Global Super-Efficiency SBM Model

We use the global super-efficiency SBM-DEA to measure the GEE with reference to Tone [30] and Chen and Yao [29] as follows:

Assume that there are k decision-making units (DMUs), which means that k cities are assumed in this study. For each DMU, there exists an input, which can be expressed as x = (x₁, .... x_N) ∈ R_N⁺. There are also M desirable outputs and I undesirable outputs, which satisfy y = (y₁, .... y_M) ∈ R_M⁺ and c = (c₁, .... c_I) ∈ R_I⁺, respectively. In general, delineating production boundaries through cross-sectional data can provide insight into the efficiency of a given unit, but it can complicate longitudinal comparability. Given that DMUs located at the production boundary have an efficiency score of 1 on their own, it is important to ensemble the super-efficiency paradigm to differentiate between optimally performing DMUs. Therefore, to ensure that efficiency metrics are consistent across cross-sections and DMUs, we attempt to construct a vertex boundary from the perspective of global DMUs. In the observation horizon of this study, the metrics for k cities are denoted as (x_k, y_k, c_k). In a closed, finite production horizon, the principles of free disposal, zero portfolio, and weak disposal of assets are followed. We utilize data envelopment analysis (DEA) to construct the model:

Fig. 1.

Research framework.

where z_k denotes the weights of the global observations, which tend to be 1 and are always positive. If the weights are not combined into a single instruction, this framework can be characterized as scale reward invariant in a production environment. Using the SBM distance function, we carve an efficiency profile for each DMU based on the characteristic of constant returns to scale:

In Equation (2), s^x ∈ R^N, s^c ∈ R^I, and s^y ∈ R^M reflect the surplus of inputs and undesirable outputs and the shortfall of desirable outputs, respectively. When ρ = 1, i.e., s^x = s^y = s^c = 0, DMU is optimally efficient. Overall, the above equations designate the DMU on the boundary as the most efficient unit. We adopt a super-efficiency calculation method to dissect the relative efficiency among DMUs.

Equation (3) takes ρ>1 as an optimization condition, i.e., DMU efficiency is optimized. Overall, this equation designates DMU on the boundary as the most efficient unit. In the above analytical framework, the efficiency indicator ρ≥1 is determined by non-expected super-efficiency SBM. To measure the GEE under the condition of variable returns to scale, condition ∑k=1Kzk=1 should be added to the third sub-equation in Equation (3).

Variable Selection and Data Source

Therefore, EPHD in the Yellow River Basin involves many aspects, such as resource utilization, ecological protection, and economic growth. Drawing on related studies [29, 54-56], this study constructs an indicator system for measuring the GEE, as shown in Table 1.

Multi-Indicator Comprehensive Evaluation Method

This study continues to define and measure GF, using a multi-indicator comprehensive evaluation method to assess GF development using 103 cities in nine provinces along the Yellow River as a research sample. Specifically, for city i, the green finance index GF_i is used to measure GF, and the calculation model of GF_i is:

where n is the total number of indicators, ω_k is the weight of indicator k (∑k=1nωk=1 in this study), and z_ik is the value of second-level indicator k assessing green financial development level after standardization and normalization.

Variable Selection and Data Source

Based on the principles of scientific, comparable, and accessible indicator selection, we select seven dimensions of green credit, investment, insurance, bond, support, fund, and equity to construct an indicator system for GF from the perspective of the green financial market. Table 2 shows the specific indicators for assessing green financial development.

Fixed Effect Model

The traditional OLS estimation will ignore the effects of both inter-cross-section and time trends in panel data, while the fixed effect model can consider both aspects at the same time. Therefore, to explore the direct impact of GF on EPHD of the Yellow River Basin, this study constructs a fixed effect model:

where GEE_i,t denotes the GEE of city i in year t, and GF_i,t is GF of city i in year t. The variable μ_i denotes a city-fixed effect, which is used to measure all influences that do not change over time. ν_t denotes time-fixed effect, which is used to measure influences that only change over time and are unrelated to the city. The variable X_i,t denotes control variables, and ε_it is a random disturbance term.

