Evaluation of Green Logistics Efficiency in Jiangxi Province Based on Three-Stage DEA from the Perspective of High-Quality Development

Abstract

The high-quality development of the logistics industry plays an important role in high-quality economic development. Green logistics is attracting more attention, owing to environmental concern. Based on the five concepts of “innovation, coordination, green, openness and sharing” of high-quality economic development, the input–output indicators of green logistics efficiency (GLE) have been designed. The correlation between the green logistics input and output indicators have been analyzed by Pearson, and the green logistics efficiency of 11 cities in Jiangxi Province has been calculated by three-stage DEA. The evolutionary characteristics of the high-quality development of green logistics efficiency in Jiangxi Province were analyzed and evaluated. The conclusions were achieved as follows: (1) Nanchang, Xinyu and Ganzhou city are at the forefront of efficiency, and are less affected by environmental factors. The logistics efficiency of other cities is obviously affected by environmental factors. (2) The comprehensive technical efficiency (TE) of green logistics in Jiangxi Province is strongly influenced by scale efficiency (SE). The poor scale efficiency reduces the comprehensive technical efficiency of green logistics. (3) There is a positive correlation between high-quality economic development and logistics efficiency. In these cities of Jiangxi province (Nanchang, Ganzhou, Jiujiang, etc.) the share of total economy and green logistics efficiency is high. This study proves that Jiangxi province is moving towards high-quality economic development due to good environmental protection. (4) In 2017, Jiangxi Province was recognized as one of the national ecological civilization pilot region by the State Council. Green logistics efficiency in Jiangxi Province showed an inflection point in 2017, indicating that the green concept is becoming more and more popular. From the government to enterprises, the province has achieved initial results in promoting the transformation of the economic development mode. It aims to provide reference for the high-quality development of other provinces.

1. Introduction

High-quality economic development is the key task of China’s 14th five-year plan. High-quality development is the transformation from extensive management with high pollution and high consumption to green, sustainable and efficient intensive management. Green logistics play an important role in the high-quality development of the logistics industry. The essence of green logistics is to pursue economic benefits without damaging the ecological environment, and realize the “two-wheel drive” of economic and social development and ecological environmental protection. The ultimate goal of green logistics is to achieve economic, social and environmental coordination, and sustainable development. However, there are still more problems in the transformation and development of green logistics in each country. The development mode of the traditional logistics industry is relatively rough and loose. Poor management efficiency and a serious waste of resources have occurred in all countries. These situations, such as unreasonable traffic arrangements and unregulated vehicle carbon emissions, have caused increasingly serious environmental pollution problems. With the development of e-commerce industry in recent years, express waste is becoming an increasingly major new source of pollution in cities. This is a problem that cannot be ignored for the sustainable green development of the logistics industry. Improving the efficiency of green logistics plays an important role in high-quality economic development, and is an important means to solve economic development and environmental pollution [1]. The quality development and green efficiency of the logistics industry is currently a concern for governments, academics and society.

In order to measure the efficiency of green logistics development, Jiangxi Province, a pilot province of ecological civilization, is selected to design input–output indicators to measure the development of green logistics, further explore the relationship between high-quality economic development and green logistics, and propose measures to improve the efficiency of green logistics.

2. Literature Review

The exit research on high-quality development is mostly focusing on economy. Some scholars defined the characteristics of high-quality economy development (Wang Yong Chang) [2]. Some scholars paid attention to the evaluation index system of high-quality economy development. For the evaluation index system of high-quality economic development, the measurement is mainly from the following aspects: first, innovation, coordination, green, open, and sharing are the five development perspectives used to establish an indicator system [3,4]. Second, for the content of high-quality development, all-factor productivity is the main measure of indicators [5]. Third, a high-quality development evaluation index system is established from the point of view of output quality, development efficiency and power transformation [6]. Fourth, the perspectives of human input, capital input, output scale and output quality are used to establish a high-quality evaluation index system [7]. One of the first models for establishing indicators is more common, but the specific indicators selected are more diverse, and the number of indicators varies. In terms of space, measurements are mainly carried out from the following spatial ranges: the regional economy from the economic circle [8,9,10,11]; measurements of high-quality economic development in China’s internal provinces [12]; measurements of the economic regions of high-quality developing countries [13,14]; analysis of the specific indicators of high-quality development from different industry perspectives [15].

Concerning green logistics efficiency (GLE), Paul R. Murphy [16] first proposed green ecology in the logistics industry. Green ecology is the current direction and trend of the development of China’s logistics industry, which includes ecological economy, ecological ethics, sustainable development, and other basic scientific ethics. Since then, research on the combination of ecology and quality development in the logistics industry has gradually increased, with most scholars focusing more on the analysis through the evolution of trends in total factor productivity. About green logistics, Wang D. F. [17] suggested high-quality development variables for including economic, environmental, trade promotion, geography, politics, culture and entry costs. Ni S. [18] believed the development factors of the green logistics system should include politics, economy, society, technology, internal management and environmental protection. The development of coal logistics is faced with the problem of diminishing resources, so the functional aspects of coal logistics, such as transportation, storage, handling and processing, need to be greened. Li A. [19] included the key factors of green coal logistics implementation are internal motivation, internal barriers, government policy system promotion, external support conditions, external related subjects and green logistics benefits. Zhang W [20] pointed out that green logistics should be analyzed from the perspective of economic development, logistics development and ecological environment. Therefore, the implementation of green logistics is beneficial to the integration of the logistics industry and ecological protection. Liu F. [21] used the Malmquist–Luenberger index method to estimate green all-factor productivity in the logistics industry, and the analysis of regional and inter-provincial heterogeneity of productivity. Zhao T. [22] estimated the factors that affect efficiency through regression models. Ren Y. [23] used a traditional data envelopment analysis model and an unintended output relaxation measurement model. In this way, economic efficiency and ecology were evaluated and compared in relation to each other. Wang L. [24] measured energy efficiency by traditional DEA, cross-efficiency DEA and game efficiency DEA methods. Chen J. X. [25] combined DEA and Shannon’s Entropy Model to measure the logistics efficiency of Chinese regions in multiple environments. The analysis of the factors affecting the efficiency of the green logistics industry, combined with the research of domestic and foreign scholars, can be defined as follows: first, the impact of technological development is greater [26,27]; second, the combination of “Internet and Logistics” has reached a low level [28]; third, investment is unreasonable and output efficiency is not up to standard [29]; fourth, the balance between logistics efficiency and ecological protection is not up to standard [30].

