Does Economic Policy Intervention Inhibit the Efficiency of China’s Green Energy Economy?

Abstract

Due to the different focus of policies in different regions, China’s energy efficiency has been unstable in recent years. The changing focus of policies at the same time has also impacted the energy system, and therefore, it is very important to explore the impact of China’s new energy policy on its oil and gas energy efficiency. The practical significance of this research is to integrate three policy intervention factors: incentive economic policy intervention, government financial intervention, and mandatory policy intervention. Through the regression of the Stochastic Frontier Approach model, the influence of these policy intervention factors on the efficiency evaluation of decision-making units is eliminated. We calculate the environmental pollution index as an undesired output to measure the efficiency of policy intervention in the green economy of China’s oil and gas energy, use Luenberger model to explore total factor productivity, and find the main reasons that affect the productivity of the green energy economy. The results show that China’s oil and gas energy construction is currently in the stage of scale economy, but the heavy dependence of China’s energy consumption on foreign imports leads to difficulties and urgency in the present stage of technological progress. After excluding the factors of policy intervention, China’s overall energy is in a slightly insufficient policy environment, and energy efficiency is in an unbalanced state.

1. Introduction

China has always been a big energy consumer, and energy efficiency is closely related to the country’s development. However, in recent years with the rapid development of its economy and the turbulence of the international energy situation, the contradiction between energy supply and demand has become increasingly prominent. According to the BP Statistical Review of World Energy, China’s natural gas consumption in 2019 was 307.3 billion cubic meters, of which 129.7 billion cubic meters were imported, up 8.69% year-on-year; its dependence on external resources was 42.21%; oil consumption was 677 million tons, of which imports amounted to 486 million tons, an increase of 8% year-on-year; and the degree of foreign dependence was 71.79%. It is expected that China’s dependence on foreign oil and natural gas energy will continue to rise, pushing it into the stage of energy structure transformation.

Traditional energy is an important support point for China’s current economic development at this stage, but fossil energy is scarce. To a large extent, the country’s long-term energy security focuses on a reliable and sufficient supply of fossil energy, rather than price fluctuations [1]. However, with the overcapacity of coal and the decline in coal consumption, China’s energy structure has changed. Its energy advantage that was originally self-sufficient has deteriorated sharply in the past decade, and it has become increasingly dependent on imported fuels [2]. Xie and Pearlman predicted China’s energy consumption and GDP trends and found that the plan to reduce energy consumption per unit of GDP is difficult to achieve, no matter from the perspective of each province or the country as a whole [3]. China still plans to continue to develop with high energy consumption as the background in the next 10 years or more.

High energy consumption is accompanied by the discharge of a large number of pollutants, resulting in a serious shortage of the carrying capacity of the natural environment. The pollution of industrial waste gas, wastewater, and solid waste in various regions is turning more and more serious, and the contradiction between energy consumption and environmental pollution has become increasingly prominent, gradually developing into a major issue restricting human development. It is estimated that by 2060, the economic cost of global air pollution will account for 1% of global GDP, of which China is one of the regions with the highest GDP loss [4]. In order to improve green energy efficiency, China should focus on pollution reduction while saving energy [5].

The above phenomenon reveals the serious problem of environmental pollution in China’s energy industry. Green development is an inevitable choice for China nowadays and has become a priority for decision-makers. With the rapid development of industrialization and urbanization, China is facing the contradiction between energy supply and demand and the pressure of environmental pollution. Fan et al. show that the focus of sustainable energy development will be reflected in optimizing production capacity structure, improving energy utilization efficiency, and promoting national energy conservation and consumption reduction in the future [6]. China’s vast territory differs greatly from region to region. At present, due to the imbalance of economic development and energy efficiency in various provinces, research by scholars such as Huang and Wang indicates that China’s energy efficiency presents a higher imbalance in the east [7]. Therefore, regional differences must be considered when formulating energy-saving policies, thus posing a difficult problem to the formulation of policies [8]. There is a long-term imbalance between China’s energy growth and environmental protection [9]. The focus of policy intervention in different regions varies a lot. Thus, the focus of this paper is on whether policy intervention is the cause of aggravating the imbalance and affecting the green efficiency of China’s energy.

The main contributions of this article are as follows. First, some current scholars discuss the correlation between policy and energy efficiency [10,11,12], but their studies are mostly based on a single policy, and most of them are discussed from the perspective of imperative policies like environmental taxes. Moreover, they fail to integrate policy and energy consumption into in-depth research. For this reason, the present article builds mandatory policy intervention, government financial intervention, and incentive economic policy intervention in three dimensions from the perspective of policy intervention. The stochastic frontier analysis method eliminates the impact of policy intervention factors on the efficiency of the green energy economy, and such efficiency before and after adjustment is compared in order to make policy. This is done so that the policy and energy efficiency are no longer seen in isolation, to verify what kind of policy environment each region is in. Second, this paper dynamically decomposes the specific causes of the green energy economy from the perspectives of productivity, scale efficiency, technological progress, and pure technological efficiency. The related research is helpful for exploring the future development trend of China’s energy consumption from different efficiency aspects and for explaining the urgency of China’s energy technology upgrading. Third, in terms of model construction, the green economy efficiency model of China’s energy is constructed, and the laws between economic growth, energy input, and pollution emission are empirically analyzed after eliminating the influence of random error and endogeneity. It shows that the province with fast economic growth may come at the expense of environmental pollution, and the negative effect of environmental pollution is much greater than the positive effect of economic growth, which leads to lower green energy efficiency. The conclusions are useful for understanding the economic growth path of energy in developing countries.

The remaining parts of the paper are organized as follows. The second part sorts out the relevant literature on energy efficiency evaluation and policy intervention. The third part introduces the process of establishing the model. The fourth part introduces the source of the dataset creation and the results of the SBM-DEA three-stage model analysis. Based on the analysis of the impact of policy intervention and the Luenberger index, an analysis of energy and green economy efficiencies is carried out, and the key factors of China’s policy intervention are discussed. The fifth part discusses the results of the empirical analysis in depth and puts forward relevant policy recommendations for the next step in China’s energy policy.

2. Literature Review

As the source of power for the development of modern industries, energy laid the foundation for the Industrial Revolution in the 19th century. Under the global oil crisis of the 1870s, oil prices rocketed up, which caused a significant reduction of industrial production in many countries and a marked slowdown in their economy. Economists since then have considered the long-term impact of energy on economic growth [13]. Scholars such as Sasana analyzed the energy of Brazil, Russia, India, China, and South Africa (BRICS), pointing out that the consumption of non-renewable energy, especially coal energy, has a positive and significant impact on economic growth [14]. On the contrary, the consumption of renewable energy has a negative impact on economic growth [14]. In the 21st century, the environmental hazards caused by increasing energy consumption have become increasingly prominent, and environmental problems caused by excessive energy consumption have begun to attract the attention of scholars. As a result, environmental impacts have also been integrated into the energy model, forming a comprehensive evaluation model of the energy-economic-environment system.

Energy, economy, and environment interact and form an inseparable whole. Energy efficiency and related environmental indicators provide important information for solving the problems of energy depletion and economic growth [13]. Because fossil energy resources are non-renewable, in order to avoid the conflicts between energy supply-demand and the deterioration of environmental pollution, it is particularly important to start with improving the efficiency of energy growth. Economic growth is no longer constrained by energy consumption. Therefore, energy efficiency has become an important issue and is regarded as a key measure to resolve the conflict between energy supply and demand as well as environmental pollution [15].

Energy, capital, and labor interact during production. Therefore, only the ratio of energy input to output as an indicator of energy efficiency will have great limitations [16]. Since the DEA method is non-parametric, there is no need to construct a production function and determine the weights of input and output indicators, and the evaluation result is more objective, especially for the evaluation and analysis of complex systems, such as energy efficiency. Therefore, Therefore, DEA has become a good tool for energy efficiency analysis problems [17,18,19].

