Research on green development measurement and regional differences in the China's Yellow River Basin based on the ecological function orientation

The ecological protection and high-quality development of the Yellow River Basin have become a national strategy. This study takes 57 cities in the Yellow River Basin as research samples, employs the green total factor productivity (GTFP) to characterizes the level of green development, introduces an improved mixed distance (MEBM) Windows model and an algorithm containing the background differences to measure the level of green development, and further uses the Dagum Gini coefficient to analyze the regional differences. It shows that the level of green development of Yellow River Basin exhibits a downward trend during the study period when the ecological function orientation (EFO) is not considered, however, the downward trend is significantly narrowed when the EFO is considered, and 91.23% of the sample rankings is changed. According to these results, the study samples are divided into three types: the controlled development zones, the moderate development zones, and the transition development zones. The inter-regional differences are the main factors that causes the regional differences in the green development of Yellow River Basin. Based on the results, we propose the relevant policy recommendations that can provide a decision-making basis for promoting the ecological protection and high-quality development in the Yellow River Basin according to the local conditions.


Introduction
The Yellow River Basin connects the Qinghai-Tibet Plateau, the Inner Mongolia Plateau, the Loess Plateau, and the North China Plain, and straddles the three major geographical steps in China.It is an important ecological function area, energy and resource enrichment area, and economic development area in China.The ecological environment condition, resource utilization efficiency, and economic development of river basin directly affect the ecosystem, resource protection, and high-quality development of China.Therefore, promoting green development in the Yellow River Basin is of great significance for China's ecological security and balance, resource protection and sustainable utilization, and strengthening the foundation of economic and social development.However, the problems of ecological environment destruction and pollution in the river basin are more prominent in recent years with the rapid economic development.For examples, the proportion of poor grade V water was 12.4% among 137 water quality sections of Yellow River in 2018, which was higher than the national average of 6.7%.And the sulfur dioxide emissions of Yellow River Basin accounted for 27.78% of the national emissions, which seriously hindered the high-quality development of river basin.In October 2021, the Outline of the Yellow River Basin's Ecological Protection and High-quality Development Plan issued by The Communist Party of China Central Committee and The State Council proposed that 'Efforts should be made to ensure the long-term stability of the Yellow River, to improve the ecological environment of the Yellow River Basin, to optimize the allocation of water resources, and to promote high-quality development of the whole basin, so as to make the Yellow River a happy river that benefits the people'.This further indicates that the economic development model in Yellow River Basin urgently needs to shift to the green development, and the key to solving the above problems lies in accurately measuring the present level of green development in the Yellow River Basin.
Ecological spatial functional zones are an important component of territorial space planning.It aims to protect and improve the regional ecological environment, and is based on a correct understanding of the characteristics of the regional ecological environment.And it adopts different development strategies for the ecological functions, ecological sensitivity differences, and the impact of human activities in different regions, to achieve the regional green development.The upper, middle and lower reaches of the Yellow River Basin have different ecological endowments, so the ecological construction priorities are different.For a long time, the ecological space function of green development in the Yellow River Basin was not received enough attention due to the lack of land space planning, which makes the measurement method of green development level unreasonable and is unable to provide path guidance for establishing a green development pattern with complementary advantages.As a result, the policy of green development in the basin is not well targeted and effective.At the Symposium on Ecological Protection and High-Quality Development in the Yellow River Basin, President Xi pointed out that 'The regions along the Yellow River should transition from the current situation to make use of the water and mountain resources, grow crops, and develop agriculture, industry or business where conditions permit'.Therefore, it is of great significance to understand correctly the ecological functional differences in the different regions of Yellow River Basin, to measure scientifically the status quo of green development, and to implement the differentiated green development according to the local conditions.
In light of these motivations, this paper mainly solves the following three tasks.First, we use the green total factor productivity to characterize the level of green development of the Yellow River Basin, introduces the MEBM-Windows model to measure the level of green development of the Yellow River Basin, which does not consider the differences in ecological function orientation (EFO).Second, to restore the real level of green development in various regions and improve the comparability of research samples, we introduce the analyzed method which is based on the resource background differences to fully consider the constraint effect of EFO in the Yellow River Basin.Finally, we employ the Dagum Gini coefficient to analyze the regional differences of green development level in the Yellow River Basin based on EFO, and then put forward the policy recommendations which is suitable for the level of green development of different functional areas.
The rest of the paper is organized as follows.Section 2 is a literature review on the green development.Section 3 introduces the data and the methods.Section 4 outlines the results and presents the discussion.Section 5 offers the conclusions, the policy implications and the limitations.

