Total-factor air environmental efficiency and its influencing factors in the areas along the Belt and Road in China: A spatial spillover perspective

Abstract

In order to promote sustainable economic development in the areas along the Belt and Road in China, it is of great necessity to reduce the negative impact of air pollutants resulting from industrialization and urbanization on the complex and fragile ecological environments of neighboring areas. First, this study estimated the total-factor air environmental efficiency (TFAEE) of 17 provinces along the Belt and Road in China from 2010 to 2017 using a slacks-based measure (SBM) model. Second, the global and local Moran indices were used to test the spatial correlations between TFAEEs. Finally, the spatial factors and spatial spillover effects influencing the TFAEEs were investigated using the spatial Durbin model with spatiotemporal double fixed effects. The results were shown as follows: (1) The total-factor TFAEEs of the areas along the Belt and Road were low and showed significant regional spatial differences during 2010–2017. (2) There was a positive spatial autocorrelation between the TFAEEs of the areas along the Belt and Road, and the spatial distribution generally clustered into High-High and Low-Low concentrations. (3) Economic development and technological innovation played significantly positive effects on TFAEEs of the areas in the Belt and Road, while energy consumption structure had negative effect on it. In addition, although industrial structure and environmental regulation were negatively correlated with TFAEEs, the coefficients were not significant. (4) The positive spatial spillover effect of the TFAEEs of the areas along the Belt and Road was mainly the result of significant environmental regulations and insignificant economic development factors, while the technological innovations, energy consumption structures and industrial structures showed insignificant negative spatial spillover effects.

Keywords

Total-factor air environmental efficiency SBM-undesirable model spatial correlation spatial Durbin model the Belt and Road

Introduction

The Chinese Belt and Road initiative was proposed by President Xi Jinping in 2013 and is a major plan to build both the “Silk Road Economic Belt” and the “Twenty-first Century Maritime Silk Road”. In his keynote speech at the Beijing summit of the China Africa Cooperation Forum on September 3 2018, President Xi Jinping explained that The Belt and Road should be a “green road”. However, the geographical and geological characteristics of the areas along the Belt and Road in China are complex and diverse, with huge differences in their spatial geographical characteristics. In addition, the areas along the Belt and Road pass through several major industrial complexes in China, which have brought rapid economic growth. However, due to the high energy consumption and low efficiency of intensive chemical industries in the past, and their high emissions of air pollutants such as SO₂, NOx and dust, air pollution (such as haze, acid rain, photochemical smog, etc.) has become more and more severe. Records show that in January 2013, “the most serious haze weather” occurred even in the central and eastern regions of China, resulting in direct economic losses of 23 billion yuan in terms of traffic and health alone.¹ Furthermore, air pollution is also the main source of acid rain in China. The Yangtze River, Pearl River Delta and Sichuan Chongqing are the main areas suffering from acid rain in China, and the total acid rain area accounts for 30% of China's land area. Records show that agricultural losses caused by acid rain in China are 1.6 billion yuan in total.² Since China promulgated the new ambient air quality standard in 2012, the photochemical smog pollution problem with O3 as the main pollutant in China has been increasing year by year.³ From the 1970s to the 21st century, the type of air pollution in China has changed from the traditional soot pollution to the regional compound pollution, such as acid rain and photochemical pollution can occur in the same region, which is rare in other countries.⁴ The increasingly acute problems associated with air pollution, such as haze and acid rain, not only harm the ecological environment, industrial production, and residents' lives, but also seriously restrict the sustainable development of a green economy in the areas along the Belt and Road.

Under this background, it is more necessary and urgent to measure the air environmental efficiency of the areas along the Belt and Road in China. However, the current researches on the areas along the line mainly focus on the regional economy of China and the countries along the line,^5,6 the Silk Road Economic Belt,^7,8 and The 21st Century Maritime Silk Road,^9,10 and lack the relevant researches on air environmental efficiency. Then, some studies on air environmental efficiency do not consider the “unexpected output” in the process of economic development, and the research method is single, without considering the spatial effect of air environmental efficiency.^11,12 In conclusion, it is urgent to enrich the research on the air environmental efficiency of the areas along the Belt and Road in China. Therefore, this study focused on air pollution along the Belt and Road in China, using accurate calculations of total-factor air environmental efficiency (TFAEE), scientific analyses of regional differences, and spatial correlations between the areas along the Belt and Road. It not only addressed the practicalities of constructing the “green road” in differing local conditions, but also provided reference data and a scientific theoretical basis for sustainable economic development and ecologically sound management of the areas along the Belt and Road initiative in China. Through accurate calculation and decomposition of the TFAEE index along the line, this study scientifically analyzed the regional differences of TFAEE and explored its spatial driving factors, which not only defined the leading factors driving the improvement of TFAEEs along the line, but also pointed out the direction of air environmental improvement, and achieved the purpose of narrowing the regional differences along the line, and It also had important practical significance for promoting the construction of ecological civilization and sustainable development in the areas along the Belt and Road in China.

This study’s contributions to the existing literature lie in the following aspects: (1) It is the first time to utilize spatial econometric model to empirically analyze the influencing factors of TFAEE, which provides more insightful information for the improvement of air quality in the areas along the Belt and Road in China. (2) A comprehensive and detailed theoretical framework supporting the study of the areas along the Belt and Road is introduced, to enrich the models analyzing the TFAEEs; (3) The fluidity and spatial effects of air pollutants, and tests of the spatial correlations between the TFAEEs, are considered using a spatial weight matrix and Moran’s I test; (4) A spatial econometric model is established to make a detailed examination of the spatial factors and spatial spillover effects influencing the TFAEEs of the areas along the Belt and Road in China.

The remainder of this paper is organized as follows: the next section presents a literature review of related studies; the Basic theory and methodology section presents the research methods used in this study; the Basic theory and methodology section presents the discussion and results; the Conclusions and policy implications section draws conclusions and discusses the implications for policy development.

This paper includes many abbreviations, they are summarized in Table 1.

Table 1.

Full names of abbreviations used in this paper.

Abbreviation	Full name
DEA	Date Envelopment Analysis
TFAEE	Total-factor air environmental efficiency
WBCSD	World Business Council for Sustainable Development
DMU	Decision-Making Unit
TFEE	Total-factor energy efficiency
EBM	Epsilon-based measure
SBM	Slacks-based measure
GML	Global Malmquist-Luenberger
GDP	Gross domestic product
SO2	Sulfur dioxide
NOX	Nitrogen oxide
DUST	Smoke and dust
CRS	Constant return to scale
PTE	Pure Technical Efficiency
SE	Scale Efficiency
OLS	Ordinary Least Square
D-W test	Durbin-Watson test
LM test	Lagrange multiplier test
R_LM test	Robust LM test
LR test	Likelihood ratio test
SYS-GMM	System Generalized Method of Moments
DIFF-GMM	Difference Generalized Method of Moments

Brief literature review

The concept of environmental efficiency was first introduced by Freeman et al. in the book “The economics of environmental policy”,¹³ in which environmental quality is described as an economic problem, and the necessity of bringing environmental resources into the economic system is emphasized. Until the early 1990s, eco-efficiency, as defined by the WBDSC, was composed of environmental and resource efficiency.¹⁴ Currently, most previous studies concentrate on the environmental efficiency of energy, water resources, land resources, et al. Scholars worldwide have published a large number of books, papers, and model methods on the various aspects of energy efficiency.^15–17 Among the existing researches on environmental efficiency, the researches on energy efficiency generally present the development process from the traditional single factor energy efficiency to the total factor energy efficiency. The measurement of single factor energy efficiency benefits from its simplicity and strong operability. But it can only measure the ratio of energy input and economic value added, it ignores the relationship between other input factors, environmental impact and other production factors.¹⁸ In this regard, in order to make up for the defects of single factor energy efficiency measurement, based on the previous research, combined with DEA method and the theory of total factor productivity, Hu and Wang developed a new energy efficiency evaluation method - total factor energy efficiency (TFEE), which overcomes the limitations of traditional research methods.¹⁹ Due to its strong universality, TFEE has been widely used by scholars at home and abroad in recent years.^20–22 In addition, many scholars have investigated environmental efficiency from the perspectives of water, bioenergy resource and land resource utilization, which have enriched the range and depth of research into environmental efficiency and verified the feasibility of efficiency measurements in different resource environments. Prominent publications on water resource utilization efficiency include: Song et al.; Gidion et al.; and Gong et al.^23–25 In addition, the water resources efficiency research based on the total factor perspective include: Yao et al.; Zhang et al.^26,27 As one of the main sources of renewable and sustainable green energy resources, bioenergy resource can reduce the negative impact on the environment in the process of conventional energy consumption. In this regard, Alsaleh et al. make a detailed study on the technical efficiency of the bioenergy industry.²⁸ Research on land resources has mainly focused on the efficiency of land resource allocation and land resource utilization efficiency. For example, Liu et al., Gai et al.^29,30 and Deng et al. have conducted research on land resource efficiency allocation,³¹ Xie et al. and Duro et al. on land resource efficiency.^32,33

Although there are now many papers on the environmental efficiency of energy, water resources, and other aspects, relatively few studies deal with the various aspects of TFAEE, and most of the researches mainly focus on air pollution assessment and management, and emission control.^34–36 What should be paid attention to is that many scholars also focus on the spatial effects of air pollutants. For example, Sun et al. pointed out that air pollutants had the characteristics of mobility and spillover, so it was necessary to build a more consistent measurement model of air pollutants.³⁷ Based on the perspective of urban agglomeration, Xu et al. discussed the spatial effects and influencing factors of air pollution in three major urban agglomerations in China by establishing a spatial econometric model, and pointed out that the spatial spillover of air pollution between cities was a problem that must be clarified at the present stage.³⁸ Ding et al. pointed out that it was necessary to consider the spatial spillover effect of air pollutants due to the close relationship between the development of regional economic network of the research objects, and used Moran's I index to verify the significant spatial dependence of industrial air pollution emission intensity in 31 provinces of China.³⁹ Therefore, based on the above research results, the spatial econometric model is more suitable for the analysis of the air environmental efficiency considering air pollutants, so as to better analyze its spatial effects and influencing factors. Besides, in their research on TFAEE, He et al. constructed an SBM-DEA model based on relaxation methods to measure the TFAEEs of districts and counties in Tianjin, and analyzed the factors influencing TFAEEs using the Tobit regression model.⁴⁰ Cai and Ye developed an SBM model to investigate the TFAEEs of China's various industrial capacities.⁴¹ However, the Tobit model used in the analysis of influencing factors ignores the mobility of air pollution and the possible spatial spillover effects, so does not reflect the spatial differentiation and dynamic evolution of TFAEEs. In addition, the SBM model does not consider the “undesirable outputs” of economic development, such as the environmental pressure constraints, so the measured TFAEE cannot accurately reflect the real situation. To address this issue, Wang et al. employed an SBM model with undesirable outputs to construct TFAEE evaluation models for the regions along the Yangtze River economic belt, and explored the causes of regional differences using the Theil index.⁴² Wang et al. examined China's provincial air pollution emission efficiency using an SBM model of undesirable outputs and the GML productivity index.⁴³ This allowed examination of the differences between provincial regions, and their dynamic evolution and influencing factors, from both static and dynamic perspectives. All of these studies on TFAEE demonstrate the usefulness of the DEA method in measuring atmospheric environmental efficiency.