With the in-depth research on energy efficiency, more and more scholars are incorporating environmental factors and other undesirable output into the empirical research process and including environmental pollution as a “bad” output into the research category, thus forming a complex system of energy-environment-economy growth. Watanabe and Tanaka proved that ignoring an undesired output such as environmental pollution will often lead to overestimation of its efficiency value [20]. On the basis of this, some scholars have developed a non-radial and non-angle SBM method based on the CCR model and the BCC model to deal with undesired output. The SBM model introduces slack variables into the objective function to solve the problem of input factors’ redundancy and efficiency evaluation under non-expected output [21,22]. Song and other scholars used the SBM-DEA model to verify that there are differences in energy efficiency between provinces and regions in China. This is because energy efficiency has significant impacts on industrial structure, energy intensity, and technology, which makes the energy efficiency of various regions fluctuate greatly, and there is regional interdependence [15]. Gokgoz and Erkul employed the SBM model to prove that European countries should not make any changes in the number of employees, total capital, or economic growth, but rather achieve an effective point by reducing fossil fuel consumption [10]. On the whole, scholars generally believe that energy consumption has a significant role in promoting economic growth and affects regional economic growth and coordinated development. Thus, a reliable, sustainable, and competitive energy policy is very important for improving energy efficiency [23,24,25].

Regarding the relevance of policy intervention and energy efficiency, many scholars have analyzed it from different policy perspectives, currently mainly from three different perspectives: legal restriction policies, economic policies, and information policies. (1) The methods used in legal restriction policies mainly include carbon emission restrictions, the establishment of energy efficiency standards, carbon labels, and the formulation of laws and regulations. For example, Fischer conducted in-depth interviews with 202 urban residents in five European countries and found that only strict regulations, significant price changes, and technological innovations can widely change the public’s energy consumption behavior and reduce energy consumption [26]. Cools carried out a study on automobile energy consumption, pointing out that policy-makers need to combine different types of transport policy to effectively change individuals’ travel behavior [27]. Upham et al. conducted analysis on carbon labeling experiments and believe that residents’ carbon emissions can be reversed through the awareness of “carbon labeling” [28].

(2) Research on the incentives and constraints of economic policies on household energy use behavior has always been the mainstream, and there are roughly two opposite views. One view affirms the incentive and restrictive functions of economic policies. Zhao conducted a questionnaire survey on household energy conservation and new energy product incentive policies in 320 households in Florida, U.S.A. That study finds for household heads that financial subsidies and tax relief have greater advantages than zero-interest loans [29]. Yang et al. presented an empirical study on China’s energy-saving and new energy equipment subsidy policies, and the results show that the subsidy policies have a more obvious effect on higher income groups, prompting a change in their energy efficiency investment attitude to investment intentions, but have no significant effect on consumers from low-income families [30]. Alberin studied the impact of tax credit policies on the improvement of energy efficiency of Italian households and found that the policy can encourage them to replace energy-efficient products, while the tax credit policy has no effect on the replacement of heating systems [31]. Another type of view denies the effectiveness of economic policies. For example, Jose et al. analyzed various economic policies of 27 EU countries and found that subsidies and tax incentives are limited by government budgets and do not have the expected effect [32]. Similarly, Zhao et al. also showed that neither tax credits nor interest-free loans have a significant effect [29]. As a whole, scholars generally believe that energy consumption has a significant role in promoting economic growth and that energy consumption will affect regional economic growth and coordinated development [23,24,25].

(3) The impact of information policy on residents’ energy-saving behavior has been a hot topic for scholars in recent years. At present, the information policies implemented by various countries are rich in content. Compared with traditional media preaching, scholars generally believe that targeted and specific information feedback and information guidance allow residents to have a clearer grasp of their energy consumption status and to understand the feasible methods to improve their behavior, stimulate their awareness of energy conservation, and thereby reduce energy consumption [11,12,33].

Asensio indicates that environmental and health information can effectively promote residents’ energy savings by 8%. Families with children can save up to 19% of their energy [34]. Lane believes that students’ environmental knowledge level and the environmental and cultural atmosphere of the school they are in have a significant impact on environmental policy and education [35]. Therefore, necessary knowledge dissemination and cultural training must be carried out for students to ensure that policy measures can be effectively implemented in this group.

The current literature mostly considers energy efficiency and economic policies independently, or only analyzes one of the policies, and does not specifically analyze which policy factors have an impact on energy efficiency. The implementation of the policy is a system that contains the common effects of multiple policies. As most policy intervention factors in each region are different, the efficiency cannot be truthfully reflected, which leads to a lack of support for policy selection and optimization. Therefore, Fried et al. proposed a three-stage DEA method based on the traditional DEA model, which can eliminate the influence of external factors and random noise on the efficiency of the decision-making unit under the policy inclination, making the efficiency evaluation more objective [36]. Fang used this method to exclude the influence of external factors such as environmental regulations and enterprise scale and evaluated the green innovation efficiency of China’s heavy pollution industries [37].

Energy efficiency evaluation is a multi-agent, multi-path interactive process. For this reason, this article integrates energy consumption, economic growth, and environmental pollution, and builds a model based on non-angle and non-radial measurement of green energy efficiency. The SFA method is adopted to exclude economic policy interference factors, to measure the efficiency of the green energy economy by provinces, and to analyze the impact of economic policy interference factors on energy efficiency. In terms of economic policy intervention indicators, the intensity of government R&D subsidies, the degree of government financial intervention, and environmental taxes are selected as economic policy intervention variables for more intuitive comparison. The research framework and measurement methods related to this article can also be applied to other developing countries and the research process of oil and gas energy policies, providing a basis for judging the impact of policies on energy efficiency.

3. Model Building

There are many methods to evaluate the efficiency of energy growth. The traditional DEA, AHP, and SFA models are widely used. Because these traditional models cannot eliminate the influence of random errors and endogeneity, it is difficult for them to truly reflect the objective status of efficiency, and so some scholars use an improved DEA method. The energy economy development process has its particularity and is greatly affected by policy intervention factors. Therefore, this paper uses the SBM-DEA model and the SFA model by Fried and other scholars to construct three stages to evaluate the efficiency of energy green economy. The optimization path proposed according to the evaluation results will be more scientific and reasonable.

3.1. Building a Three-Stage SBM-DEA

In this part, we will construct the SBM-DEA model and the SFA model for empirical analysis. The model is divided into three parts: In the first stage of analysis, the SBM-DEA model of undesired output is constructed to analyze the original input-output data. The second stage of empirical analysis eliminated the influence of external environmental factors and random noise on energy efficiency through the SFA model. Finally, the SBM-DEA model is used for evaluation again, and the evaluation results can more truly reflect the true level of energy efficiency.

3.1.1. First Stage: SBM-DEA Model

The traditional radial DEA model requires that the input and output are adjusted in the same proportion, which violates the original intention of this paper to reduce undesired output as much as possible. Therefore, this paper uses the non-radial SBM model to relax this assumption, and the non-zero relaxation variables of input and output are adjusted to handle different proportional changes. As shown in Figure 1, LL’ is the efficient frontier production line. In the radial model, since excessive and ineffective inputs can be improved at the same time, after the inefficient decision-making units E and F reduce all inputs in the same proportion, they are respectively mapped to the two points C and B on the production efficient frontier, but the point C is still not the optimal efficiency point. Under the condition that all other inputs required are equal, the energy input at point C under the radial model is significantly higher than that at point B. Therefore, in the non-radial model, due to the improper allocation of input factors, point E can eventually move to point B, which improves energy efficiency. There is a non-zero slack variable between points C and B. Therefore, the traditional DEA model ignores the existence of non-zero slack variables, which ultimately leads to a bias in the result estimation.

