44353 EASTR WORKING PAPER No. 14 Transport Sector Unit, Infrastructure Department East Asia and Pacific Region June 2007 THE WORLD BANK Domestic Trade Impacts of the Expansion of the National Expressway Network in China DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA The reports in this series are available for download from www.worldbank.org/eaptransport EASTR Working Paper 1: Roads Improvement for Poverty Alleviation in China (2000) EASTR Working Paper 2: Private Participation in Infrastructure in China ­ Issues and Recommendations for the Road, Water and Power Sectors (2003) EASTR Working Paper 3: Trade and Logistics in East Asia ­ A Development Agenda (2003) EASTR Working Paper 4: China: Building Institutions for Sustainable Urban Transport (2006) EASTR Working Paper 5: Timor-Leste Transport Sector ­ Outline of Priorities and Proposed Sector Investment Program (2005) EASTR Working Paper 6: China: Managing the Economic Interfaces in Multi- Operator Railway Environments (2006) EASTR Working Paper 7: Vietnam: Logistics Development, Trade Facilitation & the Impact on Poverty Reduction (2003) EASTR Working Paper 8: East Asia Ports in their Urban Context (2003) EASTR Working Paper 9: Private Participation in Infrastructure in China ­ Issues and Recommendations for the Road, Water and Power Sectors (2003) EASTR Working Paper 10: Timor-Leste Transport Sector ­ Outline of Priorities and Proposed Sector Investment Program (2005) EASTR Working Paper 11: China: Building Institutions for Sustainable Urban Transport (2006) EASTR Working Paper 12: China: Managing the Economic Interfaces in Multi- Operator Railways Environments (2006) EASTR Working Paper 13: China's Expressways: Connecting People and Markets for Equitable Development (2007) ­ In Review EASTR Working Paper 14: Domestic Trade Impacts of the Expansion of the National Expressway Network in China (2007) 26 June 2007 Domestic Trade Impacts of the Expansion of the National Expressway Network in China EASTR WORKING PAPER No. 14 Transport Sector Unit, Infrastructure Department East Asia and Pacific Region June 2007 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA The World Bank East Asia Transport Unit 1818 H Street NW Washington, D.C. 20433, U.S.A. Tel: (202) 458-1876 Fax: (202) 522-3573 Website: www.worldbank.org A publication of the World Bank East-Asia Transport Unit. This report is a product of the staff of the World Bank assisted by independent consultants. The findings, interpretations, and conclusions expressed herein do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of the World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. 26 June 2007 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Foreword In 2006 the World Bank prepared a report on the experience of developing its 41,000 km expressway network. This report found that there had been major economic and social benefits from the enterprise. On the basis of subsequent discussions with Chinese Government counterparts, the Bank decided to try and quantify the impact of the expressway network development on internal trade using a spatial analysis technique the Bank had developed in Africa. The results of this analysis showed that there were significant travel time savings from the network, and a major increase in interprovincial trade. The World Bank team is grateful to the Chinese Government, particularly to the staff at the MOC, for their contributions and support in the preparation of this paper. 26 June 2007 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Acronyms MOC Ministry of Communications NEN National Expressway Network NTHS National Trunk Highway System 26 June 2007 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Contents 1 Introduction ................................................................................. 1 2 Objectives and Overall Framework............................................... 1 3 Overview of the Methodology....................................................... 2 4 Data Set Construction................................................................... 4 5 Network Modeling and Estimations .............................................. 7 6 Caveats, Possible Improvements and Extensions ....................... 10 7 References ................................................................................. 13 26 June 2007 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA 26 June 2007 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA 1 Introduction 1. China's outstanding achievements in economic growth and poverty reduction over the last fifteen years have been well documented. A major element of that growth consisted of the development of its infrastructure, particularly transport. All modes of transport have seen their networks expanded, to provide the infrastructure needed to support the broader development goals. Among the surface modes (excluding pipelines or waterways), road transport has seen its share grow from 45% to 60% in terms of passenger-km and from 24% to 30% in terms of freight ton-km. 2. From 1990 to 2005, during the period of the 8th, 9th and 10th Five-Year Plans, China completed nearly 41,000 km of high-grade tolled expressways comprising the National Trunk Highway System (NTHS), or as it is now called, the National Expressway Network (NEN). During this period approximately 400,000 km of local and township roads were also improved. This was achieved by investing upwards of US$40 billion per year, with about one third of that amount allocated to development of the NEN. 3. China's expressway network has been planned following an approach that emphasized the objective of connecting all major cities to one another. The process defined the most important cities to be served and the most efficient network to connect them. The current NEN plan has been built upon the previous 1992 "5- vertical 7-horizontal" Expressway Trunk Network and now responds to the so-called 7-9-18 plan. This includes seven corridors radiating from Beijing, nine north-south corridors, and 18 east-west corridors. The strategy for defining the 7-9-18 plan consisted of combining radial and grid patterns in an attempt to maximize coverage and transport connectivity. 