Port infrastructure investment and regional economic growthin China: Panel evidence in port regions and provinces

Lili Song a,b,n, Marina van Geenhuizen c

a School of Management, Harbin Normal University, 1 Shida Road, Harbin, Chinab School of Economics and Management, Harbin Institute of Technology, 92 Xidazhi Street, Harbin, Chinac Faculty of Technology, Policy and Management, Delft University of Technology, Delft, The Netherlands

a r t i c l e i n f o

Keywords:Port investmentRegional economic growthEconomic structurePanel dataChina

a b s t r a c t

China's seaports belong to the largest in the world. The question is to what extent port infrastructureinvestment in China also contributes to growth of the regional economies involved, through mainlydirect and indirect relations. We estimate the output elasticity of port infrastructure through productionfunction, applying panel data analysis for the period of 1999–2010, and calculate the model at the levelof four port regions as well as the port province level. The results indicate clear positive effects of portinfrastructure investment in all regions, however, the strength varies considerably among the fourregions, with the Yangtze River Delta region (Shanghai) at the strongest level, followed by the Bohai Rimregion (Tianjin), the Southeast region (Guangzhou) and the Central region, where the influence is theweakest. The analysis indicates that differences are related to the character of the port (land or sea),stage of economic development of the region, international network connectivity, and the spillovereffects from adjacent regions. Overall, the weakest relation tends to be with landside transportinfrastructure density. The paper closes with some policy implications.

& 2014 Elsevier Ltd. All rights reserved.

1. Introduction

1.1. China's seaport in the world

Ports are traditionally seen as economic catalysts for theregions they serve, where the agglomeration of services andmanufacturing activities generate economic benefits and socio-economic wealth (Warf and Cox, 1989; Pettit and Beresford, 2009;Zhang et al., 2009; Danielis and Gregori, 2013). Chinese ports playa key role in the world port system of 2011, as indicated in Table 1.The 10 Chinese ports rank high with a share in total cargo volumeand container traffic of the top-20 world ports of 52.9% and 53.0%,respectively. In both rankings, China is present with three portsamong the five largest ones in the world, with Shanghai in firstplace. The rankings also show differentiation between cargo andcontainer traffic. For example, Tianjin port enjoys a higher rank(rank 3) in cargo volume and a relatively lower rank in containertraffic (rank 11), reflecting the port's specialization in raw materi-als like coal and mineral. However, the rankings are only adescription of relative size of transport flows, while this is just

one part of port activity in a situation of manifold and systematicrelationships between ports and ports' regional economies.Accordingly, the relation with local industries, economic charac-teristics of the port regions, and transport network connectivity ofthe region, etc. could also have an impact on port activity, as wellas on regional economic growth (Berechman et al., 2006; Banister,2012; Ducruet et al., 2013).

The important position of Chinese ports indicates that Chinahas made substantial capital investment in its port facilities inrecent years. What is actually less known is to what extent theport investments contribute to growth of the regional economythrough various multiplier effects, including the direct, indirectand induced effects, and whether there are large regional dispa-rities in these effects.

1.2. Port infrastructure and the regional economy: a literature review

Over the last decades, a large number of studies has focused onthe impact of transport infrastructure and accessibility in generalon regional economic growth, most of which were concerned withtransport investments, aiming to assess whether positive economicimpacts are a sufficient rationale for traffic infrastructure invest-ments (Ozbay et al., 2003; Canning and Bennathan, 2007). However,in the recent literature, impacts on the regional economy areincreasingly seen as influenced by the level of traffic infrastructure

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/tranpol

Transport Policy

http://dx.doi.org/10.1016/j.tranpol.2014.08.0030967-070X/& 2014 Elsevier Ltd. All rights reserved.

n Corresponding author at: School of Management, Harbin Normal University,1 Shida Road, Harbin, China. Tel.: þ86 13936307866.

E-mail address: rubysonglili1981@gmail.com (L. Song).

Transport Policy 36 (2014) 173–183

accumulation in the region at the start of the study period, with anemphasis on a non-linear relationship between transport infra-structure provision and economic growth (Banister, 2012). The ideahas been forwarded that below a certain level of infrastructureendowment and above a certain level, the growth effect ofexpanding transport infrastructure tends to be relatively small(Deng et al., 2013a, 2013b). Threshold values have also beenaddressed by Hong et al. (2011), but only as a lower threshold. Inthe remaining section, we discuss the statistical models used ininvestigations of the relationship between port investment and theregional economy, and studies using a broader network and valuechain view on port development, including spill-over effects.

Mainly three empirical methods are used to investigate therelationship between transport investment and regional economywhich are Cobb–Douglas production function framework, timeseries models and structural equation modeling. Most previousresearch used a Cobb–Douglas production function framework inestimating the impacts of transport investment (Blum, 1982; Biehl,1986; Nijkamp, 1986; Del Bo and Florio, 2012). The result of thesestudies is a positive relationship between transport investmentand economic growth which is now commonly accepted(Berechman et al., 2006). Yoo (2006) and Jiang (2010) investigatedthe influence of seaport infrastructure investment on economicgrowth in Korea and China respectively by applying time seriesdata. A positive impact of port investment on economic growthcould be found both in Korea and China. In addition, Jiang'sempirical findings also show regional disparities: the port invest-ments in Pearl River Delta have the highest short-term outputelasticity, whereas the short-term output elasticity in YangtzeRiver Delta is the lowest, indicating a larger amount of newconstruction and related activity in the first region compared tothe last one. Another study on China, by Deng et al. (2013a, 2013b),used structural equation modeling to unravel the different influ-ences on regional economic growth related to port investments, bydistinguishing between port supply, port demand, and valueadded-activity in ports. They observed no direct relation betweenport supply and growth in the regional economy, but port supplywas connected to this growth through the relations with portdemand and port value added activity.

Many recent studies analyze port activities and relations withthe regional or local (port city) economy fromwider network pers-pectives, including territorial embedding of port areas in commodity

flows and value chains. Ducruet et al. (2013), in a comparativestudy of almost 200 port regions in advanced economic areas,argue that port-region linkages develop in subtle interdependencies,while pointing to noticeable differences between traffic volumes,types and local economic structures, as apparent from commoditytraffic data and regional economy data. Accordingly, economicallyand demographically larger and richer regions that are specializedin (private sector) producer services, concentrate larger and morediversified traffic volumes as well as higher valued goods. Bycontrast, agricultural and industrial regions are more specializedin bulk traffic (Ducruet et al., 2010, 2013). The study of Jacobset al. (2011) on maritime advanced producer services, fits into thewider network perspective on influences on port activity andtraffic flow.

Studies paying attention to spillover effects to nearby regionsalso fit into the broader perspective. We mention Bottasso et al.(2014) who observed in 13 West European countries that a 10%increase in port throughput gave a growth in regional GDP of theport regions by 0.01–0.03%, while the effect in nearby regionsturned out to be larger, namely 0.05–0.18%. Merk and Hesse (2012)found for the port of Hamburg (Germany) not only considerableregional spillover effects, but also large distances involved. Only13% of the multiplier effects have an impact on Hamburg and itsneighboring regions, while almost a third spills over to two largesouthern regions at a distance from the port and more than half tothe rest of Germany.

