Transportation Infrastructure and Productivity Growth: Effects of Railway Speed-Up on Firm’s TFP in China

ShiZhenkai(施震凯),ShaoJun(邵军)andPuZhengning(浦正宁)

Shi Zhenkai (施震凯), Shao Jun (邵军) and Pu Zhengning (浦正宁) School of Economics and Management, Southeast University, Nanjing, China Abstract:

Improvement of transportation infrastructure quality will lead to more sufficient market competition and promote the flow of resources with greater efficiency. This paper considers China’s railway speed-up in 2007 as a quasi-natural experiment on China’s transportation infrastructure quality improvement. With the initial operation of electric multiple units (EMUs) as the basis of grouping, this research examines the effect of railway speed-up on corporate total factor productivity (TFP) growth by the differencein-differences (DID) method. Overally, the results reveal positive effects both on firms’ technological change and efficiency improvement, which lead to the increase of TFP. Based on subsamples divided by different regions and types of enterprises, further analysis indicates that the productivity of exporter, non-state and coastal firms has been mostly affected by the railway speed-up. These conclusions are verified by a placebo test. Besides, firms within “one-hour economic circle” have been shown more sensitive to the effect of railway speed increase.

Keywords:

railway speed-up, total factor productivity (TFP), data envelopment method, difference-in-differences method

JEL Classification Codes: D24; O38

DOI:1 0.19602/j .chinaeconomist.2018.11.0219

1. Introduction

World economic history since the Industrial Revolution clearly shows that transportation infrastructure improvement is essential for economic growth. For China, a large country with complex geographical characteristics, railway infrastructure is of great significance to its economic and social development. Since reform and opening up, China has made remarkable progress in its railway construction. The total length of railway operation in China increased from 53,000 kilometers in 1979 to 124,000 kilometers in 2016, resulting in a railway network with provincial cities as nodes connecting other major cities nationwide. However, China’s per capita railway network density is far below the level of developed countries. Given the insufficiency of transportation capacity, increasing train operation speed became an important solution. The former Ministry of Railway conducted six railway speed-ups from 1997 to 2007. As a result, the average train speed in China gradually increased from 48 km/h to

70 km/h. In 2007, China put into service the electric multiple units (EMU) operating at above 200 km/ h. Over a short period of time, China constructed more than 6,000 kilometers of high-speed railway , unveiling a new era of high-speed trains.

Over the years, the role of transportation infrastructure in economic growth did not receive sufficient attention in economic research. With remarkable improvement in China’s transportation infrastructure, relevant literature studies are increasing. Theoretically, transportation infrastructure improvement will accelerate the stream of people, promote fixed asset investment, increase capital stocks, enhance capital deepening, and boost economic growth rate through spillover effects. However, all these are short-term effects. Total factor productivity (TFP) is the fundamental driver of long-term economic growth. Under natural resource limitations and environmental constraints, it is unsustainable for economic growth to solely rely on production factor input; thus, economic growth increasingly needs to be driven by TFP. Some studies have examined the productivity effects of transportation infrastructure. For instance, Aschauer (1989) investigated the TFP effects of highways in the United States, and Farhadi (2015) uncovered the TFP effects of transportation infrastructure in 18 OECD countries. Both studies reveal the positive role of transportation infrastructure. With growing availability of microdata, an increasing number of studies have been carried out to investigate the corporate TFP effects of transportation infrastructure. However, most studies of this kind are focused on highways. For instance, Gibbons et al. (2012) discovered that an improvement in highway quality may have an important influence on corporate employees, labor productivity and other aspects. Li and Li (2013) discovered that an increase in highway investment will save firms’ warehousing cost and increase productivity. Holl (2016) discovered that highways promote corporate TFP growth by increasing the regional density and clustering effect. However, very few studies, especially studies on firms at a microscopic level, have been carried out to unravel the TFP effects of railway infrastructure. By addressing this topic, this paper attempts to fill this information gap in existing research literature.