The evaluation system of green logistics efficiency is currently established by selecting capital input, labor input and energy input as input indicators [31,32,33,34], and then selecting expected output (comprehensive turnover), non-expected output (carbon emissions) and logistics output. The other method is to construct the evaluation index system of the level of development of logistics industry from the perspective of economic and logistics development (the scale of logistics transportation, the level of economic development, and the logistics of infrastructure and personnel) [35,36]. Or, to explore the country’s level of logistics efficiency from the potential of information technology and innovation [37].

In recent years, researchers have used the Malmquist model to study the efficiency of logistics systems. For example, quantitative studies have been conducted to distinguish changes in the efficient production frontier by using the Malmquist productivity index [38]. Liu F. H. F. [39] used DEA–Malmquist to reveal the pattern of firm productivity change. The results of efficiency analysis using SBM and super-SBM models are compared with those of the CCR model. Klumpp M. [40] used the Malmquist model to study the impact of freight forwarding on sustainability efficiency and argued that the DEA–Malmquist index should be further explored in the future in comparison with other DEA window analyses. In the last year, more and more attention has been paid to the efficiency of logistics systems, not only in the literature but also in practice. The reason is that it helps to improve the economic and social benefits. Andrejić M. [41] used the Malmquist productivity index as an example of a retail chain to help managers in their decision making. The use of hybrid models for studying logistics efficiency will be developed further in future studies. However, limited literatures can be available for the high-quality development of green logistics efficiency.

3. Construction of a Three-Stage DEA Model

Data envelopment analysis (DEA) is a method for evaluating the relative effectiveness of decision-making units (DMU) with multiple inputs and multiple outputs. Fired et al., (2002) proposed the inclusion of environmental impacts and statistical noise in a data envelope-based assessment, and the need to separate these three impacts; hence, the construction of the three-stage DEA Model [42]. As efficiency examines the relationship between total input and total output in the logistics industry, the use of linear programming to measure efficiency does not require the determination of index weight and a uniform index dimension, and correlates multiple indicators with comparability, which can effectively deal with multiple inputs and multiple outputs, and has advantages in measuring efficiency [43,44,45].

It is assumed that the model constructed in this paper satisfies the following assumptions:

(1)

The concept of high-quality development established in this paper also exists in other countries;

(2)

The indicator system established in this paper has the same indicators in other regions;

(3)

All data are available;

(4)

The green logistics efficiency of the selected provinces is affected by the external environment and random perturbations.

As shown in Figure 1, the indicators selected in relation to high-quality development were measured. Secondly, the correlation of input–output evaluation indexes of the green logistics industry were tested to verify whether the selected indexes conformed to the use of DEA model. Finally, the three-stage DEA model was used to calculate the efficiency of green logistics in Jiangxi Province, and the total factor productivity of green logistics in Jiangxi Province was analyzed.

3.1. The First Stage of DEA Efficiency Measurement

Data envelopment analysis (DEA) was originally proposed by Charnes, Cooper and Rhodes (1979) [46] as the first DEA model, the CCR model, which was expressed as an evaluation of the combined efficiency of decision units under the assumption of constant returns to scale. Later, Banker, Charnes and Cooper [47] changed the assumption of constant returns to scale in the CCR model to one of variable returns to scale, which is known as the BCC model. This represents the assessment of the technical efficiency of a decision unit under the assumption of variable returns to scale. In DEA, the relative efficiency of firms are distributed in the (0, 1) interval, with firms at the leading edge of efficiency having an efficiency value of 1. DEA calculates allocative efficiency and technical efficiency, which can be decomposed into scale efficiency and pure technical inefficiency. Each model has both input-oriented and output-oriented forms, and can be set up with constant returns to scale (CRS) and variable returns to scale (VRS). The output-oriented DEA model is set up to maximize the value of output given a certain amount of input factors. Conversely, an input-oriented DEA model is one that minimizes the cost of inputs for a given level of output. The traditional DEA model measures the static relative efficiency of different decision units over the same period, i.e., the change in overall technical efficiency. The Malmquist index model is a dynamic analysis of the efficiency of each decision unit over time, including changes in overall technical efficiency and the lack of technological progress. It can overcome the shortcomings of the BCC model, which is only a static analysis, by measuring the efficiency gains and losses of the subject for each year of the study period to enable a dynamic analysis.

Assuming that there are multiple decision-making units (DMUs) in the BCC model, in the constraint, i = 1, 2, …, n represents the decision unit, $\theta$_i represents the combined efficiency value of the decision unit, and ${\lambda }_i$ represents the weight factor for inputs and outputs. $s^{-}$, $s^{+}$ express slack variables, $s^{-}$ represents the amount of redundancy for the input variable, and $s^{+}$ represents an insufficient amount of output. To increase the validity of judgment, and introduce non-Archimedean $\epsilon \left(\epsilon \in \left(0,\infty \right)\right).$ Include x, which represents the amount of input, and y, which represents output variable.

