Received: 5 September 2017
Revised: 26 November 2017
Accepted: 5 January 2018
DOI: 10.1111/rsp3.12114
ORIGINAL ARTICLE
Do transport infrastructure spillovers matter for economic growth? Evidence on road and railway transport infrastructure in Iranian provinces Zahra Dehghan Shabani
|
Sima Safaie
Department of Economics, Shiraz University, Shiraz, Iran
Abstract
Correspondence Zahra Dehghan Shabani, Department of Economics, Shiraz University, Shiraz, Iran. Email:
[email protected]
This study analysed the spatial spillover effects of road and railway transport infrastructure on economic growth in the provinces of Iran. More specifically, it examined the direct, indirect (spatial spillover), and total effects of such infrastructure on the output growth of the provinces. To this end, maximum likelihood was used
JEL Classification: R11; R40; C51; C31; C33
to estimate a spatial Durbin model for the provinces over the period 2001 to 2011. The results showed that main road and railway transport frameworks exert a positive and significant direct effect on the economic growth of the provinces. The main road and railway transport infrastructure in each province has a significant and positive indirect effect on the economic growth of the other provinces, indicating that improvements to individual infrastructure have spatial spillover effects on the economic growth of the other provinces. The conclusion drawn from these findings is that diffusion effects overcome agglomeration effects. Finally, total railway length has a positive spatial spillover effect on regional economic growth. KEYWORDS
spatial spillovers, transport infrastructure, Iranian provinces, economic growth, spatial Durbin model
1
|
I N T RO DU CT I O N
Studies on regional economic growth and development often focus on economic dependence among regions, generally presupposing that the economic conditions in one region influences those in other regions. Correspondingly, the assumption of mutual influence extends to economic growth. Examples of such studies are those of Chua (1993), Ades and Chua (1997), Quah (1993), and Moreno and Trehan (1997), who stated that regional growth rate depends on -------------------------------------------------------------------------------------------------------
© 2018 The Author(s). Regional Science Policy and Practice © 2018 RSAI
Reg Sci Policy Pract. 2018;10:49–63.
wileyonlinelibrary.com/journal/rsp3
49
50
DEHGHAN SHABANI AND SAFAIE
observable regional characteristics, such as labour and capital; unobservable regional characteristics, such as climate; and inter‐regional spillovers. Spillovers more effectively affect regional growth rate than do unobservable regional characteristics and can be considered as equally important as observable regional effects (Conley & Ligon, 2002). Different studies have shown that the effects of transport infrastructure on growth at the regional level generally differ from the results observed at the national level. Researchers argued that this difference is due to the presence of significant spillover effects between regions. A spillover indicates that the positive benefits of investing in infrastructure arise not only from investments made in each province but also from the positive and negative side effects of investments in the transport networks of neighbouring provinces (Yu, De Jong, Storm, & Mi, 2013). Note, however, that the spillover effects of transportation infrastructure on regional economic growth may be either positive or negative. On the one hand, transportation infrastructure enables high‐growing regions to play the role of stimulator for low‐growing regions through diffusion effects that lead to positive spillovers because such infrastructure comprises networks that connect the economic activities of different regions to an entire framework (Xueliang, 2008). On the other hand, improved transport infrastructure facilitates the migration of labour from underdeveloped to developed areas through agglomeration effects, thereby causing negative spillover effects. Therefore, the spillover effects of transportation infrastructure on a region's economic growth depend on diffusion and agglomeration effects. Previous studies on the effects of transportation infrastructure on economic growth can be divided into two strands. The first covers studies that examined the relationship between transportation infrastructure and economic growth without consideration for spatial spillover effects (Table 1). The second strand of research focused on the effects of spatial spillover from transportation infrastructure on economic growth. These studies generated three varying results: transportation infrastructure has positive spillover effects on economic growth; transportation infrastructure exerts negative spillover effects; and transportation infrastructure does not pose any spatial spillover or mixed spatial spillover effects on economic growth (Table 2). Numerous studies have been devoted to the spillover effects of transportation infrastructure on economic growth, but research on the influence of total and main transportation infrastructure is remarkably lacking.1 Some studies inquired into the effects of highways, but these highways are not the only major thoroughfares in a region. In this study, a main road network was defined as encompassing highways, freeways, wide main roads, ordinary main roads, and main railways. Main railways do not include suburban railways and manoeuvring railway lines. A clear definition is important because the extent of spatial spillover from total infrastructure differs from the magnitude of spatial spillover from main infrastructure. Accordingly, this study focused on road and railway frameworks. The current research contributes to the literature in a number of ways. It considered the spatial spillover effects of total and main road networks and total and main railway networks, deriving results that can serve as a guide for policy‐ makers in allocating transportation infrastructural investments. All other studies on transport infrastructure spillovers were conducted in developed countries, such as the United States and Spain, or in newly industrialized nations, such as China. By contrast, the present study obtained evidence on the spillover effects of transportation infrastructure in Iranian provinces, some of which were characterized by low economic development and limited transport infrastructure during most of the periods under analysis. Our findings can therefore provide lessons that may be useful for developing economies that want to enhance their transportation infrastructure. In many of the studies presented in Table 2, the researchers used transportation capital stock, transportation investment, or highway
1 Main roads include highways; freeways; wide main roads (roads with asphalt pavement or concrete for passing motor vehicles and part of national networks) with a total width of 13 to 13.30 metres; and ordinary main roads with a width of 11 to 11.30 metres. Total road networks include highway, freeways, wide main roads with a total width of 13 to 13.30 metres, ordinary main roads with a width of 11 to 11.30 metres, byways, and rural roads. Main rail networks are networks where lines are located directly along the lines between the two sides of a station. Total rail networks encompass main railways, suburban railways (lines used at station stations for acceptance and deployment of trains), and maneuvering railway lines (lines located at stations for the purpose of arranging and separating trains).