The Chinese Belt and Road initiative has been linked with the eastern, central and western regions in China macroeconomic layout, and has integrated the regional national strategy such as revitalizing the old industrial base in the northeast, the rise of the central region, and the strategy of developing the western region. By promoting and strengthening the interaction and coordinated development between regions, the rapid economic development of the regions along the line is promoted, so as to achieve the vision of alleviating the imbalance of regional economic development. It has rapidly become a hot spot for a large number of scholars to study and explore the theoretical and practical significance of the strategy from different perspectives. Among them, the researches on the areas along the Belt and Road in China mainly focus on the economic trade and investment cooperation with countries along the line;^44,45 the Silk Road Economic Belt;^46,47 the maritime Silk Road.^48,49 In addition, most of the areas along the Belt and Road in China are inland, the Silk Road Economic Belt is located in inland areas. In particular, the ecological environment along the Silk Road Economic Belt is more fragile and lack of water resources, and the rapid economic development along the belt has brought negative effects on the ecological environment. In this regard, Gu et al. and Zhu et al. emphasized the importance of promoting the green sustainable development and protecting the ecological environment in the areas along the line.^50,51 To sum up, this study expanded and applied the research results of TFEE to the research of air environmental efficiency, combined with the hot research area (i.e. area along the Belt and Road in China), and used an SBM model based on undesirable outputs to evaluate the TFAEEs of the areas along the Belt and Road in China. In addition, a spatial econometric model was established to examine the spatial factors influencing variations in local atmospheric environmental efficiencies. It will open up new ideas and methods for the study of TFAEE in this area.

Basic theory and methodology

TFAEE and the SBM-undesirable model

At present, there is no relative conclusion about the concept of air environmental efficiency, but from the existing literature researches, it mainly comes from the deepening and extension of the concept of environmental efficiency and energy efficiency. The total-factor air environmental efficiency(TFAEE) proposed in this study belongs to the category of energy and environmental efficiency. It is based on the total-factor theory, the research results of energy and environmental efficiency are extended and applied to the research of air environmental efficiency. Such as Wang et al., in his research, air environmental efficiency was defined as the ratio of economic output and air pollution emissions, and DEA method was used to solve the air environmental efficiency.³⁵ From the perspective of total-factors, Wang et al. defined “air pollution emission efficiency” as a decision-making unit to realize the minimization of factor input and air pollutant emission while pursuing the maximization of economic output, and The directional distance function (DDF) was used to measure the efficiency level.⁵² Combined with the existing researches, from the perspective of input-output description, the total-factor air environmental efficiency proposed in this study is the efficiency that a decision-making unit pursues the maximum expected output (GDP) and the least unexpected output (air pollutants) while giving input variables. In addition, the SBM-undesirable model selected in this study just fits this point, which can more reasonably and scientifically measure the total factor air environmental efficiency.

The traditional DEA method has been widely used for various efficiency measurements because of its suitability in evaluating the efficiency of multi-input and -output decision-making situations. However, the traditional DEA model is limited to evaluations of the efficiency of decision-making situations from the radial and angle levels, and ignores the slack in redundancies and insufficiencies in input and expected output elements. It was not until the non-radial and non-oriented SBM model based on relaxation measures was proposed by Tone that these problems were solved, enabling the efficiency of DMUs to be more accurately evaluated.⁵³ This study uses an SBM-undesirable model, an improvement over the non-radial and non-angle SBM model based on relaxation measures of Tone, in an improved form created by splitting the output items and introducing “undesirable outputs”.⁵⁴ This improved SBM model takes into account the “undesirable outputs” in the economic development process, including environmental pressure outputs, such as waste water and waste gas, so that the TFAEEs calculated by this model are more reasonable and accurate.

According to the “undesirable outputs” in the TFAEE, a production possibility set (PPS) including the undesirable output items was constructed. It is supposed that there are n DMUs, including M inputs and P outputs. The P outputs were divided into p₁ desirable output and p₂ undesirable output. The matrix input index was defined as $X \in R^{m \times n}$ , the desirable output index as $Y^{g} \in R^{p_{1} \times n}$ , and the undesirable output indicators as $Y^{b} \in R^{p_{2} \times n}$ . Regarding the CRS, the set of production possibilities were defined as follows:

G = {(x, y^{g}, y^{b}) | x \geq X λ, y^{g} \leq Y^{g} λ, y^{b} \geq Y^{b} λ, λ \geq 0}

(1)

Then, the efficiency of a specific DMU to be evaluated can be calculated using the following formula:

\begin{array}{l} \min ρ = \frac{1 - \frac{1}{M} \sum_{m = 1}^{M} \frac{s_{m}^{-}}{x_{m 0}}}{1 + \frac{1}{P_{1} + P_{2}} (\sum_{i = 1}^{p_{1}} \frac{s_{i}^{g +}}{y_{i 0}} + \sum_{i = 1}^{p_{2}} \frac{s_{i}^{b +}}{y_{i 0}})} \\ s . t . {\begin{cases} x_{0} = X λ + s^{-} \\ y^{g} = Y λ - s^{g +} \\ y^{b} = Y λ + s^{b +} \\ λ \geq 0, s^{-} \geq 0, \\ s^{g +} \geq 0, s^{b +} \geq 0 \end{cases} \end{array}

(2)

Where is $ρ$ the value of TFAEE to be calculated; $s^{-}, s^{g +}, s^{b +}$ are the relaxation values of factor inputs, desirable outputs and undesirable outputs; and $λ$ is the weight matrix. The inputs and outputs on the frontier can be obtained by multiplying $λ$ by the factor input and output matrices. $ρ$ is strictly monotone decreasing with respect to $s^{-}, s^{g +}, s^{b +}$ , where $0 \leq ρ \leq 1$ . If, and only if, $ρ = 1$ , the current DMU is valid. When $ρ = 1$ , the current DMU evaluated is inefficient.

$ρ$ in formula (2) is the technical efficiency (TE) of SBM undesirable under the assumption of constant returns to scale (CRS). When the constraint $e λ$ is added, the technical efficiency can be further decomposed into pure technical efficiency (PTE) and scale efficiency (SE), and the relationship between them can be expressed as: $T E = PTE \times S E$ . When $e λ = 1$ , formula (2) is the SBM undesirable model under the assumption of variable returns to scale (VRS), and the measured value is pure technical efficiency (PTE); when $e λ \leq 1$ or $e λ \geq 1$ , formula (2) is the SBM undesirable model under the assumption of diminishing returns to scale (DRS) or increasing returns to scale (IRS). At this time, it can be used to decompose the measured TFAEE index, and clarify the essence of the difference of TFAEE in the areas along the line, so as to better explore the key factors influencing the change of TFAEE of the areas along the Belt and Road in China.

Spatial autocorrelation test

Spatial weight matrix

In this study, a binary matrix based on the Queen standard was established as the weight matrix of a geographically adjacent space.

W_{i j} = {\begin{cases} 1 \begin{matrix}  \end{matrix} (region i is adjacent to region j) \\ 0 \begin{matrix}  \end{matrix} (region i is not adjacent to region j) \end{cases}

(3)

Global Moran index

Spatial correlation (spatial dependence) is one of the sources of spatial effect recognition. The premise of spatial econometric analysis is to test whether spatial correlations exist in spatial data. The commonly used test methods include Moran’s I test and Geary’s c test. According to the requirements of a particular study, it will use either global or local Moran’s I tests to test the spatial correlations between the TFAEEs of the areas along the Belt and Road in China.

The global Moran index reflects the degree of overall spatial correlation between a certain attribute value and its counterpart in a spatially adjacent or adjacent regional unit. The formula for calculating it is as follows:

Moran' s I = \frac{\sum_{i = 1}^{n} \sum_{j = 1}^{n} W_{i j} (A_{i} - \bar{A}) (A_{j} - \bar{A})}{S^{2} \sum_{i = 1}^{n} \sum_{j = 1}^{n} ω_{i j}}

(4)

where

Moran' s I

is expressed as a global Moran index;

i, j

represents the observed area, respectively;

A_{i}, A_{j}

is the value of the TFAEE in that area;

i, j

\bar{A}

is the average value of the TFAEE in each area; and

S^{2} = {(\sum_{i = 1}^{n} A_{i} - \bar{A})}^{2} / n

is the variance, and

W_{i j}

is the spatial weight matrix. The Moran index ranges between −1 and 1. If its value is between −1 and 0, it indicates the existence of a negative spatial correlation; if its value is between 0 and 1, it indicates the existence of a positive spatial correlation, and a value equal to 0 indicates no spatial correlation.