The BCC model can be described as:

$$\min \left[\theta -\epsilon \left({\displaystyle \sum _{i=1}^m{s}_i{}^{-}}+{\displaystyle \sum _{i=1}^n{s}_i{}^{+}}\right)\right]$$

(1)

$$s.t.\left\{\begin{array}{l}{\displaystyle \sum _{i=1}^n{x}_i{\lambda }_i}+s^{-}=\theta x_0\\ {\displaystyle \sum _{i=1}^n{y}_i{\lambda }_i}-s^{+}=y_0\\ {\lambda }_i\ge 0,i=1,2,\dots ,n,\\ s^{+}\ge 0,s^{-}\ge 0\end{array}\right.$$

(2)

The Malmquist index method is widely used to evaluate changes in input and output productivity. Suppose there are n DMUs to be evaluated, each DMU obtained s outputs with m inputs in t period. $x_j^t={(x_{1j}^t,x_{2j}^t,\dots ,x_{mj}^t)}^T$ represents the input indicator value of the j-th decision unit during the t-period, $y_j^t={(y_{1j}^t,y_{2j}^t,\dots ,y_{mj}^t)}^T$ represents the output indicator value of the j-th decision unit in the t-period, and they are all positive numbers.$t=1,2,\dots ,T$.

The change in productivity from period t to period t + 1 can be expressed as:

$$M(x_{}^{t+1},y_{}^{t+1},x_{}^t,y_{}^t)={\left[\frac{D^t(x_0^{t+1},y_0^{t+1})}{D^t(x_0^t,y_0^t)}\times \frac{D^{t+1}(x_0^{t+1},y_0^{t+1})}{D^{t+1}(x_0^t,y_0^t)}\right]}^{\frac{1}{2}}$$

(3)

$$\frac{D^t(x_0^{t+1},y_0^{t+1})}{D^t(x_0^t,y_0^t)}=Effcch$$

(4)

$${\left[\frac{D^t(x_0^t,y_0^t)}{D^{t+1}(x_0^t,y_0^t)}\times \frac{D^t(x_0^{t+1},y_0^{t+1})}{D^{t+1}(x_0^{t+1},y_0^{t+1})}\right]}^{\frac{1}{2}}=Tech$$

(5)

$$Tfpch=Effch\times Tech=(Pech\times Sech)\times Tech$$

(6)

If Malmquist index > 1, it indicates an increase in efficiency; if Malmquist index < 1, it indicates a decrease in efficiency.

3.2. The Second Stage of the Stochastic Frontier Approach

The production frontier is the maximum set of outputs corresponding to various proportional inputs at a certain level of technology. Stochastic frontier analysis (SFA) is a typical representative of the parametric approach in frontier analysis, which requires determining the specific form of the production frontier. Compared with non-parametric methods, its biggest advantage is that it takes into account the influence of stochastic factors on output. The problem to be solved by SFA is to measure the technical efficiency (TE) of n decision-making units (DMUs) for period T, each of which is m inputs and one output.

The second stage is to use the relaxation variable of the first stage as the explanatory variable. The SFA regression model separates the environment variables from the random noise effects. The relaxation variables of the first stage are broken down into environmental factors, management inefficiency and statistical noise. $c_i$ is the cost of $i$. $y_i$ is output, and $P_{Ki}$ is the price of element K. $u_i$ is an inefficient item, and $v_i$ is the random impact of the cost function.

$$In{c}_i={\beta }_0+{\beta }_{y}In{y}_i+{\displaystyle {\sum }_{K=1}^K{\beta }_{K}In{P}_{Ki}+v_i+u_i}$$

(7)

$$\begin{array}{l}{S}_{ni}=f(Z_i;{\beta }_n)+v_{n_i}+u_{ni}\\ n=1,2,\cdots ,n;i=1,2,\cdots ,i;\\ v~N(0,{\sigma }_v^2),u~N(0,{\sigma }_v^2)\end{array}$$

(8)

$S_{ni}$ is the relaxation value of item n input for the i decision unit. $Z_i$ is an environment variable. ${\beta }_n$ is the coefficient of the environment variable. $v_{n_i}$ and $u_{ni}$ are mixed error terms. $v_{n_i}$ represents random interference. $v~N(0,{\sigma }_v^2)$ is a random error term, and indicates the effect of random interference factors on input relaxation variables. $u_{ni}$ indicates that management is inefficient, and refers to the effect of management factors on input relaxation variables supposing it obeys the normal distribution truncated at zero $u~N(0,{\sigma }_v^2)$.

3.3. A Comparative Analysis of the Efficiency Values of the Third Stage

The purpose of SFA is to eliminate the effects of environmental and random factors on efficiency measurements. This adjusts all decision units to the same external environment.

The adjustment environment is as follows:

$$\begin{array}{l}{X}_{ni}^A=X_{ni}+\left[\max (f(Z_i;{\widehat{\beta }}_n))-f(Z_i;{\widehat{\beta }}_n)\right]+\left[\max (v_{ni}-v_{ni})\right]\\ i=1,2,\cdots ,i;n=1,2,\cdots ,n\end{array}$$

(9)

$X_{ni}^A$ is an adjusted input. $X_{ni}$ is the pre-adjustment input. $\left[\max (f(Z_i;{\widehat{\beta }}_n))-f(Z_i;{\widehat{\beta }}_n)\right]$ is to adjust for external environmental factors. $\left[\max (v_{ni}-v_{ni})\right]$ is to bring all decision units to the same environmental level.

This paper uses the ideas put forward by Jondrow et al. [48] to make estimates. The formula of Roden Yue [49], Chen Wei, etc. [50] are quoted, referring to the previous study of the three-stage DEA separation formula.

