Beijing hosts more than 10 million residents at its core
built-up area, which is served by the five ring roads and other connecting roads. The second ring roughly delineates the boundary of
the “old Beijing,” developed before the communist party took power
in 1949. Development in the 1980s mainly occurred around or within
the third ring and seldom went beyond the fourth ring. Since the
1990s, the fourth and fifth rings have enabled conversion of rural and
semirural areas into urban landscapes, stimulated by a strong demand
for housing and office space.
The expansion of the city has increased trip length as travelers
look for employment and entertainment opportunities throughout
the ever-enlarging urban region. Relocation from the inner city out
into what were once semirural areas has also increased the spatial
separation between workplaces and residences (7 ). Large volumes
of traffic increase road congestion as the capacity limit of existing
transport facilities is exceeded. It has been reported that bus speed
is about 10 km/h during peak periods.
Urban rail (surface or subway) in Beijing provides an alterative for
those who want reliable mobility but cannot afford a car. The subway system in Beijing opened to the public in 1981 (8). At that time,
the length of the system was 27.6 km, and it included only 19 stops.
By 2008, this system had eight lines, more than 200 km of tracks, and
123 stations. At least seven more lines, with 164 km of track length,
are now under construction. Figure 1 shows three rail lines. The other
five are omitted because they were completed later and are not
directly relevant to the transportation pattern described in this article.
Urban transportation in China has undergone dramatic transition. An
increase in automobile ownership, an increase in reliance on motorization for everyday trips, and an increase in congestion and air pollution in China’s megacities are well-discussed topics. A recent study
suggests that China is following the motorization path of developed
economies (1). The increase in automobile ownership in relation to
the growth in China’s gross domestic product is similar to that in the
United States, South Korea, Japan, and Germany.
As the macro trend becomes increasingly clear, researchers have
begun to pay attention to the social details of China’s urban transportation (2). Little, however, is known about the spatial details of
China’s urban transportation, partially because of the lack of relevant data. How does automobile ownership vary from the central
city to suburban areas? How does transportation mode vary among
different neighborhoods? How does commuting time vary among
different social groups? Is it similar to what has been observed in
Western cities? A better understanding of these spatial and social
characteristics of urban transportation in China can help assess
its current transportation and urban development policies (3), particularly since Chinese cities are still experiencing significant spatial transformations (4), which could have significant impacts on
mobility and accessibility (5, 6). Toward this end, this article presents a spatial analysis of urban transportation in Beijing, China’s
capital.
Program of City and Regional Planning, Architecture Building East, Room 204,
Georgia Institute of Technology, Atlanta, GA 30332-0155. jiawen.yang@
gatech.edu.
Data and Method
Transportation Research Record: Journal of the Transportation Research Board,
No. 2193, Transportation Research Board of the National Academies, Washington,
D.C., 2010, pp. 59–67.
DOI: 10.3141/2193-08
Research was enabled by the availability of a key household survey
conducted in 2006. The survey’s data were collected by John Logan
59
60
Transportation Research Record 2193
FIGURE 1
Major transportation infrastructure in study region.
and his research team at Brown University (9). The survey was conducted to understand the dynamics of housing markets, household
location, and travel behavior. Coded survey results recorded detailed
information on vehicle ownership, commuting duration and modes,
residential location, and workplace location. The 2006 survey covered
48 urban neighborhoods, with 25 households surveyed in each neighborhood. These neighborhoods were in the central city, the inner suburbs, and the outskirts beyond the fifth ring road. Figure 2 shows the
location of the surveyed neighborhoods; the size of each dot represents
the average household income in each neighborhood. The survey data
were used to develop a group of neighborhood-level indicators. Analysis of these indicators can show how car ownership and individual
travel behavior vary from one neighborhood to the other.
Being cognizant of significant suburban development and urban
rail investment in Beijing, the spatial characteristics of urban transportation were analyzed from multiple perspectives. Three distance variables were introduced to define the location of each
surveyed neighborhood: distance to Beijing’s geometric center, distance to the closest job centers, and distance to the closest subway
station.