In order to analyze the TFP effects of railway infrastructure improvement on firms, we should measure the improvement of transportation infrastructure quality, which is measured in existing studies by such criteria as increased investment, operation length, transportation capacity and reduced time of operation. For instance, Deng et al. (2014) examines the economic growth effects of transportation infrastructure based on investment and mileage data. Huang and Xu (2012) discovered that reductions in transportation cost, such as exemption of highway toll fees, are conducive to exports in interior regions. Based on China’s transportation infrastructure and investment stock data, Shi and Huang (2014) investigated whether China’s transportation infrastructure is in oversupply relative to economic development. Many studies regard China’s railway speed-up as a quasi-natural experiment for the analysis of multifaceted effects of railway infrastructure improvement. For instance, Zhou and Yu (2013) discovered that China’s railway speed-up has a more significant positive effect on secondary industry than on tertiary industry. Song and Li (2015) found that China’s railway speed-up will significantly promote population growth in cities along the rail and highway routes in the long run. Referencing the above methods, this paper identifies China’s railway speed-up as a quasi-natural experiment of transportation infrastructure improvement to examine its TFP effects.

Specifically, this paper creates a treated group and a control group based on whether a city has access to EMU trains. Based on that, the study is carried out using difference-in-differences method. Treated group and control group samples are differentiated by whether cities have access to EMU train stations rather than whether cities are along railway route, since not all cities along railway route have access to EMU trains. The most important meaning of railway speed-up is the operating route. However,

different from highways for which entrance and exit ramps can be created where necessary, railway stations may only be constructed at optimal sites due to cost considerations. Whether a city has a railway station determines the city’s access to a railway transportation network. What matters is not only whether a city is located along a railway route or the railway length within its jurisdiction. Rather, the key is whether it has a railway station with access to railway network. Cities with railway stations provide more options for the flow of people, logistics and information compared with others without railway stations.

2. Theoretical Mechanism

Improvement of firm’s TFP stems from technology progress and managerial innovation. The role of technology progress is particularly important. New technology is often created in one place, and spills over to other places (Liu et al., 2010). However, technology dissemination is both a temporal and spatial process. The longer it takes to disseminate, the more information will be distorted, i.e. longer distance diminishes the technology spillover effect (Fu, 2009). Spatial distance in the modern sense is not entirely confined to physical distance. It is also subject to temporal factor. This paper introduces real distance deflated by temporal factors into Keller’s (2002) model to determine productivity’s relationship with real distance and explains the influence of railway speed-up. Assuming that Place A and Place B mutually influence each other and are symmetric, their geographical distance is D, and railway speed-up will shorten train operation time from t0 to tr. We use t=tr/t0 to measure the impact of time on geographical distance. Smaller t means shorter real distance between the two places. Hence, the real distance of transportation deflated by t can be expressed as Dr=tD. Assuming that the output Y of firms in Place A meets:

Where, A is constant term, K is capital, and α is output elasticity, and 0< α< 1. m and n are the quantities of intermediate inputs provided by monopolistic manufacturers in Place A and Place B. c and c* are the types of input. N and N* are corresponding input portfolios. Assuming that the prices p and p* of intermediate inputs in both places are

Manufacturers in both places respectively use labor L and L* as the only input factor. w and w* are the wage levels of both places. Cost of transportation for firms in Place A to use local products is 0. However, the iceberg transport cost of transportation for use of nonlocal products is . Thus, equations and hold true under the optimal condition. When general equilibrium is reached, we may further deduce the following relational expression between m and n from p=p* and w=w*.

Namely, the elasticity of substitution between local input product m and nonlocal input product n employed by firms in Place A is 1/(1- α). Given the use of non-substitutable labor L and L* as input factor in the two places and the intermediate input portfolios of N and N* respectively, we may further arrive at:

3. Model and Data

After the first railway speed- up administered by the former Ministry of Railway for BeijingGuangzhou Railway and Beijing-Shanghai Railway in April 1997, six railway speed-ups were carried out prior to 2007. But the first five railway speed-ups were carried out only for traditional locomotives with very limited speed increases. In comparison, the railway speed-up in 2007 represents a major milestone that unveiled EMU trains that operate at over 200 km/h. This paper regards China’s railway speed-up of 2007 as a quasi-natural experiment and examines the corporate TFP effects of railway infrastructure quality improvement using the difference-in-differences method.

In conducting difference-in-differences method, the treated group and control group should be set first. According to the train schedule of April 2007, we identify 58 stations along EMU routes located in 49 cities involving 10 provinces, 3 municipalities and 1 special economic zone. Municipalities and special economic zones are more developed and highly heterogeneous compared with other cities. Thus, it is difficult to create a control group compatible with them. For this reason, municipalities and special economic zones are excluded from the examples, so there are 45 cities left.. Then, firms located in these cities are classified into treated group, and other cities in these 10 provinces are taken into the control group. We identify the geographical distribution of the treated group and control group in Table 1. Specifically, the coastal region includes 5 provinces, i.e. Hebei, Shandong, Jiangsu, Zhejiang and Guangdong. The remaining 5 provinces are Jiangxi, Henan, Hubei, Hunan and Shaanxi, which are inland regions.

