Analyzing inclusive green growth in China: a perspective of relative efficiency

Inclusive Green Growth (IGG) has important reference value for China’s ecological civilization construction and transformation of economic development. Therefore, this study assesses China’s IGG level from the perspective of relative efficiency. The IGG efficiency (IGGE) was measured at the provincial level in China from 2000 to 2020 by using Super-Epsilon-Based Measure (EBM) model that considers undesirable outputs. The spatiotemporal pattern of IGGE was analyzed by kernel density estimation and spatial autocorrelation. The results indicate a fluctuating trend from 2000 to 2020 for the IGGE of China, and significant differences between regional and interprovincial IGGE were observed. On average, the eastern region presented the highest efficiency, while the level in the central regions was lowest. There is a positive spatial autocorrelation in the IGGE distribution, and the agglomeration of spatial distribution fluctuated during the study period. The IGGE has spatial spillover effects at the provincial level according to the spatial Durbin model. Among the influencing factors, the spatial spillover effects of industrial structure, government administrative capability, and industrialization level are significant. The regression results also confirm the Environmental Kuznets Curve effect between IGG and economic growth in China. Finally, some implicit policies can be established based on the empirical analysis.


Introduction
The Millennium Development Goals (MDGs) were set by the United Nations (UN) in 2000 to end poverty, hunger, disease, illiteracy, environmental degradation, and discrimination against women (United Nations 2000). The MDGs have been the core and basic framework of global development from 2000 to 2015, and they have been decisive in promoting global development (Way 2015). However, some new problems have gradually become prominent, such as the growing gap between the rich and the poor, global warming, environmental pollution, ecological destruction, and loss of biodiversity. Therefore, in September 2015, the UN general assembly, which includes leaders from 193 countries, adopted an action plan named Transforming Our World: The 2030 Agenda for Sustainable Development Goals (SDGs). This action plan aims to achieve a harmonious development of economic growth, social inclusion, and environmental sustainability by 2030 (Johnston 2016;Omer and Noguchi 2020). With the announcement of the 2030 Agenda, many countries have begun to adopt Inclusive Green Growth (IGG) as an implementation strategy for achieving the new SDGs. The concept of IGG was first proposed in the UN Rio + 20 summit in 2012 to merge the interests of green and inclusive growth (Albagoury 2016). Therefore, IGG provides a good reference for the implementation of SDGs.
Since its reform and opening-up, China's economy has developed rapidly. Simultaneously, people's living standards, industrialization, and urbanization have also greatly improved. The World Bank (WB) statistics show that China has become the second largest economy in the world. However, the remarkable results of China's economic development have come at the expense of substantial environmental and social costs (Chai et al. 2021;He and Du 2021;Cao 2022). On one hand, extensive economic growth has led to a series of resource and environmental problems, including resource exhaustion, environmental pollution, and ecological destruction (Liu et al. 2017;Liu and Dong 2021a); on the other hand, traditional development has led to social problems, such as the unequal distribution of the fruits of economic growth, widening the gap between the rich and the poor, and worsening the income gap between urban and rural residents (Weiss and Cattaneo 2017;Sun et al. 2022;Shen et al. 2022). For example, China's Gini coefficient was 0.465 in 2019, which was still higher than the international warning line of 0.4, indicating a wide disparity in wealth distribution. The emergence of these problems shows that China's economic growth is uncoordinated, unbalanced, and unsustainable. The root cause of these problems lies in the lack of "inclusiveness" and "greenness" of the traditional economic growth model (Chen et al. 2020). Therefore, a new development concept including "innovation, green, coordination, openness, and sharing" was presented in China's 13th 5-year Plan. In 2021, the 14th 5-year Plan and the outline of the long-term goal for 2035 emphasized the realization of a comprehensive green economic and social transformation, promoting the harmony between man and nature, and improving people's livelihoods and well-being. The simultaneous achievement of economic growth, social inclusiveness, and environmental sustainability has become a challenge for the Chinese government. Coincidentally, the IGG concept is similar to the concepts of sharing and greening new developments. Thus, IGG provides a tool for China to achieve sustainable development by addressing China's current problems. Therefore, the IGG level in China should be accurately measured.
Efficiency evaluation can be used as a performance standard to evaluate the allocation of resources and quality of economic growth. Data envelopment analysis (DEA) is an effective and widely used method to evaluate the relative efficiency of decision-making units (DMUs) with multiple inputs and outputs. It requires few indicators and does not require the setting of a particular functional form in advance or subjective weighting (Wu et al. 2014;Zhang 2020). Traditional DEA models can be divided into two types, radial and non-radial (Charnes et al. 1978). Radial models include the Charnes-Cooper-Rhodes (CCR) and Banker- Charnes-Cooper (BCC), and non-radial models include Slacks-based measure (SBM). However, both radial and non-radial models have inherent limitations. For example, the radial model requires all inputs or outputs' variables to change in the same proportion, and it neglects the influence of non-radial slacks variables. In turn, the non-radial model can account for all radial and non-radial relaxation variables, but it can easily lose the direct proportional information between the input or output target value and the actual value . To overcome these shortcomings, Tone and Tsutsui (2010) proposed the Epsilon-Based Measure (EBM) model by combing both radial and non-radial characteristics (input-oriented, output-oriented, and non-radial) into a unified framework. The EBM model has been widely used in relative efficiency assessments and presents good compatibility (Mushtaq et al. 2022;Zhao et al. 2022b). To address the undesirable outputs and ranking problem of DMUs, some scholars have integrated Andersen and Petersen's super-efficiency DEA model (Andersen and Petersen 1993) based on the EBM model to establish a Super-EBM model that accounts for undesirable outputs. This model provides a good calculation method for the relative efficiency assessment of our study.
Based on the above idea, the present study aimed to measure the performance of China's IGG by evaluating the China's IGG efficiency (IGGE) from the perspective of relative efficiency assessment. The spatiotemporal characteristics and factors that influenced the IGGE in China were also identified. To achieve these research objectives, the present study established an input-output evaluation index system for IGGE. The IGGE at the provincial level in China was calculated using the Super-EBM model considering undesirable outputs. Kernel density estimation and spatial autocorrelation analysis were used to analyze the spatiotemporal characteristics of IGGE. Finally, the influencing factors and spatial spillover effects of IGGE were identified through the application of the spatial Durbin model (SDM). We hope that our findings will serve as a reference for scholars studying IGG, and for governments to formulate policies to promote IGG.
The rest of the paper is organized as follows. The "Literature review" section reviews the literature on IGG and introduces the research gaps and main contributions of this study. The "Methods and data" section presents the main methods, evaluation index system, influencing factor variables, research area division, and data source. The "Results and discussion" section analyzes the calculation results, spatiotemporal patterns, and influencing factors of IGGE. Finally, main conclusion, policy recommendations, and limitations and outlook are proposed in the "Main conclusions and policy recommendations" section.

