Impact of Port Reform, Political and Economic Events on Maritime Traffic in Chinese Ports

Transcription

1 Impact of Port Reform, Political and Economic Events on Maritime Traffic in Chinese Ports Dong Yang a1 Anthony TH Chin b Shun Chen c a Centre for Maritime Studies, National University of Singapore, Singapore b Department of Economics, National University of Singapore, Singapore c College of Transport & Communications, Shanghai Maritime University, Shanghai, China Abstract Chinese ports have undergone a series of dynamic international and domestic events, economic and political reform since the founding of the People s Republic. This paper attempts to ascertain if these political, economic events and port reforms have had an impact on maritime traffic of Chinese ports by using econometric methods. Findings suggest that foreign trade drives the increase in throughput of Chinese ports, especially foreign and coastal port throughput. In contrast, this increase in port throughput has led to an increase in domestic retail sales (or domestic demand) as well as more port investment. In its development, port traffic was influenced by multiple shocks. Among all the events, the Great Leap Forward has exerted the biggest influence to the throughput of Chinese ports, especially in short run. China s accession to WTO brought an exclusive and minor effect to port throughputs. Sudden rises in port throughput were largely a consequence of implementation of port policy but not as prominent as effect from economic and political events. Port structural reform is proved to be more efficient and long lasting than simple investment in the port infrastructure construction. Keywords: Economic and political events, Port reforms, Maritime traffic in ports, Structural break 1 Corresponding Author s Address: Research Fellow, Dong Yang, Centre for Maritime Studies, National University of Singapore, 12 Prince George s Park, Singapore yangdong@nus.edu.sg 1

2 1. Introduction. The Chinese economy has undergone several dramatic transformations since As such some of the economic indicators may not be trend stationary. Li (Li 2000) and Smyth (Smyth 2004) analyzed national and provincial GDPs and concluded that these GDPs present a stationary process with one or two breaks as opposed to a unit-root process. Wang et.al (2008) revealed that Chinese export trend is a piecewise stationary series subject to two or more breaks. Ma (2008) recognized a structure break for Chinese saving rate at The structural break in these historical data may be due to a series national political or economic events, such as the Great Leap Forward ( ), Three years of natural disasters ( ), Cultural Revolution ( ), Open Door Policy and accompanying Economic Reforms ( ), Macroeconomic Austerity Program ( ), Deng s Southern Tour (1992), China's World Trade Organization Accession (2001), SARS (2003) and so on. The throughput of Chinese port is also found be a non-stationary series. Potential shocks may come not only from above-mentioned national economic or political events but also a series of policies implemented in the port sector. In the first two decades after 1949, Chinese port throughput was influenced by domestic trade and the development of ports was slow. Foreign shipping traffic grew rapidly with the establishment of foreign diplomatic relations. The capacity of coastal ports was not able to catch up with the increase in shipping activities. In 1973 a threeyear project to upgrade the Chinese coastal port's facilities was initiated where CNY 2.3 billion has been invested in the port facilities construction. This was followed by decentralization of authority as an outcome of port structural reform after 1984 where 37 out of 38 major ports were jointly operated jointly managed by the Ministry of Communications and the local governments until The latest change in Chinese ports policy was initiated after China s accession to WTO in The expected growth in trade and the economy led to the introduction of Port Law and its complementary, regulations such as the Rules on Port Operation and Management, in A modern enterprise system was gradually introduced into the port sector allowing operators greater autonomy in operations and management. In addition to the above events, several international shocks should also be taken into account in the study of port throughput such as the Oil crises (1973, 1979 and 1990), Asian Financial Crisis (1997) and the Global Financial Crisis (2008). Table 1 gives a timeline of events which might have an impact on throughput of port. The objective of this paper is to ascertain the relative impact of the above events on the maritime traffic of Chinese ports since The study is organized into three stages. Stage one firstly tests for unit roots with the assumption of no structural breaks in Chinese ports throughput series data. If a series contains a unit root, then any certain event or reform is of limited value and long run growth in throughput is probably affected by multiple factors. However, if the series is stationary, this implies that only huge shock will have at least a semi-permanent effect on the growth path (Li, 2000). Unit root test allowing for structural break is then performed to identify the exact point of the most significant shock which affected port traffic. In stage two, we control for the effects of economics variables such as foreign trade, Chinese domestic retails sales and port investment to verify a co-integration relationship with Chinese ports throughput. If co-integration exists, then granger causality test will be employed based on error correction model estimation (VECM) in an attempt to explain the interaction between ports throughputs and those economics variables. Stage three analyzes specifically with events which may have impact on port performance. We test for co-integration between ports throughput and corresponding economic variables allowing structural break. A potential break in co-integration implies deviation. The timing and significance of break point will indicate exclusive impact of events on port throughput. 2

