Modern Economy, 2010, 1, 206-211
doi:10.4236/me.2010.13023 Published Online November 2010 (http://www.SciRP.org/journal/me)
Study the Doub le - Transfer Pa th o f Gu a ngdong Province on
Gravity Model and Cluster Analysis
——Taking Foshan’s Ceramics Industry as an Example
Yun Liang1, Xiaode Zuo2
1Guangdong University of Financ e, Guangzhou, China
2Management School, Jinan University, Guangz hou, China
Received September 12, 2010; revised October 15, 2010; accepted October 18
Abstract
Gravity model and system cluster method are integrated in this paper to generate a regression analysis on the
gravity of many transferees of Foshan’s ceramics industry, which are discussed under the background of the
industry and human resource transfer practice. Also the value and role orientation of the industries, enter-
prises and governments in the transfer process are explored, which provides theoretic guidance to the transfer
practice. The analyzing results show us, the gravity of transferees in Guangdong province is bigger when
compared to other areas.
Keywords: Industry Transfer, Gravity Model, Cluster Analysis, Regression Analysis
A large scale of enterprises has emerged since china’s
reform and opening up, which constitute to a certain in-
dustry structure. Along with the spread of financial crisis,
the mode of export-oriented economy confronts great
difficulties that never seen, putting the industry structure
upgrading and transformation at the edge. The concept of
“Double- transfer” was created in Guangdong Province
of China, which contains industry transfer and labor
transfer, means that the labor-intensive industry will
transfer from Pearl River Delta to east, west and north
Guangdong; while the labor of east, west and north
Guangdong will transfer to local secondary and tertiary
industry, some of qualified ones will transferred to Pearl
River Delta area.
Transfer from advanced economies to developing ones
is the basic path of industry transfer, with the aim of
shortening regional economic gap and promoting harmo-
nious development. Lots of factors can affect industry
transfer, natural resource, labor, capital, technology, in-
frastructure, geographical location, cultural environment,
policy environment, market environment and so on. The
following questions come whether double-transfer will
successfully carry out. What kinds of industries and labor
should be transferred? Where should they be transferred?
This paper applies the gravity model and cluster
analysis to industry transfer, tries to evaluate the influen-
tial factors to them, calculates every city’s gravity
through cluster analysis, whose outputs would be taken
as the inputs to do the electrometrical analysis, through
which all factor’s parameters can be decided, providing
theoretical guidance to the double- transfer practice.
1. Descriptions to the Gravity Model
The gravity model is based on the theory of Universal
Gravitation Principal. It goes that, the gravity between
two objects obeys to the function:
12
2
ij
M
M
ij D
FG (1)
while i
M
,
j
M
means the mass of the two objects, ij
is the distance between them, and is the gravity con-
stant.
D
G
Ever since 1960s, some scholars began to use gravity
model to research bilateral trade volume [1,2]. Reference
[3] used GDPs of two countries to replace the objects’
mass, and the distance between two countries to that be-
tween the objects. The logarithm form was adopted to
make the model linear. He analyzed bilateral trade data
of 15 developed countries and 3 developing countries of
year 1959 and found out that, the bilateral trade flow is
greatly dependent on the economic scales of two coun-
tries and the geographical distance between them. Al-
most at the same time, German economist Poyhonen
(1963) [4] studied bilateral trade with gravity model.

Y. LIANG ET AL.
207
Trading gravity model can be simply expressed as
ij
ij ij
YY
TD







[5]. while is the trade volume be-
ij
T
tween two countries, i and Y
j
Y are GDPs of country
and country , ij is the distance between the two
countries, and
ij D