After seting treated group and control group, this paper creates the following regression model based on the difference-in-differences method:

Where, Z is a dependent variable and refers to TFP-related variables obtained by DEA method, including Malmquist Index ( mqst), technology change index ( mtc) and efficiency change index ( mtec). CRH×period is the most important independent variable in the model, and its value is the product

between CRH and period. CRH is a group-specific dummy variable that describes differences between treated group and control group. If a firm is located in a region involved in the railway speed-up, it is classified as treated group with the value of 1; otherwise, the value is 0. period is time dummy variable that distinguishes the differences before and after railway speed-up, and its value is 1 during 2007-2009 and 0 for other years. h is the fixed effect of industry, u is the fixed effect of region, t is the fixed effect of time, cons is constant term, and ε is disturbance term. Besides. X mainly includes the following control variables: kl is capital density measured by the ratio between corporate capital and employees; sub is a dummy variable that denotes whether a firm enjoys government subsidy, and its value is 1 if it does; otherwise, it is 0. Firm size (scale) is a dummy variable whose value is 1 for large and medium-sized firms, and 0 for small firms. The city development ( citygdp) is measured by the GDP of the cities. Data employed in this paper are primarily from the China Industrial Enterprises Database 2001-2009 and the China City Statistical Yearbooks.

4. Regression Result and Analysis

4.1 Basic Results and Analysis

An important precondition for testing policy effects using the difference-in-differences method is to meet the “parallel trend hypothesis,” i.e. there is no systematic difference between a treated group and control group. Before the occurrence of an event, both of the groups share consistent development characteristics and trends. Otherwise, the result of difference-in-differences method is likely to be unbelievable. Referencing the methods of Moser and Voena (2012) and Tanaka (2015), this paper draws a time series plot of key variables and estimates parallel trend to assess whether treated group and control group samples share a consistent trend before railway speed-up. First, we drew a time series plot (Figure 1) for the mean values of dependent variables of treated group and control group. Through observation of Figure 1, we can discover that, despite the slight differences in the time series curves of variables prior to the railway speed-up, the increases and decreases of the treated group and control group share a consistent trend and demonstrate relatively robust parallel characteristics.

Figure 1 provides an initial assessment that treated groups and control groups share a similar development trend prior to railway speed-up implementation. However, parallel trend assumptions require a more precise test. Refer to Moser and Voena (2012), we further introduce a cross-multiplying term between the group-specific dummy variable and the time trend and report the results in Table 2. It can be found that after controlling for industry effect, time effect, regional effect and control variables, despite the positive and negative values of the regression coefficients of “treated group × year” term, the differences in such values are insignificant at the 10% statistical level and cannot reject the null hypothesis that a treated group and control group will share a consistent development trend prior to the railway speed-up. Thus, the parallel trend hypothesis test is passed. This results above mean that the

experimental grouping in this paper is appropriate, and that it is reasonable to use the difference-indifferences method to determine the effects of the railway speed-up on firm’s TFP.

Table 3 provides the regression results based on the difference-in-differences method. To make a comparative analysis, this paper simultaneously provides two results, one for core independent variables and another for inclusion of control variables. Apparently, there is no significant change in the coefficients of CRH × period. For the two regression equations of the Malmquist index ( mqst), the coefficients of CRH × period are all significantly positive. This result shows that the railway speedup of 2007 had a significantly positive effect on corporate TFP growth. Columns 4 and 5 of Table 3 are the regression results of the technology change index ( mtc) as a dependent variable. Moreover, the coefficients of CRH × period are significantly positive, i.e. the railway speed-up had a positive effect on the technology of firms in relevant regions. The last two columns of Table 3 are the estimation results for the efficiency change index ( mtec) and the coefficients of CRH×period are all insignificantly positive, meaning the railway speed- up has an insignificantly positive effect on firm efficiency improvement. While the railway speed-up significantly boosted firms’ technology progress in relevant regions, its efficiency improvement effect was not significant. A possible reason is the partially negative efficiency effect of the railway speed-up. Overall, despite the insignificant efficiency improvement

due to uncertainties of railway speed-up, the technology of firms benefited from the increased flows of passengers, logistics and information, and the positive effects of railway speed-up should be recognized.

