J. Service Science & Management, 2009, 2: 80-91
Published Online June 2009 in SciRes (www.SciRP.org/journal/jssm)
Copyright © 2009 SciRes JSSM
Research on Urban Logistics Infrastructure:
An Empirical Study of China
Changbing Jiang, Daqiang Chen
College of Computer and Information Management, Zhejiang Gongshang University, Hangzhou, China.
Email: johncabin@mail.zjgsu.edu.cn, chendaqiang@mail.zjgsu.edu.cn
Received February 26th, 2009; March 26th, 2009; accepted April 20th, 2009.
ABSTRACT
Urban Logistics Infrastructure (ULI) is an important area of urban competition capability. The connotation of Urban
Logistics Capability (ULC) is analyzed in this paper. Compared with the extensive research on ULC in developed world,
empirical work is still rare in China. In this paper the theory of ULC is firstly overviewed. Then a new evaluation index
system for ULC evaluation is set up which contains factors that reflect the market supply and demand, economic devel-
opment and transportation accessibility. Secondly, an empirical study is carried out by using Hierarchical Cluster
Analysis (HCA) and Principal Component Analysis (PCA) method to classify ULC into 3 clusters for 30 cities in Peo-
ple's Republic of China. Thirdly, according to the characteristics of the 3 clusters, suggestions are proposed for im-
proving their ULI. Finally, after comparing different ULC of 30 cities in People's Republic of China, this paper focuses
on that different logistics infrastructure including Hub, Central Distribution Center & Cross Docking Center, Regional
Distribution Center or Distribution Center should be build reasonably in order to meet the customer’s requirement in
the 3 different cluster cities.
Keywords: urban logistics, performance, hierarchical cluster analysis, principal component analysis
1. Introduction
With the accession into the WTO, modern logistics in
China possess the great development opportunity [1].
Shanghai, Hong Kong and Guangzhou, some of the ma-
jor gates to the outside world in China, plan to build
themselves into major international logistics center in 5
to 10 years [2]. Therefore the research of Urban Logis-
tics Infrastructure (ULI) has recently become a hot topic
in the logistics area.
Urban Logistics Infrastructure (ULI) including Logis-
tics Hub, Central Distribution Center, Cross Docking
Center, Regional Distribution Center and Distribution
Center is an important area of Urban Logistics Capability
(ULC) [3]. Mentzer and Konrad reviewed urban logistics
performance measurement practices from an efficiency
and effectiveness perspective [4]. Much more attention is
paid to freight transport on an interurban level, due to the
evolution of supply chain analysis, but this attention is
basically devoted to cost factors, which are to be mini-
mized in order to improve the efficiency of the urban
logistics system. However, ULI which is an important
component of ULC should be re-engineered in order to
improve the effectiveness of the urban logistics system
[5]. From 1952 to 2003, the large-scale city has in-
creased from 9 to 49. Therefore, we need to construct
different ULI according to each city’s logistics capabil-
ity.
This paper is organized into 5 sections. In Section 1, a
brief description of logistics for metropolitan cities in
China is introduced. Hierarchical Cluster Analysis (HCA)
and Principal Component Analysis (PCA) method are
explained in Section 2. After comparing and analyzing
different evaluation system of ULC and overview of
ULC theory, a new ULC evaluation system is proposed
in Section 3, which is composed of market supply and
demand, economic development and transportation ac-
cessibility. In Section 4, ULC is classified into 3 clusters
for 30 sample cities in China using Hierarchical Cluster
Analysis (HCA) and Principal Component Analysis
(PCA) method. In the final section conclusions of the
study is summarized, and further research for this study
is suggested.
2. Research Methodologies
2.1 The Method of Hierarchical Cluster Analysis
Clustering is one of the most important and primitive
activities of human beings, dating back to Aristotle.
CHANGBING JIANG, DAQIANG CHEN 81
Given a set of data objects (also known as patterns, enti-
ties, instances, observances, or units), cluster analysis
aims to explore natural and hidden data structure and to
provide insights to the questions such as, “Are there
any clusters (groups, subsets, or categories) in the data,
and if yes, how many clusters are in the data?” More
specifically, supposing we have a set of N data objects
with d features (attributes, dimensions, or variables)
1
Xx,,x,,x
jN
 , where
j12 d
xx,x,,x d
jj j

,
we have the following mathematical descriptions of two
types of clustering [6]:
1) Hard partitional clustering attempts to seek a
K-partition of X, , such that:

)(,,
1NKCCC K
;,1, KiCi
;X
1
i
K
iCY
.,,1,, jiKjiCC ji
2) Hierarchical clustering attempts to construct a tree-
like nested structure partition of X,
1,,
Q
H
HH
l
H
, such that , and m>l imply
or
(QN
i
CC
)
j
,
imj
CHC
ij
CC

for all ij . ,,,1,,imlQ
As aforementioned, clustering is generally classified
as partitional clustering or hierarchical clustering based
on the properties of clusters generated [7]. Partitional
clustering directly partitions data objects into some
pre-specified number of clusters, while hierarchical
clustering groups data with a sequence of nested parti-
tions, either from singleton clusters to a cluster including
all individuals or vice versa. The former is known as
agglomerative hierarchical clustering, and the latter is
called divisive hierarchical clustering. As the binary di-
vision of data is computationally expensive, we will fo-
cus on agglomerative hierarchical clustering, which is
more commonly used in practice.