Literature review
Scholars have conducted numerous studies on IGG that provided useful references for our study. We reviewed the literature on IGG based on its definition, measurement methods, spatiotemporal differences, and influencing factors.

Studies on the definition of IGG
IGG is a combination of inclusive and green growth . Inclusive growth was first mentioned in 2007 by the Asian Development Bank (ADB) in its report "Inclusive growth: Asia for prosperity" (Ali 2007). So far, the concept of inclusive growth has focused on empowerment, holistic approaches, non-discrimination, better equality, pro-poor, and equitable opportunity growth (Ali and Son 2007;Ianchovichina and Lundström 2009;Besley and Cord 2006;Klasen 2010;Ranieri and Ramos 2013;Whajah et al. 2019;Ofori et al. 2022). It is easy to see that the concept of inclusive growth emphasizes the relationship between the economic and social systems. Thus, inclusive growth aims to achieve social inclusion, while ensuring economic growth. Green growth is proposed in the framework of the sustainable development concept, whereby there is consensus worldwide to achieve the SDGs. The concept of "green growth" was first introduced at the Fifth Ministerial Conference on Environment and Development in Asia and the Pacific in 2005. Green growth was defined as an environmentally sustainable economic process that promotes low-carbon development and benefits all members of society (UNESCAP 2005). Compared with inclusive growth, green growth places more emphasis on the relationship between economic and environmental systems (World Bank 2012;Capasso et al. 2019;Li et al. 2019;Kamath et al. 2022).
As an upgraded version of the inclusive and green growth concept, IGG emphasizes the coordination between the economy, society, resources, and environmental systems. However, this concept does not have a unified definition as it was only proposed in 2012 . IGG can be divided into two categories based on its emphasis. First, the economic development view considers IGG as a sustainable development approach. For instance, the World Bank (2012) defines IGG as "the economics of sustainable development". UNEP considers that IGG consists of social, economic, and environmental pillars (Bassi and Sheng 2012). The International Monetary Fund (IMF) defines IGG as a paradigm that aims to achieve sustainable development by reconciling the interests of developing countries (Spratt and Griffith-Jones 2013). The second perspective is welfare economics, which states that the purpose of economic growth is to improve welfare. Albagoury and Bouma argue that IGG requires increased welfare growth for the current and future generations (Albagoury 2016;Bouma and Berkhout 2015). Furthermore, Berkhout et al. (2018) proposed that IGG should include the improved welfare for the poor. In recent years, some Chinese scholars have enriched the definition of IGG, extending the framework of IGG to specific aspects, such as economic growth, social equity, the equitable distribution of the benefits of development, resource conservation, and environmental protection (Zhou and Wu 2018;Li et al. 2021).

Studies on the measurement of IGG
It is important to measure IGG levels, especially when governments formulate IGG strategies to achieve SDGs. There is extensive literature on methodologies for measuring inclusive and green growth. Among them, the methodology for the measurement of inclusive growth mainly considers the social opportunity function (Ali 2007;Sun et al. 2018), social mobility curve (Anand et al. 2013), Bonferroni index model (Silber and Son 2010), comprehensive evaluation index (McKinley 2010;Udah and Ebi 2016;Jiang et al. 2022), and DEA models (Chen and Qin 2014;Albagoury 2021;Fan and Liu 2022). However, the methodology for measuring inclusive growth often ignores green factors (i.e., environmental factors). There are also many methods for measuring green growth, including green GDP (Hoff et al. 2020), comprehensive evaluation index Zhao et al. 2022b;Naseer et al. 2022), and relative efficiency evaluation assessment (Cui and Liu 2021;Cai et al. 2022). Compared with the methodology for the measurement of inclusive growth, there are limitations in the techniques used for measuring green growth when considering the indicators of social factors.
IGG evaluation is mostly influenced by research carried out on inclusive and green growth measurements. IGG measurement methods can be divided into two types. The first considers the establishment of an index system with multiple dimensions, including economic growth, social inclusion, and environmental protection, to obtain a comprehensive index by assigning weights to different dimensions that reflect the level of IGG (Narloch et al. 2016;Jha et al. 2018). The subjective weighting method (Albagoury 2016), entropy method (Zhou and Wu 2018;Zhang et al. 2022), factor analysis method , coupled coordination degree model (Cao 2022), and the TODIM method ) are commonly used calculation models. The second is the relative efficiency evaluation method, which constructs an IGGE input-output index system using the DEA model. When constructing an IGGE input-output index system, scholars often include non-inclusive and environmental pollution factors in the evaluation system as undesired output indicators. The SBM model that considers undesired outputs (Zhu and Ye 2018;Zhao et al. 2019;Shen et al. 2022), directional distance function (DDF) model (Sun et al. 2020), and Malmquist-Luenberger index Wang et al. 2021) are commonly used evaluation models. Owing to the advantage of using the relative efficiency evaluation method to reflect the factor allocation of economic activities, literature on evaluating the IGG from the perspective of relative efficiency has rapidly been published in recent years. Unfortunately, existing evaluation index systems and evaluation models for IGGE have shortcomings that need to be improved.

