Transportation (2011) 38:737–752

DOI 10.1007/s11116-011-9349-6

Transport infrastructure and regional economic growth:

evidence from China

Junjie Hong • Zhaofang Chu • Qiang Wang

Published online: 2 June 2011

Ó Springer Science+Business Media, LLC. 2011

Abstract This article develops a multi-dimensions measurement of transport infra-

structure and examines the linkage between transport infrastructure and regional economic
growth. A panel data model is estimated using data from a sample of 31 Chinese provinces
from 1998 to 2007. The results provide strong evidence that transport infrastructure plays
an important role in economic growth. Both land transport and water transport infra-
structure have strong and significant impacts, while the contribution of airway transport
infrastructure is weak. Furthermore, land transport infrastructure contributes more to
economic growth in locations with poor land transport infrastructure, while the investment
in water transport infrastructure contribute positively to economic growth only after the
investment scale exceeds a threshold level. These results are robust to a variety of alter-
native methods, the exclusion of possible outliers, and consideration of endogeneity.
A retrospective analysis shows that uneven distribution of transport infrastructure is an
important reason behind economic disparities across Chinese regions.

Keywords Transport infrastructure  Economic growth  Regional disparity  China

Introduction

Transport infrastructure leads to economic growth through several ways. First, investing in
infrastructure itself increases demand for goods and services. Second, transport infra-
structure improvement reduces travel time, and passenger and freight transporters gain
directly from time and cost saving (Gunasekera et al. 2008). The time savings induced by
improvement of transport infrastructure can yield economic consequences through making
producers access to distant markets and draw inputs from a larger area, and hence stimulate
local production. In addition, highway infrastructure investment can generate benefits by
lowering firms’ inventories (Shirley and Winston 2004). Third, better infrastructure attracts

J. Hong (&)  Z. Chu  Q. Wang

School of International Trade and Economics, University of International Business and Economics,
Beijing 100029, China
e-mail: [email protected]

123
738 Transportation (2011) 38:737–752

foreign direct investment (Hong 2007), which is widely regarded as an important engine of
economic growth in China. Lastly, lower transport and trade cost can accelerate industrial
agglomeration (Baldwin and Forslid 2000; Krugman 1991), and the concentration of
economic activities increases labor productivity (Ciccone and Hall 1996). Banister and
Berechman (2000) depict a general framework that describes the relationship between the
transportation system and economic growth. According to their framework, improved
transport accessibility reduces travel time and cost, increases traffic volume, and leads to
spatial redistribution of economic activities. It further leads to pecuniary externalities, as
well as allocative externalities in environment, transport network economies, labor market
and firm agglomeration, and hence cause economic growth.
In theory, the perception of the role of infrastructure as an engine of economic growth
has changed over time. Solow’s (1956) neoclassical growth model used an aggregate
production function approach and assumed exogenous technical changes, while the work
of Romer (1986) began a set of theoretical research on the endogeneity of economic
growth. During the past two decades, numerous empirical studies have been done to
examine the impact of infrastructure on economic development. Following the pioneer
study by Aschauer (1989), some economists have estimated neo-classical production
function, in which infrastructure investment is treated as an input to production (e.g.,
Berechman et al. 2006; Evans and Karras 1993; Gillen 1996; Lall 2007; Otto and Voss
1996; Seitz 1994). In recent years, simultaneous approach has been used in the literature
(e.g., Roller and Waverman 2001; Yamaguchi 2007). Most previous studies conclude that
public investment in transportation infrastructure has a strong and positive effect on output.
Some studies attempt to investigate the impact of transport infrastructure on economic
growth in China. Mody and Wang (1997) examined the determinants of coastal China’s
growth using panel data on 23 industrial sectors between 1985 and 1989, and found that
roadway transport infrastructure was an engine of economic growth. Fleisher and Chen
(1997) did not find any significant impact of transportation infrastructure on total factor
productivity and economic growth during the period of 1978–1993. Demurger (2001)
investigated the relationship between public infrastructure and economic growth in China
using panel data from a sample of 24 Chinese provinces from 1985 to 1998, and found that
infrastructure endowment did contribute to economic development.
While a number of studies have been done in the literature, several questions remain
unsolved. First, previous studies normally focus on a certain type of transport infrastruc-
ture, such as roadway (e.g., Gunasekera et al. 2008), airports (e.g., Yamaguchi 2007), and
hence cannot compare the impacts of various types of infrastructure. Second, most of
previous studies use either transportation investment (e.g., Berechman et al. 2006), or
roadway length (e.g., Demurger 2001) to measure infrastructure endowment, and hence
cannot capture the effect of infrastructure quality. Third, previous studies have ignored that
the impacts of infrastructure on economic growth may differ across regions. In a large
country like China, substantial regional differences exist. Transport infrastructure may
have different impacts on economic growth across regions. Finally, most of existing studies
in the literature assumes a linear relationship between infrastructure and economic
development. However, transportation gains (e.g., time saved) may be capitalized through
agglomeration effects, which implies a non-linear relationship between transportation and
growth.1
This research attempts to fill the above gaps in the literature. Based on a factor analysis,
we develop a comprehensive index to measure both quantitative and qualitative

1
Thanks for an anonymous referee to raise this point.

123
Transportation (2011) 38:737–752 739

characteristics of each type of transport infrastructure. This allows us to capture the impact
of both quantity and quality of transport infrastructure, and compare the contribution of
different types of infrastructure to economic growth. The square of the transport variables
are included to study the nature of returns to scale to transport investment. We also make a
growth decomposition analysis, to identify whether the impacts of various modes of
transport infrastructure on economic development differ across regions.
The rest of the article is organized as follows. Section 2 gives an overview of infra-
structure development and the Chinese economy. Section 3 describes research method-
ology and data source. Then, the results and discussions are reported. The last section
provides concluding remarks.