51
DEHGHAN SHABANI AND SAFAIE TABLE 1
Summary of studies on the relationship between transportation infrastructure and economic growth
Authors
Countries
Period
Methodology
Findings
Aschauer (1989)
US
1949–1985
OLS, TSNLS
TI ! G
Duffy‐Deno and Eberts (1991)
US
1980–1984
SPDM
GTI ! IP
Garcia‐Mila and McGuire (1992)
US
1969–1983
PD
HW ! GDP
Haque and Kim (2003)
15 developing countries
1970–1987
SEPD
GTI ! G
Boopen (2006)
South African countries Developing countries
1980–2000 1985–2000
DPD
TI ! G
Kustepeli, Gulcan, and Akgungor (2008)
Turkey
1970–2005
CA and GC
No relationship between HWE and G
Hong, Chu, and Wang (2011)
China
1998–2007
PD
T !G
Akbarian and Ghaedi (2011)
Iran
1961–2006
VAR
TI ! GDP
Daii karimzadeh, Emadzadeh, and Kamkar Delakh (2009)
Iran
1973–2008
ARDL
GTI ! G
Banerjee, Duflo, and Qian (2012)
China
1986–2000
Descriptive method
T ! GDP
Carruthers (2013)
11 southern and eastern Mediterranean countries
‐
‐
TI ! G
Bosede, Abalaba, and Afolabi (2013)
Nigeria
1981–2011
OLS
TI ! G
Sun (2013)
China
1978–2011
PD
T ! GI
Ahmed, Abbas, and Ahmed (2013)
Pakistan
‐
DCGE
TI ! G
Pradhan, Norman, Badir, and Samadhan (2013)
India
1970–2012
ARDL
T ↔ GDP
Agbelie (2014)
40 countries
1992–2010
OLS, RF
TI ! G
Afsharpour, Mehrabi, and Pahlavani (2014)
Iran
2000–2011
PD
T ! GA
Kodongo and Ojah (2016)
45 sub‐Saharan African countries
2000–2011
System GMM
TI ! G
þ
þ
þ
þ
þ
þ
þ
þ
þ
þ
þ
þ
þ
þ
þ
þ
Notes: (1) OLS = ordinary least squares; NLS = non‐linear least squares; TSNLS = two‐stage NLS; PD = panel data; SPDM = system estimation dynamic PD; CA = cointegration analysis; GC = Granger causality; ARDL = autoregressive distributed lag; RE = random effect; VAR = vector autoregressive; DCGE = dynamic computable general equilibrium model; TI = transportation infrastructure investment; G = economic growth; IP = per capita income; HW = highway; GDP = gross domestic product; GTI = government transportation infrastructure investment; HWE = highway transportation expenditures; T = transportation þ infrastructure investment; GI = growth inequality; GA = agricultural growth. ! denotes a positive relationship; ↔ denotes bidirectional causality.
capital stock as a measure of transportation infrastructure. A more effective approach is to present transport infrastructure variables on the basis of their physical forms to reflect their distribution. In countries such as Iran, the cost of building roads is not the same across all provinces; in areas where construction costs and investments are excessively high, few options for building road networks are available. The rest of the paper is structured as follows. Section 2 presents the model, methodology, and data employed in this work. Section 3 discusses the empirical results, and Section 4 concludes with a summary, policy implications, and suggestions for future work.
52
DEHGHAN SHABANI AND SAFAIE
TABLE 2
Summary of studies on the spatial spillover effects of transportation infrastructure on economic growth
Authors
Countries
Period
Transportation infrastructure measure
Findings
Boarnet (1998)
US
1969–1988
Highway capital stock
SSPH ! pro
Pereira and Roca‐ Sagales (2003)
Spain
1970–1995
Public capital stock in transportation and communication
SSP ! pro
Pereira and Andraz (2004)
US
1977–1999
Highway infrastructure
SSPH ! pro
Cantos, Gumbau‐Albert, and Maudos (2005)
Spain
1965–1995
Public capital in transport infrastructure
SSP ! G
Berechman, Ozmen, & Ozbay, 2006
US
1990–2000
Highway capital
SSPH ! G
Liu, Chen, and Zhou (2007)
11 cities of China
1994 to 2003 Highway capital
SSPH ! G
Ozbay, Ozmen, and Berechman (2007)
New York/New Jersey metropolitan area
1990–2000
Highway capital
SSPH ! pro
Xueliang (2008)
China
1993–2004
Transportation infrastructure capital stock
SSP ! G
Hu and Liu (2010)
China
1985–2006
Transportation infrastructure capital stock
SSP ! G
Liu (2010)
China
1978–2008
Highway and waterway capital stock
Before 2001:
Jiwattanakulpaisarn, US Noland, and Graham (2011)
−
þ
þ
þ
þ
þ þ
þor−
þ
þ
SSPH; SSPW ! G After 2001: − SSPH; SSPW ! G þ
1984–1997
Highway capital
SSPH ! pro þ
Álvarez‐Ayuso and Delgado‐Rodriguez (2012)
Spain
1980–2008
Public capital on high‐capacity road networks
SSPRO ! PPro
Yu et al. (2013)
China
1978–2009 1978–1990 1991–2000 2001–2009
Transportation capital stock
SSP ! G
Wang, Deng, and Wu (2014)
China
1990–2010
Transportation capital stock
SSP ! G
Chen and Haynes (2015)
US
1991–2009
þ Public highways, railways, SSPH, SSPR, SSPA ! G transit systems, no effect G SSPT ! and airports
Jiang et al. (2015)
China
1985–2012
Transportation investment
SSP ! G
Álvarez, Condeço‐ Melhorado, Gutiérrez, and Zofío (2016)
Spain
1980–2007
Transportation capital stock
SSP ! pro
Álvarez, Barbero, and Zofío (2014)
Spain
1980–2007
Infrastructure capital stock (internal and imported capital)
SSP ! G
Li, Wen, and Jiang (2017)
31 provinces in new 2005–2014 silk road Economic Belt
Density of railway and highway networks
SSPH, SSPR ! G
þ
−
þor−
þor−
þ
þ
þ
þor−
þ
Notes: ! denotes a positive relationship; ! denotes a negative relationship; ! denotes a positive or negative relationship. Pro = production, G = economic growth, PPro = private production, SSP = spatial spillover of transportation, SSPH = spatial spillover of highways, SSPR = spatial spillover of railways, SSPA = spatial spillover of airports, SSPT = spatial spillover of transit ways, SSPRO = spatial spillover of roads, U>S = United States
53
DEHGHAN SHABANI AND SAFAIE
2
|
M O D E L SP E C I F I C A T I O N
Given that this study centered on an analysis of the spatial spillover effects of transportation infrastructure on economic growth, a model that considers transportation infrastructure and spillovers as important factors was necessary. Among regional growth models, new economic–geographic versions are spatial models wherein a neighbourhood is a positive factor for efficiency and growth. In these models, the determinants of efficiency and growth are increasing returns and transportation costs. This study used a new economic–geographic model in which the concentration of transportation and industrial activities is an important stimulating factor for economic growth. We adopted a Cobb‐Douglas production function to examine the relationship between production and input factors in a region. In this production function, transportation infrastructure, along with the concentration of industrial activities on productivity, is regarded as an external factor (Boarnet, 1998; Jiwattanakulpaisarn et al., 2011). Transportation infrastructure and industrial concentration are treated as production function frontier shifters, which increase the efficiency of other inputs. The production function of region i is considered as follows: βk βh Y it ¼ Ait zw it K it Lit ;