The formula for separation management inefficiency and random interference is as follows:

$$E(\mu |\epsilon )={\sigma }_{\ast }\left[\frac{\varphi (\lambda \frac{\epsilon }{\sigma })}{\varphi (\frac{\lambda \epsilon }{\sigma })}+\frac{\lambda \epsilon }{\sigma }\right]$$

(10)

including ${\sigma }_{\ast }=\frac{{\sigma }_{\mu }{\sigma }_{\upsilon }}{\sigma }$, $\sigma =\sqrt{{\sigma }_{\mu }^2+{\sigma }_{\upsilon }^2}$, $\lambda =\raisebox{1ex}{${\sigma }_{\mu }$}\!\left/ \!\raisebox{-1ex}{${\sigma }_{\upsilon }$}\right.$.

4. Design of Green Logistics Efficiency Evaluation Index from the Perspective of High-Quality Development

This paper establishes an indicator system for high-quality economic development, and selects indicators with high relevance and factors from the input–output perspective for the evaluation of green logistics efficiency, so as to derive technical efficiency, scale efficiency and overall efficiency.

4.1. A System of Indicators of High-Quality Economy Development

Starting from the five concepts of “innovation, coordination, green, openness and sharing” of high-quality economic development, this paper combines green logistics to build a green logistics evaluation index system from the perspective of high-quality development.

Table 1. Green logistics measurement index from the perspective of high-quality development.

Characteristics	Secondary Indexes	Unit	Index Attribute
Innovation	R&D expenditure	10,000 Yuan	Negative
	Number of patent applications	parts	Positive
	Number of patents granted	parts	Positive
	Total postal and telecommunications business	Billions (Yuan)	Positive
Coordination	Value added of primary industry/GDP	%	Positive
	Value added of the secondary industry/GDP	%	Positive
	Value added of the tertiary industry	Billions (Yuan)	Positive
	Urban–rural disposable income	Yuan	Negative
	GDP	Billion Yuan	Positive
Green	Energy consumption	Tons of standard coal	Positive
	Wastewater discharge of tertiary industry	10,000 Ton/Year	Negative
	CO2 emission	Kg	Negative
	Comprehensive utilization rate of general industrial solid waste	%	Negative
Openness	Total imports and exports	Billions (dollars)	Positive
	Foreign exchange income from international tourism	Billions (dollars)	Positive
	Fixed asset investment in transportation, warehousing and postal services	10,000 Yuan	Positive
Sharing	Logistics storage land	Square kilometers	Positive
	Number of employees in transportation, storage and postal industries at the end of the year	10,000 people	Positive
	Number of Internet broadband access users	10,000 households	Positive

shows the green logistics measurement index from the perspective of high-quality development.

Innovation and development are the source and driving force of economic development and logistics efficiency. Scientific research investment and the development of the Internet have greatly promoted the efficiency of the logistics industry and promoted economic development. This paper chooses the index of innovation development from the angle of science and technology: R&D expenditure; number of patent applications; number of patents granted, total postal and telecommunications business.

Coordinated development is mainly reflected in the coordinated development of three major industries and economy. Thus, this paper selects the following indicators to measure the level of coordinated development: proportion of added value in primary industry; proportion of added value of secondary industry; the added value of the tertiary industry; urban–rural disposable income and GDP.

Green development is based on the constraints of ecological environmental capacity and resource carrying capacity. It regards environmental protection as a new development model that is an important pillar for achieving sustainable development. This article mainly selects the following indicators: energy consumption; wastewater discharge of the tertiary industry; CO₂ emissions; comprehensive utilization rate of general industrial solid waste. Currently, there are three main ways to deal with undesirable outputs. First, by inverting the undesirable output by linear transformation. Second, by taking undesirable outputs as input variables for correlation analysis. This article selects this method. Third, by analysis based on DEA-ML-supposed directional distance function method.

The carbon emission factors according to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories’ calculation of carbon emissions refer to many literatures.

This paper adopts the following calculation method:

$$C{O}_2={\displaystyle \sum _{i=1}^{7}C{O}_{2i}}={\displaystyle \sum _{i=1}^7{E}_i\times }NC{V}_i\times CE{F}_i\times CO{F}_i\times (\frac{44}{12})$$

(11)

The data comes from the “regional energy balance table” in the energy statistical yearbook from 2013 to 2019, which is converted into standard coal addition to obtain total energy consumption. The coefficients of various energy conversion standards are shown in

Table 2. Estimation of carbon dioxide emission coefficient.

The Type of Energy	NCV (kJ/kg)	CEF (kgC/CJ)	COF	Standard Coal Conversion Coefficient (kgC/kg)
Raw coal	20,908	25.8	1	0.7143
Gasoline	43,070	18.9	1	1.4714
Kerosene	43,070	19.6	1	1.4714
Diesel oil	42,652	20.2	1	1.4571
Fuel oil	41,815	21.2	1	1.4283
Liquefied natural gas	50,171	17.3	1	1.7141
Natural gas	38,930	15.2	1	1.3301

Source: IPCC National Greenhouse Gas Inventory Guidelines 2006.

.

Open development mainly reflects China’s participation in global economic development. This paper selects the following indicators to measure open development: total imports and exports; foreign exchange income from international tourism; fixed asset investment in transportation; warehousing and postal services. Fixed asset investment is the workload and associated costs of fixed asset activities for logistics built and purchased by the whole society. It mainly includes transportation, warehousing and postal logistics as part of the investment, packaging circulation processing and distribution logistics as part of the investment, and wholesale logistics as part of the investment.