Agglomerative hierarchical clustering generates a re-
sult, which is depicted by a binary tree or dendrogram,
based on the proximity matrix. The root node of the den-
drogram represents the entire data set, and each leaf node
is regarded as a data object. The intermediate nodes thus
describe the extent to which the objects are proximal to
each other, and the height of the dendrogram usually
expresses the distance between each pair of data objects
or clusters, or a data object and a cluster. The ultimate
clustering results can be obtained by cutting the dendro-
gram at different levels. This representation provides
very informative descriptions and visualizations for the
potential data clustering structures, especially when real
hierarchical relations exist in the data.
More specifically, for a data set with N samples, gen-
eral agglomerative hierarchical clustering can be sum-
marized by the following procedure:
1) start with N singleton clusters Ci (i=1,…, N) and
calculate the proximity matrix for these N clusters;
2) in the proximity matrix, search the minimal dis-
tance D(Ci, Cj)=minD(Cm, Cl) (1m,l,N, ml), where
D(· , ·) is the distance function, and combine cluster Ci
and Cj to form a new cluster Cij;
3) update the proximity matrix by computing the dis-
tances between the cluster Cij and the other clusters; and
4) repeat steps 2~3 until only one cluster remains.
2.2 The Method of Principal Component Analysis
The method used to derive the component scores using
ten indicators for reflecting Urban Logistics Capability
(ULC) is Principal Component Analysis (PCA). PCA
transforms the original set of variables into a smaller set
of linear combinations that account for most of the varia-
tions of the original set. The principal components are
extracted so that first principal component denoted by
PC(1) accounts for the largest variation in the data.
Let us consider the variables X1, X2, ..., Xp. A principal
component analysis of this set of variables can generate
p new variables, known as the principal components,
PC1, PC2, ... , PCp. The principal components can be
expressed as follows:
PC1 = b11X1 +...+ b1pXp=Xb1
.
.
.
PCp=bp1X1 +...+ bppXp=Xbp
or, in general,
PC=Xb
where b’s are the coefficients for principal component
and each column of b contains the coefficients for one
principal component. Here, the coefficient for PC1 is
chosen such that it’s variance is the largest, and PC2 is
chosen to have the second largest variance subject to the
condition that PC1 and PC2 are uncorrelated, and so on.
For any principal component, the coefficients of princi-
pal components are chosen such that 2
1
'1
p
ijj j
i
bbb
.
Now, if we consider that the sample variance-covariance
matrix of the original variables, X, is Sx then the coeffi-
cient vector, bj, can be obtained by solving the following
equations:
0
x
SIb
where λ is the vector of characteristic roots and b is a
matrix comprising of the characteristic vectors corre-
Copyright © 2009 SciRes JSSM
CHANGBING JIANG, DAQIANG CHEN
82
sponding to each characteristic root [8]. There may be p
characteristic roots, some of which may be zero if there
are linear dependence among the original variables, X. It
may be noted here that PC1 is computed by using the
characteristic vector corresponding to the largest charac-
teristic root, λ1, similarly, PC2 is computed by using
characteristic vector corresponding to the second largest
characteristic root, λ2, and so on.
It must be stressed that a principal component analysis
does not always work in the sense that a large number of
original variables are reduced to a small number of
transformed variables. Indeed if the original variables are
uncorrelated then the analysis does absolutely nothing.
The best results are obtained when the variables are cor-
related, positively or negatively [9]. One merit of PCA is
that an increase in the number of variables that one may
wish to include for deriving a composite index imposes
very little cost on the analysis and one can include many
related variables for deriving the principal components
[10].
3. Evaluation System of Urban Logistics
Capability
Lu and Yang identified the key logistics capabilities in-
dicator for international distribution center operators,
based on five key logistics capabilities including cus-
tomer response, innovation, economic scale, flexible
operation and logistics knowledge [11]. Zhang re-
searched the theory of the location planning for logistics
park and set up a new index system for logistics park
performance evaluation [12].
The comprehensive evaluation on Urban Logistics
Capability (ULC) needs a synthetic evaluation system
that takes factors as much as possible into consideration
to release the objective evaluation for different impacts
of different factors on selection of the logistics facility
location, i.e. the planning of Urban Logistics Infrastruc-
ture. In general, we select 10 factors which are grouped
and stated as the following:
3.1 Supply and Demand Capability
3.1.1 Overall Economy Level
Here we adopt Gross Domestic Product (GDP) for this
factor. GDP refers to the final products at market prices
produced by all resident units in the region during a cer-
tain period of time. Generally the higher this factor is in
a given region, the more feasible a synthetic logistics
facility locates in this region.
3.1.2 Industry Developing Level
Here we adopt Number of State-owned and state-holding
Enterprises (NSSE) for this factor. NSSE refer to
state-owned enterprises plus state-holding enterprises.
State-owned enterprises (originally known as state-run
enterprises with ownership by the whole society) are
non-corporate economic entities registered in accordance
with the Regulation of the People’s Republic of China on
the Management of Registration of Legal Enterprises,
where all assets are owned by the state.
3.1.3 Retail Market Level
Here we adopt Consumption Expenditure of Urban
Households (CEUH) for this factor. CEUH refers to total
expenditure of the sample households for consumption in
daily life, including expenditure on eight categories
such as food, clothing, household appliances and ser-
vices, health care and medical services, transport and
communications, recreation, education and cultural ser-
vices, housing, miscellaneous goods and services.