(1)
where Y represents the real gross domestic product (GDP) without oil, A is the total factor productivity, zw it denotes the concentration of transportation infrastructure (roads and railways) and industrial activities in a province, and K and L represent capital and labour, respectively. Assuming that the total factor productivity has an exponential relationship with road, railway, and industrial concentration: Ait ¼ RoadβitRo RailβitRa AgβitAG eit ;
(2)
where Road and Rail refer to the total and main road and railway lengths in each province (kilometre), and Ag represents the concentration of industrial activities. The log linear transformation of Equation (2) becomes: Lyit ¼ βro Lroadit þ βra Lrailit þ βag Lagit þ βk Lkit þ eit ; in which y is the logarithm of the real GDP per capita without oil per capita; Lroadit and Lrailit are the logarithms of total or main road length per capita and railway length per capita, respectively; and Lkit denotes the real capital stock per capita. The construction of interstate transportation infrastructure (for example) can improve a network through the efficient connection of states, thereby enabling a redistribution of existing resources for production. An improved transportation network in a state can provide a more efficient and integrated network to the state and, consequently, contribute to economic activities in spatially related states. Economic activities can also be reallocated from states with poor transportation infrastructure to states with well‐maintained transportation systems. Thus, the construction or improvement of transportation infrastructure in one state can adversely affect the output of private sectors in neighbouring states with less‐developed transportation infrastructure (Tong, Yu, Cho, Jensen, & Ugarte, 2013). In this case, the transportation infrastructure of one state exerts spatial spillover effects on other states. The spatial Durbin model (SDM) is useful for testing hypotheses on spatial spillover because the model captures the contribution of transportation infrastructure in a given region to economic growth both in and outside of the region. The SDM includes a spatially lagged dependent variable (WY) and spatially lagged explanatory variables (WX) (Vega & Elhorst, 2013). It is the only model that produces an unbiased estimator in all possible spatial data modeling processes, according to LeSage and Pace (2009). The omission of variables is also less likely to be observed
54
DEHGHAN SHABANI AND SAFAIE
in the SDM because of the inclusion of spatial dependence in the explanatory variables (Tong et al., 2013). The following equation represents the framework used for the SDM in this research: Y it ¼ α þ βX it þ ρ∑Nj¼1 W i; j Y jt þ θ∑Nj¼1 W i; j Xjt þ Ui þ γt þ eit ;
(3)
where Yit stands for the dependent variable in region i at time t, α is a constant, Xit is a vector of explanatory variables, β is a vector of parameters, Wi,j is the i, jth element in a 28 × 28 spatial weight matrix, ∑Nj¼1 W i; j Y jt refers to the spatial lag of the dependent variable, and coefficient ρ shows the effects of the regional growth of neighbouring provinces on the regional growth of province i. Furthermore, ∑Nj¼1 W i; j Xjt is the spatial lag of the explanatory variables; coefficient θ represents the effects of neighbouring provinces’ logarithm road per capita (total or main), railway per capita (total or main), industrial concentration, and real capital stock per capita on the regional growth of each province; eit pertains to an error term with a 0 mean and constant variance; Ui refers to spatial fixed effects; and γt stands for time‐period fixed effects. Spatial and time‐period fixed effects are used to account for spatial and temporal heterogeneities, respectively. The final model is expressed thus: Lyit ¼ α þ βrai Lrailit þ βroi Lroadit þ βag Lagit þ βk Lkit þ ρ∑Nj¼1 W ij Lyjt þ θ1 ∑Nj¼1 W ij Lrailjt þ θ2 ∑Nj¼1 W ij Lroadjt þ
θ3 ∑Nj¼1 W ij Lagjt
þ
θ4 ∑Nj¼1 W ij Lkjt
(4)
þ Ui þ γt þ eit ;
where i represents a province, t represents a year, and j represents other provinces (i ≠ j). The final model was estimated two times: once by using the total road and railway networks and again by using the main road and railway networks. A weight matrix based on distance decay was also used because in spatial matrices constructed on the basis of contiguity, having two units with a shared boundary yields a matrix element with a value of 1; otherwise, this value is 0. This means that a spatial unit can both affect and not affect another spatial unit. This phenomenon, however, is limited to observations with a shared boundary, that is, the absence of differences between non‐neighbouring provinces. Considering the nature of transportation infrastructure, with spillovers that are not confined to neighbouring provinces, this study used the distance matrix below:
W i; j
8 < 1 ¼ di; j : 0
9 i≠j= ; ; i¼ j
(5)
where di,j is the distance between provinces i and j (distance metric is Euclidean distance and distance variables (X‐coordinate and Y‐coordinate) are centroids). In the SDM, the inclusion of a spatial lagged dependent variable as an explanatory variable produces an endogeneity problem because of the correlation between the regressor (WY) and residual (e). As a result, traditional ordinary least squares (OLS) estimation may be biased and inconsistent. To prevent this endogeneity issue, we employed maximum likelihood estimation based on the conditional log‐likelihood function of the model (Tong et al., 2013).