Sharing this development reflects the economic achievements of sharing economic development in the whole society, and reflects the basic situation of people’s livelihoods. This paper selects the following indicators to measure the development of sharing: logistics storage land; number of employees in transportation; storage and postal industries at the end of the year; number of Internet broadband access users. The number of employees is related to production scale, labor productivity and labor results, which can directly reflect the development scale of the logistics industry.

4.2. Selection of Input–Output Index of Green Logistics Efficiency

Based on the connotation of logistics efficiency and the sample data of 11 cities in Jiangxi Province, this paper simplifies the index system of high-quality development of green logistics efficiency (Table 1). Correlation analysis and factor analysis of preliminary test indicators ensure the efficiency of green logistics and the principle of high-quality development in the logistics industry. After deleting some relevant indicators, the input–output index system was established, as shown in

Table 3. Input–output Index System of Green Logistics Industry.

Category	Specific Indictors	Variable	Index Explanation
Input	Capital	X1	Transportation, storage and postal fixed assets investment includes construction and installation projects, equipment, tools, equipment purchase and other costs.
	Energy consumption	X2	Energy consumption includes the total consumption of raw coal and crude oil and their products, natural gas and electricity.
	employees	X3	Number of employees in transportation, storage and postal industries at the end of the year.
Output	Demand scale	Y1	The total amount of post and telecommunication business reflects the total achievements of post and telecommunication work in a certain period, reflecting the demand scale of the logistics industry.
	Added value of tertiary industry	Y2	The added value of tertiary industry refers to the growth value of the circulation and service industry in the cycle (usually annual) over the previous liquidation cycle.
Environmental	GDP	Z1	GDP refers to the final results of the production activities of the permanent residence units around the region in a certain period of time.
	R&D expenditure	Z2	R&D expenditure refers to scientific research funds and the cost of scientific research.

.

(I)

Input index. From the perspective of capital input, energy input and labor input, we selected the following as input indicators: fixed asset investment in transportation; warehousing and postal services (X1); energy consumption (X2); number of employees in transportation, storage and postal industries at the end of the year (X3).

(II)

Out index. From the perspective of output scale and output quality, the following were selected as output indicators: total postal and telecommunications business (Y1); and value added to the tertiary industry (Y2).

(III)

Environmental criteria. Environmental indicators refer to factors that have an impact on the efficiency of the logistics industry and are outside the sample subjects. The environmental factors used in the logistics industry generally include the relevant institutions for logistics development, government policies, talents in logistics, infrastructure of the logistics industry, the level of information technology in logistics and the economic level. Using factor analysis and principal component analysis, we screened the indicators from the perspective of the results of regional economic production activities and the ability to innovate in science and technology. The following two indicators were selected as environmental indicators: GDP (Z1) and R&D expenditure (Z2).

5. Empirical Analysis

The calculation results were obtained for the presented input and output data in the timeframe 2013 to 2019, while applying an output-oriented DEA–Malmquist index model with Stochastic frontier analysis, for the analyzed green logistics efficiency in Jiangxi Province.

5.1. Data Sources

This paper used data from 2013–2019 for 11 cities in Jiangxi Province. The relevant data of output variables, input variables and environmental impact variables were derived from the Statistical Yearbook of Jiangxi Province, the Statistical Yearbook of China Transportation, the China Statistical Information Network and the National Economic Development Statistical Bulletin of cities. According to the China Tertiary Industry Statistical Yearbook the logistics industry refers to the transportation, warehousing, and postal industries in the National Economic Industry Classification. According to the statistics of China National Bureau of Statistics, transportation, warehousing and postal services account for about 85% of the added value of the logistics industry, which can fundamentally reflect the development of the logistics industry. Therefore, the logistics industry in this paper refers to transportation, warehousing, and postal services. For energy consumption we selected the logistics industry (transportation, warehousing and postal services) annual total energy consumption as an indicator, and data from the statistical yearbook of Jiangxi Province. In addition, the carbon emission level indicators are from the China Carbon Emission Database.

5.2. Spatial Effciency Analysis

Due to the fact that DEA uses linear programming to measure efficiency, the Pearson correlation coefficient (PPCS) is used to measure the linear correlation between two variables X and Y, with a value between −1 and 1. Therefore, the Pearson correlation coefficient was chosen to test the correlation of the variables in this paper. Conducted analysis of the variable correlation coefficients using SPSS 20.0 software and the results are shown in

Table 4. Pearson correlation coefficient of input and output variables in green logistics efficiency system of Jiangxi province.

Year	Index	Y1	Y2	Year	Index	Y1	Y2
2013	X1	0.910 ***	0.753 ***	2017	X1	0.309 *	0.355 *
	X2	0.543 *	0.661 **		X2	0.720 **	0.780 ***
	X3	0.910 **	0.742 ***		X3	0.645 *	0.643 **
2014	X1	0.912 ***	0.719 **	2018	X1	0.868 ***	0.781 ***
	X2	0.582 *	0.648 **		X2	0.730 **	0.839 ***
	X3	0.844 **	0.722 **		X3	0.768 *	0.687 **
2015	X1	0.856 ***	0.660 **	2019	X1	0.848 ***	0.847 ***
	X2	0.626 *	0.737 ***		X2	0.744 ***	0.808 ***
	X3	0.818 **	0.673 **		X3	0.851 ***	0.727 **
2016	X1	0.722 *	0.628 **
	X2	0.661 *	0.770 ***
	X3	0.760 **	0.705 **

Note: *, **, *** has significant correlation in 10%, 5%, 1%, respectively.