3.1.4 Foreign Trade Developing Level
Here we use the Number of Foreign Funded Enterprises
(NFFE) for this factor. Foreign trade enterprise is the
main participator of logistics out-sourcing service in
China, which associates with a huge amount of interna-
tional logistics service, thus the higher this factor is in a
given region, the more feasible a synthetic logistics facil-
ity locates in this region.
3.1.5 Urban Freight Traffic and Turnover Volume
Level
Here we adopt Freight Traffic (FT) for this factor. FT
refers to the volume of freight transported with various
means within a specific period of time. This indicator
reflects the service of the transport industry towards the
national economy and people’s living conditions, as well
as an important indicator used in formulating and moni-
toring transport production plans and research into the
scale and pace of transport development. Freight trans-
port is calculated in tons. Freight transport is calculated
in terms of the actual weight of the goods and takes no
account of the type of freight and distance of travel.
Generally, the huger the volume is in a given region, the
more feasible for a synthetic logistics facility locates in
this region.
3.2 Economic Development Capability
3.2.1 Average Economy Level
Here we adopt GDP per capita (GDPpC) for this factor.
GDPpC refers to the GDP of the region divided by the
population of the region. Generally the higher this factor
is in a given region, the more feasible a synthetic logis-
tics facility locates in this region.
3.2.2 Social Reproduction
Here we adopt Total Investment in Fixed Assets in the
region (TIFA) for this factor. TIFA refers to the volume
of activities in construction and purchases of fixed assets
and related fees, expressed in monetary terms. It is a
Copyright © 2009 SciRes JSSM
CHANGBING JIANG, DAQIANG CHEN
Copyright © 2009 SciRes JSSM
83
comprehensive indicator which shows the size, structure
and growth of the investment in fixed assets, providing
basis for observing the progress of construction projects
and evaluating results of investment. Total investment in
fixed assets in the whole country includes, by type of
ownership, the investment by the state-owned units, col-
lective units, individuals, joint ownership units, share-
holding units, as well as investment by businessmen
from foreign countries and from Hong Kong, Macao and
Taiwan, and by other units.
3.3 Transportation Development Level
3.3.1 Accessibility of Railway
Here we adopt Railway Density (RWD) for this factor.
RWD is computed by Length of Railway (LRW) divided
by area of the region. LRW refers to the total length of
the trunk line under passenger and freight transportation
(including both full operation and temporary operation).
The calculation is based on the actual length of the first
line even if this line has a full or partial double track or
more tracks, excluding double tracks, station sidings,
tracks under the charge of stations, branch lines, spe-
cial-purpose lines and the non-payable connecting lines.
The length of railway in operation is an important indi-
cator to show the development of the infrastructure for
the railway transport, and also the essential data to cal-
culate volume of passenger freight transport, traffic den-
sity and utilization efficiency of the locomotives and
carriages. RWD is a better way to indicate the accessibil-
ity of given region. Because different level logistics fa-
cilities need a excellent transport network to facilitate its
logistics service, railway density index is feasible for this
purpose, which is calculated by length of railroad lines in
service divides the region total land area.
3.3.2 Accessibility of Roads
Here we adopt Highway Density (HWD) for this factor.
HWD is computed by Length of Highway (LHW) di-
vided by area of the region. LHW refers to the length of
highway which are built in conformity with the grades
specified by the highway engineering standard formu-
lated by the Ministry of Communications, and have been
formally checked and accepted by the departments of
highway and put into use. HWD is another better way to
indicate the accessibility of given region, Because dif-
ferent level logistics facilities need an excellent transport
network to facilitate its logistics service: highway net-
work density index is feasible for this purpose, which is
calculated by total length of highway network in service
divides the region total land area.
3.3.3 Transportation Capacity
Here we adopt Possession of Civil Motor Vehicles
(PCMV) for this factor. PCMV refer to the total numbers
of vehicles that are registered and received vehicles li-
cense tags according to the Work Standard for Motor
Vehicles Registration formulated by transport manage-
ment office under department of public security at the
end of reference period. They are divided into following
categories according to the structure of motor vehicles:
passenger vehicles, trucks and others; and private vehi-
cles and vehicles for units use according to ownerships;
working vehicles and non-working vehicles according to
kind of usage; large passenger vehicles, medium pas-
senger vehicles, small passenger vehicles and mini pas-
senger vehicle, heavy trucks, light-heavy trucks, light
trucks and mini trucks according to sizes of vehicles.
Based on above analysis, we may obtain ULC evalua-
tion index system for macro level logistics facility plan-
ning, as showed in Table 1.
4. Classification of ULC
4.1 Sample Cities Selection and Data Statistics
Due to imperfect evaluation index system for logistics
Table 1. Evaluation index system for ULC
First-grade factor Second-grade factor Indicator Abbreviation
Overall economy level X1Gross Domestic Product (billion yuan) GDP
Industry developing level X2Number of State-owned and State-holding Enter-
prises (unit) NSSE
Retail market level X3Consumption Expenditure of Urban Households
(billion yuan) CEUH
Foreign trade developing level X4Number of Foreign Funded Enterprises (unit) NFFE
Supply and demand
capability
Urban freight traffic and turn-
over volume level X5Freight Traffic (million tons) FT
Average economy level X6GDP per Capita (yuan/person) GDPpC
Economic develop-
ment capability Social reproduction X7Total Investment in Fixed Assets (billion yuan) TIFA
Accessibility of railway X8Railway Density (km/1,000km2) RWD
Accessibility of roads X9Highways Density (km/1,000km2) HWD
Transportation acces-
sibility level
Transportation capacity X10Possession of Civil Motor Vehicles (1,000 unit) PCMV
Source: Stanley E.Fawcett (1997) [13]; David J. Closs, Thomas J. Goldsby and Steven R. Clinton (1997) [14]; Edward A. Morash, Cornelia L.M.