Infrastructure development and regional economy in China

China’s economic growth has drawn much attention due to its booming economy with an
annual GDP growth rate of about 10% in the past decade. However, the divergence of
regional economic development in China appears to be very immense, with eastern China
developed and western China lagged. Inequality in China has increased steadily and
inexorably since the opening up. In 1983, China’s overall Gini coefficient, measured on
income, was 0.28, but increased to 0.447 in 2001 (Naughton 2007). The regional inequality
has severely restricted the long-term sustainable development of the Chinese economy.
Balanced growth is one of the major challenges faced by the government. To achieve this,
China’s central government puts forward ‘‘Go West and Central China Strategy’’, taking
some preferential policies towards lagged areas in order to reduce inter-provincial
inequalities. Among these, increasing public investment in transport infrastructure in
central and western China is a key policy. As suggested by Cieslik and Kaniewsk (2004),
transportation improvement can reduce transaction costs on trade in goods. Considering
China’s huge size, the availability of an appropriate infrastructure facilitates trade and
knowledge-spillovers across regions, and hence is helpful for economic growth in lagged
areas.
At the early stage of the economic reform, infrastructure development in many regions
has lagged behind economic growth due to low government spending on infrastructure,
decreased investment incentives of state-owned enterprises, and diminished capability of
local government in mobilizing rural resources (Lin 2001). Since early 1990s, investment
in infrastructure has been regarded as a major policy priority, resulting in a substantial
increase in the share of transport in state fixed-assets investment.
In order to measure both quantity and quality of infrastructure, we select available
attributes of each type of infrastructure, and then do a factor analysis. As indicated by
Table 1, density of roadway and railway was used to characterize the quantity of land
transport, and the share of highway2 over total roadway in terms of length was used to
characterize the quality of land transport infrastructure. We used area of airport lounge,
length and PCN of runway3 in the largest airport in a province, and number of airports
over land area in the province, to measure both quality and quantity of airway transport

2
Highway is defined as the roadway with high-quality pavement in order to ensure high-speed transpor-
tation. It is the highest level in terms of road conditions among three types of roadways in China
(i.e., second-class roadway, first-class roadway, and highway).
3
PCN is a measurement of the pavement condition of runway. It, together with the length of runway,
determines what kind of aircraft an airport can service.

123
740 Transportation (2011) 38:737–752

Table 1 Measurement of transportation infrastructure using factor analysis

Constructs/measured variables Mean Standard Factor % of Internal
deviation loading variance consistency

Land transportation
Density of railway (= railway length/land 0.016 0.015 0.878 74.277 0.827
area)
Density of road (= roadway length/land area) 0.427 0.325 0.854
Quality of road (= highway length/total 0.016 0.013 0.853
roadway length)
Air transportation
Area of airport lounge 56770.719 73357.425 0.865 68.691 0.845
Length of runway 3306.129 374.522 0.729
PCN 67.597 17.516 0.908
Density of airport(= the number of airports/ 0.037 0.055 0.802
land area in a province)
Water transportation
Sum of number of berth/the distance 1.120 3.684 0.940 88.323 N/A
to seaport
Sum of number of deep-berth/the 0.330 0.827 0.940
distance to seaport

Data source: China Statistical Yearbooks, 1999–2008 and Statistical Data on Civil Aviation of China,
1999–2008

infrastructure. With regard to water transportation, the number of berths and deep-water
berths (which can berth high-tonnage ships), adjusted by the distance between the pro-
vincial capital cities and seaports, were taken to measure the quantitative and qualitative
characteristics of water transport infrastructure.
The descriptive statistics of the above variables and the results of factor analysis4 are
reported in Table 1. The indicators are loaded onto their underlying constructs using
principal components method, which is particularly appropriate for theory exploration and
has been widely used because it is not required that the correlation or covariance matrix is
nonsingular (Johnson and Wichern 2008). The Eigenvalues for these factors are above the
1.0 cut-off point, while the percentages of variation are all higher than 68%. The factor
loadings are above 0.7, which is larger than the cut-off point 0.5 (Hair et al. 1998). Values
of Cronbach’s alpha in excess of 0.8 suggest that the measurement scales are sufficiently
reliable.
Table 2 reports the development of transport infrastructure across regions in 2007.
It indicates that the infrastructure in eastern region is much better than that in central and
western region. This is due to greater demand for transportation, central government’s
priorities towards eastern region at the early stage of China’s opening up, as well as the
decentralization process, which gives local government larger autonomy to collect funds
and construct infrastructure. The decentralization process increases regional inequality in

4
Factor analysis is a technique used to reduce a large set of variables to a smaller set of underlying factors,
which helps to detect the presence of meaningful patterns among the original variables. Factor analysis is
used in this study to develop a comprehensive index that can capture both qualitative and quantitative
characteristics of each type of transport infrastructure. In addition, it also helps to avoid the multicolinearity
issue in the econometric model.

