The construction efficiency study of China National Ecological Civilization Pilot Zone with network SBM model: a city-based analysis

The Chinese government proposed the establishment of China National Ecological Civilization Pilot Zone in 2016 to further explore the coordinated development of economy and environment. Fujian, Jiangxi, and Guizhou provinces were selected as the first batch of pilot zones. After years of exploration, it is necessary to discuss and summarize the construction progress of the three pilot zones from the perspective of the city. In this study, first, the ecological civilization pilot zone construction system was decomposed into an economic construction subsystem (ECS) and an environmental optimization subsystem (EOS). Then, a two-stage network SBM model was adopted to calculate the efficiencies of the subsystems, and the Kruskal–Wallis test was used to measure the efficiency difference. Finally, a panel data regression model was applied to explore the influencing factors of both subsystems. The results show that the ECS efficiency is higher than that of the EOS, and the ECS efficiency in Fujian is significantly better than that in Jiangxi and Guizhou. However, there is no significant difference in EOS efficiency in the three provinces. Furthermore, industrial structure and population agglomeration have a significant effect on ECS efficiency, environmental regulation has a significant impact on EOS, and the technology level has a significant impact on both subsystems. Based on the results, policy implications for improving the efficiency of the two subsystems were given respectively.


Introduction
Due to historical reasons, China's economic and environmental development is unbalanced and multilevel. Therefore, there is an urgent need to establish a benchmark and explore new paths for ecological civilization construction. To this end, the Chinese government has proposed the establishment of a unified and standardized national ecological civilization pilot zone. In August 2016, the document "Opinions on Establishing a Unified and Standardized National Ecological Civilization Pilot Zone" was released, and Fujian, Jiangxi, and Guizhou were identified as the first batch of pilot zones. The document pointed out that the pilot zones need to present a relatively complete ecological civilization system and form a few major systems that can be promoted across the country.
The National Ecological Civilization Pilot Zone is of great importance in promoting ecological civilization construction in China, and its construction achievements can be used as a reference for other regions. However, there are few studies on the construction efficiency of ecological civilization pilot zones. Moreover, most of the literature studying eco-efficiency is based on single-stage DEA models or on provincial administrative regions, which cannot either scientifically characterize the logic of ecological civilization construction or targeted enough. Therefore, from the perspective of the city, this study explores the efficiency and influencing factors of the subsystems of economic construction and environmental optimization in 26 cities of the pilot zones and provides targeted policy implications for ecological civilization construction for the government.
The rest of this study is organized as follows. The "Literature review" section sorts out the relevant literature and puts forward the innovation of this study. The "Methods" section introduces the methods used in the study. The "Samples and data" section states the samples and indicators of the study. The efficiencies and influencing factors of the two subsystems are analyzed, and the policy implications are proposed in the "Results and discussion" section. Finally, research conclusions are summarized in the "Conclusions" section.

