Assessing the impact of infrastructure investments using customs data: The case of the greater Mekong Subregion corridor and the People's Republic of China

Elhan-Kayalar, Yesim; Kucheryavyy, Konstantin; Nose, Manabu; Sawada, Yasuyuki; Shangguan, Ruo Working Paper Assessing the impact of infrastructure investments using customs data: The case of the greater Mekong Subregion corridor and the People's Republic of China ADB Economics Working Paper Series, No. 710 Provided in Cooperation with: Asian Development Bank (ADB), Manila Suggested Citation: Elhan-Kayalar, Yesim; Kucheryavyy, Konstantin; Nose, Manabu; Sawada, Yasuyuki; Shangguan, Ruo (2023) : Assessing the impact of infrastructure investments using customs data: The case of the greater Mekong Subregion corridor and the People's Republic of China, ADB Economics Working Paper Series, No. 710, Asian Development Bank (ADB), Manila, https://doi.org/10.22617/WPS230589-2 This Version is available at: https://hdl.handle.net/10419/298156 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ ASIAN DEVELOPMENT BANK ASIAN DEVELOPMENT BANK 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines www.adb.org ADB ECONOMICS WORKING PAPER SERIES NO. 710 December 2023 Assessing the Impact of Infrastructure Investments Using Customs Data The Case of the Greater Mekong Subregion Corridor and the People’s Republic of China This paper presents new evidence of the effects of road construction on both domestic and international trade flows in the People’s Republic of China using customs data and information on transport investments in the region. Road construction helped to reduce trade costs significantly from 2000 to 2011. The construction of the Kunming–Bangkok Expressway, in particular, led to local economic growth and higher regional specialization in accordance with comparative advantage, facilitating market integration across borders in the Greater Mekong Subregion. About the Asian Development Bank ADB is committed to achieving a prosperous, inclusive, resilient, and sustainable Asia and the Pacific, while sustaining its efforts to eradicate extreme poverty. Established in 1966, it is owned by 68 members —49 from the region. Its main instruments for helping its developing member countries are policy dialogue, loans, equity investments, guarantees, grants, and technical assistance. ASSESSING THE IMPACT OF INFRASTRUCTURE INVESTMENTS USING CUSTOMS DATA THE CASE OF THE GREATER MEKONG SUBREGION CORRIDOR AND THE PEOPLE’S REPUBLIC OF CHINA Yesim Elhan-Kayalar, Konstantin Kucheryavyy, Manabu Nose, Yasuyuki Sawada, and Ruo Shangguan ASIAN DEVELOPMENT BANK The ADB Economics Working Paper Series presents research in progress to elicit comments and encourage debate on development issues in Asia and the Pacific. The views expressed are those of the authors and do not necessarily reflect the views and policies of ADB or its Board of Governors or the governments they represent. ADB Economics Working Paper Series Yesim Elhan-Kayalar, Konstantin Kucheryavyy, Manabu Nose, Yasuyuki Sawada, and Ruo Shangguan No. 710 | December 2023 Yesim Elhan-Kayalar ([email protected]) is an advisor at the Economic Research and Development Impact Department, Asian Development Bank. Konstantin Kucheryavyy ([email protected]) is an assistant professor and Yasuyuki Sawada (sa[email protected]o. ac.jp) is a professor at the University of Tokyo. Manabu Nose ([email protected]) is an economist at the International Monetary Fund. Ruo Shangguan (ro[email protected]) is an assistant professor at Jinan University. Assessing the Impact of Infrastructure Investments Using Customs Data: The Case of the Greater Mekong Subregion Corridor and the People’s Republic of China Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) © 2023 Asian Development Bank 6 ADB Avenue, Mandaluyong City, 1550 Metro Manila, Philippines Tel +63 2 8632 4444; Fax +63 2 8636 2444 www.adb.org Some rights reserved. Published in 2023. ISSN 2313-6537 (print), 2313-6545 (electronic) Publication Stock No. WPS230589-2 DOI: http://dx.doi.org/10.22617/WPS230589-2 The views expressed in this publication are those of the authors and do not necessarily reflect the views and policies ofthe Asian Development Bank (ADB) or its Board of Governors or the governments they represent. ADB does not guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence of their use. The mention of specific companies or products of manufacturers does not imply that they are endorsed or recommended by ADB in preference to others of a similar nature that are not mentioned. By making any designation of or reference to a particular territory or geographic area, or by using the term “country” inthis publication, ADB does not intend to make any judgments as to the legal or other status of any territory or area. This publication is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) https://creativecommons.org/licenses/by/3.0/igo/. By using the content of this publication, you agree to be bound bytheterms of this license. For attribution, translations, adaptations, and permissions, please read the provisions andterms of use at https://www.adb.org/terms-use#openaccess. This CC license does not apply to non-ADB copyright materials in this publication. If the material is attributed toanother source, please contact the copyright owner or publisher of that source for permission to reproduce it. ADB cannot be held liable for any claims that arise as a result of your use of the material. Please contact [email protected] if you have questions or comments with respect to content, or if you wish toobtain copyright permission for your intended use that does not fall within these terms, or for permission to use theADB logo. Corrigenda to ADB publications may be found at http://www.adb.org/publications/corrigenda. Notes: In this publication, “$” refers to United States dollars. ADB recognizes “China” as the People’s Republic of China and “Turkey” as Türkiye. ABSTRACT This paper provides new evidence of the effects of road construction on both domestic and international trade flows in the People’s Republic of China (PRC) using customs data and information on transport investments in the region, including those supported by multilateral