4. Behind this overall strategy, the expressway plan seeks to connect all cities with more than 200,000 people, serving as facilitator of economic and social interactions as the economy comes to rely more and more on road transport. In prioritizing the selection of cities (nodes) to be connected, the planning process has incorporated economic and transport objectives (including trade and container traffic requirements, and tourism needs), giving special consideration to poorer regions and environmental issues (e.g., avoiding environmentally sensitive areas). This will improve the regional integration of the economy and allow growth dynamics to expand from the coastal regions to the interior and western parts of the country.1 2 Objectives and Overall Framework 5. The analytical work presented here seeks to contribute to the quantification of potential benefits from network expansion. The analysis is set in the framework of the "new" economic geography models that see location of economic activities and network linkages between centers of production and consumption as important determinants of economic growth (e.g., Fujita et al. 1999). 1Wen (2001), Fujita and Hu (2001), Zhang and Kanbur (2001), Luo (2004), Catin et al. (2005), among others, discuss location of economic activities in China. 1 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA 6. Due to data limitations, estimation results presented here are likely to have a lower level of accuracy than results from studies based on international trade data. We believe that the main trends and spatial patterns that we identified are quite robust. But more accurate prediction of the magnitudes of trade expansion and other benefits would require improvements in input data as well as in the imputation of missing or incomplete information. Our main objective is therefore to outline an analytic approach that links transport sector work and economic analysis within a consistent, spatially explicit framework. We suggest several possible extensions and improvements at the end of this report that could be implemented in close cooperation with partners in Chinese transport, economics and statistical institutions. 7. Our analysis focuses on an estimation of trade expansion between Chinese cities and provinces following the construction of new expressways or upgrading of existing ones. It follows the general approach developed in Buys et al. (2006) for Africa and Shepherd and Wilson (2006) for Eastern Europe and Central Asia (ECA). For Africa, for instance, this work suggests that upgrading of the major highway system between the mainland sub-Saharan African countries could essentially triple the trade volume among these trading partners. For the ECA region, where infrastructure is already of higher quality, analysis suggests a 50 percent increase in trade. While trade volumes are not a direct measure of welfare, research by Frankel and Romer (1999), among others, suggests that each additional dollar of international trade increases incomes by between 0.5 and 2 dollars. Benefits of domestic trade may well fall into a similar range. 3 Overview of the Methodology 8. The methodology underlying the transport and trade expansion analysis in this and the previously mentioned applications is conceptually straightforward. Trade is greatly facilitated by good transport infrastructure. Our analysis seeks to establish an empirical relationship between trade flows among economic centers (major cities) and the quality of infrastructure while controlling for other factors such as the economic size of exporting and importing centers. The estimated parameters that describe this relationship can then be used in a policy simulation by predicting trade flows after improvements in infrastructure while all other factors are held constant.2 9. Transport infrastructure is modeled spatially explicitly by using geographic information systems (GIS) data of road networks. Standard spatial analysis techniques allow us to determine the most likely routes through the network along which trade will flow. This allows a decomposition of aggregate trade flows and allocation to specific network links. In an ex-ante policy application, the links with the highest predicted flows are likely candidates for priority upgrading. 10. There are two main innovations of our work compared to the Africa and ECA applications. Firstly, the analysis for China generates estimates of trade expansion at the sub-national level rather than between countries. This complicates matters considerably because very few countries collect data on internal trade flows. 2While past applications have focused on the quality of infrastructure exclusively, the framework would allow for consideration of other factors influencing trade flows such as border delays or quality of transport services. If data on these factors were consistently available, the relative impacts of improvements in each factor could be determined. 2 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA But a credible analytic framework at the subnational level would open up tremendous opportunities for client country engagement in the transport sector. Secondly, the applications for Africa and ECA based the main impact evaluation on improvements in an imputed, general quality measure of road network links that is assumed constant within each country. This was necessary, because detailed road quality information is not consistently available across countries. The analysis presented here, in contrast, measures the impact of improvements in travel times between economic centers. These are calculated using spatially referenced road network data sets that have estimates of design speeds for each road link and present the status of the Chinese road system before and after implementation of the NEN program. 