The impact of green credits on high-quality energy development: evidence from China

The implementation of green credits has become an important engine for China’s high-quality energy development (HQED). On the basis of constructing an index of HQED and the panel data of thirty provinces in China from 2008 to 2019, this study empirically investigated the effects of green credits on HQED and the action mechanisms behind it in a multi-dimensional manner using a panel fixed-effects model, mediating-effects model, and spatial Durbin model. The results indicated that green credits had significantly contributed to China’s HQED, and that conclusion still held true after a series of robustness tests were conducted. It was found that industrial structures and human capital were important channels through which green credits influenced China’s HQED. Moreover, the spatial spillover effects of green credits on HQED were also confirmed. Finally, in terms of temporal heterogeneity, the positive effects of green credits on HQED were found to have increased significantly after 2012. Also, in terms of regional heterogeneity, this study observed that the positive influence of green credits on HQED was more significantly in central and western China than in eastern China, and in southern China than in northern China. The results obtained in this research investigation will potentially provide some important insights for energy planners and policymakers to further the understanding of the drivers of HQED, and the corresponding transmission mechanisms and regional differences.


Introduction
At the present time, China's economy has shifted from a stage of high-speed growth to a stage of high-quality development, and "high quality" has become the theme of China's current and future economic construction (Pan et al. 2021). Energy is a crucial material basis for the sustainable development of national economy and the power source of modern economic growth. However, a coal-based energy resource endowment makes China's traditional energy system a high-emission system. In addition, it also exposes China to such dilemmas as insufficient supplies of renewable energy and excessive dependence on overseas oil and gas supplies (Zhang 2022;Dong et al. 2018). Following the introduction of carbon peak and carbon neutral targets, the goals of green and clean energy transformation became daunting. Furthermore, under the conditions of geopolitical conflicts, anxious international energy supply chains have also challenged the stability of China's energy supply (Mendez et al. 2022;Szép et al. 2022). Therefore, to maintain stable and sustainable economic development under environmental and energy constraints, it has become necessary for China to establish a modern energy system which is safe, reliable, economically viable, and green.
The white paper titled China Energy Development in the New Era (Wang et al. 2022a) argues that HQED should be an energy development strategy reflecting the new developmental concepts, which point the way to HQED in China. However, the concrete mechanisms for its realization are still to be explored. Green credits are a financial instrument which serves the ecological environment. Its main goal is to achieve a win-win situation for both environmental protection and economic development through the green allocation of funds (Zhu et al. 2020). Since the introduction of green Responsible Editor: Nicholas Apergis credits in China in 2007, the incentives have demonstrated remarkable results in regard to pollution control and ecological protection. Currently, green credit policies have gradually expanded and extended to the energy sector and have become an important driving force and support for HQED. By the end of 2021, the balance between China's domestic and foreign currency green credits was RMB 15.9 trillion, ranking first in the world in terms of stock size, of which the balance of loans for the clean energy industry totaled RMB 4.21 trillion, accounting for 26.48% of the balance of the green credits . The 20th CPC Congress Report also further emphasized that "Promoting green and low-carbon economic and social development are key links to achieving high-quality development." Therefore, it is important to determine how to effectively unleash the power of green credits to fuel China's HQED, and focused planning will be a widely discussed action topic by both the government and communities.
However, questions remain regarding how green credits actually drive HQED in China. If such effects are confirmed, it will become important to identify the mechanisms behind them. In addition, how the influences of green credits on HQED differ in terms of their own features and spatial patterns requires further clarification. Scientifically based answers to the aforementioned questions will provide a policy basis for China to achieve sustainable and healthy social and economic development. Moreover, as the world's largest developing country, China's experience in exploring those issues will provide useful references for the international community and contribute to the promotion of a beautiful and clean world. To be more specific, the unique attributes of HQED were taken into account in this study. An indicator system from the perspective of an "energy impossibility triangle" was constructed, and the levels of HQED in thirty Chinese provinces between 2008 and 2019 were measured. Moreover, various econometric methods were utilized to empirically examine the impacts of green credits on HQED, along with their mechanisms of action. The findings showed that green credit incentives significantly contributed to HQED of China, with industrial structures and human capital as important impact mechanisms. Furthermore, green credit spatial spillover effects emerged in the measured HQED. Those findings held true even after various robustness tests were completed, such as winsorization, one-period lagging of explanatory variables, and an instrumental variables approach.
The possible marginal contributions made by this study lie in three aspects. First of all, the research systematically distilled the new connotation of HQED from the perspective of the energy impossibility triangle, on the basis of which the HQED was measured, which enriched the theoretical study of high-quality development in China. Secondly, the main pathways through which green credits affect HQED were explored within a unified framework, and the channel roles of industrial structures and human capital in the influencing effects of green credits on HQED were comprehensively assessed, which deepened the existing literature. Finally, this study utilized a spatial Durbin model to examine the spatial spillover effects of green credits on HQED. That is to say, the model built in this study not only examined the direct impacts of green credits on HQED, but also investigated the changes of the direct and indirect impacts of green credits on HQED after spatial factors were incorporated.