3 Table 1 Timeline of Political, Economic and Port Reforms Year Events Scope Great Leap Forward National Three years of natural disasters National Cultural Revolution National Three-year project to upgrade the Chinese coastal ports Coastal Ports Open Door Policy and accompanying Economic Reforms National Reform: semi-decentralization or dual-administration system Major Ports Macroeconomic Austerity Program National 1992 Deng s Southern Tour National 1997 Asian Financial Crisis Asia 2001 China s World Trade Organization Accession National 2003 SARS National 1973, 1979,1990 Oil Crisis Global 2004 Reform: Rules on Port Operation and Management All Ports 2. Literature Review There is a lot literature on economic analysis of port development and reform. Pallis et al. (2010) reviewed 395 relevant journal papers in port economics, policy and management during the period Among which, Port Governance as one of total seven classifications, achieves a lot of research attention (61 relevant papers out of all 395 papers). Ircha,(1997, 2001) reviewed and evaluated the Canadian port reform and also outlined the evolution of strategic planning and its applicability to Canadian ports. He kept a positive attitude to the port reform in Canada and concluded that the reform will lead to further rationalization and enhanced competitiveness of Canadian ports. Everett (Everett and Robinson 1998; Everett 2007) argued in a corporatization port regime, where government has specifically chosen to retain ownership but policy prescribes that government businesses will be profitable, bureaucrat and minister impose significant constraints on achieving competitive efficiency. She also examined the impacts and constraints the state government regulator imposed on terminal expansion and operations and suggested the transfer of state government ports to a single national regulator. Hadi Baaj and Issa (2001) presented the institutional reform of the Lebanese maritime transport sector. He listed different port autonome models and asserted that the State Corporation model is the most high scoring and powerful model. Mexico s port system was centrally managed by public firms until Then reforms liberalized and decentralized it to regional port authorities to improve its efficiency. Estache (2004) measured the changes and sources of efficiency since the reforms. He suggested that port reforms were quite successful in contributing to improvements in Mexico s economic competitiveness. Serebrisky and Trujillo (2005) showed that structural reform caused significant efficiency gains to Argentine ports and also identifies outstanding issues, such as vertical mergers in the port of Buenos, could impact the long-run sustainability of the gains achieved. Castillo-Manzano et al (2008) applied an estimated econometric model to investigate whether there is an important impact of legislative changes on Spanish port traffic with econometric technology by using data over the period of He provided evidence supporting that greater port autonomy had beneficial effects for the Spanish port system as a whole. Cullinane and Wang (2006) described China s policies of economic reform since the 3

4 inauguration of its open door policy in They divide the development of Chinese ports into three distinct phases: 1978 to 1984, 1984 to 2004 and 2004 to now according to different reform contents. Still with regard to Chinese ports reform, Qiu (Qiu 2008) focused on the economic background, motivations and progress, and discusses issues associated with relevant planning events. He also recognized three phases in the development of Chinese ports. However he regarded the 2001 as the beginning of the third phase instead of 2004 which Cullinane and Wang believed. Qiu (2008) made his conclusion that the reforms are necessary for the ports industry to raise funds for infrastructure expansion and to enhance the industry efficiency but Cullinane and Wang asserted that it is still too early to tell whether the latest phase of reforms will prove to be successful in solving China s port problems. This paper builds upon the above literatures in two ways. First, the analysis in this paper is drawn from a long time series data starting from 1952 considering the impacts from not only reforms but also a series of economic and political events. Previous studies began from early 1980 s and only focus on reform. Second, most previous studies on port reform and performance tend to be qualitative. We employ an empirical approach controlling for the endogenous structural breaks. Instead of ascertaining a simple structural break test, this paper proposes both unit-root and cointegration tests allowing structural break so as to understand impacts from events in different perspectives. 3. Data Description The annual data utilized covers the period 1952 to 2009 drawn from various sources such as Comprehensive Statistical Data and Materials on 50 Years of New China (1999), Transportation Statistical Data and Materials on 50 Years of New China:1949~ 1999 (1999), China Shipping Development Annual Report 2009 (2010), China Statistical Yearbook 2010 (2010), China Trade and External Economic Statistical Yearbook 2010 (2010), China Marine Statistical Yearbook 2009 (2010) and Chinese Statistical Bulletin for Shipping and Road transportation (2001 ~ 2010). The data is divided into two groups, among which, different Chinese ports throughputs are primary research objects and economic variables as proxy. Chinese ports throughput is recorded in two classifications: main coastal ports and main inland ports. Main ports are also described as Ports above a designated size and this refers to: 1) Coastal ports with over one million handling capacity in one year; 2) Inland port with over 2 million handling capacity in one year; 3) Ports with license of foreign transportation business. These ports actually account for dominant traffic amount over all Chinese ports. For each classification of port, it contains two time series which are Domestic Trade throughput and Foreign Trade throughput. Therefore, totally, there are four records for ports throughput, which are 1) Domestic throughput of main inland ports; 2) Foreign throughput of main inland ports; 3) Domestic throughput of main coastal ports and 4) Foreign throughput of main coastal ports. We recompose these throughputs and look at five definitions of throughput time series data in this article, i.e. i) Domestic and Foreign throughput of Chinese main INland ports (DFIN); ii) Domestic and Foreign throughput of Chinese main COastal ports (DFCO); iii) Domestic throughput of Chinese main INland and COstal ports (DINCO); iv) Foreign throughput of Chinese main INland and COastal ports (FINCO); and, v) Domestic and Foreign throughput of Chinese main INland and Coastal ports (DFINCO). The economic explanatory variables used in the analysis are, Chinese foreign trade (TRADE) which contains both import and export trade, Chinese Domestic Retail Sales (RS) which also indicates the level of Domestic Demand, Chinese main INland and Coastal ports investment (INCOVE) including INland ports investment (INVE), COastal ports investment (COVE) and 4