11. The methodology employed in our analysis builds upon the estimation of a gravity model of trade between provinces in China. This model, in analogy to the theory of gravity in physics, states that the trade volume between two trading partners is proportional to their economic size, and inversely proportional to the distance of separation between them.3 Therefore we expect two bigger provinces to trade more than two smaller ones, and provinces that are located in close proximity are likely to trade more than those that are further apart. The measure of an economy's size is usually GDP or gross regional product. 12. Mathematically, this yields the following formulation of the gravity trade model: Ei M j i j (1) Tij = k dij where Tij = Trade volume between provinces i and j. Ei = Economic scale of the exporting province i. Mj = Economic scale of the importing province j. dij = Distance or travel time between provinces i and j. k = A scaling parameter. 13. Many extensions of the standard gravity model have been proposed. These incorporate additional information that may influence trade between two provinces such as whether trading partners share a common border or common language, special trade agreements or general measures of remoteness.4 14. In practice, straight line and even network distance is an incomplete measure of "friction" between two trading partners, since it ignores the quality of transport infrastructure that facilitates or hinders trade. Network quality can be incorporated in the trade model by adding a separate term that measures road quality for each network link. Our analysis instead uses GIS computed travel times between trading partners as the measure of separation, which is an aggregate measure of both, distance and road quality. 3There is a large literature on gravity models in international trade analysis. Anderson and van Wintrop 2003 & 2004 provide recent overviews. In geography and regional science an overview is provided by Fotheringham and O'Kelley (1989). 4See Frankel (1997), Rose (2000), Soloaga and Winters (2001), Carillo and Li (2002), and Coulibaly and Fontagné (2004) for recent examples. All of these analyze international trade. 3 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA 15. The initial gravity model parameters are estimated using trade flows for 1997 between Chinese provinces and travel times computed using the road network representation that does not include expressways. After applying a log transformation and retaining dij as the measure of travel time, we estimate the following regression: (2) logTij = k0 + i log Ei + i log M + log dij + ij j 16. Although this model is estimated using province level data, we are ultimately interested in specific trade flows between major cities through the road network. We therefore use the estimated parameters from the province level model together with estimated city level GDP and travel times between cities to predict baseline flows between all major cities (all cities with a population greater than 500,000 in 2005 plus provincial capitals).5 17. The main policy simulation then assesses how trade flows would increase with a reduction in travel times due to the implementation of the expressway program. Holding all other factors constant (i.e., GDP at 1997 levels), we replace base network travel times with those derived from a GIS road network that includes new and upgraded expressways which allow faster travel. The difference between the predicted before and after city-to-city flows then represent the trade expansion that is due to road network improvements. The results likely represent a lower bound estimate since they do not include intra-provincial trade increases or larger induced trade with the rest of the world due to improved access to national export/import hubs.6 18. Over the notional analysis period of 1997 ­ 2005, China's GDP roughly doubled from 953 billion to 1.9 trillion in constant 2000 USD. This growth in the economy will also have an impact on domestic trade expansion independent from the transport improvements. We estimate the potential magnitude of total trade expansion by running a third prediction of trade flows using travel times from the complete road system including expressways and estimated city level GDP for 2005. 19. Since we model trade flows explicitly using a spatial network approach, we can spatially allocate predicted flows to network links. The resulting data and maps show where within the Chinese road network we will expect the largest changes in transport flows. 4 Data Set Construction 20. The analysis presented here requires four main sets of data: GIS referenced road networks; regional and city level GDP; trade flows; and information on city location and population. Data sources and conventions used in data imputations are described below. 5Although the NEN aims to connect all cities above 200,000 in China, we use a larger threshold to reduce computational requirements. 6We only consider trade between cities in different provinces since this allows us to constrain city level trade flow predictions to levels derived in the initial provincial gravity model predictions. 