Regarding green credits
The previous research on green credits has focused on the subsequent three main areas: Firstly, the introduction of the conceptual connotation of green credits: Jeucken and Bouma (1999) thought green credits as a means for banks or financial institutions to promote environmental sustainability through the implementation of policies. It was argued that financial institutions use their informational advantages, credit resources, and financing policies to provide credit funds for green industries, which could be considered to be credit rationing under environmental constraints. In contrast, Thompson and Cowton (2004) put forward that green credits are mechanisms through which banks incorporate environment-related information regarding projects and their operating companies as an examination criterion in lending processes, through which they make the final lending decisions.
The second is the analysis of the effects of green credits: From a micro perspective, Hu et al. (2021) argued that the introduction of green credits has promoted green innovations among heavy polluters. Lai et al. (2022) argued that green credit incentives have improved the values of new energy listed companies. As the main implementers of green credit policies, banks have also received extensive academic attention. Yin et al. (2021) demonstrated the positive impacts of green credits on bank performance. Ding et al. (2022b) further investigated the significant roles and impact mechanisms of green credits on the high-quality development of commercial banks. Based on a macro perspective, numerous researchers have studied the effects of green credit incentives on carbon emissions  and environmental quality (Zhang et al. 2021a). Additionally, researchers such as Ma et al. (2021) and Tan et al. (2022) have also studied the positive effects of green credits on efficient energy usage.
Thirdly, there are studies on green credit risks. Chen and Zhao (2022) pointed out that effectively managing green credit risks leads to the sound growth and smoother functioning of the green credit market. Zhao and Chen (2022) suggested that effective green credit risk management calls for the energetic deeds of government agencies, financial institutes, green companies, and regulators.

Regarding high-quality energy development
The existing studies regarding HQED were found to mainly cover the following three aspects: first, a systematic distillation and summary of the basic connotation, main objectives, and important significance of HQED from a qualitative perspective, as well as the promotion and interpretation of the policy approaches related to HQED. Wang et al. (2022a) suggested that HQED should be a model of energy development which reflects the five new development concepts of innovation, coordination, green, openness, and sharing. Zhao et al. (2021) presented the idea that the aim of HQED is to make clean energy accessible to people of different income levels. In addition, from a national perspective, Motz (2021) emphasized that energy security should be the primary objective of energy development. The second is about the theoretical exploration of the paths to HQED should be explored. For instance, Xie et al. (2022) pointed out that on the background of "double carbon," the Beijing-Tianjin-Hebei region should be capable of promoting HQED under the novel situation, starting with energy consumption and supply, science and technology, and institutional mechanisms.
Thirdly, quantitative evaluation studies are carried out on the sub-topics of HQED, for example, energy security, green energy, and energy efficiency. The research paradigm would be mainly based on the construction of a multi-dimensional evaluation index system. Grey models and entropy methods should be used to measure the energy development levels of different spatial units and the combination of ArcGIS spatial analysis techniques adopted to reveal their spatial and temporal evolution characteristics (Peng et al. 2021;Liu et al. 2012). Regarding this, Zhang et al. (2017) constructed an energy security evaluation system from five dimensions, including governance and innovation, multi-use and security, affordability, equity, and environmental sustainability, and conducted comprehensive assessments of the levels of energy security in thirty Chinese provinces.
As can be seen from the abovementioned literature, although a growing body of theoretical and empirical research on green credits and HQED existed, studies which combined the two were rare. The few studies which had addressed both were mainly analyses of the impacts of green credits on the sub-topics of the energy system, such as carbon emissions and energy efficiency Zhang et al. 2021c;Song et al. 2021;Wen et al. 2021), which were not systematic or comprehensive. At the same time, it was observed that previous studies had not provided unified frameworks to answer the questions of what paths the promotion of green credits mainly take so as to benefit HQED. Therefore, this research investigation attempted to incorporate green credits and HQED into the same model, thereby broadening the scope of the existing research. Then, a mediating effects model and a spatial Durbin model were utilized to examine the influence paths and spatial spillover effects of green credits on HQED. The obtained results provided practical and effective paths for the facilitation of HQED using theoretical exploration methods and practical urgency.

Theoretical analysis
The effects of green credits on HQED were mainly examined and demonstrated in this study from three aspects, and research hypotheses were presented.

Direct effects of green credits on HQED
Green credits provide low interest rate loan support for the development of clean energy enterprises. Meanwhile, high polluting energy enterprises face higher financing constraints through punitive high interest rates. Such financial factors serve to promote low-carbon energy transformation and move toward achieving China's green and clean energy goals. Green credits provide loans for new energy projects with differentiated interest rates, thereby promoting the spread and use of new energy and ensuring stable energy supplies. In addition, by directly influencing the production activities of energy enterprises, green credits promote the transformation of the energy industry into a mode of economic development, drive down energy costs, improve the economic viability of energy resources, and boost China's HQED. Therefore, according to the aforementioned, the first hypothesis was proposed as follows: H 1 : The development of green credit programs facilitate HQED.

