Transcript A Spatial Econometric Analysis of the Impact of Labor Migration on

2006 International Symposium
on Contemporary Labor
Economics (LABOR2006)
December 16 - 18, 2006
1
A Spatial Econometric Analysis of
the Impact of Labor Migration on
the Disparity of Regional
Economic Development in
Mainland China
Zhengming Qian, Xiuhua Zhang,
Yangping Yu and Pengfei Guo
Xiamen University (361005)
Xiamen, Fujian
PRC
2
1. Introduction
2. Analysis of Convergence of Regional
Economic Growth of China
3. The Spatial Econometric Model
4. Data Set & Empirical Results
5. Conclusions & Suggestions
3
1. Introduction
Labor migration has significant influences on regional
economic development. At present, the combination of
economic growth and labor migration to study the impact of
labor migration on regional economic growth becomes a hot
spot.
There exist two opposing viewpoints about the impact of labor
migration on regional economic development in domestic
academy.
One view holds that labor migration presents positive impact
on regional economic convergence, and it will narrows regional
economic disparity, where labor migration could markedly
reduce the disparity of resource among regions and effectively
reduces
regional
economic
disparity
and
realizes
the‘conditional convergence’. The related studies include: 4
Duan Pingzhoug & Liu Chuangzhong (2005) point out that,
floating population does obviously contribution to regional
economic growth, and such contribution decreases gradually.
Wang XiaoLu & Fan Guang (2004) think that labor migration
among regions could narrow the disparity of labor’s return
and per-GDP among regions;
Yao Zhizhong & Zhou Shufang(2003) present that labor
migration do contribute something to narrow regional
economic disparity, but there exists great restriction on labor
migration;
Liu Qiang(2001) shows that great labor migration among
regions is a important leading factor for the economic
convergence of China’s regions;
5
The opposite viewpoint thinks that labor migration
among provinces broaden regional economic disparity.
The related studies include:
Sun ZhiFeng(2006) points out that labor migration
among provinces broadens regional economic disparity.
He Qiang(2006) presents that rural labor migration
boosts civilization, as well as broadens incoming
disparity between city and country for the reason of
human capital barrier.
6
We suggest that most of the above studies support
the effect of labor migration through the index of
net labor migration, but they haven’t separated the
labor migration into labor inflow and labor
outflow, and distinguished the East, the West and
the Middle. And what’s more, most empirical
studies are based on cross-sectional data and
neglect the influence of time effect.
So, this research constructs spatial econometric
model based on the panel data of China’s regions
to analyze how the direction and the power of
inflow and outflow of labor influence the three
region’s economic development.
7
2．Analysis of Convergence of Regional
Economic Growth of China
In statistical analysis of economic growth
convergence, researchers always like to use some
coefficient or index such as Theil coefficient. Theil
coefficient is a analytical method to measure
regional spatial diversity, and it can be used to
analyze the total change, inner-change and interchange of regional disparity and the impact of the
inner-change and inter-change of regional disparity
on the total change of regional disparity.
Therefore, we use the Theil coefficient to measure
8
the economic disparity among the East, the West
The Theil coefficient can be defined as:
N
T=  y i log( y i /p i )
（1）
i 1
Where T denotes the Theil coefficient; yi denotes the
share of the GDP of the ith region on total GDP; pi
denotes the share of the population of the ith region on
total population.
The greater T value shows the greater disparity among
regional economic development, while the smaller T
value shows a smaller disparity among regional
economic development.
9
Let province as a regional unit, and the total disparity of Theil
coefficient can be defined as:
T  
i
j
Yij
Y
log(
Yij / Y
Pij / P
)
（2）
Where
Yij
denotes the population of the jth province in the ith region,
P denotes the total population.
Pij
denotes the GDP of the jth province in the ith region,
Y denotes the total GDP,
10
Based on the equation(2), define the disparity inner the ith region as:
Ti   (
j
Where
Yij
Yi
) log(
Yij / Yi
Pij / Pi
)
(3)
Yi
denotes the GDP of the ith region,
Pi
denotes the population of the ith region
Based on the equation (3), the Theil coefficient of equation (2)
can be separated into:
T=  (
i
Yi
Y
Y /Y
)Ti   ( i ) log( i
)
Y
Y
Pi / P
i
（4）
  (Yi /Y)Ti  TBR
i
=TWR  TBR
Where TWR denotes the inner-region disparity, TBR denotes
the inter-region disparity.