2.1
|
Data
The GDP statistics without oil from 2001 to 2011 were extracted from provincial statistical yearbooks published by the Statistical Centre of Iran. We calculated the real GDP per capita using the price index and population of a province. To determine the concentration of industrial activities, we used Nakamura and Paul's (2009) index: LAgj ¼
Xj J
∑ Xj
¼
Xj j ¼ 1; …; J; X*
(6)
j¼1
where X* is the total added value in the industrial sector, Xj is the industrial value added of region/province j, and LAgj shows the industrial sector concentration in region j. Nakamura and Paul's index falls between 0 and 1. If the industrial
55
DEHGHAN SHABANI AND SAFAIE
sector is completely concentrated in one area, the value of the index is 1, and if the industrial sector is distributed in some large areas with very small shares, the indicator tends toward 0. To calculate the index, we used data from the provincial statistical yearbooks published by the Statistical Centre of Iran. To calculate the real capital stock per capita, the government credit for development and credits to the private sector were considered the investment variables for a province. The specific method of calculation used was that developed by Berlemann and Wesselhöft (2014). This method requires the calculation of the initial capital stock K0 from investments IN0, the long‐term growth rate of investments λ, and the rate of capital depreciation δ. (We assumed IN0 . We therefore used the entire time series of λ investments (INt) in province i at time t and estimated the equation Ln(INt) = α + λt + εt using the OLS method. We
that δ = 0 in the calculation of initial capital stock.) That is, K 0 ¼
also estimated λ as a measure of investment growth. After we obtained the initial capital stock using K 0 ¼
IN0 , we λ
K t−1 þ It . Taking into account a 5% capital depreciation, we can calcu1þδ late the capital stock in different years (Berlemann & Wesselhöft, 2014). Note that the real capital stock per capita was adopted the perpetual inventory method, K t ¼
calculated using a province's price index and population—data that were collected from provincial statistical yearbooks. In this study, transportation infrastructure was regarded as encompassing roads and railways (total and main networks), with related information extracted also from the provinces’ statistical yearbooks.
2.2
|
Spatial autocorrelation of provincial real GDP per capita without oil in Iran
To examine the spatial dependence of the real GDP per capita, Moran's I statistic (1950) was adopted as a useful tool for measuring the degree of dependence between a variable in one region and the same variable in a neighbouring region. Moran's I statistic is expressed as follows: n∑ni¼1 ∑nj¼1 W ij X i −X Xj −X I¼ 2 ; ∑ni¼1 ∑nj¼1 W ij ∑ni¼1 Xi −X
(7)
in which nis the number of provinces that were examined in this study (28 provinces), Xi is the real GDP per capita of the provinces, X denotes the average real GDP per capita of all the provinces, and Wrepresents the spatial weight matrix. The null hypothesis (H0) states that no spatial autocorrelation exists. The results derived on the basis of Moran's I statistic are shown in Table 3. The value of the statistic is positive and statistically significant for all the studied years. A spatial
TABLE 3
Global Moran's I statistic for regional real GDP per capita without oil
Year
Moran's I value
p‐value
2001
0.227
0.015
2002
0.219
0.012
2003
0.218
0.012
2004
0.264
0.007
2005
0.263
0.005
2006
0.285
0.009
2007
0.293
0.004
2008
0.273
0.005
2009
0.159
0.045
2010
0.159
0.049
2011
0.157
0.020
Note: Adjacency matrix was constructed based on distance.
56
DEHGHAN SHABANI AND SAFAIE
autocorrelation in the real GDP per capita of the Iranian provinces exists, with significance at the 5% level. The positive Moran's I index suggested that the real GDP per capita is similar across contiguous provinces. That is, provinces with a high level of real GDP are contiguous to one another; the same holds for provinces with a low level of real GDP. The global Moran's I statistic can be helpful in determining a general spatial autocorrelation trend, but its use would mean that obtaining data on provinces that do not follow this trend would be impossible. As an alternative, the local Moran's statistic (map) proposed by Anselin (1996) was used for a closer examination of spatial autocorrelation. The local Moran and cluster maps of Iranian provinces in 2001 and 2011 are presented in Figures 1 to 4.
3
|
EMPIRICAL RESULTS
Spatial or non‐spatial panel data models with fixed or random effects are selected on the basis of relevant available studies and some statistical criteria. We selected spatial and non‐spatial fixed effects models with reference to the Hausman test (Hausman, 1978) results presented in Tables 4 and 5. Table 4 displays the estimation results of the non‐spatial panel data model. The models included two‐way fixed effects, spatial fixed effects, time fixed effects, and a pooled OLS. The results concerning the likelihood ratio (LR) test are also provided in Table 4. This test was used to account for spatial and time‐period fixed effects. On the basis of the findings, the null hypotheses on the joint non‐ significance of the time‐period fixed effects and the joint non‐significance of spatial fixed effects are rejected. This rejection indicates that a model must include both spatial and time‐period fixed effects, thus prompting the selection of the two‐way fixed effects model. It should be mentioned here, however, that the effects of spatial interaction between the variables have not been accounted for in this model. Consequently, this finding cannot be reliable because the coefficient estimates may be biased. To determine whether either spatial lag dependence or spatial error dependence exert any statistically significant effects on labour productivity in the different provinces, we performed a Lagrange multiplier (LM) test after running the LR test. The rejection of the null hypothesis indicated that the tested spatial interaction effects should be
FIGURE 1
Cluster map of Iranian provinces, 2001
DEHGHAN SHABANI AND SAFAIE
FIGURE 2
Local Moran map of Iranian provinces, 2001
FIGURE 3
Cluster map of Iranian provinces, 2011
57
incorporated into the models (whether spatial lag or spatial error). As shown in Table 5, the null hypothesis of the non‐ spatial model has been rejected only in favor of the spatial lag model. Consequently, the spatial autoregressive (SAR) model or the SDM is recommended for use.