5 Chinese main Inland and Coastal ports Berth Number (INCOBN) including main INland ports Berth Number (INBN) and main COastal ports Berth Number (COBN). All variables are in natural logarithmic form for the purpose of scaling. Values of throughputs are valued in thousand ton in the data source. Values of variables such as TRADE, RS, INCOVE, INVE and COVE (deflated by Chinese Consumer Price Index, CPI) are in Chinese Yuan (CNY). 4. Results 4.1 Unit root tests without structural breaks Figure 1 summarizes the throughput for Chinese ports. All throughput data except for foreign throughput of Chinese main ports (FINCO) experienced a fall in the beginning of 1960s. The foreign throughput of main ports (FINCO) grew faster than throughputs of other ports. Figure 2 shows the trends of relevant economic variables. Retail sales (RS) and ports investment (INCOVE) also exhibit a dip around However, the growth of foreign trade is higher than that of all the other variables while port investment (INCOVE) fluctuates with time. 14 DINCO FINCO DFCO DFIN DFINCO 20 INCODF TRADE RS INCOBN INCOVE Figure 1 Trends of Different Chinese Ports Throughput Figure 2 Trends of Relevant Proxy Varibles We perform the Augmented Dickey-Fuller (Dickey 1979) unit root test and ensure the robustness of its results by Phillips-Perron test for all the time series data. The results of unit-root test are summarized in Tables 2 and 3. Table 2: Unit Root Test for All Derived Time Series of Chinese Ports Throughput DFIN DDFlN DFCO DDFCO DINCO DDINCO FINCO DFINCO DFINCO DDFINCO ADF t-stat Prob PP t-stat Prob Notes: ADF is the Augmented Dickey and Fuller (1981) test. The lag length of the ADF test is determined by minimizing the SBIC. PP is the Philip and Perron(1988) test. The most optimal lagged term for the model is selected according to Schwarz Information Criterion (SIC). Table 3: Unit Root Test for Potential Proxy variables TRADE DTRADE RS DRS INVE DINVE COVE DCOVE INCOVE DINCOVE COBN INCOBN ADF PP t-stat Prob t-stat Prob

6 Results show that except for foreign throughput of Chinese main ports (FINCO), the null hypothesis of a unit root cannot be rejected on the log levels of all the other series at the 1% significance level, while it is rejected on the log-first differences of all the series. These variables are then considered as I(1), i.e. integrated of order one. The FINCO rejects the null hypothesis of a unit-root at the 10% significance level by ADF test and the Phillips-Perron test at the 5% significance level. The results imply that Chinese port domestic traffic as well as Chinese port traffic as a whole are affected by multiple shocks. However, it also suggests the FINCO is some degree of trend stationary and relatively less influenced by less shocks than the domestic throughputs of Chinese main ports. 4.2 Unit root tests with structural breaks The traditional unit root test fails to reject the unit root hypothesis for the series that are actually trend stationary with structural breaks. Over fifty years after the foundation of P.R.China, as the macro-economic environment, society and ports policy experienced several significant, it is necessary to ask whether or not the traditional unit root test results are in a biased interpretation when one or two stationary alternatives are true in port throughput series and a structural break is ignored. It is interesting to ascertain what event that could has contributed a potential break. Perron (1989) firstly proposed a unit root test allowing for a structure break with three alternative models: crash model (shift in the intercept), changing growth model (change in slope) and the model containing change both in intercept and slope. However the models have been criticized for treating the time of breaks as exogenous (i.e., the time of break is known a priori). Building upon Perron s models, Zivot and Andrews (1992) developed three forms of the sequential trend break to endogenous break test (ZA model). We employ methodology A and C of ZA model. Model A aims to test short term effect and Model C treats the long run effects. Our hypothesis is that if events had short term effect it would be reflected by the intercept coefficient γ in model A and long term effect would be represented by the slope coefficient θ in model C. The following models is designed to explain the breaks in maritime traffic evolution of Chinese ports Model A: y t = c + α*y -1 + β*t + γ*du t + Model C: y t = c + α*y -1 + β*t + θ*dt t + γ*du t + DU t = DT t = Here, DU t and DT t are dummy variables used to capture the effect of shocks at break time T b. DU t indicates level (intercept) shift and DT t specifies the slope change. To apply these models, the break point is searched for over range of the sample ( T), which means the possible breaks before 1958 and after 2003 cannot be recognized due to data limitation. The t-sig approach suggested by Hall (1994) is applied to decide the optimal lag length k. The choice of sample size and t-sig approach will also be applied for the Gregory and Hansen s co-integration test in the following section. Although Zivot and Andrews (1992) provide an asymptotic critical value for this test, it may deviate substantially in terms of different sample sizes. Therefore, we calculate the exact critical values for our cases following the methodology recommended in Zivot and Andrews (1992, p.262). The results of ZA test for different Chinese ports throughputs and economic variables are shown in Table 4 and Table 5: 6