Ecological efficiency measurement study
Due to the strong Chinese characteristics of the term "ecological civilization," international scholars often use "ecological efficiency," "environmental efficiency," "sustainable development efficiency," "green efficiency," and other similar terms when studying economic, social, and ecological environment issues. Research on the efficiency of ecological civilization construction involves the integration and development of these types of efficiency. Consider that there are very few literatures that simply study the construction efficiency of China's ecological civilization pilot zones and combined with the connotation of ecological civilization construction, the "Literature review" section will cover the research literature on the efficiencies mentioned above. Sarkis (2001) introduced data envelopment analysis (DEA) into the study of ecological efficiency for the first time, which started the prelude to the wide application of the DEA method in the field of ecological civilization construction efficiency research.
Single-stage DEA methods are broadly used to explore the efficiency of ecological civilization construction. In recent years, researchers have begun to incorporate undesirable output into DEA models. Muhammad et al. (2022) evaluated the environmental efficiency of the 64 Belt and Road Initiative (BRI) by taking carbon dioxide emissions as undesirable output. Peng et al. (2020) employed a superefficiency SBM model in the energy eco-efficiency measurement of 13 prefecture-level cities in Jiangsu province from 2008 to 2017 with the total environmental pollution discharge selected as undesirable outputs. Sala-Garrido et al. (2021) used a cross-efficiency DEA model to evaluate the eco-efficiency of several English and Welsh water companies, taking greenhouse gas (GHG) as the undesirable output. Pishgar-Komleh et al. (2021) also took GHG as the undesirable output and applied a window SBM model to study the dynamic eco-efficiency of the agricultural sector in European Union countries. To sum up, only the final discharge of pollutants is covered by most single-stage DEA studies. In fact, however, these pollutants have been treated before they are discharged, leaving only some pollutants that cannot be completely eradicated due to technical limitations and eventually discharged. In other words, the logic should be that pollutants are produced by economic production and are subsequently used as inputs, along with investment in pollution control, to finally produce pollutants that have been harmlessly treated and some part untreated. However, the single-stage DEA model regarded the whole system as a "black box" which failed to identify from which subsystem the problem arose by measuring the efficiency directly according to the initial input and final output; thus, comprehensive efficiency evaluation cannot be made.
Recently, some scholars started to explore the efficiency of ecological civilization construction using the network DEA model. Wu et al. (2015a) applied a two-stage network DEA model to study the efficiency of regional energy conservation and emission reduction. Since then, an evaluation framework that divides the ecological civilization construction system into two subsystems has been gradually implemented. The subsystem naming and model establishment of various studies on the two-stage efficiencies of ecological civilization construction are different.
Firstly, the name and the definition of the subsystems are usually based on the evaluation perspective by the existing studies. Mavi et al. (2019) constructed a two-stage efficiency analysis framework that includes eco-efficiency and eco-innovation in OECD countries. Li et al. (2020a) divided economic development of China into the production stage and the processing stage. Wang and Feng (2020) applied the "production-governance" system evaluation framework to Chinese provincial industrial research. Long (2019) proposed a two-stage framework of international ecological civilization construction from the perspective of the ecological welfare, which contains ecological economic transformation and economic welfare transformation.
Secondly, the two-stage DEA models established in the studies are widely radial or non-radial DEA models. However, as a non-radial model, the SBM model can measure the efficiency of the DMUs more accurately by directly incorporating the slack variables of the indicators into the objective function; thus, the studies using network SBM model are reviewed in this paper. Song et al. (2017) developed a network SBM model to measure the production efficiency and environmental efficiency of China's coal-fired power generation industry under two different tax policies, which are the pollution tax policy and the tax deductions or exemption policy. Li et al. (2018) employed a window network SBM model in the environmental performance evaluation of Chinese industrial systems. Li et al. (2020b) applied a dynamic and network SBM model to analyze the regional energy and air pollution reduction efficiencies of China. Tang et al. (2020) conducted two network SBM models to evaluate the eco-efficiency of China's industrial system from the perspectives of outside peers and inside managers. Shi et al. (2020) established a modified dynamic two-stage SBM model for the evaluation of the efficiencies of the wastewater treatment stage and health stage of China's 30 provinces. The above studies applied a network DEA model to present the internal process of ecological civilization construction more scientifically. However, the studies were mainly aimed at countries and provinces, which were relatively macroscopic and lack exploration of small samples in specific regions, so the research was not sufficiently detailed.

Influencing factor study
Based on the PEST model of macro environment analysis, indicators of politics, economy, society, and technology are often chosen to study the influence mechanism of ecological civilization construction efficiency. In politics, environmental regulation is the most common representative indicator which quantifies the policy efforts imposed by the government on environmental governance (Chen et al. 2021;Hou and Song 2021). In economy, the indicators are usually GDP (Liu et al. 2017;Xia et al. 2021), industry structure (Xiang et al. 2021;Han et al. 2021), and external openness (Xiang et al. 2021;Sun and Wang 2022). In society, the specific indicators mainly include education level (Bonfiglio et al. 2017;Diaz-Villavicencio et al. 2017) and population density (Xia et al. 2021;Diaz-Villavicencio et al. 2017).
Meanwhile, the representative indicator of technology is the investment in science and technology (Song et al. 2022) and the research and development (Luo et al. 2021). However, most of the current measurements of influencing factors were based on the calculation results of a single-stage DEA model that regards the ecological civilization construction process as a "black box," which inevitably failed to scientifically characterize the real situation within the sample, thus reducing the credibility of the conclusions.