is a constant. From the model, the
bilateral trade volume is positively related to economic
scales of the two countries and negatively related to the
distance between them.
Gravity model improves itself along with its wide ap-
plication. For example, Linnemann added population as a
new explanatory variable to the model. While in 1970s,
many variables such as virtual variables, non-tariff cov-
erage index, bilateral exchange rate, income per capita,
common language, population intensity and so on were
introduced into the model. Although it is been doubted
and criticized for lacking of theoretical foundation, it has
won big success in practical use. [6-11]. However, grav-
ity model is seldom seen to be used in studying industry
transfer. Reference [12] conducted a research on the
transfer of Electronic and communication equipment
manufacturing industry between 2001 to 2003.
System cluster is the most widely used cluster method.
Its basic concept lies in, take each sample as a cluster,
combine the closest ones according to the sample’s simi-
larity, and calculate the distances between the combined
clusters, then combine them as above. Continue the
process until all the samples become one cluster. Among
them, the square of Euclidean distance is used to evalu-
ate the similarity among clusters, and averaging method
is used as cluster method.
2. Model Descriptions
2.1. Set up the Model
Every industry has its own characteristics and different
influential factors. Here ceramic industry of Foshan city
is chosen to illustrate the application of gravity model.
First the model brought by Tinbergen,
3 12 4
0iji jij ij
X
YY DP



 

, is written into the form of
natural logarithm:
01234 i
lnln +ln+ ln+ln+ lnP
iji jij
XYYD
j

 

(2)
Some of the parameters are re-defined according to
industry transfer practice.
1) i = 1, 2……n, indicates the cities that an industry
transfer from, j = 1,2……m，indicates the cities that an
industry transfer to. ij
X
means the gravity between two
cities.
2) i, j demonstrate industry outputs of city i and
city j. The bigger the output is, the stronger the gravity to
an industry.
Y Y
3) ij is the economic distance between the two cit-
ies, whose value is obtained from two time’s weighting
to their physical distance. The function of economic dis-
tance can be expressed as:
D
ij
DD


 (3)
D is the geographical distance,

is the geographic
weight, Which decided by the transportation situations
between two cities,

is the economy weight decided
by the ratio of GDP per capita of the transferee and
transferor cities. Table 1 shows how the economic dis-
tance is calculated.
4) ij is the transferring parameter between two cities,
measured by two governments’ beneficial policies to the
certain industry. The gravity will be bigger if the gov-
ernments offer some convenience. Here in the double-
transfer practice, ij
P = in Guangdong province, and 0
in other regions.
P
1
5) the labor volume is inneglectable, rich human re-
source is an important attractions to industry transfer, the
permanent residence is used in this function.
6) special resource gift of transferee areas. The devel-
opment of a certain industry relies on certain resources,
such as land, energy and so on. Therefore, if the region
possesses resource, it has a value of 1, 0 if it doesn’t.
Table 1. Parameters of economic distance.
Geographical distance weight

Transportation tool train van ship Train and vanTrain and shipVan and ship Train, van and ship

1 1.2 1.5 0.7 0.8 1.1 0.5
Economic distance weight

GDP per capita that industry transfers from/GDP per
capita that industry transfers to > 70% 70% ≥ ratio ≥ 45% < 45%

0.8 1.0 1.2

Y. LIANG ET AL.
208
Based on the above description, the new model can be
written as
01 23
45
ln+ln +lnln
ijj ijj
jij
F
YD la
Presj
 
 