The previous studies illustrate a myriad of interrelationshipsbetween port infrastructure investment, connectivity of the portwith land infrastructure, size and type of transport flow, valuechains and production networks embedded in the port andstretching (spilling over) in adjacent and more distant regions,and local geographical and historical specificities, like local eco-nomic specialization. This situation would mean that each estima-tion of impacts of port infrastructure investment on the regionaleconomy shows a relatively small impact and shows some differ-entiation between regions.

1.3. Research aim and questions

Most previous port investment studies have a limited scopethat is often neglecting (part of) the above indicated influences,like connected land traffic infrastructures, profile of the regional

Table 1Top 20 world ports in 2011.Source: Institute of Shipping Economics & Logistics, Containerization International Yearbook 2012.

Rank Port, Country Cargo volume (thousands of tons) Rank Port, Country Container traffic (TEUS)

1 Shanghai, China 590,439 1 Shanghai, China 31,739,0002 Singapore, Singapore 531,176 2 Singapore, Singapore 29,937,7003 Tianjin, China 459,941 3 Hong Kong, China 24,384,0004 Rotterdam, Netherlands 434,551 4 Shenzhen, China 22,570,8005 Guangzhou, China 431,000 5 Bushan, South Korea 16,163,8426 Qingdao, China 372,000 6 Ningbo, China 14,719,2007 Ningbo, China 348,911 7 Guangzhou, China 14,260,4008 Qinhuangdao, China 284,600 8 Qingdao, China 13,020,1009 Bushan, South Korea 281,513 9 Dubai Ports, United Arab Emirates 12,617,59510 Hong Kong, China 277,444 10 Rotterdam, Netherlands 11,876,92011 Port Hedland, Australia 246,672 11 Tianjin, China 11,587,60012 South Louisiana (LA), U.S.A 223,633 12 Kaohsiung, Taiwan 9,636,28913 Houston (TX), U.S.A 215,731 13 Port Kelang, Malasyia 9,435,40814 Dalian, China 211,065 14 Hamburg, Germany 9,014,16515 Shenzhen, China 205,475 15 Antwerp, Belgium 8,664,24316 Port Kelang, Malaysia 193,726 16 Los Angeles, U.S.A 7,940,51117 Antwerp, Belgium 187,151 17 Tanjung Pelepas, Malaysia 7,302,46118 Nagoya, Japan 186,305 18 Xiamen, China 6,454,20019 Dampier, Australia 171,844 19 Dalian, China 6,400,30020 Ulsan, South Korea 163,181 20 Long Beach, U.S.A 6,061,091

L. Song, M. van Geenhuizen / Transport Policy 36 (2014) 173–183174

economy and types of connected value chains. In addition, smallattention has been paid to the regional spillover effects in model-ing the impacts of port investments.

Against this background, this paper takes a broader network-related look at the effects of port infrastructure investment on theregional economy in China. Accordingly, the following questionsare addressed. What is the output elasticity of port infrastructureinvestment in the four regions and in port provinces, and whichdifferences do exist between these regions and between theseprovinces? What are the reasons behind these differences? Forexample, what is the role, aside from port capital investment, ofmulti-modal connectivity of the ports, presence of neighboringports, and the economic structure in terms of manufacturing sizein the region? To what extent are inland port areas different fromseaport areas?

The structure of the paper is as follows. The next section gives abrief descriptive analysis of the regional distribution of portfacilities, as well as an overview of port investment in China(Section 2). This is followed by analysis of the factors included inour model in Section 3. Section 4 introduces the methodology anddatabase to quantify economic effects of port infrastructure on thefour regions and 13 port provinces. The results of the modelestimations are presented in Section 5, on the regional andprovincial level, including a discussion of the results. The paperends with conclusions and policy implications.

2. Port infrastructure in China: an overview

2.1. Four port regions

China developed route schedule oriented shipping with Japanand Korea and dug the world's first canal while owning a highlyadvanced shipbuilding technology already during the Han Dynasty(256–220 BC) (Wang and Ducruet, 2013). More recently, China'sport construction is based on a development policy to graduallycreate port clusters with a hub port as the core (Li and Yuan, 2010).The sample ports in the current study are the scale ports in China,including seaports and inland ports. According to the ChineseMinistry of Transport, ‘scale ports’ are defined as ports facingthroughputs of over 15 and 10 million tons in 2002 for a sea portand an inland port respectively. Our definition of the port region(see Fig. 1) complies with the Coastal Port layout in China(MOCOPRC, 2006). The only difference is that in the current study,inland port provinces are being included as an independent portregion (named Central region or Center) to get a better under-standing of the disparity between seaports and inland ports.

We now briefly characterize the industrial specialization of thefour port regions. Bohai Rim is an important base of energy and rawmaterials production in China (for example, heavy chemical andsteel industry) and an area of abundant mineral resources. Thesetwo assets enabled the ports in Bohai Rim (Tianjin and Dalian, etc.)

Fig. 1. The four port regions and 13 port provinces in this study.

L. Song, M. van Geenhuizen / Transport Policy 36 (2014) 173–183 175

to become the most important coastal transport nodes to the southernpart of China. Next, the Yangtze River Delta region includes the port ofShanghai as one of the world's largest seaports and among China'sbiggest manufacturing and commercial services centers. Yangshanport, as an extension project of Shanghai port, is a value-added andintegrated industrial, logistics, and shipping complex rather than a soletransshipment node, with the Yangshan Bonded Port Area as amultimodal logistics center for transshipment, distribution, insurance,finance, and entrepot trade. This facility was also designed to provide alocal manufacturing base aimed at limiting truck and shipping flows toand from mainland China in addition to its first transshipmentfunction (Wang and Ducruet, 2012). In this region we find more than100 industrial parks (distributed over the urban areas of Shanghai,Hangzhou, Nanjiang, etc.) and many large enterprises (e.g. WanxiangGroup, Jinshan petro-chemical, Yangzi ethylene, Volkswagen, ShanghaiBell, EastCom) (Deng et al., 2013a, 2013b). By contrast, the Southeastregion is an important production area for imported tropical cashcrops such as rubber, cane sugar, tobacco, etc., and it is also China'searliest “opened-up” region, including open coastal economic zones(cities of Shenzhen, Zhuhai and Xiamen, etc.). In particular, Guangdongprovince benefited from the “Opening Up” policy since 1978, throughwhich many multinationals were attracted, including Canon, Epson,Samsung, Coca-Cola, Philip, etc.

And finally, there is the central region (or Center), which can beregarded as the latest developed region, as more and more ofChina's manufacturing operations are shifting inland. The resultsof the “Go West” policy “Central Rise” strategy – adopted since theearly 2000s – can now be clearly seen in inland locations, althoughspread effects may go hand-in-hand with backwash effects(Ke and Feser, 2010). Major cities like Chongqing and Wuhan arebooming as manufacturing hubs, home of some of the world'slargest manufacturers across a diverse range of products andindustries, including Ford, Intel, Hewlett-Packard, General Electric,Procter & Gamble, Siemens, and Samsung, etc.

2.2. Scale port investment

After the Chinese fiscal decentralization, in the early 1990s,many local administrative units (provincial and municipal) have

received substantial financial power (Zhang et al., 2007). Thisenabled them to have their own independency with respect to thedistribution of transport investments and investment decisions inview of their individual economic growth. However, the main linesof port investment policy are determined at the national level in asituation in which most of the scale ports are state-owned.