development banks. We find that road construction helped to reduce trade costs significantly from 2000 to 2011, supporting the catch-up of inland regions in the PRC to its coastal cities. The ad valorem rate of internal trade costs decreases by 20%, and the ad valorem rate of international trade costs decreases, on average, by 15.3%, with substantial heterogeneity of effects across sectors. Using satellite and customs data, we also document that the construction of the Kunming–Bangkok Expressway led to local economic growth and higher regional specialization in accordance with comparative advantage, suggesting the role of the road construction in facilitating market integration across borders in the Greater Mekong Subregion. Keywords: development impact, infrastructure, economic growth, trade, job creation, regional specialization, market integration JEL codes: R40, R41, F10, F13 1 Introduction Governments, international development organizations, and financial institutions invest substantially in transportation infrastructure projects around the world. For example, the People’s Republic of China’s (PRC) National Trunk Highway System costs more than $120 billion over 15 years (Faber, 2014). More recently, the PRC’s Belt and Road Initiative has promoted transportation infrastructure across Asia, Europe, and Eastern Africa. To evaluate the return on investments and inform future policy decisions, it is important to understand how improved connectivity affects the process of industrialization and the overall welfare of people living in the affected regions. This paper explores the role of road construction in shaping trade patterns in the PRC from 2000 to 2017 and in facilitating market integration across borders. To provide new evidence, we study transportation projects partially supported by the Asian Development Bank (ADB) in the Greater Mekong Subregion (GMS), an area comprising countries located along the Mekong River, i.e., Lao People’s Democratic Republic (Lao PDR), Thailand, Viet Nam, Myanmar1, and Cambodia. During the review period, the PRC’s trade expanded exponentially, and significant investments were made to improve the road network within the PRC, creating useful opportunities for researchers to study the effects of road construction on trade costs. We first document the major patterns of trade flows from 2000 to 2017. Since joining the World Trade Organization in 2021, the PRC’s international trade expanded rapidly, especially for electrical equipment and related products, which accounted for more than 40% of the PRC’s total exports after 2006. However, this total number hides the significant heterogeneity among different regions within the PRC. Infrastructure projects also contribute to the process of regional integration between the PRC and neighboring Southeast Asian countries. In this paper, we analyze the case of the Kunming–Bangkok Expressway that connects Kunming, the provincial capital city of Yunnan Province, and Bangkok. The expressway greatly facilitates road transportation between the PRC and Thailand. Using the PRC’s customs data, we document a substantial compositional change in both exports and imports between Yunnan Province and Thailand during this period. Three points are worth noting. First, agricultural products replaced chemical products as the main export from Yunnan Province to Thailand. With climate and agricultural conditions suitable for growing high-value plants in Yunnan Province, this shift implies an increase in resource allocation 1Effective 1 February 2021, ADB placed a temporary hold on sovereign project disbursements and new contracts in Myanmar. 2 efficiency consistent with regional comparative advantage. Second, the share of trade conducted via road transportation increased significantly after the construction of the expressway. Third, the customs ports in Kunming accounted for a major share of trade after the expressway was constructed. These facts are consistent with the important role of transport infrastructure in facilitating connectivity and market integration across regional borders. To complement the facts above, we use geo-coded satellite and customs data to perform a difference-in-differences (DID) analysis of the effects of road construction on economic activities. The road networks that included and did not include this expressway were used to construct separate sets of distances to the border with each neighboring county, and the difference is used to measure treatment effects. Intuitively, the larger the magnitude of distance reduction, the higher the treatment measure, and the greater the benefit from the expressway’s construction. To address the concern that the location choice of this expressway was not random, we utilize changes in the build-up level captured by daytime satellite images (1987–1995) and nighttime luminosity (1994–1997), both of which predate the road project, as the placebo test. To test the effects on trade, we employ the change in maritime exports as the placebo test since road construction is less likely to benefit maritime trade. We find consistent evidence that road construction leads to a higher share of built-up areas, higher nighttime lights intensity, and more extensive trade growth for regions that experience a higher treatment, i.e., for regions that are connected to the new road construction, and such effects are found to be nonexistent in the placebo tests. We thus provide strong evidence for a positive effect of infrastructure construction on economic activity. We also estimate the internal and international trade costs for Chinese exporters in 2000 and 2011. Our study uses information on domestic trade flows contained in the customs data. Using Chinese customs data, we measure the exports from each city in the PRC to each foreign destination via each port. Holding the origin city and destination constant, higher internal trade barriers decrease exports that go through a specific port, and this relationship allow us to extract trade costs from a standard gravity equation estimation. As a proxy for trade costs, we calculate the shortest distance from each city to each port using the Chinese road network data and the prefecture boundary provided by Baum-Snow et al. (2017). To guide the empirical analysis, we extend the Eaton and Kortum (2002) model to include an internal trade cost component and derive a gravity equation that allows us to separate the trade costs from other components that determine trade flows. 