Literature review
The traditional measurement method of economic development prospected by the GDP and economic growth rate is inapplicable under the global trend of sustainable development and green development.Considering the impact of resource constraints and environmental pollution on economic benefits, many scholars and policymakers have begun to focus on measuring the green development.The early measurement of green development mainly used the input-output method (Mardones and del Rio 2019), the energy analysis method (Wei et al 2020), and the ecosystem service value method (Ma et al 2020) to measure the economic activities of a country or region.These methods take into account the natural resources and environmental factors.Actually, it is difficult to accurately evaluate these values due to the lack of strict market prices for natural resources and environment in the green GDP accounting system.Based on this background, some countries set a series of indicators that can measure green development or sustainable development to comprehensively reflect the overall green development status of the evaluated objects.For example, the green growth measurement framework was established by the OECD and the green economy measurement indicator system was created by the UNEP.These indicator systems are designed for macro-evaluation and comparison among countries, but they are inapplicable to green development evaluation in some regions.Therefore, many countries and researchers have expanded the indicators according to the local conditions and further developed the comprehensive green development indexes suitable for the different regions.The most influential research results are mainly produced by the official, the international or the professional academic organizations, such as the Economic and Social Council of Asia-Pacific (UNESCAP), the California Green Innovation Index, and the China Green Development Index.Although the comprehensive index method has many advantages, such as the diverse evaluation scales together with intuitive and comparable data, it has a large demand for data and lacks the consensus on evaluation indicators.
Efficiency is the ratio of actual output to optimal output, which objectively measures the quality of economic development in a country or region.The productivity of green total factor aims to reflect the resource and environmental losses in a regional development through the consideration of resource and environmental costs.It is an effective measurement indicator for regional green development and is widely used in the field of regional green development evaluation (Tian and Feng 2023, Yang et al 2023a, Yu 2023).In the selection of efficiency measurement indicators, scholars pay more attention to the input indicators such as human resources, capital, energy and technological innovation, together with the output indicators such as economic benefits, social benefits and environmental damage.In terms of evaluation methods, the existing measurement of green development efficiency mainly includes the parameter stochastic frontier analysis (SFA) and the non-parametric data envelopment analysis (DEA).Because DEA method can simultaneously consider multiple inputs and outputs without setting a production function form (Su and Zhang 2020), it now becomes the mainstream method for measuring the efficiency of green development.The calculation of efficiency of green development using DEA first requires addressing the environmental pollution problem.The current solutions to this problem can be categorized into three types.① Treating environmental pollution (undesired outputs) as an input variable (Hailu and Veeman 2001).However, this method will fully reduce undesired output without affecting the increase in the desired output.It does not consider the actual production process or goes against the realistic situation and cannot reflect the essence of production process.② The data transformation of environmental pollution needs to meet the requirements of different constraints on 'bad' outputs (Zhou et al 2019).This method can better solve the efficiency evaluation problem of unexpected outputs, but data conversion will actually reduce the 'bad' outputs (Färe and Grosskopf 2004).Moreover, a strong convexity constraint is added so that the efficiency can only be solved under the variable returns to scale (VRS).③ We can consider environmental pollution as a variable with weaker treatability and introduce it into the production process simultaneously with economic output.There are mainly radial and non-radial algorithms.The radial algorithm is represented by the directional distance function (DDF) model proposed by Chung et al (1997), which is assumed that the increase in desirable outputs is proportional to the decrease in the undesirable outputs (Färe et al 2017).This method has been applied by many scholars (Zhang and Choi 2014, Amowine et al 2023), but it does not fully consider the slackness of inputs and outputs.To overcome this shortcoming, Tone (Tone 2001) proposed a efficiency measurement method of non-radial and non-angle slack-based, which allows the reduction ratio of factor inputs and the expansion ratio of desirable outputs to be different from the reduction ratio of undesirable outputs.The specific ratio depends on the slack amount (Yu et al 2021, Wen et al 2024).However, the non-radial algorithm standardizes the slack variables by evaluating the outputs and inputs of the objects, which easily ignores the statistical distribution information of the inputs and outputs of the entire sample.
With the continuous deepening of the development strategy of Yellow River Basin, its development and ecological protection have become a focus of academic attention, and related researches on green development have gradually emerged.Existing researches on the Yellow River Basin have mainly focused on the regional water resource status, distribution and governance (Zhang et al 2018a, Ren et al 2022), climate change (Sun et al 2017, Wu et al 2022), air quality (Yuan et al 2022), agricultural production and cultivated land (Aliyari et al 2021).Some scholars also regarded total factor productivity as a starting point to study the current status and regional differences of green development in the Yellow River Basin (Jiang et al 2021, Khan et al 2021, Guo et al 2022, Zhou et al 2023).For example, Ma and Wang (2022) constructed a Super-EBM model, which includes the labor input, the capital input, the energy input, and the environmental input, to measure the energy efficiency of green total factor in the Yellow River Basin, and found that the Yellow River Basin showed an upward trend during the study period.Yue et al (2023) incorporated social benefits as expected outputs of the green development measurement framework and analyzed the current status of green development in the watershed using the Super-DDF model.They also found that the overall green development in the watershed is on the rise.Liu et al (2023) pointed out that the traditional DEA method ignored the influence of external factors and random interference on the evaluation results.Therefore, the three-stage DEA model was adopted to measure the level of green development in the Yellow River Basin.The study found that the level of green development in the Yellow River Basin showed a downward trend when external factors such as the economic environment, the institutional environment, and the social environment were excluded.
In summary, there have been many discussions on the measurement of green development both domestically and internationally, and rich research results have been achieved, which laid a solid foundation for our research.However, the research gaps need to be further improved.(1) Firstly, strong sustainable development does not come at the cost of reducing the development of the environment and resources.In June 2012, it was officially recognized by the United Nations as a new green development paradigm, which has revolutionary significance in replacing the traditional brown economy paradigm.However, the theory of strong sustainable development has not been fully reflected in the green development research.In other words, the existing green development evaluations are more inclined to use the ecological damage indicators as the output indicators (ecological environment) and ignore the regulatory effect of environmental self-absorption capacity on environmental pollution.This makes it difficult to evaluate the regional green development status from the perspective of continuous optimization of macroeconomic system operation efficiency.(2) There is still room for further improvement in the calculation methods of green development efficiency, including the radial or non-radial algorithms.In general, the radial algorithms require the inputs or outputs to change in the same proportion when evaluating the efficiency, which is contrary to the realistic conditions.Non-radial algorithms ignore the original proportion information of the projected value of efficiency frontier.The difference in the calculation results is obvious when the relaxation variables with positive and zero values are selected in the linear programming method.At the same time, the traditional DEA model not considers the role of the time factor and is only suitable for analyzing the cross-sectional data, it therefore cannot be used for the vertical comparison.(3) The green development of the Yellow River Basin has not taken into account the heterogeneity of resource backgrounds.The existing researches on the measurement of green development imply a premise assumption: the assessed objects are comparable, and it is believed that the institutional supply, development constraints, and external environment faced by various regions are homologous and comparable.However, the Yellow River Basin in China has a vast territory, significant differences in natural conditions, and diverse regional functions.So far, few scholars have explored the impact of ecological functional differences on the measurement results of green development.This measurement that ignores the spatial pattern heterogeneity is not conducive to accurately identifying the natural endowments and comparative advantages of different regions.
The marginal contributions of this paper are as follows: (1) In view of the issue that the strong sustainability theory has not been fully reflected in the green development research, we construct a strong sustainable ecological environment research framework to objectively and comprehensively evaluate the level of green development in the Yellow River Basin.In this framework, the ecological environment damage and the ecological environment construction are incorporated into the ecological environment evaluation system, and the development demands without reducing the key natural capital are reflected.(2) For the shortcomings of green development efficiency measurement methods, we build a MEBM-Windows model to compensate for the shortage of radial and non-radial efficiency calculation methods, as well as to realize the dynamic comparison of the efficiency of cross-sectional decision-making units (DMUs).This can reflect the actual green development level of the Yellow River basin from a dynamic perspective.(3) For the limitations of research hypotheses in the measurement of green development, we introduce the ecological function differences to improve the existing green development assessment methods from the perspective of spatial function.Incorporating the differentiated ecological function zones of the Yellow River Basin as constraints into the research framework can more comprehensively and reasonably distinguish the restrictive effects of resource endowment differences on the sustainable green development level of the Yellow River Basin.Finally, according to local conditions, the green development strategy of differentiation (strengthening the regional dominant function advantages and completing the short board function) is proposed.