We use the SBM model to estimate green energy efficiency. Suppose there are k decision-making units ($DM{U}_i$($i=1,2\cdots \cdots k$), and these DMUs represent the provinces of China. Each DMU has the same input-output index—that is, there are $m$ input elements and $n$ outputs ($n_1$ is a desired output, and $n_2$ is a non-desired output). The i^th $DM{U}_i$ in index value $X_i$ is a desired value of output indicators $y^g$ and undesirable output index value y^b, which are respectively:

$$\left\{\begin{array}{l}{x}_i=(x_{1i},x_{2i},\cdots ,x_{mi})\in R^{m\times k}\\ y_i^g=(y_{1i}^g,y_{2i}^g,\cdots ,y_{n_{1}i}^g)\in R^{n_1\times k}\\ y_i^b=(y_{1i}^b,y_{2i}^b,\cdots ,y_{n_{2}i}^b)\in R^{n_2\times k}\end{array}\right.$$

(1)

From the construction idea of the SBM model containing undesired output, the DMU’s production technology set T is determined as: $T=\{x,y^g,y^b\left|x_i\ge X\lambda ,y_i^g\ge Y^g\lambda ,y_i^b\ge Y^b\lambda ,\right.\lambda >0,{x \mathrm{produce} y}^g{\&y}^b\}$, where $\lambda$ is the weight of the non-negative multiplier of the linear programming constructed by the production technology concentration.

The traditional DEA model uses radial measurement and ignores the slack variables. When there are slack variables, the phase efficiency may be overestimated. Based on this, Tone proposes a non-radial and non-angle SBM model, which directly considers the slack variables of input and output in the decision-making unit [22]. It can avoid the shortcomings of the traditional DEA model in terms of input and output and the same proportion of improvement. When dealing with undesired output problems such as environmental pollution, the SBM model not only can calculate the efficiency of the decision-making unit in this situation, but also point out the direction of improvement of the non-effective input. Based on Tone’s method, when considering bad output variables, the non-directional SBM model can be specified as follows:

$$\rho =\min \frac{1-\frac{1}{M}{\displaystyle \sum _{m=1}^M\frac{s_m^{-}}{x_{m0}}}}{1+\frac{1}{N_1+N_2}({\displaystyle \sum _{n_1=1}^{N_1}\frac{s_{n_1}^g}{y_{n_{1}0}^g}+{\displaystyle \sum _{n_2=1}^{N_2}\frac{s_{n_2}^b}{y_{n_{2}0}^b})}}}$$

(2)

$$\mathrm{s}.\mathrm{t}.\left\{\begin{array}{l}{\displaystyle \sum _{i=1}^k{x}_i{\lambda }_i}+s^{-}=x_0\\ {\displaystyle \sum _{i=1}^k{y}_i^g}{\lambda }_i-s^g=y_0^g\\ {\displaystyle \sum _{i=1}^k{y}_i^g}{\lambda }_i+s^b=y_0^b\\ {\lambda }_i\ge 0,s^{-}\ge 0,s^g\ge 0,s^b\ge 0\end{array}\right.$$

The variables $s^{-},s^g,s^b$ respectively represent the slack of input factors, expected output, and undesired output; and ρ is the objective function—that is, the efficiency value we seek, which is strictly decreasing about $s^{-}$, $s^g$, and $s^b$. Among them, 0 ≤ ρ ≤ 1, and when ρ = 1, the decision-making unit is efficient, which means that the slack variable is 0 (s⁻ = 0, s^g = 0, s^b = 0), and it is at the front of the production line; when ρ ≤ 1, the decision-making unit has room for efficiency improvement. The efficiency measurement of model (2) includes economic and environmental pollution factors, and so we can define it as the green economic efficiency value of oil and gas energy with undesired output.

3.1.2. Second Stage: Adjust Input and Output Based on SFA Adjustment

Fried points out that the decision-making unit is affected by management inefficiency, external interference factors, and statistical noise, resulting in a difference between the original input and the target input. Thus, it is necessary to separate these three effects [36].

Therefore, this paper uses the SFA model to eliminate policy intervention factors and random error terms. Just like for the first stage, there is a total of k DMUs, and each decision-making unit has m kinds of inputs. Here, $S_{ij}$ represents the slack change of the j^th input of the i^th DMU we get in the first stage. We then use $Z_i$to represent the value of the policy disturbance variable at the i^th DMU unit, and ${\beta }_j$ represents the coefficient of the policy interference variable. The following SFA regression function can then be constructed:

$$S_{ij}=f(Z_i;{\beta }_j)+v_{ij}+{\mu }_{ij};i=1,2\cdots ,k;j=1,2\cdots ,m$$

(3)

Here, the general expression $f(Z_i;{\beta }_j)$ is the influence of the policy intervention variable in this paper on the input slack value of element i, and ${\nu }_{ij}+{\mu }_{ij}$ is the mixed error term ${\nu }_{ij}~N(0,{\sigma }_{vj}^2)$, which represents the influence of random factors other than the linear relationship between input and output on the input slack variable. Moreover, ${\mu }_{ij}~N^{+}(0,{\sigma }_{\mu j}^2)$ represents management inefficiency—that is, the influence of management factors on input slack variables, $v_{ij}$, and ${\mu }_{ij}$ is independent of each other. We define ($\gamma =\frac{{\sigma }_{uj}^2}{{\sigma }_{uj}^2+{\sigma }_{vj}^2}$) as the ratio of the variance of management inefficiency to the total variance. When γ is close to 1, it means that the main reason for the impact is management inefficiency; otherwise, it means that the main random error is the main influencing factor.

The purpose of SFA regression is to eliminate the influence of policy interference factors and random factors on the efficiency measurement of DMUs, in which a DMU is only affected by management inefficient factors. Therefore, all DMUs can be adjusted in the same external environment.

The adjustment formula is as follows:

$$X_{ij}^{\prime }=X_{ij}+\max [f(Z_i;{\stackrel{\wedge }{\beta }}_j)]-f(Z_i;{\stackrel{\wedge }{\beta }}_j)+\max (v_{ij})-v_{ij};i=1,2\cdots ,k;j=1,2\cdots ,m$$

(4)

Here, $X_{ij}^{\prime }$ is the input after adjustment; $X_{ij}$ is the input before adjustment, $\max [f(Z_i;{\widehat{\beta }}_j)]$. means that it is in the worst economic policy intervention situation; and other DMUs make adjustments based on this; $\max [f(Z_i;{\stackrel{\wedge }{\beta }}_j)]-f(Z_{\mathrm{i}};{\stackrel{\wedge }{\beta }}_j)$ is the adjustment of economic policy intervention factors; and $\max (v_{ij})-v_{ij}$ means that the interference of random items is eliminated. In this way, all DMUs are placed in the same policy environment.

3.1.3. Third Stage: DEA Efficiency Analysis of the Adjusted Input-Output Variables

We use the adjusted input-output variables to measure the efficiency of each DMU again. At this time, the efficiency has eliminated the influence of policy intervention factors and random factors, and the calculated efficiency of a DMU is relatively true and accurate.

3.2. SBM-Luenberger Productivity Index

In the Malmquist-Luenberger production index measurement method, the measurement angle needs to be selected under the condition of minimizing cost or maximizing profit. Chambers proposes the Luenberger productivity index [38], which can be used under the condition of maximizing profit, and at the same time, it can reduce the measurement input and increase the output without choosing the measurement angle. Because the Luenberger productivity index is more general than the Malmquist-Luenberger productivity index, this article selects the Luenberger productivity index to research and optimize the Malmquist-Luenberger productivity index, so that factor analysis can be more realistic.