Summary and innovation point
From the literature review above, it can be found that there are still the following deficiencies in the research on the efficiency of ecological civilization construction.
First, most literature regarded the ecological civilization system as a "black box" by using single-stage DEA model, which failed to show the real logic of the internal operation of the system, making it difficult to discover the real reasons for non-DEA effectiveness, so the results of efficiency evaluation and influencing factor measurement are less scientific.
Second, most studies using network DEA models took samples with a wide geographical range, such as countries and provinces, as the research objects, the heterogeneity among which makes the research insufficiently detailed and the countermeasures less targeted.
At present, there is still a scarcity of articles using the network SBM model to study cities in China's National Ecological Civilization Pilot Zone. After years of construction, how efficient is the ecological civilization construction of the 26 cities in the three pilot zones? What are the factors that influence the efficiency of economic construction and environmental optimization in these cities? What experiences can be provided from the sample cities for other cities in China? These questions remain to be addressed. Therefore, this study takes 26 cities in Fujian, Jiangxi, and Guizhou provinces of China National Ecological Civilization Pilot Zone as the research object, opens the "black box" according to the essence of ecological civilization construction, and constructs a two-subsystem analysis framework of economic construction and environmental optimization. In doing so, a network SBM model was applied to measure the efficiency of ecological civilization construction, and a panel regression model was used to determine the factors affecting the efficiency of the two subsystems. The main contributions of this study are that, on the one hand, the issues in the process of ecological civilization construction in 26 cities can be found to provide countermeasures for improving the efficiency of ecological civilization construction. On the other hand, an analysis of the construction and governance paths of benchmarking cities can be a practical reference for other cities. Färe and Grosskopf (2000) were the first to open a "black box" and construct a network DEA model. Most network DEA studies use undesirable variables as the final output or intermediate variables in efficiency assessment (Maghbouli et al. 2014;Podinovski and Kuosmanen 2011). Liu et al. (2015) proposed a more flexible two-stage network SBM model, which can not only open the "black box" to deeply explore the internal situation of each subsystem but also effectively deal with the undesirable variables in the ecological civilization construction system. Therefore, this model was selected as the measurement method for efficiency. The structure of the two-stage network model used in this study is shown in Fig. 1.

Network SBM
Suppose there are n independent DMU j (j = 1, 2, ..., n) , which represent the ecological civilization construction system of the jth city. As shown in Fig. 1, the ecological civilization construction of each city consists of two subsystems: the economic construction subsystem (ECS) and the environmental optimization subsystem (EOS). The input of the ECS is denoted by X DI 1 j = (X 1j , X 2j , ..., X m DI 1 j ) T , whereas the output represented as the input to the EOS, together with additional input X The two-stage SBM model considering undesirable output is constructed as follows: Fig. 1 Structure diagram of the two-stage network model where j and j denote the decision variables of ECS and EOS, respectively, and s are the slack variables. Denote the efficiencies of ECS and EOS as E 1 and E 2 , respectively. When the optimal solution of the above function is obtained, that is, the overall efficiency E * and the slack variables s are solved, the objective functions of subsystems can be written as follows: (1)

Kruskal-Wallis nonparametric test
The Kruskal-Wallis nonparametric test is used to explore whether multiple population distributions are identical (Zhang and Wang 2014). Because the distribution form of efficiencies from DEA cannot be obtained directly, the Kruskal-Wallis test can be used to explore whether there are differences in efficiencies between the DMUs of different groups. Before constructing the test statistic, the null hypothesis is generally set as. H 0 : there is no significant difference between groups. Then, the K-W test is constructed as follows: where k is the number of sample groups, N is the number of total samples, n i denotes the sample number of ith group, and R i represents the rank sum of the ith group.
If the population distributions are not identical, that is, the H 0 is rejected, a pairwise comparison can be further made.

Panel data regression
The data used in this study are called panel data because they contain information in three dimensions: crosssectional, period, and variable. A panel data regression model covering N cross-section member equations was constructed as follows: where y i represents the T × 1 dimensional vector matrix of the dependent variables and x i represents the T × k dimensional vector matrix of the independent variables. i is the intercept, and i is the k × 1 dimensional coefficient vector which affected by different individuals. i denotes the T × 1 dimensional random error term, which satisfies the assumptions of zero mean and homoscedasticity.

