Eco-Efficiency Analysis for the Russian Cities along the Northern Sea Route: A Data Envelopment Analysis Approach Using an Epsilon-Based Measure Model

In this paper, an eco-efficiency analysis is conducted using the epsilon-based measure data envelopment analysis (EBM-DEA) model for Russian cities along the Northern Sea Route (NSR). The EBM-DEA model includes five input variables: population, capital, public investment, water supply, and energy supply and four output variables: gross regional product (GRP), greenhouse gas (GHG) emissions, solid waste, and water pollution. The pattern of eco-efficiency of 28 Russian cities along the NSR is empirically analyzed based on the associated real data across the years from 2010 to 2019. The empirical results obtained from the analysis show that St. Petersburg, Provideniya, Nadym, N. Urengoy, and Noyabrsk are eco-efficient throughout the 10 years. The results also indicate that the cities along the central section of the NSR are generally more eco-efficient than those along other sections, and the cities with higher level of GRPs per capita have relatively higher eco-efficiency with a few exceptions. The study provides deeper insights into the causes of disparity in eco-efficiency, and gives further implications on eco-efficiency improvement strategies. The contributions of this study lie in the fact that new variables are taken into account and new modeling techniques are employed for the assessment of the eco-efficiency of the Russian cities.


Introduction
Russia is one of the most significant stakeholders in the Artic region. The Russian Arctic covers an area of about 3 million m 2 (18% of the Russian territory), including 2.2 million m 2 of land [1]. It has a population of about 2.4 million, about 1.5% of the Russian population, but generates around 10% of Russia's gross domestic product (GDP) [2]. The Russian Arctic contains vast deposits of natural resources, in particular, petroleum [3,4]. The total amount of undiscovered petroleum in the Arctic area has been estimated to be 413 BBOE (billion barrels of oil equivalent), which accounts for about 22% of the world's undiscovered conventional oil and gas resources [5]. In terms of oil and gas, Russia has a larger share than other Arctic countries, with its oil accounting for 41% and its natural gas for about 70% of the total Arctic resources [5]. In addition, large quantities of Russian proven mineral resources are located in the Arctic, such as more than 96% of Russian platinum metals, over 90% of nickel and cobalt, and about 60% of copper [3].
For decades, Russia has been a major producer and exporter of natural resources and its economy growth is driven by the associated exports [6][7][8]. The natural resources in the Russian Arctic can be exported from the ports along the Northern Sea Route (NSR). ment to balance economic development and environmental protection to ensure sustainable development of the Russian Arctic. It is hence of great significance to evaluate and analyze the relationship between the economic development and environmental protection of the Russian Arctic. Eco-efficiency can be appropriate for measuring the relationship between economic development and environmental protection [21,22].
In this study, data envelopment analysis (DEA) is used to evaluate the eco-efficiency of the Russian cities along the NSR (most of them are located in the Russian Arctic), as it has been widely used in the areas regarding evaluation of economic and environmental sustainability [23]. The pattern of eco-efficiency of these cities is empirically analyzed, and the study provides further insights into the analysis results and eco-efficiency improvement strategies.
The major contributions of this study can be summarized in three aspects as follows.
Comprehensive panel data for the period of 2010-2019 (shown in Tables S1-S10 of the Supplementary Material) are collected for the evaluation and analysis of the eco-efficiency of the Russian cities along the NSR. The panel data can establish a sound foundation for further studies.
Compared with previous studies, this study introduces new variables in the evaluation of ecological total-factor energy efficiency (TFEE). It can provide a new perspective for the related stakeholders to recognize, evaluate, and analyze more comprehensively and deeply the eco-efficiency of the Russian cities along the NSR.
The epsilon-based measure data envelopment analysis (EBM-DEA) model is applied to the evaluation and analysis of sustainable development of the Russian cities along the NSR. The EBM-DEA model can effectively solve the problems, which exist in radial and non-radial direction.
The remainder of this paper is organized as follows. Section 2 reviews literatures on the definition, evaluation methods, and applications of eco-efficiency. The EBM-DEA methodology is introduced in Section 3. Section 4 presents the used data, data processing techniques, and statistical analysis of this study. Section 5 provides a presentation of results and associated discussion. Some conclusions are summarized in Section 6.

