Has the belt and road initiative boosted the resident consumption in cities along the domestic route? – evidence from credit card consumption

The orderly promotion of the B&R Initiative not only brings countries along the route closer to China (Liu, H, 2016; Du, J and Zhang, Y, 2017; Zhang Yansheng, Wang Haifeng, Yang Kunfeng, 2017) but also has a significant driving effect on the domestic regional economy and industrial economy (Zhang Jun, 2014; Yang Jirui and Luo Zhigao, 2017; Liu Zhen, 2017). Related research can be roughly divided into two directions. First, the positive directions include the construction of global value chain (Chen Jian and Gong Xiaoying, 2018), the distribution pattern of resources (Zhao Yabo, Liu Xiaofeng and Ge Yuejing, 2017), the international management methods (Fang Hui and Zhao Tian, 2017)), the location advantages (Sun Yanbo, 2017) and the cultural communication (Sun Ying, 2017). Second, the challenging directions include the trade protection (Yanwen, Wu, 2017), the financial risks (Li, W and Jin, D, 2018), the overseas political risks (Tang Lizhi and Liu Yu, 2017) and the environmental effects (Liu Naiquan and Dai Jin, 2017).

The B&R Initiative has become one of the hot topics in economic research in the past 3 years. The existing literature has conducted research from both international perspective and domestic perspective. In terms of domestic research, it can be divided into three levels, namely, specialised research involving key cities (He Yaoyin, 2015; Gao Xinfai and Yang Fang, 2015; Zhang Qi and Tong Jixin, 2016), provinces (Huang Maoxing and Ji Peng, 2015; Lu Wengang, Huang Xiaozhen and Liu Pei, 2015; Ouyang Lin et al., 2017) and inter-provincial economic belts or urban cluster (Zhou Huan and Ma Naiyi, 2016; Lu Dadao, 2017; Qiu Shan, 2017). Based on the above research results, the selection of regions or industries is equivalent to correctly identifying, inheriting and applying the classical economic theories that have advantages and conform to the economic cycle.

With the continuous promotion of the B&R Initiative, the accumulation of relevant data has become more and more abundant. The research on the B&R Initiative is no longer limited to the model (Dong Suocheng et al., 2014), the logic (Lu Feng et al., 2015) or the basic architecture (Liu Qingcai and Zhi Jichao, 2016), but the in-depth measurement (He Chun et al., 2017), the empirical research (Sun Yanbo, 2017) and even the financial products designed based on data (Li, W and Jin, D, 2018). The research from the perspective of urban economics is as follows. Xiaofeng, Li and Song Qi (2016) studied and compared the advantages of service industries in cities along the route of the B&R Initiative. Deng Jian (2017) proposed five major development measures for processing trade in border cities along the route of the B&R Initiative. Wang Lili and Xiao Wenwen (2018) studied the spatial network structure of the urban agglomeration along the B&R, which consists of five provinces in the northwest and four provinces in the southwest. Taking the China-Europe train as the research object, Wang Dongfang et al. (2018) studied the network structure of node cities along the route. The research from the perspective of industry research is as follows. Cai Zhonghua, Wang Yifan and Dong Guangwei (2016) studied the patent layout along the route of the B&R Initiative in several industries such as special (general) equipment manufacturing and chemical raw materials (products). Hu Angang et al. (2017) investigated eight representative central enterprises in the industry and provided solutions to common industry problems encountered in the construction of the B&R.

Through the above-mentioned literature review, notwithstanding the fact that it is based on the cross-sectional view of cities or regions, or the in-depth profile of industries, the research on the mechanism and effect of the B&R Initiative's driving economic development along the route has become increasingly abundant, and the research conclusions are fundamentally the same. According to the theoretical and practical results, there is no doubt that the B&R has made remarkable achievements, which also embodies the national will to explore the market vitality in the region and expand the opening-up to the outside world. However, the microscopic and detailed construction effect and empirical support still need to be further enriched.

According to the trend of the B&R, with the support of central cities along the route and key maritime ports, it aims to jointly create efficient transportation channel, international boutique tourism routes, and new tourism products to make the B&R node cities in China become commercial logistics hubs, key industries and cultural exchange bases. Under this guidance, the overall economic development of the cities along the route in China has been improved, and the consumption of livelihoods such as clothing, food, housing and transportation by residents of the cities along the route has also been driven, such as the new retail model arising from the B&R (Wang Juanjuan, 2017), the new processing trade strategy in border cities (Deng Jian, 2017) and the “Silk Road Culture” of cities (Liao Qinghu et al., 2017). Therefore, it can be inferred that the B&R not only promotes the prosperity of trade but also focuses on certain industries in the cities along the domestic routes. Based on the above-mentioned analysis, this article puts forward the following research hypothesis: The B&R Initiative can drive the economic development of domestic cities along the route. However, the driving effect on different industries may be heterogeneous.