7 Table 4 Zivot and Andrew Test for Unit Roots with One Structural Break: Model A DFIN DFCO DINCO FINCO DFINCO TRADE RS COVE INCOVE COBN INCOBN Break t α * * k γ *** 0.15 *** *** 0.25 *** *** *** *** *** *** 0.19 *** 0.55 *** Note: Exact Critical Value (t α): -5.97(1%), -5.68(5%), -5.32(10%); ***,**,* indicate significance at 1%, 5% and 10% levels, respectively. Table 5 Zivot and Andrew Test for Unit Roots with One Structural Break: Model C DFIN DFCO DINCO FINCO DFINCO TRADE RS COVE INCOVE COBN INCOBN Break t α k γ *** *** *** *** *** 0.95 *** *** 0.24 *** 0.55 *** θ *** *** *** ** *** 0.01 Note: Exact Critical Value (t α): -6.49(1%), -6.37(5%), -6.25(10%); ***,**,* indicate significance at 1%, 5% and 10% levels, respectively. The results of unit root test allowing one structural break are basically consistent with those of the standard ADF and Phillips-Perron tests. There are no evidences against accepting the unit root null hypothesis at the 5% significance level for all the time series data in both model A and model C. It confirms that most throughputs of Chinese ports on the long-run growth path have been influenced by multiple events. In model A, the most significant break point occurs following the Great Leap Forward for some time series including domestic and foreign throughput of Chinese ports (DFINCO), domestic and foreign throughput of Chinese main inland ports (DFIN), domestic throughput of Chinese main inland and costal ports (DINCO), economics variables (TRADE, Retail Sales) and ports investments (INCOVE and COVE). The foreign throughput of Chinese main ports (FINCO) is trend stationary at the 10% significance level in model A with an upward break in 2002 when China entered the WTO. The most significant break of throughput of Chinese main coastal ports (DFCO) is also in 2002, but it cannot reject the null hypothesis of a unit root at the 10% significance level. All the coefficients of dummy variables are significant at 1% significance level. The results from Model C suggest that the null hypothesis of a unit root cannot be rejected for time series at level at 10% significance level. The most significant breaks for most of the series data are around This include domestic and foreign throughput of Chinese main inland ports (DFIN); domestic and foreign throughput of Chinese main coastal ports (DFCO), domestic throughput of Chinese main inland and costal ports (DINCO), domestic and foreign throughput of Chinese main inland and coastal ports (DFINCO), Retail Sales (RS) and Chinese main ports investment (INCOVE). The coefficients of dummy variables (γ) are all significant at 1% significance level while those of dummy variables (θ) are only significant in some cases. Some of them (DFIN, DINCO and DFINCO) are not significant at the 10% significance level. A positive change in slope at 1% significance level is observed for foreign throughput of Chinese ports (FINCO) in 1998, a year after the Asian Financial Crisis, the same as the foreign trade. But the intercept coefficients of them are negative. This suggests that the crisis affected not only the trade but also the foreign throughput of Chinese ports, followed by a quick recovery. Coastal port investment is not trend stationary with one break, the intercept coefficient is significantly positive at a 1% significance level and the slope coefficient is significantly positive at a 5% significant 7

8 level given a structural break in Therefore, the implementation of the three-year project to upgrade the Chinese coastal port is proved to generate a rise of coastal ports investment. Similarly, the intercept coefficients and slope coefficient of Chinese coastal ports berth number (COBN) and the intercept coefficients of Chinese main ports berth number (INCOBN) are significantly positive at 1% significance level with a break in It confirms an increase of port construction from 1986, two years after the beginning of the first port structure reform. The Great Leap Forward has greatest influence on Chinese ports as throughputs fell together with other economic indices, foreign trade (TRADE), Domestic Demand (RS) and ports investment (INCOVE). This however didn t lead to a drastic decline to the foreign throughput of Chinese ports (FINCO) and is in short run and. Although the implement of ports project in 1972 generated substantial investment in port construction and ports structural reform during 1984 to1987 resulted in a sharply increase in berth number, a lack of evidences to support the hypothesis that these measures promote the throughput of Chinese ports in a long run significantly. The sign of port boom appeared during the turn of the new century when China recovered from the Asian Financial Crisis and accessed to WTO. 4.3 Co-integration tests without structural breaks It is noted that the ports maritime traffic series exhibit similar trend in relation to economic indications (see Figures 1 and 2). Figures 3 to Figure 8 show the cyclical trends of their changing percentage. Most series show large down turn around 1961 and almost keep positive thereafter. Three cycles are apparent around 1978 to1989, 1991 to1996 and after 1999 with some differences. Ports investments suffer from cyclical changes with greater downturns and upswings. Trade and ports throughputs are similarly affected. The retail sale is relatively stable. 1 DFINCO TRADE RS INCOVE 1 DINCO RS INCOVE Figure 3 Percent Change of DFINCO, TRADE, RS and INCOVE Figure 4 Percent Change of DINCO, RS and INCOVE FINCO TRADE INCOVE 1.5 DFIN RS INVE Figure 5 Percent Change of FINCO, TRADE and INCOVE Figure 6 Percent Change of DFIN, RS and INVE 8