Local Moran index

Because the global Moran index can only reflect the degree of spatial correlation between regional units and the entire area, it cannot be used to test the spatial correlations between local regional units. If there is a positive correlation between some local units and a negative correlation between others, the global Moran index may indicate that there are no spatial correlations anywhere in the region as a whole. The local Moran’s I index overcomes this problem because it tests whether there are spatial correlations between local areas. The formula for calculating it is as follows:

Moran' s I_{i} = \frac{(A_{i} - \bar{A})}{S^{2}} \sum_{j}^{n} W_{i j} (A_{j} - \bar{A})

(5)

where Moran’s I_i is the local Moran index. If it is positive, the TFAEE of that regional unit is similar to that of the adjacent unit, and there is a “high-high (HH)” or “low-low (LL)” aggregation; if it is negative, the regional unit and the adjacent unit have differing TFAEEs, and there is no “HH” or “LL” aggregation.

Spatial Durbin model

At present, there are three kinds of spatial econometric models;^55,56 the spatial lag model (SLM), the spatial error model (SEM) and the spatial Durbin model (SDM). The SDM is a more general expression of the SLM and SEM, and contains the general characteristics of both. The model expression is:

Y = ρ W_{i j} Y + X β + λ W_{i j} ε + μ

(6)

where Y is the vector of the explained variable; X is the matrix of the explanatory variable; W_ij is the spatial weight matrix;

W_{i j} Y

is the explained variable with spatial lag; ρ is the spatial autoregressive coefficient, which reflects the degree of diffusion and spillover effect between adjacent regions; β is the influence of the explanatory variable on the explained variable; λ is the spatial error coefficient of the explained variable vector; ε is the random error vector; and μ is the random error vector of normal distribution. When ρ=0, the spatial Durbin model is equivalent to the spatial error model, and when λ=0, the spatial Durbin model is equivalent to the spatial lag model.

Based on the above, by introducing the spatial effect and time effect, combined with the content of this study, the static spatial panel Durbin model of the spatial driving factors of TFAEEs is established. The model expression is:

\begin{array}{l} T FAE E_{i t} = α_{0} + ρ W_{i j} T FAE E_{i t} + α_{1} E D_{i t} + α_{2} T I_{i t} + α_{3} E C S_{i t} + α_{4} I S_{i t} + α_{5} E R_{i t} \\ + δ_{1} W_{i j} E D_{i t} + δ_{2} W_{i j} T I_{i t} + δ_{3} W_{i j} E C S_{i t} + δ_{4} W_{i j} I S_{i t} + δ_{5} W_{i j} E R_{i t} + υ_{i} + ξ_{t} + μ_{i t} \end{array}

(7)

In formula (7), $i, t$ represents the variables of region and year, $α$ is the coefficient of explanatory variable and explanatory variable, $ρ, δ$ is the spatial coefficient of explanatory variable, $υ_{i}$ is spatial effect, $ξ_{t}$ is time effect, and explanatory variable is economic development (ED), technical innovation (TI), energy consumption structure (ECS), industrial structure (IS), environmental regulation (ER).

As the diffusion of air pollutants has time continuity, and considering the possible endogenous problems, this study introduces the lag phase I data of TFAEE index into the static spatial panel Durbin model, constructs the dynamic spatial panel Durbin model, and uses the SYS-GMM method to test the endogenous. The dynamic spatial panel Durbin model is as follows:

\begin{array}{l} T FAE E_{i t} = α_{0} + τ TFAE E_{i t - 1} + ρ W_{i j} T FAE E_{i t} + α_{1} E D_{i t} + α_{2} T I_{i t} + α_{3} E C S_{i t} + α_{4} I S_{i t} \\ + α_{5} E R_{i t} + δ_{1} W_{i j} E D_{i t} + δ_{2} W_{i j} T I_{i t} + δ_{3} W_{i j} E C S_{i t} + δ_{4} W_{i j} I S_{i t} + δ_{5} W_{i j} E R_{i t} + υ_{i} + ξ_{t} + μ_{i t} \end{array}

(8)

It can be obtained from the above formula that when $τ = 0$ , formula (8) is a static space panel model; when $ρ = δ = 0$ , it is a dynamic panel data model; when $τ = ρ = δ = 0$ , it is a common panel data model.

Empirical results and discussion

Data

The input-output variable data, is taken from sample surveys covering the period 2010–2017, from 18 regions (provinces, municipalities and autonomous regions) adjacent to the Belt and Road in China. In order to better analyze the regional heterogeneity of TFAEE, the areas along the Belt and Road were divided into four regions. Among them, the northeast region includes four provinces (Inner Mongolia, Liaoning, Jilin, and Heilongjiang), the southeast region includes five provinces (Shanghai, Fujian, Guangdong, Zhejiang, and Hainan), the northwest region includes five provinces (Shaanxi, Gansu, Ningxia, Qinghai, and Xinjiang), and the southwest region includes four provinces (Guangxi, Chongqing, Yunnan, and Tibet). Due to the lack of data on energy consumption in Tibet in the statistical yearbook, Tibet was not included in this study, so that only 17 provinces along the Belt and Road are treated here. The locations of the four regions are shown in Figure 1.

Figure 1.

Map of the four Belt and Road regions in China.

As in previous study of Wang et al.,⁴³ this study also focuses on capital stock, labor force, and energy consumption as input variables. 1) Because data on the capital stock values of all 17 provinces were not directly available, this study adopted the perpetual inventory method using the processing methods and data of Shan,⁵⁷ and estimated the capital stock of each province from 2006–2017 based on 2006 prices. 2) The labor input was the number of employees at the end of each year in each province. 3) The energy consumption of each province is expressed in standard coal equivalents (the energy generated by burning one metric ton of coal). The outputs were defined as follows. 1) The desirable output in GDP of each province. In order to keep the data consistent, the GDP data of the provinces from 2010–2017 was also expressed in 2006 equivalent values. 2) The undesirable outputs were sulfur dioxide (SO2), nitrogen oxide (NOx) and smoke dust (DUST), the principle industrial air pollutants. Both are extremely harmful both to industrial production and residents’ lives, and are the main focus of air pollution regulations in China.

The input data (capital stock, labor force, energy consumption), and the outputs (GDP, SO₂, NOx, DUST) were compiled from the China Statistical Yearbooks (2010–2017), the China Energy Statistical Yearbooks (2010–2017) and the China Provincial Statistical Yearbooks (2010–2017). The summary input and output statistics are shown in Table 2.

Table 2.

Summary statistics for inputs and outputs 2010–2017.

Variable		Units	Min	Max	Mean	Std. Dev
inputs	Labor	Million workers	307.65	6340.79	2010.14	1378.16
	Capital stock	RMB Billion (2006)	2786.84	76427.77	22975.21	17200.71
	Energy	Million tons of coal Equivalent (Mtsce)	1314.66	31677.03	11600.37	6770.76
Desirable output	GDP	RMB Billion (2006)	1070.38	73358.35	14635.54	13596.34
Undesirable outputs	SO₂	Million tons (Mt)	1.43	140.94	49.67	31.72
	NOx	Million tons (Mt)	5.6	142.19	53.51	32.58
	Dust	Million tons (Mt)	1.5	112.07	35.55	23.55

Calculation of the atmospheric environmental efficiency in the areas along the belt and road in China

The TFAEEs of the 17 provinces along the Belt and Road from 2010–2017 were calculated according to formulas (1) and (2), and the results are shown in Figure 2 and Table 3.

Figure 2.

The average TFAEEs of the four regions along the Belt and Road in China (2010–2017).

Table 3.

The TFAEEs of 17 provinces along the Belt and Road in China (2010–2017).

Province	2010	2011	2012	2013	2014	2015	2016	2017	Average
Inner Mongolia	0.403	0.402	0.403	0.416	0.427	0.427	0.425	0.416	0.415
Liaoning	0.488	0.478	0.476	0.491	0.495	0.483	0.446	0.430	0.473
Jilin	0.451	0.444	0.460	0.464	0.471	0.473	0.477	0.466	0.463
Heilongjiang	0.535	0.515	0.514	0.526	0.519	0.514	0.506	0.506	0.517
Shanghai	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000
Zhejiang	0.667	0.633	0.641	0.632	0.660	0.661	0.660	0.620	0.647
Fujian	0.667	0.643	0.649	0.664	0.667	0.666	0.681	0.682	0.665
Guangdong	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000	1.000
Hainan	1.000	0.561	0.547	0.548	0.557	0.544	0.522	0.486	0.596
Shaanxi	0.446	0.438	0.445	0.449	0.461	0.458	0.460	0.457	0.452
Gansu	0.341	0.345	0.353	0.363	0.362	0.366	0.381	0.371	0.360
Qinghai	0.275	0.279	0.284	0.286	0.296	0.296	0.296	0.296	0.289
Ningxia	0.246	0.239	0.244	0.245	0.254	0.252	0.253	0.248	0.248
Xinjiang	0.325	0.313	0.306	0.297	0.309	0.308	0.301	0.298	0.307
Guangxi	0.455	0.459	0.468	0.472	0.474	0.474	0.479	0.474	0.469
Chongqing	0.513	0.523	0.540	0.552	0.572	0.585	0.621	0.634	0.568
Yunnan	0.330	0.323	0.329	0.349	0.351	0.357	0.357	0.374	0.346
Average	0.538	0.506	0.509	0.515	0.522	0.521	0.521	0.515	0.518

Taken overall, the average TFAEE of the regions along the Belt and Road between 2010–2017 was 0.518, and showed a 48.2% improvement in efficiency and a huge reduction in potential air pollution over this period. The average TFAEE values in the northeast, southeast, northwest and southwest regions were 0.467, 0.781, 0.331, and 0.461, respectively, showing significant regional differences where southeast > northeast > southwest > northwest. The data showed that TFAEE was closely related to the degree of regional economic development and geographical location, showing a decreasing trend from the eastern coastal regions to the inland regions.