Indirect impacts of green credits on high-quality energy development
The main problem faced by China's HQED is the imbalance between the demand and supply structure of the energy industry. A balanced demand and supply structure requires a good match and cycle between the demand side and the supply side. Green credits can facilitate high-level demands and high-quality energy supplies through two channels, industrial structures and human capital, and successfully promote HQED with synergy on both the demand and the supply sides.
On the demand side, green credits raise the demand tier for energy through industrial structures, thereby expanding the demands for high-quality energy. That in turn promotes HQED. Research findings have revealed that green credit policies direct more capital to green industries and high-technology industries which reduce the size of highly polluting enterprises, thereby promoting the rationalization and advancement of industrial structures (Zhu 2022). Green credit programs also promote the mergers and acquisitions of heavily polluting enterprises. Such actions have been found to optimize production methods, rationalize the allocation of resources, and promote green production, which further lead to the promotion of industrial structures (Zhou 2014;Hu et al. 2020). In the future, under the influences of green credit policy guidance and industrial structure upgrading and price changes, the energy-consuming sectors will increase their consumption of renewable energy or clean energy. This will support the growth of the clean energy industry through financial means so as to lessen the use of fossil fuels (Feng and Xing 2018), which optimizes the energy consumption structure. As a result, HQED will be further promoted.
On the supply side, green credits are able to upgrade the quality of energy supplies by influencing human capital, thereby promoting HQED. The introduction of green credits also puts pressure on the operations of "two high and one surplus" enterprises, signaling to the market that the development prospects of those industries are not promising. Meanwhile, high-quality human capital will be more willing to enter the sunrise industry, which is encouraged by the state. In addition, green credit programs have been found to increase the financing constraints faced by the two high and one surplus enterprises. It has been observed that when there is insufficient liquidity, enterprises may choose to hire less high-quality human capital or reduce wages to raise funds. However, high-quality human capital is conducive to improving productivity, enhancing innovation, optimizing the match between labor and physical capital, and stimulating the endogenous development of enterprises, which improves the supply capacity of clean energy and promotes HQED (Hu and Liu 2022). Therefore, according to the aforementioned, the hypotheses were proposed as follows: H 2 : Green credit programs make positive contributions to HQED through industrial structures. H 3 : Green credit programs make positive contributions to HQED through human capital.

Spatial spillover effects of green credits on high-quality energy development
When the central region promotes the development of green credits, it is expected to have spatial spillover effects on the HQED in peripheral regions as follows: first, "demonstration" effects of green credits: The introduction of green credits will bring about demonstration effects and information transferences to neighboring regions, thereby promoting industrial upgrading and investment in human capital in neighboring provinces. Such activities will assist in harmonizing the quality of energy supplies and the upgrading of demands in order to facilitate the realization of HQED (Lei et al. 2021;Yao et al. 2022). Second, radiation effects of green credits: As green credit policies are perfected and credit scales become rapidly developed, radiation effects will be experienced by the surrounding areas. Those effects will further promote the coordination of green credit resources, spread of green technologies, and the division of labor in new energy industries among the regions. As a result, the effective use of green credit resources will be realized across the regions, thereby accelerating the processes of HQED in adjacent regions. Ultimately, positive spatial spillover effects will be demonstrated Wang and Wang 2022).
Hence, according to the abovementioned expected results, the hypothesis was also proposed as follows: H 4 : Green credit policies and programs promote HQED in adjacent regions via spatial spillover effects.