Tian’anmen Square, which has always been viewed as the civic
center of Beijing, was selected as the geometric center for this study.
A geographic information system (GIS) was used to calculate the
distance from every surveyed neighborhood to this geometric center. Examination of the association between urban transportation
indicators and the distance to the geometric center revealed to what
extent a monocentric model can help interpret urban transportation
patterns.
To test whether the formation of job centers in the suburbs
affects urban transportation, the top 10 subdistricts for job density
were selected. Figure 3 shows the location of these top 10. Job
counts in each subdistrict originated from the 2002 work unit census. The selected job centers were generally located on the second,
third, and fourth ring roads. The selected subdistricts included
well-known job centers, such as Zhongguancun and Jianguomeng.
GIS was used to calculate the distance from every surveyed neighborhood to the centroid of the nearest job center. Examination of
how urban transportation indicators varied according to this distance revealed to what extent a multicenter perspective could help
explain Beijing’s transportation patterns.
The distance was calculated from each surveyed neighborhood
to the closest urban rail station. Examination of how transportation
indicators varied according to this distance revealed to what extent
access to high-end transit might affect the spatial patterns of urban
transportation.
Correlation of population and job density with the three distance
variables indicated that the above distance variables provided reasonable perspectives to interpret the urban spatial structure. As shown in
Table 1, all three categories of distance had significant impacts on spatial distribution of residences and workplaces: the shorter the distances,
the higher the density. A lower distance value, therefore, indicated a
relatively central location. Comparable correlation coefficients for the
Atlanta, Georgia, metropolitan statistical area in the United States suggest a similar pattern (Table 2). Both tables include correlation coefficients for mode split and commuting time, which are explained later in
this paper.
Yang
61
FIGURE 2
FIGURE 3
Location and average household income of surveyed neighborhoods.
Ten subdistricts of highest job density.
62
TABLE 1
Transportation Research Record 2193
Correlation Coefficients with Three Distance Variables
TABLE 2 Comparable Correlation Coefficients in the Atlanta
Metropolitan Area
Distance
Distance
Variable
Density
Population density 1982
Population density 1990
Population density 2000
Job density 2001
Income
Average household income
Vehicle ownership
Average number of motor
vehicles per household
Mode split
Percentage of walking
Percentage of biking
Percentage of transit
Percentage of driving
Duration
Commuting time
To Job
Centers
To
Geometric
Center
To Urban
Rail
Station
−.523**
−.639**
−.678**
−.579**
−.637**
−.723**
−.706**
−.599**
−.521**
−.616**
−.633**
−.532**
−.346*
−.260
−.250
.188
.128
.060
.116
.130
−.278
−.049
.218
.073
−.246
−.031
−.091
.314*
−.332*
−.022
−.145
−.318*
−.124
**Correlation is significant at the .01 level (two-tailed).
*Correlation is significant at the .05 level (two-tailed).
AUTOMOBILE OWNERSHIP
The increase in motor vehicles has been the most dynamic aspect of
China’s urban transportation development. Figure 4 illustrates the
growth of registered motor vehicles in Beijing. Two distinct phases
characterize Beijing’s motorization. The pre-1980 period saw almost
no increase, even though the city’s urban area had expanded, and
urban residents had tripled from 1949. After the mid-1990s, motor
vehicles increased at an annual rate of more than 10%. This fast
growth of automobile ownership can be attributed to many factors. As
in other developing countries, economic development and income
growth were influential factors (10). Changing travel demands as a
result of land and housing reforms also played a significant role, as
Yang reported in 2006. According to 2006 survey data, the average
household owned 0.23 cars that year.
In the United States, it is widely known that automobile ownership
is lower in central cities, because there is access to transit in highdensity areas, and a mixed-use, built-environment reduces the demand
Variable
Density
Population density
Job density
Mode split
Driving alone
Carpool
Transit
Nonmotor
Duration
Commuting time
To Job
Centers
To Geometric
Center
To Urban
Rail Station
−.557**
−.211**
−.510**
−.236**
−.523**
−.196**
.453**
−.070
−.438**
−.260**
.575**
−.112**
−.595**
−.323**
.497**
−.084*
−.503**
−.254**
.221**
.143**
.183**
NOTE: Data are from the 2000 U.S. census. The analysis units are census tracts.