Study samples
The research subjects of this study were Fujian, Jiangxi, and Guizhou, which were identified as the first batch of national ecological civilization pilot zones in China. Fujian Province is in the southeast coastal area of China which has nine cities, with Fuzhou as its capital. Fujian is the "greenest" province in China, with a good foundation for ecological civilization construction. Jiangxi Province is also located in southeast China and on the south bank of the middle and lower reaches of the Yangtze River. Jiangxi has eleven cities, with Nanchang as the provincial capital. Guizhou Province, located in the hinterland of southwest China, is a transportation hub in southwest China and an important ecological barrier in the upper reaches of the Yangtze River and Pearl River. Guizhou Province has six districts and three autonomous prefectures with Guiyang as its provincial capital.
There are three types of resource-based cities classified by state council of China in 2013, including resourcegrowing cities, resource-mature cities and resource-conserving cities. Two cities in the sample, Liupanshui and Bijie, are the supply and reserve bases of energy resources in China with great potential for resource protection and economic and social development, which belong to resource-growing cities. Twelve resource-mature cities, such as Anshun, Nanping, and Sanming, are the core areas of energy resource security in China, with strong ability of resource guarantee and high level of economic and social development. Four resource-conserving cities, namely Jingdezhen, Xinyu, Pingxiang, and Ganzhou (Dayu County), are the key areas to accelerate the transformation of the economic development due to their resource depletion, lagging economic development, prominent livelihood issues, and serious ecological pressure.

Index system
Economic development and environmental protection are the two most distinctive themes in the construction of an ecological civilization pilot zone. First, economic development requires the input of production factors such as human resources, capital, and energy to obtain material or service output to meet reproduction or consumption, finally achieving the goal of economic growth. Therefore, it was defined as an economic construction subsystem. Simultaneously, the pilot zone needs to optimize environmental factors such as air, water, and soil to ensure a good living environment; therefore, it was defined as an environmental optimization subsystem. The construction target of the pilot zone can be realized only through an organic combination of economic development and ecological protection.
Based on the connotations of the two subsystems, an index system was built as follows.

Economic construction subsystem
Labor input Labor force is a general term for the population with labor potential and is one of the important inputs of production factors. The number of employed persons at the year end X DI 1 1j was selected to represent the labor input (Demiral and Saglam 2021).
Capital input As one of the inputs of production factors, capital is the product of the development of commodity economy to a certain historical stage. Adam Smith defined capital in the Wealth of Nations as the asset that transfers the right to use in order to obtain profits. Investment in fixed assets X DI 1 2j was selected to characterize capital input (Yu et al. 2019).
Energy consumption Energy is a general term for materials that can produce energy such as heat energy and light energy and is also an important input factor in economic production. The amount of energy consumption X DI 1 3j was selected to represent the energy input (Quintano et al. 2020).

Gross regional product (GRP) GRP
is the market value of all the results of regional economic activities in a certain period and is the core indicator of national economic accounting (Yang and Zhang 2018).

Intermediates
Industrial pollution emissions Under the existing technology, it is inevitable to produce pollutants in the process of economic construction. Industrial pollution is the largest source of pollution China's economy and society. Industrial sulfur dioxide generation Z UODI 1j and industrial soot (dust) generation Z UODI 2j were selected to represent industrial pollution emissions, which were taken as the undesired output of the ECS and the input of EOS (Li et al. 2020a;Wu et al. 2015b).

Environmental optimization subsystem
Environmental financial expenditure Financial expenditure is the payment activities carried out by local governments to fulfil their functions, and its scale can reflect the policy tendency and implementation scope of the year. Expenditure for energy conservation and environmental protection X DI 2 1j was selected to characterize environmental financial expenditure (Li et al. 2020a). were selected to represent environmental benefits as the desirable output of EOS (Wang and Feng 2020).

Environmental benefits
All data were obtained from the "China City Statistical Yearbook" of 2016-2020, statistical yearbook, and statistical communique of national economic and social development of the sample cities, with some missing data filled by the moving average method. Three autonomous prefectures in Guizhou Province, Qiandongnan, Qiannan, and Qianxinan were removed because of serious data deficiencies. A descriptive analysis of the indicators is presented in Table 1.

Influencing factors
Based on the PEST analysis model, this study selected influencing factors from politics, economy, society, and technology to establish the regression model.
Environmental regulation (ER) ER may promote or inhibit the environmental protection effect of sample cities by the degree to which it is implemented. The ratio of energy saving and environmental protection expenditure to public finance expenditure was used to represent the ER (Hou and Song 2021).
Industrial structure (IS) IS can facilitate technological progress through the integration and allocation of resource elements, which is beneficial for economic development. The proportion of secondary industry in GDP was selected to represent the IS .
External openness (EO) EO may affect the processes of both subsystems by influencing the competitiveness of environment-friendly enterprises and the introduction of pollutingintensive or labor-intensive industries. The proportion of total exports and imports to GDP was chosen to characterize the EO (Xiang et al. 2021).
Population agglomeration (PA) Reasonable population agglomeration can optimize the population age structure and improve labor efficiency. Population density was selected to represent the PA (Diaz-Villavicencio et al.