Fukuyama [39] integrates the SBM measurement method and the distance function of the directionality, while considering the expected output and the undesired output. Like the traditional DEA directional distance function, the larger the SBM directional distance value is, the lower is the efficiency level, which is an indicator of the inefficiency level. Based on our definition, the objective function SBM directional distance is the maximization of the average sum of input inefficiencies and output inefficiencies:

$$\begin{gathered} S_v^t(x^{t{i}^{\prime }},y^{t{i}^{\prime }},b^{t{i}^{\prime }},g^{-},g^y,g^b)=\underset{s^{-},s^y,s^b}{\max }\frac{\frac{1}{M}{\displaystyle \sum _{m=1}^M\frac{s_m^{-}}{g_m^{-}}+\frac{1}{N_1+N_2}({\displaystyle \sum _{n_1=1}^{N_1}\frac{s_{n_1}^y}{g_{n_1}^y}+{\displaystyle \sum _{n_2=1}^{N_2}\frac{s_{n_2}^b}{g_{n_2}^b})}}}}{2} \\ s.t.\left\{\begin{array}{l}{\displaystyle \sum _{i=1}^k{x}_{im}^t{\lambda }_i}+s_m^{-}=x_{{}^{{}_{mi\prime }}}^t\\ {\displaystyle \sum _{i=1}^k{y}_{i{n}_1}^t}{\lambda }_i^t-s_{{}^{n_1}}^y=y_{n_{1}i\prime }^t\\ {\displaystyle \sum _{i=1}^k{b}_{i{n}_2}^t}{\lambda }_i^t+s_{n_2}^b=b_{n_{2}i\prime }^t\\ {\lambda }_i\ge 0,s^{-}\ge 0,s^y\ge 0,s^b\ge 0\end{array}\right. \end{gathered}$$

(5)

Just like in Equation (2), m, n₁, and n₂ represent input, expected output, and undesired output, respectively, and $S_v^t$ represents the directional distance function under VRS. If the weight variable and the constraint of 1 are removed, then $S_v^t$ is expressed as a directional distance function under CRS. Here,$(x^{t{i}^{\prime }},y^{t{i}^{\prime }},b^{t{i}^{\prime }})$ and $(s^{-},s^y,s^b)$ respectively represent the vector and slack vector of the DMU input, expected output, and undesired output; and $(g^{-},g^y,g^b)$ expresses the direction vector of compressing input, expanding expected output, and compressing undesired output. Among them, $(s^{-},s^y,s^b)$ means that the input and undesired output are excessive and the output is insufficient. When $(s^{-},s^y,s^b)$ are all greater than 0, it means that the actual input and undesired output are greater than the boundary input and undesired output, but the actual output is less than the boundary output.

Based on the DEA directional function, the Luenberger total factor productivity index reflects the dynamic changes of productivity in different periods. LTFP is expressed as:

$$\begin{array}{l}LTF{P}_t^{t+1}=\frac{1}{2}\left[S_c^t\left(x^t,y^t,b^t;g\right)-S_c^t\left(x^{t+1},y^{t+1},b^{t+1};g\right)\right]+\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\frac{1}{2}\left[S_c^{t+1}\left(x^t,y^t,b^t;g\right)-S_c^{t+1}\left(x^{t+1},y^{t+1},b^{t+1};g\right)\right]\end{array}$$

(6)

In order to further study which factors have a dominant impact on productivity, according to Grosskopf (2003), Luenberger’s total factor productivity (LTFP) continues to be divided into efficiency changes (LPEC), technology changes (LPTP), scale efficiency changes (LSEC), and technology scale changes (LTPSC). Its form runs as follows:

$$LPE{C}_t^{t+1}=S_v^t(x^t,y^t,b^t;g)-S_v^{t+1}(x^{t+1},y^{t+1},b^{t+1};g)$$

(7)

$$\begin{array}{l}LPT{P}_t^{t+1}=\frac{1}{2}\left[S_v^{t+1}(x^t,y^t,b^t;g)-S_v^t(x^t,y^t,b^t;g)\right]+\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\frac{1}{2}\left[S_v^{t+1}(x^{t+1},y^{t+1},b^{t+1};g)-S_v^t(x^{t+1},y^{t+1},b^{t+1};g)\right]\end{array}$$

(8)

$$\begin{array}{l}LSE{C}_t^{t+1}=\left[S_c^{t+1}(x^t,y^t,b^t;g)-S_v^t(x^t,y^t,b^t;g)\right]-\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\left[S_c^{t+1}(x^{t+1},y^{t+1},b^{t+1};g)-S_v^{t+1}(x^{t+1},y^{t+1},b^{t+1};g)\right]\end{array}$$

(9)

$$\begin{array}{l}LTPS{C}_t^{t+1}=\frac{1}{2}\{[S_c^{t+1}(x^t,y^t,b^t;g)-S_v^{t+1}(x^t,y^t,b^t;g)]-\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}[S_c^t(x^t,y^t,b^t;g)-S_v^t(x^t,y^t,b^t;g)]\}+\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\frac{1}{2}\{[S_c^{t+1}(x^{t+1},y^{t+1},b^{t+1};g)-S_v^{t+1}(x^{t+1},y^{t+1},b^{t+1};g)]-\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}[S_c^t(x^{t+1},y^{t+1},b^{t+1};g)-S_v^t(x^{t+1},y^{t+1},b^{t+1};g)]\}\end{array}$$

(10)

$$LTFP=LPEC+LPTP+LSEC+LTPSC$$

(11)

The calculation of the Luenberger productivity index needs to be estimated under CRS and VRS assumptions, four of which belong to CRS, and the other four are estimated under the VRS assumption. Thus, we obtain eight SBM directional distance functions. Here, LPEC, LPTP, LSEC, and LTPSC are greater than 0 (less than 0), respectively, indicating that pure technical productivity increase (decrease), technological progression (regression), scale efficiency increase (decrease), and technology deviate from the optimal scale state (moves to the optimal scale state CRS) of the DMU’s CRS from period t to t + 1.

4. Data Collection and Empirical Analysis

With the commissioning of the China-Myanmar natural gas pipeline in 2013, China’s four major energy strategic channels were officially completed. Due to data availability, this paper selects 2013 as the starting year to evaluate the data of 30 provincial administrative units across the country from 2013 to 2017. The energy data of Tibet is difficult to obtain, and so it will not be considered for the time being. The data come from China Statistical Yearbook, China Energy Statistical Yearbook, and China Environment Yearbook. The western region includes 11 provincial-level administrative units including Sichuan, Chongqing, and Guizhou; the eastern region includes 11 provincial-level administrative units including Beijing, Tianjin, and Zhejiang; and the central region includes 8 provincial-level administrative units including Shanxi, Inner Mongolia, and Jilin.

4.1. Variable Selection

Input indicators: According to the theory of economic growth, economic output is related to capital input, labor, and technological progress. Therefore, based on the research content of this article, we select capital, labor, and energy consumption as input variables. Hartwick proposes that the optimal allocation of exhausted resources and sustainable development should be measured from the total capital stock [40], and so this paper selects the inventory perpetual method to calculate the capital stock in each year as the capital investment in the study. We select the number of employees in each year as the labor input. Traditional energy is an important part of China’s energy strategy, and so we select natural gas and oil consumption as energy inputs.