The construction of the B&R Initiative is both contemporary and regional. To control the grouping effect between node cities and non-node cities as well as the time effect of the B&R Initiative, the double difference method is adopted to examine the influence of the B&R Initiative on domestic node cities. The model is set as follows: (1)consit=β0+β1Treati×OBORt+β12Treati+β13OBORt+β2GDPpcit+β3Exportit+β4FDIit+εitcon{s_{it}} = {\beta _0} + {\beta _1}Trea{t_i} \times OBO{R_t} + {\beta _{12}}Trea{t_i} + {\beta _{13}}OBO{R_t} + {\beta _2}GDPp{c_{it}} + {\beta _3}Expor{t_{it}} + {\beta _4}FD{I_{it}} + {\varepsilon _{it}} where i and t represent the city and year respectively, and cons represents the consumption of local residents. Treat is used to distinguish whether the B&R node cities, where Treat=1 means node cities, and Treat=0 means non-node cities. OBOR stands for the time node proposed by the B&R Initiative. Since President Xi Jinping proposed the strategic concept of the B&R Cooperation initiative in 2013, 2014 is set as the time node, so the variable OBOR before 2014 is 0, and the variable after that is 1. The coefficient of the cross-term Treat×OBOR is the focus of attention. The B&R construction has promoted the consumption growth of the node cities in China. GDPpc is the local GDP per capita; export is the total local exports and FDI is the actual foreign investment. The above-mentioned three variables control the impact of per capita economic output, trade volume and investment on local consumption.

There are two main types of data used in this article. One is the micro-transaction data representing the consumption level and consumption categories of various domestic cities, and the other is the data representing the economic profile of these cities, such as economy, foreign trade and FDI. Among them, urban micro-transaction data refer to the data of bank card transaction and settlement after desensitisation. In the micro-transaction data with a total of more than 23 billion each year, each transaction contains more than 200 elements of variable information, such as transaction time, location, transaction amount, consumer merchant category, issuing bank and acquiring bank. This article collects and desensitises data and further classifies them by city and industry based on these data. In terms of industry classification, the seven major industries, namely clothing, food, housing, transportation, real estate, financial services and building materials, which are closely related to the lives of the general public, have been counted, covering the livelihood consumption field. In more detail, clothing refers to the shopping consumption in shopping malls, clothing stores and other places; food refers to the relevant consumption occurring in restaurants, bars, banquets and other dining places; accommodation refers to the consumption related to the accommodation places such as hotels and hotels as well as the service consumption of villas and vacation rooms; line refers to the consumption occurring in all kinds of gas stations; real estate refers to the new house purchase transaction of urban residents along the route; financial services refer to service consumption related to withdrawal and purchase of financial service products occurring at ATM and counter of financial institutions. The consumption of building materials refers to the consumption behaviour of all kinds of building materials wholesale market in the process of improving the housing quality.

The data representing the economic profile of cities are obtained from the Statistical Yearbook of Chinese Cities from 2012 to 2017, and the total export data come from China Regional Economic Statistics Yearbook. This article not only supplements some missing data by consulting various regional statistical yearbooks and government work reports but also compares and verifies the data from the CSMAR database, China Statistical Yearbook and National Prefecture-level City Financial and Economic Statistics to ensure the reliability of the data.

Of course, in addition to the two types of data mentioned above, the dummy variables constructed in this article rely on official documents related to the B&R Initiative, such as Will and Action, which are manually collated and encoded into the existing database. The total collected data from 291 domestic cities, with the time span from 2011 to 2016, covers the economic events that have immense influence on people's livelihood and consumption in China's economic development, including from rapid growth to slower growth, from total expansion to structural transformation, anti-corruption, monetary policy that changes from loose to tight and then to loose, the “Internet +” in industrial upgrading and the rise of the big data industry. A total of 2328 valid samples are obtained after eliminating the error values.

Descriptive statistics for 2328 sample observations are listed in Table 1. Due to space limitations, this article cannot list the consumption data of each industry in detail. As shown in Table 1, the average value of household consumption is 37.99 billion yuan, and the maximum value is 119.543 billion yuan. According to the estimation of the 24 million permanent residents of Shanghai in 2015, the average annual credit card consumption per person is close to 50,000 yuan, with a minimum of 155 million yuan. According to an estimate of the 700,000 population of Changdu in the Tibet Autonomous Region in 2011, the average annual credit card consumption per person is only more than 200 yuan. The gap between the two cities is very wide. In addition, according to the value of the standard deviation, it can be seen that the dispersion of consumption values between different cities is relatively large. Among them, the city with the highest per capita GDP is Ordos, which is as high as 215,500 yuan. The city with the largest export volume is Shenzhen, which reached 1.57 trillion yuan in 2016. Tianjin ranks first in terms of foreign direct investment.