The geometric center is the census tract with the highest job density in
downtown. The job centers include all census tracts with job density higher
than 2,000 jobs per square kilometer.
**Correlation is significant at the .01 level (two-tailed).
*Correlation is significant at the .05 level (two-tailed).
for automobile ownership (11). In Chinese cities, however, it has
never been clear whether neighborhoods with high automobile ownership are clustered in a certain urban area. In Beijing, a visual examination of the surveyed neighborhoods suggests a significant spatial
variation of household vehicle ownership among different neighborhoods (Figure 5). No systematic spatial variation can be observed,
however, and it is difficult to say whether it follows a certain spatial
pattern. In a correlation of vehicle ownership with the three distance
variables (Table 1), none of the variables can significantly explain the
spatial variation of household vehicle ownership.
This spatial characteristic could be easily misinterpreted. It might
be assumed that household automobile purchasing decisions are not
affected by land use intensity, which decreases from a central location
to the periphery (Table 1). The spatial pattern, however, might be
caused by the spatial characteristics of other relevant factors such as
household income. To test this idea, a regression model was developed of vehicle ownership per household. The model used three independent variables. The first two were average household income (in
thousands) of the surveyed neighborhood and the population density
(1,000/sq km) of the subdistrict where the neighborhood was located.
Because the spatial relationship between workplace and residence
3.5
3
Millions
2.5
2
1.5
1
0.5
0
1940
1950
1960
1970
1980
Year
FIGURE 4
Registered motor vehicles in Beijing.
1990
2000
2010
Yang
63
FIGURE 5
Average household motor vehicle ownership in surveyed neighborhoods.
was an important factor that affected travel demand (12), the third
variable was controlled, namely, the percentage of households that
lived and worked in the same subdistrict. The analysis units were the
48 surveyed neighborhoods.
cars household = 0.227 + 0.005 ⴱ income
( 2.673) (3.439)
− 0.0059 ⴱ population_density
( −2.593)
− 0.154 ⴱ balance_percentage
( −1.967)
and density, therefore, work in opposite directions to shape the
spatial pattern of car ownership.
It is widely known that high-income Americans are more likely to
live in the suburbs. Why are relatively high-income Chinese households more likely to live in central locations? The question is beyond
the scope of this paper, but a likely answer has to do with the spatial
variation of the quality of public goods. In China, central locations
are generally equipped with better schools, better access to shopping
and entertainment opportunities, and better urban infrastructure. The
single-family house in the outer suburbs is attractive in terms of
housing characteristics. Its distance and the associated road congestion that separate it from the city, however, have disqualified the suburban house as a primary residence for high-income households,
even for typical automobile owners.
R 2 = .249
The regression results were no surprise. Higher ownership of motor
vehicles is significantly associated with higher household income,
lower population density, and reduced spatial balance between workplace and residence. Further examination of the spatial variation of
the population density and household income can help explain why
automobile ownership does not show a systematic spatial variation.
Population density tends to be higher at central locations (Table 1),
which reduces automobile ownership there. Household income, however, tends to increase automobile ownership in central locations
because higher income households are more likely to live there, as
suggested by the negative correlation coefficient in Table 1. Income
COMMUTING MODE
Motor vehicles are not owned only as a matter of pride but also to use
for everyday trips. The 2006 survey showed a dramatic difference in
commuting modes between car owners and nonowners (Figure 6).
Households with cars used them to make 48% of their trips. Compared with households without cars, the car owners were less likely
to commute by walking, biking, and transit riding. Commuting by
urban rail, however, was actually higher among car owners than
nonowners, which indicated that car owners could substitute driving
trips with urban rail.
64
Transportation Research Record 2193
60%
50%
40%
car owners
30%
non-owners
20%
10%
0%
walk
FIGURE 6
bike
fixed route bus employer bus
car or taxi
subway
Commuting mode split for car owners and nonowners.