Studies on the spatiotemporal differences and influencing factors of IGG
The literature on the spatiotemporal differences between inclusive and green growth is abundant. Scholars mainly use kernel density estimation , the Gini coefficient Yang et al. 2022), and exploratory spatial data analysis (Cheng and Ge 2020;Liu and Dong 2021a) to analyze the spatiotemporal evolution characteristics of inclusive growth or green growth at the national, provincial, and urban levels. In recent years, a few scholars have analyzed the spatiotemporal heterogeneity of IGG. For instance, Chen et al. (2020) used a convergence test to analyze the regional convergence of IGGE in the Yangtze River Economic Belt of China. Liu et al. (2021b) and Wang et al. (2022) used kernel density estimation and the Gagum-Gini coefficient method to analyze the spatiotemporal evolution of IGG. Unfortunately, these studies are inadequate for analyzing the spatial correlation of IGG. Similar to the literature on spatiotemporal differences, many empirical studies have been conducted on the factors that influence inclusive or green growth. According to the existing literature, the institutional environment (Agyei and Idan 2022;Nadeem et al. 2022;Ullah et al. 2022), natural resources and globalization (Majeed et al. 2022), foreign direct investment (Sylvaire et al. 2022), financial development Du et al. 2022), technological innovation (Luo et al. 2021;Zhao et al. 2022a), market environment , government expenditure (Ernawati et al. 2021;Mohsin et al. 2022), and environmental regulation (Ge and Li 2020;Yao 2021) are considered to be the main factors influencing inclusive or green growth. To date, there have been few studies on the systematic analysis of the influencing factors of IGG. A few scholars have explored the impact of economic structure optimization (Wang et al. 2021), foreign direct investment (Zhu and Ye 2018), and digital economy (Ren et al. 2022) on IGG. However, these studies seldom consider the spatial interaction of the influencing factors; that is, they ignore the spatial spillover effect of the influencing factors.

Research gaps and main contributions
In summary, academic research on IGG has yielded fruitful results. However, the following aspects can be further improved: First, IGG does not have a unified conceptual framework that is recognized by all scholars. Additionally, the evaluation index system of IGGE from the perspective of relative efficiency is not sufficiently comprehensive, especially because of the insufficient consideration of inclusive factors. Second, the evaluation method of the IGGE can be further optimized. Currently, the SBM and DDF models used in most studies have the problem of underestimating the efficiency value. Some improved methods, such as the Super-EBM model, can overcome this shortcoming. Third, the existing literature is lacking on the spatiotemporal differences and influencing factors of IGG, especially considering the lack of analysis of the spatial spillover effects of IGG and its influencing factors.
Therefore, the main contribution of our study, which attempted to remedy the above research gaps, includes three aspects: (1) We redefined the definition of IGG and constructed a relatively comprehensive indicator system to evaluate IGGE. (2) The Super-EBM model considering undesirable outputs was used to evaluate the IGGE in our study, which made the evaluation results more accurate.
(3) By adopting the spatial Durbin model, the influencing factors and spillover effects of IGGE were identified, which overcame the shortcomings of spatial correlation analysis on influencing factors in previous studies.

Methods
A Super-EBM model that considers undesirable outputs was first introduced to measure the IGGE. Based on it, Kernel density estimation and spatial autocorrelation test method were used to depict the spatiotemporal characteristics of IGGE. Finally, SDM was adopted to reveal the influencing factors and spillovers of IGGE.

Super-EBM model
The DEA model is a nonparametric method introduced by Charnes et al. (1978) to evaluate the relative efficiencies of DMUs. At present, radial and non-radial DEA models such as the CCR, BCC, and SBM models are widely used in efficiency evaluation. Meanwhile, the EBM model proposed by Tone and Tsutsui (2010) is considered the third pillar of technical efficiency measures in DEA. The EBM model integrates both radial and non-radial features in a unified framework, which accounts for the radial ratio between input frontier value and actual value, and also reflects the non-radial slacks variables that differ among inputs. Therefore, it can measure the efficiency more accurately. To solve the undesirable outputs in efficiency measurement and the ranking problems when the efficiency value of several DMUs is 1, we established a Super-EBM model with undesirable outputs by incorporating undesirable output variables and Andersen's superefficiency DEA model into the EBM model (Andersen and Petersen 1993). The Super-EBM model considers undesirable outputs under the assumption of non-oriented and constant returns-to-scale as follows: where r * is the efficiency value for the Super-EBM model in which there are DMUs to be measured; n, m, s, and q represent the number of DMUs, inputs, desirable outputs, and undesirable outputs, respectively; x, y, and z represent the inputs, desirable outputs, and undesirable outputs; λ is the intensity variable; θ is the calculated radial efficiency of the CCR model; s + r and s − p are slack variables of the desirable and undesirable output;w − i , w + r , and w − p are the weights of the input, desirable output, and undesirable output; and ε is a key parameter that combines the terms of radial and non-radial slacks.