Econometric model design
For the purpose of verifying the impacts of green credits on HQED, this paper built the basic model for the direct transmission mechanism as follows: where seg it denotes the level of HQED; green it represents the green credit level; CV it are several control variables; μ i and δ t represent individual fixed effects and time fixed effects, separately; and ε it is the random disturbance term. Aiming to eliminate the effects of potential heteroskedasticity and data volatility, each variable was logarithmized.
Next, for the purpose of discussing the possible mechanisms of green credits on HQED, it was important to test whether industrial structures (indus) and human capital (human) were mediating variables, respectively, as described in the previous section. Taking industrial structures (indus) (1) as an example, the test procedure was as follows: On the basis of the significance of the coefficient of linear regression model (1) for the green credit levels (green) on HQED (seg), a linear regression equation for the mediating variable industrial structure (indus), along with the regression equation of green credit levels (green) and the mediating variables (indus) on seg, were constructed, respectively. The presence of mediating effects was determined by the significance of the coefficients, such as β 1 , γ 1 , and γ 2 . The analysis results of the human capital impact mechanism were found to be similar. The concrete formality of the abovementioned regression model was set below: The First Law of Geography (Tobler 1979) demonstrates that all things are related to other things, but close things are more associated with each other than distant things. In this study, since both the regional levels of the green credits and the HQED were likely to be spatially associated, failure to take into account the spatial correlations between regions when conducting regression analysis would potentially bias the estimated results. Therefore, in order to further research the spatial spillover effects of green credits on HQED, a spatial interaction term between the two and other control variables was added into Eq. (1), which was then further expanded to a spatial panel econometric model as follows: where ρ represents the spatial autoregressive coefficient, and W k (k = 1, 2, 3) is the spatial weight matrix. In addition, for the purpose of improving the robustness of the empirical results, the following three methods of regression were used in this study: Economic-geographic nested matrix (W 1 ), economic distance matrix (W 2 ), and a geographical distance matrix (W 3 ), in which 1 and c are the elasticity coefficients of the core explanatory variables, and the spatial interaction terms of the control variables (Jianhong et al. 2022;Yang et al. 2022). Formula (4) includes the spatial interaction terms of the explanatory variables and is referred to as a spatial Durbin model (SDM). The spatial Durbin model was derived from the spatial autoregressive model (SAR) and the spatial error model (SEM), which filled the gaps existing in those two models (Ding et al. 2022a).

Explanatory variables
Green credit (green) indicators can generally be identified and measured in the following ways (Song et al. 2021;Lian et al. 2022;Hu et al. 2020): the proportion of energy saving and emission reduction loans to total loans, the proportion of bank loans in investments for industrial pollution control; the proportion of green credits to banks, and the inverse indicators of the proportion of interest payments in six energy-intensive industries (Wen et al. 2021). The three are all positive indicators and their data come from banks. However, due to the inconsistent statistical caliber of data published by different banks, along with the different statistical standards which may be adopted by the same bank at different periods, problems of poor accessibility and completeness exist. In view of those issues, this study used the reverse indicators of green credits for measurement purposes. The following was the calculation formula:

Explained variables
On the basis of the scientific connotation of HQED, a reasonable measurement strategy was the key focus of this study. It was found that the existing research was based on an evaluation system of the high-quality economy development, with some expansions in the connotation and construction of an indicator system (Wang et al. 2022b(Wang et al. , 2022a, which lacked focalization. It was believed that a more concrete elaboration of the meaning of HQED was urgently needed when studying specific issues. This study observed that the main challenge to HQED was the energy impossibility triangle. For example, it was difficult for an energy system to meet the three objectives of safety and reliability, economic viability, and green low carbon at the same time, and any two of the three could be mutually exclusive (Fu et al. 2021). However, throughout the entire course of energy development, the so-called impossible relationship between security, economy, and green goals was always relative and manageable. It was considered that the three were not completely opposed or separate, but mutually supportive and complementary, or a unified relationship of opposites (Yao 2022). In other words, security, economy, and green goals represent the three sides of energy. Ensuring safety and stability, low prices, and green low carbon values continue to be the unremitting pursuit of the whole society for energy development, which reflects the three basic objectives of HQED. Therefore, this study attempted to measure the levels of HQED comprehensively from the three dimensions of security and reliability, economic feasibility, and green low carbon values, originating from the issue of breaking the Energy Impossibility Triangle.
Specifically, energy security and reliability lie in the stability of energy supplies (Liu et al. 2012;Wang et al. 2022a). The energy production per capita, energy self-sufficiency rates, energy production elasticity coefficients, and energy production diversification levels were chosen to reflect the quality and quantity of energy supplies in this study. Also, the energy industry investment and technology levels were selected to reflect the sustainability of energy supplies. The economic viability of energy mainly emphasizes low energy prices. Therefore, energy efficiency, natural gas penetration rates, and the ability to pay for electricity were selected to reflect the levels of energy consumption of the population (Zhang et al. 2021b). Rural per capita electricity consumption and energy prices were chosen to reflect energy accessibility. In addition, since green and low-carbon energy focuses on energy cleanliness and environmental friendliness , this study selected sulfur dioxide emissions per capita, general industrial solid generation per capita, and carbon dioxide intensity levels to reflect energy pollution. Investment in industrial pollution treatment and forest cover were also selected as proxy variables for energy contamination control, and the proportion of green energy consumption was selected to reflect green energy consumption. Furthermore, for the purpose of eliminating the influences of subjective factors on the evaluation results, an entropy weighting method was adopted to determine the weights of each indicator. The specific indicator selection and its weighting coefficients are detailed in Table 1.

Mediator variables
This research investigation selected industrial structures and human capital as the two mediating variables of the conduction mechanism of green credit impacts on HQED according to the aforementioned theoretical analysis. The industrial structures were represented by the ratios of the values added by the tertiary sectors to the values added by the secondary  . Human capital was represented by the average years of education of the employed persons (Wen and Dai 2020).