3 We find that road construction helped to significantly reduce trade costs from 2000 to 2011. On average, the ad valorem rate of internal trade costs decreased by 20.0%, and the ad valorem rate of international trade costs declined by 15.3%. We also document a substantial heterogeneity of effects across economic sectors. Lastly, we provide further evidence supporting the role of infrastructure in improving connectivity. Because maritime transportation is one of the primary means of transporting goods overseas from the PRC, coastal areas have a natural comparative advantage in participating in international trade. Using the distance to the nearest custom port that allows maritime transportation as a measure of "inlandness," we show that trade volumes decline substantially as one moves to more inland regions of the PRC. This trend, however, is partially offset by the construction of roads that connect these inland regions to main transit corridors and reduce internal trade costs. Moreover, using the estimated trade costs, we show that access to ports has a strong explanatory power in terms of provincial outputs, with a coefficient similar to that found in Donaldson and Hornbeck (2016). 2 Background This paper is part of an effort to quantify the effects of road infrastructure development on trade in Yunnan Province and the GMS. We believe that the estimates in this paper can be valuable inputs for infrastructure projects. Asian Development Bank (2021) (2021) noted that the development of regional connectivity among international trade hubs would strengthen regional value chains, which in turn would help unlock the benefits from integration into global value chains. This ADB study highlighted the importance of (i) regions’ physical connectivity to Bangkok as well as online routing systems to evaluate and monitor up-to-date road connectivity more efficiently; (ii) reducing border crossing times, which accounted for much of the total transport time between origin and final destinations in the GMS; and (iii) connecting regional production centers to seaports through multimodal transportation as available (i.e., roads, railways, waterways, and airports). In this paper, we note the impact of these modes of transportation available for trade. In Kucheryavyy et al. (2021), we laid out a multi-sector version of Desmet et al. (2018) model that allows for a rich impact of infrastructure projects on immigration choice and industrial composition. Nose et al. (2021) adopted a reduced-form approach to evaluate the effect of highway construction between the mountainous PRC–Viet Nam border and industrial hubs in northeastern Viet Nam. They found that improved accessibility due to the GMS transport investments had significantly expanded the market potential of the treated districts, with a strong agglomeration of 4 manufacturing firms along the highway in both core and peripheral cities. Further, the agglomeration increased the value added (i.e., productivity) of firms disproportionately more in the rural peripheries. Our paper contributes to the large and growing literature on the effects of transportation infrastructure improvements on economic outcomes, as surveyed in Redding and Turner (2015) and Redding (2022). Closest to our paper are the studies that focus on the PRC. These studies can broadly be divided into two categories: the ones that used a DID approach and the ones that conducted general equilibrium analysis. Faber (2014), Baum-Snow et al. (2017), Baum-Snow et al. (2020), Banerjee, Duflo, and Qian (2020), and He et al. (2020), among others, belong to the first category employing a DID approach. One important finding in these papers is that, on average, transportation infrastructure improvements in the PRC in the 1990s and 2000s had only a small or even negative impact on economic activity in locations with improved transportation infrastructure relative to other locations. This observation holds despite the fact that different studies focused on different units of observation in the PRC: For example, Faber (2014) and He, Xie, and Zhang (2020) compared peripheral counties within the same prefecture, Baum-Snow et al. (2017) contrasted central cities across different prefectures; and Baum-Snow et al. (2020) looked across prefectures. However, the average effects estimated by these studies can conceal the heterogeneity of outcomes depending on other characteristics of locations. For example, Baum-Snow et al. (2020) observed that prefectures in the PRC that are regional population centers had positive (relative) economic outcomes, while hinterland prefectures had negative effects. He, Xie, and Zhang (2020) suggested that poorer peripheral counties within a prefecture had a positive (relative) effect of infrastructure on gross domestic product (GDP) growth, while richer counties within the same prefecture had a negative effect. We differentiate from these studies by analyzing both the domestic and international trade impact of transport infrastructure and increased connectivity along trade routes. Tombe and Zhu (2019), Fan (2019), Xu and Yang (2021), Ma and Tang (2022), and Fan, Lu, and Luo (2023), among others, belong to the second category of studies on the PRC that utilized the general equilibrium modelling approach, which allowed them to find absolute (as opposed to relative) effects of transportation infrastructure improvements on economic outcomes. Tombe and Zhu (2019) had found that reductions in internal trade and migration costs in the PRC during the 2000s had a more significant impact on regional development than declines in international trade costs. While Tombe and Zhu (2019) lacked direct measures of transport infrastructure, they estimated internal trade and migration costs based on data on aggregate trade and migration flows between regions in 11 agricultural production by solving the incentive problem underlying the People’s Commune System, the farmland became fragmented, so the adoption of modern farming techniques critical for returns-to-scale in agricultural production became difficult. Realizing this problem, the government has implemented a series of policies to promote rural land circulation since the 2000s. Consistent with the objective of these policies, Wang et al. (2018) found that the circulation rate of rural land in Yunnan Province significantly increased from 2003 to 2013. 