Kernel density estimation
Kernel density estimation is a nonparametric method for estimating unknown probability density of random variables. The advantage of this method is that it only analyzes the distribution characteristics according to the data itself and has no requirements for the form of function. We can identify the distributional pattern of efficiency values by analyzing the kernel density curve. The kernel estimator for probability density function is given as follows: where K denotes the kernel function, n is the value of sample observation, x i is the random variable, and h is the smoothing parameter named bandwidth. The selection of bandwidth h follows the basic idea of minimum integral mean square (1) error, which is related to the smoothness of the kernel density distribution. Tobler (1970) proposed the first law of geography and argued that all phenomena in space are linked, and the closer things are, the higher the correlation, thus, the farther the distance, the lower the correlation. Spatial autocorrelation is a measurement of spatial data aggregation and a prerequisite for the establishment of a spatial econometric model. Testing the spatial autocorrelation of IGGE is to analyze its distribution characteristics. Global and local spatial autocorrelations are two common tools of spatial autocorrelation analysis. The Global Moran's I was proposed by Moran to test the spatial autocorrelation of economic variables based on the phenomenon of spatial random distributions (Moran 1948).

Spatial autocorrelation analysis
Global spatial autocorrelation is usually tested by Moran's I index. Its equation is as follows: where n denotes the number of regional units; x i and x j are the feature values of the i-th and j-th region, respectively; x is the mean of x; s 2 is the variance of x; and w ij is the measure of the spatial relationship between the i-th and jth region (adjacent is 1, non-adjacent is 0). The Moran's I index ranges from − 1 to 1. Moran's I > 0 suggests positive spatial autocorrelation, Moran's I < 0 indicates negative spatial autocorrelation, and Moran's I = 0 suggests no spatial autocorrelation.
The local autocorrelation index was proposed by Anselin (1995) to identify spatial agglomeration and heterogeneity. We used the Local Moran's I index to test the local autocorrelation of IGGE. The equation of Local Moran's I index is as follows: where i, j, s 2 , x , and w ij are defined in the same way as in Eq. (3). A positive I i indicates that the values of the i-th region variables are similar to those of the adjacent regions. When I i is negative, they are not similar. In this study, we used the Local Moran's I values to analyze the local spatial aggregation. As a result, four local aggregation types can be obtained, namely high-high (H-H), high-low (H-L), lowlow (L-L), and low-high (L-H). (3)

Spatial Durbin model
The spatial econometrics theory holds that a certain economic geographical phenomenon or an attribute value of a regional spatial unit is correlated to the same phenomenon or attribute value of the adjacent regional spatial units (Anselin 2013). It is difficult to measure the spatial effect through a general regression model, but spatial econometric models can solve the problem of spatial dependence of geographical phenomena. Spatial lag model (SLM) and spatial error model (SEM) are the most commonly used spatial econometric models. However, the disadvantage of these two models is that they only consider the spatial correlation of dependent variables, and not of independent variables. The SDM proposed by Elhorst can well overcome the abovementioned disadvantage of SLM and SEM, so that SDM can be regarded as an improvement of SLM and SEM (Elhorst 2014). Accordingly, we chose SDM to analyze the spatial spillover effects and influencing factors of IGGE. The equation of SDM is as follows: where y it denotes the dependent variable (IGGE); x is the independent variable (influencing factors variable); n is the number of spatial units; m is the number of independent variables; i represents the i-th region; j is the adjacent region of the i-th unit; t denotes the period; δ is the constant term; ρ and φ are the spatial lag coefficient of the dependent and (5) w ij x jtk k + it independent variables, respectively; w ij is the spatial weight matrix; θ k represents the parameters of the k-th independent variables; and ε it is the random error term vector.
A non-linear structure occurs in the SDM due to the introduction of the spatial weight matrix. Therefore, the regression coefficient cannot reflect all the effects of the independent on the dependent variables. To overcome this limitation, Lesage and Pace (2009) proposed the concepts of total, direct, and indirect effects of independent variables by using the partial derivative matrix method. A direct effect refers to the average impact of an independent variable x on the dependent variable y in a given region. Indirect effect refers to the average impact of an independent variable x on the dependent variable y of adjacent regions. Finally, total effect refers to the average impact of independent variable x on its own region and adjacent regions.

Research area division and data sources
Considering the difficulty of data acquisition, we selected 30 provinces and cities in mainland China as study areas, excluding Tibet, Taiwan, Hong Kong, and Macao. According to existing research and the geographical proximity, the 30 provinces were divided into three regions, namely eastern, central, and western regions (Fig. 1)