11
This research uses the GDP of total 30 provinces of China from
1996 to 2004 to estimate the Theil coefficient (see in Table 1).
Table 1：Theil Coefficient（1996—2004）
YEAR
TOTA
L
EAST
MIDD
WEST
LE
TWR
TBR
TWR
Share
TBR
Share
1996
0.1024 0.0461 0.0381 0.0213 0.0394 0.0629 0.3850 0.6150
1997
0.1042 0.0470 0.0387 0.0223 0.0403 0.0639 0.3865 0.6135
1998
0.1092 0.0485 0.0367 0.0219 0.0406 0.0686 0.3721 0.6279
1999
0.1154 0.0501 0.0359 0.0206 0.0412 0.0743 0.3566 0.6434
2000
0.1054 0.0438 0.0379 0.0192 0.0381 0.0673
2001
0.1191 0.0466 0.0417 0.0235 0.0414 0.0777 0.3475 0.6525
2002
0.1213 0.0463 0.0405 0.0240 0.0410 0.0803 0.3384 0.6616
2003
0.1055 0.0394 0.0343 0.0226 0.0353 0.0702 0.3343 0.6657
2004
0.1211 0.0417 0.0396 0.0298 0.0392 0.0820 0.3235 0.6765
0.3611
0.6389
12
（Data Resource：《China Statistical Yea Book》(1996—2005), calculated）
As showed in Table 1,
Firstly, in total, the Theil coefficients increased from the 0.1024 of
1996 to 0.1211 of 2004, which shows the process of disparity which
widened among regions.
Secondly, among the three regions, the Theil coefficients change
just slightly: the East from 0.0461 to 0.0417, the West from 0.0213
to 0.0298, and the Middle from 0.0381 to 0.0396. All show a weak
disparity within the three regions.
Thirdly, the inner-region disparity TWR shows a weak decrease
during this period (from 0.0394 to 0.039), while the inter-region
disparity TBR shows a obvious increase (from 0.0629 to 0.0820).
From Table 1 we can conclude that, in mainland China, the innerregion economic development presents gradual convergence, while
the inter-region economic development presents gradual divergence.
13
Whether such a labor migration
broadens or narrows regional economic
disparity becomes an interesting issue.
This research uses spatial econometric
panel data model to analyze the impact
of inflow and outflow of labor on regional
economic disparity, and empirically
discusses the quantitative characteristics
of such impact.
14
3. The spatial econometric panel data model
3.1 Spatial Correlation Index
There are two statistical indices for testing the spatial
correlation of regional economic growth in common use, one
is Moran I which was first presented by Moran in 1950,
the other is Greary c which was first presented by Geary in
1954. And the most
used one is the Moran I, which is defined
，
as:
n
n
Wij (Yi  Y )(Y j  Y )

Moran I= i 1 j 1
n
n
（5）
2
S  Wij
i 1 i 1
Where
1 n
S   (Yi  Y ) 2
n i 1
2
1 n
Y   Yi ，Yi denotes the observed value of the ith region,
n i 1
n denotes the number of regions;
15
Wij
is the binary system neighbor spatial weight matrix, which is defined as:
1　when t he i t h r egi on nei ghbor s t he j t h r egi on；
Wij＝
0　when t he i t h r egi on unnei ghbor s t he j t h r egi on。　Based on the distribution of spatial data, the expectation and
variance of Moran I can be calculated:
We can use the following equation to test the spatial auto-correlation of
the n regions:
MoranI  E ( I )
Z (d ) 
VAR( I )
（8）
16
3.2 Spatial Econometric Model
The spatial econometric model includes spatial lag
model and spatial error model.
The spatial lag model includes explanatory variable and
spatial lag term
The spatial error model is the special case of regressive
model which includes a auto-correlated error term. In its
covariance matrix, the non-diagonal factors denote the
structure of spatial correlation, which can be defined in
different ways. The error variance-covariance matrix,
which is a parameter vector.