Expected output: Existing research on the relationship between energy, environment, and economic growth mainly uses GDP as output [41,42]. According to the environmental Kuznets curve (EKC), income level and environmental pollution present an inverted U-shape relationship [43]. Scholars such as Huang have proved that the relationship between energy efficiency and economic growth varies under different income levels [36]. This shows that there is also an inseparable relationship between environmental pollution, energy consumption, and income. China’s current situation fits in with the EKC hypothesis [44]. For this reason, we select economic growth and residents’ income as expected output variables and at the same time use the difference between the GDP of each province during periods t and t + 1 as the output variable to measure economic growth. Both of these variables reflect macroeconomic changes. At the same time, to ensure that GDP is not affected by inflation and prices, we treat GPD as follows: select 1978 as the base period, the nominal GDP of the base period = real GDP, the GDP index of the year = 100, use the actual GDP of the base period * [1 + (The GDP index of the second period − 100))/100], get the actual GDP of the second period, repeat the above operation to get the actual GDP of all periods, and select the actual GDP of 2013–2017.

Undesired output: This article refers to the practice of Fang [37] and uses the entropy method to integrate wastewater, exhaust gas, and industrial solid waste into an environmental pollution index. The specific form goes as follows.

Set k DMUs and m indices, where $x_{ij}(i=1,2,\cdots ,k;j=1,2,\cdots m)$ is the value of the j^th index in the i^th DMU. Since the three waste variables in this study are all measured in units of 10,000 tons, they are not standardized. The proportion of the j^th index in the i^th decision unit is calculated as:

$$p_{ij}=\frac{x_{ij}}{{\displaystyle \sum _{i=1}^k{x}_{ij}}}(i=1,2,\cdots k,j=1,2,\cdots m)$$

(12)

We calculate the entropy value of the j^th index according to the definition of information entropy:

$$e_j=-\frac{{\displaystyle \sum _{i=1}^k{p}_{ij}\ln (p_{ij})}}{\ln (n)}\hspace{1em}(j=1,2,\cdots m)$$

(13)

The weight is measured according to the entropy value:

$$w_j=\frac{1-e_j}{{\displaystyle \sum _{j=1}^m(1-e_j)}}\hspace{1em}(j=1,2,\cdots ,m)$$

(14)

By judging the degree of dispersion in the three waste variables, the weight of the comprehensive evaluation is determined. The greater the weight is, the greater is the degree of environmental pollution. The weights appear in

Table 1. Industrial solid waste, industrial waste gas, and wastewater weights in the environmental pollution index.

Variable	2013	2014	2015	2016	2017
Industrial solid waste	0.420	0.425	0.428	0.401	0.427
Industry exhaust	0.228	0.232	0.229	0.256	0.234
Wastewater	0.352	0.343	0.343	0.343	0.339

. It can be seen that the weight of industrial solid waste is much greater than that of industrial waste gas and wastewater.

Policy intervention variables: The external interference variables selected in the SFA model should exist objectively, but they have a significant impact on the efficiency of the DMU and interfere with the evaluation of its efficiency to a certain extent. This paper mainly examines the impact of existing policies on the efficiency of China’s green energy economy, and so these policy should be in the macro-government environment, such as energy regulations, pollution, innovative technology, and so on. Combined with the research purpose of this article, we select government R&D subsidy intensity, degree of government financial intervention, and environmental tax as economic policy intervention variables to eliminate. These economic policy intervention factors are closely related to energy policy and have an impact on energy efficiency.

(1) Incentive economic policy intervention: We select the intensity of government research and development (R&D) subsidies to measure incentive economic policy intervention. Technological innovation is an important source of an enterprise’s competitive advantage and an important guarantee for an enterprise to achieve sustainable development. The level of technological innovation directly affects the energy and economic efficiencies of various regions. When the level of technological innovation of enterprises is low and when facing environmental regulations leading to higher pollutant emission standards, it is easy for them to choose terminal treatment methods to achieve pollutant discharge standards, and terminal treatment funds are often obtained by squeezing R&D funds. This becomes an obstacle for enterprises to promote their own development through technological innovation. Therefore, R&D subsidies for companies’ green technologies will directly affect their energy and economic efficiencies. This paper uses the proportion of government funds in R&D investment for various regions to measure the intensity of government R&D subsidies.

(2) Government financial intervention: It reflects the government’s “visible hand” ability to regulate and control the economy during economic operations. Government policy tends to have an impact on economic growth to a certain extent, which may ease market friction, promote the upgrading of the industrial structure, and increase the level of consumption of residents, thereby affecting regional pollution emissions. On the other hand, local government intervention often takes promoting regional economic output as the primary task, and so it is difficult to take into account the dual goals of economic development and environmental protection, which may have counterproductive effects on environmental improvement. This will also cause greater environmental pollution. In order to more accurately reflect the effect of financial intervention, this study chooses the value of fiscal expenditure divided by nominal GDP after removing the expenditures on science, education, culture, and health to indicate the degree of financial intervention by local governments at all levels.

(3) Mandatory policy intervention: It refers to the government’s policies to influence energy consumption behavior through various administrative orders and mandatory laws and regulations, to control the waste of energy, and to reduce energy demand. This paper uses the intensity of environmental taxes in various regions to express it. Since China’s environmental protection tax was only implemented on January 1, 2018, it is measured by the proportion of each region’s pollution discharge fee income and industrial added value (

Table 2. Data descriptive index.

Index		Unit	Minimum	Max	Mean	Standard Deviation
Input variable	Capital stock	Billion	682.17	152,042.40	50,543.82	32,713.37
	employed population	Million	314.21	6766.86	2747.75	1770.14
	Natural gas supply	One hundred million cubic meters	4.27	237.69	65.66	48.42
	Petroleum energy consumption	Ten thousand tons	113.44	6211.76	1827.52	1281.34
Expected output	Economic growth	Billion	−6422.12	8850.32	1799.86	1862.37
	Residence income	yuan	10,954.40	58,988.00	22,323.07	9373.10
Undesired output	Environmental pollution	Ten thousand tons	12,474.61	324,346.40	86,105.70	64,570.45
Policy intervention variables	Incentive policy intervention	%	0.76	26.32	5.39	5.35
	Government financial intervention	%	12.08	62.69	25.00	10.16
	Mandatory policy intervention	%	19.01	57.30	43.88	8.13

).

4.2. Empirical Analysis

After processing the data, we use the SBM-DEA model and the SFA model for empirical analysis. The empirical analysis is divided into three parts: In the first stage of analysis, the initial efficiency analysis of the original input-output data is carried out using the SBM-DEA model that considers the undesired output. The second stage of empirical analysis eliminates the impact of policy factors on energy efficiency through the SFA model. According to the adjusted input value, the SBM-DEA model is again used to evaluate the green energy efficiency of various provinces in China.

4.2.1. First-Stage SBM-DEA Analysis of Undesired Output

In the first phase, Matlab software was used to calculate the initial green energy efficiency values of 30 provincial administrative units from 2013 to 2017. As is shown in

Table 3. First-stage green energy efficiency values of 30 provinces, autonomous regions, and municipalities (except Tibet) across the country.