4 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA 21. GIS road network data: Georeferenced road information for China was obtained from the Australian Consortium for the Asian Spatial Information and Analysis Network (ACASIAN; www.asian.gu.edu.au). The base network consists of 20,899 line segments with attribute information indicating the type of road represented by each link. After including the expressway network, the complete database has 31,538 segments. Although the data were of high quality, some additional work was required to make sure that the network topology (road segments connected at intersections by nodes) was correct. For each road type we determined a suitable travel speed (design speed) ranging from 10 km/h for unpaved city streets to 75 km/h for expressways (Table 1). Given the GIS calculated real world length of each segment, we computed travel times to traverse each road link which is the basis for computing the fastest inter-city network routes. Figures 1 and 2 show the network without and with expressways. 22. City data: Although the NEN aims to connect all cities above 200,000 population, we chose a higher threshold of 500,000 to reduce computational complexity. The threshold could be lowered in future applications. Unfortunately, a relatively straightforward task such as identifying population for cities in China is complicated by the lack of standard definitions for cities. For instance, one city (or shi) is listed in the Statistical Yearbook as having a population of about 30 million while other sources list the city proper population as about 4 million. Clearly, the larger figure must refer to the entire administrative unit. We utilized census data for 1990 and 2000 as well as estimates for 2007 from CIESIN (2004) and from World Gazetteer7 and interpolated 1997 and 2005 population (see Table 2). 23. GDP: Province level data on GDP (actually Gross Regional Product) are available from the annual statistical yearbooks.8 We adjusted nominal GDP to 2005 levels using the annual GDP deflators from the World Bank Indicators database. Province level data are presented in Table 3. For predicting city-to-city flows, we also require GDP estimates by city. For computational simplicity and to make it easier to enforce basic accounting constraints, we assume that all trade originates or terminates in one of the major cities. We therefore also assume that all GDP is produced in these cities such that city-level GDP is simply the population-equivalent share of total provincial GDP (Table 2). 24. Inter-provincial trade flows: Like many countries, including many industrialized nations, China does not collect consistent data on domestic trade flows. Information is available on freight-ton km transported by railways between provinces, but not on the value or volume of trade on roads. The only consistent provincial trade flow information is derived from level input-output (IO) tables that most of the Chinese provinces produce occasionally (1992, 1997 and 2002). Total domestic inflows and outflows by economic sector can be derived from the final demand columns of these IO tables, complemented by customs information that accounts for international trade.9 Such data were generated and analyzed for 1992 by Naughton (2000) and for 1987, 1992 and 1997 by Poncet (2003, 2005). 7 www.gazetteer.de 8 National Bureau of Statistics (NBS) of China: www.stats.gov.cn/english/statisticaldata/yearlydata 9 Ministry of Foreign Trade (MOFTEC), Almanac of China's Foreign Economic Relations and Trade. 5 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Figure 1: Road network (baseline) Figure 2: Road network with expressways 6 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA 25. Sandra Poncet of the Université Paris 1 kindly provided her 1997 data set of total domestic trade between each province and all other provinces in 22 economic sectors. Some of these sectors, such as mining, are unlikely to ship many of their products by road. Unfortunately, we do not have precise information on the proportion of products in each sector that will be moved through the road network. For now, we therefore determine proportions ad hoc (Table 4). These could be adjusted in future applications using information available in the country (transport statistics or at least expert opinion). To make trade flows comparable over time and compatible with the GDP data, the resulting aggregate trade flows were adjusted to 2005 levels using GDP deflators. The province level domestic inflows and outflows are given in Table 3. 26. Trade estimates were available for 24 of the 31 provinces in China. Heilongjiang, Anhui, Shandong, Hainan, Chongqing, Guizhou and Tibet did not generate IO tables in 1997. This provides a complication for the trade flow estimation since the reporting provinces contain imports and exports from and to the non-reporting provinces. We therefore had to impute trade values for the missing provinces, which in all but two cases relied on a simple regression of trade on a number of explanatory variables.10 For the very small economies of Hainan and Tibet, the predictions yielded negative estimates, so we instead assumed the same proportion of trade to GDP as the average of the three reporting provinces that were closest in size of their economies. 27. Rather than aggregate trade data--i.e., the marginal row and column totals of a complete trade flow matrix--we require the actual province-to-province flows in the estimation of the gravity model and subsequent predictions. We therefore implemented an intermediate step that uses a gravity type spatial interaction model to generate a flow table that is consistent with the marginal totals (see Fotheringham and O'Kelly 1989). This is essentially a balancing algorithm that iteratively adjusts flows to match row/column totals until some convergence criteria is reached. As in the standard gravity model, flows are assumed to be a function of distance (travel time) of separation and are influenced by the size of the origin and destination regions. Since the estimation parameters need to be given, their choice will obviously determine the results from the gravity model. Empirically, final gravity model results are much more sensitive to the choice of the friction parameter compared to the parameters for economic size. We therefore fine tuned the interaction model to yield a gravity model parameter of about 2 as in the classic formulation that has been found to hold in numerous international and domestic trade studies. 