4.2 Heterogeneous Effects

(1) We examine the effects of railway speed-up on SOE and non-SOE’s TFP, and the results are shown in Table 4. It can be found that the railway speed-up had a significant positive effect on the technology and efficiency of non-SOEs; the regression coefficients of CRH × period are 0.0093 and 0.0023 respectively, which promote corporate TFP growth under their coupled effect. While the railway speed-up boosted SOE ’s technology, the TFP is insignificant due to significant efficiency reduction. To some extent, the effect is negative. As mentioned before, railway speed-up may significantly increase

market competition in relevant regions. Due to the agent-principal relationship of SOEs and other problems like soft budgetary constraints, it takes a long time for SOEs to adapt to a new competitive environment. Despite technology improvement after railway speed-up, SOE’s efficiency reduced sharply due to negative effects. Under the goal of profit maximization, private firms are more sensitive to changes in market structure and externalities arising from railway speed-up, which is conducive to their TFP growth.

(2) We classify samples into coastal and inland regions for regression and present regression results in Table 5. Judging by the result of regression of mtc, the coefficients of the CRH × period are all significantly positive for both coastal or inland regions. In the regression of mtec, the coefficients of CRH × period are all insignificant. However, the values of the coefficients are positive for coastal

regions and negative for inland regions, and the absolute values of the latter are more significant. In summary, the coupled effect of mtc and mtec influences change in mqst. By the results of the mqst model, we may find that the TFP growth of firms in coastal cities covered by the railway speed-up campaign responded positively to the railway speed-up. However, the effect is limited for firms located in inland regions.

(3) Table 6 lists the regression results for firms whose target market is the domestic market and firms whose target market is the export market. The absolute value of the CRH × period in the mtc model of export firms is greater than the non-export firms and is more significant. Railway speedup has differentiated effects on the efficiency improvement of non-export firms and export firms. The differences may have to do with the transportation modes used by the two types of firms. At present,

marine transportation remains the primary mode of transportation for exports. On key international marine transportation routes, most goods suitable for container transportation are transported using containers. When inland manufacturers export goods, they will opt for full container loads (FCLs) or less container loads (LCLs) to ship containers directly to ports. Transportation capacity released by railway speed-up is favorable to container transportation and helps export companies to accelerate inventory turnover and increase efficiency. In addition, existing studies discover that an increase of commodity transportation time by one day is equivalent to an increase of commodity price by 0.6%–2.1%. In addition, long transportation delays will significantly reduce the success rate of exports (Hummels and Schaur, 2013). Railway speed-up will reduce the cost of trade for export firms and is favorable to their efficiency improvement. For firms whose target market is the domestic market, long-haul trucks using

highways are more convenient than railway transportation. Therefore, railway speed-up has a relatively limited effect for non-export firms with insignificant impact on their efficiency improvement

4.3 Robustness Test

(1) Placebo test. Some other policies or stochastic factors may also affect firm’s TFP. If such an impact is not correlated with railway speed-up, CRH×period should remain significant in other years. This implies that the contribution of CRH×period to corporate TFP should be attributable to factors other than railway speed-up. So we conduct a placebo test by altering policy implementation time to identify the doubt above. Without changing the treated group and control group, we shift the railway speed-up carried out to an earlier data, as 3 to 5 years. After controlling for industry effect, time effect and region

effect, Table 7 shows the regression results of placebo test. Obviously, there are slight differences in the coefficients of CRH×period in various models of hypothetical implementation year of railway speedup. However, such differences remain insignificant at 10% statistical level. These results exclude the possibility of any impacts of factors other than railway speed-up on the TFP of treatment-group firms, and reflect our core conclusions are robust.

(2) Test based on panel SFA method. We recalculate firm’s TFP using panel stochastic frontier analysis ( SFA) in stead of DEA, and retest the effectS of railway speed- up on firm’s TFP by the difference-in-differences method. Table 8 reports the regression results. In Table 8, the coefficient of CRH×period for all samples is significantly positive, which shows that railway speed-up has promoted the TFP of firms along the route and verifies the core conclusions above. Sample-specific regression results show that the CRH×period is highly significant and positive for non-SOEs, firms in coastal regions and export firms; moreover. This result coincides with our previous conclusions. Based on the above results, we may find that remeasuring firm’s TFP by a new method will not cause any significant impacts on our key conclusions.