The size of port investment (port infrastructure investment,including container terminals, cargo terminals, road and rail in portarea, cranes, etc.) per region suggests substantial regional differencesbetween seaports, but also between inland ports, and of course partof the differences can be understood based on the different numbersof ports per region (Table 2). The investments in the Yangtze Riverregion – 11 ports – are the largest at the start of the period (23,800million RMB) and also in 2010 (150,700 million RMB). In addition, theinitial port investment in Center area, with merely inland ports andonly four of them, is the lowest in 1999 (2600 million RMB), but –due to a quick growth – equals investments in the Southeast regionin 2010 (30400 million RMB). The annual growth rate of portinvestment in Center area is higher than the Yangtze River andSoutheast areas, as witnessed by 22.8% versus 16.6% and 15.7%,respectively. The same holds for Bohai Rim, with merely sea ports,nine of them (23.7%). By contrast, Southeast, including seven seaportscompares with Bohai Rimwith regard to size of initial investments in1999 but tends to stay behind in 2010. Whether the high increase ofport investment in the Center and Bohai Rim, with annual growthexceeding 20%, have resulted and result in a stronger growth of theregional economy is still not known.

3. Factors of influence

3.1. Regional economic structure

As previously indicated, economically and demographicallylarger and richer regions specialized in the commercial servicesector tend to be involved with larger and more diversifiedtransport volumes as well as with more value-added goods. Bycontrast, agricultural and heavy manufacturing regions tend to bemore specialized in bulk transport. Moreover, the level of value

Table 2Scale port investment per port region and province in China.Source: China Port Statistical Yearbook (2000–2011).

Region Province Accumulated port investment per region (100 million RMB) Annualgrowth (%)

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Bohai Rim 57 62 72 89 133 186 250 285 343 444 583 732 23.7Tianjin 11 12 14 16 21 25 30 37 48 68 100 130 22.6Liaoning 6 8 9 11 16 24 34 45 64 83 113 146 29.7Hebei 14 13 13 14 19 26 36 45 59 76 115 143 21.5Shandong 25 28 35 47 78 110 149 158 172 217 255 313 23.3Yangtze River 238 255 286 349 483 598 716 819 946 1079 1283 1507 16.6Shanghai 97 97 104 114 128 161 185 205 232 252 275 277 9.2Jiangsu 109 121 135 177 276 337 412 483 569 663 817 1018 20.4Zhejiang 32 37 47 59 79 101 119 132 144 164 192 212 17.1Southeast 52 58 61 67 84 101 119 141 176 199 240 300 15.7Fujian 17 20 21 24 31 39 47 60 86 98 115 151 19.7Guangdong 34 37 39 43 53 62 72 81 90 102 125 149 13.0Center 26 29 34 38 51 64 80 100 131 170 228 304 22.8Chongqing 14 16 19 22 31 39 49 60 77 98 129 169 23.3Anhui 2 2 2 3 4 5 6 8 13 18 26 36 27.4Hubei 10 11 12 13 15 19 25 31 40 52 70 87 19.8Hunan 0.3 0.4 0.4 0.4 0.6 0.6 0.8 1.0 1.3 1.8 2.6 12.4 36.4

Notes: Port regions are bold. SP denotes seaports and IP inland ports.Bohai Rim: Dalian (SP), Yingkou (SP), Qinhuangdao (SP), Huanghua (SP), Tangshan (SP), Qingdao (SP), Rizhao (SP), Yantai (SP) and Tianjin (SP).Yangtze River: Lianyungang (SP), Nanjing (IP), Zhenjiang (IP), Suzhou (IP), Nantong (IP), Taizhou (IP), Wuxi (IP), Shanghai (SP), Ningbo (SP), Hangzhou (IP) and Huzhou (IP).Southeast: Fuzhou (SP), Quanzhou (SP); Xiamen (SP), Shenzhen (SP), Guangzhou (SP), Zhanjiang (SP) and Foshan (IP).Center: Chongqing (IP), Wuhu (IP), Wuhan (IP) and Yueyang (IP).

L. Song, M. van Geenhuizen / Transport Policy 36 (2014) 173–183176

adding relatedness between flows of good and local economies is astrong component of the wealth of port regions, with higher levelsof local value added activity bringing higher incomes (Ducruetet al., 2013). Unfortunately, the available statistics on regionaleconomic structure in China can only picture a broad pattern. Theeconomic structure in port provinces in 1999 and 2010 (Table 3)indicates that the share of the manufacturing sector in GDPremained stable in the port provinces in Center area from 1999to 2010 except for Chongqing, with an increase at the level of 5%;in addition, this share significantly decreased in the port provincesin Yangtze River Delta and Southeast regions, with Shanghai facingthe largest decrease, namely, at the level of 8%. For Bohai Rim, this

share obviously increased in all port provinces within the region.Moreover, the share of the tertiary industry in GDP, includingtransport and other services, increased or remained stable in mostof the port provinces. The only provinces facing a substantialincrease, are in Yangtze River Delta, namely, Shanghai, Jiangsu andZhejiang, the last one at a level of 10%. These provinces may alsoattract more diversified and value-added port-related activity, andaccordingly a larger output elasticity of port investment.

Though the Chinese economy will remain being driven by theexpansion of manufacturing sector for a long period, since China isstill in the early stage of industrialization, the economic structuretransition is emerging in some regions in China, mainly Yangtze

Table 3Changes in economic structure per port province in 1999 and 2010.Source: China Regional Economy Statistical Yearbook (2000 and 2011).

Province Primary industry in GDP (%) Secondary industry in GDP (%) Tertiary industry in GDP (%)(including transport)

Manufacturing Non-manufacturing

1999 2010 1999 2010 1999 2010 1999 2010

Bohai RimTianjin 5.0 2.0 39.0 41.0 12.0 11.0 45.0 46.0Liaoning 12.0 9.0 37.0 41.0 11.0 13.0 40.0 37.0Hebei 18.0 13.0 36.0 40.0 12.0 13.0 34.0 35.0Shandong 16.0 9.0 37.0 41.0 12.0 13.0 35.0 37.0

Yangtze RiverShanghai 2.0 1.0 41.0 33.0 10.0 9.0 51.0 57.0Jiangsu 13.0 6.0 43.0 40.0 13.0 13.0 36.0 41.0Zhejiang 11.0 5.0 46.0 39.0 13.0 13.0 34.0 44.0

SoutheastFujian 18.0 9.0 36.0 37.0 11.0 14.0 40.0 40.0Guangdong 11.0 5.0 43.0 40.0 11.0 10.0 42.0 45.0

CenterChongqing 17.0 9.0 35.0 40.0 12.0 15.0 41.0 36,0Anhui 28.0 14.0 37.0 38.0 10.0 14.0 30.0 34.0Hubei 20.0 13.0 42.0 41.0 11.0 13.0 34.0 38.0Hunan 24.0 14.0 33.0 34.0 10.0 12.0 37.0 40.0

Notes: Regions are in bold. According to National Industry Classification, Chinese industry is divided into three industries: Primary industry (including Agriculture, Forestry,Animal husbandry and Fishing); Secondary industry (including Mining, Manufacturing, Production and distribution of electricity, gas and water, and Construction); Tertiaryindustry (including all sectors except for Primary industry and Secondary industry).

Table 4Transport infrastructure density in 1999 and annual growth rate from 1999 to 2010 for port provinces and four regions.Source: China Statistics Yearbook (2000 and 2011).