. The figure shows that input and output variable for each year are positively correlated at the 10% significance level, satisfying the requirement of homoscedasticity. The Pearson correlation coefficients of X1 and Y1, Y2 in 2017 were 0.390 and 0.355, respectively, which was a weak correlation. The correlation coefficient between X2 and Y1 in 2013, X2 and Y1 in 2014 was between 0.4 and 0.6, which was in the medium intensity correlation. The correlations of some variables were strong correlations between 0.6 and 0.8. Most of the variable correlations were very strong correlations between 0.8 and 1. Therefore, it can be inferred that the data collected in this paper have the same direction and meet the premise of using three-stage DEA model to measure efficiency.

5.3. Numerical Analysis

5.3.1. Empirical Study of the First Stage BCC Model

In this paper, the BCC model was used in the first stage of DEA to evaluate the green logistics efficiency in terms of integrated technical efficiency (TE), pure technical efficiency (PTE) and scale efficiency (SE). The comprehensive efficiency reflects the maximum output ratio that the regional logistics industry can achieve under the current technical conditions, reflecting the ability to transform resource input into economic benefits. Pure technical efficiency refers to the influence of input factors on comprehensive efficiency under the influence of scale efficiency.

The data of input indicators (X1, X2, X3) and output indicators (Y1, Y2) of 11 cities in Jiangxi Province from 2013–2019 were brought into the BCC model as decision units, using DEAP2.1 software. The results are shown in Figure 2.

The Figure 2 shows that the overall efficiency of green logistics in Jiangxi Province was in incremental efficiency, when environmental factors and random interference factors were excluded. However, the comprehensive technical efficiency and scale efficiency of Jingdezhen, Yingtan, Pingxiang, Yichun and Jian are decreasing. The comprehensive technical efficiency of green logistics efficiency in Jiangxi Province was in the range of 0.881–0.914, which is low overall. The average value of pure technical efficiency shows a fluctuating upward trend, and the average value of scale efficiency shows a fluctuating downward trend. Among them, the fluctuations of technical efficiency and scale efficiency in 2015 and 2016 were larger. However, the main reason for the fluctuation of overall efficiency instead is the increase in pure technical efficiency. In addition, the average value of pure technical efficiency of green logistics in Jiangxi Province was smaller than the scale efficiency for seven consecutive years. This indicates that although the pure technical efficiency is gradually improving, the key reason that restricts the improvement of comprehensive technical efficiency is the low pure technical efficiency.

The change of total factor productivity of green logistics in 11 cities of Jiangxi Province from 2013–2019 can be applied to Deap software. The calculation results are shown in Figure 3. The calculation results include technical efficiency change, technical change index, pure technical efficiency, scale efficiency and total factor productivity. Pure technical efficiency was less than 1, indicating that the utilization of input factors remains to be improved without considering scale efficiency factors. Pure technical efficiency was equal to 1, indicating that without considering the influence of scale efficiency factor, the utilization of input factor has reached the maximum without improvement. Scale efficiency reflects the direct coordination relationship between input and output. When the scale efficiency is equal to 1, it shows that the input and output have reached the optimal state and the profitability is the best. It can be seen from Figure 3 that Effch showed an overall fluctuation trend from 2013 to 2019, and there was a fluctuation increase from 2016 to 2017. The volatility of Techch in 2014–2015 and 2017–2018 indicates that technical efficiency instability is the main cause of technological change. In general, the efficiency of the logistics industry Techch in Jiangxi Province has long suppressed TFP.

5.3.2. Analysis of the Second-Stage SFA Regression Results

Stochastic frontier analysis (SFA) was applied to estimate the effects of environmental factors and random disturbances on green logistics efficiency. Gross regional product (Z1) and R&D internal funding expenditure (Z2) were used as explanatory variables. The values of γ are above 0.9 as seen in

Table 5. SFA regression analysis results of the second stage of green logistics efficiency measurement in Jiangxi Province.

Independent Variable	Year	Constant	Z1	Z2	σ2	γ	Log Likelihood Function	LR Test of the One-Sided Error
X1	2013	−57,074.24	52.04	−0.33	84,923,892.00	0.9999	−133.75	5.4276
	2014	−51,148.01	32.16	−0.11	351,466,110.00	0.9999	−140.22	7.6720
	2015	−36,987.75	13.41	−0.04	559,527,500.00	0.9999	−142.32	8.6230
	2016	−31,474.15	7.99	−0.01	217,073,610.00	0.9999	−137.19	8.4209
	2017	−903,826.26	85.41	0.84	27,186,879,000.00	0.9999	−164.35	7.422
	2018	−69,412.96	−22.02	0.16	616,301,710.00	0.9999	−144.14	5.9971
	2019	−166,724.22	−9.83	0.32	984,244,690.00	0.9999	−145.77	5.3177
X2	2013	−75.80	0.06	0.00	18,265.83	0.9999	−61.39	5.4354
	2014	−68.89	0.05	0.00	23,303.35	0.9999	−62.38	6.8700
	2015	−44.90	0.03	0.00	25,782.08	0.9999	−61.96	8.8061
	2016	−63.15	0.04	0.00	59,356.24	0.9999	−66.48	8.9452
	2017	−68.69	0.02	0.00	51,549.45	0.9999	−66.19	7.9655
	2018	−106.94	0.04	0.00	40,770.65	0.9999	−64.56	8.8656
	2019	−94.92	0.00	0.00	39,385.37	0.9999	−64.62	3.7720
X3	2013	0.02	−0.00	0.00	244.54	0.9999	−36.95	5.435
	2014	−0.02	0.00	0.00	301.28	0.9999	−37.56	8.7403
	2015	−6.05	0.00	0.00	344.32	0.9999	−38.67	7.9069
	2016	−18.62	0.01	0.00	1896.03	0.9999	−48.31	7.4038
	2017	−25.86	0.00	0.00	4670.73	0.9999	−53.89	6.1528
	2018	−18.68	0.01	0.00	675.83	0.9999	−42.28	8.1052
	2019	−12.49	0.00	0.00	462.35	0.9999	−40.40	9.2546

, indicating that the technical inefficiency variance accounts for a large share of the total variance. The influence of management factors on input slack dominates. Gross regional product (Z1) and R&D internal funding expenditure (Z2) have a significant effect on the slackness of inputs. Therefore, it is reasonable to use the frontier model by significance frontier analysis, which does not accept the original hypothesis of an inefficiency term. If the likelihood ratio test of the SFA model rejects the original hypothesis of the existence of the inefficiency term, there is no need to use the SFA regression, and it is sufficient to use the Tobit regression directly.