(3) Test based on financial data. The above section estimates the technology and efficiency of firms using the data envelopment analysis (DEA) method. To further test the effects of railway speed-up on community businesses, we conduct regression using the new product output and financial data, such as general and administrative expenses of firms in Table 9. Column 2 of Table 9 shows the regression results of all samples. Judging by the regression coefficient of CRH×period, railway speed-up leads to a significant improvement in the firm’s new product output value. Increased market competition resulting from shortened temporal and spatial railway distance may induce firms to adopt new technology for new product manufacturing or develop new products using new design approaches, which leads to an increase in new product value. A sample-specific regression result is consistent with the above key conclusions. Namely, railway speed-up has a significant effect on the new product value of private firms in coastal regions; however, it has an insignificant impact on SOEs in inland regions, but the difference is also insignificant between non-export firms and export firms. Meanwhile, total sample regression results show that railway speed-up significantly reduced corporate managerial cost. However, consideration of sample heterogeneity leads to a different conclusion, i.e. railway speed-up significantly reduces the general and administrative cost of private firms, firms in coastal regions and export firms, but the effect is insignificant or negative for state-control enterprises, firms in inland regions and non-export firms. The level of general and administrative cost reflects the level of a company’s managerial efficiency and corporate efficiency. These findings support the core conclutions of this paper.

4.4 Further Discussion: Inferences Based on Geographical Spheres

This paper classifies firms into different geographical spheres based on their distance to high-tech cities like Beijing, Shanghai, Guangzhou and Shenzhen to test the heterogeneous effects of railway speed-up on the firm’s TFP within different spheres, and the test results are shown in Table 10. Notably, railway speed-up has the most significant effects on the mtc of firms within a distance of around 150 kilometers to high- tech cities like Beijing and Shanghai. Given the EMU train operation speed is between 160 to 200 km/h and deducting the time of train stops at stations, we may get the following conclusion: Railway speed- up expanded “one- hour economic circle” from 100 km determined by ordinary railway or highway to 150 km. Agglomeration and competition effects of “one-hour economic circle” promoted corporate development within this sphere. Sample-specific regression results show that railway speed-up has significant effects on the technology of private and export-oriented firms . Technology progress of firms benefits from the “one-hour economic circles” of central cities connected through high-speed railway or metro. Alarmingly, the mtec of firms does not exhibit a positive effect consistent with mtc. This implies that while railway speed-up promotes corporate technology, it can also bring about some negative effects on firms, especially in terms of efficiency improvement..

5. Conclutions

Railway speed- up is an important event for China’s transportation infrastructure quality improvement. Specifically, the EMU trains launched during railway speed-up and the rapid development of high-speed trains are milestones in the history of China’s railway development, and profoundly

influenced its economic development trend over recent years. This paper regards railway speed-up as a quasi-natural experiment in China’s transportation infrastructure quality improvement. With EMU train stations as grouping criterion, we estimates the firm’s TFP using the data envelopment method and then examines the TFP effects of railway speed-up using the difference-in-differences method. Finally, we further decomposesthe TFP into two influence factors as technology change index and efficiency change index. Our study results reveal that railway speed-up has significantly promoted the technology of firms in relevant regions, but its efficiency improvement effect is insignificantly positive. Their coupled effect has promoted corporate TFP growth and is of more positive significance for private firms, firms in coastal regions and export firms. Despite the different effects of railway speed-up on corporate TFP due to heterogeneity of regions and firms, these different effects do not create any significant shocks to our core conclusions. According to the regression result based on the geographical sphere of firms determined by their distance to high-tech cities, railway speed-up has the most significant effect within the “one-hour economic circle”.

Despite high costs of railway speed-up and high-speed railway construction and criticism under the fiscal budget constraint, railway infrastructure quality improvement has positive and sufficiently verified effects on macroeconomic development and corporate TFP. Railway infrastructure quality improvement remarkably shorten the real distance between regions, accelerate the cross-regional flows of passengers, logistics and information, resource allocation, and increase market competition. Currently, significant disparities in transportation infrastructure still exist across various regions in China. Railway speed-up and high-speed railway construction remain concentrated in China’s eastern and central regions. As a result, not all regions in China can benefit from railway speed-up. In future planning, attention should be given to the quality of railway infrastructure in China’s less developed regions. Reasonable planning and infrastructure construction guidance should be offered to promote the positive effect of infrastructure quality improvement and reduce development disparities across regions. Considering the positive significance of the “one-hour economic circle” for productivity growth, priority should be given to the development and improvement of high-speed dedicated passenger lines and intercity railways to enhance the spillover effects of central cities on neighboring cities.