To observe the impact of the B&R Initiative on the local resident consumption more clearly, the test results of the difference in local resident consumption before and after 2014 are statistically analysed, as shown in Table 2. As can be seen from Table 2, the average value of credit card consumption of urban residents in the treatment group before 2014 was 97.51 billion yuan, but increased to 204.509 billion yuan after 2014, with the F value of 4.55. In other words, the credit card consumption of urban residents in the treatment group increased significantly after the B&R Initiative was proposed, while there was no significant difference in the control group. This preliminary statistical result is consistent with the hypothesis of this article.

Table 3 reports the regression results of the B&R initiative on local resident consumption. It should be noted that the data of local consumption, GDP per capita, export trade volume, and FDI were logarithmised to make the data more stable.

Column (1) of Table 3 is the estimated result of dummy variables. The coefficient of cross-term Treat×OBOR is 1.055, which is significant at the level of 1%, indicating that the B&R Initiative has a significantly higher driving effect on consumption in cities along the B&R Initiative than in cities outside the route. Column (2) is the re-estimation after adding control variables such as local per capita GDP GDPpc, total local exports, and actual foreign investment FDI. Among them, there is a significant positive correlation between per capita GDP, total local exports and FDI and local resident consumption. The cross-term Treat×OBOR coefficient becomes 0.617, which is slightly smaller than before but still significant at the level of 1%, indicating that the measurement pattern has no significant change.

Considering that the driving effect of the B&R construction on residents’ consumption in cities along the line may vary with different industries, this article divides the sub-samples according to consumer consumption data classified and summarised by industry to re-examine the policy effect. As shown in Table 4, columns (1) to (7) respectively show the regression results of sub-samples of resident consumption in the seven industries of clothing, food, housing, transportation, real estate, financial services and building materials.

According to the regression results mentioned above, the B&R Initiative has different driving effects on consumption in different industries. In detail, apart from the insignificant effect of the B&R initiative on real estate, it is significant in the six industries of residents’ clothing, food, housing, transportation, financial services and building materials, among which the significance level of the clothing industry is slightly lower. The possible reason for the insignificance of the real estate industry is that the real estate investment boom is sustained and widespread, which does not show a structural change due to the B&R Initiative. From the perspective of economic significance, among the six industries, the B&R Initiative has the greatest impact on the travel of urban residents along the route, while the impact on clothing and dining consumption is relatively small. Regardless of industries, the three control variables have a significant positive correlation with the consumption of residents in cities along the route.

Considering that different levels of local economic development may have different sensitivities to the construction of the B&R Initiative, all cities are divided into two groups based on the total GDP. In addition, a grouping dummy variable GDP50 is constructed, in which the top 50% cities set GDP50 as 1, and the bottom 50% cities set GDP50 as 0. Table 5 shows the estimated results of policy effects under the different levels of economic development.

Regression results show that grouping does not change the significance of the Treat×OBOR coefficient. As for the economic grouping variable GDP50, all the six industries except travel industry are significantly positive, which indicates that the higher the level of economic development, the greater the promotion effect of this initiative on resident consumption. The insignificance of travel indicates that the travel livelihood consumption is not sensitive to the level of economic development. On the whole, the more developed a city is, the more likely its local resident consumption in various industries is to be positively affected.

To ensure the robustness of the estimated results, the model is re-estimated by replacing the consumption amount with the transaction amount of domestic urban resident credit card consumption. The setting form of the model and the processing method of the data remained unchanged. The robustness test for the baseline regression is shown in Table 6. Column (1) is the result of only estimating the dummy variables. The coefficient of the cross-term Treat×OBOR is 1.042, which is significant at the level of 1%. Column (2) is the re-estimate after adding control variables such as local GDP per capita (GDPpc), local export volume (Export) and foreign actual investment volume (FDI). The cross-term Treat×OBOR coefficient is 0.674, which is significant at the level of 1%. The levels of significance and measurement patterns are consistent with the results in Table 3, which shows that the estimated results are robust.

For the sub-samples of different industries, it is also re-estimated by replacing the consumption amount with the transaction amount of domestic urban resident credit card consumption. The results are shown in Table 7. Columns (1)–(7) respectively show the regression results of the sub-samples of resident consumption in the seven industries of clothing, food, housing, transportation, real estate, financial services and building materials. Compared with Table 4, it is found that the policy variables have a significant effect on the consumption of real estate industry of urban residents along the route, but only at the edge of 10% level, which may be the reason for the data. The revaluation results of other industries are consistent with the previous ones, which show that the model is robust.

Furthermore, the city groups are re-divided into three groups according to the economic volume of the city. In addition, two grouping dummy variables GDP35 and GDP70 are constructed. Among them, the GDP35 of the top 35% cities is 1, and the others are 0. The top 70% of cities have a GDP70 of 1 and the rest have a GDP of 0. Table 8 lists the estimated results of policy effects under different economic development levels. The regression results show that all the six industries except travel are significantly positive. On the whole, the higher the degree of economic development, the more likely the local resident consumption in various major industries is to be positively affected. After grouping, the significance of Treat×OBOR variable coefficients does not change, and there is no deviation in the direction, which is consistent with the estimation in Table 5.