Definition of Eco-Efficiency
In the traditional sense, eco-efficiency is a term used to describe the quantity of economic benefits per unit of ecological energy [24]. Higher eco-efficiency requires a country to generate more economic output with a lower cost of ecological resources. However, eco-efficiency has been given different meanings [25]. The Business Council for Sustainable Development defined it as being achieved by providing a competitively priced product or service, which satisfies a high standard of living such that negative impact of economic development on the environment is at a tolerable level throughout the life cycle [26]. The Organisation for Economic Cooperation and Development (OECD) defined it as the efficiency with which ecological resources are used to meet human needs [27]. It can be considered as a ratio of an output divided by an input [27]. This definition extends the application of eco-efficiency to governments, industries, and other sectors from the perspective of input and output [27]. Although there are various definitions of eco-efficiency, the overall target of being eco-efficient is to obtain the maximum economic benefit with the minimum cost of environment and ecology.

Eco-Efficiency and DEA
Eco-efficiency is currently a research focus due to its theoretical value and practical significance [28,29]. It has been applied to a wide variety of industrial and regional contexts [25]. There are various methods for eco-efficiency evaluation, which include life cycle analysis [30,31], ecological footprint [32,33], energy analysis [34,35], and ratio method [36]. DEA is also a major method of evaluating eco-efficiency, which takes into account economic benefits and ecological performance. Hailu and Veeman [37] used DEA to analyze the eco-efficiency of the Canadian paper industry, and they proposed a non-parametric analysis method to incorporate undesirable or pollutant output into productivity growth. The DEA method is applied by Wursthorn et al. [38] to perform the analysis of environment-economic trade-off and eco-efficiency of industrial processes. Wang et al. [39] took Xinfa eco-industrial parks as a case study and developed a matrix network of DEA model to evaluate ecological industry chain efficiency, which takes into account energy, economic, and environmental constraints.
Many DEA studies have also focused on regional or international eco-efficiency. Zhou et al. [40] proposed two slack-based efficiency measures for modeling of environmental performance of 30 OECD countries, and four variables, i.e., energy supply, population, GDP, and carbon dioxide (CO 2 ) emission are included in the associated model. Li and Hu [41] constructed slacks-based measure data envelopment analysis (SBM-DEA) models to calculate the ecological TFEE of 30 regions in China from 2005 to 2009. These models take total energy consumption, total capital stock, and total labor force as inputs and GDP, CO 2 , and sulfur dioxide (SO 2 ) as outputs. Zhang et al. [42] used the SBM-DEA models to calculate the ecological TFEE of 30 provinces of China. Based on CO 2 and SO 2 emissions and chemical oxygen demand (COD) in China from 2001 to 2010, they carried out an empirical analysis of regional ecological energy efficiency. Based on the SBM-DEA models, Choi et al. [43] analyzed the efficiency of CO 2 emission and energy, potential CO 2 emission reduction, and marginal cost of CO 2 emission in 30 provinces of China from 2001 to 2010. Li et al. used DEA to study the eco-efficiencies of China at provincial levels and the associated driving factors [44]. Lorenzo-Toja et al. [45] extensively analyzed 113 wastewater treatment plants across Spain using the methodology that combines life cycle assessment (LCA) and DEA to determine the operational efficiency of each plant in order to obtain environmental benchmarks for inefficient plants. Halkos and Petrou [29] studied the eco-efficiency of the 28 EU countries in 2008, 2010, 2012, and 2014, which uses DEA and directional distance functions to deal with undesired outcomes.
It is evident from the above studies that DEA has been widely used to evaluate ecoefficiency. In addition, DEA has been proven to be a useful and valuable tool for decision makers. Hence, DEA was used in this study to evaluate the eco-efficiency of environmental governance and economic development of Russian cities along the NSR.