A comparison of the 2006 survey with a similar survey conducted
in 1996 (12) indicated that the overall commuting mode pattern had
changed significantly in the context of motorization (Figure 7). From
1996 to 2006, the biking share had decreased more than 10%, partially
because of cyclists’ decreased road safety and exposure to polluted
air. Despite the decrease, however, biking remained the primary mode
of commuting and accounted for 43% of commuting trips. Driving
(including taxi) as a share of the commuting mode was about 10%,
much lower than the corresponding numbers in American urban
regions such as Atlanta (95%) and Boston, Massachusetts (84%).
(The Boston and Atlanta percentages were calculated on the basis of
the 2006 Public Use Microdata Samples (PUMS) data set.)
A unique and interesting feature of Beijing’s urban transport is that
walking is the third primary mode of commuting, with a mode share
of 15%. Its increase from 1996, according to Yang et al., was related
to the housing location preference of a group of relatively highincome households (13). They chose to live close to their workplaces
and to walk to their offices.
Similar to what has already been observed about automobile ownership, no spatial variation of commuting modes was obvious. Any
obvious spatial pattern of mode split could hardly be detected from
Figure 8. The correlation analysis (Table 1) can confirm this. The
only two significant coefficients were associated with the distance
to the closest urban rail stations: the closer to the stations, the higher
the percentage by transit and the lower the percentage of biking.
This association makes sense, because better access to urban rail services tends to reduce the chance of commuting by bike and increase
the chance of commuting by transit. Again, this makes Beijing’s urban
transportation different from what has been observed in Atlanta, where
commuting mode split varied systematically according to distances to
central locations (Table 2).
No American-style, suburb–central city division exists in Beijing.
Given Beijing’s urban development history, the study area was divided
into two parts—those within the third ring and those beyond. Since
those residences beyond the third ring were mainly developed during
the motorization period, this area was treated as equivalent to the suburbs in a U.S. metropolitan area. Delineation of the central city from
the suburbs was different from that of existing studies of population
suburbanization in Beijing (14), which have treated areas beyond the
second ring as suburbs. This difference was justified by the purpose of
the research, which was to study the spatial characteristics of urban
transportation. To delineate the city–suburb boundary on the basis of
the stage of motorization made sense in this context. In addition, the
grouping split the surveyed neighborhoods almost in half, which
allowed for the largest possible sample size in each subset.
The null hypothesis was then tested that commuting mode selection was statistically the same between the central city and the suburbs. The statistics testing did not reject the null hypothesis. The test
60%
1996
50%
2006
40%
30%
20%
10%
0%
walk
bike
bus
exclusive
motorized modes
others
FIGURE 7 Commuting mode split in Beijing: 1996 and 2006; exclusive motorized modes include
private auto, taxi, and motorcycle.
Yang
65
FIGURE 8
Commuting mode split in surveyed neighborhoods.
result was definitive. A further test was conducted to see whether
neighborhoods that had high or low percentages of motorized trips
clustered together. A spatial analysis conducted with the Moran Index
(use of an inverse distance neighborhood) showed the percentage of
driving was –0.06, or close to random. The percentage of transit rides
was 0.04, which was also likely to be random. The test for the average number of vehicles per household was 0.03, which had a 5%
chance of being random.
The same tests were run with Atlanta’s 2000 census data, by using
census tracts as the analysis units. The null hypothesis could be
rejected at a 99% confidence level. That is to say, a statistical difference between driving and transit use by commuters existed between
those that resided in the city of Atlanta and those that resided outside
it. The Moran index (use of inverse distance for neighborhoods) for
transit use was 0.35%, with less than a 1% chance of randomness.
Similarly, the index for driving alone was 0.31%, with a less than 1%
chance of randomness. The above statistical testing suggests no significant differences between urban transportation among those that
reside in the central part of the city and those that reside in the surrounding suburbs. This spatial characteristic makes Beijing different
from its Western counterpart.