Research area
The Yellow River flows through nine provinces, including Qinghai, Sichuan, Gansu, Ningxia, Inner Mongolia, Shanxi, Shaanxi, Henan, and Shandong.Thirteen major tributaries converge into the main river channel, forming the vast water of the Yellow River.Referring to the Ecological Protection and High-quality Development Plans of the Yellow River Basin prepared by eight provinces of Inner Mongolia, Shaanxi, Gansu, Shandong, Shanxi, Ningxia, Qinghai, and Sichuan, as well as the 'Henan Provincial People's Congress Standing Committee on Promoting Ecological Protection and High-quality Development in the Yellow River Basin', this study selects the 70 cities along the main river and tributaries of the Yellow River Basin as the research objects (table 1).And some areas are excluded from this study due to the lack of data, including Yangling, Jiyuan, Haidong, Gannan Tibetan Autonomous Prefecture, Linxia Hui Autonomous Prefecture, Tibetan Qiang Autonomous Prefecture of Ngawa, Tibetan Autonomous Prefecture of Garzê, Tibetan Autonomous Prefecture of Haibei, Tibetan Autonomous Prefecture of Huangnan, Tibetan Autonomous Prefecture of Hainan, Tibetan Autonomous Prefecture of Golog, Yushu Tibetan Autonomous Prefecture, Haixi Mongolian and Tibetan Autonomous Prefecture.In addition, Laiwu city was revoked and merged into Jinan city in January 2019.Due to the late merger, this study still takes Laiwu city as a separate research sample.Finally, the study sample includes 57 prefecture-level cities, autonomous prefectures, and leagues that the main river and tributaries of the Yellow River Basin flow through (figure 1).

A measure of green development without considering ecological function orientation: MEBM-Windows model
The current calculation methods of total factor productivity (TFP) can be divided into the radial and nonradial methods.The radial method requires the inputs and outputs to vary in the same proportion when evaluating the efficiency, which is contrary to the reality.The non-radial method avoids the assumption that the factor input shrinks and expands in the same proportion because of the inclusion of non-radial relaxation variables, but it ignores the original proportion information of the projected value of efficiency front.To compensate for the shortcomings of the radial and non-radial efficiency calculation methods, Tone and Tsutsui (Tone and Tsutsui 2010) proposed the hybrid model of Epsilon-Based-Measure (EBM).However, Cheng and Zervopulos (Cheng and Zervopoulos 2014) believed that the value range of radial and non-radial parameters in the EBM model was not defined.When the input-oriented CRS (constant returns to scale) is used for the programming solutions, the optimal efficiency value might be larger than 1, and the projection value of the invalid input indicators in the DMU might be higher than their original value.There lead to a wrong conclusion that the improvement target is the 'increase inputs'.Similarly, there exist the same problem in the output-oriented and non-oriented models.Therefore, they proposed an improved EBM model, namely the MEBM model.However, most literatures dynamically compared the crosssectional data of each period and ignored the heterogeneity of the production frontier, leading to the intertemporal results were not comparable.The DEA-Windows analysis method proposed by Charnes et al (1984) can be used to solve this problem.By comparing all DMUs horizontally and vertically, the sample size was enlarged to make the measurement results more accurate.In summary, according to the relevant researches (Halkos and Polemis 2018), this study combines the MEBM model with the Windows model to build the MEBM-Windows model and measures the level of green development of Yellow River Basin without considering the EFO.The MEBM formula is as follows: Where x, y, and bindicate the inputs, desirable outputs, and undesirable outputs, respectively.y* indicates the optimal efficiency value.
s , The parameters e and - w i in the model are calculated by the Pearson correlation coefficient method.Cheng and Zervopulos (Cheng and Zervopoulos 2014) believed that the correlation index calculated by this method is more reasonable and satisfies the principles of identity, symmetry, and unit-invariance.l i is the linear combination coefficient.q and j indicate the radial efficiencies of inputs and outputs, respectively.Since the Model (1) measures the static efficiency values and cannot reflect the trend of change for green development efficiency in various regions over time, it is necessary to construct an MEBM-Windows model based on the Model (1).
The meaning of each variable in the equation is the same as in the Model (1), where l and n represent that the variable is at the time point within the lth window.For the setting of window widthW , this paper refers to the works of Zhang et al (2018b) and Zhou et al (2020) and sets the window width as 3 to ensure the credibility and stability of efficiency measurement.