Region PortProvince (a)

Railway in 1999(m/km2)

Railway growthrate (%)

Road in 1999(m/km2)

Road growthrate (%)

Inland waterwaysin 1999 (m/km2)

Inland waterwaysgrowth rate (%) (b)

Bohai Rim 22.04 1.83 357,05 8.90 7.55 �6.49Tianjin 51.02 2.43 748,30 4.43 34.01 �10.91Liaoning 25.68 1.04 304,05 7.01 5.41 �5.61Hebei 21.31 1.71 310,08 8.46 0.53 0.00Shandong 17.18 2.89 431,40 10.71 15.91 �5.93

YangtzeRiver

10.91 4.94 342,60 11.70 172,72 �0.07

Shanghai 47.31 2.43 662,36 9.14 331,18 0.39Jiangsu 8.77 6.42 269,98 15.13 232,94 0.10Zhejiang 10.81 4.19 395,87 8.74 102,16 �0.58

Southeast 10.31 3.71 485,01 5.62 48.23 0.28Fujian 10,48 4.08 404,84 5.08 29.84 �1.20Guangdong 10,01 3.44 531,66 5.90 60.06 0.74

Center 12.59 3.21 298,21 11.74 40.83 1.35Chongqing 7.28 1.71 341,01 8.46 27.91 0.00Anhui 15.78 2.33 293,34 11.40 40.16 0.00Hubei 11.83 3.69 297,87 11.57 39.25 1.08Hunan 13.22 2.35 285,14 11.71 47.68 1.09

Notes: (a) Port regions are in bold. (b) Negative developments may be due to the filling of superfluous, narrow, canals.

L. Song, M. van Geenhuizen / Transport Policy 36 (2014) 173–183 177

River, in which the share of manufacturing in GDP goes down infavor of services.

3.2. Landside traffic infrastructure and air traffic

A well-developed landside traffic infrastructure enables the emer-gence of regional economic growth derived from goods from overseas.It allows the development of multimodal chains based on an increasedaccessibility. According to the China Port Statistical Yearbook (2011),the percentage of the goods transported between ports and hinterlandby road, waterways and railway are 84%, 14% and 2% respectively,meaning that almost all goods in ports are delivered through road andwaterways. By taking rail, road and waterways into account – in termsof density – the following broad pattern becomes clear concerning theyear 1999 (Table 4): Bohai Rim has the best developed rail system,especially the province of Tianjin, while the road system is at amedium level and the waterway system at a very low level. However,railways to date turn out to be an irrelevant mode in China for freight.Yangtze River has by far the best developed system of waterways, witha medium level as far as roads are concerned, and a relatively weaklevel of railway development, this with the exception of Shanghaiwhich is endowed with relatively high densities. Southeast has thebest developed road system, especially the province of Guangdong,with inland waterways at a medium level, and railways at a low level.Given the most often use of roads, Southeast is the best endowed area.Central has relatively low scores, with waterways and railways at amedium level, and a road system that is clearly behind.

We may understand this situation regarding port-relatedeconomic growth as follows. The Center, as an inland port region,can be assumed not having reached the threshold value in theexisting traffic infrastructure in 1999 that allows a quick contribu-tion of port investment to economic growth (Banister, 2012). Therelatively strong investment in the road system here in the yearsthat followed (Table 4), may not yet be effective in the early 1990s.By contrast, both Yangtze River and Southeast, the first endowedwith a well-developed waterway system and medium-level roadsystem, and the second with a well-developed road system, can beassumed to have crossed this threshold already before 1999,potentially causing a relatively strong impact of port investmentson economic growth. Bohai Rim may be positioned somewhere in-between the last two regions and the Center.

Connectivity is not only involved with landside freight trafficwithin the port region and regions nearby and on larger distances, italso deals with other parts of the world, in particular with largecities where major economic decisions are taken. Internationalairports connect the port regions with first-tier world cities, mainlyconcerning passengers as decision-makers in various domains, suchas multinational companies' strategies, location and composition ofglobal freight flows, and global financial services (e.g. Matsumoto,2007). This is the reason why international air traffic flows areincluded in the analysis. Data constraints, however, make us tofocus on aircraft movements from/to international airports in thefour regions and respective provinces, although we realize that thisis a very broad indicator of world city linkages, because origin anddestination are not known and the purpose of flights (passengers)are also not given (Derudder and Witlox, 2005). The region facingthe highest number of air craft movement in international airportsis the Yangtze River, and the region facing the lowest level is theCenter. Within the regions, the following provinces stand out withhigh levels: Liaoning (Bohai Rim), Shanghai (Yangtze River),Chongqing (Center) and Guangdong (Southeast).

3.3. Regional spillover effects

In principle, economic growth from seaports might generatespillover effects in adjacent port provinces and sometimes also

across not adjacent areas (Notteboom and Rodrigue, 2005). Suchpatterns depend on the spatial organization of the value chainsinvolved and are connected with the presence of strong economicactivity in adjacent areas and at larger distances in the ‘hinterland’.With regard to China, we may assume still weak but increasingspillover effects over large distances, because there are not (yet)many strongly developed inland provinces. However, we doassume that multiplier effects spill over to adjacent (port) pro-vinces. To include this in our analysis would be a study in itself,therefore we count for each province the number of scale ports inadjacent provinces as a proxy, and may expect positive impacts ifthe ports are specialized to a certain extent. The number of scaleports in adjacent provinces is largest for Zhejiang province andsmallest for Chongqing.

4. Methodology

4.1. Model specification

The starting point of the analysis is a production function andpanel data (1999–2010). The baseline empirical model is con-structed derived from a production function as:

Y ¼ f ðK ;MAN; TID; S; ICÞ ð1Þwhere Y denotes output, K represents port infrastructure capitalstock, MAN is the size of manufacturing sector, TID representsaggregate land traffic density, S stands for spillover effect fromadjacent provinces, and IC represents the international connectivity.

We estimate model (1) at the whole port region level, fourregions level, as well as the port provincial level. In addition, forthe whole port region, dummy variables are applied to explore theexistence of regional disparities, as shown in Eq. (2). For the fourregions and 13 port provinces, Eq. (3) will be estimated. As usual,in the log-linearized reduced version of production function(Mera, 1973), the estimated parameters can be thought of as Yelasticities to each regressor:

ln Yit ¼ β0þβ1ln Kitþβ2ln MANitþβ3ln TIDitþβ4ln Sitþβ5ln ICit

þβ6D1itþβ7D2itþβ8D3itþεit ð2Þ

ln Yit ¼ β0þβ1ln Kitþβ2ln MANitþβ3ln TIDitþβ4ln Sitþβ5ln ICitþεitð3Þ

where Y denotes real gross domestic product; i and t are theindices port province and year respectively; K is actual portinfrastructure investment stock; MAN is the share of manufactur-ing output in the gross domestic product; TID is the aggregateddensity of road, railway and inland waterways per port province, Sstands for spillover effects from the ports in adjacent portprovinces measured as a proxy using number of scale port inadjacent provinces. IC is the number of total aircraft movements ofinternational airports in the port provinces, which is used as aproxy for international connectivity of the port provinces. Takingthe central region as a reference, D1, D2, D3 in Eq. (2) are dummyvariables to indicate the other three individual regions: Bohai Rim,Yangtze River and Southeast.