58
DEHGHAN SHABANI AND SAFAIE
FIGURE 4 TABLE 4
Local Moran map of Iranian provinces, 2011 Estimation results of panel data without spatial interaction effects Model 1
Variable
Spatial fixed
Model 2 Time‐ period
Spatial and time‐period
Spatial fixed
Time‐ period
Spatial and time‐period
LK
0.623*** (0.000) 0.505*** (0.000)
0.528*** (0.000)
0.568*** (0.000) 0.516*** (0.000) 0.544*** (0.000)
LAg
0.147*** (0.000) 0.066*** (0.000)
0.062*** (0.000)
0.111 (0.191)
0.090*** (0.000) 0.078*** (0.000)
0.078 (0.458)
0.005 (0.589)
Lrailt
–0.836 (0.204)
Lroadt
–1.395** (0.036) 0.397** (0.047) –0.775*** (0.002)
Lrails
LR spatial fixed effects test
0.562** (0.019)
0.058* (0.109)
–0.471*** (0.000) 0.228*** (0.000) 0.305*** (0.000)
Lroads Hausman
0.133** (0.026)
11.60** (0.020) 4.35*** (0.000)
LR time‐ 186.71*** (0.000) period fixed effects test
31.32*** (0.000) 5.07*** (0.000)
207.48*** (0.000)
Notes: The p‐values are enclosed in parentheses. ***, **, and * denote statistical significance at the 0.01, 0.05, and 0.1 levels, respectively. Lrailt, Lroadt, Lrails, Lroads refer to the logarithms of total rail length per capita, the logarithms of total road length per capita, the logarithms of main rail length per capita, the logarithms of main road length per capita, respectively.
Table 6 presents the estimation results on the effects of transportation infrastructure spillovers on regional economic growth. We used the SDM with both spatial and time‐period effects. On the basis of the Hausman test, the fixed effects estimation method was chosen for both models 1 and 2 (Table 6). In model 1, we considered the
59
DEHGHAN SHABANI AND SAFAIE TABLE 5
Diagnostic test results on spatial interaction effects
Test
Statistic
p‐value
0.315
0.751
13.201***
0.000
Spatial error LM Spatial lag LM
Note: ***, **, and * denote statistical significance at the 0.01, 0.05, and 0.1 levels, respectively.
TABLE 6
Estimation results of Equation (4) Model 1
Variables
Coefficient
Model 2 t‐stat
p‐value
Coefficient
t‐stat
p‐value
LK
0.566***
6.53
0.000
0.588***
7.76
0.000
LAg
0.055*
1.63
0.103
0.069**
2.13
0.033
Lrailt
0.813***
4.18
0.000
Lroadt
1.029***
3.21
0.001
Lrails
0.112***
6.52
0.000
Lroads
0.345***
4.53
0.000
W*Lk W*LAg
0.275
1.24
0.216
–0.171
–0.93
0.351
W*Lrailt
7.538***
3.63
0.000
W*Lroadt
1.911
1.19
0.233
0.785*** –0.200*
3.01
0.003
–1.64
0.100
0.000
W*Lrails
1.335***
4.78
W*Lroads
0.810*
1.72
0.086
32.82
0.000
W*Lrgdp
0.855***
R^2
0.433
Log likelihood Hausman's specification test Wald spatial lag test Wald spatial error test
59.95
0.000
0.810*** 0.400
6.099
6.099
1607.23***
0.000
470.55***
0.000
244.79***
0.000
108.82
0.000
49.78***
0.000
92.10
0.000
Notes: The logarithm of real GDP per capita is a dependent variable. ***, **, and * denote statistical significance at the 0.01, 0.05, and 0.1 levels, respectively. The numbers of observations equal the numbers of years in each period multiplied by 28 provinces. Lrailt, Lroadt, Lrails, Lroads refer to the logarithms of total rail length per capita, the logarithms of total road length per capita, the logarithms of main rail length per capita, the logarithms of main road length per capita, respectively.
effects of total road and railway networks on regional economic growth, whereas in model 2, we examined the effects of main road and railway networks on regional economic growth. A Wald test was conducted to evaluate the suitability of the SAR model and the SDM. As Table 4 indicates, the p‐value generated by the Wald spatial lag test is less than 0.05. Therefore, the SAR model is rejected and the SDM is accepted. As shown in Table 5, each explanatory variable has a direct, an indirect, and a total effect on the dependent variable (i.e., economic growth). Table 7 presents the estimation results on direct, indirect, and total effects. The explanatory variable's direct effect on regional growth reflects how a change in the variable in province i might affect the average regional growth in that province. The explanatory variable's indirect or spillover effect on regional growth indicates how a change in the variable in province i might affect the average regional growth in other provinces. In other words, the spillover effect shows the consequence of the variable's spatial spillover on the regional growth of the other provinces. The explanatory variable's total effect on regional growth can also show how a change in the
60
DEHGHAN SHABANI AND SAFAIE
TABLE 7
Direct, indirect, and total effects of the explanatory variables Model 1 Coefficient
t‐stat
p‐value
Coefficient
t‐stat
0.567*** 0.210 0.777**
6.56 0.72 2.24
0.000 0.472 0.025
0.586*** 0.655* 1.241***
7.57 1.75 3.14
0.000 0.081 0.002
1.83 −0.93 −0.48
0.067 0.354 0.354
2.49 −1.79 −0.74
0.013 0.074 0.461
0.759*** 6.727*** 7.487***
3.95 3.01 3.10
0.000 0.003 0.002
1.018*** 1.552 2.570
3.57 1.04 1.50
0.000 0.298 0.134 0.102***
6.64
0.000
Indirect effect (spillover)
1.181***
3.62
0.000
Total
1.283***
3.86
0.000
Direct effect
0.348***
4.89
0.000
Indirect effect (spillover)
0.706*
1.68
0.092
Total
1.054**
2.17
0.030
Variable
Effects
Lk
Direct effect Indirect effect Total
LAG
Direct effect Indirect effect (spillover) Total
Lrailt
Direct effect Indirect effect (spillover) Total
Lroadt
Direct effect Indirect effect (spillover) Total
Lrails
Direct effect
Lroads
Model 2
0.057* −0.148 −0.091
0.075** −0.164* −0.089
p‐value
Notes: The logarithm of real GDP per capita is a dependent variable. ***, **, and * denote statistical significance at the 0.01, 0.05, and 0.1 levels, respectively. Lrailt, Lroadt, Lrails, Lroads refer to the logarithms of total rail length per capita, the logarithms of total road length per capita, the logarithms of main rail length per capita, the logarithms of main road length per capita, respectively.