Green development measurement considering ecological function orientation: dynamic evaluation method reflecting background differences
In this study, the dynamic evaluation method reflecting the background differences is used to adjust the efficiency value of the MEBM-Windows model and reduce the impact of ecological function factors on the evaluation results (Guo et al 2012).The calculation formulas are as follows: Where h i is the comprehensive contribution coefficient of m i 's external factors to the adjustment of its evaluation value, and If there is no special factor for the value of a , ih When y i (that is, GTFP values calculated by the MEBM- Windows model) and k ih change in the positive direction, then l = .
n i is the final evaluation value, where s 1 and s 2 are the preference coefficients of the adjusted value, which satisfy s s If there is no special preference,s s = = 0.5, 1 2 or other methods can be used to obtain them.
In this paper, the area of nature reserves is used as the background factor.Considering the different areas under the jurisdiction of each province, the 'proportion of the area of nature reserves to the area under the jurisdiction' is used as the external factor indicator k ih to characterize the background differences.Nature conservation is divided into the state-level and the local-level.The larger the proportion of nature reserves in the area under the jurisdiction, the larger the key ecological function areas of the jurisdiction, and the greater the impact of ecological function factors on economic development.According to the National Nature Reserve Directory, this study manually sorts out the area data of nature reserves in each city.The missing year's data are replaced by the average value of the adjacent years.

Dagum gini coefficient
The current methods for studying regional differences mainly include: intuitive comparison method (Ma et al 2019), coefficient of variation method (Yang et al 2022), Thiel index method (Wang et al 2021a), and Dagum Gini coefficient method (Dagum 1997).The intuitive comparison method and the coefficient of variation method can only measure the degree of regional differences and cannot explain the sources of differences.The Thiel index method can decompose regional differences into the intra-regional and the inter-regional differences, and can further determine the contribution of these two differences to overall differences.However, this method ignores the contribution of cross-overlapping parts between the different regions to overall differences.The Dagum Gini coefficient method can handle the problem of information overlap between sub sample data while decomposing the regional differences, and it overcomes the limitation of the Theil index that cannot describe the distribution status of sub samples, and achieves complete identification of the contribution of overall regional differences (Gao and Yuan 2022).Therefore, we use the Dagum Gini coefficient to analyze the regional differences.The equations are as follows: where G is the overall Gini coefficient, G w is the intra-regional differences contribution, G nb is the differences contribution of inter-regional net value, and G t is the hypervariable density contribution, which is used to represent the contribution degree of the cross overlap between regions j and h to the overall green development efficiency differences, and them meet the equation t k is the number of groups of cities in the Yellow River Basin (57 cities are divided into three groups: the controlled development zones, the moderate development zones, and the transition development zones).( ) y y ji hr is the green development efficiency value of any city in region ( ) j h , n is the number of cities ( ) = n 57 , ( ) n n j h is the number of cities in group ( ) j h , ȳ is the mean value of the urban green development efficiency level in the Yellow River Basin, ( ) p p j h is the proportion of the number of cities in area ( ) j h to the total number, ( ) s s j h is the proportion of urban green development efficiency in the region ( ) j h to the total green development efficiency, D jh is the relative influence between regions j and h, d jh is the mathematical expectation of the sum of all - > y y 0 ji hr samples in regions j and h, and p jh is the mathematical expectation of the sum of all - < y y 0 ji hr samples in regions j and h.

Index selection and data source
(1) Capital stock: Based on the research of Cheng et al (2021), this paper selects the total investment in fixed assets of the whole society as the investment index of the year and uses the perpetual inventory method to calculate the capital stock of various regions in the Yellow River Basin.Relevant data are from the Statistical Yearbooks of each province.
( (3) Energy input: According to the study of Liu et al (2022), the annual electricity consumption of the whole city is taken as the proxy variable of energy input.Relevant data are from the China Urban Statistical Yearbook.
(4) Technology input: Considering the role of innovative technology in promoting green development, science and technology expenditure is selected as the index of technology input (Guo et al 2020).Relevant data are from the China Urban Statistical Yearbook.
(5) Desirable output: From the perspective of development, the output value created by the region every year is the most intuitive economic output (Cao et al 2022).This paper takes GDP as the index of desirable output and converts the variable price data of each year into the constant price in 2003 according to the GDP index of different regions.In addition, this paper uses the total retail sales of consumer goods to measure social output benefits (He and Hu 2022).Relevant data are from the China Urban Statistical Yearbook.
(6) Undesired output: The existing assessment of ecological environmental pollution mainly focuses on the level of environmental damage, and often neglects the impact of environmental self-absorption capacity on environmental quality, resulting in the imbalanced and distorted ecological assessments.However, the strong sustainability theory requires the continuous improvement for the absorption capacity (selfpurification) of environment when the ecological environment is damaged.This can ensure that the total amount of environmental welfare does not decrease and simultaneously take into account the interaction between environmental damage and environmental absorption (Li et al 2022).With this background, this paper brings the ecological environment damage and the ecological environment construction into the research framework and comprehensively evaluates the environmental status.The dimension of the ecological environment damage includes six indicators: the industrial wastewater discharge, the industrial sulfur dioxide emissions, the chemical fertilizer use, the domestic waste clearing and transporting volume, the average PM2.5 concentrations, and the carbon dioxide emissions.The dimension of the ecological environment construction includes five indicators: the total water resources, the green coverage rate of built-up areas, the harmless treatment rate of domestic garbage, the urban sewage treatment rate, and the comprehensive utilization rate of industrial solid waste.At the same time, the vertical and horizontal scatter degree method which can maximize the difference in measured objects, is used to calculate the ecological environment pollution index of the basin.The data of each indicator are from the China Urban Statistical Yearbook, China Urban Construction Statistical Yearbook, and China Regional Economic Statistical Yearbook.PM2.5 data come from the global PM2.5 concentration mean raster data published by the Socioeconomic Data and Applications Center (SEDAC) at Columbia University.We use the calculation method of Chen et al (2020) to obtain the carbon emissions data of counties, and further add up the affiliated counties to obtain the carbon emissions data of each city.Some missing data are filled using the average annual growth rate method.
The descriptive statistics of indicators are shown in table 2. It can be seen from the table 2 that the variation range of all observed values is very large.The coefficients of variation in the capital stock input, the human input, the energy input, the technology input, the economic benefits, the social benefits, and the ecological environmental pollution index are 1.305, 0.868, 1.36, 0.983, 1.143, 1.185, and 0.698, respectively.Energy input has the largest difference among 57 cities, followed by the capital stock.The coefficient of variation between the human input and the ecological environment pollution index is relatively small, which indicates that the unit human input and the ecological environment quality change little during the study period.(2) In terms of the upper, middle and lower reaches of the Yellow River Basin, the level of green development in the lower reaches is relatively high from 2003 to 2020, with an average annual efficiency of 0.821, which is higher than that of the whole basin.This is inseparable from the strong economic foundation, the reasonable industrial structure, and the leading green innovation technology of the basin (Guo et al 2022).
The level of green development for the upper and middle reaches in the Yellow River is lower than that of the whole basin, which is consistent with the results of Xu et al (2022  (3) For the urban dimension, the level of green development of 24 cities in Zhengzhou, Laiwu, Taiyuan, Xi'an, Tongchuan, Xining, Lanzhou, Shizuishan, Hohhot, Wuhai, Puyang, Sanmengxia, Jining, Tai'an, Dezhou, Heze, Changzhi, Xianyang, Baiyin, Zibo, Dongying, Wuwei, Zhongwei, and Baotou show an upward trend, The first level of green development is within [0.75, 1], whereby the level of green development is characterized as 'good' and it includes Zhengzhou, Kaifeng, Jinan, Xi'an, Hohhot, Ordos, Luoyang, Sanmengxia, Jining, Tai'an, Binzhou, Dezhou, Liaocheng, Heze, Xianyang, Weinan, Dingxi, Zibo, Dongying Yinchuan, Baotou, and Ulanqab, accounting for 38.6% of all prefecture-level cities.These cities can reasonably allocate the various input factors in the process of economic development, improve their utilization efficiency, and then improve the level of urban green development.
The third level of green development is within [0.25, 0.5), and the level of green development is characterized as 'low', and it includes Yulin, Shizuishan, Changzhi, Yan'an, Guyuan, and Zhongwei, accounting for 10.53% of all prefecture-level cities.