To estimate the model, unit root tests and co-integration testsneed to be performed to ensure the reliability of the regressionresults. LLC (Levin et al., 2002) test is employed in this study forunit root test. If the unit root test indicates the series are non-stationary, which will result in a spurious regression, then we haveto test whether the series are integrated of the same order d or forshort I (d) process, if so, Kao (1999) co-integration test needs to beemployed to examine the long-term equilibrium relationships ofthe non-stationary panels, namely to investigate whether the earlychanges in port infrastructure, economic structure, land traffic

L. Song, M. van Geenhuizen / Transport Policy 36 (2014) 173–183178

density, spillovers from ports in adjacent provinces and interna-tional connectivity can effectively explain the changes in economicgrowth in the entire port region and the port regions andprovinces individually. Ultimately, we estimate the coefficients ofthe models by the ordinary least squares (OLS) regression proce-dure using the panel data model as implemented in Eviews7.

4.2. Data collection and preparation

The sample of scale ports and four port regions is derived fromthe National Coastal Port Layout (MOCOPRC, 2006). AlthoughHong Kong Port is an important port in Southeast region, dataon this port are not given in the China Port Statistical Yearbook,therefore, in this study, Hong Kong port is excluded from South-east in the analysis of ports.

The data in this research are collected from various official Chinesesources, including the China Statistical Yearbook (2000–2011), ChinaRegional Economy Statistical Yearbook (2000–2011), China Port Sta-tistical Yearbook (2000–2011) and China Airport Production StatisticsBulletin (1999–2010). Data are used from panels of 10 provinces and3 municipalities where the 31 scale ports are located, and from panelsof four port regions at the aggregate level, for the period 1999–2010.We calculate the port infrastructure capital stock based on investmentdata according to the perpetual inventory method (Goldsmith, 1951).The values of GDP, port investment and gross industrial output are theabsolutes in 1999–2010 and will be recalculated based on the price in1999 in such a way that the factors influencing the price in this periodare removed. Table 5 gives the descriptive statistics of the variablesapplied in the model estimations. Note that the level of detail of somedata is limited, meaning that the broad scope adopted in the paper, issometimes narrowed down due to lack of data.

Before calculating Eqs. (2) and (3), unit root test and co-integrationtest are applied to test the data used in the regression to ensure theaccuracy of regression results. The results of unit root test (Table 6)show that the variables in the model are non-stationary for levels atexcept for traffic infrastructure density, which is stationary at 5% levelof significance. However, non-stationary can be rejected for first-

differences of all variables at 10% level of significance meaning that theseries are integrated of the same order 1. The results of co-integrationresidual test (Table 6) clearly imply that during the research period,the five independent variables can effectively explain the GDP growth.Therefore, we can conduct the regression analysis to estimate thecontributions of port infrastructure, the size of manufacturing sector,traffic infrastructure density, spillover effect and international con-nectivity to the economic growth in the whole port area, four regionsand 13 port provinces.

As a final step in the preparation, we check for multi-collinearity among the five independent variables (port invest-ment, economic structure, transport infrastructure density, portspillover effect and international connectivity). All the correlationcoefficients are below 0.50, meaning that there is no seriousconcern about multi-collinearity.

5. Impacts from port investments

5.1. Estimation results: regional level

Estimation results at the level of the entire port region, withoutdummy variables for the region (Model I), are listed in Table 7,indicating that port infrastructure investment, economic structure,traffic infrastructure density, port spillover effects and internationalconnectivity positively influence the growth of regional economy.Next, by including the regional dummy variables (Model II), the modelresults (R2) improve from 0.428 to 0.827, and the coefficients of D1, D2,and D3 are all significant at the 5% level. We may thus conclude thatthe model including the regions can better explain the relationship ofport infrastructure investment, economic structure, traffic infrastruc-ture density, port spillover effects and international connectivity withregional economic growth compared to the model without theregions, pointing to some relevant differences between the fourregions. Overall, the coefficient of port investment (Model II) (0.191)indicates a positive impact on economic growth, which is in line withresults on port supply by Deng et al. (2013a, 2013b).

Table 6Results for panel unit root test and co-integration test for all data.

Unit root test Residual co-integration test (Kao)

Levels LLC First-differences LLC t-Statistics Prob.

ln Y 9.218 Δln Y �4.797*** ADF �2.703 0.0034ln K 3.007 Δln K �4.115*** Residual variance 0.007ln MAN �1.271 Δln MAN �8.572*** HAC variance 0.009ln TID �2.766** Δln TID �11.252***ln S 3.851 Δln S 9.146***ln IC 1.170 Δln IC �6.398***

Note: **Statistical significance at the 5% level, ***at 1% level.

Table 5Variables per port province and descriptive statistics.

Variables Indicator Descriptive statistics

Average S.D. Max Min

Y GDP in 100 million RMB 10,811.4 8858.0 46,013.0 1450.1K Port capital investment stock in 100 million RMB 93.89 140.46 1017.71 0.30MAN Share of manufacturing in GDP 0.43 0.063 0.54 0.28TID Average traffic infrastructure density (road, railway and inland waterways) in m/ km2 848.3 440.8 2302.5 331.9S Number of scale ports in adjacent provinces 6.308 3.157 12 2IC Number of aircraft movements in international airports 116,843.1 122,272.9 58,3762 6296

Note: The units for Y, K and MAN are 100 million RMB; the unit for TID is m/km2.

L. Song, M. van Geenhuizen / Transport Policy 36 (2014) 173–183 179

The results of the model for the four port regions (Table 8)show that the coefficients of port investment, port spillover effectsand international connectivity are significant at the 5 or 1% levelfor all four port regions. However, the coefficients of the share ofmanufacturing in GDP are only statistically significant for BohaiRim and Center at the 1 and 5% level of significance respectively,while these coefficients are not significant in other regions.Remarkably, the coefficients of landside transport infrastructuredensity are all not significant in the four port regions, this may bedue to the previously indicated non-linear relationships with port-related economic growth, which cannot be grasped in our linearmodels.

With regard to the effects of port infrastructure investment, weobserve the following implications. For Bohai Rim area, the GDPwill increase by 0.54% if the port infrastructure increases by 1%,and the GDP will increase by 3.81% if the size of manufacturingsector increases by 1%. This is also the region with the compara-tively highest manufacturing share in GDP in 2010. For YangtzeRiver, endowed with a modest road density in 1999 and facing arelatively strong traffic infrastructure density increase in theresearch period, the relatively high coefficient of port investment(0.1428) is in line with the idea that a better developed land trafficinfrastructure system, ‘produces’ higher impacts of port invest-ment on economic growth due to benefits from network effects,primarily accessibility (Banister, 2012). The development towardsa relatively small manufacturing sector in Yangtze River in favor ofthe services sector, however, is not sufficiently strong to give anegative coefficient that is significant for manufacturing share,most probably because the indicator used does only partially‘grasp’ size of advanced producer services. For the Southeastregion, as the best endowed area with road infrastructure in1999, the contribution of land traffic infrastructure is the lowest(a coefficient of 0.024), most probably because the Southeast hassurpassed the threshold of accumulated transport infrastructure in1999, and the extra investments in transport infrastructurescannot generate higher economic growth in the region. For theCenter part, the contribution of port infrastructure investment ison the lowest level (a coefficient of 0.09), reflecting a differentcharacter of the region, endowed only with land-ports. Howeverthis may change in the next coming years due to a relatively strongimprovement of the road system (Table 4). The contribution of thesize of manufacturing in the Center is significant (a coefficient of

0.318), indicating that the quickly growing manufacturing (shift tothe West) has a positive impact on regional economic growth here.