Regarding the individual provinces, it is shown in Table 3 that between 2010–2017, Shanghai and Guangdong consistently had the highest TFAEEs, with averages of 1.000. Shanghai and Guangdong have achieved a leap forwards in their control of air pollution and emission reduction while maintaining their rapid economic development, achieving a win-win economic and environmental situation. Zhejiang, Fujian, and Hainan, located in the southeast coastal region, are in the middle level, with average TFAEEs of 0.596–0.665. These provinces have significant geographical advantages and have benefited from changes in national policies, largely driven by their ability to prevent and control air pollution. However, Liaoning, Jilin, Heilongjiang, and Inner Mongolia in the northeast region were lower than the median, with average TFAEEs of 0.415–0.517. These areas were once the industrial bases which gave birth to China’s new industrial economy. However, as their economies have reformed and opened up their institutional and structural contradictions have become increasingly acute. In addition, they have many sources of air pollution, which place great pressure on their attempts to improve their atmospheric environments. Shaanxi, Gansu, and other the provinces in the northwest region and Guangxi, Chongqing, and Yunnan provinces in the southwest region had average TFAEE values, lying between 0.248–0.568. Because the northwest and southwest regions of China are located inland, their ecological conditions are relatively complex, so that their accumulated air pollutants do not disperse easily. In addition, the economic foundations of some of these provinces remain poor and their technical levels are not very advanced, so that their atmospheric environmental efficiency has been low for a long time.

Although the average TFAEE of all areas along the Belt and Road decreased from 0.538 to 0.518 during the sampling period, it fluctuated slightly by about 0.500 annually. Clearly, while the guiding concept of the “green road” has eased the conflicts between economic growth and environmental protection in China, there is still a long way to go to coordinate development between them.

Decomposition of the TFAEEs of the areas along the belt and road in China

In order to better explore the key factors of TFAEEs of the areas along the Belt and Road in China change, TFAEEs is divided into pure technical efficiency (PTE) and scale efficiency (SE), and the relationship between them is expressed as TE = PTE EeSE. The decomposition results are shown in Figures 3 to 5.

From the overall analysis, Figure 3 shows that during the sample investigation period, the decomposition index of TFAEEs along the line is SE> PTE > TE, with the average values of 0.808, 0.641 and 0.518 respectively. In addition to the PTE trend showing a slow upward trend, the trend of SE is consistent with the general characteristics of TE, showing a mild downward trend year by year. From the perspective of regional analysis, Figure 4 shows that the decomposition indexes PTE and SE of TFAEEs in the four regions showed significant regional differences, in which PTE showed Southeast > Northwest > Northeast > Southwest, with the average values of 0.879, 0.594, 0.514 and 0.499 respectively. The main reason is that the southeast coastal area has obvious advantages in green technology innovation ability compared with other areas. Under the combined effect of various factors (policy inclination, geographical location and strong financial resources), it not only lays the foundation for superior technical ability, but also provides technical guarantee for the improvement of air environmental efficiency. Compared with the southeast region, the PTE index of the other three regions are lower. The reason is that the level of economic development is constrained by China’s unbalanced regional strategy and the geographical location of the inland areas. In addition, the environmental cost of backward pollution control technology is higher, which leads to the PTE level of the three regions far behind the southeast coastal areas. In addition, except for the northwest region, the overall level of SE in the other three regions is higher and the difference is not obvious, showing Northeast > Southwest > Southeast > Northwest, with the average values of 0.936, 0.896, 0.889 and 0.558 respectively, Compared with PTE, the value of SE is close to the efficiency frontier, and the improvement potential is limited. If we continue to blindly improve the production scale, it will lead to diminishing marginal returns, thus inhibiting the improvement of TFAEEs in the areas along the line. Therefore, the four regions need to improve the allocation efficiency of resources and optimize the industrial structure to achieve the goal of energy conservation and emission reduction, so as to promote the sustainable development of green economy along the line. The decomposition results of provinces and cities are shown in Figure 5.

Figure 3.

Change trend of the TFAEEs and its decomposition of the areas along the Belt and Road in China (2010–2017).

Figure 4.

Decomposition of the TFAEEs of the four regions along the Belt and Road in China (2010–2017).

Figure 5.

Efficiency decomposition distribution diagram of the TFAEEs of the areas along the Belt and Road in China (2010–2017).

From the perspective of provinces, Figure 5 shows that Shanghai, Zhejiang, Fujian and Guangdong are the provinces with high PTE from 2010 to 2017. The four provinces are the core areas of economic development in the eastern coastal areas with high level of science and technology, and the prevention and control of air pollution are relatively excellent. In addition, Hainan, Qinghai and Ningxia are special. Although PTE is high, SE is very low. Because of the imbalance of natural endowment conditions and resource allocation, and the lack of factor input, SE is far away from the frontier of efficiency. Inner Mongolia, Heilongjiang and other nine provinces are mainly concentrated in the areas with low PTE and high SE, which indicates that the low level of TFAEEs of the areas along the Belt and Road in China is caused by the low PTE. As Alsaleh et al. emphasized, the choice of energy sources is one of the most important determinants of pure technical efficiency, while the availability of energy resources is related to scale efficiency.⁵⁸ Therefore, in order to better implement the green concept and improve the TFAEEs, the areas along the Belt and Road in China should focus on the improvement of PTE, and improve the existing management measures of energy conservation and emission reduction by introducing cutting-edge air pollution prevention and control science and technology. In addition, compared with PTE, SE is closer to “effective”, and the space for improvement is relatively limited. On the contrary, too large scale of production investment, such as excessive investment of social capital, and too large scale of production at the cost of environment, will lead to the imbalance of resource allocation and the phenomenon of “diseconomy of scale”. Therefore, in order to further promote the improvement of TFAEEs in the areas along the line, while improving the ability of resource allocation and environmental management, we should focus on the innovation of green high-tech, so as to effectively play the technical effectiveness and scale effectiveness of resource utilization, and let the two produce synergistic effect.

Testing the spatial correlations between TFAEEs in the areas along the belt and road in China

Global Moran index test results

Formula (3) establishes the weight matrices of geographically adjacent spaces, and the global Moran index can be used to test the spatial autocorrelations between the TFAEEs of areas along the Belt and Road in China. Hainan Province is not directly adjacent to any of the other provinces sampled so it was changed manually in the spatial weight matrix analysis to be treated as adjacent to Guangdong Province, its nearest neighbor. The test results are shown in Table 4.

Table 4.

Moran’s I test results for the TFAEEs (2010–2017).

Year	Moran's I	P-value	Z-score
2010	0.663***	0.004	3.171
2011	0.461**	0.025	2.342
2012	0.456**	0.027	2.316
2013	0.445**	0.029	2.265
2014	0.470**	0.024	2.37
2015	0.459**	0.025	2.321
2016	0.438**	0.029	2.211
2017	0.376**	0.049	1.941

Note: **, *** represent significance levels of 5%, 1%, respectively.

Table 4 shows that the TFAEEs of the areas along the Belt and Road were positive and passed the significance test, showing that the TFAEEs of the provinces were positively correlated and highly concentrated in space. Between 2010–2017, the overall Moran index decreased from 0.663 to 0.376 The degree of spatial aggregation in the TFAEEs was relatively unstable and showed a converging trend. The largest decline in the period 2010–2011 was mainly due to the unprecedented levels of key air pollutant control in the regional areas of China since 2010. For example, the “11th Five Year Plan” for the prevention and control of acid rain and sulfur dioxide pollution in China was initiated in 2010, and greatly reduced the regional differences in TFAEE between 2010 and 2011.

Analysis of Moran scatter plots

Moran scatter plots present a more intuitive picture of the change trends in regional concentration and spatial correlations of TFAEEs in the areas along the Belt and Road. Moran scatter plots and distribution tables for 2010, 2013, 2015 and 2017 are shown in Figure 6 and Table 5 and present a summary overview of the results of this study.Figure 3. Change trend of the TFAEEs and its decomposition of the areas along the Belt and Road in China (2010—2017).

Figure 6.

Table 5.

Moran scatter distribution of TFAEEs (2010, 2013, 2015, and 2017).

Quadrant	2010	2013	2015	2017
First quadrant (High-High)	Shanghai, Zhejiang, Fujian, Guangdong, Hainan	Shanghai, Zhejiang, Fujian, Guangdong, Hainan	Shanghai, Zhejiang, Fujian, Guangdong, Hainan	Shanghai, Zhejiang, Fujian, Guangdong
Second quadrant (Low-High)	–	Heilongjiang, Chongqing	Chongqing	Chongqing
Third quadrant (Low-Low)	Inner Mongolia, Liaoning, Jilin, Heilongjiang, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang, Yunnan, Chongqing	Inner Mongolia, Liaoning, Jilin, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang, Yunnan	Inner Mongolia, Liaoning, Jilin, Heilongjiang, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang, Yunnan	Inner Mongolia, Liaoning, Jilin, Heilongjiang, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang, Yunnan
Fourth quadrant (High-Low)	Guangxi	Guangxi	Guangxi	Guangxi, Hainan

In a Moran scatter plot, the first and third quadrants showed positive spatial correlations, and the second and fourth quadrants showed negative spatial correlations. Figure 6 and Table 5 show that most provinces are concentrated in the first and third quadrants, indicating significant spatial differentiation of the TFAEEs of the areas along the Belt and Road. The TFAEEs of Shanghai, Zhejiang, Fujian, and other southeast coastal provinces were concentrated in the first quadrant because these provinces have better geographical locations, more advanced green energy production and emission reduction technology, and are more effective in preventing and controlling air pollution. In recent years, they have not only improved their own TFAEEs, but also contributed to improvements in surrounding areas. Chongqing was in the second quadrant of low-high aggregation. Inner Mongolia, Liaoning, Jilin, and other provinces in the northeast and southwest regions were mainly concentrated in the third quadrant. These provinces are located in inland, economically underdeveloped areas, and their energy consumption structure is relatively limited, their technical level is poor and their ecological environments are fragile, which has seriously hindered the improvement of their TFAEEs. Guangxi was in the fourth quadrant of high-low concentration. Although it is surrounded by provinces with high TFAEEs, it lacks good communications, both internally and between surrounding regions. Provinces with high TFAEEs often failed to play a leadership role, increasing the tendency for inter-regional polarization.

