As a typical transportation infrastructure, CRE is a cross-border land transportation channel built under the leadership of the Chinese government and has similar characteristics of a public good as HSR. CRE is a transnational cargo transport agreement, a new framework for international cargo transport cooperation built on the existing railroad network, in contrast to HSR for passenger transport (Zhou and Zhang, 2021). CRE has a significant transportation cost advantage, as Jiang et al. (2018) found that the average shipping time from a Chinese central city to a Western European port city is four times longer than that of CRE transportation. In the background of global “carbon neutrality,” with its unique cost advantage and railroad network, CRE replaces the freight volume originally transported across the border by road, water and air. It also integrates “energy saving and emission reduction” into transportation. Especially for Chinese inland cities, the opening of CRE makes up for the inherent disadvantage of location, and allows goods and commodities to be transported directly across the border. It reduces the transit process of other types and enables the gradual transformation of the local low-carbon economy.

Whether rail transportation is low-carbon is a topic of academic debate. For example, Liu et al. (2015a) compared the carbon emissions of each transportation mode in China in 2012. He found that CO₂ emissions from roads, waterways, and air reached 65.21%, 29.07%, and 3.61%, respectively, while CO₂ emissions from railroad transportation accounted for only 2.11%. Meanwhile, Li et al. (2021) also found that the share of railroad carbon emissions in total transportation carbon emissions in China in 2019 was only .68%, much lower than the share of carbon emissions from roads (86.76%), waterways (6.47%), and aviation (6.09%). In addition, Li et al. (2019) found that transportation carbon emissions from roads, waterways, and aviation in China have increased, while rail transportation carbon emissions have decreased, with roads being the main source of transportation carbon emissions growth in China. It can be seen that railroads play an important role in reducing overall greenhouse gas emissions. It is more environmentally friendly than other modes of transportation. (Baker et al., 2010). As a result, rail has emerged as the alternative for moving freight that is least harmful to the environment. It has significantly decreased direct carbon emissions thanks to the rise in freight orders from the CRE. This means that when CRE opens, the drop in non-rail freight also means a significant drop in carbon emissions. In conclusion, we believe that CRE opening reduces urban carbon emissions through the transportation substitution effect.

As a cross-border land transportation channel and international trade cooperation mechanism, CRE compresses time and space costs, and influences industrial structure adjustment. The theory of new economic geography states that CRE encourages the development of the opening city to take advantage of the core region’s comparative advantage by lowering the cost of factor transportation and luring in new sources of factor resources. It also encourages the improvement of the industrial structure under the influence of circular accumulation (Li et al., 2021b). The industrial structure has been improved and developed in the direction of services at this time, and services like transportation, wholesale and retail, and business services—all of which are directly tied to CRE—have developed quickly. Therefore, CRE promotes the transformation and upgrading of industrial structure with liner transportation and deepens bilateral industrial cooperation. Then it promotes international production capacity cooperation, promotes trade and investment integration, and realizes the reform goal of industrial upgrading and optimization.

Furthermore, industrial structure adjustment is a key factor in carbon emissions. The consumption of energy and natural resources decreases along with the proportion of the industry, which in turn lowers carbon emissions. High-energy and high-pollution businesses are gradually removed and transformed as China’s regional economic development, green technology levels, and environmental legislation advance, and industrial restructuring aids in hastening the attainment of China’s carbon emission intensity targets. Zhang et al. (2018) found that carbon emission reductions due to the restructuring of the three major industries in China accounted for 28.22% of the total national carbon emission reductions. There is a significant difference in the contribution of different industrial sectors to carbon emission reduction, with the structural transformation of agriculture and industry playing a positive role in carbon emission reduction. Combined with the region’s industrial advantages, high-value-added and low-pollution services have been developed rapidly, gradually replacing high-pollution industries, thereby reducing pollution emissions (He et al., 2018). In addition, the focus of industrial restructuring is to enhance the added value of products, thereby reducing energy consumption and carbon emission intensity per unit of added value. The increase in the added value of products mainly relies on the development of the producer service industry, which can reduce the production cost of the manufacturing industry, and extend the industrial value chain. It also reduces carbon emissions in the process of promoting the transformation and upgrading of the manufacturing structure (Han and Xie, 2017). In conclusion, we think that the CRE’s opening will encourage the modernization of industrial infrastructure and thus lower urban carbon emissions.