Mediation Effect Model

To further verify the impact path of GF to support EPHD in the Yellow River Basin, a mediation effect model is constructed:

where Media_i,t denotes the mediating variable. Mediating variables include green innovation (GI), environmental regulation (ER), and industrial structure (IS). The mediation effect test needs to satisfy these conditions [57]: 1) The coefficient of β₁ in Equation (5) is statistically significant. 2) The absolute value of θ₁ in Equation (7) is smaller than β₁ in Equation (5) in the case that both α₁ in Equation (6) and δ in Equation (7) are significant.

Variable Selection and Data Source

(1) Explained variable: green economy efficiency (GEE). GEE reflects both the region’s economic and environmental development. Of these, economic development reflects efficiency improvement and technological progress; environmental development reflects green development. Therefore, this study uses the GEE to indicate EPHD in the Yellow River Basin. This study uses the Data Envelopment Analysis (DEA) based on global super-efficiency SBM to measure the GEE, and the resulting indicator is the GEE explored in this study (see next paragraph for specific accounting method). At the same time, this study also measures the GEE (GEEv) using the same method based on the case of variable returns to scale to further enhance the credibility of conclusions. In addition, due to the lack of data for some cities, based on data availability, this study uses a total of 103 cities in the Yellow River Basin from 2011 to 2021. Relevant data are from the China Urban Statistical Yearbook and the China Energy Statistical Yearbook.

(2) Key explanatory variable: green finance development level (GF). This study uses a multi-indicator comprehensive evaluation method to assess GF. This index is measured from the green financial market perspective by choosing seven dimensions. See next paragraph for specific measurements. All the above data are from the China Statistical Yearbook and the CSMAR database.

(3) Mediating variables: green innovation (GI), environmental regulation (ER), and industrial structure (IS). To test the impact path of the GF on the GEE, the above three mediating variables are introduced for examination. Among them, this study uses the number of green patents granted in each city as an indicator to measure local green innovation, which is from the China Research Data Service Platform (CNRDS). Compared with green patent applications, green patents granted are more reflective of actual technological enhancement and effectively reflect local green innovation quality. Referring to previous literature [58], this study constructs an environmental regulation intensity variable based on the frequency of “environmental protection” related words in the government work report. For industrial structure, the value added of tertiary industry as a share of GDP is used to measure industrial structure upgrading, which is from the China Urban Statistical Yearbook.

(4) Control variables: to restore as comprehensively as possible the differences in economic and social environments faced by each city, and taking into account existing relevant literature, economic development (Pgdp), energy structure (ener), green coverage (Gre), urbanization level (Urban), and labor reserve (Pop) are added as control variables. The above data come from the China Urban Statistical Yearbook, the China Energy Statistical Yearbook, and the CSMAR database.

A total of 103 cities in nine provinces along the Yellow River are studied, and balanced panel data of each city for 2011-2021 are used for this study. This study uses linear interpolation to fill in the gaps for some of the missing data. The relevant variables we used are described in Table 3, and the descriptive statistics are shown in Table 4.

Spatial and Temporal Evolution of Green Economy Efficiency

This study uses the global super-efficiency SBM-DEA to calculate the GEE of cities in the Yellow River Basin from 2011 to 2021. The results show that the GEE of the cities in the Yellow River Basin varies considerably but generally shows an upward trend. To show the evolution process of the GEE in the Yellow River Basin more clearly, we focus on the GEE for the two years 2011 and 2021 (Fig. 2 and Fig. 3)⁵. The results show that in 2011, the cities in the northern part of Sichuan Province had the highest GEE. Areas with a good ecological environment, low pollution emissions, and fewer undesirable outputs have more regional GEE. The GEE was also higher in cities in the middle and lower reaches of the Yellow River Basin, which in turn were driven by economic development. Whereas in 2021, the cities in Qinghai Province, eastern Shaanxi Province, and southern Inner Mongolia had higher GEE, which was closely related to the economic development of these cities. There is a relatively high level of economic development in Shandong Province and Henan Province. The gradual increase in emphasis on environmental protection and pollution control reduced undesirable outputs. This contributed to higher GEE.