17
3.3 Panel Data Model
Panel data is a data structure which mixes the two
dimensions of time and space. The panel data model can be
defined as:
The most important step in panel data analyses is model
selection. The estimation of the panel data equation depends
on the assumption that is made of intercept and slope of the
model. We first investigated whether the parameters of the
dependent variable Yit were constant for all of the crosssectional units and periods.
The panel data model involves the following three situations,
depending on both the intercept and the slope.
18
In Situation 1, both the intercept and the slope are the same
(i = j, i = j); therefore, it is a pooled regression model.
In Situation 2, the slope is the same but the intercept is
different (i = j, i = j), and thus it is a variable intercept
model.
In Situation 3, both the intercept and the slope are different
(i  j, i  j), and thus it a variable coefficient model.
Based on the characteristics of sample data, the variable
intercept model and the variable coefficient model can be
separated into fixed-effect model and random-effect
model. The former model is suit for the case inferred from
sample, while the latter model suits for the case inferred
from the total.
The analysis results indicate that the panel data fit nicely into
Situation 2.
19
Panel data model can effectively solve the problems caused
by small sample, and we could not use the traditional OLS
method to estimate the panel data model because the panel
data model doesn’t satisfy the classical assumption.
So, in order to guarantee the validity of model estimation, this
study uses GLS to estimate the model.
20
4．The spatial econometric analysis of the impact of labor
migration on regional economic development of China
4.1 Data Set and Empirical Analysis Method
We study the data of 30 provinces in mainland China from
2000 to 2004 (here we merge Chongqing into Sichun and
include 30 provinces), then,
We group these 30 provinces into three regions (the East, the
Middle and the West ). The differences we intend to study are
reflected mainly in the intercept, as previously indicated, and
we adopt the model with fixed effects.
21
4.2 Analysis of Spatial Auto-Correlation of China’s Economic Growth
Firstly we test the dependence among provinces’ economic development by
Moran I and Z value (See Table 2).
Table 2：Moran I and Z value for the Per-GDP of Regional Economic
Growth in China
Moran I
E(I)
VAR(I)
Z-Value
P-Value
2000
0.2220
-0.0345
0.0230
1.6921
0.0455
2001
0.2247
-0.0345
0.0230
1.7097
0.0436
2002
0.2341
-0.0345
0.0230
1.7715
0.0384
2003
0.3202
-0.0345
0.0230
2.3396
0.0096
2004
0.2377
-0.0345
0.0230
1.7952
0.0359
In Table 2, we can see that all Z value over the critical value (1.64) at 5% level,
which means that there exists obvious positive spatial correlation among provinces’
economic growth of China, so they are spatially dependent. Therefore, it is
necessary to integrate the spatial correlation factor into the model.
22
4.3 Model Setting, Estimation and Analysis
First, we found that there existed obvious spatial
dependence among China’s economy from 4.2. We adopt
the spatial lag model as well as the spatial lag factor to
construct the panel data model. To separate the respective
impact of labor inflow and labor outflow on regional economic
development, we set the spatial panel data model as follows:
LnGDP=   Ln(MIGIN)   Ln(MIGOUT）  WLn(GDP)
（16）
Where the per-GDP is the dependent variable, the
independent variables includes labor inflow (noted as MIGIN)
which measures the population inflow from the other provinces,
labor outflow (noted as MIGOUT) which measures the
population outflow from the province to other provinces, and
the spatial lag factor (noted as WLn (GDP)) which is the
product of regional spatial weight matrix by regional per-GDP.
23
We use the GLS to estimate the spatial panel data model and the result
showed in Table 3.