Analysis of green development level based on ecological function orientation difference
The ecological endowments of Yellow River Basin are quite different, and the ecological function assigned by the main function planning is also different.The level of green development of Yellow River Basin considering the difference in EFO7 shows a fluctuating downward trend (table 4), which decreases from 0.261 in 2003 to 0.249 in 2020, with an average annual decrease of 0.28%.Compared with the level of green development considering the EFO, the downward trend is significantly narrowed.The level of green development of the upper, middle, and lower reaches of Yellow River Basin without considering the EFO are 0.263, 0.236, and 0.281, respectively.Although the lower reaches still maintain a high level (figure 3), the advantages are not obvious when compares with the upper and middle reaches, which shows that the differences in green development between the upper and middle reaches of the Yellow River Basin are narrowing after considering the constraint effect of EFO on the economic and social development.According to these results, we can also observe that, although the efficiency values of green development are different, the change trend is same, indicating that the external impact of EFO on various regions is balanced and does not distort the overall trend of green development.From the ranking of the two calculation results of 57 cities in the Yellow River Basin, and the comparison between the ranking change in the evaluation results of each city, together with the proportion of natural reserves in the area under their jurisdiction (see table 5)8 , 91.23% of the ranking has changed after considering the difference in EFO.And 21 cities have risen in ranking.Xining, Jincheng, and Wuwei have risen rapidly, from 45th, 39th, and 35th to 15th, 24th, and 3rd, respectively.The proportion of natural reserves of these cities in the area under their jurisdiction is also at the forefront of the Yellow River Basin.After considering the factors of ecological function constraints, the level of green development of Wuwei and other cities has improved significantly, further confirming that the EFO indeed has a great influence on the economic development.The rankings of Xi'an, Weinan, Yinchuan, Wuzhong, and Zhongwei have not changed due to the different reasons.The development quality of Yinchuan is relatively high and the proportion of nature reserve area is relatively high, which leads to maintaining a high level after considering the EFO factor.The area proportion of nature reserve of Xi'an is relatively low, but its development quality is high, so its ranking change is small after considering the ecological factors.The development quality of Weinan, Wuzhong, and Zhongwei is low, and its ecological function is smaller than that of Wuwei, Xining, and other cities, so the ranking changes little after considering ecological factors.The ranking of 31 cities is declined, with Puyang and Tianshui experiencing the largest decline, dropping from 31st and 33rd places to 45th and 46th, respectively.The main reason is that the proportion of nature reserves in these cities is relatively low in the whole basin.When the cities with great background influence rise in the ranking, the ranking of these cities will decline.
According to the calculation results, combined with the economic development level of each city, the proportion of the area of nature reserves to the area under the jurisdiction and the carrying capacity of resources and environment, cities can be divided into three categories: the controlled development zones, the moderate development zones, and the transition development zones.
The first category is the controlled development zones9 , and it includes Zhengzhou, Jinan, Taiyuan, Xi'an, Xining, Hohhot, Hebi, Jining, Binzhou, Dongying, Jincheng, Yan'an, Shangluo, Wuwei, Yinchuan, Bayannur, and Ulanqab.Among them, Taiyuan, Xining, Hohhot, Hebi, Binzhou, Dongying, Jincheng, Wuwei, and Yinchuan have the greatest improvement in the ranking of green development when considering EFO.At the same time, the proportion of nature reserves to the jurisdictional area in these areas is relatively high, indicating that these areas have strong ecological comparative advantages.Therefore, the large-scale development intensity should be controlled, and the ecological protection should be strengthened to provide more ecological products.Zhengzhou, Jinan, and Xi'an are national regional central cities with larger population scales and higher densities.These cities are overloaded and clustered, and their resource carrying capacity has reached its upper limit.Therefore, control development should be carried out.Yan'an, Bayannur, and Ulanqab are important areas for returning farmland to forests and grasslands.The Jining's nature reserve occupies a high proportion of the area.It is an important area for ecological regulation and wetland protection of the Yellow River along the east route of the South-to-North Water Diversion Project.Shangluo is the source of the Danjiang River The second category is the moderate development zones, which includes Kaifeng, Laiwu, Tongchuan, Lanzhou, Ordos, Luoyang, Sanmenxia, Tai'an, Dezhou, Liaocheng, Heze, Baoji, Xianyang, Weinan, Dingxi, Zibo, Jinzhong, Qingyang, and Baotou.The level of green development of these cities without considering the EFO is middle, and the ranking of green development level considering the EFO difference has improved or dropped slightly.This area is located in the transition zones between the controlled development zones and the transition development zones, and it has a certain economic development foundation and environmental comprehensive carrying capacity.But compared to economic development, its ecological comparative advantage is more prominent, so it can be moderately developed.
The third category is the transition development zones, which includes Yulin, Pingliang, Shizuishan, Wuhai, Anyang, Xinxiang, Jiaozuo, Puyang, Datong, Shuozhou, Yuncheng, Changzhi, Baiyin, Tianshui, Xinzhou, Lvliang, Yangquan, Linfen, Wuzhong, Guyuan, and Zhongwei.The calculation results of the two levels of green development in these cities are lower than the average value, and the ranking of green development level considering the EFO shows a downward trend or a small improvement.These cities are important energy and heavy industry centers in China, but their economic development efficiency is low, so they should adopt the transition development.In addition, the comparative advantages of urban development quality and ecological function are insignificant, and the endowments and disadvantages of economic growth and natural resources are equally obvious.Facing the dual pressures of economic transformation and ecological protection, it is necessary to change the mode of economic development while maintaining the economic advantages.4.2.Analysis of regional differences in green development level in the Yellow River Basin According to Formulas (6) -( 14), the Dagum Gini coefficient and its decomposition index of the level of green development of Yellow River Basin based on the differences in EFO are calculated.From the changing trend of figure 4(a), the Dagum Gini coefficient shows a fluctuating upward trend, rising from 0.145 in 2003 to 0.208 in 2020.This phenomenon indicates that the overall differences in the green development of Yellow River Basin have tended to become significant in recent years.From the perspective of the contribution rate (figure 4(b)), the contributions of intra-regional differences, inter-regional differences, and hypervariable density to the overall differences are 25.23%, 58.15%, and 16.62%, respectively, which indicates that inter-regional differences are the main source of the overall gap.This contradicts the conclusion of Li and Fang (Li and Fang 2021) that did not consider the ecological function factors in the regional differences, indicating that the development of the three types of functional zones has their own emphasis, which in turn leads to regional development disharmony.In addition, it also validates the practicality of grouping by the ecological function.Further analysis shows that the overall contribution of intra-regional differences in the Yellow River Basin remain relatively stable, the contribution of inter-regional differences decreases from 59.76% in 2003 to 51.31% in 2020, and the contribution of hypervariable density increases from 13.97% in 2003 to 21.57% in 2020.
It can be seen from figure 5(a) that the intra-regional differences in the controlled development, the moderate development, and the transition development zones show a fluctuating upward trend.The difference of green development in the controlled development zones is the most significant, and the change is largest, which rises from 0.104 in 2003 to 0.212 in 2020 with an average annual growth rate of 4.29%.The main reason is that due to the largest difference in the ecological function constraints within the controlled development zones, the level of green development of cities, such as Taiyuan, Hohhot, Dongying, and Wuwei, have a relatively high proportion of the area under their jurisdiction, thereby showing a rapid increase during the study period.And the average annual growth rates of them are 1.61%, 1.62%, 1.3% and 5.69%, respectively, which are significantly higher than those of the other 13 cities, thus widening the regional gap.The intra-regional differences in the moderate development zones rise from 0.095 in 2003 to 0.136 in 2020, with an average annual growth rate of 2.13%.The transition development zones have the smallest differences, and them rise from 0.146 in 2003 to 0.166 in 2020, with an average annual growth rate of 0.76%.It can be seen from figure 5(b) that the differences in green development between the controlled development zones and the transition development zones are the largest, with an annual Gini coefficient of 0.238 and a growth rate of 1.8%.The reason is that the controlled development zones is mainly composed of provincial capital cities and economically developed cities, and the industrialization level and technological innovation ability of this type of region are higher than those of the transition development zones, leading to a trend of gradually widening the gap in green development between the two types of regions.