Table 8 also indicates that the spillover effects from scale portsin neighboring provinces are rather different. The Yangtze Riverregion enjoys the highest positive spillovers (a coefficient of 3.026)from scale ports in neighboring port provinces. While the BohaiRim region and Southeast are facing spillover effects of a similarmodest size (coefficients of 0.600 and 0.596 respectively), thiscoefficient in the Center is negative (�0.814), which implies thatthe port development in adjacent provinces tends to hindereconomic growth in here. A similar result is also found by Yuet al. (2013), in that the transport infrastructure spillover effectsare negative in central China, including Hubei, Hunan, Anhui,Jiangxi, Henan and Shanxi provinces.

With regard to the international connectivity, in the Center, thecontribution to the regional economy is the highest (a coefficientof 0.719), while the impact of international connectivity in theSoutheast is at a medium level (a coefficient of 0.596) compared tothe two other port regions, Bohai Rim and Yangtze River (coeffi-cients of 0.227 and 0.256, respectively). The trend of stronginfluence of international connectivity in the Center may beunderstood in the context of the shift to the West, which is mainlyundertaken by large multinationals and has increased flights fromand to international airports here.

5.2. Estimation results: provincial level

The coefficients concerning port investment of all provinces aresignificant at the 5% level (Table 9). Within the regions of BohaiRim, Yangtze River and Southeast region, there are no considerabledisparities between port investment outputs per province; incontrast, within the Center, a huge gap exists between the outputin Chongqing at the 0.903 level and that in Hunan which is at the0.092 level. Considering the size of manufacturing, only in Tianjin,Liaoning, Hebei, Shanghai and Hubei the coefficients are positiveand significant. These provinces, except for Shanghai and Hubei,enjoy a comparatively larger increase of the size of manufacturingin the period 1999–2010.

With respect to transport infrastructure density, only Tianjinand Shandong, holding a comparatively high road density in 1999and also a high growth rate of road in Shandong, show coefficientsthat are significant. This means that only in these two provinces an

Table 7Regression results (coefficients) of regional economic growth.

ln K ln MAN ln TID ln S ln IC D1 D2 D3 Adj. R2

Model I 0.132 (2.71) *** 0.757(2.02)** 0.478 (4.75) *** 0.383 (4.42) *** 0.362 (6.51) *** 0.428Model II 0.191 (6.02) *** 0.611 (2.75) *** 0.227 (3.12) *** 0.534 (9.84) *** 0.489 (10.78) *** 0.212 (2.50) ** 0.369 (2.92) *** 0.295 (2.14) ** 0.827

Note: t-statistics in parenthesis.nn Statistical significance at 5% level.nnn Statistical significance at 1% level.

Table 8Regression results (coefficients) of regional economic growth for four port regions.

Port region ln K ln MAN ln TID ln S ln IC Adj. R2

Bohai Rim (SP) 0.541 (6.588)*** 3.811 (6.843)*** 0.107 (1.333) 0.600 (4.673) *** 0.227 (3.680)*** 0.850Yangtze River Delta (SP and IP) 1.428 (11.994)*** 1.163 (1.370) 0.386 (1.676) 3.026 (12.939)*** 0.256 (2.081)** 0.789Southeast (SP) 0.506 (6.585)*** 1.039 (1.987) 0.024 (0.098) 0.596 (3.103)*** 0.598 (5.816)*** 0.782Center (IP) 0.093 (2.426)** 0.318** (1.331) 0.041 (0.220) �0.814 (�6.097)*** 0.719 (7.094)*** 0.759

Note: t-statistics in parenthesis.nn Statistical significance at 5% level.nnn Statistical significance at 1% level.

L. Song, M. van Geenhuizen / Transport Policy 36 (2014) 173–183180

increase of transport infrastructure density tends to result in localeconomic growth. In addition, various disparities in the impactfrom spillovers among provinces could be found, Liaoning, Hebei,Jiangsu, Guangdong, Chongqing and Hubei enjoy positive andsignificant coefficients, meaning that in these provinces, thedevelopment of the ports in neighboring port provinces tends topromote the economic growth in these provinces.

In contrast, Anhui province enjoys negative and significant spil-lovers. The adjacent province of Anhui is Jiangsu which is a well-developed province in the Yangtze River region, indicating that theport development in Jiangsu tends to absorb production factors fromneighboring area (Anhui province), and this is confirmed by apositive and significant coefficient for regional spillovers for Jiangsu.Furthermore, regarding the international connectivity, some obviousdisparities could be detected, namely, only five provinces (Tianjin,Liaoning, Fujian, Guangdong and Hunan) enjoy a positive andsignificant coefficient, indicating that in these provinces, thestrengthening of international connectivity tends to stimulate eco-nomic growth.

5.3. Discussion

Port infrastructure investment has a clear influence on theeconomy of the four port regions in China. Using a broader networkperspective in regression analysis, in this study, the coefficients ofport infrastructure investment turn out to be significant both on theregional and on the provincial level. However, the influence issubstantially different in strength between the four regions and thatneeds to be seen in relation to various other significant influenceson the regional economy.

The Yangtze River enjoys the highest port investment output,with highest spillover effects. The Bohai Rim and Southeastregions are somewhat behind with lower output elasticity. ForBohai Rim, this is most probably related to absence of a denseinland waterway system and less dense highway infrastructure,but also not yet a sufficient level of diversification with highvalued manufacturing and producer services in the port-city andhinterland. For Southeast, though endowed with well-developedtraffic system, the relatively low output of port investmentprobably because of the low spillovers from neighboring ports,while the port of Hong-Kong falls outside the current analysis. For

a better understanding, it needs to be mentioned that the previousthree regions, apart from the Center, are the most importanteconomic zones in China benefiting from the reform and openingup policy that started in 1978, aiming to introduce capitalistmarket principles. Especially Guangdong province in Southeastand Yangzte Delta have the priority in carrying out the economicreform, with the result three decades later of market factorsplaying an essential role in the regional economy and value chainsstretching around the globe. In contrast, Bohai Rim is a newgrowth pole developed only in the late 20th century, later thanSoutheast and Yangzte Delta regions, and is still in a transitionphase from a planning economy to a market economy, of whichthe maturity and competition of the market are behind Southeastand Yangzte Delta. The reason why the port investment outputelasticity of the Center is relatively low, is mainly the concentra-tion of the inland scale ports here, inhibiting a much smaller scaleand efficiency than many sea ports and specifically Yangtze riverinland ports (Wang and Meng, 2013), where market competitionhas developed for decades (Yuen et al., 2013). Meanwhile, thenegative spillovers might be another reason for the low level ofport infrastructure output. A situation of rapid economic growthdoes not apply for the Center at the beginning of the researchperiod in this study, though today this region is catching up.

In addition, China's regions are facing different stages of theindustrialization process (Chen et al., 2006), according to Chen et al.(2012), eastern China has accomplished the industrialization pro-cess, and the rest of China is in the late industrialization phase. Liuand Li (2002) pointed out that for developing countries which havenot accomplished their industrialization stage, economic growth isessentially driven by manufacturing. Our empirical findings showthat the size of the manufacturing sector in Bohai Rim (a coefficientof 3.81), which is in the late industrialization phase, has a largerimpact on economic growth compared to Yangzte Delta and South-east (coefficients of 1.163 and 1.039, respectively) which havealready finished the industrialization stage, except for Fujianprovince. Regarding Central area, which is an underdevelopedregion that benefits from the Central China Plan in 2009, manymanufacturing companies have shifted from coastal areas to thisregion, most probably including relatively low value-added manu-facturing activity. Hence, our findings further underpin the role ofmanufacturing in the regional economy (a coefficient of 0.318).