9 DFCO TRADE COVE 1.5 INCODF INCOVE INCOBN Figure 7 Percent Change of DFCO, TRADE and COVE Figure 8 Percent Change of DFINCO, INCOVE and INCOBN Given the apparent changes, we now proceed to test the relationships between port throughputs and other economic indices. Although the Johansen (Johansen 1988) co-integration test has seen a commonly used, we employ the Engle and Granger (1987) s approach 2 to test the co-integration without structural break in this article because the methodology complements the Gregory-Hansen approach which will be used later in the article. Table 6 Engle and Granger co-integration test Endogenous t e c µ TRADE µ RS µ INCOVE β Critical value (t e ) DFINCO TRADE+RS+INCOVE * *** *** (1%) -3.89(5%) -3.56(10%) TRADE+RS ** *** *** (1%) -3.44(5%) -3.12(10%) RS *** *** *** (1%) -2.91(5%) -2.59(10%) TRADE *** 0.33 *** *** -4.12(1%) -3.49(5%) -3.17(10%) INCOVE * 6.37 *** *** (1%) -2.91(5%) -2.59(10%) DINCO RS+INCOVE * *** *** (1%) -3.44(5%) -3.12(10%) RS ** *** *** (1%) -2.91(5%) -2.59(10%) INCOVE *** *** (1%) -2.91(5%) -2.59(10%) FINCO TRADE+INCOVE * *** ** (1%) -3.44(5%) -3.12(10%) TRADE ** 5.03 *** 0.12 ** *** -4.12(1%) -3.49(5%) -3.17(10%) INCOVE *** (1%) -2.91(5%) -2.59(10%) DFIN (INVE) RS+INVE ** ** *** 0.19 *** (1%) -3.44(5%) -3.12(10%) RS ** *** *** (1%) -2.91(5%) -2.59(10%) INVE ** 6.04 *** *** (1%) -2.91(5%) -2.59(10%) DFCO (COVE) TRADE+INCOVE * 2.84 *** 0.59 *** * (1%) -3.44(5%) -3.12(10%) TRADE ** 5.57 *** 0.27 ** (1%) -2.91(5%) -2.59(10%) COVE * 6.35 *** *** (1%) -2.91(5%) -2.59(10%) Note: ***,**,* indicate significance at 1%, 5% and 10% levels, respectively. The bold indicate a co-integration with above 5% significance level. 2 Engle and Granger expression: et = y t - c - µ 1 x 1t - µ 2 x 1t - µ n x nt βt, the exact critical values of e i are calculated according to James (2010) in terms of different number of repressors and the existence or non-existence of constant and trend items. 9

10 The choice of repressors for port throughput is based on the following assumptions: 1) Domestic ports throughput and inland ports throughput interact with Chinese Retail Sale; 2) Foreign ports throughput and coastal port throughput interact with Chinese foreign trade; 3) Ports throughput interacts with corresponding Chinese ports investment, for example, throughput of all ports (DFINCO) corresponds to all ports investment (INCOVE), throughput of coastal ports (DFCO) corresponds to coastal ports investment (COVE) and 4) Because the different in the capacities of berths, berth number is not taken into consideration in co-integration test. Table 6 shows the results of Engle and Granger co-integration test for different pairs of variables. According to the test results, the throughputs of DFINCO, DINCO and DFIN which are relevant to domestic ports throughput are all co-integrated with Retail Sales (RS) at 5% significance level. In addition, a co-integration relationship at the 5% significance level is also observed between the total throughput of inland ports (DFIN) and inland ports investment (INVE). Similarly, the coastal ports relevant throughput (FINCO and DFCO) all show a co-integration relationship with foreign trade (TRADE) at 5% significance level. When TRADE and RS are tested for cointegration with total throughput of all ports (DFINCO) simultaneously, the RS coefficient µ RS is significant at 1% significance level while TRADE coefficient µ TRADE is not significant at 10% significance level. It shows the Chinese total ports throughput (DFINCO) has a more similar pattern with the Chinese Retail Sales (RS) in trend compared to foreign trade (TRADE) which maybe due to the lock States of China before 1980s. 4.4 Granger Causality test on the lead-lag relationship We next employ VECM (Vector Error Correction Model) to identify the lead-lag causal relationship among variables where a co-integration relationship exists ( Granger 1988). Table 7 summarizes the results of Granger causality test performed, it shows that Chinese main ports all throughput (DFINCO), domestic throughput of Chinese main ports (DINCO), throughput of Chinese inland ports (DFIN) granger cause Retail Sales (RS) at the 10% significance level without feedback effect. Meanwhile, throughput of Chinese inland ports (DFIN), throughput of Chinese coastal ports (DFCO) granger cause inland ports investment (INVE) and coastal ports investment (COVE) at the 10% significance level without feedback effect respectively. The causality from the throughput of Chinese coastal ports (DFCO) to coastal ports investment (COVE) is found to be significant at the 1% significance level. A unidirectional causality effect can be observed from Chinese foreign trade (TRADE) on foreign throughput of main ports (FINCO) at the 1% significance level. A bilateral causality can be recognized between Chinese main coastal ports throughput (DFCO) and foreign trade (TRADE). Table 7: Granger-Causality Test Causality Test Port throughput to proxy variables 1 Statistics (Prob.) Proxy variables to port throughput 2 Statistics (Prob.) DF DFINCO RS 4.25(0.039) 0.025(0.87) 1 DINCO RS 4.24(0.039) 0.004(0.95) 1 FINCO TRADE 3.65(0.16) 3.65(0.009) 2 DFIN RS 5.78(0.055) 0.87(0.65) 2 DFIN INVE 5.21(0.0738) 0.756(0.685) 2 DFCO TRADE 5.41(0.0668) 5.109(0.0777) DFCO COVE 9.19(0.0024) 0.572(0.45) 1 Notes: 1 means the null hypothesis is port throughput does not granger cause the proxy variables; 2 means the null hypothesis is that proxy variables do not granger cause port throughout. Figures in (.) indicate the exact significance levels. 2 10