Control variables
For a more comprehensive analysis of the impacts of green credits on the HQED, this study also selected control variables which could potentially influence the HQED Xian and Leng 2016;Liu et al. 2022b). Those variables included the following: technological innovation (tech), expressed as the sum of the number of invention patents and utility model patents in a region; financial development (fina), expressed as the ratio of the balance of deposits and loans of financial institutions to the regional GDP (Ren et al. 2021); educational development (educ), expressed as the ratio of the number of students in regional universities to the total regional population; population size (popu), expressed as the regional year-end residential population; economic development (econ), expressed as the regional GDP per capita; openness (open), expressed as the total import and export trade to the regional GDP; and level of urbanization (urban), expressed as the ratio of the regional urban population to the total regional population.

Data sources
Green credit policy of China was proposed in 2007. However, the time delay of the policy and the availability of the data were taken into consideration in this study, and the panel data of thirty provincial units (excluding Hong Kong, For the data gaps of individual years, a linear fitting method was used to fill in the missing data. Table 2 details the results of the descriptive statistics of the variables. It can be seen in the table that the natural logarithm of the HQED index had a maximum value of − 1.289, a minimum value of − 2.912, a mean value of − 2.308, and a standard deviation of 0.336, indicating that the levels of HQED were found to vary significantly across the regions. The natural logarithm of the levels of green credits also showed a large standard deviation. The mediating and control variables were observed to differ across the regions as well. Table 3 reveals this study's results of the benchmark regression of green credits (lngreen) on HQED (lnseg). As can be seen in the "Methodology and data sources" section, ordinary least squares (OLS) and fixed-effects (FE) models were used to estimate Eq. (1), without control variables and with all control variables, respectively. In previous related studies, in order to address the possible simultaneous existence of between-group heteroskedasticity, serial correlation, and cross-sectional correlations, the XTSCC command proposed by Driscoll and Kraay (Driscoll and Kraay 1998) had been successfully adopted to deal effectively with the robust standard errors of FE models. Therefore, this study made the decision to use a fixed effects estimation with Driscoll and Kraay standard errors (XTSCC) method to estimate Eq. (1). In the table, columns (1), (2), and (3) display the estimation results for the ordinary least squares model (OLS), fixed-effects model (FE), and the fixed-effects model with Driscoll and Kraay standard errors (XTSCC), respectively. The results of the benchmark regressions indicated that all of the models fit well and had achieved high confidence levels overall. Meanwhile, the estimated coefficients for green credits (lngreen) were all observed to be significantly positive at least at the 10% level, suggesting that green credits had significantly contributed to HQED in China. The results of the XTSCC model further showed that the estimated coefficient for green credits (lngreen) was 0.09 and was significantly positive at the 1% level. In other words, it was revealed that green credits (lngreen) had increased the levels of HQED (lnseg) by 0.09% for every 1% increase in green credits (lngreen) on average. With regard to control variables, the coefficient of technological innovation (lntech) was determined to be negative and significant at the 5% level, which indicated that the higher the level of technology innovation, the lower the level of HQED. A possible explanation for those findings could be that the technological research and development was characterized by large investment amounts, long investment cycles, uncertain returns, and high investment risks (Dai and Zhen 2022). Another factor could have been that the technology innovations for energy companies had led to higher costs for their energy production. Meanwhile, the safety and cleanliness of the energy often could not be enhanced quickly within a short period through technology innovation, making new technology innovation disincentives for HQED. However, it is considered that in the long run, as energy technologies continue to be introduced, the safety and cleanliness of energy will be improved, thereby promoting HQED (Murad et al. 2019). The coefficient of population size (lnpopu) was determined to be significantly negative, which indicated that the bigger the population size, the lower the level of HQED. The probable cause for those results was that under the existing conditions, increases in population size may have led to increased energy consumption and the deterioration of environmental quality, which in turn weakened the quality of HQED. The coefficient of the level of urbanization (lnurban) was also significantly negative, which suggested that the higher the level of urbanization in each province, the lower the level of HQED. It was believed that this was mainly due to the fact that urbanization was driving a sharp increase in energy demands, while resource endowments of coal-rich, oil-poor, and gas-poor resources led to a low elasticity of energy supply in China, making it vulnerable to energy shortages. At the same time, the industrial character of urbanization had resulted in large amounts of pollutant emissions, which had negative effects on HQED. In addition, the coefficients of financial development (lnfina), educational development (lneduc), economic development (lnecon), and openness (lnopen) were observed to be insignificant, indicating that their impacts on HQED remain uncertain.