5 Network Modeling and Estimations 28. We compute the fastest routes through the network between each city pair using a standard shortest-path (Dijkstra) algorithm. A simple initial measure of the importance of each road network link is the number of times each link is used in the total of 6328 connections. Figures 3 and 4 show the resulting maps for the 10Of the explanatory variables used in these regressions only total GDP was significant in predicting trade flows. The share of industry in GDP, the share of business services, population, and the GDP weighted travel time to all other regions did not contribute significantly to the predictions. 7 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA network without and with expressways. The major network arteries in the eastern regions of China appear prominently in these maps. 29. Summarizing network travel times by provinces11, the average estimated travel time drops from 38.2 hours in the base network to 28.6 hours in the network that includes expressways (i.e., by 9.6 hours or 25 percent; the median reduction is 8.5 hours). The maximum travel time decreases from 104.0 to 76.2 hours. These travel times are GIS calculated and assume non-stop travel at maximum speed. Actual times will be somewhat higher. 30. The gravity model parameters estimated in a lognormal regression on the imputed trade flows, GDP for importing and exporting province, and baseline travel times are as follows: (3) lnT^ij = -5.128 + 0.741ln Ei + 0.8735ln M - 2.016ln dij j 31. The model is estimated using standard OLS (Table 5). The model fit, with an adjusted R2 of 0.42 is acceptable by most econometric standards but lower than is common in international trade studies. The estimated model forms the basis of prediction of trade flows for the 6328 city pairs. This actually involves 12,656 calculations since each city is both an importer and exporter and the parameters for importer and exporter GDP are not identical. We first predict base line flows and subsequently flows in a system with much improved travel times. 32. Prediction is straightforward by entering observed values into equation (3) and taking exponentials to obtain estimated trade flows. However, lognormal prediction can be suspect when predicting very large or very small values (Lyles et al. 1997). We therefore apply an additive correction factor of half the variance (squared RMSE), 2, which in this model has a value of 4.74.12 Trade flows between cities in a given pair of provinces are constrained to equal the predicted aggregate flows between the two provinces. The resulting city level total trade flows are somewhat higher than the "observed" flows. However, rather than using constrained estimation, we use these flows as the baseline since we are only interested in relative changes and because we would not know the appropriate adjustment for the prediction with new travel times. 11We compute the average of travel times between cities in each pair of provinces. 12Thanks to Tim Thomas (DECRG) for pointing out the need for a correction and providing the reference. 8 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Figure 3: Most significant network links (baseline) Figure 4: Most significant network links (baseline plus expressways) 9 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA 33. The resulting predictions suggest an increase of trade flows after implementation of the expressway network from 4,814 to 10,619 billion Yuan or 120.6 percent. These total effects could be further disaggregated to yield, for instance, regional trade expansion estimates. The more than doubling of trade following expressway construction is based on prediction using 1997 level GDP for both before and after completion of the NEN. However, during this period, GDP also increased significantly which should have a further effect on trade. To provide an indication of the effect of economic growth on the estimates, we run a further prediction with new travel times as well as 2005 GDP values for importer and exporter cities. The resulting aggregate predicted trade flows of 39,839 billion Yuan are 728 percent of base line flows. This would suggest that economic growth has a higher impact on trade expansion than improvements in transport infrastructure. This may well be the case. However, economic growth as reflected in rising GDP figures is itself influenced by the quality of infrastructure, so that it becomes very difficult to attribute relative impacts to one or the other. 34. For each trade flow between a city pair we know the fastest route through the network on which road transport is likely to occur. We can therefore allocate each predicted flow to the network links that connect the two cities. Figures 5 and 6 show maps of the base line flows and those after NEN completion. Far fewer routes are used in the latter as the higher quality expressways "pull" traffic away from smaller roads. While this has likely benefits for efficiency of the economic system, for many smaller cities and regions this may mean that they are no longer fully integrated into the national trading network. 6 Caveats, Possible Improvements and Extensions 35. The analysis presented here illustrates a general methodology for estimating potential trade expansion that can be attributed to road construction or upgrading. Because of data limitations, the empirical results are likely to be less robust than those from similar studies of international trade. The biggest data gap is the lack of reliable inter-provincial trade figures. It is unlikely that better information on trade flows is available in China, but the imputed estimates from IO table final demand columns could probably be improved with more sophisticated spatial allocation models augmented by other domestic trade related information. An obvious first improvement would be better estimates of the proportion of trade flows from and to various economic sectors which were specified in a fairly ad hoc manner. 