Droge, Shawnee K. vichery (1996) [15]; Daniel F.Lynch, Scott B.Keller, John Ozment (2000) [16], arranged by author.
CHANGBING JIANG, DAQIANG CHEN
84
static in China, and the statistical indicators are inade-
quate in Urban Statistical Yearbook of China, it is im-
possible for the sample we selected to contain all the
influence indicators, i.e. to take factors from the avail-
able statistical data into consideration to covering influ-
ence indicators that we discussed as much as possible.
According to the difference inherent attributes of dif-
ferent city and distribution channel, 30 major cities in
China are selected, which exclude Hong Kong, Macao,
Taiwan and Lhasa for Statistical Indicator and method
difference. 30 major cities include:
(1) North China economic region with Beijing as the
center, covering Tianjin, Shijiazhuang, Taiyuan,
Hohhot;
(2) Northeast China economic region with Shenyang as
the center, covering Changchun, Harbin;
Table 2. Synthetic scores of 30 cities in China
Indicator
City
X1 X2 X3 X4 X5 X6 X7 X8a X9b X10
Beijing 428.3 4324 197 696 299.9 37058 252.8 68.7 893.7 176.9
Tianjin 293.2 5378 105 1010 372.8 31550 125.9 57.0 905.9 119.3
Shijiazhuang 163.3 1555 53 504 117.5 17871 71.7 25.0 375.3 706.4
Taiyuan 64.0 521 23 12 166.8 18804 33.5 20.1 420.9 324.4
Hohhot 51.2 203 16 17 55.4 26321 31.5 5.5 66.3 240.6
Shenyang 190.1 1836 81 309 150.4 27487 97.1 28.7 360.8 388.9
Changchun 153.5 580 50 90 115.8 21285 46.0 18.7 245.6 172.1
Harbin 168.0 765 71 49 99.4 17463 53.3 12.4 148.5 239.3
Shanghai 745.0 12316 245 2938 687.1 55307 308.5 41.8 1237.7 188.2
Nanjing 191.0 2165 71 272 169.4 33050 120.2 15.9 775.2 443.6
Hangzhou 251.5 5607 70 413 189.0 38858 120.5 12.2 459.9 518.6
Hefei 59.0 492 24 135 57.7 13378 36.1 16.8 511.3 298.3
Fuzhou 154.8 2345 58 406 90.4 23444 52.7 11.9 461.1 221.8
Nanchang 77.0
632 23 38 42.0 17238 35.2 13.6 369.8 181.0
Jinan 161.9 1510 69 107 150.3 27610 65.1 21.3 506.9 721.7
Zhengzhou 137.8 1801 56 51 86.7 21233 61.3 24.7 457.2 473.1
Wuhan 195.6 1403 96 94 170.5 24963 82.2 13.4 481.7 297.9
Changsha 113.4 1149 53 48 110.7 18036 66.8 13.4 413.7 295.5
Guangzhou 411.6 4727 168 541 352.0 56271 134.9 12.3 629.4 1010.7
Nanning 58.9 561 24 22 67.9 9126 26.1 11.6 252.1 184.3
Haikou 25.3 231 10 60 34.7 17928 11.9 11.3 614.3 54.4
Chongqing 266.5 2634 106 120 381.7 9608 162.2 8.7 391.9 146.2
Chengdu 218.6 1872 88 123 181.7 20777 108.5 6.1 233.4 406.6
Guiyang 44.4 674 18 31 57.4 12683 29.3 10.7 261.7 145.7
Kunming 94.2 672 37 11 116.5 18773 43.5 6.1 435.1 359.8
Xi'an 109.6 776 51
64 110.0 14081 64.0 15.3 256.0 190.2
Lanzhou 50.5 950 23 9 57.9 16479 23.2 5.7 100.6 88.8
Xining 17.5 163 7 7 23.7 8484 9.9 1.5 39.2 43.9
Yinchuan 18.9 226 7 14 20.1 17668 17.2 15.3 240.2 57.8
Urumqi 48.4 301 20 16 120.8 22820 17.6 1.7 53.2 201.5
Note: a computed by Total Length of Railways divide by area of the province and, b computed by Total Length of Highways divide by area of the
province.
Source: National Bureau of Statistics of People's Republic of China (2005); China Statistical Yearbook (2005); China City Statistical Yearbook
(2005), arranged by author.
Copyright © 2009 SciRes JSSM
CHANGBING JIANG, DAQIANG CHEN 85
(3) East China economic region with Shanghai as the
center, covering Nanjing, Hangzhou, Hefei, Fuzhou,
Nanchang, Jinan;
(4) Central China economic region with Guangzhou as
the center, covering Zhengzhou, Wuhan, Changsha,
Nanning, Haikou;
(5) Southwest China economic region centered in
Chongqing, covering Chengdu, Guiyang, Kunming,
Lhasa excluded;
(6) Northwest China economic region centered in Xi'an,
covering Lanzhou, Xining, Yinchuan, Urumqi.