123
Transportation (2011) 38:737–752 741

Table 2 Development of transport infrastructure across regions in 2007

Land transport Air transport Water transport

Eastern China 0.849 0.773 0.731

Central China -0.114 -0.454 -0.274
Western China -0.702 -0.406 -0.488
The measures for land, air and water transport infrastructure are based on calculations in Table 1. The values
are standardized and range between -1 and 1. A positive number indicates the score is greater than the mean
value, while a negative number implies the reverse
The factor analysis in Table 1 is based on 10-year data during the study period. Table 2 only reports the
results in 2007. Therefore, the values for each type of transport infrastructure do not sum up to 0 in Table 2

infrastructure endowment because the ability to raise funds mainly depends on local
government revenues and their ability to negotiate with the central government (Demurger
2001). The eastern region with stronger ability to raise funds gains more from the
decentralization process.
The government’s priority towards various types of infrastructure is different across
time. Before the economic reform, the Chinese government favored railway construction,
which can carry a large volume of raw materials at lower cost. However, after 1990s, the
government seems to put more efforts in roadway construction. As indicated by Fig. 1,
from 1998 to 2007, the increase rate of road infrastructure was the highest. Airway
infrastructure was improved substantially from 1998 to 2000, but slowed down since then.
On the other hand, the development of waterway infrastructure was stagnant before 2004,
but the investment has been growing fast afterwards. Table 3 reports the market share of
various transport modes across regions in 2007. It indicates that roadway transport dom-
inated the passenger transport market. In freight transport market, roadway and waterway
accounted for more than 90% of the market in eastern region, while in central and western

Fig. 1 Development of transport infrastructure from 1998 to 2007. Note: The measures for land, air and
water transport infrastructure are based on calculations in Table 1. The values are standardized and range
between -1 and 1. A positive number indicates the score is greater than the mean value, while a negative
number implies the reverse. There is no unit after standardization

123
742 Transportation (2011) 38:737–752

Table 3 Market share of various transport modes across regions in 2007

Rail transport (%) Road transport (%) Water transport (%) Air transport (%)

Passenger transport
Eastern China 6.177 90.255 1.109 2.459
Central China 7.352 91.509 0.435 0.704
Western China 4.532 92.537 1.424 1.507
Freight transport
Eastern China 8.085 71.220 20.629 0.066
Central China 22.182 72.420 5.392 0
Western China 16.377 80.190 3.411 0.022
Data source: China Statistical Yearbooks, 1999–2008 and Statistical Data on Civil Aviation of China,
1999–2008

China, roadway and railway played a dominant role. The market share of air transport was
fairly low in both passenger and freight market across all regions during the period.

Econometric model and data source

The impact of infrastructure endowment on economic growth has been examined in the
framework of new growth theory (Aschauer 1989; Barro 1990). Better transport infra-
structure helps to reduce transaction costs on trade, and leads to pecuniary and allocative
externalities (Banister and Berechman 2000). It is easier for entrepreneurs to adopt new
technologies where the infrastructure is developed. Following previous studies (e.g.,
Demurger 2001; Cieslik and Kaniewsk 2004), we assume that transport infrastructure is an
input that appears in the production function. We estimate a growth model by using Barro-
type framework, which allows us to test conditional convergence.5 The growth equation
at time t in region i is give as follows:
git ¼ ait þ a lnðqi;t1 Þ þ bkit þ clit þ /hit þ stit þ eit ð1Þ
where g represents the annual growth rate of real GDP per capita; q represents GDP per
capita; k, l and h denote physical capital, labor force and labor quality respectively; t is
infrastructure endowment in the region; e is the error term, and a; a; b; c; /; s are coeffi-
cients to be estimated.
Panel data are used from a sample of 31 provinces between 1998 and 2007.6 Standard
fixed and random-effects techniques will be utilized to estimate Eq. 1. Compared to simple
5
An implicit assumption for new growth theory by Barro (1990) is that the market structure is competitive.
The authors believe that this assumption is held in China at the macroeconomic level. According to the first
economic census in China, there were 3.25 million enterprises (among them, only 0.192 million were state-
owned), and 214.604 million workers at the end of 2004. The huge amount of enterprises, employees and
consumers implies a competitive market. Furthermore, the importance of state-owned enterprises becomes
weaker in the Chinese economy. Except several strategic sectors (such as telecommunications, oil and
petroleum), non-state-owned enterprises have become the driving force and dominant power in most sectors.
6
The data after 2007 were not available when we did this research. In addition, The Chinese government
invested heavily in infrastructure to deal with the Asian financial crisis in 1997. To remove the outliers, we
selected the year of 1998–2007 as the study period of this research. Another reason for selecting this study
period is that some key variables are not available before 1997.