Table 3：Estimating Result of the Spatial Panel Data
MIGIN
EAST
MIDD
LE
WES
T
MIGOUT WLn(GDP)
coefficient
0.1143
-0.1016
0.3945
Std.E
0.0486
0.0212
0.0120
t
2.3509
-4.7990
32.8111
p
0.0236
0.0000
0.0000
coefficient
-0.1047
0.7065
0.1605
Std.E
0.0492
0.1009
0.0560
t
-2.1285
7.0033
2.8640
p
0.0419
0.0000
0.0077
coefficient
0.0860
0.3820
0.1820
Std.E
0.0422
0.0474
0.0122
t
2.0376
8.0629
14.8903
p
0.0481
0.0000
0.0000
R2
Adj R2
P(F)
DW
0.9999
0.9999
0.0000
1.4983
0.9999
0.9999
0.0000
1.1174
0.9999
0.9999
0.0000
1.7707
24
From Table 3, we can see that the R values, F value and t value all
pass the tests. The DW value shows unobvious serial correlation,
which means a good fit of the model. Thus we can conclude that the
model with fixed effects performs well. The labor migration
significantly influences regional economic development. And the t
value of spatial lag factor passes tests at 1% level, which confirms the
spatial dependence among regional economic development.
Then we analyze the impacts of labor inflow and labor outflow on
regional economic development for the East, the West and the Middle.
First - the impact on the East. The coefficient of labor inflow is
positive, which means the labor inflow boosts the economic
development of the East. The 1% increase of labor inflow can
generate a 0.1143% growth of GDP for the East. Whereas, the
coefficient of labor outflow is negative, which means the labor
outflow counteracts the economic development of the East.
The 1% increase of labor outflow would make a 0.1016%
25
decrease of GDP for the East.
Second - the impact on the Middle. The coefficient of labor
inflow is negative, which means the labor inflow counteracts
the economic development of the Middle. The 1% increase of
labor inflow can generate a 0.1047% decrease of GDP for the
Middle. Whereas, the coefficient of labor outflow is positive,
which means the labor outflow boosts the economic
development of the Middle. The 1% increase of labor outflow
would make a 0.7065% growth of GDP for the Middle.
Third - the impact on the West. The coefficient of labor inflow
is positive, which means the labor inflow boosts the economic
development of the West. The 1% increase of labor inflow can
generate a 0.0867% growth of GDP for the West. The
coefficient of labor outflow is also positive and is much greater,
which means the labor outflow can greatly boosts the
economic development of the Middle. The 1% increase of
labor outflow would make a 0.3820% growth of GDP for the
26
West.
4.4 Factor Analysis
We can see from the findings of the empirical study that
both labor inflow and outflow have different effects on
different regions. Even within the same region, such effects
are also different for labor inflow and outflow.
We find that the two main reasons which cause such
differences: (1) the issue of human capital associated with
labor flows; and (2) the distribution of flow associated with
labor migration.
So, Next step, we analyze the reason for such different
impact of labor migration on regional economic
development. There are two different factors: (1)Human
Capital Barrier to the Labor Migration . (2) Unbalanced
Distribution of Labor Migration (please see my paper). 27
5. Conclusions and Suggestion
Since the opening and reform policy implemented, China’s
regional economic development has presented an obvious
‘club convergence’, and the disparity among the three
regions becomes wider. The labor migration did impact the
regional economic disparity. In particular, the inflow and
outflow of labor have different impacts on regional economic
growth and their combined effect results in the disparity of
the regional economic development.
On the whole, the labor inflow leads to a difference of labor
in both quantity and quality, which in turn broadens the
disparity of regional economic development. The labor
outflow transfers the income from the East to the West and
the Middle, which in turn narrows the disparity of regional
economic development.
28
Because of the opposite impact of labor inflow and outflow
on regional economic development, the negative impact on
regional economic convergence of unreasonable labor
inflow finally counteracts the positive impact on regional
economic convergence of labor outflow. Thus the total
impact of labor migration on regional economic convergence
is negative, which means the labor migration totally widens
the disparity of regional economic development.
In order to ensure the healthy impact of labor migration on
regional economic development, the government needs to
balance the direction, quantity and quality of labor migration.
29
The government should implement some feasible policies to
direct labor migration.
Firstly, It has to put more attention to the ‘Big’ West
Development, encourage and allure much more high
qualified labor to the West, which can produce the great
effect of human capital.
Secondly, it has to put the same attention to the Middle by
addressing the low quality and quantity of labor migration. It
needs to encourage and allure more high qualified labor to
the Middle, and to reverse the impact of the direction of labor
inflow on regional economy.
30
Thank you for
participation
31