 
b
(4)
We can conclude 5 factors that affect a city’s gravity
to an industry, they are industry output, economic dis-
tance, labor volume, beneficial policies and natural re-
source gift. The data of these 5 factors are used to cluster
analysis to measure each city’s gravity.
3. Double Transfer of Foshan’s Ceramics
Industry
14 cities are selected as potential areas that the industry
will transfer to: Qingyuan, Zaoqing, Heyuan, Jiajiang,
Jingdezhen, Gaoan, Fengcheng, Liling, Huanggang, Zibo,
linyi, Faku, Jinjiang, and Yining.
First we calculate the economic distance between the
above cities and Foshan, the results are shown in Table
2.
Other parameter’s values are shown in Table 31:
SPSS16.0 is adopted to do the cluster analysis. Stan-
dardize all variables and make them easy to be com-
pared.
The classfing results are shown as in Figure 1 and
Figure 2:
Dendrogram using Average Linkage (Between Groups)
Figure 2 shows the situations when the 14 cities been
classified to 2, 3, 4, 5 and 6 clusters. It is quite obvious
that if been classified by gravity, from bottom to up, to 2
clusters, Faku and Yining are in one cluster and the other
12 cities in the other group; if been classified into 3 clus-
ters, Yining in one cluster, Faku another, and the remain-
ing cities the third. If 6, Yining the first cluster, then Faku,
Linyi the third, Jingdezhen the forth, Huanggang, Jiajiang,
Liling, Jinjiang, Gaoan, Fengcheng and Zibo the fifth, and
Qingyuang, Zaoqing Heyuang the sixth cluster.
Only four factors’ influences have been considered in
the model. After cluster analysis, the industry transferring
direction is quite clear and practical: the potential areas
with biggest gravity are in group one, Qingyuang, Zaoqing
and Heyuang. That is because of short economic distance
and governments’ great push on double-transfer policy.
Jiajiang and Liling have sound transportation conditions
and are rich in resources, some places such as Gaoan and
Fengcheng have already formed the ceramics industry
chain. While with a long distance from Foshan, weak
transportation situation and low GDP level, Yining has
little gravity to the ceramics industry even it possess many
natural resources and has a great market coverage. It is
noted that, although Jingdezhen is a famous city with de
Table 2. Economic distance between Foshan and potential
cities.
No. Potential
cities D (km)


Dij
1 Qingyuan74.06 Train, van 0.7 1.20 62.2104
2 Zaoqing 70.35 train, van 0.7 1.20 59.094
3 Heyuan 185.73train, van 0.7 1.20 156.0132
4 Jiajiang 1210.15train, van 0.7 1.20 1016.526
5 Jingdezhen804.7 train, van 0.7 1.20 675.948
6 Gaoan 641.97van 1.2 1.20 924.4368
7 Fengcheng629.71train, van 0.7 1.20 528.9564
8 Liling 516.39train, van 0.7 1.20 433.7676
9 Huanggang845.96van 1.2 1.20 1218.182
10Zibo 1611.15train, van 0.7 0.80 902.244
11Linyi 1431.44train, van 0.7 1.20 1202.41
12Faku 2367.24van 1.2 1.20 3408.826
13Jinjiang 590.52Van, ship 1.1 1.00 649.572
14Yining 3714.63Van 1.2 1.20 5349.067
Table 3. Parameter’s value of potential areas.
No. Potential
cities
Y
(100 million
Yuan)
Dij (km) Lab
(1000 per)Pij Res
1 Qingyuan21.86 62.2104 365.87 1.50
2 Zaoqing 9.01 59.094 375.20 1.50
3 Heyuan 6.14 156.0132 281.82 1.50
4 Jiajiang 50.00 1016.526 35.00 1 0
5 Jingdezhen42.00 675.948 155.44 1 0
6 Gaoan 12.46 924.4368 80.80 1 1
7 Fengcheng10.00 528.9564 133.30 1 0
8 Liling 90.00 433.7676 98.39 1 0
9 Huanggang8.00 1218.182 666.70 1 0
10Zibo 18.00 902.244 419.59 1 0
11Linyi 17.00 1202.41 1027.50 1 0
12Faku 113.40 3408.826 45.00 1 1
13Jinjiang 70.00 649.572 158.70 1 0
14Yining 3.00 5349.067 44.22 1 1
1Data of GDP per capita and labor volume are from statistical yearboo
k
of 2008.