DEA Models
Compared with other methods, DEA has the following advantages. (1) It does not require the estimation of the production function in advance [43]. (2) It gives objective weights to different environmental factors based on data and does not depend on human judgment [39]. (3) It can describe the effective production frontier and provide a benchmark for the efficiency improvement of invalid decision-making units (DMUs) [46]. (4) It explains multi-input and multi-output systems for efficiency measurement [47]. Based on these properties, non-parametric frontier analysis represented by DEA has been widely applied to efficiency measurement due to its unique flexibility and applicability [48].
The DEA model mainly evaluates the relative efficiency of DMUs. It generates the efficiency by analyzing the frontier of input and output variables. It has several variants, e.g., the Charnes-Cooper-Rhodes (CCR) model, the Banker-Charnes-Cooper (BCC) model, the SBM model, and the EBM model. The CCR model proposed by Charnes et al. [47] assumes that the return on scale is constant, but there is rarely a constant return on scale in the real world. Banker et al. [49] proposed the BCC model by extending the CCR model. The BCC model assumes a scale return for variables. The BCC model accepts scale return in constant or decreasing marginal productivity. The CCR and BCC are radial models when all of their input and output variables change in the same proportion. The conventional models do not consider the non-radial slacks, so the results ignore some inefficiency impacts. The SBM model proposed by Tone [50] is a non-radial model. It adds slacks into the objective function, which deals with the problem of undesired output. Compared with the CCR and BCC models, the SBM model averts the deviation and influence caused by radial and angle differences, which can better reflect the essence of efficiency evaluation. Although the SBM model can describe all slacks information, it ignores the overall proportional changes of variables. Another issue comes from the SBM model's property, i.e., linear programming, where the optimal loose case presents a strong contrast between positive and zero values [51]. This leads to an underestimation that is inconsistent with the actual situation.
The CCR and BCC model are both radial DEA models, where non-radial relaxation variables are ignored. Moreover, the SBM model fails to consider the characteristics of the radial model. Tone and Tsutsui [51] proposed the EBM model, which integrates radial and non-radial features in a unified framework. It reflects the difference between the optimal observed value and the real value. In addition, the EBM model takes slacks into consideration to reflect the difference between the non-radial parts of inputs and outputs. The results of the EBM model consider the framework of both the CCR and SBM model, and it can thereby calculate the efficiency of DMU more accurately.

Methodology
Suppose that there are n DMUs in this study. Each DMU denoted by DMU j (j = 1, . . . , n) has m inputs (i = 1, . . . , m) and s outputs (r = 1, . . . , s). The input and output of DMU j are denoted by X = x ij ∈ R m×n and Y = y ij ∈ R s×n , respectively. It is assumed that X > 0 and Y > 0. Based on the terminology introduced above, the CCR, SBM, and EBM model are briefly introduced in the following part of this section.

CCR Model
Under the assumption of constant returns to scale, the input-oriented CCR model is used to evaluate the technical efficiency θ * of DMU based on the following linear optimization program.
In Equation (1), the range of θ is 0 ≤ θ ≤ 1. s − and s + represent the non-radial slacks of each input and output of DMUs, respectively. λ indicates the intensity vector.

SBM Model
Under the assumption of constant returns to scale, the input-oriented SBM model proposed by Tone [50] can be used to evaluate the efficiency τ * of DMU using the linear optimization program shown in Equation (2).

EBM Model
The EBM model proposed by Tone and Tsutsui [51] has both radial and non-radial features in a unified framework. The objective function of EBM is shown as follows: In Equation (3), γ * is the optimal efficiency score of EBM model. θ is the radial efficiency value calculated by the CCR model. λ represents the weight vector.
should be provided prior to efficiency measurements. ε x combines the radial θ and nonradial slacks. The properties and related definitions of EBM model are shown as follows.

Definition 2. (EBM projection). Let the optimal solution to Equations
Tone and Tsutsui (2010) defined the projection of DMU (x 0 , y 0 ) as follows:

Data Preparation and Model Framework
EBM models are used for the eco-efficiency analysis of the 28 Russian cities along the NSR in this study. The variables used in the models include population, capital, public investment, water supply, energy supply, gross regional product (GRP), GHG emissions, solid waste, and water pollution. The first five of the abovementioned variables are used as the inputs of the EBM models, and the rest are taken as the associated outputs. Figure 1 exhibits such a model framework. These variables are measured in the units displayed in Table 1. Annual data on these variables for 28 Russian cities along the NSR in the period from 2010 to 2019 are obtained from the Russian Federal Statistics Service. These data are used as inputs and outputs of the models constructed in this study. The descriptive statistics of these inputs and outputs for all the years are exhibited in Table 2.
as the inputs of the EBM models, and the rest are taken as the associated outputs. Figure  1 exhibits such a model framework. These variables are measured in the units displayed in Table 1. Annual data on these variables for 28 Russian cities along the NSR in the period from 2010 to 2019 are obtained from the Russian Federal Statistics Service. These data are used as inputs and outputs of the models constructed in this study. The descriptive statistics of these inputs and outputs for all the years are exhibited in Table 2.