5 Economic Effects of Kunming-Bangkok Expressway This section analyzes the effects of the Kunming–Bangkok Expressway on economic activities in Yunnan Province. We first discussed the distance measure that serves as the treatment indicator for different regions in Yunnan Province, then presented evidence showing that the construction of the expressway impacts economic activities, as measured by satellite images and customs data. We define DistMhit as the distance to Mohan port along the road network in year t for each county iin Yunnan Province. Two networks are considered to represent two different time points. In the first network, we used only the national highway in 1999 to represent the network available before the construction of the expressway. In the second network, we added the expressway G8511 to the national highway in 1999 to capture the change in the road network after the construction of the expressway. We assume that the driving speed for the expressway is twice as fast as the national highway’s to capture the quality difference. After calculating the distance for both the before and after periods, ∆ln DistMhiis calculated as the log difference of the distance under two networks. We therefore define TreatInti=−∆ln DistMhias the intensity of treatment from the construction of the expressway. For counties with a larger magnitude of distance reduction, the potential benefit tends to be higher. Figure 4 shows the distribution of TreatIntiacross counties in Yunnan Province. Naturally, the difference is higher in magnitude for counties near the expressway. Conditional on the arc distance to the expressway, the counties with a better connection to the expressway through the national highway will also have a larger difference and thus more intensive treatment. To measure the change in economic activities at the county level, we relied on daytime and nighttime satellite images. Figure 5 plots the amount of built-up area for each county in Yunnan Province over time. To save space, we show only plots for 1987, 1995, and 2019. For each county, we counted the number of pixels that are classified as buildings by the machine learning algorithm, then divided the count by the total number of pixels. The 12 counties with red color are those that have a greater change in the share of built-up areas, and counties with white color are those for which we do not have valid observations. It is clear from the figure that there has been a significant increase in the share of built-up areas during the review period. Figure 6 plots the average NTL intensity for each county over the years. Again to save space, we plot for 1999, 2004, 2009, and 2019 only. The gradual increase in the NTL intensity is visually clear, with the brightest region corresponding to Kunming city. The dramatic increase in the NTL intensity in 2019 may be the result of the switch of satellite discussed earlier instead of real economics. Therefore, in later analysis, we separate the sample periods that use different satellites. 5.1 Level Regressions To estimate the effect of expressway construction on local economic outcomes, we estimate the following regression: yit =α0+α1TreatInti×A f terConst+α0 2Xit +ψi+ψt+eit,(1) where for the outcome variable yit, we use ln (NTLit +1)and BShareit to represent the NTL intensity and building share respectively; A f terConstis a year dummy equal to 1 if year tis later or equal to 1998, ψiand ψtare the county and year fixed effects, respectively; and Xit is the vector of county attributes. The underlying identification strategy is DID. For each county, we calculate the difference in economic activity over time. Then we compare counties that are more affected by the expressway construction with those that are less affected. Because there are other expressways that affect the connectivity of counties, we restrict the sample to counties that are within 50 kilometers of the expressway. Results using NTL intensity are reported in Table 9. The first column shows the estimate of α1without controlling for county-specific attributes, and the second column shows the estimate of α1controlling for the land area and total population in each county in year t−1. Both estimates report a positive, significant estimate of α1, supporting the hypothesis that counties with a larger reduction in the distance to the border due to the expressway’s construction tend to have greater NTL intensity. Since we control for county fixed effects and year fixed effects, any time-invariant county attributes that confound the estimation result are already controlled for. Due to data limits, we have only county attributes since 2001 for a subset of the counties. Therefore, the number of observations is smaller in the second column. That the two columns report similar estimates is reassuring. 13 Table 10 reports the estimates using the share of built-up areas as the outcome variable. The estimate in the first column is positive and significant, showing that counties with higher treatment tend to have a larger share of built-ups, consistent with the finding using NTL intensity. When we controlled for lagged county land area and total population, the effect has become only marginally significant. We also note that the sample size has shrunk due to data limitation. 