Y. LIANG ET AL.
209
Rescaled Distance Cluster Combine
6 ─┬─────┐
7 ─┘ ├───┐
9 ───┬───┘ │
10 ───┘ ├───────┐
8 ─┬─┐ │ │
13 ─┘ ├───────┘ ├─┐
4 ───┘ │ │
5 ───────────────────┘ ├───────┐
2 ─┐ │ │
3 ─┼───────────────────┘ ├───────────────────┐
1 ─┘ │ │
11 ─────────────────────────────┘ │
12 ───────────────────────────────┬─────────────────┘
14 ───────────────────────────────┘
Figure 1. Cluster tree chart.
Cluster Membership
Case 6 Clusters 5 Clusters 4 Clusters 3 Clusters 2 Clusters
1 1 1 1 1 1
2 1 1 1 1 1
3 1 1 1 1 1
4 2 2 1 1 1
5 3 2 1 1 1
6 2 2 1 1 1
7 2 2 1 1 1
8 2 2 1 1 1
9 2 2 1 1 1
10 2 2 1 1 1
11 4 3 2 1 1
12 5 4 3 2 2
13 2 2 1 1 1
14 6 5 4 3 2
Figure 2. Cluster membership.
-veloped ceramics industry, it is catalyzed into the third
group for resource, policies reasons.
Based on the result of cluster analysis, evaluate the
gravity of different group in an up-down order as 6, 5, 4, 3,
2, 1. Then do an electrometrical analysis to the influential
factors in Function 3. Regression based on Least Squares
method is adopted and Eviews 5.0 is used. The regression
results are shown in Table 4.
From the table, R2 = 0.914419, the adjusted R2 =
0.860930, which tells that the variables we chose can
explain more than 86% of the gravity. F = 17.09564,
which is bigger than critical value 2.73. to every single
variable, it is found that t-Statistic is not significant, such
as Res. That probably because ceramics industry doesn’t
rely on resources very much. So the variable Res is
kicked out, and regression is done again as shown in Ta-
ble 5:
R2 of this time is 0.914054, and the adjusted one is
0.875856, the value of F is 23.91928, bigger than F’s
value last time, which means the explaintory variables
has a better explantation to the explained variable. a new
function can be obtained based on function 3:
ln6.0267520.007076 ln0.00917 ln
0.001140 ln0.534150
j
ij
j
F
jY
lab Pj
 

D
(5)
Here are the explanations: 1) the ceramics industry
output is negatively related to the gravity, which is quite
reasonable since a big output means a full covered mar-
ket, providing great barrier to new entrants; 2) economic
distance’s effect is negative, but the parameter’s value is
small. It’s easy to understand since with the improve-
ments of transportation and economy, economic distance
has little effect on ceramics industry, especially this in-
dustry doesn’t need much physical transfer, most of the
cases are building factories and purchasing equipments
in the places that will transfer to. What transfer in the
process are just technologies, capital and expertise. Also,
economic gap is just a compe concept which can arativ

210 Y. LIANG ET AL.
Table 4. Regression analysis of ceramics industry.
Dependent Variable: F
Method: Least Squares
Date: 07/04/09 Time: 01:34
Sample: 1 14
Included observations: 14
Variable Coefficient Std. Error t-Statistic Prob.
C 6.036283 0.423007 14.26993 0.0000
Y –0.007120 0.005169 –1.377319 0.2057
DIJ –0.000893 0.000176 –5.080806 0.0010
LAB –0.001187 0.000668 –1.777148 0.1134
P 0.538737 0.424937 1.267804 0.2405
RES –0.113089 0.612503 –0.184635 0.8581
R-squared 0.914419 Mean dependent var 4.500000
Adjusted R-squared 0.860930 S.D. dependent var 1.506397
S.E. of regression 0.561767 Akaike info criterion 1.982066
Sum squared resid 2.524653 Schwarz criterion 2.255948
Log likelihood –7.874465 F-statistic 17.09564
Durbin-Watson stat 2.171311 Prob(F-statistic) 0.000435
Table 5. Regression results without Res.
Dependent Variable: F
Method: Least Squares
Date: 07/04/09 Time: 14:39
Sample: 1 14
Included observations: 14
Variable Coefficient Std. Error t-Statistic Prob.
C 6.026752 0.396677 15.19312 0.0000
Y –0.007076 0.004879 –1.450367 0.1809
DIJ –0.000917 0.000114 –8.057946 0.0000
LAB -0.001140 0.000585 –1.950037 0.0830
P 0.534150 0.400801 1.332708 0.2154
R-squared 0.914054 Mean dependent var 4.500000
Adjusted R-squared 0.875856 S.D. dependent var 1.506397
S.E. of regression 0.530766 Akaike info criterion 1.843461
Sum squared resid 2.535412 Schwarz criterion 2.071696
Log likelihood –7.904230 F-statistic 23.92918
Durbin-Watson stat 2.082530 Prob(F-statistic) 0.000082

