Logistics Efficiency Evaluation and Empirical Research under the New Retailing Model: The Way toward Sustainable Development

Abstract

As an innovative retailing mode, “new retailing” is making the distinction between physical and virtual commerce obsolete, where offline stores attract customers and supply them with the opportunity of experiences, and online stores provide services for placing orders and transactions. However, many new retailing companies are beset by their logistics efficiency evaluations because evaluation indicators and methods differ significantly from those with traditional business patterns. In this way, considering the sustainable development principle, this study focuses on the approach of logistics efficiency evaluation and empirical analysis under the new retailing model, explores the main factors related to logistics efficiency improvement, and proposes corresponding measures to reinforce the core competence of companies. We first establish a novel logistics efficiency evaluation index system by word frequency and correlation analysis from a sustainable development view, including six first-level indicators and twenty-seven second-level indicators; then, the logistics efficiency evaluation framework is presented: the static evaluation is made using cross-efficiency DEA and IAHP entropy; the dynamic evaluation is performed using the Malmquist index. After that, a demonstrated analysis of logistics efficiency is conducted with 15 listed companies in China. Furthermore, efficiency-influencing factors are identified using Tobit regression, and countermeasures are proposed to promote the development of new retailing. Comparison results show that the proposed logistics efficiency evaluation framework can be scientific and reliable, helping managers to identify the weaknesses and influencing factors in enterprise logistics operations, therefore improving the performance and competitiveness of new retailing companies.

1. Introduction

Under the reform of the supply and demand sides, the traditional e-commerce operation model is being challenged severely by the decrease in online user growth rate [1,2]. Compared to online shopping, offline shopping has incomparable advantages with experience, feeling, and immediacy. At the same time, the combination of online e-commerce and offline retail stores is gradually becoming a new trend with the rapid development of new technologies, promoting the upgrading of the traditional retailing industry [3,4,5]. In detail, the “new retailing” model provides an all-around service for customers using the integration of online and offline based on the most advanced technology, such as big data, mobile Internet, artificial intelligence, blockchain, modern logistics, and so on. In other words, “new retailing” has bridged the difference between physical and virtual commerce, where offline stores attract customers and supply them with the opportunity of experiences; simultaneously, online stores provide services for placing orders and transactions. This can be summarized as the integration of three parts, described as an equation: new retailing = online + offline + logistics, whose essence is the sharing of consumer-focused data involving payment, inventory, delivery, and so on [6,7,8]. As the background of the new retailing, producers, supervisors, and customers propose new demands for logistics: (i) a reasonable and accurate distribution plan; and (ii) a platform for logistics data processing and storage. However, there are still some problems in the logistics in new retailing companies: the high cost of logistics infrastructure; the high proportion of distribution and warehousing in total costs; a low stock with fast-selling goods while excess inventory with slow-selling goods, and so on [9,10]. At present, evaluating and measuring logistics efficiency is an urgent problem that needs to be solved to impress customers, drive customer retention and loyalty, reduce logistics operating costs, and promote market expansion.

Logistics efficiency describes whether or not the logistics system meets customer satisfaction with reasonable cost under the premise of appropriate service quality and comprehensively reflecting the actual operation of the enterprise logistics. As a result, it is important to evaluate the enterprise logistics efficiency and optimize the logistics system from two aspects, namely reducing logistics cost and improving logistics efficiency, helping problem-solving effectively, such as imbalance between input and output, low quality of logistics, and so on. Furthermore, in recent years, it has become an important task of the global logistics industry to promote the sustainable development of modern logistics. To truly achieve sustainable development, logistics managers must consider the economic impact of their sustainability-related decisions on enterprise profits and losses [11]. In essence, logistics efficiency improvement is shown in the background of the conception of sustainable development. Nevertheless, most existing studies related to logistics efficiency mainly involve evaluation from a macroscopical aspect, focusing on comparisons among regions or provinces. A few researchers have put less emphasis on enterprise logistics efficiency, especially with the background of new retailing. Moreover, logistics efficiency studies from sustainable development have rarely been involved in previous research. Under the new retailing model, customers’ logistics demands are becoming increasingly individualized, therefore affecting evaluation indicators and methods. More than that, there are many uncontrollable elements in enterprise logistics efficiency evaluation, and evaluations from experts and personnel in the enterprise, which are only subjective, are not sufficient because of the dynamic changing and updating of the enterprise [12,13].

To sum up, previous studies have not comprehensively presented the logistics efficiency evaluation framework under the new retailing model, and no one has proposed corresponding strategies for improving logistics efficiency from static and dynamic views. Based on the analysis above, considering the sustainable development principle, this study focuses on the logistics efficiency evaluation and empirical analysis in the new retailing era, explores the main factors related to logistics efficiency improvement, and proposes corresponding measures to reinforce core competence for companies. As a result, this paper first proposes a logistics efficiency evaluation index system according to the new retailing mode, including six first-level indicators and twenty-seven second-level indicators. Next, the logistics efficiency evaluation model is constructed using cross-efficiency based on DEA and IAHP entropy, and the optimal efficiency value is calculated using data envelopment analysis [14,15,16]. After that, the total factor productivity (TFP) is evaluated using the Malmquist index [17,18]. The main influence factors and their correlations are identified using Tobit regression. According to the evaluation results, the bottleneck of enterprise logistics efficiency improvement is analyzed and identified for putting forward corresponding suggestions and countermeasures.

The remainder of the paper is arranged as follows. Section 2 reviews the relevant literature on research issues and proposes a research framework. Then, a logistics efficiency evaluation index system by word frequency and correlation analysis from a sustainable development view is established in Section 3. After that, we present the approaches to evaluate the logistics efficiency under the background of new retailing with cross-efficiency DEA, IAHP entropy, and the Malmquist index in Section 4. After that, the logistics efficiencies of 15 Chinese listed companies are computed and analyzed in Section 5. Finally, Section 6 presents the conclusions, limitations, and future research prospects.

2. Literature Review

In this section, we review the relevant literature on research issues and present a research framework, including the logistics efficiency evaluation indicators and their measurement approaches.

2.1. Logistics Efficiency Evaluation Indicators

Logistics efficiency evaluation mainly refers to the specific value of total input and total output, which is used as a measurement of enterprise efficiency and competitiveness. Aiming to more objectively and comprehensively expose the effect of enterprise logistics, much attention has recently been paid to logistics efficiency evaluation indicators from several perspectives. Logistics efficiency evaluation indicator selection is an issue of fundamental importance, and many researchers have presented logistics efficiency evaluation indicators from different points of view. Zhu and Liu proposed evaluation indicators of coal logistics from perspectives internal and external to the enterprise as well as related policies and terms [19]. Besides input cost, nonfinancial indicators were included in the enterprise logistics performance evaluation by Gunasekaranb [20]. Hamdan constructed a warehousing logistics evaluation index system, where working hours, warehouse area, technology, and machinery were used as input indicators, and freight volumes, warehouse utilization, and orders completed were used as output indicators [21]. To evaluate the logistics efficiency of fresh food and the cold chain, Soysal established an indicator system, taking perishability and the cost of fresh agricultural products into account [9]. In addition, 12 indicators, including cost, the consumer complaints rate, the damage rate, and the warehouse area, were selected to evaluate the efficiency of fresh products cold-chain logistics [22]. To further improve the evaluation reliability and effectiveness, Zhao and Ma estimated the efficiency of cold-chain distribution enterprises in reference [23], while the workforce, freezer, and distribution costs were specified as the inputs, with inventory turnover delivery on schedule rate and profit as output. Furthermore, combining existing input/output indicators, the authors in [24] proposed a cold-chain logistics efficiency evaluation system, including investment in fixed assets, the volume of freight transport, and so on.