5.2 First-difference Regressions To complement the regressions in level, we ran the following regressions in first difference: ∆yit =β0+β1TreatInti+ψt+eit (2) where for the outcome variable yit, we included ln (NTLit +1)and BShareit to represent the NTL intensity and building share respectively. Running the regression in first difference helps to cancel out the factors that are time-invariant and helps to mitigate the concern in our identification strategy. If the expressway is chosen to pass through counties with higher growth potential, then we would overestimate the effect of expressway construction. To alleviate this concern, we used the periods before the expressway construction as the placebo test. Because the expressway has not yet been constructed, we should not see any significant effects on our treatment. Because of the potential concern in data measurement, we utilized the periods before 2013 for NTL regression. Table 11 reports the estimation results for the NTL intensity. The first column shows that the treatment did not have a significant effect for the period before 1997 when construction began on the Kunming–Mohan Expressway. Between 1998 and 2012, there is a significant treatment effect. The counties with a greater change in distance tended to have a larger increase in NTL intensity. Table 12 shows the regression results using the changes in building share as the measure of economic outcome. Similarly, for the periods before the expressway’s construction, there is no evidence that counties with higher treatments tended to have a higher growth rate in building share. But the effect is estimated to be significant after the expressway’s construction. 5.3 Effects on Trade Flows Finally, we investigate the effects of expressway construction on trade flows. For that purpose, we replace the outcome variable with changes in export value. For each year, 14 our panel contains detailed export values from each customs region to Thailand, for each HS 8 code, each kind of transportation method, and each port. In light of the over trends reported in subsection 4.2 of this paper, we aggregated the categorical variables to capture the major differences. Therefore, in the regressions reported below, the index irepresents a cell defined by its geographical region, type of product being agricultural or nonagricultural, method of transportation being road or maritime, and port of clearance being Kunming and non-Kunming. To construct a balanced panel for each cell, we imputed a zero trade value for unobserved cells each year. This makes sense since the customs data recorded universe of all export transactions. Because the expressway also connects to the Lao PDR, we include exports from Kunming to the Lao PDR as well. We reported estimation results separately for road and maritime transportation, as the expressway’s construction was more likely to benefit road transportation. The regressions using maritime transportation can thus be viewed as a placebo test. We controlled for the destination, port, product category, and year fixed effects in all the regressions using changes in trade values. Table 13 reports the estimation results. As shown in the first column, the regions that experienced a higher reduction in the distance tended to have a larger increase in the export value. The elasticity is around 6, which means that a 10% decrease in distance could lead to a 60% increase in exports. This result suggests that the high growth of trade during the data period is concentrated in regions that benefit more from the expressway’s construction. Moreover, the second column shows that such an effect is much smaller and statistically insignificant for maritime transportation, supporting that our estimates do capture the benefits of road construction. This estimate, however, do not distinguish between generating trade or sorting (reallocating) trade. This is important since the impacts on welfare will be more positive if the road generates more trade instead of reallocating trade from elsewhere. Nonetheless, given the magnitude of trade growth, it is very likely that the generating effect is present. 6 Estimation of Trade Costs In this section, we use nationwide data to infer the reduction of trade costs due to the construction of high quality expressways. In the first subsection, we proposed a simple model to guide the analysis, and in the second subsection, we performed the analysis. Later subsections will explore the implications of the estimation results. 15 6.1 Model We rationalized the data with an extension of the standard Eaton and Kortum (2002) model that featured an internal trade cost component between origin cities and ports. In the model, locations include Chinese cities and foreign countries. A continuum of goods is indexed by ω. City ican produce any good with a constant marginal cost ci. For goods to be exported from a Chinese city to a foreign destination, a port must be chosen. To explain the fact that each city exports from different ports to the same country, we assumed that there is a city–port shock to the trade costs, in addition to the standard iceberg costs. The bilateral trade costs thus have three components: (i) the cost from the origin city to the port, (i) the cost from the port to the foreign destination, and (iii) an origin–port shock. Formally, the trade costs from the origin city ito a foreign destination jare equal to τidτdj νid(ω). We assumed perfect competition, thus the price of a good produced in city ifaced by a buyer in destination jand shipped via port dis equal to pidj (ω)=ciτidτdj νid (ω).(3) Effectively, the shock also incorporated the efficiency differences in producing goods across cities. We made the standard assumption that νid follows an independent and identically distributed Frechet distribution across goods, with location parameter Tid and shape parameter θ. Representative consumers in destination jchoose consumption quantity Q(ω)to maximize the the constant elasticity of substitution (CES) utility function: U="ˆ1 0 Q(ω)σ−1 σdω#σ σ−1 ,(4) where σis the elasticity of substitution. The buyers in the destination jchooses the minimum price for each good, pj(ω)=min i,dpidj (ω).(5) Let Xidj denote export value from origin ito destination jvia port d, and Xj=∑l∑kXlkj as the total expenditure of country j. Following the same procedure as in Eaton and Kortum 16 (2002), we obtained a closed-form solution for the share of trade from origin ito destination jvia das πidj ≡Xidj Xj =Tid ciτidτdj−θ Φj ,(6) where Φj=∑l∑kTlk clτlkτkj−θ. To close the model, we let the marginal cost equal the wage in city i, or ci=wi, and write the labor market clearing condition as wiLi=∑ d ∑ j πidjwjLj.