Input and output variables
The definition of inclusive growth originally refers to growth with equal opportunities. The aim is to maintain high economic speed and sustained growth, and also to promote social equity and inclusiveness by reducing and eliminating inequality of opportunities. The connotation of inclusive growth is relatively comprehensive, and it comprises poverty reduction, inequality reduction, productive employment, welfare sharing, gender equality, income distribution, good governance, and human development. Green growth further emphasizes the promotion of economic growth while ensuring that natural assets continue to provide resources and environmental functions on which human well-being depends. Based on previous researches, we endorse that IGG is the combination of inclusive and green growths. In other words, while maintaining sustained economic growth, we should emphasize the rational distribution of income, equality of development opportunities, sharing of development fruits, efficient and intensive utilization of resources, and development of a friendly ecological environment, so as to achieve sustainable development of society, economy, resources, and environment (Chen et al. 2020;Liu et al. 2021b). Considering the definition of IGG we proposed, inclusive and green factors should be considered when evaluating IGG from the perspective of efficiency evaluation. Therefore, in the selection of evaluation indicators, we chose indicators that reflect inclusive and green factors.
Based on the above analysis, we established an IGGE evaluation system in terms of input and output. The principles of scientificity, objectivity, systematicness, and data availability were followed in the construction of the evaluation system in Table 1. Labor, capital, and resources are basic production input elements. According to the existing literature, the year-end employment was selected as the labor input indicator (Zhu and Ye 2018), fixed asset investment was selected to represent capital input (He and Du 2021), and total energy and water consumption were selected as the resource input index (Zhu and Ye 2018;Cao 2022). In view of the gradually obvious role of science and technology in economic activities, full time equivalent for R&D personnel and intramural expenditure on R&D was selected as the technical input index (Zhao et al. 2021).
The output indicators can be divided into desirable and undesirable. In green growth efficiency, GDP is often used as a desirable output indicator . Meanwhile, we selected the per capita total retail sales of consumer goods, the number of health technical personnel per 10,000 population, and the teacher-student Total energy consumption (10,000 tons of standard coal), total water consumption (100 million cubic meters) Technology Full time equivalent for R&D personnel (person-year), intramural expenditure on R&D (10 thousand Yuan) Desirable outputs Socio-economic output GDP (100 million Yuan), per capita total retail sales of consumer goods (Yuan), number of health technical personnel per 10,000 population (person), teacher-student ratio in primary and secondary school (%) Undesirable outputs Non-inclusive output Ratio between per capita disposable income of urban and rural residents (%), ratio between consumption expenditure of urban and rural residents (%), unemployment rate (%) Environmental pollution Total volume of wastewater discharge (10 thousand tons), volume of sulfur dioxide emission (10 thousand tons), smoke (powder) dust emission (10 thousand tons), general industrial solid waste production (10 thousand tons) ratio in primary and secondary school to characterize the social well-being from economic activities in terms of resident consumption, medical, and education services Sun et al. 2022). For undesirable output indicators, we chose factors of non-inclusive and environmental pollution generated by economic activities. Referring to Cao (2022), the ratio between the per capita disposable income of urban and rural residents was selected to represent the unreasonable factors of income distribution. The larger the ratio, the more unreasonable the income distribution. The ratio between the consumption level of urban and rural residents was also used (He and Du 2021). Larger ratios indicate that the development fruits are less shared. The unemployment rate was selected to reflect the equality of development opportunities . The higher the unemployment rate, the more unequal the opportunities. These three indicators can together reflect the non-inclusive factor in economic growth. To increase the green factor, environmental pollution index was selected as an undesirable output. With reference to existing literature Guo et al. 2022;Chen et al. 2020), total wastewater discharge, sulfur dioxide emissions, smoke (powder) dust emissions, and general industrial solid waste production were selected as specific indicators of environmental pollution. These four indicators reflect the damage of economic activities to the environment. The smaller the index value, the less damage to the environment, and the more conducive the achievement of green growth.

Influencing variables
To explore the impacts of economic, social, policy, international trade, and other variables on IGGE, we selected some possible influencing variables as follows in Table 2.
(1) Economic development level: we adopted per capita GDP (PGDP) to reflect the economic development level (Guo et al. 2020). To verify whether there was an EKC (Environmental Kuznets Curve) effect on IGGE, the square of the logarithm of per capita GDP (PGDP 2 ) was also included in the regression equation (Goldin 1966). (2) Industrial structure (INDS): considering the great impact of the secondary industry on the environment, its proportion to regional GDP was selected to reflect its level (Ge and Li 2020). (3) Government administrative capability (GOV): the per capita general public expenditure was chosen to represent this variable (Han et al. 2021). (4) Industrialization level (INDU): industries contribute significantly to economic development, but they also have a great impact on the environment, so proportion of industrial added value to regional GDP was selected as a proxy variable (Wang and Zhang 2018). (5) Opening-up degree (OPEN): the entry of foreign enterprises has a significant impact on China's economy and environment, so the total investment of foreign-invested enterprises was chosen to discuss whether there was a pollution paradise hypothesis (Copeland and Taylor 2014). (6) Foreign trade dependence (TRAD): the proportion of total import and exports to GDP was selected to reflect this variable (Zhao et al. 2019). (7) Environmental regulation (ER): investments on industrial pollution treatment were selected to reflect the government's environmental regulation intensity (Ge and Li 2020). (8) Technological innovation (TI): the proportion of expenditure for science and technology to general public expenditure was incorporated into the regression equation . To avoid the problem of heteroscedasticity, the original data of per capita GDP, per capita general public expenditure, total investment of foreign-invested enterprise, and investments on industrial pollution treatment were processed by a logarithmic method.   Fig. 2, the average value of IGGE in China has increased from 0.496 in 2000 to 0.748 in 2020, and the efficiency type also improved from middle to middle-tohigh efficiency. The evolution of the average IGGE in China shows an evolutionary trend of decreasing and then increasing, with stage characteristics. IGGE in China can be divided into 3 stages: decline development stage (2000)(2001)(2002)(2003), IGGE decreased from 0.496 to 0.437, and the efficiency type decreased from middle to low-to-middle efficiency. Stable development stage (2003)(2004)(2005)(2006)(2007)(2008)(2009), IGGE changed from 0.437 in 2003 to 0.431 in 2009, and the efficiency type belonged to the low-to-middle efficiency. The reason is that during this stage, China changed its economic development model and takes a new type of industrialization path, and promoted the rapid development of productivity in all provinces and cities, which has improved IGGE to a certain extent, but the pollution problems caused by the rapid economic development have not been solved, the ecological environment problems are prominent, and the green development momentum is insufficient, which inhibits the rapid improvement of IGGE. Volatility upward stage (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019)(2020), the efficiency value increased to the highest value of 0.774 in the study period, and the rate of increase gradually accelerated, but there was a slight downward trend at the end of the study period, and IGGE fell to 0.748. The efficiency type evolved to the middle-to-high efficiency. Figure 2 also presents that there are significant differences between IGGE in eastern, central, and western China. During the study period, the efficiency value of the eastern region was between 0.506 and 0.878, the central region was between 0.337 and 0.697, and the western region was between 0.387 and 0.739, showing the imbalance of eastern > western > central. Under the dual promotion of location conditions and policy advantages, the eastern region has completed the optimization and upgrading of the industrial structure in the process of achieving high-quality economic development, and the proportion of high-tech industries and service industries has increased significantly to promote green economic and social development; meanwhile, local governments pay more attention to the improvement of people's well-being and social equity to enhance social development inclusiveness. Therefore, IGGE is higher in the eastern region.
According to the IGGE shown in Table 3, Hainan (0.911), Qinghai (0.839), and Tianjin (0.824) have higher levels and belong to the high efficiency type. Most central and western provinces such as Guizhou (0.359), Gansu (0.373), Yunnan (0.382), Anhui (0.383), and Shaanxi (0.396) have efficiency values much lower than the national average, belonging to the low-to-middle efficiency type. From 2000 to 2020, the IGGE value of Chinese provinces and cities has improved, but there are still significant gaps between provinces.