As an emerging business model, new retailing can be regarded as the integration of online/offline and logistics, promoting precise service for customers. The new retailing business model involves interaction among offline users, suppliers, and consumers on each side of the e-commerce platform. In this way, many enterprises have built warehouses in different cities, and local delivery service exceeds remote delivery in this model [25]. Using big data, new logistics can accurately predict the sales volume and arrange distribution. With the background of the application of e-commerce, the new retailing model is described in Figure 1.

As can be seen from Figure 1, the so-called new retailing model establishes a new consumption mode of life. In contrast to the traditional retailing model, the new retailing can give full play to the advantages of physical and online retail. The latest example is a strategic partnership announced between Alibaba and Bailian Group, in which they will share resources, including offline retail branches, merchandising capability, logistics facilities, and Internet technologies, to develop new retail formats [26].

2.2. Logistics Efficiency Evaluation Methods

At present, stochastic frontier analysis (SFA) [27], data envelopment analysis (DEA) [28], the entropy method, principal component analysis (PCA) [29], the fuzzy comprehensive evaluation method [30], TOPSIS [31], and so on are the most popular approaches of logistics efficiency evaluation, and many achievements have been made using them. As a single output evaluation method, stochastic frontier analysis is frequently used to evaluate logistics efficiency by researchers. For example, Qinmei and Rui computed the logistics efficiency of the core area of the Silk Road Economic Belt using SFA [10]. Similarly, the authors in [32] evaluated the efficiency of logistics delivery to the military. Trujillo and Tovar took European port firms as an object of the research, evaluating their logistics efficiency using SFA [33]. As a non-parametric method with multiple outputs, DEA has been widely used by scholars in the process of efficiency evaluation. Researchers first used it to evaluate logistics efficiency [34]. To better and objectively evaluate efficiency, a series of extended approaches based on DEA have been proposed with many advances. Tan and Guo analyzed the impact of various regulatory environments on operations with a two-step DEA method based on the DEA super-efficiency model [35]. Quintano studied the impacts of the greenhouse effect on the logistics of port firms and put forward corresponding optimization strategies based on the results [36].

However, DEA only exposes the static efficiency of DMU and cannot reveal the dynamic changing of efficiency. Some scholars have filled this gap by combining Malmquist and DEA. The authors in [37] made an empirical analysis of the static logistics efficiency difference and dynamic changing trend using the combination of DEA and the Malmquist model. The authors in [38] measured the operation efficiency of logistics enterprises using the DEA–Malmquist index method, proving that technology plays an important role in the changing of operating efficiency. In reference [39], Deng et al. calculated the logistics efficiency of 16 ports between China and Mongolia using BCC and the Malmquist index. Based on the DEA–Malmquist, Wu et al. analyzed the factors affecting the efficiency using Tobit and evaluated the logistics efficiency of 15 coal enterprises using super-efficiency and the Malmquist model [40]. Zheng and Yang calculated the logistics efficiency of ports in the major coastal cities in a China-based game cross-efficiency model, then analyzed the affecting factors of efficiency using Tobit [41]. Huang evaluated the financing efficiency of logistics firms using DEA and Tobit and proposed relevant measures for efficiency improvement [42]. In addition, some researchers proposed extended methods from different views [43,44,45,46].

2.3. Review

From the existing studies above, related research findings on logistics efficiency provided a rich theoretical and practical basis for this study [47,48,49,50]. With current research, the following conclusions can be drawn from recent achievements.

(i) The research objects of logistics efficiency mainly involve inter-enterprise, inter-industry, and inter-regional comparisons, and the logistics efficiency evaluation indicators focus on cost, profit, workforce, and so on.

(ii) SFA and DEA are the most popular methods in logistics efficiency evaluation, and many approaches based on DEA have been improved and applied to different industries.

(iii) Most studies took cold-chain logistics or port logistics as examples to evaluate logistics efficiency, and the validity and correctness are proved with empirical data.

Nevertheless, there are still some limitations if we apply the current evaluation methods to logistics efficiency from a sustainable development view, as can be summarized as follows.

(i) At present, research findings on logistics efficiency evaluation index systems for new retailing companies are still rare. The evaluation indicators in logistics efficiency with the new retailing model differ significantly from traditional e-commerce environments, where on-time delivery, information technology investment, and so on are more important.

(ii) There are many efficiency evaluation methods, including the analytic hierarchy process (AHP) and the fuzzy comprehensive evaluation (FCE), whose results are greatly influenced by subjective factors. By comparison, the entropy weight method has advantages in weight setting, which are especially suitable for the application of efficiency evaluation.

(iii) The existing research on logistics efficiency is increasingly in-depth and comprehensive; however, current studies are less related to the contents of sustainable development. Science and innovation play an important role in the logistics industry.

Based on the analysis above, this paper aims to propose an evaluation framework for measuring the logistics efficiency of new retailing companies and attempts to establish a scientific and reasonable evaluation index system for the sustainable development of logistics, then calculate the logistics efficiency of 15 listed companies with the new retailing in China. After that, some suggestions and countermeasures based on analysis of the logistics efficiency will be proposed to promote the sustainable development of logistics.

3. Construction of Logistics Efficiency Evaluation Index System under the Background of New Retailing from Sustainable Development View

In this part, we will propose a new logistics efficiency evaluation index system for new retailing companies from the perspective of sustainable development. Also, descriptive statistics of 15 listed companies in China are shown for analysis of logistics efficiency.

3.1. First-Level and Second-Level Indicators of Logistics Efficiency Evaluation

In this study, the collection of candidate logistics efficiency evaluation indicators is carried out through literature analysis and on-site investigation of experienced executives and experts. We first retrieved related literature with “logistics efficiency”, “sustainable development”, and “new retailing”, as keywords in the CNKI database (https://www.cnki.net, accessed on 16 January 2023) and the Web of Science (http://www.webofknowledge.com, accessed on 16 January 2023) over the past five years. The abstracts of the top 200 papers are text-preprocessed using GooSeeker, including word segmentation processing and counting [51]. Then, some frequently appearing evaluation indicators are selected and sorted for the evaluation of logistics efficiency under the background of new retailing. In addition, we further make a useful supplement to the indicators from the perspectives of logistics experts.

Table 1. First-level indicators of logistics efficiency.

First-Level Indicators	Explanation
B1: Delivery [10]	Delivery is the critical factor of the supplier chain, whose core elements involve cost, speed, and so on.
B2: Warehousing [14]	Warehousing is an indispensable component of the modern logistics system, which is an important indicator of service quality and economic benefits.
B3: Operation [39]	Operation is closely related to logistics efficiency, including operational efficiency and the profitability of human resources.
B4: New retailing	The new retailing merges online and offline resources, which closes the gap between them.
B5: Innovation [19]	Innovation reflects the development potential in the near future.
B6: Environmental protection	Environmental protection is an important part of continuous economic development.

presents the first-level indicators and their explanation of logistics efficiency evaluation.