(7) Later to solve the change, we also expressed the trade share in the “hat” form as ˆ πidj = ˆ Tid ˆ wiˆ τid ˆ τdj−θ ∑l∑kπilk ˆ Tlk ˆ wlˆ τlk ˆ τkj−θ,(8) where for any variable Xand its new value X0we define ˆ X≡X0 X. The labor market clearing condition can also be written as ˆ wiˆ Li=∑ d ∑ j γidj ˆ πidj ˆ wjˆ Lj,(9) where γidj ≡πidjwjLj wiLi. Under the assumption that ˆ Tid =0,ˆ Li=0, we have ˆ wi=∑ d ∑ j γidj ˆ wiˆ τid ˆ τdj−θ ∑l∑kπilk ˆ wlˆ τlk ˆ τkj−θˆ wj,(10) so given the values of γidj and changes in trade costs, we can solve for change in wages. 6.2 Estimation Equation (6) suggests the following specification: ln Xidj =−θln τid −θln τdj +ψid +φj+eidj,(11) where ψid =Tidc−θ iand φj=XjΦjrepresent origin and country fixed effects. In practice, we made use of the least-cost route distance between the origin iand port dto measure 17 the internal trade costs τid, and the arc distance between port dand destination jas the measure of international trade costs τdj, and assume that −θln τid =γ1ln (Distid +1), −θln τdj =γ2ln Distdj. Moreover, there are many origin–port pairs that had positive trade values in 2011 but had zero trade value in 2000. We found it important to incorporate these zeros into the estimation, otherwise the pairs with higher trade costs between origin cities and ports tend to be selected out of the sample, and the estimate of distance elasticity will become positive, contradicting the theory. Since our data also contain information on the product category and transportation method, we employed that information as well. The regression we estimate is thus ln Xidjsrt +1=γ0+γ1ln (Distidt +1)+γ2ln Distdj +ψi+ψd+φj+eidj,(12) where Xidjsrt represents total export value from origin city ito destination jvia port dfor products in sector s(defined using HS2 code) under transportation method r. To obtain estimates of τid and τdj, we need to take a stance on how large is θ. We take θ=4from Simonovska and Waugh (2014). Table 14 shows the estimation results. From 2000 to 2011, the distance elasticity decreased in absolute value from –0.261 to –0.151, consistent with the assumption that improvements in road infrastructure reduce internal trade costs at the same distance. On the other hand, the distance elasticity with respect to the distance between port and foreign destination declined in absolute value from –0.320 to –0.285, showing that the overall costs as well as the cost for travel between ports and foreign destinations decline. The magnitude of the decline in the distance elasticity is larger for the internal distance, consistent with the assumption that road construction significantly reduces domestic trade barriers. Since the construction of roads helped to reduce the effective distance from an origin city to the nearest port, the reduction in trade costs cannot be read from the change in distance elasticity only. Figure 7 shows the distribution of trade costs in 2000 (first row) and in 2011 (second row). The mean of internal trade costs decreased from 1.402 to 1.202, equivalent to a 20% decrease in tariffs. This is also depicted in Figure 7. The change in the extensive margin is substantial. In both observation years, there are 461 origin cities, 41 ports, and 209 unique destinations. In total, there were 252,563 unique 18 combinations that had positive trade in either 2000 or 2011. Out of all combinations, 176,737 combinations had zero trade value in 2000 and a positive trade value in 2011, and 15,309 combinations had a positive trade value in 2000 but zero trade value in 2011. Only 60,517 combinations had a positive trade value in both years. Since high trade barriers are likely a cause of zero trade, a sample excluding zero observations tends to select origin–destination pairs with low trade costs and thus cause an underestimation of trade costs. 6.3 Heterogeneity in Cost Reduction This section explores heterogeneity in the cost reduction effects of road construction projects. Table 15 reports the gravity equation estimate for 10 sectors that have the highest total export values in 2017. Columns (3) and (6) report the average inferred trade costs for each sector. There are substantial differences in trade costs. Sector 16 (machinery and mechanical appliances, etc.) has the highest internal trade costs. This may reflect the difficulty of shipping large machines via trucks. On the other hand, Sector 6 (products of chemical or allied industries) has the lowest internal trade costs. This is consistent with there being lower costs for standardized chemical goods. Table 16 reports the same estimate using the sample of 2011. When comparing across years between Table 15 and Table 16, the relative sizes of internal trade costs across sectors are very similar to those in 2000. For example, sector 16 and sector 6 are still the sectors with the highest and the lowest trade costs. This supports the notion that the estimated trade costs reflect the intrinsic nature of different goods. Moreover, we see a consistent decrease in internal trade costs across sectors. The external trade costs, however, change more dramatically. For example, for sector 16, the average external trade cost was 1.677 in 2000, which is nearly the lowest value. But it increased to 3.289 in 2011, ranking the second highest in that year. This probably reflects the change in destination compositions and is related to the change in global value chains. For example, if the PRC were to export more machine-related goods to trade partners in closer proximity, the model would interpret this change as an increase in external trade costs. 6.4 Effects on the Greater Mekong Subregion This section documents the estimates for the origins and destinations located in the GMS, where ADB has supported several road infrastructure projects. The selected destinations include the Lao PDR, Thailand, Viet Nam, Myanmar, and Cambodia. The selected origin 19 provinces in the PRC include Guangdong, Guangxi Zhuang Autonomous Region, and Yunnan Province. Table 17 reports the estimated results. The estimated internal distance elasticity significantly decreased from –0.282 to –0.041 from 2000 to 2011, providing support to the claim that road construction in the region promotes trade. 7 Other Evidence In this section, we show two further pieces of evidence that infrastructure projects promote connectivity and economic development across regions. 