According to the new economic geography, the combination of market size, cross-regional factor flows, and transportation costs determine the concentration of regional economic activities under monopolistic competition (Krugman, 1980; Krugman, 1991). Transportation infrastructure has been shown to lower transportation costs, improve the availability of goods, capital, labor, and other elements, and alter the geographic distribution of economic activities (Donaldson and Hornbeck, 2016; Ahlfeldt and Feddersen, 2018). CRE development must take into account elements like cheap cost and the city’s need for global commodity trading as a new kind of international land transportation. It has a significant influence on where industrial and service firms choose to locate. Relying on cross-border rail networks, the opening of CRE has effectively expanded market boundaries, greatly reduced transportation costs, enhanced the accessibility of goods, and thus changed the economic quality and agglomeration of cities in time and space (Wong et al., 2021). In addition, CRE connects Asia-Europe transportation routes, making the economic gathering and radiation function along the routes stronger and stronger. CRE not only brings convenience and benefits to the people of each country, but also injects new vitality into the development of enterprises and industries in the countries along the routes.

Higher economic density leads to stronger spillover effects of sharing, matching, and learning from agglomeration economies in areas with higher levels of economic agglomeration (Ingstrup and Damgaard, 2013). This leads to an efficient allocation of resources, resulting in higher utilization of human capital and equipment and emission reduction effects. The massive agglomeration of production factors creates conditions for energy saving and emission reduction. On the one hand, the positive externalities and economies of scale of economic agglomeration are increasingly prominent. They are conducive to accelerating technological change and exchange, generating green technology spillover effects, and prompting enterprises to improve their energy-saving and emission-reduction performance. For example, the degree of energy-saving and emission reduction in urban areas with higher economic agglomeration is higher than that in rural areas with lower agglomeration (Glaeser and Kahn, 2010). On the other hand, economic agglomeration facilitates unified environmental management by government agencies and reduces the cost of government environmental regulation of enterprises. It promotes the development of agglomeration of professional environmental protection enterprises, and thus limits the growth of carbon emissions (Han et al., 2018; Shao et al., 2019a). In addition, Yu et al. (2020) empirically showed that urban agglomeration activities significantly reduced carbon emissions, i.e., for every 1% increase in urban population, carbon emissions would be reduced by .22%. Thus, CRE opening promotes economic agglomeration, which in turn leads to carbon reduction.

The CRE is a new international land transportation corridor that opens up new opportunities for cross-border freight collaboration on the Eurasian continent and considerably boosts the development of commodity import and export in Chinese cities (Zhou and Zhang, 2021). To promote trade facilitation and increase the gravitational pull of the bilateral market, CRE discovered the need for “point-to-point” transit between opening cities and destination nations. Compared with road, air, and sea, CRE has the advantages of low-cost and stable transportation, which attracts a large number of cross-border commodities and cargoes. CRE can reduce the uncertainty in transit and maximize trade convenience, to attract a large number of cross-border goods and sources. Fang et al. (2020) found that the opening of CRE would significantly reduce trade costs, unleash the effect of scale economies and external economies in related industries, and improve the competitiveness of export products and expand demand for imported products, thereby expanding the degree of trade openness.

The impact of international trade opening on carbon emissions is complex and two-sided (Shahbaz et al., 2017). On the one hand, increased exports result in higher domestic output, which uses a lot of energy, fossil fuels, and other natural resources and increases carbon emissions (Bosupeng, 2016). On the other side, as imports increase, the manufacturing of domestic products is replaced, which results in a decrease in the resources and energy used in production and a consequent reduction in carbon emissions. Long et al. (2018) found that China’s embodied carbon emissions in exports reached 2318.73 Mt in 2014 compared to 2000, much higher than the 757 Mt of embodied carbon emissions in imports in the same period. This indicates that China was a typical net exporter of embodied carbon. Meanwhile, the expansion of imports has also brought knowledge and technology spillover, prompting domestic enterprises to transform technology and improve production efficiency, and achieve cleaner production (Kozul-Wright and Fortunato, 2012). In addition, Yang et al. (2018) argue that import activities introduce advanced technologies and equipment that contribute to the gradual improvement of labor productivity and resource use efficiency, while the import of resources and energy relieves production pressure, thus reducing carbon emissions in the long run. If the import is more open than the export, trade opening can lower carbon emissions, and vice versa, it can increase carbon emissions. As a result, we think that expanding CRE can increase trade openness, which in turn influences urban carbon emissions.