11 This result reveals that the foreign trade is the main driving force to Chinese ports throughput, especially the foreign throughput and coastal port throughput. The foreign throughput is counteractive to the foreign trade. The rise of ports throughput boosts domestic demand and lead to substantial port investment. 4.5 Co-integration tests with structural breaks We now consider the long run relationship between Chinese ports throughput and other economics variables by allowing a structure break indicating an exclusive impact on port. On the basis of Engle and Granger co-integration test and unit-root test allowing structure break suggested by Perron (1987), Gregory and Hansen (1996) propose three models for co-integration test allowing one endogenous structural (GH model). The break in GH model implies a maximal deviation between ports throughput and proxy variables in this study. The GH model has been broadly applied in many research fields. For example, Budget balance (Keho 2010), Public investment in agriculture (Lee and Hsu 2009), Fiscal Synchronization (Murat Aslan 2009), Energy consumption (Altinay and Karagol 2004), Fiscal policy (Hjelm and Johansson 2002), Japanese demand system (Ogura 2011), Saving and investment nexus for China (Narayan 2005) and so on. We control for our economic indices in the co-integration test models with break as: Model A(C): allow a change in the intercept which is represented as a level shift y t = c + + γ*du t + ε t Model C(C/S): it is called regime shift where changes in both intercept and slope coefficients are included y t = c + + γ*du t + * DU t +ε t Here, DU t is different from DU t in ZA models, DU t =1 when t=t b instead of t=t b -1. Castillomanzano (2008) used world maritime traffic and Spanish GDP as proxy economic variables to study the effect of legal structural reform on Spanish ports. We hence noticed and tested the international level economic variables such as world or regional maritime traffic, trade and Chinese GDP. Analysis shows that there are very large deviations between international or regional maritime traffic, economics indices such as regional trade and GDP, and all sorts of Chinese port throughputs. The possible reason is that China was isolated from the world economy prior to As an alternative, this paper controls for the domestic economic variables as proxy which have been used in Engle-Granger co-integration test. By doing this, the exclusive effect brought especially by events relevant to ports can be identified through comparing the results of Engle-Granger test and Gregory and Hansen test. The results of Gregory-Hansen test can be seen in Table 8. In contrast to the results of Engle and Granger test, in most cases the null hypothesis of no cointegration between variables cannot be rejected by GH model. The implication is that there is no overwhelming shock to throughputs brought by implementation of port policy or reform. In all pairs of variables, a co-integration relationship at 1% significance level is only found between foreign throughput of Chinese ports (FINCO) and foreign trade (TRADE) in both model A and model C. The break occurred in 1961 following the Great Leap Forward. A positive intercept coefficient (θ TRADE ) at 1% significance level in model A and a negative slope coefficient (γ TRADE ) in mode C imply that the foreign throughput of Chinese ports (FINCO) suffers much less than foreign trade during the Great Leap Forward. But in a long run from 1961 to now, it grows slower than the foreign trade. Throughput of Chinese inland ports (DFIN) has a co-integration 11