According to the analysis between candidate indicators, the second-level index is established as the logistics efficiency evaluation indicator under sustainable development, as can be seen in

Table 2. Evaluation indicators of logistics efficiency under the new retailing background.

First-Level Indicators	Second-Level Indicators	Type
B1: Delivery	C11: Delivery fee rate [10]	Input
	C12: Cold-chain vehicle rate	Input
	C13: On-time delivery rate [10]	Output
	C14: The average vehicle capacity rate	Output
B2: Warehousing	C21: Warehouse fee rate	Input
	C22: Cooling warehousing rate [14]	Input
	C23: Inventory turnover rate	Output
	C24: Warehouse utilization rate	Output
B3: Operation	C31: Logistics professionals rate [14]	Input
	C32: Logistics cost rate	Input
	C33: Logistics profit margins rate [19]	Output
	C34: Customer satisfaction rate	Output
	C35: Omni-channel market share	Output
B4: New retailing	C41: Offline investment rate	Input
	C42: Online active users rate [19]	Output
	C43: Mobile orders rate	Output
	C44: Turnover per square meter	Output
B5: Innovation	C51: R&D investment rate	Input
	C52: Intellectual property rights (IPR) rate [39]	Input
	C53: Technician rate	Input
	C54: Investment in information technology	Input
	C55: Information-sharing level	Input
B6: Environmental protection	C61: Green package rate	Input
	C62: New energy vehicles rate	Input
	C63: Environmental protection investment rate [39]	Input
	C64: Pollutant discharge rate	Output
	C65: Three-waste disposal rate	Output

, where the third column describes the type of indicator.

It should be pointed out that the proposed evaluation index system involves not only logistics efficiency with the background of new retailing but also logistics efficiency evaluation of its sustainable development. Specifically, first-level indicators in “B4: New retailing” mainly describe the combination of online and offline transactions. “C44: Turnover per square meter” is used to calculate the operational efficiency of the offline stores, whose value is the total sales volume divided by the area of the store. Furthermore, “B5: Innovation” and “B6: Environmental Protection” are used to describe sustainable development progress. In the aspect of innovation, investment in research and design is especially important to its future development, given that science and technology-oriented development will be the future trend. The volume of R&D investment is an economic base for innovation in logistics enterprises. In this study, “C51: R&D investment rate”, “C52: Intellectual property rights (IPR)”, and so on, are selected as their sustainable development capacity. Although logistics is not a high-polluting industry, those with high environmental protection capacity have more development prospects and commercial potential.

In addition, correlation analysis is necessary to ensure candidate indicators’ independence from each other and eliminate repeated semantics. In this way, the correlation coefficient between each pair of second-level indicators is computed using Equation (1).

$$r_{ij}=\frac{{{\displaystyle \sum }}_{k=1}^n\left(Z_{ki}-{\overline{Z}}_i\right)\left(Z_{kj}-{\overline{Z}}_j\right)}{\sqrt{{{\displaystyle \sum }}_{k=1}^n{\left(Z_{ki}-{\overline{Z}}_i\right)}^2{\left(Z_{kj}-{\overline{Z}}_j\right)}^2}}$$

(1)

where r_ij is the correlation coefficient of the ith indicator and the jth indicator, Z_ki is the value of the ith indicator of object i; ${\overline{Z}}_i$ is the average value of the ith indicator. Furthermore, a threshold M (0 < M < 1) is set to exclude those indicators whose absolute correlation value is bigger than M. The calculation results indicate that each pair of indicators is independent of the other. Therefore, the indicators shown in

Table 3. Descriptive statistics of logistics efficiency under sustainable development.

First Level	Second Level	Unit	Average	Standard Deviation	Min	Max
B1	C11	%	1.34	1.16	0.4	4.6
	C12	%	60.9	26.4	10.1	97.0
	C13	%	94.3	3.5	89.2	99.1
	C14	%	82.0	6.6	69.0	92.0
B2	C21	%	0.7	0.4	0.1	1.3
	C22	%	11.0	14.1	0.1	57.1
	C23	%	13.0	5.2	3.9	21.6
	C24	%	92.0	2.7	89.0	97.2
B3	C31	%	9.0	7.3	2.1	29.0
	C32	%	1.6	1.7	0.3	10.0
	C33	%	2.8	2.8	0.1	6.9
	C34	%	92.0	3.3	87.1	97.3
	C35	%	0.6	1.3	0.01	5.1
B4	C41	%	22.0	7.4	12.1	34.8
	C42	%	51.0	63.3	3.1	117.0
	C43	%	76.1	19.2	42.3	96.0
	C44	Ten thousand per square meter	2.8	3.8	1.2	6.8
B5	C51	%	3.4	2.1	0.6	51.1
	C52		3.1	1.5	10.2	1.0
	C53	%	6.3	8.0	0.3	30.1
	C54	%	0.6	1.1	0.1	40.5
	C55		3.1	2.4	2.0	4.1
B6	C61	%	46.1	34.3	24.0	58.0
	C62	%	38.3	27.4	16.3	57.8
	C63	%	36.4	31.5	14.4	60.4
	C64	%	41.8	60.4	35.4	90.8
	C65	%	58.1	48.3	39.6	89.8

are selected as the second-level indicators for logistics efficiency evaluation from the view of sustainable development.

3.2. Data of Decision Units

To analyze the changes in logistics efficiency under the new retailing background in the past 5 years, 15 listed new retailing companies are regarded as samples for comparisons, and Table 3 lists their descriptive statistics of second-level indicators from the view of sustainable development.

4. Methodologies for Logistics Efficiency Evaluation under the New Retailing Background from a Sustainable Development Perspective

In this section, we will present the methodologies for logistics efficiency evaluation with the new retailing background under sustainable development, including the static evaluation method, dynamic evaluation method, and regressive analysis, as can be seen in Figure 2.

4.1. The Optimal Logistics Efficiency Calculation Using Cross-Efficiency DEA

There are many data envelopment analysis (DEA)-based methods, including traditional DEA (BCC, CCR), super-efficiency DEA, the Malmquist index model, cross-efficiency, and so on [52,53,54,55]. In particular, each DMU can use the optimal weight combination in efficiency evaluation when CCR, BCC, and super-efficiency are employed. However, those DEA-based approaches usually overstate the influence of input and output, resulting in 1 as efficiency. The cross-efficiency evaluation method uses self-evaluation and other evaluations when determining the weights, making up for the shortcomings of traditional approaches.