2017).
Technology level (TL) TL can affect the efficiency of both subsystems by changing the mode of production and environmental protection. The number of patents granted was chosen to characterize the TL (Luo et al. 2021).
The descriptive analysis of the influencing factors is shown in Table 2.
According to the above description, IS, EO, PA, and TL were selected as ECS variables, while ER, EO, and TL were selected as EOS variables.

Efficiency calculation
The network SBM model, based on the assumption of constant returns to scale, was applied to calculate the overall efficiency and the efficiencies of the two subsystems, and the results are shown in Table 3. It can be seen from Table 3 that the average value of the overall efficiency of ecological civilization construction during the study period is 0.562, which is far from the DEA effective level, indicating that the construction efficiency of ecological civilization pilot zone was unsatisfactory. The ecological civilization construction of sample cities still needs further exploration and development.
The average efficiency of ECS is 0.708, which is higher than the average efficiency of EOS (0.623), indicating that the sample cities have not achieved coordinated economic and environmental development. Moreover, ArcGIS was applied to draw the spatial distribution of the efficiency for each subsystem, with the natural breakpoint method  Fig. 2, it is obvious that the ECS efficiency presents a trend of "gradually decreasing from east to west." The average efficiency of Fujian Province is 0.972, which is higher than that of Jiangxi (0.615) and Guizhou (0.484), indicating that the overall efficiency of Fujian is close to the benchmark level, which sets up an advanced economic development experience for Jiangxi and Guizhou to absorb. However, as shown in Fig. 3, the EOS efficiency had no obvious spatial distribution, with the efficiencies of Fujian, Jiangxi, and Guizhou being 0.690, 0.624, and 0.521, respectively. Compared with the ECS, the efficiency gap of the EOS narrowed considerably between the three provinces, leaving room for improvement.

Efficiency difference test
Because the efficiency calculation is not able to identify whether there are significant differences between the three pilot zones and to verify the necessity of selecting cities as the object in this study, the Kruskal-Wallis nonparametric tests of the two subsystems were applied. First, the following research hypotheses are proposed: H 1 : There is no significant difference in the efficiency of the ECS among the three ecological civilization pilot zones. H 2 : There is no significant difference in the efficiency of the EOS among the three ecological civilization pilot zones.
The test results are shown in Table 4. As shown in Table 4, the p-value of ECS is 0.000, rejecting the null hypothesis, indicating that there are significant differences in the efficiency of ECS in the three pilot zones. A pairwise comparison was made among the three provinces to further analyze the differences, and the results are shown in Table 5. Table 5 shows that the p-value of Guizhou-Jiangxi with the adjustment of the Bonferroni correction method is 0.507, indicating no significant difference in the ECS between these two provinces. On the contrary, the p-value of Guizhou-Fujian and Jiangxi-Fujian is 0.000 and 0.005 respectively. Combined with the efficiency measurement results in Table 3, it is shown that the efficiency of Fujian is significantly better than that of Jiangxi and Guizhou.
However, according to Table 4, the p-value of EOS is 0.206, thus accepting the null hypothesis. There is no significant difference in the efficiency of EOS in the three provinces. The results indicated that the EOS statuses of the three provinces are similar, and it is necessary for cities to further explore and improve the level of ecological optimization.
From the discussion above, the difference in the EOS efficiency is insignificant, while there is a significant difference in the efficiency of ECS, and the efficiency of Fujian is significantly higher than that of Jiangxi and Guizhou, verifying the distribution law obtained from the spatial distribution. Therefore, it is necessary to analyze the specific situation of the two subsystems from the perspective of the city and propose more realistic and targeted countermeasures.