COMMUTING TIME
Commuting time, which is a measure of access to the workplace, has
always been an important measure of urban transportation. Table 3
presents average commuting time for different social groups. The
2006 Beijing survey data can be compared with the 1996 survey and
TABLE 3 Average Commuting Time by Social Groups
and Mode
Commuting Time (min)
Beijing
Group
1996
2006
Atlanta
2006
Boston
2006
Average
Housing ownership
Tenant
Owner
Gender
Female
Male
College education
No
Yes or above
Household composition
One worker
Biworker
Multiworker
Mode
Walk
Bike
Fixed-route bus
Employer bus
Driving
Urban rail
39.0
36.5
31.5
28.2
38.4
39.1
35.7
38.7
30.8
31.1
28.4
27.6
40.0
38.0
35.7
37.1
29.7
33.1
26.2
30.0
38.5
40.0
34.6
39.0
31.2
32.0
25.3
31.4
40.1
37.3
44.6
38.9
35.1
38.5
32.8
31.7
29.8
29.7
29.2
25.2
13.0
31.1
63.9
42.4
33.9
14.6
26.3
57.8
54.6
33.1
50.3
12.8
19.8
56.9
12.3
24.5
42.4
30.7
56.7
26.5
47.6
NOTE: Atlanta and Boston indicators are extracted from PUMS, U.S. Census
Bureau.
66
with the 2006 Boston and Atlanta PUMS data. Several things can be
observed.
First, the gender difference in average commuting time for Beijing
reversed itself from 1996 to 2006, which made it similar to Atlanta
and Boston. In 1996, the commuting duration of female workers was
2 min longer than for male workers. In 2006, however, the average
commuting time for male workers was 1.4 min longer than for female
workers. This change can be probably explained by the change in
housing location selection. Before 1996, the housing market was not
well developed, and employers still played an important role in housing subsidy and location selection. If a household had a working husband and a working wife, the husband was more likely to have a
higher status than the wife and therefore was more likely to receive
housing support from his work unit. The couple’s housing location
thus was typically closer to the husband’s workplace than the wife’s,
which led to a shorter commuting time for the husband. As market
power has penetrated, however, residential selection and housing
consumption have come to depend more on the housing market and
household income. Housing location is now more likely to be closer
to the wife’s workplace because of her relatively lower physical
strength and her larger share of responsibilities for childcare and
housework.
In a market economy, housing location decisions become increasingly separated from the collective decision of the work unit. Households have gained more autonomy in residence selection. Under this
new circumstance, couples are more likely to select a residence convenient for the wife so that she can have more time and energy to
cook, clean and take care of children. In contrast, the husband is
expected to make a longer commute, given his relatively greater
physical strength and lesser involvement in childcare and housework.
Second, the number of workers in a household also has effects on
commuting time. A dual-worker household tends to have a shorter
commuting time than a single-worker household. Those households
with three or more workers, however, commute longer, which indicates a difficulty in balancing home and work location. The Atlanta
and Boston data indicate that the more workers in a household, the
shorter the commuting time. The result appears to challenge the traditional assumption that workers from dual-worker households find it
more difficult to achieve a balance between workplace and residence
and thus tend to commute longer distances than those in single-worker
households. Sultana once explained this contradiction with housing
affordability (15). Even though multiworker households tend to have
more difficulty than single-worker households in the achievement of a
balance between workplace and residence, the lack of affordable housing could overshadow this difficulty. The more workers a household
has, the higher the income it can earn, and the more likely it is that its
members can afford a better location to live. Given Beijing’s high
housing price relative to income, it is no surprise that this affordability
problem is reflected in commuting time.
Third, commuting time in Beijing according to modes has become
similar to its American counterparts. The two primary commuting
modes in Beijing are biking and fixed route bus. Average commuting
time by bike decreased from 31.1 min in 1996 to 26.3 min in 2006,
which made it more comparable to that in Boston. The shortening of
biking commuting time could be caused by the replacement of long
biking trips with motorized trips. The average commuting time by
fixed-route bus decreased from 63.9 min to 57.8 min, which was close
to the 56.9 min commuting duration in Atlanta.