12 relationship with Retail Sales at 5% significance level if a break at 1963 is considered both in model A and model C. The reason is unclear. Measuring by Retail Sales and foreign trade simultaneously, the null hypothesis of no co-integration is rejected by throughput of Chinese main ports (DFINCO) at a 5% significance level in model C. The break year is The intercept coefficient (θ TRADE+RS ) is significantly positive at 1% significance level while the slope coefficient for both foreign trade and retails sales (γ TRADE and γ RS ) are negative and insignificant. This break coincides with Three-year Project in It suggests that the project has had a positive impact in raising ports throughput in short run but not in the long term. Table 8 Gregory and Hansen test for co-integration with structural break Model Break ADF α TRADE α RS α INCOVE γ θ x1 θ x2 k Critical Value (1%, 5%, 10%) DFINCO TRADE +RS m=2 A(C) * 1.32 *** *** C(C/S) ** *** *** DFINCO RS m=1 A(C) * *** *** C(C/S) *** DFINCO TRADE m=1 A(C) *** C(C/S) *** DFINCO INCOVE m=1 A(C) *** *** C(C/S) *** *** 0.34 *** DINCO RS m=1 A(C) *** *** C(C/S) *** DINCO INCOVE m=1 A(C) *** *** C(C/S) *** *** 0.33 *** FINCO TRADE m=1 A(C) *** 0.77 *** *** C(C/S) *** 1.25 *** * FINCO INCOVE m=1 A(C) *** 0.95 *** C(C/S) * DFIN RS m=1 A(C) ** *** *** C(C/S) ** *** *** -0.9 *** DFIN INVE m=1 A(C) *** 0.38 *** C(C/S) *** *** 0.33 *** DFCO TRADE m=1 A(C) *** C(C/S) *** DFCO COVE m=1 A(C) *** *** C(C/S) *** *** 0.37 *** Note: SIC is used as lag selection standard. Maxlag is selected according to t-sig method. Critical value are extracted from Table in Gregory and Hansen (1996, p. 109). 12

13 When Retail Sale (RS) is regarded as a benchmark for different corresponding ports throughputs in Gregory and Hansen test, the most significant break can be identified around 2002 (or 2001) for throughput of Chinese main ports (DFINCO) and domestic throughput of Chinese main ports (DINCO). The intercept coefficients (γ RS ) are significantly positive at 1% significance level in model A and the slope coefficients (θ RS ) are insignificantly positive in model C. The same break point was can be The most significant break is also found in 2002 in model C for the co-integration relationship between the trade and port throughputs such as throughput of Chinese main ports (DFINCO), throughput of Chinese coastal ports (DFCO). But the breaks of model C are in The intercept coefficients in model A and slope coefficients (θ TRADE ) are insignificantly positive in model C. These results illustrate that China accession to WTO (2001) resulted in exclusive positive impact on ports throughput compared to its impact on domestic demand and foreign trade. The Asian Financial Crisis (1997) had a greater negative impact on trade than its on port throughput. Thereafter the port throughput grew faster than the foreign trade. Qiu (2008) considered the 2001 as the beginning of third boom of Chinese port industry. This statement is proved to be valid according to our analysis results based on the historic data. Breaks between throughput of Chinese main ports (DFINCO), domestic throughput of Chinese main inland and costal ports (DINCO), throughput of Chinese main coastal ports (DFCO) and their corresponding ports investment (INCOVE, COVE) are observed as 1972 in model A and 1986 in model C. The coefficients of intercept (θ incove ) in model A are significantly negative at 1% significance level and the slope coefficients (γ incove ) in model C are significantly positive at 1% significance level. It shows the investment in ports brought by port structural reform is proved to be more efficient to ports development. 5. Conclusion Chinese ports throughput is largely influenced by foreign trade and retail sales (domestic demand) in a long run. There is also correlation between throughput and port investment. This seems to suggest that foreign trade impacts port throughput positively through foreign and coastal throughput which in turn increases domestic retail sales (domestic demand). Port investment lags behind port throughput. The multiple economic and reform events have an influence on maritime traffic. For example, the implementation of 3-year project to upgrade the Chinese coastal ports in 1972 can be regarded as a watershed in the history of port development. Prior to 1972, domestic inland ports accounted for almost all of throughput volume. The Great Leap Forward had a profound negative impact on the nation s economics indices in 1961 in particular domestic port throughput in a short term. Post 1972, the foreign throughput takes over the dominant amount over domestic port throughput. The largest impact to coastal port throughput was after accession to WTO in 2001 as foreign trade grew by leaps and bounds and this impact is exclusive to port throughput and trade. As such, the impact of changes in reform in of port policy was probably minimal due to the phenomenal growth in foreign trade. Perhaps The 3-year project to upgrade the Chinese coastal ports facilitated the increase in ports investment and the growth of investment was observed much faster than the port throughput. In 1986, two years after the first port reform was introduced, port throughput out grew port investment suggesting that the investment in ports and structural change in port governance led to higher efficiency. 13