Conclusions
This paper selects relevant data from 2003 to 2020, measures the level of green development of the Yellow River Basin using the MEBM-Windows model.Based on the results, the background differences algorithm is used to eliminate the impact of ecological functional differences on the green development measurement.Then the Dagum Gini coefficient is adopted to analyze the regional differences in the Yellow River Basin.The main conclusions are as follows: (1) The level of green development without considering the EFO shows a downward trend, which is consistent with the results of Wang et al (2021b) on the urban ecological efficiency in the Yellow River Basin.In terms of the upper, middle, and lower reaches, the level of green development in the lower reaches is relatively high.From the urban perspective, the level of green development of 22 cities is relatively high, that of 29 cities is moderate, and that of 6 cities is relatively low.
(2) The level of green development considering the differences in EFO also shows a trend of fluctuation and decline, but the downward trend is significantly narrowed when compares with the level of green development without considering the EFO.At the same time, 91.23% of the samples in the ranking of green development level with considering EFO have changed, indicating that considering EFO has a significant impact on the level of green development of the research area.This is consistent with the important role of ecological function in the regional green development emphasized by the 'Opinions on Strengthening Ecological Environment Protection comprehensively and Fighting the Battle of Pollution Prevention and Control', which is issued by the Central Committee of the Communist Party of China and the State Council in 2018.(3) The inter-regional differences are the main source of the overall differences in green development in the Yellow River Basin, which also confirms the accuracy of grouping according to the EFO.