Provincial Administrative Unit	2013	2014	2015	2016	2017	Mean
Inner Mongolia	0.851 (16)	0.926 (13)	0.182 (27)	0.801 (24)	1 (1)	0.742 (21)
Liaoning	0.675 (26)	0.837 (19)	0.072 (27)	1 (1)	0.827 (19)	0.682 (25)
Jilin	0.877 (12)	0.917 (14)	0.671 (21)	1 (1)	0.307 (29)	0.754 (20)
Heilongjiang	0.857 (15)	0.857 (18)	0.098 (28)	0.808 (23)	0.825 (20)	0.689 (24)
Shanghai	0.877 (12)	1 (1)	1 (1)	1 (1)	1 (1)	0.975 (9)
Zhejiang	0.820 (20)	0.878 (16)	0.843 (17)	0.96 (19)	0.837 (18)	0.868 (15)
FuJian	0.832 (18)	0.801 (22)	0.899 (13)	1 (1)	1 (1)	0.906 (12)
Guangdong	0.458 (29)	0.539 (26)	0.548 (22)	1 (1)	1 (1)	0.709 (22)
Guangxi	1 (1)	1 (1)	1 (1)	1 (1)	0.286 (30)	0.857 (16)
Hainan	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
Chongqing	1 (1)	1 (1)	1 (1)	1 (1)	0.788 (21)	0.958 (10)
Yunnan	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
Shaanxi	0.753 (25)	0.764 (24)	0.807 (19)	0.759 (26)	1 (1)	0.817 (18)
GanSu	0.971 (11)	0.955 (12)	1 (1)	0.870 (21)	0.844 (17)	0.928 (11)
Qinghai	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
Ningxia	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
Xinjiang	0.785 (24)	0.776 (23)	0.215 (26)	0.829 (22)	0.773 (22)	0.676 (26)
Beijing	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
Tianjin	1 (1)	1 (1)	0.999 (12)	1 (1)	0.924 (14)	0.985 (8)
Hebei	0.816 (21)	0.812 (21)	0.781 (20)	0.576 (28)	0.465 (26)	0.690 (23)
Shanxi	0.827 (19)	0.213 (30)	0.014 (30)	0.795 (25)	1 (1)	0.570 (27)
JiangSu	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
AnHui	0.834 (17)	0.873 (17)	0.835 (18)	0.923 (20)	0.750 (23)	0.843 (17)
Jiangxi	0.874 (14)	0.889 (15)	0.865 (14)	1 (1)	0.896 (15)	0.905 (13)
Shandong	0.537 (27)	0.534 (27)	0.531 (23)	0.578 (27)	0.445 (27)	0.525 (28)
Henan	0.429 (30)	0.449 (29)	0.464 (24)	0.515 (29)	0.431 (28)	0.458 (30)
Hubei	0.795 (23)	0.722 (25)	0.847 (16)	1 (1)	0.720 (24)	0.817 (19)
Hunan	0.803 (22)	0.818 (20)	0.857 (15)	1 (1)	0.874 (16)	0.870 (14)
Sichuan	0.465 (28)	0.488 (28)	0.442 (25)	0.454 (30)	0.637 (25)	0.497 (29)
Guizhou	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
Population mean	0.838	0.835	0.732	0.896	0.821	0.824
Central region	0.787	0.717	0.581	0.880	0.725	0.738
Eastern region	0.820	0.855	0.788	0.919	0.863	0.851
Western region	0.893	0.901	0.786	0.883	0.848	0.866
Number of provinces on the production frontier	10	11	11	18	13	7
Percentage of total provinces	33.34%	36.67%	36.67%	60%	43.33%	23.33%

, it can be seen that the average green energy efficiency of the central region from 2013 to 2017 is 0.738, which is far lower than the national average of 0.826. The average efficiencies of the central and western regions are 0.851 and 0.866, respectively. The development of energy and green economic efficiencies in various regions is uneven. The province with the lowest efficiency value is Henan at only 0.458. However, since the influence of policy intervention factors is not excluded in the first stage, it does not reflect the effective value of DMU efficiency, and so this paper continues to use the second-stage SFA for analysis.

4.2.2. Second-Stage SFA Model Analysis

The second-stage empirical analysis uses the SFA model to eliminate the interference of exogenous policy interference variables and random error terms. With Frontier 4.1 software, this paper uses the input slack variable generated in the first stage as the dependent variable and the incentive economic policy intervention, government financial intervention, and mandatory policy intervention as independent variables to perform SFA regression to test whether the policy intervention factors are slack on the input factors and to then make adjustments. The results are in

Table 4. SFA estimation for the second stage.

Slack Variable	Capital Stock Slack Variable		Slack Variable		Natural Gas Consumption Slack Variable		Oil Consumption Slack Variable
	Parameter Value	t Value	Parameter Value	t Value	Parameter Value	t Value	Parameter Value	t Value
Constant term	13,842.55	12,731.55 ***	486.35	1.25	−8.25	−1.67 *	462.11	1.44
Incentive economic policy intervention	−56,199.01	−54,953.95 ***	−1733.31	−2.47 **	−7.24	−0.21	−692.38	−15.25 ***
Government financial intervention	−39,410.89	−36,005.67 ***	−317.92	−2.59 ***	−5.21	−0.36	−564.55	−0.86
Mandatory policy intervention	−39.98	−1.45	−9.27	−1.78 *	0.15	1.71 *	−9.07	−1.70 *
σ 2	176,112,850	176,112,850 ***	445,982.97	331,756.41 ***	368.76	4.29 ***	577,061.8	134,855.15 ***
γ	0.61	12.59 ***	0.74	22.09 ***	0.70	9.63 ***	0.71	19.01 ***
LR unilateral error test	30.45 ***		57.84 ***		63.96 ***		56.69 ***

Note: ***, **, and * represent significance at the 1%, 5%, and 10% levels, respectively.

. The one-sided likelihood ratio LR test results are respectively 30.45, 57.84, 63.96, and 56.69, which are all placed outside the 99% confidence interval, pass the significance test, and reject the assumption that there is no management inefficiency item. In terms of γ, when γ is close to 1, it indicates that the actual efficiency value is mainly affected by management inefficiency. When γ is close to 0, it indicates that random errors have a greater impact on the actual efficiency. In this paper, the γ values are all above 0.6, meaning the main reason for the errors is due to inefficient management. Based on this, it is reasonable to set up the SFA model.

In the SFA model, the input slack variable is the difference between the original input value and the target input value. When the parameter value of each policy interference variable negatively correlates with the input slack variable, it means that an increase of the policy interference variable will lead to an increase of the input slack variable or a decrease in undesired output. An increase in the utilization rate of input factors with an increase of policy interference variables will directly promote the efficiency of energy growth.

The slack variables of incentive policy intervention significantly and negatively correlate with capital stock, employed population, and oil consumption at the 1%, 5%, and 10% levels, respectively. The degree of government R&D subsidies closely relates to technology and investment. The higher is the degree of government R&D subsidies, the more enhanced international scientific and technological cooperation can be, and the more fixed assets can be invested and built to provide more employment for the population. According to the research of Li and Lin [45], the degree of government R&D subsidies has a positive effect on China’s energy efficiency. However, the degree of government R&D subsidies and natural gas supply slack variables do not pass the significance test, indicating that government R&D subsidies only play a role in green economic growth through petroleum energy.

The slack variables of government financial intervention significantly and negatively correlate with capital stock and the employed population at the 1% level, indicating that government intervention is beneficial to the improvement of green energy efficiency. Government intervention often has an impact on the allocation of market resources. In recent years, the state has increased financial support to a region through strategic energy channels such as the Central Asia Natural Gas Pipeline, adjusted the industrial structure, and introduced a series of environmental regulations and systems. A dynamic market mechanism can indirectly improve energy performance, and an appropriate macro-level control will promote and balance energy and economic development.

The slack variable of mandatory policy intervention significantly and negatively correlates with the employed population and oil energy consumption at the 10% level and significantly and positively correlates with natural gas consumption at the 10% level. This shows that in recent years, China has levied corporate pollution discharge fees to encourage relevant companies to improve their pollution discharge capabilities and reduce energy consumption. It also promotes the optimization of the country’s economic structure. For example, in 2018, China’s secondary industry accounted for 36.1% of GDP, while the tertiary industry accounted for 59.7%. The tertiary industry with less pollutant emissions is becoming a pillar industry in China. Generally speaking, moderate intervention by the China government at this stage is conducive to green economic growth.

4.2.3. Analysis of the Third-Stage Empirical Results

According to the parameters after the second-stage regression model, the value of each input element variable in the first stage is adjusted. The input element variable value after eliminating policy interference factors and random errors is then brought into the SBM-DEA model for estimation, and the calculated results are real green energy efficiency. The calculation results are in

Table 5. Comparison of green energy efficiency values between the first stage and third stage of 30 provincial administrative units across the country.