Table 9Regression results (coefficients) of regional economic growth for 13 port provinces.

Port province ln K ln MAN ln TID ln S ln IC

Bohai RimTianjin 0.399(2.74)*** 1.252(3.66)*** 1.136 (1.73)** 1.86(0.49) 0.173(2.61)**Liaoning 0.282(3.72)*** 0.998(5.57)*** 0.081(1.24) 12.41(2.18)** 0.493(2.33)**Hebei 0.521(6.09)*** 2.204(2.02)** 0.052(0.26) 5.48(5.16)*** 0.110(1.39)Shandong 0.333(2.25)** 0.464(0.90) 0.212(2.67)*** 0.415(0.39) 0.482(1.60)

Yangtze RiverShanghai 0.685(2.90)*** 0.716(2.46)** 0.041(0.14) 0.015(0.02) 0.368(1.65)Jiangsu 0.661(6.00)*** 1.150(1.91) 0.240(1.27) 1.589(2.86)*** 0.026(0.67)Zhejiang 0.452(1.70)** 0.183(0.20) 0.295(1.59) 0.722(0.71) 0.304(0.75)

SoutheastFujian 0.315(2.94)*** 0.209(0.90) 0.116(1.17) 1.09(0.86) 0.889(3.73)***Guangdong 0.494(2.84)*** 0.510(1.55) 0.237(1.52) 3.454(1.99)** 0.831(3.19)***

CenterChongqing 0.903(3.81)*** 0.519(1.04) 0.128(1.25) 12.411(2.18)** 0.473(1.07)Anhui 0.406(9.35)*** 0.052(0.14) 0.004(0.07) �2.656(-4.67)*** 0.149(1.36)Hubei 0.752(12.4)*** 0.830(4.34)*** 0.064(0.81) 5.176(8.76)*** 0.049(0.47)Hunan 0.092(2.57)** 0.504(1.17) 0.009(0.08) �0.412(�0.37) 0.922(5.11)***

Note: t-statistics in parenthesis. Adj. R2 for the panel model of 13 port provinces is 0.698.nn Statistical significance at 5% level.nnn Statistical significance at 1% level.

L. Song, M. van Geenhuizen / Transport Policy 36 (2014) 173–183 181

6. Conclusion, policy suggestions and issues for futureresearch

In this paper, we evaluated the economic impact of portinfrastructure in China, which has attracted attention amongresearchers only very recently. Based on this study, we mayconclude that port infrastructure investment has a positive impacton regional economic growth in China, but with obvious differ-ences at the regional and provincial level. These differences couldbe connected with differences between seaports and inland ports,different densities in land transport infrastructure, phases ofeconomic development, spillovers from ports in neighboringregions, as well as reform policies carried out by the centralgovernment.

In terms of policy making, we may forward some interestingsuggestions and ideas. Based on the empirical results, we believethat at the regional level, central government and local govern-ment should balance the investments in port infrastructurethrough specific policies, particularly in Center, where outputelasticity of port investment is relatively low. Improving portefficiency is much more important than the increase of physicalinfrastructure only, meaning better port management and inlandtransport connectivity, but also a much quicker integration of theports with supply chains connected to the regional economy. Thiscould be enhanced by the creation of specialized, high value-added, clusters of port-related economic activity, includingadvanced maritime producer services, but this is a long-termeffort. On the short term, feasible strategies of fast establishingthe linkages between ports and existing clusters or growth poles,through which value added can be increased (Pettit and Beresford,2009) should be carried out, this is in the frame of an overallurban policy for knowledge-based development (Geenhuizen andNijkamp, 1998). In such scenario, also so-called breeding places fordevelopment and experimentation in new port and maritimetechnology and logistics need to be established in the port areain connection with local research institutes (universities), this toenhance a stronger emergence and growth of small technology-based firms.

In addition, the empirical results on the landside traffic infra-structure density indicate that the government should pay moreattention on investment in connecting traffic infrastructuresbecause at the entire port region level, the landside trafficinfrastructure density has a positive and significant impact oneconomic growth, but at the provincial level, only two provinces(Tianjin and Shandong) tend to benefit from landside trafficinfrastructure. Therefore, for the port provinces, intermodalityamong various traffic modes should be considered to improvethe connectivity of transport network within the province ratherthan merely increasing investment in the ports.

Furthermore, for the Yangtze and Southeast regions, our resultsindicate that the size of the manufacturing sector is somewhatdecreasing in favor of services, pointing to the need of a reconsi-deration of port development in the next coming years. There isthe question of reaching the optimal level in the near future, abovewhich additional port investments only produce small growth.Also, and related, there is the question of increased competitionbetween ports in the same region, like between Hong Kong,Shenzhen and Guangzhou in container transport, not yet depictedin the current study. The magnitude of these developments andtheir implications need to be investigated before new decisions onlarge infrastructure investment are taken. This is connected with ageneral increasing uncertainty that is replacing the stability of theglobal business environment and growth of the Chinese economy,whereas ports have a long lifetime and port investments are oftenirreversible. This situation requires to pay a stronger attention tohow the implications of uncertainty can be incorporated in the

way of planning and design of ports, like flexibility in design(adaptive port planning) and real options analysis (Walker et al.,2001; De Neufville and Scholtes, 2011).

Despite the interesting results, the study also suffers from someweaknesses which could be addressed in future research. Firstly,we adopted the idea of non-linear economic growth inhibitingthreshold values, only by assuming that the relatively smaller portareas have not exceeded a lower threshold, among others due toless developed land transport infrastructure. In future research,this idea, including a second threshold as an optimal level abovewhich growth diminishes, could be incorporated in a dynamicmodel inhibiting non-linearity, like in the study of Deng et al.(2013a, 2013b) on highways in China. Secondly, due to datalimitations, we used various relatively broad indicators, of whichwe mention the number of scale ports in neighboring portprovinces to indicate regional spillovers effects and the regionaleconomic structure to indicate structural shifts but in whichadvanced producer services remained unidentified. In futureresearch, spillovers should be disaggregated to specific portactivities to explore the spillovers in greater detail and the servicessector should be measured at a more disaggregated level as well.Thirdly, our study has remained broad as we made no distinctionbetween different value chains and different port activity in themodeling. In future research, differences could be made betweeninvestments in ports predominantly active in bulk, containertransport, other cargo, or a mix, and ports connected with (petro)chemical industry, steel industry, food industry, assemblage indus-try, etc. As a fourth point we forward that the observations in thisstudy are the scale ports in China, which do not offer a completepicture of port development in China, many inland ports are nottaken into account, which also have an impact on the regionaleconomy. Hong Kong port as part of China could also be includedin future research.

Acknowledgments

We thank the anonymous reviewers and the editor for theirconstructive comments. We express our appreciation for theVisiting Study Funding of China Scholarship Council (Grant no.201207167011). We thank Mr. Liang Yan for his contribution to themaking the map.

References

Banister, D., 2012. Transport and economic development: reviewing the evidence.Transp. Rev. 32 (1), 1–2.