7.1 Inland Regions and Trade Since water transportation is the primary method of international trade, it is often argued that coastal regions have a comparative advantage over inland regions in international trade. Since a custom port that allows for water transportation must be chosen for such trade, we measure the degree of “inlandness” of each custom region, defined by the fivedigit customs code, as the distance between the region and the nearest port that allows for water transportation, which we denote as “Inlandi” for region i. For each year, we regressed the logarithmic value of total exports of each region, denoted as ln Xit, on the logarithmic value of inlandness, or ln Inlandi. To minimize the selection effect, we ensured that the sample of regions is the same across all years. Specifically, we include the set of regions with positive trade for at least one year and for each region iin this set; zero trade value is imputed for year tif iis not observed with positive trade. Table 18 reports the estimated results. Inlandness is a strong indicator of export values with an elasticity of about –1.0. Moreover, the magnitude of the elasticity declined substantially from –1.666 in 2000 to –0.923 in 2011, consistent with the view that the improvement of inland transportation infrastructure has a significant effect on reducing internal trade friction over time. 7.2 Port Access and Economic Development Since most international trade is transported by water, access to ports is critical for regions to benefit from globalization. In this subsection, using the inferred trade costs from each origin to port, we explored the potential importance of port access to economic 20 development. Using the idea populated by Donaldson and Hornbeck (2016), and also consistent with our model, we calculated the port access by PAit =∑ d τ−θ idt Xdt,(13) where τidt is the trade costs estimated in the previous section, θ=4, and Xdt is the total export value via port din year t. Then we used the port access term to explain the GDP in each region. Since we only have GDP data at the province level, we average port access within each province to also determine province-level port access. The result is shown in Table 19. Both GDP and port access are measured in US dollars. Since we control for province fixed effects, the identification comes from variation across years within each province. The relationship is significant both statistically and economically. A 10% increase in port access is associated with a 6% increase in provincial GDP. The magnitude is similar to the result reported in Donaldson and Hornbeck (2016). 8 Conclusion In this paper, we explore the relationship between road construction and trade patterns in the PRC. We document that improved transport infrastructure has helped inland regions’ trade growth by facilitating connectivity and lowering internal trade costs. We note that there may have been a sample selection bias, i.e., when individuals or groups in the sample studied differ from the population, for road transport versus sea transport given the nature of goods being transported, and therefore focus on changes in trade volumes and value for goods transported by roads in our analyses. Similarly, within country migration dynamics among coastal and inland areas in the PRC, which might have contributed to the clustering of commercial and residential areas around newly developed road corridors, are outside the scope of this paper and warrant further attention in the future. Finally, we also show evidence that investments in an enhanced road network have played a role in integrating regional markets in the PRC’s Yunnan Province and the GMS. 27 Table 7: Export Value from Yunnan Province to Thailand by Port 2000 2011 2017 Port Value ($ billion) Value Share Port Value ($ billion) Value Share Port Value ($ billion) Value Share Huangpu 52 0.007 0.323 Kunming 0.264 0.658 Kunming 0.454 0.559 Zhanjiang 67 0.007 0.300 Nanning 0.047 0.118 Shenzhen 0.233 0.287 Nanning 72 0.006 0.268 Shenzhen 0.035 0.088 Nanning 0.080 0.098 Kunming 86 0.001 0.043 Hangzhou 0.016 0.040 Hangzhou 0.009 0.011 Shanghai 22 0.001 0.028 Huangpu 0.015 0.037 Huangpu 0.008 0.010 Guangzhou 51 0.001 0.027 Zhanjiang 0.011 0.027 Nanjing 0.007 0.009 Dalian 09 0.000 0.007 Qingdao 0.008 0.019 Zhanjiang 0.007 0.008 Shenzhen 53 0.000 0.004 Nanjing 0.002 0.004 Tianjin 0.004 0.004 Tianjin 02 0.000 0.001 Shanghai 0.001 0.004 Shenyang 0.004 0.004 Fuzhou 35 0.000 0.001 Tianjin 0.001 0.002 Xiamen 0.002 0.003 Note: only the top 10 ports in terms of export value are shown. Source: Authors’ calculations. 28 Table 8: Export Value from Yunnan Province to Thailand by Transportation Method 2000 2011 2017 Transportation Value ($ billion) Value Share Transportation Value Value Share Transportation Value Value Share Water 0.022 0.978 Water 0.208 0.519 Water 0.422 0.520 Road 0.000 0.011 Road 0.175 0.437 Road 0.389 0.479 Airway 0.000 0.009 Airway 0.018 0.044 Airway 0.000 0.001 Source: Authors’ calculations. 29 Table 9: Nighttime Lights Regression in Level ln (NTLit +1)t-value ∆ln (NTLit +1)t-value TreatInti×A f terConst0.486 4.487 1.411 4.555 Lagged County Atrributes No Yes No. Obs 494 198 Adj. R20.974 0.966 Note: Year and county fixed effects are controlled. Source: Authors’ calculations. Table 10: Building Share Regression BShareit t-value BShareit t-value TreatInti×A f terConst0.056 3.067 0.023 1.534 Lagged County Atrributes No Yes No. Obs 176 32 Adj. R20.726 0.826 Source: Authors’ calculations. Table 11: Nighttime Lights Regression ∆ln (NTLit +1)t-value ∆ln (NTLit +1)t-value TreatInti0.029 0.550 0.054 2.016 No. Obs 78 390 Adj. R2-0.004 0.225 Period 1994 to 1997 1998 to 2012 Source: Authors’ calculations. 30 Table 12: Building Share Regression ∆BShareit t-value ∆BShareit t-value TreatInti0.004 1.360 0.022 2.296 No. Obs 44 88 Adj. R20.022 0.072 Period 1987 to 1995 2000 to 2019 Source: Authors’ calculations. Table 13: Export Level Regression ∆ln Expoit t-value ∆ln Expoit t-value TreatInti6.119 2.439 3.695 1.474 No. Obs 288 288 Adj. R20.050 0.074 Trans Road Maritime Source: Authors’ calculations. Table 14: Gravity Equation Estimates ln Xidjsrt +1t-value ln Xidjsrt +1t-value ln (Distidt +1)-0.261 -158.605 -0.151 -90.679 ln Distdj -0.320 -31.096 -0.285 -28.299 Year 2000 2011 No. Obs 1913098 1913098 Adj. R20.184 0.094 Source: Authors’ calculations. 