Figure 1 illustrates the relationship among all the mechanisms. Based on the above mechanism analysis, we propose the following two theoretical hypotheses:

Hypothesis 1:

The opening of CRE significantly reduces urban carbon emissions.

Hypothesis 2:

The opening of CRE reduces urban carbon emissions by exerting the transportation substitution effect, upgrading the industrial structure, enhancing the economic agglomeration effect, and promoting trade opening.

The explained variable C O E   i s measured by the logarithm of the total carbon emissions of each city. It should be noted that China does not currently have official statistics on carbon emissions at the city level, and the accuracy of published data such as energy consumption and CO₂ emission factors is low (Liu et al., 2015b). Meanwhile, most of the current city-level carbon emissions are obtained by the method of back calculation from socio-economic factors (Song et al., 2015), but the obtained measurement results have large errors due to factors such as statistical caliber and data sources. To avoid the above problems that lead to biased empirical results, we apply a top-down approach to measure the total CO₂ emissions of cities at the prefecture level and above (Meng et al., 2014).

The explanatory variable C R E × P o s t is the interaction between the city dummy variable C R E for whether C R E is opened and the time dummy variable P o s t for when C R E is opened. By sorting out the data on CRE opening, we found that 77 cities in China have opened CRE by the end of December 2019. Therefore, we consider the 77 cities as the treated group and the remaining cities as the control group. Considering the opening of some cities in the second half of the year, the initial operation of the opening is less frequent, which may not affect the city’s carbon emissions for a short period. Therefore, we then take June of the actual opening year as the cut-off point, and take the time dummy variable P o s t to be 1 from the current year for C R E opened before June, and P o s t to be 1 from the next year for those opened after June, and 0 for the rest of the time.

Economic development ( E C D ). Increased industrial output can be sparked by higher levels of city economic development, which could also ultimately result in higher carbon emissions (Zhang and Da, 2015). We use the logarithm of GDP per capita to measure economic development, and this variable is expected to increase CO₂ emissions.

Population density ( P P D ). Theoretically, population density may both increase carbon emissions through scale effects (Verhoef and Nijkamp, 2002) and optimize environmental quality through agglomeration effects (Glaeser and Kahn, 2010). Therefore, we use the total population per unit of urban administrative area to express. The direction of the influence remains uncertain.

Human Capital ( H M C ). The development of human capital supports the advancement of green technology, which enhances environmental quality. Limited by the availability of data, we use the number of urban education sector employees as a share of total urban employees to express that the expected sign is positive.

Energy consumption ( E G C ). Fossil energy is the main source of CO₂ emissions. Due to limited data availability and drawing from Sun and Li (2021), we use the logarithm of the total urban society-wide electricity consumption to express. The expected impact is positive.

Environmental regulation ( E M R ). Environmental regulation is an important factor influencing the environmental quality and usually includes nationally uniform environmental regulations (Shao et al., 2011) and municipal environmental regulations (Bai et al., 2018). As a result, we build a new environmental regulation variable that accounts for both national and local regulations. To be specific, we use the dummy variable of China’s energy conservation and emission reduction policy implemented in 2006 to denote the national-level environmental regulation E M R 1 , and the average of the comprehensive utilization rate of urban general industrial solid waste, sewage treatment rate, and domestic waste treatment rate to denote the city-level environmental regulation variable E M R 2 . Then, we use the interaction term of E M R 1 and E M R 2 to represent the environmental regulation with a positive expected impact.

Foreign investment ( F D I ). FDI has more complicated effects on the environment because it is a significant indicator of economic openness. There are two hypotheses for the relationship between the two: “pollution halo” and “pollution haven” (Shao et al., 2019b). Hence, we use the amount of foreign investment per capita to express the expected impact is uncertain.