Local Moran index test results

In order to further explore the spatial distribution characteristics of TFAEEs among the 17 provinces along the Belt and Road in China, the local spatial correlations of TFAEEs among provinces was tested by drawing a Lisa cluster map and a Lisa significance map. For brevity, Figure 7 only shows the specific results for the 2010 and 2017 Lisa cluster and significance maps.

Figure 7.

LISA cluster and significance maps of TFAEE (2010, 2017).

As shown in Figure 7, in the 2010 Lisa cluster diagram, Zhejiang, Fujian, and Guangdong were in the high-high cluster area, Xinjiang and Gansu provinces were in the low-low cluster area, while there were no local spatially autocorrelated provinces in the low-high and high-low cluster areas. In the 2010 Lisa significance map, only Gansu Province passed the 1% significance level test, four provinces (Zhejiang, Fujian, Guangdong, and Xinjiang) passed the 5% significance level test, while other areas along the Belt and Road were not significantly clustered.

In the Lisa cluster map for 2017, Zhejiang and Fujian provinces were in the high-high cluster area, Hainan Province was in the low-high cluster area, Xinjiang and Gansu provinces were in the low-low cluster area, and there were no local spatially autocorrelated provinces in the high-low cluster area. In the Lisa significance map for 2017, Hainan and Gansu provinces passed the 1% significance level, three provinces (Zhejiang, Fujian, and Xinjiang) passed the 5% significance level, while other areas along the Belt and Road were not significantly autocorrelated.

According to the results of Figures 6 and 7, the high-high cluster area is mainly located in the southeast along the line, while the low-low cluster area is in the northwest and northeast inland areas, and there are few provinces in the low-high and high-low cluster area, especially in the Lisa cluster map, which further verifies the significant “spatial imbalance” of TFAEEs of the areas along the Belt and Road in China. Especially for the northwest and northeast inland areas with low TFAEEs value, they have been in the core area of low-low concentration from 2010 to 2017, which are easy to fall into the dilemma of slow economic growth and hindered improvement of air environment. Luo and Li pointed out that the air pollutants often present regional broadcasting effect, and the air pollutants between cities along the line often pass on each other through the feedback effect of interaction.⁵⁹ In this regard, Liang et al. emphasized that promoting the development of new urban agglomerations is conducive to reducing the intensity of air pollutants. When the urbanization rate exceeds the critical point of 68.6%, industrial agglomeration presents emission reduction effect, thus improving the efficiency of economic development.⁶⁰ In addition, the number of provinces with high-level cluster decreased from 2010 to 2017, and the trend of low-high or high-level concentration became more significant. The adjacent areas of Guangxi, Chongqing and Hainan Provinces are mostly the areas with high or low TFAEEs values, which reflected that the differences of TFAEEs in adjacent areas were gradually increasing. Therefore, Ding et al. pointed out in the relevant research that the air pollution spillover is strong, and there are great differences in the air pollution situation and economic development level between regions in China. It is necessary to adjust measures to local conditions and strengthen the environmental governance responsibility of local governments.⁶¹ In order to reverse the current situation of the areas along the line and change the development concept in time, new challenges are put forward to local governments, it is necessary for the relevant departments to do a good job in the prevention and control of air pollution and improve the quality of economic development of the areas, which is the rational choice to realize the sustainable development of the areas along the Belt and Road in China.

Analysis of influencing factors

Selection of spatial factors influencing the variables

The changes in TFAEE in areas along the Belt and Road were affected by many factors, including socio-economic factors and spatial geographical effects. However, the traditional econometric model method not only ignores the fluidity and spatial effects of air pollutants, but also cannot accurately reflect the spatial effects of TFAEE. The spatial econometric model extends the traditional econometric model by considering spatial effects, and improves the analysis results making them more scientific. Therefore, the spatial factors and spatial spillover effects influencing the TFAEEs in areas along the Belt and Road were more deeply explored using the spatial econometric model used in this study.

Based on existing research combined with the characteristics of the areas along the Belt and Road in China, we shall now explore the spatial factors influencing the TFAEEs from the following five perspectives：

Economic development (ED): This study used the per capita GDP of each province as an index to measure economic development as an explanatory variable. In order to eliminate the influence of price factors and heteroscedasticity, the per capita GDP index was expressed in the logarithm value of 2016 prices.

Technological innovation (TI): This study used the proportion of internal R&D expenditure of industrial enterprises as a proportion of GDP in each province as an index to measure technological innovation as an explanatory variable.

Energy consumption structure (ECS): This study used coal consumption as a proportion of total energy consumption in each province as an index to measure energy consumption structure as an explanatory variable.

Industrial structure (IS): This study used secondary industry as a proportion of GDP in each province as an index to measure industrial structure as an explanatory variable.

Environmental regulation (ER): This study used the proportion of GDP invested in environmental pollution control in each province as an indicator to measure environmental regulation as an explanatory variable.

Data for these five explanatory variables were obtained from the China Statistical Yearbooks (2010–2017), the China Science and Technology Statistical Yearbooks (2010–2017), the China Energy Statistical Yearbooks (2010–2017), and the China Environment Statistical Yearbooks (2010–2017).

Ordinary panel regression model and the LM test

Before constructing the spatial econometric panel model, we used the Huasman test to judge whether the panel data model should use the fixed effect model or the random effect model. The test results showed that the Hausman test statistic was 42.030, and had passed the 1% significance level. Therefore, the original random effect hypothesis was rejected and the spatial econometric model with fixed effect was used.

According to the spatial econometric model selection criteria proposed by Anselin et al., the ordinary panel regression model should be carried out first, and the LM Test and robust LM Test should be conducted based on the model residuals, to determine which spatial econometric model is applicable.⁶² The regression results and LM test results are shown in Table 6.

Table 6.

Regression results of the ordinary panel data model and the LM test results.

Explanatory variables	OLS			Spatial fixed effect			Time fixed effect			Spatiotemporal double fixed effect
Explanatory variables	Coefficients	t-statistic	P-value	Coefficients	t-statistic	P-value	Coefficients	t-statistic	P-value	Coefficients	t-statistic	P-value
ED	0.273***	7.829	0.000	–0.016	–0.442	0.659	0.216***	7.794	0.000	0.491***	4.063	0.000
TI	10.090***	3.225	0.002	0.443	0.185	0.854	14.559***	7.120	0.000	3.649	1.509	0.134
ECS	–0.158**	–2.548	0.012	–0.149**	–2.230	0.027	–0.105***	–2.680	0.008	–0.049	–0.736	0.463
IS	0.435***	3.979	0.000	–0.017	–0.123	0.902	–0.887***	–7.482	0.000	–0.259*	–1.659	0.100
ER	–2.886	–1.529	0.129	0.155	0.182	0.856	–4.462***	–3.709	0.000	–0.124	–0.157	0.876
R-squared	0.576		0.041		0.833		0.153
sigma^2	0.020		0.002		0.008		0.001
loglik	74.357		246.877		137.936		259.025
D-W	1.117		2.014		1.333		2.214
LM_lag	36.811***		0.000	0.062		0.803	0.541		0.462	0.064		0.800
R_LM_lag	18.505***		0.000	1.852		0.174	0.122		0.727	0.340		0.560
LM_error	18.375***		0.000	0.232		0.630	3.097*		0.078	0.127		0.721
R_LM_error	0.069		0.793	2.021		0.155	2.677		0.102	0.404		0.525

Note: The t-test value are given in brackets: *, **, and, *** represent significance levels of 10%, 5%, 1%, respectively.

Table 6 shows that the LM lag and LM error test results of the OLS model were significant, while the robust LM Test was only significant for the R-LM-lag test, indicating that the spatial lag model should be used in this paper. In addition, considering that the fixed effect model was divided into the spatial fixed effect model, the time fixed effect model, and the time and space double fixed effect model, the choice between these three should be made on the basis of an LR test. The LR test results are shown in Table 7.

Table 7.

LR test results.

Inspection items	Test value	Variance	P-value
LR test of spatial fixed effect	242.177***	17	0.000
LR test of time fixed effect	24.296***	8	0.002

Note: *** represents significance level 1%.

Table 7 shows that the LR test results for the spatial fixed effect and time fixed effect models both passed the 1% significance level, indicating that the spatial econometric model could be used to establish the spatial and temporal double fixed effect model. This section examines the influence of the spatial factors influencing the TFAEEs in the areas along the Belt and Road in China. It is important not only to consider the spatial differences between different regions, but also to investigate the dynamic changes in each region during the investigation period. Therefore, it is more reasonable to choose the spatial and temporal double fixed effect model, which has theoretical and practical significance.

Fitness test and regression analysis of the spatial Durbin model

This section is based on the formula (7) and the test results presented in the previous section, combined with the existing research.⁵⁶,⁶³ We established a spatial Durbin model based on spatiotemporal double fixed effects, and then used the Wald test and the LR test to examine whether the spatial Durbin model would degenerate into a spatial lag or spatial error model. The model regression and test results are shown in Tables 8 and 9.

Table 8.

Regression results for the spatiotemporal double fixed effect spatial Durbin model.

Explanatory variables	Static panel_SDM (Adjacency matrix)
Explanatory variables	Coefficients	t-statistic	P-value
ED	0.264 **	2.413	0.016
TI	5.301***	2.671	0.008
ECS	–0.101*	–1.789	0.074
IS	–0.067	–0.501	0.616
ER	–0.675	–1.013	0.311
W*ED	0.118	0.536	0.592
W*TI	–3.788	–1.469	0.142
W*ECS	–0.024	–0.307	0.759
W*IS	–0.189	–0.868	0.385
W*ER	6.137 ***	7.164	0.000
R-squared	0.982
log-likelihood	290.308

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

Table 9.

Wald test and LR test results.