Berechman, J., Ozmen, D., Ozbay, K., 2006. Empirical analysis of transportationinvestment and economic development at state, country and municipalitylevels. Transportation 33 (6), 537–551.

Biehl, D. 1986 The Contribution of Infrastructure to Regional Development: FinalReport. Office for Publications of the European Communities (Luxembourg/Washington DC).

Blum, U., 1982. Effects of transportation investments on regional growth: atheoretical and empirical investigation. Pap. Reg. Sci. 40 (1), 169–184.

Bottasso, A., Conti, M., Ferrari, C., Tei, A., 2014. Ports and regional development: aspatial analysis on a panel of European regions. Transp. Res. A: Policy Pract. July2014, 44–45.

Canning, D., Bennathan, E, 2007. The Rate of Return to Transportation Infrastruc-ture. World Bank Working Paper no. 3293.

Chen, J., Huang, Q., Zhong, H., 2006. The synthetic evaluation and analysis onregional industrialization. Econ. Res. 6, 4–15.

Chen, J., Huang, Q., Lv, T., Li, X., 2012. The Report on China's Industrialization. SocialSciences Documentation Publishing House, Beijing.

Danielis, R., Gregori, T., 2013. An input–output-based methodology to estimate theeconomic role of a port: the case of the port system of the Friuli Venezia GiuliaRegion, Italy. Marit. Econ. Logist. 15, 222–255.

Del Bo, C., Florio, M., 2012. Infrastructure and growth in a spatial framework:evidence from the EU regions. Eur. Plan. Stud. 20 (8), 1393–1414.

Deng, P., Lu, S., Xiao, H., 2013a. Evaluation of the relevance measure between portsand regional economy using structural equation modeling. Transp. Policy 27,123–133.

L. Song, M. van Geenhuizen / Transport Policy 36 (2014) 173–183182

Deng, T., Shao, S., Yang, L., Zhang, X., 2013b. Has the transport-led economic growtheffect reached a peak in China? A panel threshold regression approach.Transportation, (published online 15 October).

De Neufville, R., Scholtes, S., 2011. Flexibility in Engineering Design. The MIT Press,Massachusetts, The United States.

Derudder, B., Witlox, F., 2005. An appraisal of the use of airline data in assessing theworld city network: a research note on data. Urban Stud. 42, 2371–2388.

Ducruet, C., Koster, H., Beek, D. van der, 2010. Commodity variety and seaportperformance. Reg. Stud. 44, 1221–1240.

Ducruet, C., Itoh, H., Joly, O., 2013. Ports and the local embedding of commodityflows. Papers in Regional Science, (published online: 22 Dec. 2013).

Geenhuizen, M., Nijkamp, P., 1998. Improving the knowledge capability of cities:the case of Mainport Rotterdam. Int. J. Technol. Manag. 15 (6/7), 691–709.

Goldsmith, S.B., 1951. A Perpetual Inventory of National Wealth, Studies in Incomeand Wealth 14. NBER, New York.

Hong, J., Chu, Z., Wang, Q., 2011. Transport infrastructure and regional economicgrowth: evidence from China. Transportation 38, 737–752.

Jacobs, W., Koster, H., Hall, P., 2011. The location and global network structure ofmaritime advanced producer services. Urban Stud. 48 (13), 2751–2771.

Jiang, N., 2010. Seaport investment and economic development in China. DalianMarine University Press, Dalian.

Kao, C., 1999. Spurious regression and residual-based tests for co-integration inpanel data. J. Econom. 90 (1), 1–44.

Ke, S., Feser, E., 2010. Count on growth pole strategy for regional economic growth?Spread- backwash effects in Greater Central China. Reg. Stud. 44 (9), 1131–1147.

Levin, A., Lin, C.F., Chu, C., 2002. Unit root tests in panel data: asymptotic and finite-sample properties. J. Econom. 108, 1–24.

Li, D.S., Yuan, L.F., 2010. Efficiency measurement of Port Group Logistics System—

based on multiple subsystem fuzzy DEA model. J. Beijing Jiaotong Univ. (Soc.Sci. Ed.) 9 (2), 42–47.

Liu, W., Li, S.R., 2002. Industry structure and economic growth. China Ind. Econ. 5,14–21.

Matsumoto, H., 2007. International air network structures and air traffic density ofworld cities. Transp. Res. Part E: Logist. Transp. Rev. 43, 269–282.

Mera, K., 1973. Regional production functions and social overhead capital. Reg.Urban Econ. 3, 157–186.

Merk, O., Hesse, M., 2012. The Competitiveness of Global Port-cities: The Case ofHamburg. OECD, Paris.

MOCOPRC (Ministry of Communications of the People's Republic of China), 2006.The Coastal Port Layout Planning. Chinese Government Nets. ⟨http://news.xinhuanet.com/politics/2006-11/20/content_5352716.htm⟩ (accessed 20.11.06).

Nijkamp, P., 1986. Infrastructure and regional development: a multidimensionalpolicy analysis. Empir. Econ. 11, 1–21.

Notteboom, T., Rodrigue, J.P., 2005. Port regionalization: towards a new phase inport development. Marit. Policy Manag. 32, 297–313.

Ozbay, K., Ozmen-Ertekin, D., Berechman, J., 2003. Empirical analysis of therelationship between accessibility and economic development. J. Urban Plan.Dev. 129 (2), 97–119.

Pettit, S.J., Beresford, A.K.C., 2009. Port development: from gateways to logisticshubs. Marit. Policy Manag. 36 (3), 253–267.

Walker, W.E., Rahman, S.J., Cave, J., 2001. Adaptive policies, policy analysis andpolicy making. Eur. J. Oper. Res. 128 (2), 282–289.

Wang, C., Ducruet, C., 2012. New port development and global city making:emergence of the Shanghai-Yangshan multilayered gateway hub. J. Transp.Geogr. 25, 58–69.

Wang, C., Ducruet, C., 2013. Regional resilience and spatial cycles: long-termevolution of the Chinese port system (221BC–2010AD). J. Soc. Econ. Geogr.104 (5), 521–538.

Wang, L., Meng, H., 2013. Efficiency comparison between inland port and seaport inChina. Soft Sci. 27 (3), 90–95.

Warf, B., Cox, J., 1989. The changing economic impacts of the port of New York.Marit. Policy Manag. 16 (1), 3–11.

Yoo, S.-H., 2006. Seaport infrastructure investment and economic growth in Korea.Int. J. Crit. Infrastruct. 2 (1), 1–9.

Yu, N., de Jong, M., Storm, S., Mi, J., 2013. Spatial spillover effects of transportinfrastructure: evidence from Chinese regions. J. Transp. Geogr. 28, 56–66.

Yuen, A., Zhang, A., Cheung, W., 2013. Foreign participation and competition: a wayto improve the container port efficiency in China? Transp. Res. Part A 49,220–231.

Zhang, J., Gao, Y., Fu, Y., Zhang, H., 2007. Why does China enjoy so much betterphysical infrastructure? Econ. Res. 4 (3), 4–19.

Zhang, Z., Wen, H., Shen, Y., 2009. Granger causality analysis on the economy andtransportation infrastructure construction. J. Wuhan Univ. Technol. (Transp. Sci.Eng.) 33 (4), 699–702.

L. Song, M. van Geenhuizen / Transport Policy 36 (2014) 173–183 183