Mediating effect results
The preceding sections of this study described the theoretical analysis of the conduction mechanism of the impacts of green credits on HQED from the point of industrial structures. For the purpose of testing the hypothesis of the role of the mechanism, a mediating effects model was chosen for  (1), the positive effects of green credits on HQED were confirmed.
The results obtained using model (2) verified the green credits promoted improvements of industrial structures. The regression coefficients of green credits were 0.133 and 0.106, separately, and at least at the 5% level, both were significantly positive. Finally, the mediating variable of industrial structures was put back into the regression equation of the impacts of green credits on HQED. The coefficient values of the core explanatory variables were judged by observing the coefficient values and their changes in significance. The coefficients of green credits and industrial structures in model (3) were all significantly positive, with the direct effects of green credits being 0.117, which was smaller than that of the total effects. Therefore, it was indicated that the industrial structure upgrading was a mechanism of green credits in the promotion of HQED. Furthermore, the effects of human capital as a mediating variable was tested in this study using the same approach as previously described. Column (4) reveals a significantly positive relationship between the effects of green credits on HQED, which indicated that the testing of the mediating effects could be continued. As can be seen in column (5), the coefficient of green credits was significantly negative at the 1% level. In other words, increases in the levels of green credits had inhibited the improvement of human capital. The probable cause was that the green credit policies imposed strict credit constraints on the traditional energy sector. On one hand, this had pushed firms to upgrade through technological innovations in order to avoid the restrictions on their credit granted by the green credit policies. This had generated a demand for highly skilled labor to match technology studies and development plans and promoted the optimization of human capital (Fan and Li 2022). However, on the other hand, it had also reduced the funds available to hire highly skilled personnel for research and development and innovations, which further inhibited the optimization of human capital. The energy industry in China is characterized by a difficult working environment and a long talent training cycle, which results in the supply of highly skilled energy professionals being relatively scarce (Yu et al. 2021). Meanwhile, the internal management problems of energy businesses, such as the emphasis on hardware but not software, along with the emphasis on construction but not operations and maintenance, coupled with the principle of profit maximization, has led to relatively limited effects of green credits on the improvement of human capital, with more prominent inhibiting effects observed. Column (6) shows that the coefficients of green credits and human capital were significantly positive, with the coefficient value of green credits being 0.134, which was greater than the coefficient value of green credits in column (4). Those results suggested that green credits' promotional roles on HQED were significantly increased after the inclusion of human capital as a mediating variable. In other words, it was confirmed that human capital plays a part in masking the mediating effects in the relation between green credits and HQED. The calculated indirect effects accounted for 19.84% of the total effects.

Spatial effects results
As the linkages between regional economic development continue to grow, the levels of green credit incentives in neighboring regions can also potentially influence the HQED in a region. Therefore, based on the aforementioned, the following spatial measures were adopted in this study to further verify the relevance of green credits and HQED in China.
It was necessary to test whether there were spatial effects for the research objects before conducting spatial econometric analyses. This study conducted a spatial autocorrelation test on the levels of green credit development and the HQED index. The spatial effects for each year under the economic-geographic nested matrix were calculated by using the Moran'I Index method, and the outcomes are shown in Table 5. The Moran'I Index for the levels of green credit development and the HQED index under the economic-geographic nested matrix weights from 2008 to 2019 had both reached a significance level of 5%. Those findings had indicated that the levels of green credit development and the HQED index in each region of China from 2008 to 2019 had significant spatial autocorrelation. In other words, the two displayed clustering in spatial distribution.
The results of the spatial regression model of the green credits effects on HQED for three different spatial weight matrices are detailed in Table 6. Prior to this investigation, a combination of "specific to general" and "general to specific" approaches consisting of LM tests, SDM model fixed effects, Hausman tests, and SDM model simplification tests in accordance with the Elhorst test (Elhorst 2014) was used in turn. The SDM model with double spatiotemporal fixed effects was identified as the optimal choice. Therefore, this study's subsequent analysis processes focused on the SDM model. For the purpose of comparing the robustness of the estimates, the results of the spatial lag model (SAR) and spatial error model (SER) for the spatial-temporal fixed effects were also reported.
Depending on the results shown in Table 6, the spatial autoregressive coefficients for HQED in the SDM model under all three spatial weight matrices were significantly positive at the 1% level. Meanwhile, the coefficient of the spatial interaction term for green credits was positive, which indicated the existence of not only exogenous green credit interaction effects, but also endogenous interaction effects for HQED in the sample regions in space. Nevertheless, the regression coefficient values of the spatial interaction terms could not be used directly to research the marginal impacts of green credits on HQED, since analyzing the spatial spillover effects between regions through simple point regression results would produce incorrect estimates. Consequently, a partial differential explanation of the change in variables (for example, the use of direct and indirect effects) was required to expound the impacts of independent variables in one region on dependent variables in the same region, as well as in other regions. Table 6 details where the direct effects, indirect effects, and total effects of the green credits on HQED were observed to be positive and had passed the significance tests, respectively. It was indicated that higher levels of green credits promoted HQED through direct as well as indirect effects. It was found that green credits in a province not only had positive effects on the HQED in that province, but also the implementation of green credits could break through inter-regional restrictions and send green development signals to neighboring provinces, thereby promoting the HQED in those neighboring provinces. That is to say, spatial spillover effects of the impacts of green credits on the HQED were evident in this study's results.