Based on National Bureau of Statistics of People's
Republic of China (2005), China Statistical Yearbook
(2005) and China City Statistical Yearbook (2005), ten
indicators of the evaluation index system is analyzed
above as statistical variables, we have the data showed in
Table 2.
4.2 Hierarchical Cluster Analysis of ULC
Hierarchical cluster analysis is a statistical method for
finding relatively homogeneous clusters of cases based
on measured characteristics. The aim is to maximize
between-group variance and to minimize within-group
variance. It starts with each case in a separate cluster and
then combines the clusters sequentially, reducing the
number of clusters at each step until only one cluster is
left. In this paper, we apply the Hierarchical cluster
method in Statistical Package for Social Sciences to
analysis the 30 major cities. The final result is showed in
Figure 1.
According to Figure 1, Shanghai and Guangzhou are
in one cluster district with rescaled distance cluster com-
bine between 7 and 24, i.e. these two cities have the first
cluster with high logistics capacity, and the rest cities are
in the other cluster. For rescaled distance cluster com-
bine between 4 and 7, these 30 cities can be classified
into 3 clusters, Shanghai and Guangzhou as the first
cluster, Beijing, Hangzhou, Tianjin and Nanjing as the
second cluster, and the 24 rest city as the third cluster.
And for rescaled distance cluster combine as 3, the third
cluster can be classified into 2 clusters, and then the 30
sample city can be classified into 4 clusters.
4.3 Principal Component Analysis of ULC
Since most of the indicators suffer from simultaneity and
multi-colinearity, Principal Component Analysis (PCA)
is best suited for removing such difficulties because it
maximizes the variance rather than minimizing the least
square distance where any other technique (such as re-
gression analysis) fails to do so.
In this paper, we apply the “Factor Analysis” method
in Statistical Package for Social Sciences (SPSS) to
analysis the 30 major cities. The results shown in Table 3
suggest a two-factor solution. The eigenvalues clearly
show that only two common factors are present by using
the criterion of “eigenvalue greater than 1” and it is fur-
ther confirmed by the fact that the break point occurs at
the three eigenvalue of the scree plot (see Figure 2). This
being the case, the two-factor solution would appear to
be acceptable. Table 4 shows the two Component load-
ings. From Table 4 we see that RWD and PCMV carried
less weight than the others in case of ranking of cities.
We also find that the first component explains 73.550%,
and the second component explains 11.692% of the total
variation in the data. Since both the eigenvalue of the
first component and the second component (in case of
ten variables) are greater than 1, in the present case the
two components are used to calculate component score
for each city to determine the ranking of selected cities.
The two Principal Components (PC) explain about
85.242% of the variations in the level of ULC. The vari-
ables like GDP, NSSE and CEUH played a major role in
classifying the cities in terms of ULC compared to the
variables such as RWD and PCMV.
In order to calculate the ranking of the selected cities,
the principal components can be expressed as follows:
PC1=0.362*X1+0.351*X2+0.351*X3+0.332*X4
+0.345*X5+0.308*X6+0.35*X7+0.258*X8
+0.318*X9+0.093*X10
PC2=0.039*X1-0.032*X2+0.053*X3-0.172*X4
-0.042*X5+0.34*X6-0.067*X7-0.31*X8
-0.097*X9+0.859*X10
PC=0.317*X1+0.298*X2+0.31*X3+0.263*X4
+0.292*X5+0.313*X6+0.293*X7+0.18*X8
+0.261*X9+0.198*X10
Based on the Hierarchical Cluster Analysis and Prin-
cipal Component Analysis of ULC mentioned above, we
can calculate the ranking of the selected cities. Table 5
shows the ranking of the selected cities based on Princi-
pal Component (PC) scores as well as ranking based on
GDP. Figure 1 shows the result of the hierarchical cluster
analysis.
From Figure 1 and Table 5, We can classify ULC into
3 clusters among the selected cities in People's Republic
of China. For the purpose of analysis conveniently, here
we define the first cluster as the high LIC cities, the sec-
ond cluster as the medium ULC cities, the third cluster as
the low ULC cities. The high ULC cities including
Shanghai, Beijing, Guangzhou and Tianjin, the medium
Copyright © 2009 SciRes JSSM
CHANGBING JIANG, DAQIANG CHEN
86
Figure 1. Hierarchical cluster analysis of 30 major cities
Table 3Eigenvalues of the correlation matrix (ten variables)
Initial Eigenvalues Extraction Sums of Squared Loadings
Component
Total % of Variance Cumulative % Total % of Variance Cumulative %
1 7.355 73.550 73.550 7.355 73.550 73.550
2 1.169 11.692 85.242 1.169 11.692 85.242
3 .608 6.082 91.324
4 .343 3.430 94.754
5 .206 2.058 96.813
6 .159 1.590 98.403
7 .078 .785 99.187
8 .052 .517 99.705
9 .026 .257 99.962
10 .004 .038 100.000
Source: Calculated by the author.
ULC cities including Hangzhou, Nanjing, Chongqing,
Shenyang, Jinan, Wuhan, Shijiazhuang and Chengdu. the
low ULC cities including Zhengzhou, Fuzhou, Changsha,
Changchun, Harbin, Taiyuan, Kunming, Xi’an, Hefei,
Nanchang, Haikou, Hohhot, Urumqi, Guiyang, Nanning,
Lanzhou, Yinchuan and Xining.