As an improved DEA model, cross-efficiency evaluation was first proposed by Sexton, Silkman, and Hogan in 1986 [56] and was later investigated by Doyle and Green [57,58]. Currently, it has been widely accepted as a discriminative assessment tool for data envelopment analysis (DEA). It determines a unique set of input and output weights for each DMU and then calculates its efficiencies using all the sets of weights. Given m inputs, s outputs, n DMUs, the input-output vector of the kth DMU can be described as (x_k, y_k), where x_k = (x_1k, x_2k, …, x_mk)^T, y_k = (y_1k, y_2k, …, y_mk)^T; input weight vector v = (v₁, v₂, …, v_m)^T; output weight vector u = (u₁, u₂, …, u_s)^T; and the ratio of total input and total output for DMU_k can be calculated using Equation (2).

$$H_{kk}=\frac{{{\displaystyle \sum }}_{r=1}^s{u}_r{y}_{kr}}{{{\displaystyle \sum }}_{i=1}^m{u}_i{x}_{ki}}=\frac{u^T{y}_k}{u^T{x}_k},k=1,2,\dots ,n.$$

(2)

Cross-efficiency matrix H can be described as follows in Equation (3).

$$H=\left[\begin{array}{cccc}{h}_{11}& h_{12}& \dots & h_{1n}\\ h_{21}& h_{22}& \dots & h_{2n}\\ \dots & \dots & \dots & \dots \\ h_{n1}& h_{n3}& \dots & h_{nn}\end{array}\right],$$

(3)

where the element h_kk of the principal diagonal of matrix H is the value of self-evaluation; off-diagonal elements h_kj (k ≠ j) is the value of cross-evaluation.

For each DMU_k, the ratio of the input and output can be calculated using the following Equation (4).

$$\begin{gathered} {\mathit{MaxH}}_{\mathit{kk}}=\frac{u^T{y}_k}{u^T{y}_k}, \\ S.t.\left\{\begin{array}{c}{H}_{kj}=\frac{u^T{y}_j}{v^T{x}_j}\\ v^T{x}_k=1\\ u\ge 0, v\ge 0\end{array}\right. \end{gathered}$$

(4)

Once the optimal weights u_k and v_k are obtained, the self-evaluation values h_kk = u_k^Ty_k. The cross-efficiency matrix H can be calculated according to Equation (4), where h_kk is the optimal value in the traditional DEA model.

4.2. Weights Calculation Using IAHP Entropy

IAHP entropy is the combination of the entropy weight method and interval analytic hierarchy process (IAHP), where the former is used to calculate the index weights objectively, and the latter is used to calculate the index weights subjectively [59,60]. In recent years, the combination of the entropy weight method and IAHP has been widely applied in the efficiency evaluation of enterprises.

The objective weights values can be calculated using the entropy method according to DMU data, where X_ij is the value of the jth index of the ith DMU, P_ij is the index proportion, E_ij is the entropy value, D_ij is the coefficient of variation, W_ij is entropy weight.

(1) Weight calculation using Equation (5).

$$P_{\mathit{ij}}=\frac{X_{ij}}{{{\displaystyle \sum }}_{i=1}^n{X}_{ij}},i=1,2,\dots ,n;j=1,2,\dots ,m.$$

(5)

(2) Entropy values calculation using Equation (6).

$$E_{\mathit{ij}}=-1/\mathit{lnm}{\displaystyle (}{{\displaystyle \sum }}_{j=1}^m{P}_{ij}ln{P}_{ij}{\displaystyle )},I=1,2,\dots ,n;j=1,2,\dots ,m.$$

(6)

(3) Coefficient of variation calculation using Equation (7).

D_ij = 1 − E_ij, I = 1, 2, ..., n; j = 1, 2, ..., m.

(7)

(4) Entropy weight calculation using Equation (8).

$$W_{\mathit{ij}}=D_{\mathit{ij}}/{{\displaystyle \sum }}_{j=1}^m{D}_{ij},I=1,2,\dots ,n;j=1,2,\dots ,m.$$

(8)

The interval analytic hierarchy process is an improved approach by replacing the weights in traditional AHP with fuzzy values, and it uses interval numbers when describing the importance of the objective, overcoming the disadvantage of subjectivity with precise values [61]. The steps of IAHP in this study can be described as follows.

(i) Evaluating each index with the grade and establishing a hierarchy structure model;

(ii) Constructing interval judgment matrix A as in Equation (9)

$$A=\left[\begin{array}{cccc}\left[1,1\right]& \left[b_{12}^{-},b_{12}^{+}\right]& \dots & \left[b_{1\mathrm{j}}^{-},b_{1\mathrm{j}}^{+}\right]\\ \left[b_{12}^{-},b_{12}^{+}\right]& \left[1,1\right]& \dots & \left[b_{2\mathrm{j}}^{-},b_{2\mathrm{j}}^{+}\right]\\ \dots & \dots & \dots & \dots \\ \left[b_{\mathrm{i}1}^{-},b_{\mathrm{i}1}^{+}\right]& \left[b_{\mathrm{i}2}^{-},b_{\mathrm{i}2}^{+}\right]& \dots & \left[1,1\right]\end{array}\right],$$

(9)

where b_ij = [$b_{ij}^{-}$, $b_{ij}^{+}$], and 1/9 ≤ $b_{ij}^{-}\le b_{ij}^{+}$ ≤ 9; $b_{ij}^{-}$ = 1/$b_{ij}^{+}$; A = [A⁻, A⁺], $A^{-}$ = $\left(b_{ij}^{-}\right)$_n×n.

(iii) Calculating interval weights using the following sub-steps:

(a) Calculating the eigenvector that corresponds to the maximum eigenvalue of the judgment matrix;

(b) Calculating the values of k and m using Equations (10) and (11), respectively.

$$k=\sqrt{{{\displaystyle \sum }}_{j=1}^n\frac{1}{{{\displaystyle \sum }}_{j=1}^n{b}_{ij}^{+}}},$$

(10)

$$m=\sqrt{{{\displaystyle \sum }}_{j=1}^n\frac{1}{{{\displaystyle \sum }}_{j=1}^n{\mathrm{b}}_{\mathrm{ij}}^{-}}}.$$

(11)

The judgment matrix A is highly consistent if and only if 0 ≤ k ≤ 1 ≤ m; otherwise, the evaluation scores need to be modified.

(c) Calculating the weight of IAHP according to the values of k, m, and eigenvector.

(iv) Carrying out the consistency verification and determining the weights.

The objective weights of the first-level index are calculated by IAHP; the subject weights of the first-level index are calculated by the entropy weights method; and comprehensive weights can be obtained using Equation (12).

$$B=\frac{a_j{w}_j}{{{\displaystyle \sum }}_{j=1}^n{a}_j{w}_j } j=1,2,3,\dots ,n.$$

(12)

4.3. Dynamic Evaluation of Logistics Efficiency Using the Malmquist Index

Malmquist can reflect the dynamic change in efficiency by observing the changes in productivity between this term and the next term [62]. Under technical conditions of term t, the technical efficiency changing from the term t to t + 1 can be calculated using Equation (13).

$$M^t=\frac{D_C^t\left(x^{t+1},y^{t+1}\right)}{D_C^t\left(x^t,y^t\right)},$$

(13)

Similarly, under technical conditions of the term t + 1, the technical efficiency changing from term t to t + 1 can be calculated using Equation (14).

$$M^{t+1}=\frac{D_C^{t+1}\left(x^{t+1},y^{t+1}\right)}{D_C^{t+1}\left(x^t,y^t\right)},$$

(14)

Productivity changing from the term t to t+1 can be calculated using Equations (13) and (14) as in Equation (15).

$$M(x^{t+1},y^{t+1},{\mathrm{x}}^t,y^t)=\sqrt{M^t\times M^{t+1}}=\sqrt{\frac{D_C^t\left(x^{t+1},y^{t+1}\right)}{D_C^t\left(x^t,y^t\right)}\times \frac{D_C^{t+1}\left(x^{t+1},y^{t+1}\right)}{D_C^{t+1}\left(x^t,y^t\right)}}.$$

(15)

4.4. Efficiency-Influencing Factors Identification Using Tobit

As one of the limited dependent variable models, Tobit can conduct effective regression analysis on limited variables. By comparison, DEA can reflect the efficiency of input-output indicators, but it cannot identify the affecting factors of logistics.