variable in province i might affect the average economic growth rate in all the provinces, including province i (Atella, Belotti, Depalo, & Mortari, 2014). In models 1 and 2, the direct and total effects of stock capital per capita on economic growth is positive and significant (p < 0.05), indicating that an increase in the capital of each province elevates the economic growth in that province and in the entire country. This finding aligns with our theoretical expectations. The results of model 1 showed that the direct effect of total road and railway per capita is positive and significant (p < 0.05), which means that an increase in transportation infrastructure translates to an increase in economic growth. On the basis of the proposed theoretical framework, these results can be explained by two aspects that require consideration with respect to the development of transportation infrastructure. On the one hand, this development leads to the growth of industries that provide products and services needed in transportation infrastructure. On the other hand, the transportation costs of entities decrease, thereby driving increased production. Moreover, access to larger markets is provided, thus attracting manufacturers and economic activities to a given region. Higher industrial concentration is established, which leads to economies of agglomeration. Such agglomeration, in turn, reduces product prices and increases demand. The spillover effect of total railway length on economic growth is positive and significant (p < 0.01), indicating that the development of rail transportation infrastructure in a province results in an increase in the economic growth of neighbouring provinces. By contrast, the spillover effect of total road length is non‐significant, indicating that total road length does not have any spillover on other provinces. The total effect of total railway length is significant and positive, whereas that of total road length is non‐significant (Table 4). The direct effects of main road and railway networks are significant and positive in model 2 (Table 6). The indirect effects (spatial spillover) of railways and main roads are significant (p < 0.05 and p < 0.1, respectively). The proposed theoretical framework indicates that with an increase in neighbouring provinces’ transportation infrastructure, diffusion and agglomeration effects occur. Diffusion means that the development of transportation infrastructure in
DEHGHAN SHABANI AND SAFAIE
61
each province drives its growth (direct effect) and that growth in each province leads to growth in other provinces (indirect effect). Agglomeration means that the development of the transportation infrastructure of a province causes the work force and manufacturers to migrate to that province, which in turn, diminishes growth in other provinces. This is negative spillover on economic growth. Positive spillover happens because the diffusion effects dominate the agglomeration effects. The total effect (summation of direct and indirect effects) of main road and railway length on economic growth is positive and significant (p < 0.05). This result indicates that the development of transportation infrastructure in province i prompts, on average, an increase in economic growth in other provinces. As Table 5 shows, the direct effect of industrial concentration on economic growth is positive. This finding is attributed to the availability of abundant labour reserves, which promotes the efficiency of the local labour market by creating a match between employers and employees. In addition, the availability of useful knowledge spillovers promotes intra‐ and inter‐industry productivity. Industrial concentration can therefore be accompanied by a higher level of productivity, increased real wages, better standards of living, and improved economic growth (Baldwin & Martin, 2004).
4
|
C O N CL U S I O N
Much of the evidence on transport infrastructure spillovers has been reported for states and counties in developed countries, such as the United States and Spain, where a severe lack of infrastructural endowments may not be a problem. Deviating from the traditional approach in the literature, we derived evidence on the spillover effects of transport stock in the Iranian context, where some provinces suffered from inferior economic development and limited transport infrastructure in most of the periods under analysis. The total railway spillover effect is positive and significant. The result is similar Chen and Haynes (2015) and Li et al. (2017). The spillover effects of main road and railway transportation infrastructure on economic growth are positive and significant, but the spillover from railways is greater than that from roads. The result of the main road transport (that include highway) impact is similar to Cantos et al. (2005), Berechman et al. (2006), Liu et al. (2007), Ozbey (2007), Chen and Haynes (2015) and Li et al. (2017). This research presents important policy implications for investment in transportation infrastructure in Iran and other developing economies. The results showed that investment in rail and road infrastructure—specifically, construction and maintenance—is imperative not only to economic growth within a region but also to economic growth in other regions. The findings can also assist the government in prioritizing rail and road structure investment given the limited availability of investment resources under the current budget deficit of Iran. On the basis of the empirical results, we recommend that investment policy be directed primarily to the development of cross‐regional transport networks instead of intra‐regional networks. Our findings likewise suggested that investment decisions on road infrastructure should be geared toward the construction of main roads as these thoroughfares contribute to economic growth both in and outside of a region. The results highlight the essentiality of investment in rail (main and total) infrastructure in Iran. Iran's railway infrastructure network is shorter than its road infrastructure, yet the results indicated that the former has had a major impact on growth in the region and other regions. Under the current budget deficit of Iran, we advise policy‐makers to allocate investment in the construction and maintenance of the country's rail infrastructure. A limitation of this study is that the spatial weight matrix used in the SDM may be more intuitive than the spatial contiguity matrices used in our analysis. A spatial weight matrix based on intuitive knowledge about the spatial linkage of transportation networks across states would have expanded our examination. However, a matrix that can accommodate flow to and from major states of production is extremely complicated to construct. Other researchers can explore a spatial empirical model that is grounded in the aforementioned intuitive knowledge in relation to production. The spatial spillover analysis of transportation infrastructure in this study can also be applied to other issues of importance to a sector. For instance, scholars can probe into the spatial spillover of transportation infrastructure on the outputs of agricultural, industrial, and service sectors.