Policy implications
(1) Improving resource input conversion rate and technological innovation.
From the perspective of factor input, the government should first consider the cities with higher efficiency and greater economic development potential when allocating investment to improve the capital utilization rate due to the scarcity of fixed asset investment.In terms of the labor force, cities with sufficient human capital such as Xi'an and Zhengzhou should fully leverage their advantages in the talent aggregation.For regions with serious brain drain such as Lanzhou, the government should increase capital expenditure, introduce scientific research talents through various preferential policies, and increase the accumulation of human capital.For example, governments first should sign a 5-10-year agreement of talent protection to avoid serious talent loss in the underdeveloped areas.Then, they should improve the welfare benefits of special or urgently needed talents and make their income higher than the regional average through the tax reduction or cash subsidies.Next, they should evaluate the attractiveness of talent policies over a two-year evaluation cycle by collecting and analyzing data on the quantity, quality, and field distribution of introduced talents.On this basis, further analysis is conducted on the development status and social effects of regional industries after talent introduction, in order to verify whether the implementation effect of policies has enhanced the development of the regional economic and people's livelihood.For the energy input, cities should reasonably customize and need to avoid redundancy according to their own development.For the technology investment, we can create the 'Lanzhou-Xi'an-Zhengzhou-Jinan' innovation chain in the Yellow River Basin, and then drive the development of surrounding cities through the spillover effect of technological progress in the technologically advanced areas.The scientific and technological investment and technical assistance in areas with the low green development efficiency such as Shizuishan, Yangquan, and Guyuan should be increased.The digestion and absorption of advanced technology in areas with high scientific and technological investment and low green development efficiency such as Yulin and Changzhi should be strengthen.And the transformation ability of regional scientific and technological achievements should be improved.From the perspective of output, it is necessary to strengthen the awareness of ecological environment protection and pollution disposal capabilities while promoting the urban economic growth and the social benefits within the watershed.And it should develop the reasonable pollutant reduction plans based on the corresponding inputs to improve the environmental quality of the watershed and to enhance the efficiency of green development.
(2) Utilizing the comparative advantages and promoting ecological protection and high-quality development in the Yellow River Basin according to the local conditions.Part of the controlled development zones, such as Taiyuan, Xining, and Hohhot, have a relatively high proportion of nature protection areas, which have a large ecological comparative advantage and are suitable for the development of ecological products.For these areas, the large-scale, high-intensity, intensive economic development activities, and urbanization development should be strictly limited.The environmental education and science popularization activities should be vigorously carried out to increase the public awareness of ecological resources.The enterprises or individuals should be actively guided and encouraged to collect and utilize the natural resources in a sustainable manner, which can provide more ecological products.For example, we can promote the efficient development of organic ecological agriculture by developing and promoting the ecotourism attractions.And we can improve and protect the ecological environment through the ecological engineering projects, thereby increasing the proportion of renewable and clean energy in the energy market.However, Jinan, Zhengzhou, and Xi'an have large economic scales, rapidly expanding populations, and increasing pressure on the comprehensive carrying capacity of the cities, especially the restriction of water resources, thus they cannot withstand the intense large-scale urbanization development.Therefore, this kind of area should control the development intensity, optimize the urbanization layout, and improve the development quality and livability of the city based on the existing scale.Although the ecological advantages of moderate development zones are not obvious, they have certain economic development capabilities.On the premise of maintaining ecological advantages, economical construction can be moderately accelerated, but high pollution and high energy consumption industries should be eliminated.The moderate development zones such as Kaifeng, Lanzhou, Ordos, Dezhou, and Zibo, have relatively good economic foundations.They are encouraged to undertake more economic development functions and their accelerated development can provide more economical products.The transition development zones should change the mode of economic development, establish a negative list of industrial access, close the 'five small' enterprises with high pollution emissions in a timely manner, eliminate backward production capacity, implement the 'green+' development model, promote vigorously the adjustment of industrial structure, and change the industrial characteristics of 'black, rough, heavy and hard.' (3) Establishing a development incentive and assessment mechanism suitable for EFO.
First, the Yellow River Basin should be guided to establish an ecological compensation mechanism.The controlled development zones have obvious ecological advantages and they mainly undertake the responsibility of developing ecological products and the ecological security of the Yellow River Basin, but they will affect the local economic development, so a scientific ecological compensation mechanism should be established.For example, the government should formulate the ecological compensation pricing standard, evaluate the constraint effect of ecological function on economic development, and consider the ecological construction cost, the economic environment cost, and the constraint the overall cost.Additionally, the Yellow River Basin can learn from the 'Shandong-Henan Ecological Compensation Model' and the 'Xin'anjiang Model' to establish a whole basin ecological compensation and an ecological product purchase mechanism, whose price is based on the quality and two-way constraints.Second, the government should establish a differentiated assessment system.For a long time, GDP and its growth, fiscal revenue, and other economic indicators are the main assessment methods of the central and provincial governments.Because the three types of regions have undertaken different development responsibilities, it is impossible to implement a unified assessment standard.For example, the controlled development zones should focus on assessing the indicators such as the scale and quality of ecological space, the value of ecological product, and highlight the differences in ecological functions such as water resources conservation, soil and water conservation, and windbreak and sand fixation.The moderate development zones should simultaneously focus on the ecological protection and economic development, and emphasize the assessment of economic costs per unit of GDP.Therefore, the performance evaluation indicators should focus on the resource and environmental carrying capacity, GDP, energy consumption per unit of GDP, and so on.The transition development zones should focus on evaluating the effectiveness of promoting the transformation of economic development mode and power transformation.So the performance evaluation indicators should focus on the change of industrial structure, contribution rate of scientific and technological progress, and energy consumption.