The following prerequisites should be satisfied in the Tobit model, as in Equation (16).

y = βx + ε and y = max(0, y*).

(16)

Also, they can be described as follows in Equation (17), where y* is the variable being explained, x is an explanatory variable, and β is a regression parameter.

y* = βx + ε, ε~N(0, σ²),
y = y*, when y* > 0;
y = 0, when y* ≤ 0.

(17)

In this study, cross-efficiency DEA and IAHP entropy are combined to calculate the comprehensive logistics efficiency; then, the dynamically changing logistics efficiency is analyzed using the Malmquist index. After that, the efficiency-influencing factors are identified using the Tobit regression model for logistics firms.

5. Logistics Efficiency Evaluation and Analysis with the New Retailing Background under Sustainable Development

In this section, we will present the logistics efficiency evaluation and analysis using the data and methods proposed in the previous sections, including its static and dynamic evaluation, respectively. Then, the influencing factors of logistics efficiency are identified, and some suggestions and countermeasures for further conversion will be given.

5.1. Static Evaluation and Analysis of Logistics Efficiency

5.1.1. The Optimal Efficiency Using Cross-Efficiency DEA

Fifteen Chinese logistics listed companies were chosen to calculate their efficiency values in the past five years using MATLAB 2018.

Table 4. The values of logistics efficiency from 2018 to 2022.

DMU	2018	2019	2020	2021	2022
DMU1	0.1464	0.1437	0.1183	0.0979	0.1093
DMU2	0.215	0.177	0.1182	0.1184	0.1747
DMU3	0.1886	0.2296	0.1676	0.1253	0.1013
DMU4	0.1331	0.1128	0.1215	0.1142	0.0977
DMU5	0.1173	0.1084	0.0841	0.0884	0.1325
DMU6	0.1876	0.2112	0.1285	0.1576	0.0941
DMU7	0.1599	0.1293	0.1159	0.0996	0.0985
DMU8	0.2131	0.1771	0.155	0.1473	0.1285
DMU9	0.1046	0.1072	0.0808	0.0868	0.0954
DMU10	0.1718	0.2047	0.0967	0.112	0.1142
DMU11	0.2174	0.1919	0.1287	0.1293	0.1487
DMU12	0.1683	0.1783	0.1038	0.1068	0.0994
DMU13	0.1792	0.1273	0.1151	0.0851	0.076
DMU14	0.2822	0.2141	0.1282	0.0982	0.1118
DMU15	0.1175	0.1109	0.093	0.0881	0.0897

presents the results calculated using cross-efficiency DEA.

Table 5. The logistics efficiency values of the first-level indicator in 2022.

Logistics Efficiency	B1: Delivery	B2: Warehousing	B3: Operation	B4: New Retailing	B5: Innovation	B6: Environmental Protection
DMU1	2.7500	2.8406	3.2786	3.4606	3.5487	2.7754
DMU2	3.7273	4.9983	4.0173	2.6997	3.5215	3.6247
DMU3	3.3070	3.5981	2.8515	3.4243	4.2158	3.6685
DMU4	2.3914	2.5654	2.7139	4.4099	4.0158	3.5547
DMU5	3.2555	2.0380	3.5333	2.6250	3.5417	2.6985
DMU6	4.5656	3.9922	2.6109	3.9481	3.2411	3.5224
DMU7	2.8304	2.6865	2.4150	3.5145	2.9660	3.5214
DMU8	5.5540	3.6753	4.3551	2.8289	2.8982	3.2584
DMU9	2.0379	2.7583	1.5179	2.8109	3.5749	3.6258
DMU10	4.3980	3.9572	2.1477	3.1999	2.9965	3.0547
DMU11	2.5885	2.2699	4.1180	3.5334	3.1149	3.6528
DMU12	3.0429	2.8705	1.7299	3.8197	3.1187	3.2588
DMU13	3.4653	2.9815	2.1848	3.2333	3.5249	2.9574
DMU14	3.9939	3.8648	3.6550	3.0705	3.2471	3.6254
DMU15	4.1730	2.4883	3.0587	2.9279	3.2574	3.6258

presents the results of the first-level index using the cross-efficiency DEA method in 2022.

As can be seen in Table 4 and Table 5, there are significant differences among these logistics companies. In particular, there is a big gap between the efficiency values of delivery and operation. In the first three years, the new retailing is at its beginning in China, and logistics efficiencies of the new retailing companies are actually increasing and have changed little in the next few years. Due to COVID-19, the efficiency values of most companies with the new retailing did not show a sharp decline because community group buying was strongly supported by the government, and a relatively complete and efficient logistics system was soon established.

5.1.2. Weights Calculation Using IAHP Entropy

(1) Objective weight calculation of first-level index using IAHP

We first invited experts, including logistics specialists and staff, to score the importance between the first-level index using an interval, then the interval judgment matrix can be obtained as follows in

Table 6. The interval judgment matrix.

A	B1	B2	B3	B4	B5	B6
B1	[1,1]	[2,3]	[1/3,1/2]	[1,2]	[1/5,1/4]	[1,2]
B2	[1/3,1/2]	[1,1]	[1/5,1/4]	[1/3,1/2]	[2,3]	[1/4,1/3]
B3	[2,3]	[1/4,1/3]	[1,1]	[1/4,1/3]	[1/4,1/3]	[1/4,1/3]
B4	[1/5,1/4]	[1/4,1/3]	[1/6,1/5]	[1,1]	[1/4,1/3]	[3,4]
B5	[3,4]	[2,3]	[1/5,1/4]	[1/5, 1/4]	[1,1]	[1,2]
B6	[1,2]	[2,3]	[3,4]	[4,5]	[1/4,1/3]	[1,1]

.

According to Equation (9), A = [A⁻, A⁺], A⁻ = ($b_{ij}^{-}$)_n*n, A⁺ = ($b_{ij}^{+}$)_n*n.

$$A^{-}={\displaystyle (}{b}_{ij}^{-}{{\displaystyle )}}_{n*n}=\left[\begin{array}{cccccc}1& 2& 1/3& 4& 1/5& 3\\ 1/3& 1& 1/5& 4& 3& 4\\ 2& 4& 1& 1/4& 3& 1/4\\ 1/5& 1/4& 1/6& 1& 1/4& 3\\ 3& 2& 4& 1/5& 1& 4\\ 4& 2& 3& 4& 1/4& 1\end{array}\right]$$

And the λ = 2.1935, eigenvector x⁻ = [0.4469, 0.3445, 0.5055, 0.1877, 0.6063, 0.1522].