Heterogeneity analysis results
This subsection examines the heterogeneous impacts of green credits on HQED in both the temporal and spatial dimensions.
At the temporal level, the issue of the Green Credit Guidelines in 2012 regulated the credit decisions of banks and the environmental performances of firms, thereby placing higher demands on commercial banks' corporate governance and credit risk management processes. Therefore, this study chose 2012 as the cut-off point and divided the sample into two sub-samples, 2008 to 2012 and 2013 to 2019, for further examination. The regression results in columns (1) and (2) of Table 7 represent that the coefficient values of green credits within the two subsamples for 2008 to 2012 and 2013 to 2019 were 0.0275 and 0.169, respectively. However, only the coefficient values of green credits within the sub-sample for 2013 to 2019 had passed the significance test at the 5% level. Those findings indicated that before the introduction of the Green Credit Guidelines, green credit programs had not achieved significant positive impacts on HQED. Then, after the guideline's introduction, the contribution of green credits to HQED had notably increased. The possible reason for those results was that although the green credit policies had been introduced in 2007, they had not been really taken seriously by commercial banks at the time, which had led to the policies' limited role. Meanwhile, the green credit policies were still in the initial exploration period, and the supporting measures and other aspects were not comprehensive, thereby rendering the green credit programs to be ineffectively implemented at the initial stages (Guo 2014). The issue of the Green Credit Guidelines in 2012 prompted the banking industry to elevate green credits to a strategic level, and to guide the optimization of the country's economic development structures with a green upgrade of the credit structure, which greatly contributed to HQED.
At the spatial level, China has a large territory with various resource endowments and industrial structures in different regions, which could potentially lead to regional heterogeneity in the effects of green credits on HQED. Therefore, based on such possibilities, this study conducted a systematic analysis of the differences in the impacts of green credits on HQED from the East to Midwest regions and the South to North regions, respectively. This study first divided the thirty provinces into two groups , East and Midwest, and then ran the regressions separately. In Table 7, columns (3) and (4) display the regression results for the eastern and midwestern subsamples, separately. It can be seen in the table that the positive effects of green credits on HQED were not significant in the eastern regions. Meanwhile, the green credits had significant positive effects on HQED in the midwestern regions. Therefore, considering the regional heterogeneity, the positive effects of green credits on HQED were stronger in the midwestern regions. The probable causes could be related to the fact that the financial sectors were more developed in the eastern region, and the financing channels of the enterprises were not only limited to bank credits, but also obtained through the issuing of bonds, stocks, and other means. Consequently, the impacts of green credit policies in the east were smaller Liu et al. 2022a). Meanwhile, since the financial channels in the midwestern regions are relatively scarce, enterprises tend to rely more on obtaining funds from banks. Consequently, the impacts of green credit policies in the midwestern regions were observed to be larger. In addition, the eastern region is considered to be more technologically advanced and has more technology-intensive emerging energy enterprises. Conversely, the midwestern regions are dominated by resource-intensive and labor-intensive traditional energy enterprises. As a result, HQED in the eastern region may not be significantly influenced by the promotion of green credits.
Recently, the regional gap between the northern and southern regions of China has shown a persistent trend ). Consequently, this study further examined the heterogeneity in terms of two regions, the South and the North. Using the Qinling-Huaihe line as the boundary, China was roughly divided into the southern regions and the northern regions. Column (6) details the coefficient of the green credits for the southern region, which was 0.365 and held a 1% level of significance. Simultaneously, according to the results shown in column (7), the coefficient of the green credits for the northern region held a 10% level of significance, with a value of 0.0907, which is only approximately a quarter of the former. Those findings suggested that the contribution of green credits to HQED was more significant in the southern regions. The main reason was that the share of traditional fossil energy-related industries with high pollution and emissions was lower in the south than in the north (Zhuang and Mi 2022). In addition, the objective resource environment for the development of renewable energy sources, such as hydropower, was more advantageous in the south. It was found that both scenarios together had resulted in a greater marginal effect of green credits on HQED in the southern regions.