36. Such improvements would result in better estimates of inter-province flows that would in turn yield more reliable gravity model results. Our study uses a very simple gravity model formulation that only includes importer and exporter GDP and travel time as a measure of trade friction. Better data would justify using an extended model formulation that could include neighborhood effects, coastal location (export orientation), generalized measures of remoteness, and so on. These additional independent variables should improve model fit and therefore reduce the standard error in predictions. 10 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Figure 5: Predicted baseline trade flows on the road network (100 million yuan) Figure 6: Predicted trade flows after expressway expansion (100 million yuan) 11 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA 37. Our study used 1997 as the baseline from which predictions of future trade flows are derived. In addition, IO tables were also produced by a number of provinces for 1992 and 2002, and will presumably be constructed for 2007. While trade flows have already been derived for 1992 by Naughton and Poncet, this has not yet been done for 2002. Such data would provide a good means of cross- checking predictions of trade induced by both transport infrastructure investments and other agents of economic growth. 38. A further extension of the model would be to include intra-provincial trade (i.e., trade among cities in the same province) and international trade which has grown significantly in recent years. Aggregate data on international imports and exports from provinces exists as does information on export flows through the main ports. These and other export hubs could thus be treated as additional demand and supply centers similar to the cities in the present application. Including intra-province and international trade should greatly increase the trade expansion benefits of improved transport infrastructure. 39. A more ambitious extension to the current model would be to generate a spatially explicit multi-modal model of the Chinese transportation system. GIS data on rail lines and waterways exist and some statistics for these transport modes seem to be better developed than for road transport. Network modeling would be more complicated because mode switching would not be feasible at every intersection of, for example, road and rail lines. But tools exist in the field of transport GIS to address these issues. 40. Besides trade expansion benefits, policy makers are interested in the welfare benefits of infrastructure investments such as reduction in poverty rates. Trade itself induces economic growth which tends to raise incomes across welfare groups. Solid empirical analysis that quantifies the direct impact of infrastructure investment on welfare, however, is quite rare. This is largely because of the unknown time lag between investments and benefits, the difficulties in isolating infrastructure effects from other dynamics, and problems in determining the direction of causality (e.g., do better roads lead to successful local economies, or are better roads typically placed in high growth areas?). Most studies therefore use aggregate state or province level information and econometric specifications that relate aggregate infrastructure stock or investment to economic growth or welfare changes (e.g., Fan and Chan-Kang, 2005, for rural roads in China; Luo, 2004, for western provinces in China; and Hulten et al., 2006, for states in India). 41. For good identification, one would ideally have detailed information about the timing and placement of road investments, as well as time series data on outcome measures. The most direct outcome measures would be changes in poverty rates, although the mechanisms by which major highways affect poverty reduction may be very indirect ­ i.e., through economic growth, increased employment and better social investments due to increase in local government revenue. 42. Alternatively, one could limit estimation of impacts to economic performance of a region which could be measured by output, productivity growth or reductions in input costs ­ the last of these is perhaps most directly related to better transport infrastructure. Such information may be available from surveys at the individual firm level or in aggregate form at the county or province level. The finer the spatial disaggregation, the better. Detailed specification of the model will depend on data availability. Options include panel data models of the change in 12 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA outcome measures as a function of road investments (or related measures such as travel time to key input source or demand centers) and controlling for other factors. Alternatively one could model the effect of improved accessibility on firm productivity (e.g., Lall et al. 2004) or location choice by firms (e.g., Deichmann et al. 2004). 