123
Transportation (2011) 38:737–752 743

Table 4 Definitions and descriptive statistics of variables

Variables Definitiona Unit Mean Std.
deviation

Land A comprehensive index based on density of Standardized 0 1

transport railway, roadway, and quality of roadway using factor
analysis
Air transport A comprehensive index based on area of airport Standardized 0 1
lounge, length and PCN of runways, and density using factor
of airport analysis
Water A comprehensive index based on sum of number Standardized 0 1
transport of berth/the distance to seaport, and sum of using factor
number of deep-berth/the distance to seaport analysis
Per capita GDP per capita Chinese yuan 5754.913 4444.329
GDP per person
Per capita Real capital stock net of transportation capital per Chinese yuan 2431.041 1876.436
investment capita per person
Population Ratio of total population to land area 10,000 persons/ 0.038 0.049
density km2
Labor Ratio of population with college-degree or above Percentage (%) 5.3 4.2
quality to total population

Please refer to Table 1 for the detailed information on definitions of land, air and water transport

OLS methods, panel data model reduces the effects of the omitted variable bias that may
arise in cross-section regressions. This is particularly important when we use regional data,
since many regional characteristics can not be properly identified. The data used in this
research were from two sources, including various issues of China Statistical Yearbook
(State Statistical Bureau 1999–2008), and Statistical Data on Civil Aviation of China (Civil
Aviation Administration of China 1999–2008).
The dependent variable (git) was measured by annual growth rate of real GDP per
capita. Following previous studies (e.g., Roller and Waverman 2001), real capital stock net
of transportation capital per capita was used as a proxy for investment (kit). Per capita GDP
and per capita investment were converted into fixed prices.7 Labor force (lit) was proxied
by population density, and human capital (hit) was measured by the share of population
with college-degree or above in a province. The measures of transport-related variables
were based on the calculations of factor analysis (see Table 1), which allows us to capture
both quantity and quality of infrastructure endowment, and compare the impacts of various
types of transport infrastructure. Definitions and descriptive statistics of the variables are
reported in Table 4.

Results and discussions

General estimation results

Table 5 reports the estimation results based on pooled OLS, random-effect and fixed-effect
panel data model, respectively. To control for possible heterogeneity, the estimated

7
The GDP deflators were obtained from the World Bank official website. Per capita investment at current
prices was obtained from China Statistical Yearbook.

123
744 Transportation (2011) 38:737–752

Table 5 Estimation results

Pooled Random Fixed effect Fixed effect, Fixed
OLS (1) effect (2) (3) 2SLS (4) effect,
2SLS (5)

Constant 9.701*** 9.733*** 7.727*** 8.223*** 5.995***

(26.70) (21.94) (7.05) (6.14) (3.91)
Per capita GDP (-1) -0.00007 -0.00015*** -0.00038*** -0.00044*** -0.00006
(-1.24) (-3.35) (-6.36) (-4.74) (-0.46)
Per capita investment 17.383*** 18.399*** 21.117*** 22.099**** 18.722***
(7.91) (12.23) (11.15) (7.05) (5.91)
Population density -17.654*** -26.499*** 21.359 13.790 -9.974
(-3.56) (-4.35) (0.68) (0.26) (-0.18)
Labor quality -26.696*** -7.807 32.455*** 35.016*** 29.829***
(-5.81) (-1.21) (3.88) (3.37) (3.13)
Land transport 0.554*** 0.938*** 1.331*** 1.456*** 2.757***
(2.65) (3.68) (3.92) (2.64) (4.02)
Air transport -0.427** -0.287 0.092 1.846 1.425
(2.40) (-1.17) (0.28) (0.88) (0.72)
Water transport 0.184*** 0.318* 0.615*** 0.938*** -9.015**
(1.42) (1.74) (3.01) (4.12) (-2.48)
Square of land transport -1.038***
(-5.34)
Square of air transport -0.289
(-0.65)
Square of water transport 1.234**
(2.56)
Hausman test 30.40***
No. of observations 310 310 310 279 279
R-square 0.580 0.675 0.782 0.698 0.669
Hansen J-statistic of 2.851 (0.091) 1.015
overidentifying (0.314)
restriction (p-value)

(The abbreviation 2SLS indicates two-stage least-squares. Instruments are one-year lagged values of land,
air and water infrastructure, one and two-year lagged values of real per capita GDP, and other non-transport
variables in Eq. 1
*, **, *** Denote significance at the 10, 5 and 1% levels, respectively. t-statistics are in parentheses. The
estimated standard deviations have been corrected using a White matrix

standard deviations have been corrected using a White matrix. Although pooled data and
simple OLS assume unrealistically that individual effects are constant and equal across
regions, we use the results as a point of reference. Fixed effects model relax the assumption
of equality of constant terms, while random effects model further assume that individual
effects might be randomly generated. The use of different panel data techniques is
important to extract as much information as possible from the data. It is also an important
way to test the robustness of the estimation results. Hausman statistic of 30.40 indicates
that the fixed-effect model (model 3) is preferred.
There is possible causality issue between transportation infrastructure and economic
growth. The improvement of transportation infrastructure may cause economic growth
through reduction of travel time and cost, increase of traffic volume, and induced spatial
redistribution of economic activities (Banister and Berechman 2000). On the other hand,