Limitations and future work
This study introduces the EFO into the green development research framework, uses GTFP to characterize the level of green development and to measures the green development in the Yellow River Basin, and analyzes regional differences and their causes.However, this study still some research limitations to be addressed.
First, it is well known that the MEBM-Windows model yields results between 0 and 1.If there are multiple valid DMUs in the results of efficiency (that is, the efficiency value is 1), this method cannot further compare the advantages and disadvantages of effective DMUs.Fortunately, this issue was not encountered during our research process.Besides, the MEBM-Windows model is mainly based on the assumption that the production front is homogeneous.In fact, there are some objective differences in the resource endowments and the economic conditions among different DMUs, which may lead to some deviation in the measurement results.This is beyond the scope of our present research and will be discussed in our future research.The combination of Meta frontier model and Super-MEBM-Windows model in the future perhaps can be used to accurately calculate the efficiency of green development.This combination may not only overcome the problem of unable to distinguish comparisons when the DMUs are fully effective by using the Super-MEBM-Windows model, but also take into account the heterogeneity among different groups by the Meta frontier model.
Second, in the calculation of green development efficiency, we use the number of employees to represent the labor input, which is equivalent to assuming that labor is homogeneous (that is, there is no significant individual difference among all workers).However, due to the differences in the skills, work experience, and educational level of workers, their contributions and efficiency in the production process also vary.Therefore, only considering the quantity of labor and ignoring the quality of labor may lead to inaccurate evaluation of efficiency in calculating the green development efficiency.If the data on the education level of employed people in various cities can be released in the future, the product of the number of employed people and the average years of education can be used to represent the quality and quantity of labor input, which can provide more accurate measurement result of the green development level.
Third, we find that the consideration of EFO has a great impact on the study of green development level in different regions.This study can be also extended to other green development studies with similar geographical environments, such as the Yangtze River Economic Belt.Similar to the Yellow River Basin, the Yangtze River Economic Belt is also an important economic and ecological protection zone based on rivers.It spans the three major regions of east, central, and west China in space, and there are significant differences in the natural resource endowment, the ecological environment quality, and the economic development within the region.Therefore, the green development of the Yangtze River Economic Belt is also an important area for China's ecological civilization construction.Based on the same background, we plan to extend this study to the measurement of green development in the Yangtze River Economic Belt in the future.For example, we can compare the changes and regional differences in the green development level between the Yangtze River Economic Belt and the Yellow River Basin under the consideration of differences in EFO, to accurately measure the green development level of various types of regions and then to propose the differentiated green development plans.

Figure 1 .
Figure 1.The geographic location of our study area.
slack of the ith input, the rth desirable output, and the mth undesirable output, respectively.- w i represents the weight of the ith factor input.
and k ih change in the opposite direction, then l = . ih Analysis of the level of green development of the Yellow River Basin 4.1.1.Measurement of green development level in the Yellow River Basin without considering ecological function orientation Table3and figure2show the level of green development and the evolution track of cities in the Yellow River Basin without considering the EFO 6 .(1)Overall, the level of green development without considering the EFO shows a downward trend during the study period, from 0.748 in 2003 to 0.680 in 2020, with an average annual decline of 0.56%, which is consistent with the latest results of Wang et al (2021b) ,Khan et al (2021), andYang et al (2023b).The reduction in the level of green development mainly comes from two aspects.On the one hand, the unique geographical location of the Yellow River Basin makes it rich in the fossil energy.High pollution and high energy consumption industries account for a large proportion of economy.The development mode is also dominated by the extensive economic development.Investment plays a significant role in driving the green development, especially in the high energy consumption investment.Therefore, there is still room for further improvement and enhancement in the green development(He et al 2021).On the other hand, most cities in the Yellow River Basin are located in the central and western regions.After the industrial transformation and upgrading in the eastern region, some resource and pollution-intensive industries have gradually shifted to the central and western regions.Most of these industries are resource-intensive and labor-intensive industries with high investment, and the economic development efficiency is low.Although inefficient economic development has brought economic growth to the central and western regions, it has also reduced overall development efficiency, thereby affecting the level of green development.Fortunately, the level of green development has improved significantly since 2017 (from 0.518 to 0.566), which indicates that China began to adjust its economic development mode and industrial structure during this period, and the environmental quality of the Yellow River Basin has been effectively improved under the background of eco-friendly development.
and Liu et al (2023).During the study period, the level of green development in the upper reaches of the Yellow River first increase and then decrease, and the values start from 0.689 in 2003 to 0.741 in 2016 and then to 0.672 in 2020.The main reason is that the upper reaches of Yellow River are mostly economically and technologically underdeveloped cities, and the low output of TFP affects the efficiency value, making the level of green development relatively low.The level of green development in the middle reaches of Yellow River show a straight downward trend from 0.756 in 2003 to 0.576 in 2020, with an average annual decrease of 1.59%.The main reason is that there are many resource-based cities in the middle reaches, and the traditional model of resource-dependent economic development affects the green development.From the above analysis, improving the level of green development for cities in the middle and upper reaches is the key to enhancing the green development of the entire Yellow River Basin.

Figure 2 .
Figure 2. Trend of green development level in the upper, middle and lower reaches of the Yellow River Basin without considering the EFO.

Figure 3 .
Figure 3. Trend of green development level in the upper, middle and lower reaches of the Yellow River Basin considering the difference in EFO.

Figure 4 .
Figure 4. Trend differences and contribution rates in the Yellow River Basin.

Figure 5 .
Figure 5. Regional variation trend of level of green development in the Yellow River Basin.
2) Human input: According to the study of Zhao et al (2022) together with Liu and Zhu (Liu and Zhu 2022), this paper uses the number of employees to represent human input.Relevant data are from the China Urban Statistical Yearbook.

Table 2 .
Descriptive Statistics of Indicators from 2003 to 2020.

Table 3 .
The level of green development of cities in the Yellow River Basin without considering the EFO from 2003 to 2020.
Fan et al (2021)ou and Wuhai, whose green development level rise from 0.589 and 0.29 in 2003 to 1 in 2020, respectively.Most of the above cities belong to the upstream regions.With the national attention to the ecological construction in the central and western regions, various policies of returning farmland to forest and grassland are implemented.The ecological environment quality of the upstream area also improves, and the value of GTFP also increases.To further analyze the level of green development of each city, this study refers to the method ofFan et al (2021)and divides the level of green development of the study sample into three levels:

Table 4 .
The level of green development of the Yellow River Basin considering the difference in EFO from 2003 to 2020.

Table 5 .
Ranking of two calculation results in the Yellow River Basin and proportion of natural reserve area to area under the jurisdiction.Danjiangkou Reservoir in the middle route of the South-to-North Water Transfer Project.It undertakes the important mission of 'sending clean water from a river to Beijing'.