7 References Anderson, J.E. and E. Van Wincoop, (2003), Gravity with Gravitas: A Solution to the Border Puzzle, American Economic Review, 93, 1:170-192. Anderson, J.E. and E. Van Wincoop (2004), Trade Costs, Journal of Economic Literature, 42, 3:691-751. Buys, P., U. Deichmann and D. Wheeler (2006), Road network upgrading and overland trade expansion in sub-Saharan Africa, Development Research Group, World Bank, Washington D.C. Catin, M., X. Luo and C. Van Huffel (2005), Openness, industrialization and geographic concentration of activities in China, Policy Research Working Paper # 3706, Development Research Group, World Bank, Washington D.C. Carrillo, Carlos and Carmen Li, 2002, "Trade Blocks and the Gravity Model: Evidence from Latin American Countries," University of Essex, Discussion Paper. CIESIN (2004) Global Rural-Urban Mapping Project (GRUMP), Palisades, NY: Center for International Earth Science Information Network, Columbia University (CIESIN); International Food Policy Research Institute (IFPRI), the World Bank; and Centro Internacional de Agricultura Tropical (CIAT). Available at http://sedac.ciesin.columbia.edu/gpw. Coulibaly, Souleymane and Lionel Fontagné, 2004, "South ­ South Trade: Geography Matters," CEPII Working Paper No. 2004-08, CEPII Research Center (forthcoming, Journal of African Economies ­ access online: October, 2005). Deichmann, U., M. Fay, J. Koo and S. Lall (2004), Economic structure, productivity, and infrastructure quality in Southern Mexico, Annals of Regional Science, 38, 361-385; also World Bank DECRG Policy Research Working Paper No. 2900, 2002. Fan, S. and C. Chan-Kang (2005), Road development, economic growth, and poverty reduction in China, International Food Policy Institute, Washington, D.C. Fotheringham, A.S. and M.E. O'Kelly (1989), Spatial interaction models: Formulations and Applications, Kluwer, Dordrecht. Frankel, Jeffrey A., 1997, Regional Trading Blocs in the World Trading System, Institute for International Economics: Washington D.C. Frankel, J.A. and D. Romer (1999) Does trade cause growth? American Economic Review, 89, 3:379-399. 13 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Fujita, M., P. Krugman and A.J. Venables (1999), The spatial economy. Cities, regions and international trade, MIT Press, Cambridge, MA. Fujita, M. and D. Hu (2001). Regional disparity in China 1985-1994: The effects of globalization and economic liberalization. Annals of Regional Science, 35, 3­ 37. Hulten C.R., E. Bennathan, and S. Srinivasan (2006), Infrastructure, Externalities, and Economic Development: A Study of the Indian Manufacturing Industry, World Bank Economic Review, 20, 2:291­308. Lall, S.V., Z. Shalizi and U. Deichmann (2004), Agglomeration economies and productivity in Indian industry, Journal of Development Economics, 73, 643- 673. Luo, X. (2004), The Role of Infrastructure Investment Location in China's Western Development, Policy Research Working Paper # 3345, Development Research Group, World Bank, Washington D.C. Lyles, R.H., L.L. Kupper and S.M. Rappaport (1997), On prediction of lognormal- scale mean exposure levels in epidemiologic studies, Journal of Agricultural, Biological and Environmental Statistics, 2, 4:417-439. Naughton, B. (2000), How much can regional integration do to unify China's markets?, Center for Research on Economic development and Policy Reform, Working Paper No. 58, Stanford University, Palo Alto, CA. Poncet, S. (2003), Measuring Chinese domestic and international integration, China Economic Review, 14, 1-21. Poncet, S. (2005), A fragmented China: Measure and determinants of Chinese domestic market integration, Review of International Econommics, 13, 3: 409-430. Shepherd, B. and J.S. Wilson (2006), Road Infrastructure in Europe and Central Asia: Does Network Quality Affect Trade?, Development Research Group, World Bank, Washington D.C. Soloaga, Isidro and L. Alan Winters, 2001, "Regionalism in the Nineties: What Effect on Trade?" North American Journal of Economics and Finance 12, 1-29. Wen, M. (2001). Relocation and Agglomeration of Chinese Industry. http://rspas.anu.edu.au/economics/publish/papers/wp2001/2001- 07%20MeiWenWP2v.pdf Zhang, X. and Kanbur, R. (2001). What difference do polarisation measures make? An application to China. Journal of Development Studies. 37: 85-98. 14 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Table 1: Road types and assumed feasible travel speed Road type Pavement Assumed type speed (km/h) City street paved 10 Local road unpaved 10 paved 15 Motorway unpaved 20 paved 30 National highway unpaved 35 paved 50 Provincial highway unpaved 35 paved 50 Expressway paved 75 15 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Table 2: Population and GDP of cities above 500,000 population and provincial capitals Estimated Population Estimated GDP City Province 1997 2005 1997 2005 Bengbu Anhui 489,219 571,369 389 682 Hefei Anhui 1,237,797 1,388,815 985 1,658 Huaibei Anhui 793,225 926,424 631 1,106 Huainan Anhui 928,354 1,084,244 739 1,295 Maanshan Anhui 454,407 530,711 362 634 Beijing Beijing 6,457,008 7,477,178 2,106 6,886 Chongqing Chongqing 2,740,138 3,966,978 1,571 3,070 Fuzhou Fujian 1,009,472 1,178,983 1,736 3,266 Xiamen Fujian 500,021 583,985 860 1,618 Zhangzhou Fujian 520,572 607,987 895 1,684 Lanzhou Gansu 1,294,484 1,415,888 909 1,934 Foshan Guangdong 578,499 675,641 539 1,463 Guangzhou Guangdong 3,067,815 3,144,812 2,861 6,809 Jiangmen Guangdong 474,414 554,078 442 1,200 Maoming Guangdong 450,628 526,297 420 1,139 Rongcheng Guangdong 1,002,126 1,170,404 935 2,534 Shantou Guangdong 1,048,993 1,334,455 978 2,889 Shaoguan Guangdong 565,050 659,934 527 1,429 Shenzhen Guangdong 944,981 1,103,663 881 2,390 Zhanjiang Guangdong 559,932 653,956 522 1,416 Zhuhai Guangdong 434,305 507,233 405 1,098 Guilin Guangxi 569,796 665,477 711 1,235 Liuzhou Guangxi 644,465 752,684 804 1,397 Nanning Guangxi 665,512 777,265 830 1,443 Guiyang Guizhou 985,463 1,150,942 923 1,979 Haikou Hainan 533,148 622,674 477 895 Baoding Hebei 874,778 1,021,671 488 1,040 Cangzhou Hebei 463,741 541,613 259 552 Handan Hebei 1,138,110 1,328,459 635 1,353 Langfang Hebei 603,374 704,693 336 718 Luancheng Hebei 500,324 584,339 279 595 Qinhuangdao Hebei 655,810 765,934 366 780 Shijiazhuang Hebei 1,356,295 1,993,532 756 2,030 Tangshan Hebei 1,479,055 1,596,156 825 1,626 Xingtai Hebei 551,628 644,258 308 656 Zhangjiakou Hebei 627,406 732,761 350 746 Harbin Heilongjiang 2,813,777 3,229,209 1,699 2,948 Hegang Heilongjiang 629,962 735,746 380 672 Jiamusi Heilongjiang 461,291 538,751 279 492 Mudanjiang Heilongjiang 565,297 660,222 