123
Transportation (2011) 38:737–752 745

higher economic growth leads to more government revenue and greater demand on good
infrastructure, which stimulates further improvement of transport conditions. Therefore,
the causality direction can run both ways, and results in endogeneity problem. Following
previous studies (e.g., Demurger 2001), a two-stage least-squares (2SLS) procedure8 (that
is, a fixed-effects panel data model with instrumental variables) is used to deal with the
possible endogeneity issue.9 We instrument the transport-related variables with their own
one-year lagged values, one and two-year lagged values of per capita GDP, and other
exogenous variables10 in Eq. 1. As shown in column 4 of Table 5, Hansen’s J-statistic of
2.851 is insignificant at 5% level,11 indicating that the instruments are acceptable.
The subsequent interpretations and further analyses will be based on the fixed-effect
(2SLS) regression results, reported in column 4 of Table 5. The estimation results show
that land transport infrastructure has significant impact on economic development. The
coefficient for land transport equals to 1.456, indicating a strong impact of land transport.
Table 5 also shows that water transport infrastructure has significant and positive effect on
economic growth, with coefficient estimate of 0.938 and t-value of 4.12. However, the
impact of air infrastructure is insignificant. To conclude, after taking both quantitative and
qualitative characteristics of transport infrastructure into consideration, we find that land
and water transport contributes significantly to regional economic growth, which is con-
sistent with the previous studies (e.g., Demurger 2001; Gunasekera et al. 2008). However,
airway transport has insignificant impact, in contrast to the findings by Yamaguchi (2007).
It should be noted that in this research, we can only identify the overall impact of each type
of transport infrastructure to economic growth, but cannot separately examine the role of
quantitative and qualitative attributes of transport infrastructure.
The above estimation results show a clear picture that in China, land transport infra-
structure has greatest impact on economic development, followed by water transport, while
the impact of airway transport infrastructure is insignificant. This picture is consistent with
the market share of various transport modes across Chinese regions. As shown by Table 3,
land transport (including both roadway and railway) plays a dominant role in both pas-
senger and freight transport market, followed by water transport, while the market share of
air transport is very low. One possible interpretation for the above estimation results is that
since land transport dominates the market, an improvement of land transport condition can
generate greater impacts in terms of travel time and cost reduction, traffic volume increase,
spatial redistribution of economic activities, and hence has strongest influence on economic
growth.
We now turn to the relationship between non-infrastructure variables and economic
growth. As expected, per capita investment and labor quality have significant and positive
effects on economic growth, confirming the impact of physical capital and labor quality.
On the other hand, population density has insignificant effect on economic growth, as
shown in column 4 of Table 5. Population density was used to measure the availability of
labor force, but it may capture crowding effect in a province. The insignificant coefficient
for population density suggests that the crowding effect offsets the labor-availability effect.
Lagged log value of real GDP per capita was included in the model in an attempt to test for
conditional convergence. The coefficient is found to be negative and significant, indicating

8
We use Stata routine xtivreg2 developed by Schaffer (2007).
9
The 2SLS fixed-effects estimation is a standard method to deal with causality issue in the literature.
10
As usual, all other variables except transport-related variables are treated as exogenous.
11
Hansen’s J-statistics are insignificant at 10% level in column 5 of Table 5, columns 4 and 5 of Table 6.

123
746 Transportation (2011) 38:737–752

that provinces with lower levels of GDP per capita tend to grow at a higher rate. Therefore,
the result here supports the conditional convergence hypothesis, although there is no
evidence of absolute convergence between provinces.
The relationship between transportation and growth is likely to be non-linear. Following
Datta and Agarwal (2004), we include the square of the transport infrastructure variables
into the model to examine the nature of returns to scale of transport infrastructure. Column
5 in Table 5 reports the estimation results. The significant and negative coefficient estimate
of the square of land transport indicates diminishing returns. In another words, the effect of
land transport infrastructure is inversely related to its prior level. This implies that land
transport infrastructure contribute more to economic growth in provinces with poor land
transport infrastructure. This result is consistent with previous studies (e.g., Demurger
2001). The estimation results also show that there is a diminishing return to air transport
infrastructure, but the effect is insignificant. The significant and positive coefficient esti-
mate of the square of water transport, together with the negative coefficient of water
transport, suggests that the relationship between water transport infrastructure and eco-
nomic growth is U-shaped. A possible interpretation is due to huge amount of setup cost of
water transport infrastructure. The investment in water transport infrastructure will con-
tribute positively to economic growth only after the investment scale exceeds a threshold
level. Therefore, although column 4 shows that both land transport and water transport
have positive contribution to economic growth, column 5 indicates that the relationships
between the two types of transport infrastructure and economic development are very
different. Further research should be done to identify the exact reasons behind the
divergence.

Robustness tests

Table 5 provides some evidence for robustness of our models by applying a variety of
methods. To estimate Eq. 1, OLS, 2SLS, standard fixed and random-effects panel data
techniques have been utilized. It shows that except pooled OLS, the results based on other
estimation techniques are quite consistent, especially the estimations computed through
fixed-effect (model 3) and fixed-effect 2SLS procedure (model 4). One possible reason for
the inconsistent estimations by pooled OLS is its unrealistic assumption that individual
effects are constant and equal across regions.
In order to further test the robustness of models, we apply the same specification to
subsamples of our dataset. One concern is that in China, some provinces (including Inner
Mongolia, Tibet and Xinjiang) are much larger than others in terms of land area, but are
associated with much poorer transport infrastructure due to the weak demand and geo-
graphical reasons. For the robustness test, we adjusted the sample by excluding the above
three provinces. The estimation results in Table 6 show that despite differences in the
number of observations, coefficient estimates are quite consistent in terms of magnitude
and significance. Although the coefficient for population density is positive in Table 5 but
becomes negative in Table 6, both tables show that its impact is insignificant. Taken
together, the above tests lead us to conclude that our results are quite robust, and not
merely a result of the methods and sample chosen.