First, from the regression coefficients of Z1 and Z2 in Table 5, the external variables are regressed on the input slack variables. Therefore, when the regression coefficients are negative, there is a negative correlation between environmental variables and input slack variables, and an increase in environmental variables is beneficial to reduce input slack variables. Conversely, when the regression coefficient is positive, there is a positive relationship between environmental variables and input redundancy. The increase in external environmental variables is accompanied by an increase in input redundancy variables. As shown in Table 5, the coefficient difference between the regional GDP (Z1) and the three input variables was small. The coefficients of R&D internal funding expenditure (Z2) were small, and less significant with the three input variables, but overall, they showed a negative relationship with the input variable slack. The greater the government support, the less the redundancy of inputs, which is conducive to the formation of scale efficiency of green logistics. The more expenditures on science and technology, the more beneficial to reduce the waste of resources. Improving the efficiency of green logistics is beneficial to the development of the whole logistics industry.

The stochastic frontier regression (SFA) analysis shows that the external environment has an impact on the internal efficiency of green logistics in Jiangxi Province. However, the degree and direction of impact are different. Therefore, the green logistics efficiency differs between similar regions at the same time. In the analysis of green logistics efficiency of 11 cities in Jiangxi province, the influence of environmental factors and random interference terms were excluded, and further adjustments were made according to input factors. The green logistics efficiency under the same environment was measured.

5.3.3. The Third-Stage Efficiency and Difference Analysis

The third stage was calculated by stochastic frontier analysis, after separating the effects of the management inefficiency term and the random disturbance term. The input variables (X1, X2, X3) and output variables (Y1, Y2) of green logistics in 11 cities in Jiangxi province from 2013 to 2019 were adjusted, and the efficiency of logistics industry was calculated again by using Deap2.1 and Acrgis10.8, in combination with the original output data. The adjusted values of integrated efficiency, pure technical efficiency, and scale efficiency were, obtained. The results are shown in Figure 4. Comparing the adjusted DEA analysis with the pre-adjusted efficiency analysis after removing the influence of environmental factors and random errors on the efficiency of green logistics in Jiangxi Province, there was a significant change.

The analysis was performed from the perspective of integrated efficiency through the calculation of a three-stage DEA model. After adjusting for calculations, five cities were on the effective frontier surface from 2013 to 2019. The combined efficiency of Nanchang, Yingtan, Xinyu and Ganzhou remained unchanged. The combined efficiency of Jingdezhen was adjusted from DEA invalid to DEA valid, indicating that its green logistics efficiency was underestimated due to environmental factors. Figure 4 shows that there was a significant change in the adjusted average combined efficiency from 2013–2017, with the largest change of 0.643 in 2017. In 2019, the adjusted comprehensive efficiency of green logistics had increased substantially in Jingdezhen, Yingtan, Fuzhou and Yichun, whereas Jiujiang, Shangrao and Ji’an saw a small decline. However, Pingxiang had a substantial decline. In summary, the redundancy of input factors will make the overall efficiency of green logistics decrease, whereas Nanchang, Shangrao, Xinyu and Ganzhou ranked higher and were more dynamic.

The analysis of technical efficiency can reflect the level of development of green logistics technology in the region. Figure 4 shows that the average technical efficiency of Jiangxi province all had an increase before the adjustment and after the DEA adjustment. More than half of the cities were on the effective frontier. This phenomenon suggests that the technical efficiency of some cities is underestimated, subject to environmental and stochastic factors. After DEA adjustment, the technical efficiency of nearly half of the cities (Nanchang, Jiujiang, Jingdezhen, Yingtan, Shangrao, Fuzhou, Xinyu, Ganzhou, and Ji’an) tended to be close to 1. The technical efficiency of green logistics in Pingxiang and Yichun was in a downward trend, indicating that it was more influenced by environmental factors.

The analysis of the scale efficiency of green logistics can be obtained from the fact that five cities were in the effective frontier surface, after the adjustment in 2019. Figure 4 indicates that Nanchang, Xinyu, and Ganzhou remain unchanged, whereas Jingdezhen and Yingtan were adjusted from DEA invalid to DEA valid. In addition, Jiujiang, Shangrao, Pingxiang, Yichun, and Ji’an decreased to 0.953, 0.997, 0.395, 0.744, and 0.925, respectively, and Fuzhou increased to 0.998. In 2013, the pure technical efficiency of Jingdezhen and Yingtan was 1.000. However, the scale efficiency was weak, making the overall efficiency only 0.535 and 0.557, which was at the lower value of 11 local cities. This was due to the weak development of local logistics infrastructure and logistics industry. Between 2013 and 2017, the combined efficiency of Jingdezhen and Yingtan ranged from 0.500 to 0.600. By 2018 and 2019, it reached DEA validity. This illustrates the large influence of environmental and stochastic factors on the scale efficiency of Jiangxi Province, which eventually affects its evaluation of the overall efficiency.