Spatiotemporal characteristic of IGGE
To examine the dynamic evolution and distribution characteristics of IGGE in China, the kernel density estimation was employed. We selected data from four time points (2000, 2006, 2012, and 2020) to depict the kernel density distribution map in Fig. 3. Some valuable conclusions can be drawn from Fig. 3. Firstly, in terms of distribution location, the center of the kernel density curve of IGGE turned to the left from 2000 to 2006, which indicates a declining pattern in the efficiency value during this period. The central position of the kernel density curve substantially moved to the right from 2006 to 2020, thus revealing a significant improvement in the efficiency value. This also confirms that the IGGE showed a fluctuating trend of decline followed by rise. Secondly, the kernel density curve shape shows that the curve presented negative skewed distribution characteristics from 2000 to 2006, which indicates that most regions were classified under low efficiency. Among them, the kernel density curve in 2006 showed a multi-peak shape, and the kernel density corresponding to the first wave peak was much higher than that corresponding to the other wave peaks on the right. This illustrates that the IGGE presented a trend of multipolar differentiation, and the proportion of low efficiency areas was significantly higher than that of high efficiency. In 2012, the kernel density curve was characterized by a normal-like distribution, indicating that the inter-provincial distribution developed towards the direction of equalization. The kernel density curve has a rightskewed distribution in 2020, and there was a right trailing characteristic of the kernel density curve, which indicated that the number of provinces with high efficiency level increased. In addition, from 2000 to 2020, the kernel density distribution of IGGE shows an overall change from wide-narrow-wide, thus indicating that the interprovincial differences of IGGE in China increased, and the spatial non-equilibrium increased significantly.

Fig. 4 Spatial pattern of IGGE in China
According to the classification results of the Jenks optimal natural fracture method, we selected data from four time points (2000, 2006, 2012, and 2020) to draw the spatial distribution map of IGGE in Fig. 4. From 2000 to 2020, there were significant spatial disparities of IGGE among the provinces in China. In 2000, the high efficiency type was distributed in Tianjin, Qinghai, and Hainan; middle efficiency was mainly distributed in northeast regions; and the low and low-to-middle efficiency type were contiguously distributed in most areas of the eastern coast, central, and southwestern regions; and IGGE was significantly lower in the provinces of southern than in the north regions. In 2006, the vast majority of Chinese provinces and cities were in the low and low-to-middle efficiency types, with only Hainan and Tianjin in the high efficiency type, and the overall low efficiency equilibrium stage. In 2012, IGGE has improved, but the overall level is not high, there is no high efficiency type, only Beijing, Tianjin, Shanghai, Hainan, and Qinghai are middle-to-high efficiency types. Meanwhile, a middle efficiency type distribution belt is formed in the eastern region, but the western region is still dominated by lowto-middle and low efficiency types. In 2020, the number of middle-to-high and high efficiency types will increase significantly, with the formation of high efficiency agglomerations in Bohai Rim Region, Yangtze River Delta, and Pearl River Delta in the eastern coastal region. The low-to-middle efficiency types decrease and are only distributed in some provinces and cities in the northeast, northwest, and southwest. To sum up, the areas with high efficiency level are mainly distributed in coastal and policy-oriented areas, which have strong economic vitality and high resource allocation rates, and attach importance to high-quality economic development and coordinated development of ecological environment and people's well-being, with high IGGE efficiency; low efficiency types mainly distributed in some provinces in the northwest and southwest, the industrial development in these areas started late, the foundation is weak, and they pay more attention to economic development, with insufficient investment in ecological improvement and livelihood protection, which hinders the improvement of IGGE.