From Table 5, we see that, Shanghai topped the list in
terms of ULC followed by Beijing and Guangzhou. For
the rate of GDP and PC score are quite high, Shanghai is
the front runner among the selected cities. The high ULC
cities are almost in Bohai Bay region (including Beijing
and Tianjin), Yangtze River Delta Region (Shanghai)
and Pearl River Delta Region (Guangzhou). The low
ULC cities are almost among the Northwest China eco-
nomic region (including Urumqi, Lanzhou, Yinchuan,
Xining, etc.). as illustrated in Table 6 and Figure 3.
Copyright © 2009 SciRes JSSM
CHANGBING JIANG, DAQIANG CHEN 87
Table 4. Component loadings (Eigenvectors)
Component
PC1 PC2
GDP 0.981 0.042
NSSE 0.952 -0.035
CEUH 0.953 0.058
NFFE 0.901 -0.186
FT 0.935 -0.046
GDPpC 0.836 0.367
TIFA 0.949 -0.073
RWD 0.698 -0.335
HWD 0.862 -0.105
PCMV 0.252 0.929
Figure 2. Scree plot of eigenvalues
Source: Calculated by the author Note: * Only factors with eigenvalues > 1 are retained
Table 5. Ranking of 30 cities in China on indicators of environment (using 10 variables)
name PC1
score
PC1
ranking
PC2
score
PC2
ranking
PC
score
PC
ranking GDP GDP
ranking
Shanghai 10.209 1 -1.310 28 8.629 1 745 1
Beijing 5.041 2 -1.483 29 4.146 2 428.3 2
Guangzhou 3.891 3 3.731 1 3.869 3 411.6 3
Tianjin 3.659 4 -1.760 30 2.916 4 293.2 4
Hangzhou 1.607 5 1.311 3 1.566 5 251.5 6
Nanjing 1.100 6 0.736 5 1.050 6 191 9
Chongqing 0.886 7 -0.840 25 0.649 7 266.5 5
Shenyang 0.536 8 0.273 10 0.500 8 190.1 10
Jinan 0.254 9 1.765 2 0.461 9 161.9 13
Wuhan 0.199 10 0.212 11 0.201 10 195.6 8
Shijiazhuang -0.001 11 1.258 4 0.172 11 163.3 12
Chengdu -0.033 12 0.712 6 0.069 12 218.6 7
Zhengzhou -0.325 14 0.567 7 -0.203 13 137.8 16
Fuzhou -0.284 13 -0.194 16 -0.272 14 154.8 14
Changsha -0.815 15 0.014 13 -0.701 15 113.4 17
Changchun -0.960 16 -0.405 18 -0.884 16 153.5 15
Harbin -1.091 17 -0.055 15 -0.949 17 168 11
Taiyuan -1.104 18 0.003 14 -0.952 18 64 21
Kunming -1.208 20 0.451 8 -0.980 19 94.2 19
Xi'an -1.197 19 -0.498 20 -1.101 20 109.6 18
Hefei -1.400 21 -0.229 17 -1.239 21 59 22
Nanchang -1.614 22 -0.424 19 -1.451 22 77 20
Haikou -1.785 23 -0.943 27 -1.670 23 25.3 28
Hohhot -2.008 24 0.351 9 -1.685 24 51.2 24
Urumqi -2.077 27 0.175 12 -1.769 25 48.40 26
Guiyang -2.025 25 -0.606 23 -1.830 26 44.4 27
Nanning -2.038 26 -0.560 22 -1.836 27 58.9 23
Lanzhou -2.183 28 -0.542 21 -1.958 28 50.5 25
Yinchuan -2.232 29 -0.877 26 -2.046 29 18.90 29
Xining -2.999 30 -0.832 24 -2.702 30 17.50 30
Source: Calculated by the author
Copyright © 2009 SciRes JSSM
CHANGBING JIANG, DAQIANG CHEN
88
Table 6. Profile of LIC
Region High
P
C
2.90
Medium
0.00PC<2.90
Low
PC<0.00
North China economic region Beijing, Tianjin Shijiazhuang Taiyuan, Hohhot
Northeast China economic region Shenyang Changchun, Harbin
East China economic region Shanghai Jinan, Hangzhou,
Nanjing
Fuzhou, Hefei,
Nanchang
Central China economic region Guangzhou Wuhan Zhengzhou, Changsha,
Haikou, Nanning
Southwest China economic region Chongqing, Chengdu Kunming, Guiyang
Northwest China economic region
Xi'an, Urumqi,
Lanzhou, Yinchuan,
Xining
Source: Arranged by the author
The result of classification shows that there are three
basic logistics capacity clusters, which is useful for ap-
plication in China. The between-group variance of the
third and fourth cluster is the smallest, i.e. logistics ca-
pacities in cities belong to these two clusters are similar
to each other and can be modified as one cluster. While
the between-group variance of the first, second and third
cluster have significant difference, i.e. cities belong to
these three clusters, respectively, are different in the lo-
gistics capacity.
For the purpose of analysis conveniently, here we de-
fine the first cluster city as the Hub/CDC (Hub/ Central
Distribution Center & Cross Docking Center) city, the
second cluster as the CDC/RDC (Central Distribution
Center & Cross Docking Center / Regional Distribution
Center) and the third cluster as the RDC/DC (Regional
Distribution Center / Distribution Center) city, as illus-
trated in Figure 3.