Provincial Administrative unit	2013	2014	2015	2016	2017	Mean (Before Adjustment)	Mean (After Adjustment)
Inner Mongolia	0.835 (20)	0.891 (20)	0.202 (26)	0.500 (28)	1 (1)	0.742 (21)	0.727 (24)
Liaoning	0.677 (27)	0.836 (24)	0.064 (29)	1 (1)	0.827 (21)	0.682 (25)	0.700 (26)
Jilin	0.864 (17)	0.946 (18)	0.724 (23)	0.894 (19)	0.306 (28)	0.754 (20)	0.747 (21)
Heilongjiang	0.769 (25)	0.818 (26)	0.092 (28)	0.489 (30)	0.710 (24)	0.689 (24)	0.576 (29)
Shanghai	0.800 (23)	1 (1)	1 (1)	1 (1)	1 (1)	0.975 (9)	0.960 (9)
Zhejiang	0.830 (21)	0.885 (21)	0.937 (15)	1 (1)	0.894 (15)	0.868 (15)	0.909 (16)
FuJian	0.916 (16)	1 (1)	1 (1)	0.972 (17)	1 (1)	0.906 (12)	0.978 (7)
Guangdong	0.785 (24)	1 (1)	1 (1)	1 (1)	1 (1)	0.709 (22)	0.957 (10)
Guangxi	1 (1)	1 (1)	1 (1)	0.926 (18)	0.298 (29)	0.857 (16)	0.845 (18)
Hainan	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
Chongqing	0.860 (18)	1 (1)	1 (1)	1 (1)	0.863 (19)	0.958 (10)	0.945 (14)
Yunnan	1 (1)	1 (1)	0.950 (14)	1 (1)	1 (1)	1 (1)	0.990 (6)
Shaanxi	1 (1)	0.963 (17)	0.798 (19)	0.759 (22)	1 (1)	0.817 (18)	0.904 (17)
GanSu	0.849 (19)	0.861 (23)	1 (1)	0.846 (21)	0.520 (27)	0.928 (11)	0.815 (19)
Qinghai	1 (1)	0.982 (16)	0.575 (25)	0.552 (26)	0.188 (30)	1 (1)	0.659 (27)
Ningxia	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
Xinjiang	0.810 (22)	0.884 (22)	0.133 (27)	0.567 (25)	0.773 (22)	0.676 (26)	0.633(28)
Beijing	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
Tianjin	1 (1)	1 (1)	1 (1)	1 (1)	0.740 (23)	0.985 (8)	0.948(12)
Hebei	0.669 (28)	0.831 (25)	0.781 (21)	0.869 (20)	0.847 (20)	0.690 (23)	0.799(20)
Shanxi	0.694 (26)	0.206 (30)	0.011 (30)	0.497 (29)	1 (1)	0.570 (27)	0.482 (30)
JiangSu	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
AnHui	1 (1)	0.902 (19)	0.835 (18)	0.978 (16)	0.895 (14)	0.843 (17)	0.922 (15)
Jiangxi	1 (1)	1 (1)	1 (1)	1 (1)	0.894 (15)	0.905 (13)	0.979 (7)
Shandong	1 (1)	0.699 (28)	0.720 (24)	0.679 (24)	0.629 (26)	0.525 (28)	0.745 (22)
Henan	0.626 (30)	0.812 (27)	0.772 (22)	0.728 (23)	0.685 (25)	0.458 (30)	0.725 (25)
Hubei	0.975 (15)	1 (1)	0.905 (16)	1 (1)	0.885 (17)	0.817 (19)	0.953 (11)
Hunan	1 (1)	1 (1)	0.857 (17)	1 (1)	0.874 (18)	0.870 (14)	0.946 (13)
Sichuan	0.667 (29)	0.669 (28)	0.796 (20)	0.542 (27)	1 (1)	0.497 (29)	0.735 (23)
Guizhou	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)	1 (1)
Population mean	0.888	0.906	0.772	0.860	0.828	0.824	0.853
Central region	0.866	0.836	0.650	0.823	0.781	0.738	0.791
Eastern region	0.88	0.932	0.864	0.956	0.903	0.851	0.909
Western region	0.911	0.932	0.769	0.79	0.786	0.866	0.841
Number of provinces on the production frontier	14	15	13	7	14	7	5
Percentage of total provinces	0.615	0.538	0.385	0.538	0.385	23.33%	23.08%

.

Figure 2 shows the comparison of green energy efficiency before and after the adjustment. It can be seen from Figure 2 that after excluding policy interference factors, the overall fluctuations in the efficiencies of the energy and green economies in the three regions are relatively obvious. The average efficiency before and after the adjustment shows a downward trend in 2014. One of the reasons may be that in 2014, the energy development strategy action plan was released in China, and the initial stage of the policy was not stable. Due to the country’s re-arrangement and adjustment of the energy structure at the initial stage of implementation of the plan, some energy industries were clearly impacted. Compared with the efficiency before the adjustment, the efficiency values of most provinces after the adjustment have changed significantly, which shows that the policy interference factors and random error factors are the main reasons for the changes in green energy economy efficiency.

After excluding policy interference factors, we divide the efficiency value into four levels, the first level is efficiency value $1\ge uni{t}_i>0.75$, the second level is $0.75\ge uni{t}_i>0.5$, the third level is $0.5\ge uni{t}_i>0.25$, and the fourth level is $0.25\ge uni{t}_i>0$. Among them, there are 20 provinces with the first level of energy and green economic efficiency, and only one province in Shanxi is at the third level. This indicates that China’s energy green economy efficiency is at a better level. From the point of view of regional average, after setting all regions back to the same external environment, the mean value of China’s overall efficiency distinctly increased, but the efficiency value of the western region declined slightly. The western region dropped from 0.866 before the adjustment to 0.841 after the adjustment. The central and eastern regions rose from 0.738 and 0.851 before the adjustment to 0.791 and 0.909 after the adjustment, respectively. The decline in the western region was smaller than the increase in the central and eastern regions. After adjustment, from 2013 to 2017, the energy and green economic efficiencies of the eastern region were always greater than that of the western region, which stayed basically the same in 2014. The provinces with the highest growth rate are Guangdong, Sichuan, and Henan, all of which are above 0.2. Qinghai has the largest decline in efficiency, with a decrease of 0.341. Shanxi, which has the lowest efficiency value after adjustment, has an efficiency value of only 0.482. From the above, we can see that the efficiencies of China’s energy and green economies at the emergence stage is not balanced, and there is regional heterogeneity. The overall situation of green energy economy efficiency is eastern region > western region > national average > central region.

The policy interference factors we select are all real and have an impact. On the whole, China’s policy is slightly insufficient, and the energy efficiency difference before and after the adjustment is 0.029. From the comparison of efficiency before and after the adjustment, we find that the external environment is obviously better in the western region. In contrast, the central and eastern regions are at a disadvantaged policy environment. The advantageous policy environment under policy preference has a greater impact on energy and economic efficiencies and has a more significant impact on promoting economic growth and reducing environmental pollution emissions. Better policy factors are the main reason for the higher efficiency in the western region before the adjustment. The study of energy consumption by [33] found that the carbon emission density of eastern China is significantly higher than that of other regions. For green energy efficiency that contains environmental pollution as an undesired output, this may be the reason for a region to be doing better in terms of economic growth.