A comparative analysis of the results of the first and third stages shows that the overall mean value of the combined technical efficiency decreased from 0.895 to 0.765, the technical efficiency increased from 0.938 to 0.974, and the scale efficiency decreased from 0.941 to 0.784. This indicates that environmental factors and random disturbances were masking the true efficiency value of green logistics efficiency in Jiangxi Province. In summary, Nanchang, Xinyu and Ganzhou were all on the efficiency frontier surface. The common feature of these three cities is the faster transformation of the economic development mode. Their industrial upgrading is faster, so the contribution of green logistics total factor productivity to the high-quality economic development is higher. The improvement of comprehensive efficiency after DEA adjustment is mainly due to the improvement of scale efficiency, which indicates that the external environment and random errors affect the improvement of technical efficiency. It can be shown that the main reason for the low overall efficiency of green logistics efficiency in Jiangxi Province is the low-scale efficiency. In addition, comparing the average efficiency values of each year, it can be discovered that the magnitude of efficiency changes of cities in Jiangxi Province before and after the adjustment gradually becomes smaller. This shows that the influence of environmental factors and random factors on the green logistics efficiency of Jiangxi Province is gradually weakening.

6. Conclusions and Suggestions

6.1. Conclusions

In the context of a low-carbon environment, based on the relevant data of 11 cities in Jiangxi Province, this paper selects the indicators related to high-quality development and green logistics efficiency, establishes the green logistics efficiency evaluation index through correlation analysis, factor analysis and high selection coefficient, and measures the comprehensive technical efficiency, pure technology efficiency and scale efficiency of 11 cities in Jiangxi Province through the three-stage DEA model.

The research shows that:

(I)

There are great differences in green logistics efficiency and development quality between regional cities. Logistics efficiency is obviously affected by environmental factors, but the effect is different between different cities. Nanchang, Xinyu and Ganzhou city are at the forefront of efficiency, which are less affected by environmental factors. Logistics efficiency of other cities are obviously affected by environmental factors.

(II)

The comprehensive technical efficiency (TE) of green logistics in Jiangxi Province is strongly influenced by scale efficiency (SE). The poor scale efficiency reduces the comprehensive technical efficiency of green logistics.

(III)

There is a positive correlation between high-quality economic development and logistics efficiency. In these cities of Jiangxi province (Nanchang, Ganzhou, Jiujiang, etc.) the share of total economy and green logistics efficiency are high. The study proves that Jiangxi province is moving towards high-quality economic development, due to good environmental protection.

(IV)

In 2017, Jiangxi Province was recognized as a national ecological civilization pilot region by the State Council. Green logistics efficiency in Jiangxi Province showed an inflection point in 2017, indicating that the green concept is becoming more and more popular. From the government to enterprises, the province has achieved initial results in promoting the transformation of the economic development mode. It aims to provide reference for the high-quality development of other provinces.

For future research, we will strive to establish a globally applicable and efficient measurement and evaluation index system for green logistics. The logistics indicator systems of developed and developing countries are considered in the evaluation system to explore the coupling and coordination of green logistics and industrial digitalization in the context of high-quality development. In addition, some representative algorithms can be used to solve such problems, such as cluster lasso model, clustering algorithm, differential evolution algorithm and genetic algorithm. We will try to study them in our future work.

6.2. Suggestion

The development of green logistics is closely related to government actions. All countries with a high efficiency of green logistics have benefited from the active advocacy of the government. The role played by governments in promoting the development of green logistics is mainly manifested in additional investment to promote the development of environmental protection, the organization of forces to supervise the implementation of environmental protection, and the development of special policies and decrees to guide the environmental protection behavior of enterprises. In this paper, we made recommendations based on the results of the study, combined with the actual government-led behavior perspective.

For cities affected by environmental factors, the government can actively guide the development of tertiary industries by improving the areas of technology investment. The government guides the transfer of traditional industries to technology-intensive industries, clarifies technology routes and strengthens technology-balanced investments in logistics. In order to achieve logistics forecasting and standardized management, carbon emission technology management needs to be integrated. Moreover, the complex uncertainties in carbon neutral potential technologies need to be actively addressed.

For the cities in the first line of efficiency, it is necessary to deepen the logistics standardization of logistics enterprises and refine the management. The combination of government promotion and market-oriented operation for standardized management and unified planning and construction of logistics platforms. In addition, we need to support the research and development technology of enterprises and universities, and provide them with policy support and financial support to encourage the introduction of modern information technology. We must also enhance the organization and reliability of the supply chain and reduce the risk of supply chain disruption.

Furthermore, in order to improve the efficiency of regional logistics, scale efficiency needs to be scaled up along with technical efficiency. The government should strengthen the investment in scientific research and scale, according to different regions and economic nature, planning and investment of scale construction. Governments should pay attention to the development orientation of the region, and promote the development of logistics industry according to local characteristics. The government and enterprises should encourage logistics enterprises and R&D centers to develop new technologies for the sake of the coordinated development of the green logistics industry and economy, and accelerate the speed of upgrading technical equipment. In terms of infrastructure, the government should increase the scale of investment in the logistics industry and the construction of logistics infrastructure, and strengthen the financial support for the construction of special lines, multimodal yards, logistics centers, industrial parks and other infrastructure.

Finally, the regional government should strengthen the awareness of low-carbon and environmental protection within the logistics industry in order to achieve the high-quality development of green logistics. Meanwhile some new technologies of energy saving and emission reduction should be utilized into all aspects of the logistics industry. In order to achieve the high-quality development of green logistics, it is nec-essary to integrate energy-saving and emission-reducing technologies. The government and enterprises should work together to explore an economically efficient transformation path to reduce energy consumption. Developed countries should focus on the research of carbon emission control technology and introduce energy-saving and emission-reducing logistics facilities and technologies.