Overall, from 2011 to 2021, the GEE of cities in the upper and middle reaches of the Yellow River Basin increased significantly. The color corresponding to the GEE of the cities in eastern Inner Mongolia noticeably lightened, which is related to its more backward economic output. Meanwhile, in 2011, the GEE of Shandong Province was relatively high. In contrast, the GEE of Xining, Ordos, Yulin, and Xi’an was later at the forefront of the Yellow River Basin in 2021.

Spatial and Temporal Evolution of Green Finance Development Level

The measurement results of the multi-indicator comprehensive evaluation method show that GF in the Yellow River Basin has improved from 2011 to 2021. The green finance development index grew from 0.4342 in 2011 to 0.5699 in 2021 (Fig. 4), an increase of more than 30%. Among them, the year with the fastest growth in GF is 2017. This is due to the Guiding Opinions on Building a Green Financial System issued in August 2016, which clearly set out eight tasks, including the development of green credit, securities, development funds, and insurance, as well as the improvement of environmental rights and interests markets, the development of local pilots, and the promotion of international cooperation. All of these tasks were practically implemented in 2017, which became an important force driving the remarkable growth of China’s Green Finance Index in 2017.

Specifically, Shandong, Shaanxi, and Gansu Province have relatively high green financial development levels, with annual average GF indices reaching 0.7903, 0.7930, and 0.8070, respectively. Among them, Shandong Province’s green finance development started earlier, had a better foundation, and was ranked first in the Yellow River Basin in 2011. However, it was slightly weaker than the Shaanxi and Gansu Provinces in the following years. Since 2012, the highest level of GF in the Yellow River Basin has appeared in Gansu Province and has maintained a trend of higher growth year by year. Except for 2018-2019, when it was surpassed by Shaanxi Province, Gansu Province has always maintained the first place in the Yellow River Basin in all the years of GF. In contrast, the GF of Qinghai Province, Inner Mongolia Autonomous, and Ningxia Hui Autonomous Region still lags behind, and the GF of these three provinces in 2021 is still less than the average level of the Yellow River Basin in 2011. For these three provinces, GF still has significant space for development.

Similar to the changing trend of GF in the Yellow River Basin, the GF of the 103 cities in the Yellow River Basin is also increasing (Fig. 2 and Fig. 3). Compared with 2011, GF realized an increase in 2021, and the dots that characterize the level of GF in Fig. 2 and Fig. 3 all increase in general. Specifically, in 2011, compared with other cities, cities in Shandong Province had relatively high levels of GF, with Qingdao, Weihai, and Binzhou standing out, while cities in Inner Mongolia had lower levels of GF. By 2021, in addition to the still high level of GF in cities of Shandong Province, cities like Xi’an, Baoji, and Yan’an in Shaanxi Province, and Lanzhou and Jinchang in Gansu Province gradually came later, exceeding 80% of the cities. However, the GF development level in Inner Mongolia is still relatively backward, especially in Ulanqab and Bayannur. In addition, from 2011 to 2021, Baoji in Shannxi Province, Jinchang in Gansu Province, and Liaocheng in Shandong Province had the largest increases in GF, and GF in the eastern cities of Sichuan Province also increased considerably.

Existence Test for Effects

Using Equation (5) for estimation, this study finds that GF positively contributes to EPHD in the Yellow River Basin (Table 5). Columns (1)-(2) of Table 5 demonstrate the impact of the GF on the GEE (based on constant returns to scale). Column (1) presents the estimation results without control variables included. It shows that the estimated coefficient of GF is significant at the 1% level of 0.5805, which suggests that GF has a positive contribution to the GEE during the observation period selected in this study. Column (2) shows the model estimation results after adding five control variables: economic development, energy structure, green coverage, urbanization level, and labor reserve. It shows that the estimated coefficient of GF is 0.5837, which is positive at the 5% level. For every one-unit increase in GF, EPHD of the Yellow River Basin, as measured by the GEE, there will be an increase of 0.5837 units. This is due to the increase in green finance, which means that local government guides the popularization of green financial instruments, strengthens the environmental access threshold for micro-entrepreneurs to obtain loans from financial institutions, and provides financing for environmentally friendly projects, which has a positive effect on the local environment. On the other hand, it also indicates that in the process of GF, the structure of the finance market is gradually improved, which reduces the financing constraints faced by real enterprises and has a favorable effect on local economic development. Combining the above two aspects, GF can positively affect EPHD of the Yellow River Basin, which verifies H1.