Transportation alternatives T P S . The establishment of CRE has improved access to international trade in the city, partially replacing the amount of freight moved via roads, waterways, and the air. Therefore, we use the logarithm of non-rail freight per capita to express the transportation substitution effect, where non-rail freight includes roads, waterways, and air.

Industry structure ( I D S ). The opening of CRE results in low-carbon development while also changing the status of the city’s industrial structure and maximizing output. Normally, a decrease in the share of secondary industry, which is mainly industry, can effectively reduce carbon emissions (Sun and Li, 2021). Hence, we use the value added of the secondary sector as a share of GDP to measure the industry structure.

Economic Agglomeration ( E C A ). In order to assist the decrease in carbon emissions, the openness of CRE has resulted in the mobility of components like labor, capital, and commodities as well as economic agglomeration within a specific range. Following Combes et al. (2007), we use an employment density indicator for the measure, which is the ratio of the total number of employees in the secondary and tertiary industries to the administrative area of the city.

Trade opening ( T O P ). The opening of CRE promotes the growth of urban import and export and has a differential impact on carbon emissions. To facilitate the identification of such differential impacts, we use the logarithm of the ratio of import dependence degree to export dependence degree to represent the trade opening.

Drawing on Chetty et al. (2009) and Li et al. (2016), we used a placebo test to ensure the robustness of the benchmark result. First, we randomly generate a list of CRE-opened cities and assign the corresponding opening years, thus forming a sample of errors. Then, based on the benchmark model (1), we empirically estimate the above false sample using the DID method, leading to a false estimation coefficient α f a l s e for CRE opening. Finally, we repeat the above process 3,000 times and thus obtain 3,000 false estimation coefficients α f a l s e . We next show the distribution of the above 3000 false coefficients α f a l s e and their p-values (Figure 3), where the red dashed line is the correct coefficient α of the benchmark estimate. Theoretically, the coefficient α f a l s e should be close to 0, i.e., the opening of CRE under the false sample cannot significantly reduce carbon emissions. Figure 2 shows that the false coefficients α f a l s e are mainly distributed around the zero value with normal distribution, and the p-values are mainly greater than .1. Although the correct coefficient α of the benchmark regression intersects with the α f a l s e distribution, the percentage of overlap is only .14%, which is consistent with the expected requirements of the placebo test. Therefore, the estimated results of the benchmark are robust and are not adversely affected by other unobservable factors.

There is another prerequisite for the DID estimation method, which is that the selection of CRE opening cities needs to be random. To exclude the interference of sample selection problem, referring to Jia et al. (2021) and Yang et al. (2019), we apply Propensity Score Matching (PSM) method to re-screen the control group and perform DID estimation on the matched samples. First, given that cities open CRE at different times, we use the year-by-year matching method, and use control variables in model (1) as matching characteristic covariates. We then use nearest neighbor matching method to construct control groups for CRE opening cities, and estimate it by using the Logit model and estimate the corresponding propensity scores.

To ensure the validity of the matching results, we perform a balance test. Table 3 shows that each characteristic covariates is significantly different at the 1% level before matching, while there is no significant difference after matching. Therefore, it is reasonable to believe that the matched covariates are appropriate and the results of the nearest neighbor matching method are reliable.

Finally, the results in column (1) of Table 4 are obtained with the support of the PSM-DID method. The estimated coefficient of the matched C R E × P o s t is still significantly negative (−.0076) at the 1% statistical level, which is consistent with the baseline results. Therefore, after excluding the possible self-selection problem of the sample, the opening of CRE significantly reduces urban carbon emissions.

In this section, we change the measures of carbon emissions and CRE opening variables, and re-estimate the empirical result. For the explanatory variables, drawing on Shan et al. (2020) and Guan et al. (2021), we use urban carbon emission data published by the China Carbon Accounting Database (Carbon Emission Accounts and Datasets, CEADs) for the measurement. For the explanatory variables, we no longer distinguish the cut-off point of the opening time, June, but assign a value of 1 to the year of opening and thereafter, and 0 to the rest of the year. At this moment, we could obtain the new CRE opening variable ( C R E × P o s t 1 ). The results in columns (2)–(3) of Table 4 indicate that the coefficients of C R E × P o s t ; C R E × P o s t 1 are both significantly negative at the 1% statistical level (−.0483 and −.0067). Therefore, the benchmark results still hold after replacing key variables.