31 Table 15: Gravity Equation Estimates across Sectors, 2000 (1) (2) (3) (4) (5) (6) sector ln Distidt t-value τid ln Distdj t-value τdj 6 -0.179 -31.923 1.261 -0.312 -9.125 1.973 7 -0.297 -52.243 1.475 -0.326 -8.022 2.056 11 -0.277 -62.128 1.432 -0.350 -14.419 2.153 12 -0.278 -37.031 1.436 -0.476 -9.232 2.880 13 -0.269 -45.501 1.438 -0.344 -8.553 2.138 15 -0.245 -58.443 1.373 -0.284 -10.273 1.868 16 -0.384 -80.597 1.665 -0.234 -7.193 1.677 17 -0.246 -27.213 1.368 -0.329 -4.778 2.073 18 -0.314 -46.790 1.486 -0.205 -4.638 1.571 20 -0.367 -62.660 1.624 -0.438 -10.968 2.649 Source: Authors’ calculations. Table 16: Gravity Equation Estimates across Sectors, 2011 (1) (2) (3) (4) (5) (6) sector ln (Distidt +1)t-value τid ln Distdj t-value τdj 6 -0.143 -25.228 1.192 -0.345 -10.414 2.123 7 -0.246 -43.519 1.358 -0.397 -10.165 2.409 11 -0.125 -28.844 1.164 -0.548 -24.043 3.329 12 -0.141 -18.922 1.187 -0.407 -8.207 2.473 13 -0.097 -15.609 1.131 -0.369 -9.026 2.258 15 -0.121 -28.070 1.159 -0.221 -8.050 1.626 16 -0.436 -102.896 1.743 -0.539 -19.224 3.289 17 -0.268 -30.026 1.386 -0.441 -6.718 2.654 18 -0.252 -38.488 1.350 -0.364 -8.737 2.236 20 -0.264 -46.287 1.391 -0.346 -9.139 2.154 Source: Authors’ calculations. 32 Table 17: Gravity Equation Estimates ln Xidjsrt +1t-value ln Xidjsrt +1t-value ln (Distidt +1)-0.282 -19.083 -0.041 -2.403 ln Distdj -2.264 -5.049 -2.602 -5.344 Year 2000 2011 No. Obs 21055 21055 Adj. R20.218 0.094 Source: Authors’ calculations. Table 18: Total Exports and Inlandness ln (Xit +1)t-value ln (Xit +1)t-value ln (Xit +1)t-value ln (Xit +1)t-value ln (Inlandi+1)-1.666 -11.211 -1.083 -11.802 -0.923 -10.730 -0.984 -9.118 Year 2000 2006 2011 2017 No. Obs 474 474 474 474 Adj. R20.209 0.226 0.194 0.148 Source: Authors’ calculations. Table 19: Port Access and Economic Development ln (GDPit)t-value ln (PAit)0.677 38.255 FE Province No. Obs 62 Adj. R20.980 FE = fixed effects, GDP = gross domestic product. Source: Authors’ calculations. 33 Figure 1: Network of National Highways in Yunnan Province, 1999 Source: Baum-Snow et al. (2017); Authors’ calculations. Disclaimer: The boundaries, colors, denominations, and any other information shown on the maps presented in this paper do not imply, on the part of the Asian Development Bank, any judgment on the legal status of any territory, or any other endorsement or acceptance of such boundaries, colors, denominations, or information. 34 Figure 2: Network of High Quality Expressways in Yunnan Province–1999, 2010, and 2021 Source: Baum-Snow et al. (2017); National Catalogue Service For Geographic Information; Authors’ calculations. 35 Figure 3: Example of Shortest Distance Calculation Source: Baum-Snow et al. (2017); Authors’ calculations. 36 Figure 4: Treatment Intensity across Counties in Yunnan Province Note: The figure plots the treatment intensity TreatIntifor different counties in Yunnan Province. Source: Authors’ calculations. 43 Nose, M., Y. Sawada, and T. Nguyen (2021). From battlefield to marketplace: Connectivity, industrialization, and spatial convergence in the Greater Mekong Subregion. Mimeo. Redding, S. J. (2022). Chapter 3: Trade and geography. In G. Gopinath, E. Helpman, and K. Rogoff (Eds.), Handbook of International Economics: International Trade, Volume 5, pp. 147–217. Elsevier. Redding, S. J. and M. A. Turner (2015). Chapter 20: Transportation costs and the spatial organization of economic activity. In G. Duranton, J. V. Henderson, and W. C. Strange (Eds.), Handbook of Regional and Urban Economics (1st ed.), Volume 5, pp. 1339–1398. Elsevier. Rothenberg, A. D. (2013). Transport infrastructure and firm location choice in equilibrium: Evidence from Indonesia’s highways. Mimeo, Department of Economics, University of California, Berkeley. Simonovska, I. and M. E. Waugh (2014). The elasticity of trade: Estimates and evidence. Journal of International Economics 92(1), 34–50. Tombe, T. and X. Zhu (2019). Trade, migration, and productivity: A quantitative analysis of China. American Economic Review 109(5), 1843–72. Volpe Martincus, C., J. Carballo, and A. Cusolito (2017). Roads, exports and employment: Evidence from a developing country. Journal of Development Economics 125, 21–39. Wang, Y., X. Li, L. Xin, M. Tan, and M. Jiang (2018). Spatiotemporal changes in Chinese land circulation between 2003 and 2013. Journal of Geographical Sciences 28(6), 707–724. Xu, Y. and X. Yang (2021). Access to ports and the welfare gains from domestic transportation infrastructure. Journal of Urban Economics 126, 103392. ASIAN DEVELOPMENT BANK ASIAN DEVELOPMENT BANK 6 ADB Avenue, Mandaluyong City 1550 Metro Manila, Philippines www.adb.org ADB ECONOMICS WORKING PAPER SERIES NO. 710 December 2023 Assessing the Impact of Infrastructure Investments Using Customs Data The Case of the Greater Mekong Subregion Corridor and the People’s Republic of China This paper presents new evidence of the effects of road construction on both domestic and international trade flows in the People’s Republic of China using customs data and information on transport investments in the region. Road construction helped to reduce trade costs significantly from 2000 to 2011. The construction of the Kunming–Bangkok Expressway, in particular, led to local economic growth and higher regional specialization in accordance with comparative advantage, facilitating market integration across borders in the Greater Mekong Subregion. About the Asian Development Bank ADB is committed to achieving a prosperous, inclusive, resilient, and sustainable Asia and the Pacific, while sustaining its efforts to eradicate extreme poverty. Established in 1966, it is owned by 68 members —49 from the region. Its main instruments for helping its developing member countries are policy dialogue, loans, equity investments, guarantees, grants, and technical assistance. ASSESSING THE IMPACT OF INFRASTRUCTURE INVESTMENTS USING CUSTOMS DATA THE CASE OF THE GREATER MEKONG SUBREGION CORRIDOR AND THE PEOPLE’S REPUBLIC OF CHINA Yesim Elhan-Kayalar, Konstantin Kucheryavyy, Manabu Nose, Yasuyuki Sawada, and Ruo Shangguan