Model	Wald test		LR test
Model	Test value	P-value	Coefficients	P-value
SDM against SAR	78.939***	0.000	62.355***	0.000
SDM against SEM	78.645***	0.000	61.719***	0.000

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

First, Table 9 shows that both the Wald and LR tests were significant at the 1% significance level, showing that the spatial Durbin model of spatiotemporal double fixed effect did not degenerate into a spatial lag model and a spatial error model. Second, the R-squared and log likelihood values of the Wald test were 0.982 and 290.308, respectively, which were much higher than the LR test values of 0.576 and 74.357, indicating that the spatial Durbin model with spatiotemporal double fixed effects was a better fit to the data. The following conclusions can therefore be drawn from the regression results in Table 8.

The coefficient of economic development was positive, with a value of 0.264, and passed the 5% significance level. Economic growth therefore had a positive effect on the improvement in TFAEE in areas along the Belt and Road. The improvement in residents’ income levels in areas along the Belt and Road had a positive effect on their awareness of the need for green environmental protection. This is consistent with the results of Wang et al.; Lv and Deng.^12,52 Furthermore, improvements in economic level could promote prevention and control of air pollution by governments, industry, enterprise, and institutions in the areas along the Belt and Road and also provide an economic argument for promoting the improvement of TFAEE.

The coefficient of technical innovation was positive, with a value of 5.301, and passed the 1% significance level, showing that investment in research and development was also a principal factor improving the TFAEE in areas along the Belt and Road. The regression results are consistent with the conclusion (pure technical efficiency is the main reason restricting the TFAEEs of the area along the line ) in the Decomposition of the TFAEEs of the areas along the belt and road in China section, and also with the research results of Wang et al.⁶⁴ The regression results further proved that the level of technological R&D (such as the promotion of low-carbon green technology, energy conservation, and emission reduction) was a fundamental means of improving the efficiency of the air environment.

The coefficient of energy consumption was negative, with a value of − 0.101, and passed the 10% significance level, proving that coal consumption as a proportion of the total energy consumption hindered the improvement of TFAEE in areas along the Belt and Road. This is consistent with study Wang et al., which is pointed out that coal combustion is the main cause of the deterioration of air environment in China, which is mainly due to the low utilization efficiency of coal itself and the large amount of pollution emissions caused by coal combustion.⁵² It is clearly urgent to optimize energy consumption and improve energy utilization efficiency in the areas along the Belt and Road. The emission of air pollutants can only be controlled at the source by increasing the proportion of clean, renewable energy so as to improve TFAEE in the areas along the Belt and Road in China.

The coefficient of industrial structure was negative, with a value of −0.067. Clearly, an increase in the proportion of secondary industry hindered improvements in the TFAEE, but not significantly. Although the development of secondary industries in the areas along the Belt and Road causes air pollution, the impact was not significant. This is consistent with Lv and Deng, which analyzes the reasons for the inconspicuousness, which may be due to the economic growth brought by the traditional chemical industry and steel industry and the emission reduction effect caused by the industrial structure adjustment.¹² However, continued reasonable development and optimization of the industrial infrastructure in the areas along the Belt and Road cannot be ignored.

The coefficient of environmental regulation was negative and not statistically significant. The regression results showed that investment in environmental pollution control did not play a role in promoting the improvement of TFAEE, This is consistent with the results of Wang et al.⁶⁴ This may be because environmental pollution control efforts in recent years had not been enough to make up for, or effectively improve, the historical environmental debt due to past extensive economic development and existing severe environmental problems. The investment made to remedy environmental pollution control had clearly not been sufficient. The governments of areas along the Belt and Road should take further measures to optimize investment in environmental pollution control and improve the efficiency of environmental management.

Analysis of deconstruction of the total effect results of the spatial Durbin model

This study used the partial differential method proposed by Lesage and Pace,⁶⁵ to deconstruct the total spatial effects demonstrated by the spatial Durbin model, and reveal the direct effect, indirect effect, and total effect of each explanatory variable on TFAEEs. The direct effect represented the changes influencing the current regional explanatory variables on the TFAEEs of the regions; the indirect effect represented the impact of changes in the current regional explanatory variables on the TFAEEs of the adjacent areas (i.e., the spillover effect). The research on spatial spillover effect originated in foreign countries and has been systematically. Hirshman put forward the polarization trickle down theory, which described the interaction and influence among regions with different economic development levels.⁶⁶ On this basis, Richardson et al. called the diffusion (i.e. trickle down) effect between regions as positive spillover effect, and correspondingly, the reflux (i.e. polarization) effect as negative spillover effect.⁶⁷ The regression results are shown in Table 10.

Table 10.

Results of the deconstructed spatial Durbin model effects.

Explanatory variables	Direct effect		Indirect effect		Total effect
Explanatory variables	Coefficients	t-statistic	Coefficients	t-statistic	Coefficients	t-statistic
ED	0.269**	2.44	0.112	0.511	0.381*	1.895
TI	5.258**	2.682	–3.825	–1.545	1.433	0.497
ECS	–0.104*	–1.88	–0.023	–0.292	–0.128	–1.342
IS	–0.065	–0.481	–0.198	–0.936	–0.263	–1.014
ER	–0.698	–1.081	6.171***	6.818	5.473**	5.099

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

Comparing the coefficients of direct effects in Table 10, with the coefficients of explanatory variables in the regression results in Table 8, the positive and negative correlations and significance test results of the coefficients in the two tables were basically consistent. The economic development and technological innovation coefficients were 0.269 and 5.258, respectively, and had significant positive impacts on the TFAEEs of the areas along the Belt and Road. The coefficients of energy consumption structure, industrial structure and environmental regulation were −0.104, −0.065, −0.698, respectively, and had negative correlations with the TFAEEs. This is consistent with the results of spatial Durbin regression in Table 8, which further verifies the impact mechanism of the changes of the current regional explanatory variables on the regional TFAEEs.

According to the coefficient results of the indirect effects in Table 10, the economic development and environmental regulation coefficients of 0.112 and 6.171, respectively, had positive spatial spillover effects, and the spillover effect of environmental regulation was significant. Thanks to the promotion of The Chinese Belt and Road initiative, a stable economic connection has been established between inland and coastal areas, and an “economic corridor” has been constructed, which lays an economic foundation for the treatment of air pollution between regions. In addition, the environmental pollution control along the line has a significant diffusion effect, which promotes the improvement of air environment in adjacent areas. The coefficients of technological innovation, energy consumption structure, and industrial structure were −3.825, −0.023, −0.198, respectively, but showed no significant negative spatial spillover effects. It shows that the areas along the line are still in the polarization (agglomeration) effect of technological innovation, and the driving effect of the developed areas (such as the southeast coastal areas) at the core of technology on the improvement of TFAEEs in other areas (such as the northwest inland areas) is still quite limited. While the agglomeration industry and energy consumption drive the current regional economic development, it brings resistance to the ecological environment governance of adjacent areas,This is consistent with the research of Shao et al., which pointed out that the higher the technical level, the lower the pollutant emission.⁶⁸ However, considering the potential impact of energy rebound effect, the result may not meet the expected reduction of energy consumption, and the pollutant emission may even increase. As Ran and Xu point out that the industrial agglomeration effect also makes the spatial adjacent areas become the first choice and pollution radiation places for inter provincial industrial transfer.⁶⁹ With the continuous expansion of the scale of regional industrial agglomeration, the flow and diffusion of air pollutants discharged by industrial enterprises have brought pressure on the air environment of adjacent areas. Although China’s energy structure has changed to diversification in recent years, but coal combustion is the main source of air pollutants. To sum up, the agglomeration of industry and energy consumption is the main reason for the decline of TFAEEs in adjacent areas. However, the non significant side of the regression results shows that the diffusion and dilution of air pollutants is a macro dynamic process, and the impact on adjacent areas is also different.

Endogenous and robustness test

Endogenous test

Considering that the static panel spatial Durbin model established in this study may have the endogenous problem of explanatory variables, in order to further eliminate the bias of regression results caused by the endogenous problem, this section selects the first-order lag term of TFAEEs as a tool variable to test the endogenous problem that may exist in the model. Based on the formula (8) and using SYS-GMM estimation method can effectively overcome the endogenous problem in regression model. Compared with DIFF-GMM, which is influenced by weak instrumental variables, SYS-GMM is more efficient.⁷⁰

In Table 11, AR (1) of the autocorrelation test results of the SYS-GMM disturbance item passes the 5% significance test, and AR (2) fails the test, indicating that there is a first-order autocorrelation of the disturbance item, but there is no second-order autocorrelation. Sargan test results show that the instrumental variables set in the model are effective, and there is no problem of over identification. L.TFAEEs is the first-order lag term of TFAEEs, and the regression coefficient is 0.259, which has passed the significance test of 10%. It shows that the air environmental efficiency of the current region is largely affected by the previous period’s air environmental efficiency, and the continuity and mobility of air pollutants along the line have been verified. This is consistent with the study of Shao et al.,⁷¹ and he points out that the air pollution in adjacent areas display the pattern being bound together for good or ill. Therefore, in the air pollution control, it needs the coordination and cooperation of the areas along the line to achieve the purpose of better governance.

Table 11.

Results of endogenous test.

Explanatory variables	Dynamic panel_SDM (SYS-GMM)
Explanatory variables	Coefficients	t-statistic	P-value
L.TFAEEs	0.259	1.690	0.095
ED	0.104	1.260	0.210
TI	24.140	2.580	0.011
ECS	–0.055	–0.610	0.544
IS	–0.218	–0.580	0.563
ER	–5.588	–1.070	0.286
W*ED	0.005	0.120	0.901
W*TI	–0.599	–0.180	0.860
W*ECS	0.012	0.290	0.775
W*IS	–0.050	–0.570	0.573
W*ER	–1.074	–0.390	0.700
AR（1）		–2.394	0.017
AR（2）		0.810	0.418
Sargan		13.746	0.132

By compared with the estimation results of the static spatial Durbin model in Table 8, the regression coefficients and significance of other explanatory variables are basically consistent, which verifies the rationality of the model in this study.