Robustness tests
The main finding of this study was that green credits make significant contributions to HQED. However, so as to ensure the reliability of the findings, this study performed robustness tests from the following three perspectives: winsorization, one-period lag of the explanatory variables, and an instrumental variables approach.
It was believed that when the significant differences in the levels of HQED across the regions were considered, this may have led to outliers in the HQED index. Therefore, for the purpose of eliminating the negative impacts of outliers and non-randomness on the measurement results, this study winsorized 1% of the top and 1% of the bottom main explanatory variables. The results are presented in column (1) of Table 8. The coefficients of the green credits remained significantly positive, which was the same as the previous benchmark regression results.
In addition, in order to attenuate the effects of reverse causality, the lag period of the green credit policies was chosen as the explanatory variable in order to be re-estimate the study results. The re-estimation results are displayed in column (2) of Table 8. The one-period lag of the green credits was significantly positive at the 1% level, which was in line with the results of the benchmark regression.
This study also re-estimated the model results using an instrumental variables approach so as to test if the regression results were subject to endogenous interference. The related analysis results revealed that there was a strong relationship between the one-period lag of the green credit policies and the green credits themselves, which satisfied the correlation characteristics of the instrumental variables. Meanwhile, there was no direct relationship observed between the lag period of the green credit policies and the HQED, which satisfied the exclusion characteristics of the instrumental variables. Therefore, the one-period lag of the green credit policies was chosen as the instrumental variable for the 2SLS estimations. Since the validity of the instrumental variable selection required verification, tests for underidentification, weak instrumental variables, and over-identification were separately conducted in this study. As can be seen in column (3) of The instrumental variables in this study had corresponded one-to-one with the endogenous variables. Therefore, there were no problems of over-identification of the instrumental variables. After accounting for any endogeneity issues, the estimated coefficients of the green credits under the instrumental variables approach were confirmed to be positive and had passed the significance test at the 1% level. In other words, the green credits were found to significantly promote HQED, showing no differences to the main conclusions of the previous section. In summary, based on the abovementioned three aspects of the regression results, the research in this study had successfully passed the tests with good stability and reliability.

Conclusions and policy recommendations
Green credit policies are the earliest, largest, and most mature part of China's green finance sector and have gradually become a vital engine for China's HQED. In this research investigation, based on panel data of thirty provinces in China from 2008 to 2019, empirical tests were conducted regarding the impacts of green credits on HQED. The mechanisms of actions and the spatial spillover effects of green credits were explored in a multi-dimensional manner, using a panel fixed-effects model, spatial Durbin model, and a mediating-effects model, on the basis of constructing an index of HQED. The following are the main findings of this study: Firstly, green credits were found to significantly contribute to China's HQED and have become a primary force in the new era for China to break the Energy Impossibility Triangle. This conclusion had still significantly held after the completion of this study's robustness tests, including Winsorization, one-period lag of the explanatory variables, and an instrumental variables approach. Second, industrial structures and human capital were confirmed to play mediating roles in the relation between green credits and HQED. In other words, there existed partial mediation effects and masking effects, respectively. Third, the spatial spillover effects of green credits on HQED were also verified in this study's results, which indicated that the introduction of green credits helped to enable a new pattern of coordinated development between regions. Fourth, in terms of temporal heterogeneity, the positive effects of green credits on HQED were determined to increase significantly after 2012, while non-significant positive effects were observed before 2012. Regarding regional heterogeneity, the effects of green credits on HQED were significantly positive in the midwestern regions, while there were no significant effects observed in the eastern regions. Furthermore, the positive impacts of the green credits in the southern regions were significantly greater than those in the northern regions.
Apart from offering a series of empirical evidence for green credit contributions to HQED, the policy implications arising from the findings of this study are as follows: Firstly, under the fact that green credit policies can become a new driving force for HQED, it is important to fully promote green credit investment efforts in clean and efficient coal power, new energy, pumped storage, and hydropower. The planning and construction of a new energy system should be accelerated, as well as further consolidating the advantages of green finance in bringing dividends to the energy cleanliness. Second, the role of industrial structures and human capital as transmission mechanisms ought to be fully exploited. Further improvements should be made to the policy system in order to facilitate industrial restructuring, continued optimization, and adjustments to industrial structures and encourage the promotion and construction of an environmentally friendly industrial structure system. In addition, financial support for the attraction and cultivation of new energy talent should be increased and a talent supply system completed in order to make up for any shortcomings in the talent supply policies. Once again, the spatial spillover effects of green credits on HQED suggest that exchanges, coordination, and cooperation among regions should be increased, and a cross-regional linkage mechanism for green credit policies and energy and environmental governance should be established in order to fully realize green credit spatial contribution capacity to HQED. Finally, considering the regional heterogeneity of green credit policy contributions to HQED, a dynamic, differentiated, and precise green credit strategy are supposed to be tailor-made to suit local conditions. Furthermore, policy support for lagging regions should be increased in order to allow green credit incentives to effectively reduce regional development imbalances and form a good situation for the joint impetus of HQED in China.
Overall, the results obtained in this study enrich the horizon of green credit research and expand the research on the drivers of HQED, thereby laying a theoretical framework for further research. The findings also will potentially provide valuable references for future decision-making in order to steadily and comprehensively promote HQED. However, this study also had some deficiencies. For instance, limited by the availability of data, some very representative indicators were not included in the evaluation system of HQED, such as investments in energy conservation and environmental protection, and energy subsidies. Therefore, the meaning of HQED may not have been fully interpreted and comprehensively measured.

Funding
We are thankful to the Natural Science Foundation of China (No. 41901205), and the Natural Science Foundation of Jiangsu Province (No. BK20190482).

Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations
Ethics approval and consent to participate Not applicable.

Competing interests
The authors declare no competing interests.