Similarly, the matrix A⁺ = ($b_{ij}^{+}$)_n*n can be described as follows.

$$A^{+}={\displaystyle (}{b}_{ij}^{+}{{\displaystyle )}}_{n*n}=\left[\begin{array}{cccccc}1& 3& 1/2& 5& 1/4& 4\\ 1/2& 1& 1/4& 5& 4& 5\\ 3& 5& 1& 1/3& 4& 1/3\\ 1/4& 1/3& 1/5& 1& 1/3& 4\\ 4& 3& 5& 1/4& 1& 5\\ 5& 3& 4& 5& 1/3& 1\end{array}\right]$$

The corresponding values of =2.3337, eigenvector x⁻ = [0.4448, 0.3532, 0.4718, 0. 2262, 0.6167, 0.1520].

Using Equations (10) and (11), the values of k and m can be obtained.

$$\begin{gathered} k=\sqrt{{{\displaystyle \sum }}_{j=1}^n\frac{1}{{{\displaystyle \sum }}_{i=1}^n{a}_{ij}^{+}}}=0.7521; \\ m=\sqrt{{{\displaystyle \sum }}_{j=1}^n\frac{1}{{{\displaystyle \sum }}_{i=1}^n{a}_{ij}^{-}}}=1.0246. \end{gathered}$$

It can be concluded that matrix A is consistent according to the relationship ≤ k ≤ 1 ≤ m.

        k_x− = k(0.4469, 0.3445, 0.5055, 0.1877, 0.6063, 0.1522)^T
=(0.336,0.2591, 0.3802, 0.1412, 0.4559, 0.1145)^T
             m_x+ = m(0.4448, 0.3532, 0.4718, 0. 2262, 0.6167, 0.1520)^T
   =(0.4557, 0.3619, 0.4834, 0.2318, 0.6319, 0.1557)^T.
                                    W = 1/2(k_x− + m_x+) = (0.3959, 0.3105, 0.43179, 0.1865, 0.5439, 0.1351)^T.

After normalization, the weights of the first-level index can be obtained as follows.

w = (0.1976, 0.1550, 0.2155, 0.0931, 0.2715, 0.0674).

(2) Subjective weight calculation of first-level index using entropy weight

According to values of optimal efficiency, subjective weights can be calculated. Furthermore, the value of E_j can be obtained according to Equation (6).

E_j = [1.3862, 1.5418, 1.7968, 1.9610].

Similarly, the matrix D_j can be described as follows using Equation (7).

D_j = −[0.3862, 0.5418, 0.7968, 0.9610].

Finally, the subjective weights are calculated using Equation (8)

W_j = [0.1544, 0.1749, 0.1474, 0.1979, 0.2034, 0.1218].

The comprehensive weights using IAHP entropy can be described in

Table 7. Comprehensive weights.

Weight	B1	B2	B3	B4	B5	B6
Objective	0.1976	0.1550	0.2155	0.0931	0.2715	0.0674
Subjective	0.1544	0.1749	0.1474	0.1979	0.2034	0.1218
Comprehensive	0.1760	0.1649	0.1815	0.1455	0.2375	0.0946

.

5.1.3. The Optimal Logistics Efficiency and Analysis for Fifteen Companies

Based on the comprehensive weights calculated using IAHP and the optimal efficiency calculated using cross-efficiency DEA, the total logistics efficiency values in 2022 can be obtained, as in

Table 8. The logistics efficiency values of the first index in 2022.

Logistics Efficiency	B1	B2	B3	B4	B5	B6	Total
DMU1	0.2056	0.1077	0.2247	0.0604	0.4652	0.9845	0.3071
DMU2	0.4302	0.3401	0.3506	0.4094	0.5874	0.9875	0.4879
DMU3	0.3298	0.2466	0.4558	0.1076	0.4456	0.9874	0.3963
DMU4	0.2248	0.0582	0.9925	0.0418	1.3541	0.7487	0.6278
DMU5	0.3572	0.0957	0.1647	0.0831	1.5748	0.8587	0.5758
DMU6	0.5581	0.2347	0.3967	0.1262	1.6654	0.9854	0.7160
DMU7	0.2825	0.1689	1.2039	0.1026	0.8745	0.2574	0.5430
DMU8	0.7154	0.2632	0.4127	0.1234	1.7415	1.5748	0.8247
DMU9	0.1512	0.1981	0.4498	0.073	2.8741	0.8741	0.9168
DMU10	0.5264	0.2317	0.9347	0.1314	0.5417	2.1455	0.6512
DMU11	0.2388	0.0959	1.1343	0.1606	0.8741	1.5489	0.6412
DMU12	0.3312	0.1886	0.7590	0.1701	1.1647	1.3246	0.6538
DMU13	0.3984	0.2314	0.9530	0.1115	1.3506	1.0967	0.7219
DMU14	0.5310	0.2317	0.8032	0.093	1.4558	1.2039	0.7506
DMU15	0.5470	0.0565	1.4553	0.0814	0.7925	0.5774	0.6244

.

According to Table 8, the following conclusions can be concluded.

(1) There is much difference among these DMUs in the sum of logistics efficiency, and nearly half of them have obtained satisfactory economic efficiency.

(2) “B5: Innovation” is the one with the highest weight of the six first-level indicators, whose comprehensive efficiency varies considerably.

(3) “B4: New retailing” has the characteristics of stable variation, indicating that most logistics companies have begun to realize the importance of sustainable development.

In addition, we have compared the results of logistics efficiency values with other approaches, including CCR [53] and aggressive and benevolent cross-efficiency [55]. The simulations are conducted using software of MATLAB, and the results can be seen in

Table 9. Comparisons of logistics efficiency with different approaches.

Logistics Efficiency	CCR	Benevolent Cross-Efficiency	Aggressive Cross-Efficiency	This Study
DMU1	0.8969	0.9889	0.1231	0.3071
DMU2	0.9969	0.9960	0.1607	0.4879
DMU3	0.9603	0.9888	0.1625	0.3963
DMU4	0.8971	0.9810	0.1159	0.6278
DMU5	0.9900	0.9769	0.1061	0.5758
DMU6	0.9801	0.9815	0.1558	0.7160
DMU7	0.899	0.9782	0.1206	0.5430
DMU8	1.0000	0.9940	0.1642	0.8247
DMU9	0.7687	0.9683	0.0950	0.9168
DMU10	0.9245	0.9851	0.1399	0.6512
DMU11	1.1421	0.9880	0.1632	0.6412
DMU12	0.8821	0.9849	0.1313	0.6538
DMU13	0.8252	0.9777	0.1165	0.7219
DMU14	0.9353	0.9840	0.1669	0.7506
DMU15	1.0000	0.9908	0.0998	0.6244

.