Spatial autocorrelation characteristic of IGGE
According to Eq. (3), the Global Moran's I index of IGGE in China from 2000 to 2020 can be obtained by using the GeoDa software in Table 4. The results showed that the Global Moran's I index of IGGE was positive during this period, and most years passed the significance test and showed different degrees of significance. This basically explains that the IGGE of the 30 provincial units in China has a certain positive spatial autocorrelation, and presented a spatial agglomeration mode. Furthermore, the Global Moran's I index showed a fluctuating trend of up-down-updown, much like the "M" letter, which indicates a spatial autocorrelation trend of strengthening-weakening-strengthening-weakening. That is to say, the spatial distribution of IGGE alternates between agglomeration and decentralization, but the dispersion trend is more significant.
The Global Moran's I index analyzes spatial autocorrelation characteristics from a global perspective, and it cannot indicate the local spatial aggregation pattern of IGGE . Therefore, in this section, we used the Local Moran's I index to test and draw the spatial agglomeration map of IGGE by combining it with Moran's I scatter diagram. As shown in Fig. 5, Chinese IGGE mainly exhibits the characteristics of high-high and low-low spatial agglomeration, which also confirms the above conclusion that IGGE exhibits positive spatial autocorrelation characteristics. In 2000, high-high (H-H) spatial agglomeration was less and was mainly distributed in Beijing, Jilin, and Xinjiang, while low-low (L-L) spatial agglomeration was distributed in the eastern coastal, central, and southwestern regions. High-low (H-L) and low-high (L-H) spatial agglomeration were distributed in the northeast and northwest regions. In 2006, the distribution area of L-L expanded rapidly, and it was distributed in the northeast, central, southwest, and eastern coastal regions of China. The number of H-L and L-H aggregation types decreased. In 2012, IGGE is mainly dominated by H-H and L-L spatial agglomeration types, and the number of H-H has increased; it is mainly distributed in the eastern coastal area. In 2020, the number of H-H is very small, and it was distributed in Tianjin, Shanghai, Fujian, and Guangdong. The distribution area of L-L was reduced, and it was mainly in northeast, southwest, and central regions.
During the study period, the number of H-H was very small, and it was mainly distributed in eastern coastal regions; compared with other provinces, these provinces have a high level of economic development, a strong environmental awareness, and a good social security system; thus, IGGE is high. The distribution area of L-L was large and showed a trend of shrinking. The H-L and L-H were distributed between H-H and L-L distribution regions, with At the same time, L-L aggregation regions also need to continually improve their IGGE.

SDM estimation results
The spatial correlation test showed a significant spatial correlation in the IGGE in China (Table 5). Consequently, the spatial econometric model should be used to analyze the influencing factors of IGGE.    Table 6. As shown in Table 6, the spatial autoregressive coefficient of IGGE was 0.510, and it was statistically significant at the 1% significance level, which indicates that IGGE has significant spatial spillover effects among Chinese provinces and cities.
In the regression results, the coefficient of INDU was significantly positive. This factor had a positive impact on the IGGE of the provinces where it is located. The coefficient of INDS, GOV, and ER were significantly negative, thus indicating that these three factors have a negative influence on IGGE. The spatial lag coefficient of GOV and INDU were significantly negative, which suggests that GOV and INDU have an inhibiting effect on the IGGE of neighboring regions. The spatial lag coefficient of INDS and ER were significantly positive, which indicates that optimization of industrial structure and strengthening of environmental regulation in a region can promote the IGGE of neighborhood regions. However, the regression coefficient of OPEN and TRAD was not significant.
An interesting regression result is that the coefficient of PGDP was significantly negative, but the coefficient of PGDP 2 was significantly positive. This result indicates that the relationship between economic development and IGGE is not a simple linear relationship, but a U-shaped relationship. That is, upon economic development, IGGE tends to decline first and then rise. If we regard IGGE as a policy tool for the government to realize sustainable development, the improvement of IGGE would help reduce non-inclusive factors and environmental pollution. Therefore, IGGE was considered a negative alternative indicator of environmental pollution and social non-inclusiveness. Thus, the EKC theory used to explain the relationship between economic development and natural environment can also be used to understand the U-shaped relationship between IGGE and economic development (Destek and Sarkodie 2019). In other words, the relationship between economic development and IGG in China is consistent with the EKC theory. Furthermore, as the SDM regression coefficient cannot fully reflect the overall influence of the independent variable on the dependent variable, we decomposed the spatial spillover effects.

Decomposition of spillover effects
According to the SDM results and the concepts of direct, indirect, and total effects proposed by Lesage and Pace (2009), we calculated the direct, indirect, and total effects of the factors influencing IGGE in China. As illustrated in Table 7, the INDS, GOV, and INDU coefficients of indirect effect were significant. This result indicates that the independent variables of INDS, GOV, and INDU have spillover effects, whereas the other variables do not have significant spillover effects.
The PGDP and PGDP 2 coefficients of direct effect were significant, and statistically significant at the 1% level, while the indirect effect coefficients were not . This suggests that the U-shaped relationship between economic growth and IGGE only exists in the province and has no spillover effect on neighboring provinces. It should be noted that the relationship between China's economic growth and IGGE from 2000 to 2020 is on the left of the U-shaped curve (Guo et al. 2020). The direct effect of INDS was significantly negative, the indirect effect was significantly positive, and the total effect coefficient was positive but not significant (Wang et al. 2021). This showed that the larger the proportion of the secondary industry, the stronger the hindrance to IGGE. While the secondary industry promoted economic development, the development of high pollution and energy consumption industries will bring serious resource, environmental, and social problems, inhibit the coordinated and sustainable development of economy, society, and environment, and hinder the improvement of IGGE. The indirect effect coefficients are significantly positive, which indicates that INDS has a positive spillover effect on neighboring areas, driving the improvement of IGGE in the surrounding provinces and cities.
The regression coefficients of total, direct, and indirect effects of GOV were significantly negative . This showed that the government administrative intervention has a significant inhibitory effect on the promotion of IGGE in a province and its surrounding areas. The government intervention and regulation of social and economic activities are conducive to the optimal allocation of resource elements. However, the government administrative intervention is often oriented to economic development, which will lead to insufficient investment in the improvement of livelihood, social equity, and environmental protection, which hinders the improvement of IGGE to a certain extent.
The coefficient of INDU indirect and total effects was significantly negative; the direct effect was significantly positive. The industrialization is conducive to promoting local economic growth, expanding employment, and increasing people's income. Consequently, the emphasis on reducing the industrial scale is not conducive to the transformation of the development pattern. On the contrary, actively promoting high-tech industries is beneficial to the development of a green economy and reduction of unemployment. The spillover effect of provinces and cities with high level of industrialization to the surrounding areas will increase the resource consumption and pollution discharge of the neighboring regions, which is not conducive to the green development of the neighboring provinces and cities, thus hindering the IGGE of the neighboring provinces and cities.
The regression coefficients of direct effects of ER were significantly negative, and the total and indirect effect were non-significant (Ma et al. 2022). This result indicates that the environmental protection systems and policies of local government have a negative effect on the IGGE of a province. This conclusion does not support the "Porter hypothesis" (Porter and Van 2017). As a conventional policy tool of the local government, EPI aims to promote a coordinated development of economic growth, resource utilization, and environmental protection. However, the enhancement of environmental regulation intensity can increase the costs of environmental governance and economic burden to the regulated enterprises, which would affect the expansion of enterprises and increase of employment, thus inhibiting the improvement of IGGE in the province.
The regression coefficients of the total, direct, and indirect effect of OPEN, TRAD, and TI were not significant, thereby indicating that opening up, increasing trade quotas, and improving of technology level do not significantly promote or inhibit the IGGE of a province and its neighbors (Guo et al. 2020;Song et al. 2022).