The impact of infrastructure across regions

One of the important characteristics of the Chinese economy is unequal development

across regions: the developed eastern region, the developing central region, and the lagging

123
Transportation (2011) 38:737–752 747

Table 6 Robustness test: results with the sample excluding Inner Mongolia, Tibet, and Xinjiang province
Pooled OLS Random Fixed Fixed effect, Fixed
(1) effect (2) effect (3) 2SLS (4) effect,
2SLS (5)

Constant 10.223*** 10.194*** 9.145*** 9.869*** 6.529***

(32.01) (26.36) (7.74) (6.77) (4.01)
Per capita GDP (-1) -0.00001 -0.00006 -0.00028*** -0.00034*** -0.000006
(-0.35) (-1.34) (-4.64) (-4.23) (-0.06)
Per capita investment 13.729*** 14.792*** 16.647*** 17.214*** 16.326***
(8.59) (9.14) (7.89) (6.56) (6.29)
Population density -21.174*** -26.271*** -13.566 -25.176 -9.621
(-5.37) (-5.29) (-0.44) (-0.52) (-0.20)
Labor quality -29.515*** -21.473*** 25.599*** 27.146*** 20.050**
(-6.95) (-3.70) (3.04) (2.84) (2.34)
Land transport 0.870*** 1.109*** 1.749*** 2.096*** 2.899***
(4.61) (4.52) (5.08) (3.86) (4.69)
Air transport -0.199 -0.183 0.077 (0.25) 1.915 (0.91) 1.171
(-1.24) (-0.87) (0.60)
Water transport 0.146 0.223 0.542*** 0.817*** -7.623**
(1.27) (1.28) (2.80) (3.90) (-2.32)
Square of land transport -1.015***
(-5.52)
Square of air transport -0.290
(-0.70)
Square of water transport 1.033**
(2.37)
Hausman test 30.40***
No. of observations 280 280 280 252 279
R-square 0.577 0.653 0.713 0.689 0.669
Hansen J-stat of 2.512 0.803
overidentifying (0.113) (0.370)
restriction (p-value)

The abbreviation 2SLS indicates two-stage least-squares. Instruments are one-year lagged values of land, air
and water infrastructure, one and two-year lagged values of real per capita GDP, and other non-transport
variables in Eq. 1
*, **, *** Denote significance at the 10, 5 and 1% levels, respectively. t-statistics are in parentheses. The
estimated standard deviations have been corrected using a White matrix

western region. Since the late 1990s, the government has put much effort to deal with the
problem of regional inequality, by taking ‘‘Go Western and Central China Strategy’’. One
of the most important policies taken by the government is to develop infrastructure in
Western and Central China. Some policy-makers and researchers argue that construction of
infrastructure in lagged area may contribute more to economic development and reduce
regional inequality (e.g., Demurger 2001). To identify the impact of infrastructure on
regional economic growth, we classify all provinces into three regions based on their
geographic location. These are Eastern China (Beijing, Tianjin, Hebei, Liaoning, Shanghai,
Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and Hainan), Central China (Shanxi,
Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, and Hunan), and Western China (Inner
Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu,
Qinghai, Ningxia, and Xinjiang). A retrospective analysis is used to analyze economic

123
 Table 7 Contribution to regional economic growth gaps relative to the national mean
748

Growth (2SLS) Land transport Air transport Water transport Initial GDP Investment Population density Labor quality Fixed effect

123
Eastern China 0.647 1.249 1.421 0.655 -1.542 2.919 0.468 0.830 23.693
Beijing 0.276 3.289 4.583 0.161 -3.829 9.397 0.664 5.443 -15.284
Tianjin 2.011 3.504 0.180 2.024 -2.662 5.422 0.727 1.926 -6.239
Hebei -0.168 0.504 -0.540 -0.144 0.235 -0.804 -0.026 -0.484 0.958
Liaoning -0.106 0.759 -0.326 -0.156 -0.514 1.301 -0.120 0.653 -1.255
Shanghai 0.307 3.292 6.661 3.019 -6.214 10.958 3.209 3.372 -17.205
Jiangsu 1.462 0.764 1.598 -0.015 -1.001 2.045 0.478 -0.134 -1.252
Zhejiang 1.325 0.310 1.411 0.155 -1.466 4.087 0.122 -0.059 -1.836
Fujian 0.238 -0.284 1.485 -0.295 -0.618 -0.060 -0.119 -0.507 1.471
Shandong 1.499 1.006 0.260 -0.044 -0.391 0.853 0.304 -0.428 0.049
Guangdong 1.187 0.143 -0.080 2.572 -1.108 0.676 0.097 -0.260 0.477
Hainan -0.917 0.449 0.396 -0.069 0.609 -1.765 -0.191 -0.386 -0.512
Central China -0.487 -0.202 -0.811 -0.270 0.670 -1.884 -0.115 -0.364 1.806
Shanxi -0.268 0.077 -0.412 -0.262 0.769 -1.695 -0.225 -0.254 0.916
Jilin -0.054 -0.442 -2.139 -0.320 0.440 -0.618 -0.318 0.212 2.701
Heilongjiang -1.039 -1.028 -0.360 -0.386 0.082 -1.758 -0.406 -0.139 3.074
Anhui -0.953 0.057 -1.448 -0.127 1.002 -2.425 0.101 -0.770 1.567
Jiangxi -0.311 -0.234 -0.921 -0.251 0.969 -2.161 -0.165 -0.574 1.916
Henan 0.175 0.508 0.647 -0.234 0.783 -2.187 0.264 -0.731 0.328
Hubei -0.666 -0.188 0.001 -0.278 0.504 -1.741 -0.083 -0.157 0.801
Hunan -0.782 -0.370 -1.860 -0.305 0.808 -2.487 -0.090 -0.496 3.148
Western China -0.268 -1.010 -0.761 -0.421 0.967 -1.420 -0.352 -0.519 2.181
Inner Mongolia 3.857 -1.522 -1.727 -0.325 0.435 0.937 -0.489 0.001 5.992
Guangxi -0.511 -0.488 -1.855 -0.338 1.081 -2.862 -0.241 -0.786 3.853
Chongqing -0.424 -0.200 -1.654 -0.422 0.827 -0.907 -0.011 -0.752 1.736
Transportation (2011) 38:737–752
 Table 7 continued