Quadrant analysis of subsystem efficiency
To uncover the reasons for the efficiency differences in the sample cities, we further studied the efficiency distribution of the two subsystems. With the average efficiency of the two subsystems as the dividing point, the efficiency of the ECS as the horizontal axis, and the efficiency of the EOS as the vertical axis, the subsystem analysis Boston matrix was constructed, as shown in Fig. 4. (

1) Cities with high efficiency in both subsystems
As shown in the first quadrant, there are six cities laid in the first quadrant, namely Xiamen, Putian, Sanming, Xinyu, Longyan, and Nanchang, whose both subsystem efficiencies are higher than the sample means.
Xiamen, for example, ranked first with an efficiency of 1 in the ECS and 2nd with an efficiency of 0.988 in the EOS. In terms of economic development, Xiamen supported the development of high-tech enterprises by building innovation demonstration zones, introducing innovative talent, and subsidizing innovation finance. In addition, the government insisted on building typical green manufacturing enterprises and carrying out green credit interest discounts to optimize the environment for green financial development. In terms of environmental protection, Xiamen has focused on pollution prevention and control by fighting the battle for blue sky, clear water, and clean land and repeatedly carrying out the "100-day campaign to protect the blue sky." Besides, it released six ecological laws in 2019, including "Water Pollution Prevention and Control Law" and "Air Pollution Prevention and Control Law," which successfully provided legal protection for pollution control. (

2) Cities with low ECS efficiency but high EOS efficiency
There are three cities in second quadrant, namely Yingtan, Jingdezhen, and Anshun, which are all listed as resourcebased cities by the State Council of China. They are below average in the efficiency of ECS but above average in EOS.
Further analysis of Yingtan shows that the EOS efficiency is 0.997, which ranks first, but the EOS efficiency is only 0.563, which ranks 17th. ECS inefficiency is related to Yingtan's industrial homogeneity and small economic volume. As the "copper capital of the world," Yingtan took non-ferrous metal smelting as its only main industry, which reduced the market innovation vitality and slowed the urbanization process, finally made its ECS less efficient. In contrast, focusing on the goal of building a "green world copper capital," the government took green and recycling as the main line to lead industrial development, strengthening the environmental protection assessment of the copper industry. In addition, the authority implemented the "one enterprise, one policy" control principle, vigorously promoting industrial source desulfurization, dust removal, and sewage treatment.
(3) Cities with high ECS efficiency but low EOS efficiency Six cities are located in the fourth quadrant: Fuzhou (Fujian), Ningde, Quanzhou, Zhangzhou, Nanping, and Fuzhou (Jiangxi). The economic construction of these six cities achieved excellent results but simultaneously paid a relatively high ecological price.
For example, the efficiencies of Fuzhou's (Fujian) ECS and EOS are 1.000 and 0.531, respectively, with the former ranking first and the latter 18th, respectively. In the EOS, the manufacturing industry of Fuzhou has been dominated by traditional industries, such as textiles, chemical fibers, and metallurgy, which drive economic development accompanied by resource consumption and environmental pollution. However, the high efficiency in ECS may be related to a high level of openness. As a capital city of a coastal province and an important window to Taiwan, Fuzhou was approved to establish the "Fuzhou New Area" and "Fuzhou area of China (Fujian) Pilot Free Trade Zone" in 2015, laying a solid foundation for the city to broaden the field of external openness. The policy and location advantages have helped Fuzhou attract many foreign-funded enterprises to settle down, which has led Fuzhou into the "high-speed era" of open economy development.

(4) Cities with low efficiency in both subsystems
There are up to 11 cities located in the third quadrant, whose efficiencies of both subsystems are below the average. The cities are Liupanshui, Ganzhou, Bijie, Tongren, Yichun, Pingxiang, Zunyi, Jiujiang, Shangrao, Guiyang, and Ji'an, among which the first ten cities are resource-based cities.
Taking Bijie as an example, its ECS efficiency is 0.508, ranking 21st, and its EOS efficiency is 0.343, ranking 26th. From the EOS perspective, inefficiency is related to the weak power of transformation and development. Influenced by the historical legacy of a resourcebased city, the urbanization rate in Bijie was less than 48% at the end of 2019, with more than 670 thousand people living on subsistence allowances. In addition, the high dependence on mineral resources, together with the lack of high-tech industries, made this resource-growing city weak in gathering capital and talent, which led to serious labor outflow problems. From the ECS perspective, inefficiency is related to lagging mine geological environment management. The long-term mining of mineral resources has brought about a series of ecological problems in Bijie, such as land salinization and vegetation destruction.

Analysis of the influencing factors
Based on the background of economic globalization, the ecological civilization construction in China is affected by many external factors; thus, it is necessary to explore the factors that promote or inhibit the construction process of ecological civilization pilot zone.
A panel data regression model is implemented to test the relationship between each influencing factor and the ECS and EOS efficiency of the 26 cities. The p-values of the F and LM tests of the two subsystems were both significant at the 1% level, rejecting the null hypothesis; the Housman test also rejected the null hypothesis. Therefore, the fixed effects model was chosen to test the two subsystems. The test results are listed in Table 6.
As shown in Table 6, IS is found to have a significant positive impact on the efficiency of ECS, while TL and PA have a negative impact. ER has a significant negative impact on the efficiency of EOS, while TL has a positive impact.