From Table 9, the average logistics efficiency using the benevolent cross-efficiency approach is 0.9843; the second place is 0.9398, calculated by CCR; and the proposed approach ranks third with a value of 0.6292. Furthermore, the logistics efficiency can be distinguished by them. Also, their trends are almost identical, and the results by IAHP entropy are lower than those of CCR and benevolent cross-efficiency approaches. Also, as can be seen, the results computed using IAHP entropy are smaller than the CCR and benevolent cross-efficiency approaches, and they are more objective, real, and reasonable compared to the benevolent and aggressive cross-efficiency, which introduced a process of self-appraisal and a process of peer-appraisal, respectively. The proposed method overcomes the problem of overestimating efficiency values due to unreasonable weight allocation. In the evaluation using benevolent and aggressive cross-efficiency methods, both take the self-interest maximum as the precondition, but each decision unit considers the weight of other DMUs. The entropy weights are first obtained, and then the weights of indicators are calculated, giving equal treatment to each DMU.

Furthermore, the proposed evaluation method exhibits advantages in computing the values of logistics efficiency compared to the traditional approach.

(1) The cross-efficiency method solves the problem that effective decision-making units with values of 1 cannot be sorted, improving the accuracy of parameter identification. Therefore, the results supply much information in logistics efficiency evaluation.

(2) The IAHP entropy method not only presents the subjective preference of business decision-makers but also controls too large values of subjective preference by the entropy weight method.

(3) The main body of logistics efficiency evaluation is the logistics activities in enterprises, which have lots of diversified kinds of forms. The cross-efficiency IAHP method gives enough consideration to relations between DMUs, taking into account the effect of subjective factors and objective factors, so it is suitable to evaluate logistics efficiency under the new retailing mode.

5.2. Dynamic Evaluation and Analysis of Logistics Efficiency

In the previous subsection, logistics efficiency static evaluation is conducted for 15 DMUs. However, static evaluation cannot reveal the changing rules of logistics efficiency. In this way, the Malmquist index is employed to measure the logistics efficiency change using DEA in the past five years. The results can be seen in

Table 10. The values of the Malmquist index from 2017 to 2022.

Year	effch	techch	pech	sech	tfpch
2017–2018	1.1012	0.8317	1.0000	1.1421	0.9123
2018–2019	1.1314	0.8639	1.0000	1.1335	0.9711
2019–2020	0.8755	0.9925	0.9715	0.9096	0.8633
2020–2021	1.0123	1.0000	1.0000	1.0108	1.0124
2021–2022	1.0309	1.0425	1.0396	1.0000	1.0819
Average	1.0303	0.9460	1.0022	1.0392	0.9682

. And comparisons of TFP decomposition results from 2017 to 2022 can be seen in Figure 3.

Table 11. The values of DMUs using the Malmquist index from 2017 to 2022.

DMU	effch	techch	pech	sech	tfpch	Rank
DMU1	1.00	0.96	1.00	1.00	0.96	9
DMU2	0.95	0.80	1.00	0.95	0.76	14
DMU3	1.00	1.01	1.00	1.00	1.01	7
DMU4	1.11	1.06	1.00	1.10	1.17	1
DMU5	1.08	1.00	1.03	1.05	1.08	3
DMU6	0.79	0.80	0.99	0.80	0.63	15
DMU7	1.07	0.89	1.00	1.07	0.95	10
DMU8	1.11	1.05	1.00	1.10	1.15	2
DMU9	0.99	0.92	0.99	1.00	0.91	12
DMU10	0.99	1.00	1.00	0.99	0.99	8
DMU11	0.95	1.13	0.90	1.05	1.07	5
DMU12	1.02	0.90	1.02	1.00	0.92	11
DMU13	0.98	1.10	1.00	0.98	1.08	4
DMU14	0.91	0.89	0.90	1.01	0.81	13
DMU15	1.00	1.06	1.00	1.00	1.06	6

presents the Malmquist index of 15 DMUs from 2017 to 2022, where the average value of total factor productivity changing (tfpch) did not reach 1. In addition, the average value of technical changing (techch) is greater than 1, indicating that the increase of total factor productivity change is subject to the technical changing index to some extent. In technical efficiency, the pure technical efficiency changing (pech) index tends to rise, while the scale efficiency changing (sech) index tends to decline. The decline of the pure technical efficiency changing index drove the technical changing index down to 6%. In this way, the technological efficiency change is mainly influenced by the scale efficiency index. During the period of the second half of the year 2019, the distribution of warehouse and store openings gathered pace with the economic recovery.

5.3. Factor Identification Using the Tobit Regression Model

In this study, the regression analysis is conducted by the panel data model based on the data from statistical yearbooks and their annual reports.

5.3.1. Hausman Test

Before the analysis, it is necessary to examine the relation between the individual effect and the explanatory variable to determine the used model: fixed effects model or random effects model. Furthermore, the Hausman test is used to test and analyze data. According to the result using Eviews 10, the value satisfies the following inequality: p < 0.01, indicating that a null hypothesis should be rejected under the 99% confidence level. The individual effect is not correlated with the explanatory variable with the hypothesis of the Hausman test. Therefore, the fixed effects model is selected because of the rejected hypothesis. The fixed effects regression model can be described as follows in Equation (18).

y_it = μ_i + x_itβ + ε_it, ε_it ~ N(0, σ_t²)

(18)

where x_it is independent variables, μ_i is the individual effect satisfying E(μ_i/x_it) ≠ 0.

5.3.2. Result and Analysis

The fixed effects regression model is employed based on the Hausman test, and the Tobit regression equation can be described as follows in Equation (19).

y = μ_i + x₁β₁ + x₂β₂ + x₃β₃ + x₄β₄ + x₅β₅ + x₆β₆ + x₇β₇ + ε,

(19)

where y is the investment difference; μ_i is a constant; x₁, x₂, x₃, x₄, x₅, x₆, x₇ are indicators such as online retail sales, road mileage, number of chain stores, number of broadband Internet access users, enterprise size, and so on.

Table 12. The Tobit regression results using Eviews10.

Explanatory Variable	Correlation Coefficient	Standard Deviation	Statistic	p (Significance Level)
x1	0.0954	0.0349	2.9900	0.0081
x2	0.0368	0.0116	2.4923	0.0945
x3	−0.0017	0.0001	−1.5712	0.0613
x4	0.0003	0.0002	1.0935	0.0770
x5	−0.0868	0.2741	−1.9835	0.0151
x6	0.0159	0.0310	0.7701	0.0492
Cons	0.2659	0.0219	5.3973	0.0001

presents the Tobit regression results using Eviews10.

Based on the results in Table 12, it can be proved that there is a direct correction between the explanatory variables and logistics investment. Such conclusions can be drawn as follows.

(1) Online retail sales (x₁) and logistics investment are positively correlated with values greater than zero. It can be found that they are at the 1% level significantly, indicating that the increase in online retail sales will result in the redundancy of logistics. On the other hand, the more online retail, the better the development of the retail economy and the higher the motivation of consumers, which motivates new retail companies and increases the investment in logistics.

(2) Regression coefficient of road mileage (x₂) and delivery investment are both positive values under 10% significantly. Nevertheless, they are weak correlative, showing that the increase in roads influences investment of delivery with slight importance.

(3) Government Subsidy (x₃) is negatively correlated with indicators except for the new retailing mode under 10% significantly. However, the absolute value of the correlation coefficient is relatively small, showing a weak correlation between them. The results show that the government should increase subsidies to reduce redundancy of delivery, warehouse, and operation.