Fig. 2.

The GEE and GF of cities in the Yellow River Basin in 2011.

Fig. 3.

The GEE and GF of cities in the Yellow River Basin in 2021.

Fig. 4.

The level of the GF of provinces in the Yellow River Basin in 2011-2021.

To further explore the impact of the GF on the GEE, we test this impact effect again using the GEE measured by SBM-DEA based on variable returns to scale [29], and results are shown in Columns (3)-(4) of Table 5. As can be seen from the results, the coefficient of GF is positive, regardless of whether control variables are included. This again verifies H1 that GF can significantly enhance EPHD.

Results of Robustness Tests

To test the robustness of the above findings, we use three methods: transforming explanatory variables, changing sample size, and transforming robust standard errors method. We also test whether GF’s supportive role for EPHD in the Yellow River Basin still exists.

First, transforming the explanatory variable. In the above, this study measured the GEE based on global super-efficiency SBM-DEA with constant returns to scale and variable returns to scale, respectively, which was used to measure its EPHD. While there are various methods to measure the GEE, it is unknown whether the final conclusions obtained are consistent with this study if other measurement methods are used. Therefore, to make the benchmark conclusions of this study more robust, the Cobb-Douglas production function [59] is used to remeasure the GEE (GEE_cd) and conduct a robustness test. The results are shown in Column (1) of Table 6. These show that the effect of the GF on the GEE is still positive and significant, indicating that GF can significantly enhance EPHD in the Yellow River Basin, which is consistent with the conclusion of benchmark regression. This also further verifies the robustness of the basic conclusions.

Second, changing the sample size. Compared with general cities, municipalities, separately listed cities, and provincial capitals may have more outstanding performance in terms of economic development, policy implementation, and environmental protection. Therefore, to reduce the biased nature of the sample, this study separately lists cities and provincial capital cities after eliminating the municipalities and re-regresses the remaining 94 cities, and the results are shown in Column (2) of Table 6. These show that GF enhances the GEE at the 1% significance level, i.e., GF still positively contributes to EPHD in the Yellow River Basin. This once again verifies the robustness of this study’s conclusions.

Third, transforming the robust standard errors method. In the benchmark regression, this study uses heteroskedasticity-robust standard errors for estimation. Whereas there may be autocorrelation in the perturbation terms of different years in the same city, the clustered robust standard errors can overcome such problems to a certain extent. Therefore, to ensure the validity of the estimation method, this study re-estimates the empirical model using city-individual clustering robust standard errors, and the results are shown in Column (3) of Table 6. In Column (3), the coefficient of GF remains significantly positive; this robustness test also does not change the baseline regression results, and the conclusion remains robust.

Results of the Endogeneity Test

For this study, cities with high levels of GF may be more inclined to focus on ecological and economic development, and there may be bidirectional causality between the explanatory variable (GF) and the explained variables (GEE). This mutual causality leads to endogeneity to some extent. Therefore, this study uses the Instrument Variables (IV) method to test it. Referring to previous literature [19, 20], this study uses the level of GF in the lagged period as IV to deal with endogeneity. GF in the previous period provides a reference for the development of GF in the current period, and the two are correlated. It is difficult for the current period to impact the value of the previous period, which has already occurred. This ensures that IV and the explanatory variables are independent of each other.