62
DEHGHAN SHABANI AND SAFAIE
RE FE R ENC ES Ades, A., & Chua, H. B. (1997). Thy neighbor's curse: Regional instability and economic growth. Journal of Economic Growth, 2, 279–304. Afsharpour, M., Mehrabi, B., & Pahlavani, M. (2014). Effects of transportation infrastructure development on value added of Agricultural Sector. Journal of Agricultural Economics Research, 6, 115–134. Agbelie, B. R. (2014). An empirical analysis of three econometric frameworks for evaluating economic impacts of transportation infrastructure expenditures across countries. Transport Policy, 35, 304–310. Ahmed, V., Abbas, A., & Ahmed, S. (2013). Public infrastructure and economic growth in Pakistan: A dynamic CGE‐ microsimulation analysis. In J. Cockburn, Y. Dissou, J. Y. Duclos, & L. Tiberti (Eds.), Infrastructure and economic growth in Asia (pp. 117–143). Cham: Springer. Akbarian, R., & Ghaedi, A. (2011). Investment in economic infrastructures and economic growth: The case of Iran. Quarterly Journal of Economic Growth and Development Research, 6, 11–48. Álvarez, I. C., Barbero, J., & Zofío, J. L. (2014). A spatial autoregressive panel model to analyze road network spillovers on production. Transportation Research Part A: Policy and Practice, 93, 83–92. Álvarez, I., Condeço‐Melhorado, A., Gutiérrez, J., & Zofío, J. L. (2016). Integrating network analysis and interregional trade to study the spatial impact of transport infrastructure using production functions. Regional Studies, 50, 996–1015. Álvarez‐Ayuso, I. C., & Delgado‐Rodriguez, M. J. (2012). High‐capacity road networks and spatial spillovers in Spanish regions. Journal of Transport Economics and Policy (JTEP), 46(2), 281–292. Anselin, L. (1996). The Moran Scatterplot as an ESDA tool to assess local instability in spatial association. In M. Fisher, H. J. Scholten, & D. Unwin (Eds.), Spatial analytical perspectives on GIS (pp. 111–125). London: Taylor and Francis. Aschauer, D. A. (1989). Is public expenditure productive? Journal of Monetary Economics, 23, 177–200. Atella, V., Belotti, F., Depalo, D., & Mortari, A. P. (2014). Measuring spatial effects in the presence of institutional constraints: The case of Italian local health authority expenditure. Regional Science and Urban Economics, 49, 232–241. Baldwin, R. E., & Martin, P. (2004). Agglomeration and regional growth. In Handbook of regional and urban economics (Vol. 4, pp. 2671–2711). Elsevier. Banerjee, A., Duflo, E., & Qian, N. (2012). On the road: Access to transportation infrastructure and economic growth in China. Working Paper 17897. National Bureau of Economic Research. Berechman, J., Ozmen, D., & Ozbay, K. (2006). Empirical analysis of transportationinvestment and economic development at state, county and municipality levels. Transportation, 33, 537–551. Berlemann, M., & Wesselhöft, J. E. (2014). Estimating aggregate capital stocks using the perpetual inventory method. Review of Economics, 65, 1–34. Boarnet, G. M. (1998). Spillovers and the locational effects of public infrastructure. Journal of Regional Science, 38, 381–400. Boopen, S. (2006). Transport infrastructure and economic growth: Evidence from Africa using dynamic panel estimates. The Empirical Economics Letters, 5, 38–52. Bosede, A., Abalaba, B., & Afolabi, D. (2013). Transport infrastructure improvement and economic growth in Nigeria. International Journal of Humanities and Social Science Invention, 2(8), 23–31. Cantos, P., Gumbau‐Albert, M., & Maudos, J. (2005). Transport infrastructure, spillover effects and regional growth: evidence of the Spanish case. Transport Reviews, 25, 25–50. Carruthers, R. (2013). What prospects for transport infrastructure and impacts on growth in southern and eastern Mediterranean countries? Medpro Report No. 3, Medpro. Chen, Z., & Haynes, K. E. (2015). Regional impact of public transportation infrastructure: A spatial panel assessment of the US Northeast megaregion. Economic Development Quarterly, 29, 275–291. Chua, H. B. (1993). Regional spillovers and economic growth. Economic growth center, Discussion Paper, N0.700, Yale University, September. Conley, T. G., & Ligon, E. (2002). Economic distance and cross–country spillovers. Journal of Economic Growth, 7, 157–187. Daii karimzadeh, S., Emadzadeh, M., & Kamkar Delakh, H. (2009). Public Investment in the Transportation Sector and Economic Growth in Iran (1970–2008). Quarterly Journal of Economical Modeling, 3, 63–82. Duffy‐Deno, K. T., & Eberts, R. W. (1991). Public infrastructure and regional economic development: a simultaneous equations approach. Journal of Urban Economics, 30, 329–343. Garcia‐Mila, T., & McGuire, T. J. (1992). The contribution of publicly provided inputs to states’ economies. Regional Science and Urban Economics, 22, 229–241.