(4) The number of Internet access users (x₄) is negatively correlated with logistics investment under 10% significantly. The increase in Internet users signifies the improvement of the network popularization rate, which is weakly correlated with the logistics investment.

(5) The number of chain stores is negatively correlated with the investment of delivery and warehouse under 5%, while positively correlated with the investment of operation and new retailing modes. The increase in chain stores will reduce the redundancy of delivery and warehouse while increasing the redundancy of investment of logistics personnel and offline investment.

(6) Enterprise size (x₆) is positively correlated with input slack variables of delivery, warehouse, and new retailing mode while negatively correlated with the operation, indicating that the bigger logistics companies depend more on sufficient funds during an economic transition period.

5.3.3. Suggestions and Countermeasures on Logistics Efficiency Improvement

According to the results of the static and dynamic evaluation of logistics efficiency, the following suggestions and countermeasures should be taken to improve logistics efficiency under the new retailing model at present [63,64].

(1) Making full use of information technology and opening logistical information-sharing platforms.

Under the new retailing model, the logistics service capability largely depends on the information technology and logistical information platform. On the one hand, information technology can supply high-quality service in logistics; on the other hand, it can promote the operation efficiency of the logistical system, therefore promoting logistics efficiency. In particular, the “last kilometer” is an urgent issue for consumers in the new retailing environment.

(2) Strengthening digitalization in the field of logistics.

Affected by the new retailing model, logistics companies provide convenient services, such as the same- and next-day delivery service, which has such characteristics as high frequencies, short distance, timeliness, and so on. In such a situation, it is important to make full use of big data to schedule routes and time. In particular, the great power of information integration is needed with the new retailing model, tracking goods from entry and out of the warehouse using big data.

(3) Optimizing the purchasing process of logistics.

Based on the optimization of purchasing, the frequency of purchasing and transportation would be dramatically decreased, therefore reducing the cost of vehicles and personnel. Using the analysis and intelligent prediction of big data, purchasing plans can be automatically generated, minimizing the cost of overstocking.

(4) Promoting and developing offline platforms.

Most of the new retailing companies have not paid much attention to offline platforms, which results in an imbalance between online and offline development. According to the analysis using Tobit regression, the number of offline stores is closely related to logistics, and the expansion of offline store size and operation efficiency would promote the improvement of logistics efficiency.

(5) Training and improving the quality of personnel.

According to the analysis results, those with high management ability and quality companies are usually ranked in the top position in logistics efficiency. Instead, experienced personnel in some companies make up a very small portion. It is an urgent problem to be solved that logistics companies should further introduce and train technical talent to improve efficiency.

6. Conclusions and Limitations of the Study

In the era of new retailing, the combination of online and offline stores is becoming increasingly popular, forcing a paradigm shift for companies. On the one hand, the enormous gap that exists between consumers and sellers is disappearing; on the other hand, the online and offline integration process is accelerating around the world. As an indispensable part, logistics plays a crucial role in enhancing the competitiveness of the company, improving profitability, and avoiding risk in the current environment. In this study, the logistics efficiency of 15 Chinese listed companies with the new retailing model is empirically analyzed from the view of sustainable development, and suggestions and countermeasures are proposed to improve the progress of high quality. Specifically, a logistics efficiency evaluation index system is established for new retailing companies, including six first-level indicators and twenty-seven second-level indicators. Then, the static and dynamic logistics efficiency evaluation methodologies are presented, including IAHP entropy and the Malmquist index. After that, the logistics efficiency values of new retailing companies are evaluated using the proposed approaches. Finally, efficiency-influencing factor identification is conducted using the Tobit regression, and countermeasures are proposed to meet the development of the new retail model.

6.1. Theoretical Contributions

For any research to be beneficial and practically applicable, it must justify its relevance and relationship with broader areas of knowledge being considered [65,66,67]. This study makes a significant contribution to the new retailing industry, sustainable development, and logistics while adding to theoretical concepts of efficiency evaluation. The evaluation indicators of logistics efficiency in the new retailing model differ significantly from traditional e-commerce environments, where on-time delivery, information technology investment, and so on are more important. We have extended the previous logistics efficiency evaluation index system to establish a new one, which not only truly reflects the development under the new retailing model currently but also guides the direction of the near future because the sustainable development index is included [10,14,39]. Second, we advance research that the combination of cross-efficiency and IAHP entropy can effectively evaluate logistics efficiency from the static and dynamic perspectives. Most of the current research work has covered the impacts of one aspect, whose results are greatly influenced by subjective factors. By comparison, the entropy weight method has advantages in weight settings that are especially suitable for the application of efficiency evaluation. The proposed evaluation framework can comprehensively and objectively evaluate logistics efficiency for new retailing companies. Third, the Tobit regression model is used to analyze and evaluate the influencing factors, including online sales amount, the number of chain stores, and so on. Little of the existing research provides analysis of influencing factors. In this study, the Tobit regression model is used to identify influencing factors from a sustainability view for new retailing companies.

6.2. Practical Implications

To evaluate the logistics efficiency of the new retailing companies in the past five years, the authors collected data from 15 listed companies located in the east of China. For these companies, logistics efficiency is one of the most important aspects currently because logistics is playing an increasingly important role in improving their management level, enhancing company competitiveness, and thus increasing company profits. What is most desired is a way to evaluate logistics efficiency objectively. Considering the positive aspects, this study could provide useful insights to researchers, policymakers, and managers working in this field.

(1) Fifteen Chinese logistics listed companies were chosen to calculate their logistic efficiency values using MATLAB 2018. Based on the comprehensive weights calculated using IAHP and the optimal efficiency calculated using cross-efficiency DEA, the total logistics efficiency values in 2022 can be obtained.

(2) Static evaluation cannot reveal the changing rule of logistics efficiency. Because of this, the Malmquist index is employed to measure the logistics efficiency change using DEA in the past five years. The total factor productivity has been increasing, indicating that companies should put increased emphasis on technological innovation.

(3) The Tobit regression analysis is conducted by the panel data model based on the data from statistical yearbooks and their annual reports. The results show that there is a direct correction between the explanatory variables and logistics investment.

(4) Corresponding countermeasures and suggestions are given to improve the logistics efficiency for new retailing companies according to analysis results, including the optimization of the purchasing process, the construction and management of information-sharing platforms, and so on.

6.3. Limitations and Future Scope

The study exhibits a few limitations. (1) The first-level logistics efficiency evaluation index for companies with the new retailing model from the view of sustainable development includes three aspects—logistics, new retailing, and sustainable developments—while the management abilities of administrators are not involved because it is not easy to measure. (2) This study focused on the companies in the east of China and was unable to obtain information from the other regions at the more micro level. Companies located in relatively undeveloped regions should be considered in the near future. (3) To make the calculation results more objective and reasonable, it is better to introduce new approaches in logistics efficiency evaluation, such as game theory, approximate ideal solution, and so on. (4) Finally, the findings from the evaluation approach are limited to a single country, and their applicability in a cross-country context must be addressed to generalize the findings. As a result, these are the main areas on which to focus in the future. Further studies could also be undertaken based on other efficiency evaluation issues related to logistics in different countries, especially for the migrant labor force.