Results and Discussion
The eco-efficiency scores of the cities along the NSR are generated based on the EBM model, which are exhibited in Table 3. As shown in Table 3, St. Petersburg, Provideniya, Nadym, N. Urengoy, and Noyabrsk are eco-efficient across all the years from 2010 to 2019. Onega is moderately eco-efficient, and the eco-efficiency scores range between 0.345 and 0.420. The other cities in Table 3 have relatively low levels of eco-efficiency. From the perspective of the temporal pattern of eco-efficiency scores (those of the cities along the western, central, and eastern section of the NSR are displayed in Figures 2-4, respectively), it is evident that the scores of most cities remain stable or fluctuate slightly around a certain level. The exceptions are Naryan-Mar, Novodvinsk, and Salekhard, which all experience steady rise and sharp decline.       Regarding the spatial distribution of the eco-efficiency of these cities (shown in Figure 5), it can be seen that the cities along the central section of the NSR are generally more eco-efficient than those along the eastern and western section. This regional disparity in eco-efficiency exhibits a similar pattern to the GRPs per capita of the cities. As can be seen from Figure 6, the cities with higher GRPs per capita also have higher eco-efficiency scores with the exception of St. Petersburg and Provideniya (denoted by "1" and "8", respectively). A possible explanation for the association is based on natural resources, populations, and policies. The neighboring areas of the central section of the NSR containing rich reserve of natural resources such as oil and gas can generate relatively high GRPs, and the areas have small populations, which leads to comparatively high GRPs per capita. In addition, the Russian economy relies heavily on the export of natural resources, and the Russian government provides a strong support for the development of the energy industry. This all indicates that the authorities and related stakeholders can invest more resources in the fields such as capital, technology, and management to the cities along the eastern and western section, so as to make these cities use resources more efficiently and discharge fewer pollutants. Regarding the spatial distribution of the eco-efficiency of these cities (shown in Figure 5), it can be seen that the cities along the central section of the NSR are generally more eco-efficient than those along the eastern and western section. This regional disparity in eco-efficiency exhibits a similar pattern to the GRPs per capita of the cities. As can be seen from Figure 6, the cities with higher GRPs per capita also have higher eco-efficiency scores with the exception of St. Petersburg and Provideniya (denoted by "1" and "8", respectively). A possible explanation for the association is based on natural resources, populations, and policies. The neighboring areas of the central section of the NSR containing rich reserve of natural resources such as oil and gas can generate relatively high GRPs, and the areas have small populations, which leads to comparatively high GRPs per capita. In addition, the Russian economy relies heavily on the export of natural resources, and the Russian government provides a strong support for the development of the energy industry. This all indicates that the authorities and related stakeholders can invest more resources in the fields such as capital, technology, and management to the cities along the eastern and western section, so as to make these cities use resources more efficiently and discharge fewer pollutants. Figure 5. Spatial distribution of eco-efficiency of the cities along the NSR. Note: the dark green, green, yellow, and orange markers in Figure 5 indicate that the associated cities are fully, highly, averagely, and poorly eco-efficient, respectively.  Figure 6 are the numbers in the "No." column of Table 3, which represent the respective cities.
To look at how the eco-efficiency of the cities might be improved, it is necessary to generate the results of inputs and outputs optimization. The average annual percentage changes in the inputs and outputs for these cities are shown in Table 4. The "S-", "S+", and "SB" in Table 4 indicate the excesses of inputs, shortfalls of positive outputs, and excesses of negative outputs, respectively, according to the efficiency (optimal solution) achieved by the EBM-DEA models. It is clear from Table 4 that 18 out of 28 cities need to have a more than 50% reduction in both capital and public investment to achieve ecoefficiency. This leads to the inefficiency in capital and public investment of many Russian Figure 5. Spatial distribution of eco-efficiency of the cities along the NSR. Note: the dark green, green, yellow, and orange markers in Figure 5 indicate that the associated cities are fully, highly, averagely, and poorly eco-efficient, respectively.  Figure 5. Spatial distribution of eco-efficiency of the cities along the NSR. Note: the dark green, green, yellow, and orange markers in Figure 5 indicate that the associated cities are fully, highly, averagely, and poorly eco-efficient, respectively. Figure 6. The relationship between GRP per capita and eco-efficiency of the cities along the NSR. Note: the number on the orange point in Figure 6 are the numbers in the "No." column of Table 3, which represent the respective cities.
To look at how the eco-efficiency of the cities might be improved, it is necessary to generate the results of inputs and outputs optimization. The average annual percentage changes in the inputs and outputs for these cities are shown in Table 4. The "S-", "S+", and "SB" in Table 4 indicate the excesses of inputs, shortfalls of positive outputs, and excesses of negative outputs, respectively, according to the efficiency (optimal solution) achieved by the EBM-DEA models. It is clear from Table 4 that 18 out of 28 cities need to have a more than 50% reduction in both capital and public investment to achieve ecoefficiency. This leads to the inefficiency in capital and public investment of many Russian Figure 6. The relationship between GRP per capita and eco-efficiency of the cities along the NSR. Note: the number on the orange point in Figure 6 are the numbers in the "No." column of Table 3, which represent the respective cities.