Robustness test

In order to verify the robustness of the double fixed effect spatial Durbin model, we replace different spatial weight matrices to test the robustness of the regression results. The spatial weight matrix based on geographical distance and economic distance is adopted, and the calculation results are shown in Table 11. By comparing the regression results of the spatial Dubin model based on the spatial weight of adjacency distance in Table 8, the regression results in Table 12 show that the coefficients and significance of the explanatory variables are basically the same under different spatial weight matrices, which indicates that the main conclusions of this study have good robustness.

Table 12.

Results of spatial Durbin model based on different spatial weight matrices.

Explanatory variables	Geographical distance weight (SDM)			Economic distance weight (SDM)
Explanatory variables	Coefficients	t-statistic	P-value	Coefficients	t-statistic	P-value
ED	0.463***	4.099	0.000	0.526***	4.441	0.000
TI	4.609**	2.092	0.036	3.242	1.336	0.182
ECS	–0.098*	–1.764	0.078	–0.105	–1.390	0.164
IS	–0.265**	–1.974	0.048	–0.380**	–2.108	0.035
ER	0.037	0.057	0.955	0.119	0.156	0.876
W*ED	0.153	0.593	0.553	0.057	0.225	0.822
W*TI	7.741**	1.989	0.047	6.673	1.127	0.260
W*ECS	0.761***	3.729	0.000	–0.097	–0.778	0.437
W*IS	–0.537*	–1.873	0.061	–0.569*	–1.742	0.082
W*ER	8.049***	5.118	0.000	–2.452	–1.247	0.212
R-squared	0.981			0.974
log-likelihood	283.530			263.554

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

Conclusions and policy implications

Conclusions

In this study, the SBM-undesirable model was used to estimate the TFAEEs of 17 provinces along the Belt and Road in China. The global and local Moran indices were used to test the spatial correlations between air quality and different areas along the Belt and Road. Based on the above data and analyses, a spatial Durbin model with spatiotemporal double fixed effects was established to analyze the spatial factors influencing the direct and indirect effects on the TFAEEs in the areas along the Belt and Road in China. The above conclusions still show reliability and robustness after the spatial dynamic panel SYS-GMM estimation and the robustness test of replacing the spatial weight matrix. The main conclusions were as follows:

From 2010 to 2017, the average TFAEEs of the regions along the Belt and Road in China were 0.518, and the potential for air pollution reduction was huge. The TFAEEs of the four regions along the Belt and Road were, in order: southeast 0.78; northeast 0.467; southwest 0.461; and northwest 0.331. The results showed that the TFAEEs were to a great extent closely related to the development of regional economies and geographical location, showing a decreasing trend from the eastern coastal areas to the inland areas.

(2)The index of TFAEEs along the Belt and Road of China were decomposed into PTE0.641 and SE0.808, and the relationship between the decomposition indexes was SE > PTE > TE. By drawing the spatial distribution map of PTE and SE of air environment in the areas along the line, the results showed that the change of TFAEEs in the areas along the line occured under the coupling effect of PTE and SE of air environment, and the main reason for the low level of TFAEEs in the areas along the Belt and Road in China was the low level of PTE.

There were positive spatial autocorrelations and high concentrations of the TFAEE in 17 provinces along the Belt and Road in China. The results of the global Moran index analysis, which decreased from 0.663 in 2010 to 0.376 in 2017, showed a downward trend over time. The regional Moran index analysis showed that the spatial distribution of TFAEEs in the areas along the Belt and Road were clustered into high-high concentration and low-low concentration areas, indicating that that the spatial differentiation of TFAEEs was significant.

The analysis of the spatial factors influencing the TFAEEs in the areas along the Belt and Road in China, showed that increasing economic development and technological innovation had significant positive effects on the TFAEEs, while the energy consumption structure, industrial structure and environmental regulation had a negative effect, although the effects of industrial structure and environmental regulation were not significant.

Deconstructing the total effect of the spatial Durbin model, showed that the positive spatial spillover effect of TFAEE in the areas along the Belt and Road in China was affected by environmental regulation and economic development factors. However, the polarization effect of technology level, energy consumption structure and industrial structure hindered the governance of ecological environment in adjacent areas, and the emission and diffusion of air pollutants in concentrated areas limited the improvement of TFAEEs in surrounding areas.

Policy implications

These findings led to some policy recommendations to improve TFAEE in areas along the Belt and Road in China.

In view of the considerable spatial heterogeneity in the TFAEEs in the areas along the Belt and Road in China, measures should be taken to strengthen information exchange and resource sharing between the different provinces, while at the same time improving regional economies. For example, the areas along the Belt and Road should build an information exchange platform to share information on talent and its circulation, resources and technology, resource waste solutions, and win-win options for sustainable development, so as to plug information gaps and narrow the differences in TFAEEs between regions.

Throughout the development process of the Chinese Belt and Road initiative, although the concept of green development has been put forward for a long time, the TFAEEs of the areas along the line remain at a low level. In order to better promote the construction of ecological civilization and sustainable development strategy, we should not only take green as the background of the development of the areas along the line, but also focus on the improvement of the PTE of the economic development of the areas along the line, Improving technical expertise and scientific and technological innovation is the fundamental means of improving the TFAEEs in areas along the Belt and Road in China. It is necessary to introduce advanced scientific and technological methods of air pollution prevention and control, and to improve current management and technical measures to improve implementation of the green road concept and improve air quality.

In order to optimize the energy consumption structure in the areas along the Belt and Road in China, it is necessary to vigorously develop the natural resource advantages found in the eastern and western areas, to increase their proportion of clean energy consumption, and to further improve their energy efficiency. Governments can promote the reform of energy consumption structures and introduce and promote green and other new technologies focusing on energy conservation, emission reduction and low-carbon technologies, so as to prevent and control air pollution at source. For example, the western and northern areas need to actively promote the construction of “coal to gas” and “coal to electricity” projects, promote central heating in autumn and winter, and reduce the personal use of coal and other high carbon heating fuels. Intensive chemical enterprises in the eastern and southern areas of China should focus on desulfurization and removal of Dust emissions.

In order to promote the sustainable development of green economies in the areas along the Belt and Road in China, it is necessary to improve resource allocation efficiency and optimize industrial structure. In addition, by controlling the emission of SO2, NOx and Dust in key areas along the Belt and Road, reduction of the main pollutant emissions can be enhanced and the TFAEEs in the areas along the Belt and Road can be improved.

Government departments in areas along the Belt and Road in China should adhere to the implementation of the “green Belt and Road” initiative, improve air quality supervision and management, capitalize on local strengths, and formulate scientific and effective air pollution prevention and control measures suitable for local conditions. Financial support should be increased for green and low carbon industries, upgrading and transforming energy consuming and polluting industries, publicity campaigns and education on air pollution prevention and control, enhancing residents’ understanding and support for the need for environmental protection, and the formation of new green initiatives involving cooperation between governments, industrial enterprises, and residents.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Ke-Liang Wang

Ni-Ni Yin: She graduated from the school of economics and management of Anhui University of Science and Technology with a master's degree in management. Her research interest is resource and environmental economics.

Ke-Liang Wang: He is a professor and doctoral supervisor of school of economics, Ocean University of China. He is mainly engaged in environmental efficiency assessment.

Zhen Yao: He graduated from the school of economics and management of Anhui University of Science and Technology with a master's degree in management. His research interest is mainly in low-carbon economics.

Li-Li Ding: She is a professor and doctoral supervisor of school of economics, Ocean University of China. Her main research fileds are marine economics and environmental economics.

Zhuang Miao: He is a professor and doctoral supervisor of west china economic research center of Southwest University of Finance and Economics. His main research interests are environmental economics and air pollution economics.

References

Zhang

SQ.

An evaluation of the economic loss due to the heavy haze during January 2013 in China. China Environ Sci 2013; 33: 2087–2094.

Zheng

Yan

Qian

YH.

Estimation of economic losses of environmental pollution in China in the mid-1990s. Manag World 1999; 2: 189–197 + 207.

Zhao

Pan

, et al. Photochemical smog pollution status and proposal for its monitoring network construction in China. Environ Monit China 2018; 34: 81–87.

Wang

Gao

Zhang

, et al. Review of Chinese atmospheric science research over the past 70 years: atmospheric physics and atmospheric environment. Sci China Earth Sci 2019; 62: 1903–1945.

Zhang

Wang

WJ.

Impact of China-West Asia FTA: an empirical analysis based on the dynamic GTAP. World Econ Stud 2016; 270: 70–78 + 100+136.

Cao

, et al. Does the increase of openness promote regional economic growth? An empirical analysis based on the panel data of countries along the “one Belt and one Road”. Econ Survey 2019; 36: 64–71.

Gao

Yang

Analysis of temporal-spatial changes of urban economic ties on the silk road economic Belt-From the perspective of urban flow intensity. J Lanzhou Univ (Soc Sci) 2015; 43: 9–18.

Ariken

Zhang

Chan

, et al. Coupling coordination analysis and spatio-temporal heterogeneity between urbanization and eco-environment along the silk road economic belt in China. Ecol Indicators 2021; 121: 107014–107025.

Yang

Tian

YY.

An empirical study on agricultural trade between China and countries along the 21st century maritime silk road from the perspective of competition,

complementarity and trade potential. Mod Econ Res 2018; 440: 54–65.

10.

Sun

SQ.

The interactive effect of trade openness and high quality economic development – an empirical study based on China and countries along the “maritime silk road”. J Henan Normal Univ (Philos Soc Sci Ed) 2020; 47: 71–78.

11.

Ding

SS.

Evaluation of atmospheric environment efficiency and regional differences in Zhejiang. J Saf Environ 2019; 19: 1075–1085.

12.

Deng

QW.

Analysis on spatial-temporal difference and influence factors of atmospheric environment efficiency in Shandong province based on Super-SBM model and GML index. Ecol Econ 2020; 36: 193–199.

13.

Freeman

Haveman

Kneese

AV.

1973. The economics of environmental policy. New York, NY: John Wiley and Sons.

14.