5. Urban Logistics Network Structure
Logistics network is an integration of organizations and
facilities within the logistics procedure, and also is an
integration network that consists of physical network of
product that flows between facilities such as hub, central
distribution center, regional distribution center and other
kinds of logistics facility, and network of information
that flows parallel with the product flow. The aims of
urban logistics network structure is to define the location
and number of logistics facilities and control the system-
atic cost with certain service supply level by network
analysis and optimal methods.
5.1 Category of Urban Logistics Facility
There are four categories urban logistics facility in China,
such as hub, central distribution center, regional distribu-
tion center and distribution center.
5.1.1 Hub
According to DOD in USA, hub refers to an organization
that sorts and distributes inbound cargo from wholesale
supply sources (airlifted, sealifted, and ground trans-
portable) and/or from within the theater. Suppliers can
arrange material and product in hub to supply the large
hub or logistics center in service destination by long dis-
tance transportation to concentrate the supply, take ad-
vantage of common transport and combined loading,
improve the logistics active efficiency and productivity,
and decrease the procurement and supply cost.
5.1.2 CDC (Central Distribution Center & Cross
Docking Center)
Cross Docking Center is the facility where the material
or products are received from suppliers, sorted directly to
be shipped to a consolidated batch (often including other
orders from other suppliers) to the customers by the
same vehicle or different without putting them in storage.
Its particular advantages reside at the minimization of
warehousing and economies of scale in outbound flows
(from the distribution center to the customers), and it
helps reduce operating costs, increase throughput, re-
duces inventory levels, and helps in increase of sales
space. The material or products handled in CDC are usu-
ally of large-size, small-item, and low-frequency.
5.1.3 RDC (Regional Distribution Center)
A Regional Distribution Center (RDC) is a collection
and consolidation center for finished goods, components
and spare parts to be distributed to the distribution center
belongs to dealers, importers or other unrelated organi-
zations within or outside the region. Among the func-
tions involved are information network service, repack-
aging and labeling, and distribution. The material or
products handled in CDC are usually of small-size, mul-
Copyright © 2009 SciRes JSSM
CHANGBING JIANG, DAQIANG CHEN 89
tiple-item, and personality.
5.1.4 DC (Distribution Center)
Distribution Centers are foundation of urban logistics
network, which usually is a model “warehouse” or other
specialized building which is stocked with products to be
re-distributed to retailers or wholesalers. In the urban lo-
gistics network discussed in this paper, up-level facilities
will ship truckloads of products to the distribution center,
and then the distribution center will store the product until
needed by the retail location and ship the proper quantity
to the retails, stores, even the final consumers.
5.2 Planning of Urban Logistics Facility
Based on above quantitative analysis of the sample data,
and qualitative analysis of four categories urban logistics
facility, we locate different facility in different city clus-
ters, and draw up different developing policy respectively.
5.2.1 Developing Policy for Hub/CDC City
In Hub/CDC city, four tier logistics network is designed,
in which goods is transported from the Hub to the CDC,
and then distributed to RDC, finally reaches DC, as il-
lustrated in Figure 4.
There is a relatively perfect operation system in
Hub/CDC city, the focus of urban logistics infrastructure
planning is on enhancing the improvement and integra-
tion of the logistics systematic function, strengthening the
mechanization, automatization and informationization,
improving the capacity of handling and efficiency of dif-
ferent logistics facility, developing multimode transporta-
tion, facilitating efficient connection between facilities
and transport line, optimizing operation process to control
the cost.
In practice, advanced operation model should be in-
troduced, i.e. synthetic logistics facility should be built to
attract supplier of components and spare parts, and inte-
grated operation system with supply-produce-sale and
quick response should be instructed.
Note: cluster 1: Hub/CDC city, includes 4 cities: Shanghai, Beijing, Guangzhou and Tianjin;
cluster 2: CDC/RDC city, includes 8 cities: Hangzhou, Nanjing, Chongqing, Shenyang, Jinan, Wuhan,
Shijiazhuang and Chengdu;
cluster 3: RDC/DC city , includes 18 cities: Jinan, Hohhot, Harbin, Urumqi, etc.
Figure 3. Classification of 30 major cities in China according to ULC
Copyright © 2009 SciRes JSSM
CHANGBING JIANG, DAQIANG CHEN
90
Hub CDC
DC
RDC
DC
RDC
DC
TRA
RDC
customer
DIS DEL
customer
customer
TRA: transportation
DIS: distribution
DEL: delivery
TRA
TRA
Figure 4. Logistics network for Hub/CDC city
DC
RDC
DC
RDC
DC
RDC
CDC
customer
DIS
DEL
TRA: transportation
DIS: distribution
DEL: delivery
TRA
customer
customer
Figure 5. Logistics network for CDC/RDC city
DC
RDC
DC
RDC
DC
RDC
customer
DIS DEL
TRA: transportation
DIS: distribution
DEL: delivery
TRA
customer
customer
Figure 6. Logistics network for RDC/DC city
5.2.2 Developing Policy for CDC/RDC City
In CDC/RDC city, three tier logistics network is de-
signed, in which goods is transported from the CDC to
the RDC, and then distributed to DC, as illustrated in
Figure 5.
In this system, larger RDCs should be built for out-
board transport, which should be located around the in-
dustry park with convenient transportation network with
the surrounding area, i.e. goods or material can distrib-
uted to the surrounding area directly through this facility.