Based on this, this study uses the two-stage least squares (2SLS) method for estimation [60], and the results are shown in Table 7. Columns (1) and (3) demonstrate the results of the first stage, which indicate that IV satisfies the correlation. Furthermore, regardless of whether control variables are included, the P-value of the Kleibergen-Paap rk LM statistic remains 0.000, rejecting the null hypothesis of “under-identification of the instrumental variable”. This indicates that the IV used in this study does not suffer from under-identification issues. Similarly, both the Cragg-Donald Wald F statistic and the Kleibergen-Paap rk Wald F statistic are significantly greater than 10, rejecting the null hypothesis of “weak instrumental variable”. Columns (2) and (4) show the results of the second stage, in which the coefficient values of IV are significantly positive. This indicates that based on the instrumental variable method, the promotion effect of GF on EPHD in the Yellow River Basin still exists. This is consistent with this study’s basic conclusion and further proves our findings’ robustness.

Green Innovation Increasing Effect

GF’s capital financing and capital-oriented functions can effectively promote factor integration and resource allocation and potentially impact green innovation. This study verifies the impact of GF on EPHD of the Yellow River Basin through the green innovation enhancement effect, and the results are shown in Columns (1)-(2) of Table 8. As can be seen from the table, the coefficient GF of green innovation measured by green patent authorization is 6.1380 and passes the significance test, which means that GF can increase green innovation. From this, it can be inferred that the improvement of GF can guide capital flow to low-pollution green industries and force high-energy-consuming and high-pollution enterprises to carry out green innovation and clean production [61], significantly improving local green innovation. In Column (2) of Table 7, the coefficient of GI is significantly positive, which indicates that the increase in the number of green innovation results has a role in promoting the GEE. The coefficient of GF is 0.5498, which is significantly lower than in Table 5 (0.5837). This implies that green innovation does produce a mediating effect. To protect the environment and promote green development, GF provides financing and services for environmentally friendly green industries, giving related enterprises more funds for green innovation, thus promoting the green development of society. Green innovation contains the double connotation of green and innovation, which can improve enterprise productivity and development. Therefore, GF can improve green innovation and further contribute to EPHD in the Yellow River Basin.

Environmental Regulation Enhancing Effect

Currently, the government regards environmental regulation as an important measure to protect the environment and realize the synergistic development of the environment and economy. Based on this, this study verifies the impact of GF on EPHD of the Yellow River Basin through the environmental regulation adjustment effect, and the results are shown in Columns (3)-(4) of Table 8. As seen from this table, there is a significant positive correlation between GF and environmental regulation in the Yellow River Basin. This is because GF directs capital flows to low-polluting enterprises and restricts the development of enterprises with poor environmental performance. GF has a reverse selection effect on local industries and significantly increases local environmental regulation levels [49]. In Column (4), the coefficient of ER is positive, indicating that there is also a positive relationship between the GEE and environmental regulation. Therefore, cities with a high level of GF will have a higher level of environmental regulation. High environmental regulation means that it is difficult for local enterprises to raise funds and has high barriers to entry, forcing them to undergo green transformation, thereby promoting EPHD. This proves that GF improves the GEE by enhancing local environmental regulation. In addition, the coefficient of GF in Column (4) is 0.5251, which is also lower than in Table 5, 0.5837, suggesting that environmental regulation plays a mediating role in the process of GF affecting the GEE. The hypothesis of “innovation compensation” suggests that appropriate environmental regulation will increase the incentives for enterprises to engage in technological innovation and promote green innovation, thereby compensating for their environmental costs. This increases the likelihood that enterprises will have access to green funds, easing the pressure on financing and allowing more resources and capital to be invested in the production process. While environmental protection is being promoted, the high-quality development of the local economy will also grow. Therefore, GF plays a role in environmental regulation and, thus, plays a positive role in promoting local GEE.

Industrial Structure Upgrading Effect

The primary task of building a modernized socialist country is high-quality development, and industrial structure optimization means high-quality development. In this context, this study also verifies the mediating effect of industrial structure upgrading in EPHD affected by GF, and the results are shown in Columns (5)-(6) of Table 8. This table shows that GF has a role in promoting industrial structure upgrading, while industrial structure upgrading can further enhance city GEE. This is because GF can alleviate financing constraints of green and environmentally friendly enterprises, increase the cost and difficulty of obtaining credit funds for high-polluting and energy-consuming enterprises, and incentivize local enterprises to engage in green production. It can also strengthen the market selection mechanism of survival of the fittest and promote resource flow between different industries, thus adjusting the ratio of secondary and tertiary industries and optimizing the industrial structure [53]. In addition, the coefficient of GF in Column (6) (0.5340) is lower compared with Column (2) in Table 5 (0.5837), indicating that industrial structure upgrading does have a mediating effect. Adjustment of industrial structure usually implies an increase in the share of greener tertiary industry, which can bring higher added value and promote high economic growth, reduce energy consumption and pollution output, realize green development of the environment, and ultimately promote local EPHD.