Despite the observable increase in congestion in Beijing, the survey
data do not reveal any increase in average commuting time. This could
be caused by the imperfection of the data. The size of the sample
Transportation Research Record 2193
TABLE 4 Average Commuting Time by
Age of Neighborhood
Post-2000
1990s
1980s
1970s
Beijing
2006
Atlanta
2006
Boston
2006
29.4
33.1
39.4
39.3
35.5
32.1
31.2
29.4
32.0
30.1
28.9
29.0
NOTE: Atlanta and Boston indicators are extracted
from PUMS, U.S. Census Bureau. Commuting time
is in minutes.
may be too small; however, it could also reflect the actual situation.
Households could respond to congestion by switching to highermobility options, such as driving. In addition, they can select to live
closer to the workplace (13). Furthermore, the continuous addition of
the ring roads, particularly the fourth and fifth rings, has helped to
reduce congestion. As suggested by the correlation coefficients of
commuting time (Table 1), the commute from a suburban location on
average tends to be shorter than it is from a central location. This could
be a Chinese version of the American commuting paradox (16). It also
confirms an observation made by Yang (17): the effect of decentralized development on commuting depends on the format of suburban
development. Beijing’s high density and clustered development in
suburban areas might have helped to reduce commuting time.
Fourth, incremental urbanization has not necessarily lengthened
commuting time in Beijing. In American metropolitan areas, a close
association exists between commute duration and age of neighborhood. The newer the neighborhood, the further away it is from the
center of the city, and the longer the commute (18). In Beijing,
neighborhoods developed in the 1970s and thereafter have a reversed
pattern of commuting duration: the newer the neighborhood, the
shorter the commute time (Table 4). As for mode split, Table 5 indicates that the newer the neighborhood, the higher the percentage of
nonmotorized modes (including walking and biking).
Beijing’s difference can be partially attributed to the spatial characteristics of its new housing construction. In the United States, redevelopment and infill development in established urban areas are
generally limited, compared with opportunities for new development
in suburban areas. Therefore, new neighborhoods are more likely
to be located in peripheral areas and require longer commutes and a
higher percentage of driving. In China, significant redevelopment
occurs in the central part of the city. Consequently, a relative higher
percentage of new neighborhoods are actually developed in the central city. Higher density and mixed use there enable a higher percentage use of nonmotorized modes and create the convenience of being
able to reach destinations within a short travel time. As for new neighborhoods in suburban Beijing, the addition of ring roads enables those
TABLE 5
Mode Split by Age of Neighborhood in Beijing
Age of
Neighborhood
Post-2000
1990s
1980s
Pre-1980
Car or
Taxi (%)
Nonmotorized
(%)
Walking
(%)
10.0
18.5
13.8
11.6
58.0
49.7
47.4
53.9
18.0
13.0
10.3
8.9
Yang
that rely on motorized modes to reach their destinations fast, despite
the possibility of a long commuting distance.
CONCLUSION
Since the late 1980s, land and housing reforms have reshaped urban
space in China (4, 6). Economic growth has increased motorization.
This article presents an updated view of the spatial and social details
of China’s urban transportation, which stems from the existing
knowledge base. Information extracted from a household survey
indicates that Beijing is becoming similar to its Western counterparts in terms of driving behavior by automobile owners, and access
to workplaces according to different commuting modes and different social groups. This trend toward similar outcomes is no surprise,
as individuals make decisions on vehicle ownership, motorized
travel, and residential location selection on the basis of a similar set
of constraints and objectives.
Despite the trend, China’s urban transportation shows distinct
spatial characteristics. The survey data reveal no American style, city–
suburb division of urban transportation. They show no significant nor
systematic variation in vehicle ownership and commuting modes by
distance to central locations. The regression model of automobile
ownership suggests that residence location preference of high-income
households could help explain Beijing’s spatial characteristics of
urban transportation. Relatively high-income households are more
likely to live in central locations, which increase automobile ownership there. It is still unknown if this spatial difference will last long. It
could disappear as China continues to change its planning practices
for land development, urban transportation, and public goods supply.