14 In conclusion, the political events such as the Great Leap Forward and economic events like China s accession to WTO had greater impact on port throughput. Port investment and infrastructure and port structural reform merely facilitated port capacity if not acted as catalyst to increase port throughput performance. On the whole port reform did not play a major role. Improvements to the study can be summarized as follows: Methods for detecting unit root and co-integration test are available for two breaks ((Lumsdaine and Papell 1997), (Hatemi-J 2008)). Lumsdaine and Papell noted that the number of structural breaks is not definitive. Perhaps, we may need to explore further methodology that is able to identify numbers of breaks at various levels of significance. As ports are heterogeneous in terms of level of technology employed and management structures perhaps different functional forms should be employed to analyze port performance. Cargo types (bulk, container, oil) by different time periods can also provide finer analysis. 6. Reference Altinay, G. and E. Karagol (2004). "Structural break, unit root, and the causality between energy consumption and GDP in Turkey." Energy Economics 26(6): China Communication Press (1999). Transportation Statistical Data and Materials on 50 Years of New China:1949~ 1999, Beijing, China. China Ocean Press (2010). China Marine Statistical Yearbook 2009, Beijing, China. China's Ministry of Transport (2010). China Shipping Development Report Beijing, China National Bureau of Statistics of China (2010). China Statistical Yearbook Beijing, China. China's Ministry of Transport (2001 ~ 2010). Bulletin of statistical analysis of highway and waterway development in year 2001 ~ 2010, Beijing, China. China Statistics Press (2010). China Trade and External Economic Statistical Yearbook Beijing, China. China Statistics Press(1999). Comprehensive Statistical Data and Materials on 50 Years of New China, Beijing, China. Cullinane, K. and T. Wang (2006). "Chapter 15 Port Governance in China." Research in Transportation Economics 17: Dickey, D. A., Fuller, W. A. (1979). "<Distribution of the Estimators for Autoregressive Time Series with a Unit Root.pdf>." Journal of American Statistical Association 74(366): 54. Estache, A. (2004). "Sources of efficiency gains in port reform: a DEA decomposition of a Malmquist TFP index for Mexico." Utilities Policy 12(4): Everett, S. (2007). "Port reform in Australia: regulation constraints on efficiency." Maritime Policy & Management 34(2): Everett, S. and R. Robinson (1998). "Port reform in Australia: issues in the ownership debate." Maritime Policy & Management 25(1): Granger, C. W. J. (1988). "Some recent development in a concept of causality." Journal of Econometrics 39(1-2): Gregory, A. W. and B. E. Hansen (1996). "Residual-based tests for cointegration in models with regime shifts." Journal of Econometrics 70(1): Hadi Baaj, M. and J. Issa (2001). "Institutional reform in ports of developing countries: the case of Lebanon part I: the plan." Maritime Policy & Management 28(4):

15 Hall, A. R. (1994). "Testing for a Unit Root in Time Series with Pretest Data-Based Model Selection." Journal of Business & Economic Statistics 12(4): Hatemi-J, P. A. (2008). "Tests for cointegration with two unknown regime shifts with an application to financial market integration." Empirical Economics 35(3): Hjelm, G. and M. W. Johansson (2002). Structural Change in Fiscal Policy and The Permanence of Fiscal Contractions - The Case of Denmark and Ireland, Lund University, Department of Economics. Ircha, M. C. (1997). "Reforming Canadian ports." Maritime Policy & Management 24(2): Ircha, M. C. (2001). "Port strategic planning: Canadian port reform." Maritime Policy & Management 28(2): James, G. M. (2010). Critical Values for Cointegration Tests, Queen's University, Department of Economics. Johansen, S. (1988). "Statistical analysis of cointegration vectors." Journal of Economic Dynamics and Control 12(2-3): Keho, Y. (2010). "<Budget balance through revenue or spending adjustment Econometric analysis of the Ivorian budgetary process, pdf>." Journal of Economics and International Finance 2(1): Lee, C.-C. and Y.-C. Hsu (2009). "Endogenous structural breaks, public investment in agriculture and agricultural land productivity in Taiwan." Applied Economics 41(1): Li, X.-M. (2000). "The Great Leap Forward, Economic Reforms, and the Unit Root Hypothesis: Testing for Breaking Trend Functions in China's GDP Data." Journal of Comparative Economics 28(4): Lumsdaine, R. and D. Papell (1997). "Multiple Trend Breaks And The Unit-Root Hypothesis." The Review of Economics and Statistics 79(2): Murat Aslan, M. T. (2009). "Is Fiscal Synchronization Hypothesis Relevant for Turkey? Evidence from Cointegration and Causality Tests with Endogenous Structural Breaks." Journal of Money, Investment and Banking(12): Narayan, P. K. (2005). "The saving and investment nexus for China: evidence from cointegration tests." Applied Economics 37(17): Ogura, M. (2011). "Testing for structural break in Japanese demand system after the bubble era." Structural Change and Economic Dynamics. Pallis, A. A., T. K. Vitsounis, et al. (2010). "Port Economics, Policy and Management: Review of an Emerging Research Field." Transport Reviews 30(1): Perron, P. (1989). "The Great Crash, the Oil Price Shock, and the Unit Root Hypothesis." Econometrica 57(6): Qiu, M. (2008). "Coastal port reform in China." Maritime Policy & Management 35(2): Serebrisky, T. and L. Trujillo (2005). "An assessment of port reform in Argentina: outcomes and challenges ahead." Maritime Policy & Management 32(3): Smyth, R. (2004). "Is Chinese provincial real GDP per capita nonstationary? Evidence from multiple trend break unit root tests." China Economic Review 15(1): Zivot, E. and D. W. K. Andrews (1992). "<Further evidence on the great crash, the oil price shock, and the unit root hypothesis.pdf>." Journal of Business & Economic Statistics 10(3): 19. ` 15