Some pilot policies on carbon emissions and the opening of HSR may also affect the baseline results (Jia et al., 2021; Sun and Li, 2021). To avoid the interference of other policies and the opening of HSR, we collect some disturbing factors, such as the Fiscal Comprehensive Pilot Policy (FCPP) for energy conservation and emission reduction, Emissions Trading Pilot Policy (ETPP), Low Carbon City Pilot Policy (LCCPP), and the opening of HSR. Then, we include the above-mentioned confounding factors into model (1) for estimation, and the results are reported in column (4) of Table 4. At this point, we find that the coefficient of C R E × P o s t is still significantly negative (-.0067) at the 1% statistical level. This indicates that the opening of CRE still has a significant carbon reduction effect after excluding the interference from other policy and HSR.

The endogeneity problem in this paper arises mainly from the non-random nature of sample selection for the opening of CRE. To further overcome the potential endogeneity problem, we construct an instrument variable and use the method of Two-stage Least Squares (TSLS) to estimate. The corresponding results are reported in columns (5)–(6) of Table 4.

Specifically, referring to Jiang et al. (2018), we use the shortest geographic distance from each city to China’s three major corridor ports (Alataw Pass, Erlianhaote and Manzhouli) of the CRE as the instrumental variable (IV). In order to realize the dynamic change of the instrumental variable, following Nunn and Qian (2014), we use the product of the above shortest geographic distance and the national railroad freight volume in the previous year to construct a time-varying instrumental variable. Then, in terms of the rationality of IV, on the one hand, the above three continental border crossings are closely related to the ancient Silk Road and the CRE operation channel, and the closer the distance to the crossings, the more competitive the location characteristics and operation costs are. Then, those cities are more likely to open CRE, thus satisfying the correlation requirement. On the other hand, geographical distance is a typical natural factor, not affected by economic and social factors, and can only affect the carbon emission status by influencing the CRE opening status (such as transportation distance and time cost, etc.). Meanwhile, the rail freight volume at the national level in the previous year is independent of the carbon emission of each city in the current period, so the requirement of exogeneity is satisfied.

As shown in columns (5)–(6) of Table 4, the first stage results show that the estimated coefficient of IV is significantly negative at the 1% statistical level, indicating that there is a significant negative correlation between IV and the opening of CRE, which is in line with theoretical expectations. The second stage results show that the endogeneity test results strongly reject the null hypothesis that C R E × P o s t is an exogenous variable at the 1% significance level. Thus necessitating the use of instrumental variables estimation to overcome the endogeneity problem. The Minimum Eigenvalue Statistics of Stock and Yogo (2002) is 39.319, which is much higher than the threshold value (16.38) at the 10% statistical level, and strongly rejects the null hypothesis of “weak instrumental variable”. Meanwhile, the first-stage F value of 16.35, which is greater than 10, indicating the absence of weak instrumental variables. The Kleibergen-Paap rk LM test strongly rejects the null hypothesis of “unidentifiable instrumental variables” at the 1% statistical level, indicating that the selected instrumental variables satisfy the requirement of correlation and there is no requirement of weak instrumental variables. At this point, the coefficient of C R E × P o s t is significantly negative at the 5% statistical level (-.1360), which is larger than the benchmark coefficient in column (5) of Table 2 (-.0078). This may be due to the instrumental variables address the non-randomness of transportation infrastructure construction in real situations, and thus the coefficient of the potential transportation facility variable has increased (Duranton and Turner, 2012). Hence, after overcoming the endogeneity problem, CRE opening still significantly reduces carbon emissions, and thus we consider the baseline results robust.