Main conclusions
From the perspective of relative efficiency, this study developed an IGGE evaluation index system based on the redefined definition of IGG. We measured the IGGE in China from 2000 to 2020 by adopting a Super-EBM model that considered undesirable outputs and analyzed the spatiotemporal pattern characteristics and factors that influenced IGGE. According to our research, China's IGGE has significantly improved over the past 20 years. There are significant gaps and spillover effects among various regions in the IGGE, indicating the necessity of promoting coordinated regional development strategies. This study provides a reference index system and measurement method to evaluate IGG levels. It can thus complement theoretical research on IGG and can provide a reference for the Chinese government to formulate policies to coordinate economic development, social inclusion, and environmental protection. The main conclusions are as follows: Firstly, the IGGE in China showed an evolutionary trend of first falling and then rising from 2000 to 2020, with the efficiency value rising from middle to middle-to-high efficiency. There were obvious differences of IGGE among the eastern, central, and western regions, showing a spatial imbalance of eastern > western > central. Meanwhile, the interprovincial differences of IGGE showed a widening trend. The spatial imbalance increased significantly, and the number of provinces with medium-to-high and high efficiency also increased, forming high efficiency agglomerations in the Bohai Rim, Yangtze River Delta, and Pearl River Delta.
Secondly, there was a significant positive spatial autocorrelation of IGGE. The spatial distribution pattern of IGGE was an agglomeration mode. The number of H-H areas was small, and it was mainly distributed in eastern coastal regions of China. The number of L-L areas was large and decreased. The number of H-L and L-H areas was small.
Thirdly, based on the regression results of SDM, the IGGE had a spatial spillover effect at the provincial level. IGGE was influenced by the efficiency of adjacent regions, and it affected the efficiency of adjacent regions. Among the influencing factors, INDS, GOV, and INDU presented significant spatial spillover effects. The relationship between economic development and IGGE in China was consistent with the EKC theory.

Policy recommendations
Based on the empirical analysis results of this study, the following policy implications may be extracted. First, the Chinese government should fully leverage the spatial spillover effects of IGG. Regional exchanges and cooperation regarding human resources, technology, and capital should be encouraged. The cross-regional flow of production factors can be realized through industrial and technological diffusion. It is also necessary to further improve the infrastructure for transportation and communication. Second, different regions should play different roles in achieving IGG. At present, developed provinces on the east coast should make efforts to enhance the inclusiveness of the benefits of development and strengthen the radiation effect on the central and western regions. The central and western regions should enhance their economic strength and undertake industrial relocation from the eastern region to create more job opportunities. When undertaking industrial transfers from the east, thresholds for access should be set while considering environmental protection. Third, the optimization and upgrading of industrial structures and the establishment of modern industrial systems should become important tasks for local governments. Blindly reducing the proportion of industry and excessive capital inflow into the virtual economy also needs to be avoided. For local governments, actively developing high-tech industries and promoting the green and intelligent transformation of traditional industries will also be of great help to the IGG. Fourth, the policy system for environmental regulations should be improved. The application and intensity of environmental regulations should be adjusted to ensure the congruent development of economic and environmental goals. Regional coordination and cooperation of environmental regulation policies should also be emphasized. Finally, local governments should improve their degree of openness, pay more attention to the supervision of foreign direct investments, and encourage industries to move from the mid-low end of the global value chain to the mid-high end. In addition, establishing an industry-university-research collaborative innovation system could be valuable for speeding up the commercialization of scientific and technological achievements for IGG.

Limitations and outlook
There are limitations that need to be studied further. First, the evaluation index system of IGGE can be further improved. For example, in this study, we used data from statistical yearbooks, which lacked data on residents' subjective feelings, especially on their perception of inclusiveness. In the future, data sources could be expanded by issuing questionnaires. Second, the TFP of IGG was not measured, and the power source of IGG could potentially be fully investigated through the measurement of TFP. In addition, the "U" relationship between economic growth and IGGE was confirmed. Future research could use threshold regression analysis to further explore the threshold effect between economic growth and IGGE.