And lager DC should be built for distributing goods to
retails, stores, even the final consumers.
For transportation network constructing, it is the long
distance transportation system that should be constructed
with urban distribution network improvement as a sup-
plement, i.e. to construct a smooth outboard transport
system.
5.2.3 Developing Policy for RDC/DC City
In RDC/DC city, two tier logistics network is designed,
in which goods is distributed transported from the RDC
to the DC, as illustrated in Figure 6.
In this system, DCs should be built to enforce the
capacity of small-size, multiple-item, and high-frequency
goods order-picking, to improve the capacity of auto-
matic handling, and to quick the response activity. And
RDCs should also be built to accept the goods supplied
by surrounding area and distribute the goods to DCs
timely.
For transportation network constructing, it is the l ur-
ban distribution network that should be constructed by
expansion of traffic road network and traffic capacity
improvement, thus a distribution channel, in which the
routing optimization within urban area is the focus and
long distance transportation is a supplement, should be
setup based on urban GIS and freight characters.
6. Conclusions
The empirical evidence presented in Section 4 gives a
better method of classifying ULC. The following con-
clusions can be drawn:
(1) 30 sample cities’ ULC in China are classified into 3
clusters based on hierarchical cluster analysis
method.
(2) We find that China has two Hub/CDC cities includ-
ing Shanghai and Guangzhou, four CDC/RDC cities
including Beijing, Tianjing, Nanjing and Hangzhou,
and the other 24 RDC/DC cities including Jinan,
Hohhot, Harbin, Urumqi, etc.
(3) The Hub/CDC and CDC/RDC cities are in coastland
area including Bohai Bay region (Beijing), Yangtze
River Delta Region (Shanghai, Nanjing and Hang-
zhou) and Pearl River Delta Region (Guangzhou).
The RDC/DC cities are almost among inland area
(Changsha, Guiyang, Harbin, Urumqi, etc.).
(4) The evaluation framework of ULC is partly tested by
empirical study and needs further and deeper re-
search.
7. Acknowledgement
It is a project supported by philosophy and social science
in Zhejiang Province (07CGLJ018YBX), the results of
Center for Research in Modern Business, Zhejiang
Gongshang University (the important research base for
high school social and science of High Education De-
partment) and the normal project of philosophy and so-
cial science in Hangzhou (D07GL07).
REFERENCES
[1] M. Goh and C. Ling, “Logistics development in China,”
International Journal of Physical Distribution & Logistics
Management, 33(10), pp. 886-917, 2003.
[2] Shanghai Information Center, “Logistics to get boost in
south China city,” People Daily, No. 1, June 5, 2003.
Copyright © 2009 SciRes JSSM
CHANGBING JIANG, DAQIANG CHEN 91
[3] J. Muñuzuri, J. Larrañeta, L. Onieva, and P. Cortés, “So-
lutions applicable by local administrations for urban logis-
tics improvement,” Cities, 22(1), pp. 15-28, 2005.
[4] Mentzer and Konrad, “An efficiency/effectiveness ap-
proach to logistics performance analysis,” Journal of
Business Logistics, 12(1), pp. 33-61, 1991.
[5] B. Jiang and E. Prater, “Distribution and logistics devel-
opment in China: The revolution has begun,” International
Journal of Physical Distribution & Logistics Management,
32(9), pp. 783-798, 2002.
[6] P. Hansen and B. Jaumard, “Cluster analysis and mathe-
matical programming,” Mathematical Programming, 79(1),
191-215, 1997.
[7] R. Xu and D. C. Wunsch II, “Recent advances in cluster
analysis,” International Journal of Intelligent Computing
and Cybernetics, 1(4), 484-508, 2008.
[8] J. R. Harris, “A primer of multivariate statistics (3rd Edi-
tion),” Lawrence Erlbaum Associates Publishers Press, 2001.
[9] B. Manly, “Multivariate statistical methods: A primer,”
Chapman and Hall Press, 1986.
[10] R. Ram, “Composite indices of physical quality of life,
basic needs fulfillment and income: A principal compo-
nent representation,” Journal of Development Economics,
11(1), pp. 227–247, 1982.
[11] C.-S. Lu and C.-C. Yang, “Evaluating key logistics capa-
bilities for international distribution center operators in
Taiwan,” Transportation Journal, 45(4), pp. 9-27, 2006.
[12] X. D. Zhang, “Research on the theory of the location
planning for logistics park,” China Logistics Publishing
House, Beijing (in Chinese), 2004.
[13] S. E. Fawcett, “Developing a logistics capability to im-
prove the performance of international operations,” Jour-
nal of Business Logistics, 18(2), pp. 101-127, 1997.
[14] D. J. Closs, T. J. Goldsby, and S. R. Clinton, “Information
technology influences on world class logistics capability,”
International Journal of Physical Distribution & Logistics
Management, 27(1), pp. 4-17, 1997.
[15] E. A. Morash, C. L. M. Droge, and S. K. Vichery, “Stra-
tegic logistics capabilities for competitive advantage and
firm success,” Journal of Business Logistics, 17(1), pp.
1-22, 1996.
[16] D. F. Lynch, S. B. Keller, and J. Ozment, “The effects of
logistics capabilities and strategy on firm performance,”
Journal of Business Logistics, 21(2), pp. 47-67, 2000.
Copyright © 2009 SciRes JSSM