ACKNOWLEDGMENTS
This research was supported by the Lincoln Institute of Land Policy. The author thanks John Logan for the survey data, and Ge Song
for valuable research assistance.
REFERENCES
1. Schipper, L., and W. S. Ng. Urban Transport Options in China: The
Challenge to Choose. EMBARQ. World Resources Institute, Center for
Transport and Environment. http://www.embarq.org/en/urban-transportoptions-china-the-challenge-choose.
67
2. Zegras, P. C., and S. Srinivasan. Household Income, Travel Behavior,
Location, and Accessibility: Sketches from Two Different Developing
Contexts. In Transportation Research Record: Journal of the Transportation Research Board, No. 2038, Transportation Research Board of
the National Academies, Washington, D.C., 2007, pp. 128–138.
3. Gakenheimer, R. Social Effects of Potential Motorization Pathways. In
Personal Cars and China (Chinese Academy of Engineering and National
Research Council of the National Academies) National Academies Press,
Washington, D.C., 2003, pp. 131–149.
4. Zhou, Y. X., and L. Ma. Economic Restructuring and Suburbanization
in China. Urban Geography, Vol. 21, No. 3, 2000, pp. 205–236.
5. Shen, Q. Urban Transportation Problems in Shanghai, China: Problems
and Planning Implications. International Journal of Urban and Regional
Research, Vol. 21, No. 4, 1997, pp. 589–606.
6. Yang, J. W., and R. Gakenheimer. Assessing the Mobility and Accessibility Consequences of Land Use Transformation in Urban China. Habitat
International, Vol. 31, 2007, pp. 345–353.
7. Yang, J. Transportation Implications of Land Development in a Transitional Economy: Evidence from Housing Relocation in Beijing. In Transportation Research Record: Journal of the Transportation Research
Board, No. 1954, Transportation Research Board of the National Academies, Washington, D.C., 2006, pp. 7–14.
8. Major Events for Beijing Subway: 1981–1990. http://www.bjsubway.
com/zjdt/dtdsj/2007-05-15/2976.shtml. Accessed July 1, 2009.
9. Urban China Research Network, Multi-City Study of Urban China.
http://www.s4.brown.edu/ChinaProject/.
10. Gakenheimer, R. Urban Mobility in the Developing World. Transportation Research Part A, Vol. 33, 1999, pp. 671–689.
11. Newman, P., and J. Kenworthy. Sustainability and Cities: Overcoming
Auto Dependence. Island Press, Washington, D.C., 1999.
12. Li, S. M. Life Course and Residential Mobility in Beijing, China. Environment and Planning A, Vol. 36, No. 1, 2004, pp. 27–43.
13. Yang, J. W., J. Feng, and R. Gakenheimer. The Physical and Social
Dimension of Job-Housing Balance in Urban China. In Smart Growth
in China, Chapter 11. Lincoln Institute of Land Policy Press, Cambridge,
Mass., 2009.
14. Zhou Y. X. On the Suburbanization of Beijing. Chinese Geographical
Science, Vol. 7, No. 3, 1997, pp. 208–219.
15. Sultana, S. Effects of Married-Couple Dual-Earner Households on Metropolitan Commuting: Evidence from the Atlanta Metropolitan Area.
Vol. 26, No. 4, 2005, pp. 328–352.
16. Gordon, P., H. Richardson, and M.-J. Jun. The Commuting Paradox:
Evidence from the Top Twenty. Journal of the American Planning
Association, Vol. 57, 1991, pp. 416–420.
17. Yang, J. W. The Commuting Impacts of Spatial Decentralization: A
Comparison of Boston and Atlanta. Journal of Regional Analysis and
Policy, Vol. 35, 2006, pp. 69–78.
18. Sultana, S., and J. Weber. Journey-to-Work Patterns in the Age of Sprawl:
Evidence from Two Midsize Southern Metropolitan Areas. The Professional Geographer, Vol. 59, No. 2, 2006, pp. 193–208.
The Transportation in the Developing Countries Committee peer-reviewed this
paper.