To look at how the eco-efficiency of the cities might be improved, it is necessary to generate the results of inputs and outputs optimization. The average annual percentage changes in the inputs and outputs for these cities are shown in Table 4. The "S-", "S+", and "SB" in Table 4 indicate the excesses of inputs, shortfalls of positive outputs, and excesses of negative outputs, respectively, according to the efficiency (optimal solution) achieved by the EBM-DEA models. It is clear from Table 4 that 18 out of 28 cities need to have a more than 50% reduction in both capital and public investment to achieve eco-efficiency. This leads to the inefficiency in capital and public investment of many Russian Arctic cities.
In fact, this does not mean that capital and public investment in these cities need to be reduced. It is worth noting that the relative efficiency benchmarks (e.g., St. Petersburg) in these cities are insufficient in their use of capital and public investment. It is recommended to diversify the local industrial structure and extend the economic functions, which is one of the ways for inefficiency to be resolved in the Russian cities along the NSR. In addition to the conventional industries such as oil, gas, and mining industries, the authorities can vigorously develop the fishery economy based on the considerable fishery resource in the Russian Arctic. Arctic tourism can be a promising sector of the Russian economy, which has a multiplicative effect for the development of the infrastructure, social services, and employment in the Russian cities along the NSR. In addition, support tools should be used to attract investment to these Russian cities. These tools include lower profit tax rates; reduced severance tax coefficients for oil, gas, and mineral development; a notifying procedure for value-added tax refunds; a simplified procedure for land plots supply; and invariable terms for investment projects implementation.
In terms of population, some cities such as Onega, Vladivostok, and Olenegorsk need to reduce their populations significantly to achieve eco-efficiency according to the results of EBM-DEA models. The authorities need to develop a set of preferential policies to encourage immigration between the associated cities, so that the cities that have a real demand of population can obtain population growth and those that have excessive populations can obtain fewer population to achieve eco-efficiency. Regarding water supply, the cities such as Murmansk, Kandalaksha, Dudinka, Vanino, Nakhodka, Norilsk, and Olenegorsk need to take measures to reduce the water supply substantially to achieve eco-efficiency. With regards to energy supply, in order to be eco-efficient, 10 out of 28 cities need to cut their energy supply by more than 40%. In respect of GRP, a remarkable fact is that the GRPs of most cities need to be increased substantially (the cities, e.g., Dudinka, Vorkuta, Revda, and Nikel even need to have an increase of more than 1000% on GRPs), so that these cities can achieve eco-efficiency. This suggests that according to the efficiency generated by the models, most of the Russian cities along the NSR are economically inefficient, which is generally attributed to the factors such as outdated infrastructure, tiny populations, harsh natural environment, simple economic structure, and insufficient financial and technological resources. It is recommended to update these cities' outdated infrastructure, and develop more regional infrastructure and transportation projects. The "North Latitude Passage" is one of the related key infrastructure projects. It will advance the effective development of the rich natural resources in the Russian Arctic areas (e.g., Polar Urals, Yamal, and the north of Krasnoyarsk territory). The authorities need to continue developing the communication and coastal infrastructure (e.g., navigational and hydrometeorological aids, and port facilities) along the NSR, which will ensure safety of commercial transits through the NSR. The successful functioning of the NSR will bring more development opportunities to the cities along the NSR. The logistical hubs at the end points of the NSR (e.g., the ports of Murmansk and Petropavlovsk-Kamchatsky) can be created to serve domestic and international shipping. The Russian Arctic has great potential for economic development due to its rich natural resources and geographical significance. Its infrastructure construction and oil, gas, and mineral resource development have become the main contributors to current economic growth [53]. It is suggested to promote the further construction and development of the related projects (e.g., the projects of Prirazlomnoye oil field, Novy Port oil field, Bovanenkovo gas field, Kharasaveyskoye gas field, Yamal LNG, and Arctic LNG 2), which can potentially help these cities attract more residents and business opportunities, so that the economy of these cities can be stimulated and boosted.
Regarding the excesses of negative outputs (denoted by "SB" in Table 4), including GHG, solid waste, and water pollution, more than half of the cities need to have a reduction of more than 40% on these outputs to achieve eco-efficiency. The ecology of the Russian Arctic is fragile, and the impact of environmental damage is greater than that in other regions. The circular economy will become the sustainable development solution for Russian Arctic cities in the future. It can reduce the emissions of various forms of pollution and waste, and ensure the sustainable development of the region as much as possible [54]. To maintain the balance between economic development and Arctic environmental protection, the authorities need to continue conducting a major clean-up of the environmental damage in the Russian Arctic areas which was accumulated through the economic activities in the past decades. They also need to develop a system of specially protected natural territories and reserves in the Russian Arctic for better environmental protection. Education and science centers need to be established in the Russian cities along the NSR to ensure the development of fundamental research and help address the practical tasks of Arctic sustainable development. International research teams and alliances of high-tech companies should be encouraged to take part in joint research projects in the fields such as shipbuilding, navigation safety, environmental protection, oil, gas, and mineral production, and marine bioresources harvesting.