Environmental regulation
In the EOS, the coefficient of ER is − 2.8972 and significant at the 5% level, indicating that every 1% increase in environmental regulation intensity will reduce efficiency by 2.90%, that is, the more government intervention, the lower the efficiency. There may be two reasons. First, from the raw data, 88% of the cities spent less than 5% of public finance expenditures on energy conservation and emission reduction in 2019, which was unreasonable. Second, the environmental regulatory measures adopted by the government may be poorly targeted and inadequately regulated.
Industrial structure The coefficient of IS in ECS is 1.0615, which passes the 1% significance level test, indicating that for every 1% increase in IS, the efficiency of ECS increases by 1.06%. There are two possible explanations for this observation. First, most cities still relied on the secondary industry as the pillar industry in the study period, with economies mainly driven by manufacturing, such as Jingdezhen, which is famous for its ceramic industry. Second, the manufacturing industry plays a role in improving the efficiency of economic construction through industrial rationalization, ultimately promoting high-quality economic development.

Population agglomeration
The coefficient of PA in ECS is − 0.9967, which is significant at the 10% level. A reasonable explanation is that when the population growth rate is greater than the capacity of economic growth to absorb, population density harms economic growth. Seventeen of the 26 sample cities belong to the central and western regions, the population agglomeration of which is characterized by a higher aggregation of the low-end labor force, making it difficult to exert the population agglomeration effect. Therefore, population agglomeration in pilot zones failed to promote economic development.
Technology level TL has a significant impact on the two subsystems. In the ECS, the correlation coefficient is − 0.0820, passing the significance test of 1%. This is because there are problems in the transformation of scientific research achievements in most cities, such as poor channels for transformation to enterprises and a lack of transformation motivation and platforms. However, TL has a positive impact on the EOS efficiency and can be divided into two reasons. First, pollution monitoring and treatment technology not only improved the efficiency of pollution treatment in enterprises but also provided data and technical support for urban environmental monitoring stations. Second, the construction of municipal ecological information platforms, such as the air pollution prevention and control platform and the "River chief System" information management platform, not only promoted the intelligent development of municipal ecological environmental management but also boosted the joint prevention and control of regional pollution.

Policy implications
According to the above analysis, the policy implications for strengthening the ecological civilization construction for cities in China based on the perspectives of the two subsystems are proposed.

Economic construction subsystem
(1) Promote industrial transformation and development.
The industrial structure, represented by the proportion of the secondary industry, has a promoting effect on the efficiency of the economic construction subsystem, indicating that the real economy is still the key point for economic development in the ecological civilization pilot zone. Twelve of the fourteen cities whose ECS efficiency is lower than the average are resource-based cities. Based on local industry structure, cities should follow the "3R" principle (reduce, reuse, recycle) and take the road of circular economy development.
Resource-growing cities should focus on industrial diversification while consolidating and upgrading traditional industries. Resource-mature cities should focus on the construction of industrial clusters and cultivation of alternative industries. Resource-declining cities should accelerate the transformation of old and new momentum and develop non-resource-dependent industries.
(2) Accelerate the application of scientific research achievements. The development of science and technology is an important part of China's high-quality development, and the measurement of influencing factors showed that the effect of technological development on efficiency is restricted by the low level of scientific achievement transformation, so the speed of scientific achievement transformation should be accelerated in the future. Universities and research institutions should be given greater autonomy, including the right to flexibly use funding for research projects and dispose of scientific and technological achievements. Enterprises should be encouraged to pay more attention to the "Industry-University-Research," establish cooperative alliances with universities and research institutions, and strive to turn research achievements into real productive forces. For example, Sanming, Ningde, and other cities in Fujian have stimulated the enthusiasm of scientific research personnel through policy guarantees, financial incentives to promote the transfer and implementation of scientific and technological achievements. (3) Optimize the structure of urban talent. According to the theory of environmental carrying capacity, human activities should not exceed the carrying range of environmental carrying capacity. It can be seen from the empirical analysis that the insufficient capacity of economic growth to absorb population and the scarcity of high-end talents in the central and western cities are the important reasons for the negative impact of population agglomeration on the efficiency of the economic construction subsystem. Cities should change the way talent is introduced by innovating talent incentive mechanisms, exploring the transition from policy attraction to environmental attraction, and then to cultural attraction. For example, Putian successfully introduced thousands of high-level talents through the implementation of the "Hulan project." What calls for special attention is that central and western cities should avoid the phenomenon of "policy failure," which worsens the effectiveness of talent attraction.