Growth (2SLS) Land transport Air transport Water transport Initial GDP Investment Population density Labor quality Fixed effect

Sichuan -0.355 -0.940 0.559 -0.437 1.001 -2.284 -0.278 -0.766 1.530
Guizhou -0.979 -0.774 -0.352 -0.413 1.588 -3.403 -0.213 -0.863 1.779
Yunnan -2.180 -0.919 -0.422 -0.452 1.120 -2.527 -0.365 -1.043 1.321
Tibet 0.695 -2.006 -0.325 -0.513 1.032 -0.547 -0.514 -1.559 4.039
Shaanxi -0.243 -0.448 -0.804 -0.354 1.003 -1.664 -0.270 -0.065 1.219
Gansu -1.075 -1.300 -0.241 -0.433 1.266 -2.831 -0.438 -0.701 2.185
Transportation (2011) 38:737–752

Qinghai 0.060 -1.809 -1.580 -0.446 0.893 -0.562 -0.507 -0.336 3.485
Ningxia -0.428 -0.048 -1.023 -0.398 0.950 -0.197 -0.398 -0.102 -0.413
Xinjiang -1.631 -1.663 0.286 -0.514 0.405 -0.192 -0.501 0.748 -0.551
The above results are based on the fixed effect panel data model (2SLS) in Table 4
The bold numbers are computed by averaging the values of all provinces included in the region
749

123
750 Transportation (2011) 38:737–752

growth of Chinese regions in light of their infrastructure. We firstly compute the gap
between the predicted per capita GDP growth rate of each province and the national
average, and then decompose this gap into the independent variables.
The estimation results are presented in Table 7. It shows that eastern China has had
much higher per capita GDP than the national average level. The above-average perfor-
mance comes from a relatively developed transport infrastructure as well as higher
investment and labor quality. Beijing, Tianjin and Shanghai are the biggest beneficiary in
developing land transport network, while water transport contributes most to economic
growth in Shanghai, Tianjin and Guangdong province. Per-capita GDP growth rate in
Central and western China is lower than the national average. Poor transport network has
affected negatively their economic performances. For instance, four remote provinces
(Tibet, Qinghai, Xinjiang and Inner Mongolia) have suffered most from poor land transport
conditions. The above results suggest that in the past decade, the lack of adequate transport
infrastructure endowment is an important reason behind lagged economic development in
Western and Central China.

Conclusions

This article investigates the impact of infrastructure on economic growth in China. Using
panel data from a sample of 31 Chinese provinces from 1998 to 2007, the estimation
results provided strong evidence that differences in transport infrastructure, physical
capital and labor quality did account for a significant part of the observed variation in the
economic growth of provinces. Both land transport (including roadway and railway) and
water transport infrastructure had significant impacts, but the contribution of airway
transport infrastructure was weak. Furthermore, land transport infrastructure contributed
more to economic growth in provinces with poor land transport conditions, while the
investment in water transport infrastructure contributed positively to economic growth only
after the investment scale exceeded a threshold level. The above results were robust to a
variety of alternative methods, the exclusion of possible outliers, and consideration of
endogeneity.
We also found that various modes of infrastructure contributed to economic perfor-
mance differently across regions. Improved transport conditions (including land, and water
transport) played an important role in economic success in eastern region. In lagged
western region, poor land infrastructure were important factors behind its weak economic
performance. The above results suggested that economic policies that improve transport
infrastructure had a nonnegligible impact in promoting per capita income convergence
among Chinese regions. Moreover, expanding and upgrading the network of land trans-
portation in lagged areas might have higher growth payoff, and would be particularly
useful to allow for the development of efficient market and knowledge diffusion in China.
The relationship between China’s transport improvement and its economic performance
is worth examining, not only because the formulation of its own future policies may
depend on these experiences, but also because China is the stellar example for other
developing countries to emulate. The results in this article imply that in a developing
country like China, the improvement of transport conditions does play an important role in
developing regional economy, and uneven distribution of transport infrastructure results in
regional disparities. Of course, there is a need for future research to establish the validity of
the findings reported here in different country and temporal contexts. Due to data

123
Transportation (2011) 38:737–752 751

limitation, the exact economic process through which transportation improvement induces
economic growth is left for future research agenda as well.

Acknowledgments This research is supported by the Program for New Century Excellent Talents in
University (NCET-09-0301), Program for Innovative Research Team, ‘‘211 Program’’ and Excellent
Doctoral Dissertation Naturing Program in UIBE.