In order to confirm the mechanism, referring to Baron and Kenny (1986), we construct a three-step mediating effect model for mechanism testing. The model is as follows. C O E i t = c 1 + α 1 C R E i × P o s t t + β 1 Z i t + γ 1 i + θ 1 t + ϵ 1 i t (3) M E D i t = c 2 + α 2 C R E i × P o s t t + β 2 Z i t + γ 2 i + θ 2 t + ϵ 2 i t (4) C O E i t = c 3 + α 3 C R E i × P o s t t + π M E D i t + β 3 Z i t + γ 3 i + θ 3 t + ϵ 3 i t (5) where M E D i t represents the mediating variables, including transport substitution T P S , industrial structure I D S , economic agglomeration E C A and trade opening T O P , and π is the estimated coefficient of the corresponding mediating variable. α 1 is the total effect of the opening of CRE on carbon emissions, α 2 is the effect of the opening of CRE on mediating variables, and α 3 is the effect of the opening of CRE on carbon emissions after controlling the mediating variable M E D i t . If α 2 , α 3 and π pass the significance test and α 3 is smaller than α 1 , then the mediating effect holds and the mediating variable plays a partially mediating role.

Based on the models (3)–(5), we report the results of the mechanism tests in Tables 5; Table 6, where column (1) is the baseline estimation result. For transport substitution T P S , the coefficient of C R E × P o s t in column (2) of table (5) is -.1170 and significant at the 10% level, indicating that the opening of CRE significantly reduces the volume of non-rail freight, i.e., reduce the volume of domestic road, waterway and air freight. The coefficient of T P S in column (3) is significantly positive (.0075) at the significant level of 1%, indicating that the increase in non-rail freight traffic significantly increases carbon emissions. Meanwhile, the coefficient of C R E × P o s t is significantly negative at 1% level (−.0076) and the absolute impact is smaller than the baseline coefficient of column (1) (−.0078). Thus, the transport substitution plays a partially intermediary role, with the opening of CRE replacing non-rail freight traffic to reduce carbon emissions. As one of the railroad transportation ways with cost advantages and low-carbon green advantages, the opening of China CRE provides a new channel for urban cargo transportation and cross-border transactions, replacing traditional non-low-carbon transportation type of transportation (such as road transportation), which helps achieve the goal of carbon emission reduction.

For industrial structure I D S , as shown in column (4) of Table 5, the coefficient of C R E × P o s t is significantly negative (-.0054) at the 10% statistical level, indicating that the opening of CRE significantly reduces the share of output in the secondary industry, which leads to the upgrading of the industrial structure. Meanwhile, the coefficient of C R E × P o s t is significantly negative at 1% level (−.0074) and the absolute impact is smaller than the baseline coefficient of column (1) (−.0078). Therefore, we believe that industrial structure plays a partly intermediary role, and the opening of CRE reduces urban carbon emissions by promoting industrial structure upgrading. In fact, the normal operation of CRE promotes the rapid development of urban transportation, wholesale and retail, business services and other service industries. This can accelerate the upgrading of service industrial structure, which can transform the traditional industrial structure with high energy consumption and pollution, and ultimately reduce carbon emissions.

For economic agglomeration E C A , the coefficient of C R E × P o s t in column (2) of Table 6 is significantly positive (.022) at the 1% level, indicating that the opening of CRE significantly increases the level of urban economic agglomeration. Column (3) shows that the coefficient of ECA is significantly negative (−.0553) at the 1% level, indicating that urban economic agglomeration significantly reduces carbon emissions. Also, the coefficient of C R E × P o s t is significantly negative (−.0066) at 1% statistical level and the absolute impact is smaller than the coefficient of column (1) (−.0078). Therefore, the partial mediating effect of economic agglomeration holds, which means that CRE opening reduces urban carbon emissions by promoting economic agglomeration. Relying on the ever-improving railroad network, the opening of CRE changes the spatial layout of urban economy and production factors. This can create increasingly significant positive externalities and economies of scale, which is conducive to achieving green technology spillover effects and prompting enterprises to improve their ability of energy saving and emission reduction performance.

For trade opening T O P , column (4) of Table 6 shows that the coefficient of C R E × P o s t is significantly positive (.1250) at the 10% level, indicating that the opening of CRE significantly promotes the degree of trade openness of the city. The coefficient of TOP in column (5) is significantly negative (−.0019) at 1% level, indicating that the increase in openness of import can effectively reduce carbon emissions. Moreover, the coefficient of C R E × P o s t is significantly negative at 1% statistical level (−.0076) and the absolute impact is smaller than the coefficient of column (1) (−.0078). Thus, the opening of CRE increases urban carbon emissions by increasing the degree of trade opening. There is no doubt that the opening of CRE provides a new platform for the growth of China’s foreign trade and accelerates the increase in the level of trade opening. Compared to export openness, increasing import opening reduces domestic energy and resource consumption. At the same time, the import also brings spillover of foreign knowledge and technology, prompting domestic enterprises to transform technology and improve production efficiency to achieve low-carbon and clean production.