Y. LIANG ET AL.
211
make up by beneficial policies and low costs; 3) Human
resource, replaced by permanent residence, also has a
negative but small effect on the gravity, which is quite
understandable since ceramics industry doesn’t need
large number of qualified labor, making every city’s hu-
man resources are sufficient for this industry; 4) gov-
ernment policy plays an important role here with a coef-
ficient of 0.538737. industry transfer has great influence
on both regions that transfers from and transfers to,
which deeply relied on goverments’ policies. From this,
the importance of Guangdong’s double-transfer to pro-
mote industry transfer is quite obvious.
4. Conclusions
The influential factors are decided through gravity model,
which are taken as the input to cluster analysis, whose
results are given different group values as the value of
different gravity levels. Then Eviews 5.0 is used to do
the regression analysis. The following conclusions are
drawn: 1) governments’ beneficial policies play impor-
tant roles in industry transfer process, that’s why recipi-
ent governments enacted all kinds of polices to attract
industries transferring in; 2) as to economic distance, its
role in industry transfer is reduced since great improve-
ment of transportation and economy conditions national
wide; 3) Only five variables are considered in the model
and cluster analysis, which maybe cannot explain the real
situation sufficiently. Some variables are just hard to be
quantified but still have huge influences on industry
transfer; 4) also just the ceramics industry of Foshan has
been studied in this paper, which is a small fraction of
Guangdong’s industry transfer business, more industries’
practice can be research further. Moreover, some of data
used in this paper are not so precise because of hard ac-
cess.
5. References
[1] Anderson and E. A. James, “Theoretical Foundation for
the Gravity Equation,” The American Economic Review,
Vol. 69, No. 1, 1979, pp. 106-116.
[2] D. J. Peng. “Ethnic Chinese Business Network and the
Asia Pacific Economic Integration,” Journal of Asian&
African Studies, Vol. 35, No. 2, 2000, pp. 229-250.
[3] J. Tinbergen, “Shaping the World Economy: An Analysis
of World Trade Flows,” New York Twentieth Century
Fund, Vol. 5, No. 1, 1962, pp. 27-30.
[4] P. Poyhonen, “A Tentative Model of the Volume of
Trade between Countries,” The American Economic Re-
view, Vol. 2, 1963, pp. 93-99.
[5] S. Bin and M. Z. Liao, “Trade Volume and Export Poten-
tial of China,” World Economy, Vol. 2, 2004, pp. 3-12.
[6] J. H. Zhang, “Study on Investing Country Characteristics
and Investment Intensity to China,” World Economy, Vol.
1, 2004, pp. 16-22.
[7] X. Ruan and J. H. Zhao, “FDI Location Selection from
the View of Gravity Model,” world Economy, Vol. 2,
2004, pp. 67-71.
[8] C. C. Zhang, “Analysis on Investment Environment Fac-
tors That Affect FDI,” Management World, Vol. 2, 2002,
pp. 32-41.
[9] K. J. Gu, “International Economics’ Development and
Application on the Gravity Model,” World Economy, Vol.
2, 2001, pp. 14-25.
[10] H. X. Shi and S. F. Liu, “Study on the transfer of elec-
tronic and communication equipment industry based on
gravity model,” Industry Technology Economics, Vol. 8,
2008.
[11] W. Sun, “Cluster Analysis on Equipment Manufacturing
Industry in 29 Provinces of China,” China Technology
Forum, Vol. 6, 2005.
[12] D. R. Cheng and J. Li, “Study on the Interactivity of City
Economics under Mulnucipal Economic Cycles Con-
struction,” Technology and Economics, Vol. 2, 2007, pp.
42-44.