References

Aschauer, D.A.: Is public expenditure productive? J. Monetary Econ. 23(2), 177–200 (1989)
Baldwin, R.E., Forslid, R.: The core-periphery model and endogenous growth: stabilizing and destabilizing
integration. Economica 67, 307–324 (2000)
Banister, D., Berechman, J.: Transport Investment and Economic Development. UCL Press, London (2000)
Barro, R.J.: Government spending in a simple model of endogenous growth. J. Polit. Econ. 98(5), 103–125
(1990)
Berechman, J., Ozmen, D., Ozbay, K.: Empirical analysis of transportation investment and economic
development at state, county and municipality levels. Transportation 33, 537–551 (2006)
Ciccone, A., Hall, R.: Productivity and density of economic activity. Am. Econ. Rev. 86, 54–70 (1996)
Cieslik, A., Kaniewsk, M.: Telecommunications infrastructure and regional economic development: the case
of Poland. Reg. Stud. 38(6), 713–725 (2004)
Civil Aviation Administration of China: Statistical Data on Civil Aviation of China. China Civil Aviation
Press, Beijing (1999–2008)
Datta, A., Agarwal, S.: Telecommunications and economic growth: a panel data approach. Appl. Econ. 36,
1649–1654 (2004)
Demurger, S.: Infrastructure development and economic growth: an explanation for regional disparities in
China? J. Comp. Econ. 29, 95–117 (2001)
Evans, P., Karras, G.: Is government capital productive? Evidence from a panel of seven countries.
J. Macroecon. 16, 271–279 (1993)
Fleisher, B.M., Chen, J.: The coast–noncoast income gap, productivity and regional economic policy in
China. J. Comp. Econ. 25(2), 220–236 (1997)
Gillen, D.W.: Transportation infrastructure and economic development: a review of recent literature. Logist.
Transp. Rev. 32(1), 39–62 (1996)
Gunasekera, K., Anderson, W., Lakshmanan, T.R.: Highway-induced development: evidence from Sri
Lanka. World Dev. 36(11), 2371–2389 (2008)
Hair, J.F., Anderson, R.E., Tatham, R.L., Black, W.C.: Multivariate Data Analysis. Prentice Hall, Engle-
wood Cliffs (1998)
Hong, J.: Transport and the location of foreign logistics firms: the Chinese experience. Transp. Res. A 41,
597–609 (2007)
Johnson, R.A., Wichern, D.W.: Applied Multivariate Statistical Analysis, 6th edn. Prentice Hall, Upper
Saddle River (2008)
Krugman, P.: Increasing returns and economic geography. J. Polit. Econ. 99, 483–499 (1991)
Lall, S.V.: Infrastructure and regional growth, growth dynamics and policy relevance for India. Ann. Reg.
Sci. 41(3), 581–599 (2007)
Lin, S.: Public infrastructure development in China. Comp. Econ. Stud. XLIII(2), 83–109 (2001)
Mody, A., Wang, F.: Explaining industrial growth in coastal China: economic reforms and what else? World
Bank Econ. Rev. 11(2), 293–325 (1997)
Naughton, B.: The Chinese Economy: Transitions and Growth. MIT Press, Cambridge (2007)
Otto, G., Voss, G.: Public capital and private production in Australia. South. Econ. J. 62, 723–738 (1996)
Roller, L.H., Waverman, L.: Telecommunications infrastructure and economic development: a simultaneous
approach. Am. Econ. Rev. 91(4), 909–923 (2001)
Romer, P.M.: Increasing returns and long-run growth. J. Polit. Econ. 94(5), 1024–1037 (1986)
Schaffer, M.E.: xtivreg2: Stata module to perform extended IV/2SLS, GMM and AC/HAC, LIML and
k-class regression for panel data models. http://ideas.repec.org/c/boc/bocode/s456501.html (2007)
Seitz, H.: Public capital and demand for private inputs. J. Public Econ. 54, 287–307 (1994)
Shirley, C., Winston, C.: Firm inventory behavior and the returns from highway infrastructure investments.
J. Urban Econ. 55, 398–415 (2004)
Solow, R.M.: A contribution to the theory of economic growth. Quart. J. Econ. 70, 65–94 (1956)

123
752 Transportation (2011) 38:737–752

State Statistical Bureau: China Statistical Yearbook. China Statistical Publishing House, Beijing
(1999–2008)
Yamaguchi, K.: Inter-regional air transport accessibility and macro-economic performance in Japan. Transp.
Res. E 43, 247–258 (2007)

Author Biographies

Junjie Hong is Professor at the School of International Trade and Economics, University of International
Business and Economics, Beijing, China. Dr. Hong obtained his PhD degree in economics from the National
University of Singapore. His research interests cover Transport and Logistics, Urban Economics and
International Economics. His publications have appeared in international refereed journals, such as
Transportation Research Part A, Transportation Journal, World Economy, Journal of Regional Science,
Regional Studies, Urban Studies.

Zhaofang Chu is a doctoral student in Transport and Logistics Management at the School of International
Trade and Economics, University of International Business and Economics. His research interests include
Transport, Logistics and Supply Chain Management. He has published in Transportation Journal, Industrial
Management & Data Systems.

Qiang Wang is Professor at the School of International Trade and Economics, University of International
Business and Economics, Beijing, China. Dr. Wang received his PhD degree from City University of Hong
Kong. His research interests include Chinese Logistics and Supply Chain Management. He has published in
such journals as Journal of the Operational Research Society, International Journal of Production
Research, Journal of Supply Chain Management, Transportation Journal and Supply Chain Management:
An International Journal.

123