DEHGHAN SHABANI AND SAFAIE
63
Haque, M. E., & Kim, D. H. (2003). Public investment in transportation and comunication and growth: A dynamic panel approach. The School of Economics Discussion Paper 31, University of Manchester. Hausman, J. A. (1978). Specification tests in econometrics. Econometrica, 46, 1251–1271. Hong, J., Chu, Z., & Wang, Q. (2011). Transport infrastructure and regional economic growth: evidence from China. Transportation, 38, 737–752. Hu, A., & Liu, S. (2010). Transportation, economic growth and spillover effects: The conclusion based on the spatial econometric model. Frontiers of Economics in China, 5, 169–186. Jiwattanakulpaisarn, P., Noland, R. B., & Graham, D. J. (2011). Highway infrastructure and private output: evidence from static and dynamic production function models. Transportmetrica, 7, 347–367. Kodongo, O., & Ojah, K. (2016). Does infrastructure really explain economic growth in Sub‐Saharan Africa? Review of Development Finance, 6, 105–125. Kustepeli, Y., Gulcan, Y., & Akgungor, S. (2008). Transportation expenditures, growth and international trade. Discussion Paper, N0. 08/03, Dokuz Eylül University, April. LeSage, J. P., & Pace, R. K. (2009). Introduction to spatial econometrics. Statistics, textbooks and monographs, CRC Press. Li, J., Wen, J., & Jiang, B. (2017). Spatial spillover effects of transport infrastructure in Chinese new Silk Road economic belt. International Journal of e‐Navigation and Maritime Economy, 6, 1–8. Liu, N., Chen, Y., & Zhou, Q. (2007). Spatial spillover effects of transport infrastructure on regional economic growth. Journal of Southeast University (English Edition), 23, 33–39. Liu, Y. (2010). Transport infrastructure investment, regional economic growth and the spatial spillover effects‐based on highway and marine's panel data analysis. China Industrial Economics, 12, 37–46. Moran, P. A. (1950). Notes on continuous stochastic phenomena. Biometrika, 37(1/2), 17–23. Moreno, R., & Trehan, B. (1997). Location and the growth of nations. Journal of Economic Growth, 2, 399–418. Nakamura, R., & Paul, C. J. M. (2009). Measuring agglomeration. In: R. Capello, & P. Nijkamp (Eds.), Handbook of regional growth and development theories (Chapter 16, pp. 305–328). Northampton: Edward Elgar Publishing. Ozbay, K., Ozmen, D., & Berechman, J. (2007). Contribution of transportation investments to county output. Transport Policy, 14, 317–329. Pereira, A., & Roca‐Sagales, O. (2003). Spillover effects of public capital formation: Evidence from the Spanish regions. Journal of Urban Economics, 53, 238–256. Pereira, A. M., & Andraz, J. M. (2004). Public highway spending and state spillovers in the USA. Applied Economics Letters, 31, 785–788. Pradhan, R. P., Norman, N. R., Badir, Y., & Samadhan, B. (2013). Transport infrastructure, foreign direct investment and economic growth interactions in India: The ARDL bounds testing approach. Procedia ‐ Social and Behavioral Sciences, 104, 914–921. Quah, D. (1993). Empirical cross‐section dynamics in economic growth. European Economic Review, 37, 426–434. Sun, Z. (2013). Explaining regional disparities of China's economic growth: Geography, policy and infrastructure. Berkeley (Unpublished Thesis): Department of Economics University of California. Tong, T., Yu, T. H. E., Cho, S. H., Jensen, K., & Ugarte, D. D. L. T. (2013). Evaluating the spatial spillover effects of transportation infrastructure on agricultural output across the United States. Journal of Transport Geography, 30, 47–55. Vega, S. H., & Elhorst, J. P. (2013, August). On spatial econometric models, spillover effects, and W. In In 53rd ERSA Congress. Palermo: Italy. Wang, X., Deng, D., & Wu, X. (2014). Stimulate economic growth by improving transport infrastructure–a lesson from China. Transport Problems, 9, 63–72. Xueliang, Z. (2008). Transport infrastructure, spatial spillover and economic growth: Evidence from China. Frontiers of Economics in China, 3, 585–597. Yu, N., De Jong, M., Storm, S., & Mi, J. (2013). Spatial spillover effects of transport infrastructure: Evidence from Chinese regions. Journal of Transport Geography, 28, 56–66.
How to cite this article: Dehghan Shabani Z, Safaie S. Do transport infrastructure spillovers matter for economic growth? Evidence on road and railway transport infrastructure in Iranian provinces. Reg Sci Policy Pract. 2018;10:49–63. https://doi.org/10.1111/rsp3.12114
DOI: 10.1111/rsp3.12114
Resumen. Este estudio analizó los efectos de spillover espacial de la infraestructura de transporte por carretera y ferrocarril en el crecimiento económico de las provincias de Irán. Más específicamente, el estudio examinó los efectos directos, indirectos (spillover espacial) y totales de dicha infraestructura en el crecimiento del producto de las provincias. Con este fin, se utilizó la máxima verosimilitud para estimar un modelo espacial de Durbin para las provincias durante el período 2001 a 2011. Los resultados mostraron que los principales marcos de transporte por carretera y ferroviario ejercen un efecto directo positivo y significativo sobre el crecimiento económico de las provincias. La infraestructura principal de transporte por carretera y ferroviario de cada provincia tiene un efecto indirecto significativo y positivo sobre el crecimiento económico de las otras provincias, lo que indica que las mejoras en la infraestructura individual tienen efectos de spillover espacial sobre el crecimiento económico de las otras provincias. La conclusión extraída de estos hallazgos es que los efectos de difusión superan a los efectos de aglomeración. Finalmente, la longitud total del ferrocarril tiene un efecto positivo de spillover espacial sobre el crecimiento económico regional.
抄録: 本稿では、イランの州(province)の経済成長に対する、道路および鉄道輸送の交通 インフラの空間的波及効果を分析する。具体的には、州の生産高の増加に対する、道路お よび鉄道輸送インフラの、直接的、間接的(すなわち空間的波及効果)、総合的な効果を検 討する。その方法として、2001年から2011年までの各州の空間ダービンモデルを最尤法に より推定する。結果は、主要な道路および鉄道輸送のフレームワークが、州の経済成長に 対して大きくかつ直接的なプラスの効果を与えていることを示す。各州の主要な道路およ び鉄道輸送インフラは、他州の経済成長に対して、大きなプラスの効果を間接的に与えて おり、各インフラの改善が他州の経済成長に波及効果を及ぼすことを示している。以上の 知見から、拡散効果は集積効果を上回るという結論が得られる。さらに、鉄道の全長は地 域の経済成長に対してプラスの空間的波及効果を与えている。
© 2018 The Author(s). Regional Science Policy and Practice © 2018 RSAI