Environmental optimization subsystem
(1) Improve the environmental protection system. Cities should change the concept of legislation and constantly update ecological laws and regulations. Meanwhile, the ecological justice system should be completed by improving the publicity of environmental information, building enterprise environmental credit mechanisms, and increasing penalties for violations. For example, Guiyang has issued several original programs tailored to local conditions, such as the "Clean Water Protection Plan of Guiyang," and has incorporated ecological environmental protection into the performance assessment of officials.
(2) Standardize the expenditure for energy conservation and environmental protection. There is a negative correlation between environmental regulation and the efficiency of EOS, mainly because of the unreasonable proportion and unreasonable use of environmental protection funds in most cities. According to the ecological compensation theory, the government should focus on environmental governance and spend more funds in areas such as ecological infrastructure construction, land remediation, and mine ecological restoration. Simultaneously, it is necessary to give full play to the role of the market in environmental protection and encourage private capital to participate in environmental protection. (3) Enhance pollution control technologies. The advanced mode and technology of pollution control make the efficiency of the environmental optimization subsystem benefit from the level of technology. Cities should promote excellent pollution control cases, such as learning Xinyu to implement the new mode of urban and rural pollution control, which is mainly distributed and supplemented by centralization. In addition, it is of great importance to update pollution control technologies, especially advanced waste gas treatment technologies, such as nanotechnology, microwave catalytic oxidation, and sewage treatment technologies, such as microbial treatment and chemical oxidation. Finally, cities should cooperate with each other to optimize the efficiency of joint prevention and control of regional pollution through information sharing.

Conclusions
This study took the 26 cities in Fujian, Jiangxi, and Guizhou provinces of the first batch of China National Ecological Civilization Pilot Zones as the research objects and decomposed the ecological civilization construction system into two closely related subsystems, economic construction and environmental optimization, by opening the "black box" from the essence of ecological civilization construction. The network SBM model was applied to calculate the efficiency of the two subsystems from 2015 to 2019, and the Kruskal-Wallis nonparametric test was used to explore the differences in efficiency across the test areas. The panel data regression model was then selected to measure the factors influencing subsystem efficiency. The main conclusions are as follows.
First, the efficiency of the economic construction subsystem was better than that of the environmental optimization subsystem in the sample cities. The efficiency of the economic construction subsystem in Fujian was significantly better than that in Jiangxi and Guizhou, but there was no significant difference in the efficiency of the environmental optimization subsystem in the three provinces. Six of the 26 cities achieved "double high" efficiency of the two subsystems, while another nine cities had only one subsystem with an efficiency higher than the sample average, and the remaining eleven cities had both subsystems with efficiency lower than the average. Second, the regression results show that the efficiencies of the two subsystems of the sample cities were affected by a variety of factors, among which the efficiency of the ECS was significantly positively influenced by the industrial structure and negatively influenced by the population agglomeration and technology level; the efficiency of the EOS was positively influenced by the technology level and negatively influenced by environmental regulation.
There are still some deficiencies in this study. Firstly, the limited environmental protection data disclosed by some samples led to the omission of the evaluation index system of ecological civilization construction and the exclusion of several cities in Guizhou Province. Secondly, the serial correlation problem in DEA model and the lack of description of a coherent data generating process of the panel data regression model used in the study may result in bias in empirical analysis results. Future study will further update the index system and the date of research data, exploring the application of bootstrap regression model in panel data of the study, so the comprehensiveness and scientific nature of the evaluation of China's ecological civilization pilot zones can be improved.
Author contribution Dan Liu: structural design, work coordination and supervision, literature collection and review, data collection and analysis, methodology, writing guidance and review.
Tiange Liu: literature collection and review, data collection, calculation and analysis, methodology, software, manuscript writing and modification, writing review and editing.
Qi Zhang: literature collection and review, data collection, calculation and analysis, writing review and editing.
Funding This study was supported by the National Social Science Foundation of China (Grant No. 19BGL012).

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.