341 603 Qiqihar Heilongjiang 747,916 873,506 452 797 Anyang Henan 661,867 773,008 394 894 Jiaozuo Henan 440,587 514,570 262 595 Kaifeng Henan 483,105 564,229 288 652 Luoyang Henan 1,327,737 1,389,218 791 1,606 Pingdingshan Henan 739,979 864,236 441 999 Puyang Henan 603,235 704,531 359 815 Xinxiang Henan 618,482 722,338 368 835 Xinyang Henan 1,170,481 1,613,598 697 1,865 Zhengzhou Henan 1,923,102 2,012,151 1,145 2,326 Huangshi Hubei 582,410 680,209 492 826 Shashi Hubei 437,697 511,196 370 620 Wuhan Hubei 3,729,185 4,180,264 3,152 5,074 Changde Hunan 456,693 533,381 329 595 Changsha Hunan 1,619,656 2,073,669 1,165 2,314 Hengyang Hunan 629,739 735,485 453 821 Shaoyang Hunan 548,751 640,898 395 715 Xiangtan Hunan 558,927 652,783 402 729 16 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Estimated Population Estimated GDP City Province 1997 2005 1997 2005 Yueyang Hunan 437,848 511,372 315 571 Zhuzhou Hunan 588,084 686,836 423 767 Changzhou Jiangsu 806,570 942,010 676 1,533 Huaiyin Jiangsu 462,476 540,135 388 879 Nanjing Jiangsu 2,497,183 3,086,091 2,093 5,021 Nantong Jiangsu 554,951 648,139 465 1,054 Suzhou Jiangsu 936,389 1,343,769 785 2,186 Taizhou Jiangsu 510,059 595,708 427 969 Wuxi Jiangsu 965,466 1,127,587 809 1,834 Xuzhou Jiangsu 1,041,865 1,216,816 873 1,980 Yancheng Jiangsu 523,458 611,357 439 995 Yangzhou Jiangsu 449,554 525,043 377 854 Zhenjiang Jiangsu 526,881 615,356 442 1,001 Nanchang Jiangxi 1,387,748 1,871,003 1,996 4,057 Changchun Jilin 2,017,984 2,536,822 884 1,839 Jilin Jilin 1,333,037 1,882,401 584 1,365 Siping Jilin 491,840 574,430 215 416 Anshan Liaoning 995,874 1,163,101 356 722 Benxi Liaoning 852,892 996,110 305 619 Dalian Liaoning 1,698,545 2,034,384 607 1,264 Dandong Liaoning 539,448 630,032 193 391 Fushun Liaoning 1,306,437 1,399,190 467 869 Fuxin Liaoning 577,919 674,963 206 419 Jinzhou Liaoning 497,459 580,993 178 361 Liaoyang Liaoning 590,150 689,248 211 428 Panjin Liaoning 539,456 630,041 193 391 Shenyang Liaoning 3,266,902 3,508,646 1,167 2,179 Yingkou Liaoning 502,446 586,817 179 365 Baotou Inner Mongolia 1,091,608 1,274,911 797 2,437 Hohhot Inner Mongolia 653,481 763,214 477 1,459 Xining Qinghai 637,995 745,127 235 543 Xian Shaanxi 2,898,628 3,953,023 1,181 2,912 Xianyang Shaanxi 887,595 1,036,641 362 764 Jinan Shandong 1,733,358 2,068,438 2,671 6,502 Qingdao Shandong 1,420,056 1,641,300 2,188 5,159 Taian Shandong 681,066 795,431 1,050 2,500 Yantai Shandong 627,849 733,277 968 2,305 Zhangdian Shandong 558,457 652,234 861 2,050 Shanghai Shanghai 11,350,971 14,595,886 3,909 9,154 Changzhi Shanxi 591,988 691,395 259 650 Datong Shanxi 882,102 1,030,225 386 969 Taiyuan Shanxi 2,458,647 2,720,674 1,077 2,560 Chengdu Sichuan 3,836,588 3,947,443 3,048 5,666 Neijiang Sichuan 453,248 529,357 360 760 Zigong Sichuan 571,859 667,886 454 959 Tanggu Tianjin 465,272 543,401 158 467 Tianjin Tianjin 3,781,448 3,759,999 1,285 3,231 Shihezi Xinjiang 487,339 569,173 342 714 Urumqi Xinjiang 1,255,129 1,507,206 880 1,890 Lasa Tibet 100,770 117,692 90 251 Kunming Yunnan 878,111 1,025,564 1,913 3,473 Hangzhou Zhejiang 1,737,007 1,878,422 3,056 7,355 Ningbo Zhejiang 603,836 705,232 1,062 2,762 Wenzhou Zhejiang 726,139 848,073 1,278 3,321 Yinchuan Ningxia Hui 401,146 468,507 245 606 Estimated GDP in 100 mill yuan: Provincial GDP distributed to cities in the province in proportion to their population. Adjusted to 2005 levels using GDP deflators. 17 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Table 3: Trade flows and GDP for provinces Domestic Domestic GDP Num Province imports exports GDP 1997 2005 1 Beijing 768 111 2106 6886 2 Tianjin 876 974 1443 3698 3 Hebei 6522 7437 4600 10096 4 Shanxi 315 86 1722 4180 5 InnerMongolia 452 707 1273 3896 6 Liaoning 460 775 4061 8009 7 Jilin 704 588 1683 3620 8 Heilongjiang 1111 1114 3151 5512 9 Shanghai 1920 1040 3909 9154 10 Jiangsu 3845 4351 7772 18306 11 Zhejiang 535 948 5396 13438 12 Anhui 1093 1095 3106 5375 13 Fujian 203 29 3491 6569 14 Jiangxi 469 511 1996 4057 15 Shandong 2991 3099 7737 18517 16 Henan 996 1300 4746 10587 17 Hubei 669 783 4014 6520 18 Hunan 571 502 3482 6511 19 Guangdong 3057 2439 8511 22367 20 Guangxi 731 749 2345 4076 21 Hainan 158 136 477 895 22 Chongqing 463 430 1571 3070 23 Sichuan 329 358 3863 7385 24 Guizhou 197 150 923 1979 25 Yunnan 431 384 1913 3473 26 Tibet 30 25 90 251 27 Shaanxi 590 556 1543 3676 28 Gansu 285 292 909 1934 29 Qinghai 73 66 235 543 30 Ningxia 91 62 245 606 31 Xinjiang 486 325 1222 2604 Trade and GDP in 100 million Yuan at 2005 levels. 18 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Table 4: Assumed proportions of economic sector trade transported by road Num Sector Proportion 1 Agriculture 0.75 2 Coal mining and processing 0.00 3 Crude oil and natural gas products 0.00 4 Metal ore mining 0.00 5 Non-ferrous mineral mining 0.00 Manufacture of food products and tobacco 6 processing 1.00 7 Textile goods 1.00 Wearing apparel, leather, furs, down and related 8 products 1.00 9 Sawmills and furniture 0.75 Paper and products, printing and record medium 10 reproduction 0.75 11 Petroleum processing and coking 0.25 12 Chemicals 0.50 13 Nonmetal mineral products 0.50 14 Metals smelting and pressing 0.75 15 Metal products 1.00 16 Machinery and equipment 0.50 17 Transport equipment 0.50 18 Electric equipment and machinery 1.00 19 Electronic and telecommunication equipment 1.00 20 Instruments, meters, cultural and office machinery 1.00 21 Maintenance and repair of machinery and equipment 0.00 22 Other manufacturing products 1.00 19 DOMESTIC TRADE IMPACTS OF THE EXPANSION OF THE NATIONAL EXPRESSWAY NETWORK IN CHINA Table 5: Gravity estimation results of trade flows by provinces Trade flows Importer GDP 0.741 (10.56)** Exporter GDP 0.874 (12.51)** Travel time -2.016 (15.74)** Constant -5.128 (5.15)** Observations 928 R-squared 0.42 Absolute value of t statistics in parentheses * significant at 5%; ** significant at 1% 20