In summary, we find that the opening of CRE reduces urban carbon emissions through four mechanisms: exerting the transportation substitution effect, promoting industrial structure upgrading, strengthening economic agglomeration, and promoting trade opening. Therefore, the estimation results of the mediating effects model validate the second theoretical hypothesis of this paper.

Referring to Sun and Li (2021), we divide the provinces where the sample cities are located into three major regions, and then construct three location dummy variables for whether the cities are in the eastern region E R , the central region C R , and the western region W R . Then, we construct three interaction terms C R E × P o s t × E R , C R E × P o s t × C R , and C R E × P o s t × W R , and take them into the benchmark model (1) for estimation, respectively, and the empirical results are reported in columns (1)–(3) of Table 7. It can be seen that the coefficient of C R E × P o s t × E R is significantly negative (−.005) at the 5% level, the coefficient of C R E × P o s t × C R is significantly negative (−.005) at the 10% level, while the coefficient of C R E × P o s t × W R does not significant. Hence, the opening of CRE significantly reduces carbon emissions in eastern and central cities, but cannot effectively reduce carbon emissions in western cities. This may be due to the relatively strict environmental regulations in the central and eastern cities compared to the western cities, and the greater number of cities opened by the CRE, which produce more significant carbon emission reduction effects through transportation substitution effect, industrial structure upgrading effect, economic agglomeration effect and trade opening effect.

Based on the 2019 China City Level, we construct the China city level C C L as a dummy variable, which takes the value of 1 when the city belongs to Tier 1, New Tier 1 and Tier 2 cities, and 0 when it belongs to Tier 3, Tier 4 and Tier 5 cities. Then, we include an interaction term of CRE opening and C C L in the benchmark model (1) for estimation, and the empirical results are reported in column (5) of Table 7. The C R E × P o s t × C C L coefficient is significantly negative (-.015) at the 10% level, indicating that the opening of the CRE effectively reduces carbon emissions in larger cities. Generally speaking, larger cities have stronger economic agglomeration effects, industrial structure upgrading effects and technological innovation advantages, which are conducive to optimizing resource allocation to alleviate congestion and achieve energy saving and emission reduction (Yang et al., 2019; Jia et al., 2021). By now, compared to smaller cities, the opening of the CRE is effective in reducing carbon emissions through the advantages of larger cities in terms of agglomeration, structure and innovation. Another point to note is that although larger cities usually have more obvious agglomeration effects, they also face more traffic congestion and environmental pollution problems. With the effect of transportation substitution and economic agglomeration, the opening of CRE can effectively optimize the agglomeration effect and alleviate the congestion effect. And CRE can optimize resource allocation and accelerate technology spillover, thus achieving energy saving and emission reduction. Therefore, the opening of CRE can be more conducive to curbing carbon emissions in larger cities.

Based on the National List of Resource-based Cities published by the State Council of China, we construct the resource-based cities R B C as dummy variable, we take 1 when the city is a resource-based city and 0 otherwise. Next, we include the interaction terms R E × P o s t × R B C in the benchmark model (1) for estimation, and the empirical results are reported in column (5) of Table 7. The C R E × P o s t × R B C coefficient is significantly positive (.007) at the 10% level, indicating that the opening of the CRE significantly contributes to carbon emissions in resource-based cities, so it effectively suppresses carbon emissions in non-resource-based cities. Resource-based cities often rely heavily on resource development and heavy industrial processes, which lead to higher CO₂ emissions and thus face the serious “carbon curse” (Friedrichs and Inderwildi, 2013). Therefore, compared with non-resource-based cities, resource-based cities are mainly dominated by traditional and resource-driven heavy industries, while the development of modern service industries is relatively lagging behind, resulting in the opening of CRE cannot reduce carbon emissions by industrial upgrading.