Conclusions and Policy Implications
In this study, the EBM-DEA model is used to analyze the eco-efficiency of the 28 Russian cities along the NSR for 10 years from 2010 to 2019. This is a meaningful attempt to study the eco-efficiency of these Russian cities through quantitative analysis. This study expands the traditional TFEE model to make it suitable for the evaluation of eco-efficiency of these Russian cities. Compared with the conventional TFEE models, more pollutant variables (negative outputs) are added to the model in this study. Regarding the policy implications, in order to achieve higher eco-efficiency, the related authorities and stakeholders should devote more resources in multiple fields to the cities along the eastern and western section of the NSR. They need to develop renewable energy or eco-efficient projects such as using wind and geothermal energy resources in these cities. It is necessary to develop more industrial sectors (e.g., fishery, shipbuilding, and tourism) that are not limited to traditional energy industries based on the regional characteristics in these cities, and extend economic functions of these cities. Investment support tools should be used to attract investment to these Russian cities. It is important to renovate obsolete infrastructures, and carry out more regional infrastructure and transportation projects. It is also suggested to promote the further construction and development of the projects about oil, gas, and mineral extraction in the Russian Arctic. Particular emphasis should be placed on the circular economy for the sustainable development of the Russian Arctic cities. The clean-up of environmental damage and establishment of specially protected natural territories and reserves are effective measures for the balance between economic development and environmental protection. More resources should be invested to the related research for the sustainable development of the Russian cities along the NSR.
However, there are still unconsidered issues in this study, which can be further studied. There are many variables in the eco-efficiency model that have a carryover effect. The dynamic model will be improved to study the effect. In addition, the single process model usually ignores the conflicts between different departments within the process. The network methods will be used to conduct further research on existing models to discover the impact of conflicts between departments. Finally, the existing evaluation methods are based on the assumption of diminishing marginal productivity. It ignores that the existence of economies of scale may cause evaluation errors. Finding out how to consider the effect of economies of scale and the effect of diminishing marginal productivity within an evaluation method requires further attention.