World Business Council for Sustainable Development (WBCSD). Measuring eco-efficiency: a guide to reporting company performance. Geneva: World Business Council for Sustainable Development, 2000.

15.

Alsaleh

Abdul

SAR

Othman

, et al. Frontier production function and cost efficiency empirical analysis of bioenergy industry in EU28 region. Soc Sci 2016; 11: 5318–5332.

16.

Alsaleh

Abdul

SAR.

Determinants of cost efficiency of bioenergy industry: evidence from EU28 countries. Renewable Energy 2018; 127: 746–762.

17.

Ibrahim

AT.

The underlying drivers of economy-wide energy efficiency and asymmetric energy price responses. Energy Econ 2021; 98: 105222.

18.

Wei

Shen

MH.

Energy efficiency and energy productivity: a comparison based on the panel data by province. J Quant Tech Econ 2007; 24: 110–121.

19.

Wang

SC.

Total-factor energy efficiency of regions in China. Energy Policy 2006; 34: 3206–3217.

20.

Borozan

Technical and total factor energy efficiency of European regions: a two-stage approach. Energy 2018; 152: 521–532.

21.

Yang

Wei

XX.

The measurement and influences of china's urban total factor energy efficiency under environmental pollution: based on the game cross-efficiency DEA. J Clean Prod 2019; 209: 439–450.

22.

Peng

SY.

Chinamattotal factor energy efficiency evaluation: based on three stage global undesirable-hybrid-slack-based-model. Econ Probl 2020; 496: 11–19.

23.

Song

Wang

Zeng

XQ.

Water resources utilization efficiency and influence factors under environmental restrictions. J Clean Prod 2018; 184: 611–621.

24.

Gidion

Hong

Adams

MZA

, et al. Network DEA models for assessing urban water utility efficiency. Util Policy 2019; 57: 48–58.

25.

Gong

Zhang

XQ.

Spatio-temporal evolution and influencing factors of water resources utilization efficiency of cities along the middle and lower reaches of the yellow river. Sci Geogr Sin 2020; 40: 1930–1939.

26.

Yao

Feng

Zhang

, et al. Measurement and decomposition of industrial green total factor water efficiency in China. J Clean Prod 2018; 198: 1144–1156.

27.

Zhang

Wang

Xue

HF.

China’s industrial green total factor water efficiency under the constraints of environment and resource. China Environ Sci 2020; 40: 5079–5091.

28.

Alsaleh

Abdul

RAS

Mohd

SHO

, et al. An empirical analysis for technical efficiency of bioenergy industry in EU28 region based on data envelopment analysis method. Int J Energy Econ Policy 2016; 6: 290–304.

29.

Liu

Zhang

Zhou

Efficiency of construction land allocation in China: an econometric analysis of panel data. Land Use Policy 2018; 74: 261–272.

30.

Gai

Cheng

Zhu

, et al. Can land rent improve land allocation’s efficiency? Evidence from national fixed point survey. China Econ Q 2020; 20: 321–340.

31.

Deng

Zhao

HR.

Innovation of land allocation system and TFP of industrial enterprises: evaluation based on synthetic control method. Zhejiang Soc Sci 2021; 1: 34–43+156.

32.

Xie

Chen

, et al. Spatial-temporal disparities, saving potential and influential factors of industrial land use efficiency: a case study in urban agglomeration in the middle reaches of the Yangtze river. Land Use Policy 2018; 75: 518–529.

33.

Duro

Lauk

Kastner

, et al. Global inequalities in food consumption, cropland demand and land-use efficiency: a decomposition analysis. Glob Environ Change 2020; 64: 102124.

34.

Moutinho

Madaleno

Macedo

The effect of urban air pollutants in Germany: eco-efficiency analysis through fractional regression models applied after DEA and SFA efficiency predictions. Sustainable Cities Soc 2020; 59: 102204.

35.

Wang

Meng

Yang

, et al. Regional differences and influencing factors of China’s air pollution emission efficiency considering technological heterogeneity. China Popul Resour Environ 2017; 27: 101–110.

36.

Jiang

Pan

YZ.

Studying on improving the operation efficiency of joint prevention mechanism of air pollution. East China Econ Manag 2019; 33: 49–58.

37.

Sun

Wang

HM.

Fiscal decentralization, policy synergy and efficiency of air pollution governance – analysis based on the panel data of urban agglomeration in Beijing-Tianjin-Hebei and the surrounding areas. China Soft Sci 2019; 8: 154–165.

38.

Liu

Zhang

HY.

Study of spatial effects and influencing factors of air pollution from perspective of urban agglomeration. Sci Technol Manag Res 2020; 40: 244–251.

39.

Ding

Zang

SY.

Import trade, environmental effects of technology spillovers and industrial air pollution – analysis of mediating path based on provincial panel data. J Beijing Inst Technol (Soc Sci Ed) 2021; 231–17.

40.

Liu

Guo

SL.

Empirical study on evaluation and determinants of atmospheric environmental efficiency of Tianjin. J Arid Land Resour Environ 2016; 30: 31–35.

41.

Cai

AZ.

Research on the interactive relationships among industrial atmospheric environmental efficiency, factor mobility and economic output. Soft Sci 2019; 33: 47–52.

42.

Wang

Shi

Liu

, et al. Study on the temporal-spatial heterogeneity and influencing factors of air environmental efficiency for Yangtze river economic zone. Resour Environ Yangtze Basin 2018; 27: 453–462.

43.

Wang

Miao

Zhao

, et al. China’s provincial total-factor air pollution emission efficiency evaluation, dynamic evolution and influencing factors. Ecol Indic 2019; 107: 105578–105511.

44.

Foo

Lean

Salim

The impact of china’s one belt one road initiative on international trade in the ASEAN region. North Am J Econ Fin 2020; 54: 101089.

45.

Luo

QX.

The construction of B&R FTA network and its economic effects simulation. Int Econ Trade Res 2020; 36: 4–19.

46.

Baniya

Rocha

Ruta

Trade effects of the new silk road: a gravity analysis. J Dev Econ 2020; 146: 102467.

47.

Guo

Zuo

Research on trading efficiency between China and countries of Eurasian economic union and potentials –based on stochastic frontier gravity model. Inquiry into Economic Issues 2021; 3: 100–110.

48.

Schinas

Westarp

AG.

Assessing the impact of the Maritime silk road. J Ocean Eng Sci 2017; 2: 186–195.

49.

Zhao

Sun

CZ.

Analyzing the spatial-temporal characteristics of the marine economic efficiency of countries along the Maritime silk road and the influencing factors. Ocean Coast Manag 2021; 204: 105517.

50.

Che

Cai

, et al. Spatial statistics analysis of the decoupling level between economic growth and environmental pollution: a case study of the 9 provinces of the silk road economic belt. Ecol Econ 2017; 33: 145–149.

51.

Zhu

Gong

Zhao

, et al. Coupling relationship between urbanization and eco-environment in the core area of the silk road economic belt: a review. Acta Ecol Sin 2019; 39: 5149–5156.

52.

Wang

Liu

Meng

, et al. China’s provincial air environmental efficiency measurement. Stat Decis 2017; 492: 97–101.

53.

Tone

A slacks-based measure of efficiency in data envelopment analysis. Eur J Oper Res 2001; 130: 498–509.

54.

Tone

Dealing with undesirable outputs in DEA: a slacks-based measure (SBM) approach. In: NAPW iii, Toronto, 2004,01.01.

55.

Anselin

Spatial econometrics: methods and models. Dordrecht, The Netherlands: Kluwer Academic Publishers, 1988.

56.

Elhorst

PJ.

Applied spatial econometrics: raising the bar. Spatial Econ Anal 2010; 5: 9–28.

57.

Shan

HJ.

Re-estimating the capital stock of China: 1952–2006. J Quant Tech Econ 2008; 25: 17–31.

58.

Alsaleh

Abdul

RAS

Mohd

SHO.

Determinants of technical efficiency in the bioenergy industry in the EU28 region. Renewable Sustainable Energy Rev 2017; 78: 1331–1349.

59.

Luo

JM.

Do industrial agglomeration and traffic links aggravate the spatial spillover effects of haze? Analysis from the perspective of spatial distribution of industries. Ind Econ Res 2018; 4: 52–64.

60.

Liang

Yang

XG.

Interaction between agglomeration and urban haze pollution. Urban Probl 2017; 9: 83–93.

61.

Ding

Sun

Mei

ZW.

Fiscal decentralization, local government behavior and environment pollution – empirical analysis based on SO2 emissions in 30 provinces. Inquiry Econ Issues 2019; 11: 37–48.

62.

Anselin

Gallo

Jayet

Spatial panel econometrics. The econometrics of panel data. Germany: Springer Berlin Heidelberg, 2008.

63.

Elhorst

JP.

Dynamic spatial panels: models, methods, and inferences. J Geogr Syst 2012; 14: 5–28.

64.

Wang

Liu

Yang

BC.

Regional differences, dynamic trends and influencing mechanism of Beijing-Tianjin-Hebei urban agglomeration’s air environmental efficiency. Areal Res Dev 2019; 38: 135–140.

65.

LeSage

Pace

RK.

Introduction to spatial econometrics. Boca Raton, FL, USA: Taylor & Francis, 2009.

66.

Hirshman

AO.

Economic development strategy. Economic Science Press, 1991, pp.169–172.

67.

Richardson

HW.

Growth pole spillovers: the dynamics of backwash and spread. Reg Stud 2007; 41: S27–S35.

68.

Shao

Cao

, et al. China’s economic policy choices for governing smog pollution based on spatial spillover effects. Econ Res J 2016; 51: 73–88.

69.

Ran

LN.

Environmental regulation, inter-provincial industrial transfer and pollution spillover effect-based on spatial Durbin model and dynamic threshold panel model. East China Econ Manag 2019; 33: 5–13.

70.

Che

Tao

, et al. The impact of income on democracy revisited. J Comp Econ 2013; 41: 159–169.

71.

Shao

Yang

Huang

Theoretical model and experience from China of energy rebound effect. Econ Res J 2013; 48: 96–109.