In summary, the empirical analysis results of the mediating effect show that GF’s impact on EPHD of the Yellow River Basin mainly works through increasing green innovation, enhancing environmental regulation, and upgrading industrial structure, which verifies H2.

Based on Environmental Pollution Status

Environmental pollution status reflects the base level of regional EPHD. Referring to previous literature [62], this study uses the value of PM_2.5⁶ to represent city environmental pollution status and examines the similarities and differences in the promotion of EPHD by GF in cities with different baseline levels of the environment. According to the value of PM_2.5, the environmental pollution level of sample cities is sorted from high to low, and sample cities with an environmental pollution level higher than the median are defined as high-pollution cities, while those lower than the median are defined as low-pollution cities to analyze the impact of the GF on the GEE of cities with different environmental pollution conditions. As shown in Columns (1)-(2) of Table 9, GF can improve local EPHD in cities with either high or low environmental pollution, which is significant at least at the 5% level. This is due to the green attributes of GF itself, which can promote local green development regardless of urban environmental pollution degree. In addition, this result passes the grouping test, which shows that the coefficients of these two groups can be compared; that is to say, compared with the highly polluted areas, GF enhances EPHD of low-pollution cities more obviously. This is due to the attributes of low-pollution cities; such cities usually have lower pollution control costs and environmental compliance costs, and enterprises are more likely to obtain green funds from banks, so GF has a more obvious role in enhancing the EPHD of low-pollution cities.

Based on Fiscal Expenditure Level

Fiscal policy has a guiding and promoting role for GF, which is an important support for GF, and the comprehensive use of fiscal and financial instruments has become an important measure to promote regional EPHD. To examine whether the impact of GF on local GEE is heterogeneous under different levels of fiscal expenditure, we conduct a relevant heterogeneity test. This study uses local fiscal budget expenditures to characterize the fiscal expenditure level of cities, which are ranked in order of size, and the median is taken. If a city’s fiscal expenditure level is smaller than the median, it is defined as a low fiscal expenditure city; otherwise, it is defined as a high fiscal expenditure city. The results of heterogeneity analysis based on fiscal expenditure levels are shown in Columns (3)-(4) of Table 9. The results show that the enhancement of GF on EPHD is more obvious in cities with higher fiscal expenditure levels, which suggests that fiscal expenditure strengthens the role of GF in enhancing the GEE. This is because cities with high fiscal expenditures have more capital supply, hold more resources to invest in environmental and economic development, provide strong financial support for local ecological environment restoration, pollution control, and production, and are more capable of realizing the improvement of local EPHD.

Based on the Marketization Level

GF is a market-oriented institutional arrangement; therefore, this study examines whether the role of GF on local EPHD is heterogeneous under different regional marketization levels. Here, the marketization index is used to characterize regional marketization development, and sample cities are divided into two groups according to marketization level for the heterogeneity test. In this study, the marketization level is ranked from high to low, with the median as the cut-off point. If a city’s marketization level is higher than the median, it is defined as a high marketization city; if the marketization level is lower than the median, it is a low marketization city. In particular, this study constructs the marketization process indicator [63]. The results are shown in Columns (5)-(6) of Table 9, where the increase of GF can significantly enhance EPHD in highly marketized cities, while the effect on the GEE of low marketized cities is not significant. This may be because, compared with low-marketized areas, the market development of high-marketized cities is more mature. The government’s intervention in bank and enterprise lending is lower, and it can give full play to the positive role of the market mechanism in financial market lending. It can also effectively provide enterprises with green funds to achieve local green economy development.