In 2010, the National Development and Reform Commission organized the first batch of low-carbon projects to carry out pilot work in five provinces, formulated low-carbon development plans, and realized industrial greening. They are Yunnan, Hebei, Shaanxi, Liaoning, and Guangdong, the successful implementation of low-carbon projects has enabled Yunnan to do a better job in environmental protection and pollution reduction, with less undesired output. On the contrary, the efficiency of the other four provinces has increased after removing the policy factors, indicating that the low-carbon project policies implemented in 2010 did not have a substantial effect on these four provinces. Some cities in the west, such as Guizhou and Ningxia, have always been at the forefront of production lines. According to the “Evaluation Report on the Construction of Ecological Civilization in China’s Provinces” in recent years, both Guizhou and Ningxia have lagged behind in the evaluation of ecological civilization construction. At the same time, Guizhou and Ningxia are also ranked low in economic rankings and require less energy input. From the perspective of efficiency, it shows that the lower economic rankings of Guizhou and Ningxia are more suitable for the ecological environment. Low input and low output are more suitable, resulting in this high-efficiency situation.

4.3. Luenberger Productivity Index Analysis

For the specific causes of green energy economy, we need to further carry out dynamic decomposition. With the help of Matlab software, this paper uses Luenberger total factor productivity to dynamically analyze the relative position changes between provinces and production boundaries, where LTFP, LSEC, LPTP, and LPEC respectively represent productivity, scale efficiency, technological progress, and pure technical efficiency. Since these policy intervention factors exist objectively, we use data that do not exclude policy intervention factors for measurement.

Table 6. Total factor Luenberger productivity and its composition.

	ltfp	lsec	lptp	lpec
2013–2014	0.083	−0.016	−0.009	0.026
2014–2015	0.037	0.061	−0.023	0.026
2015–2016	−0.070	0.016	−0.029	−0.013
2016–2017	0.147	0.056	−0.003	0.042
Population mean	0.049	0.029	−0.016	0.020

shows the Luenberger total factor average productivity of the two regions from 2013 to 2017. Total factor productivity and economic growth rate have a monotonically increasing relationship [5]. In Table 6, we see that the national productivity growth rate is 4.9%, except for 2015–2016, when the productivity of the exogenous green energy economy was positive. The negative productivity in 2015–2016 was mainly caused by the negative value of technological progress. Scale efficiency has the most important contribution to the improvement of total factor efficiency, accounting for 59.18% of productivity. This shows that China’s oil and gas energy market is currently in the stage of economies of scale and has not yet reached its optimal equilibrium point. Energy strategic cooperation has great potential, and China can continue to expand energy strategic cooperation. What is surprising is that from the four stages of 2013–2017, national technological progress is negative. This shows that China’s oil and gas energy is still mainly imported from other countries at this stage, and the two regions are relatively dependent on external energy. Moreover, the current policy environment may intensify this degree of dependence on foreign energy import, and neither region has established a complete set of energy technology systems.

5. Conclusions

This paper takes 30 provincial administrative units in China as the research object and utilizes the three-stage SBM-DEA model for analysis. The model is based on undesired output and uses the SFA method to eliminate the influence of policy interference factors and error terms. The next step calculates the green economic efficiency of oil and gas energy and then compares the efficiency values before and after adjustment. Employing the Luenberger productivity index to analyze the impact of Luenberger’s dynamic total factor productivity and its constituent factors, this paper draws the following conclusions.

(1) Under China’s current development conditions, incentive economic policy intervention, government financial intervention, and mandatory policy intervention all have an impact on its provincial-level oil and gas green energy efficiency. When policy interference factors are not removed, the overall efficiency average is 0.824. After adjustment, the overall efficiency value increases to 0.853. Only the western region shows a slight decrease in efficiency after the adjustment, which indicates that China overall is in a slightly disadvantaged policy environment.

(2) No matter whether before or after the adjustment, the energy and green economic efficiency gaps in the various regions are large. The difference between the highest and lowest efficiencies before the adjustment is 0.542, while the difference after the adjustment is 0.518. Regional efficiency development is uneven, and the efficiency value of the central region is always lower than the other two regions.

(3) Ningxia and other areas have a backward economy, less energy input, and low ecological environment scores, but China’s energy green efficiency is 1, which we call “false frontier of production efficiency”.

(4) The average value of scale efficiency is 4.9%, and the average value of technical efficiency is −1.6%. This indicates that China’s oil and gas energy development is still in the stage of economies of scale. The energy strategy is dominated by direct input, with heavy external dependence, and energy technology is still at a relatively backward stage.

6. Policy Recommendations

Based on the above research results, we make the following recommendations: (1) Formulate environmental policies in line with the development of various regions. We can see that in the environmental policy of the low-carbon project pilot implemented by the country in 2010, the efficiency values of the four provinces increased slightly after eliminating policy interference, indicating that the policy planning and actual situation do not match, the system is not strong, and local resources and funds have not been effectively integrated. There is an urgent need to establish a systematic green energy consumption policy system. In this regard, various provinces should introduce different management methods, notices, and guidelines in accordance with the different natures of regional economic development and environmental and ecological development.

(2) Prevent the existence of high efficiency caused by low input and low output. Taking Ningxia and Guizhou as examples, when backward economic conditions and low ecological environment rankings occur simultaneously, production efficiency may be at the forefront. In this regard, all provinces should prevent the occurrence of such “high efficiency caused by low input and low output”, maintain a certain economic growth rate, encourage low-carbon consumption, transform economic growth patterns, and call for faster green development. While ensuring the coordinated development of the economy and the environment, it also guarantees its economic growth rate.

(3) Adjust the policy direction so that allow the coexistence of oil and gas input as well as technological development. With the rapid advancement of China’s industrialization and urbanization, the demand for energy in various regions has grown rapidly. The original energy industry system is difficult to change for a while. China’s oil and gas energy has insufficient domestic energy supply and is heavily dependent on external sources. At present, China’s energy strategic channels such as Central Asia and China–Myanmar can, to a certain extent, enable the country to increase oil and gas energy input and ensure sufficient energy supply. At present, China’s foreign oil and gas energy strategy is mainly based on the abundant energy resources of neighboring countries, using the capital and technology of countries along the route to build international oil and gas transmission cooperation corridors, and promote the integration of oil and gas energy industries while ensuring the safety of domestic oil and gas energy supply [46]. However, excessive external dependence means that China’s energy technology is still in the backward stage at present. Cultivating advanced scientific and technological concepts, improving oil and gas energy efficiency, and reducing pollution emissions are important ways to achieve coordinated development of energy, environment, and economy. Blindly relying on energy import has laid a huge hidden danger to China’s energy security. Finding alternative new energy sources and diversifying energy consumption are a new path for the development of China’s energy economy.

(4) As China is still in the stage of economies of scale, it should further expand energy openness, strengthen oil and gas energy cooperation, and guide more oil and gas production companies to join the investment. Scholars such as Chen have studied and concluded that in the early stages of energy policy deployment, the competitiveness of new technologies is often lower because of their higher costs. Therefore, favorable investment policies are needed, including fiscal subsidies, tax cuts, and production quotas [47]. From energy investment promotion to “investment selection”, we should accelerate the green transformation of the energy economy, improve energy utilization, promote the sustainable development of oil and gas trade cooperation, and realize the revolution of energy consumption and utilization under the new era policy.

(5) Currently, the “One Belt and One Road” energy strategy platform is actively being built, and the energy and green economic efficiency of the “Belt and Road” construction area that contains policy interference factors is 0.076 higher than other regions. The “One Belt and One Road” strategic platform provides an ideal trading environment and has a positive impact through integrated regional cooperation among trade [48]. We combine them to verify that the western region construction area is currently in a dominant policy. The Chinese government can take the “One Belt and One Road” construction and western regions as a supporting point, actively explore green economic growth mode, promote the application of new energy technologies, and at the same, increase international cooperation, especially building a favorable policy environment in the central region, guiding all regions to share development concepts, and promoting the common development of all regions.