Monitoring Eco-Efficiency and Its Convergence Towards Sustainability in the European Rubber and Plastics Industry Through Circular Economy Transition

Abstract

Eco-efficiency is crucial for the European rubber and plastics industry to minimize production costs through effective resource management (e.g., energy management) and reduce environmental impacts like greenhouse gases (GHGs) emissions. Circular economy (CE) solutions can support the industry’s competitiveness while aligning with sustainability goals and regulatory requirements. In the present research, we employ a hybrid window data envelopment analysis (WDEA) methodology to measure panel data eco-efficiency via the application of the moving average principle. The examination of 27 European countries as decision-making units (DMUs), in the period 2014–2022, led to the conclusion that the average eco-efficiency is 70.33%, showing that most of the DMUs can ameliorate their performance regarding pollution control. The highest eco-efficiency in 2014 can be monitored in Ireland, Switzerland, Norway, and Poland, but in 2022, only Ireland and Switzerland kept their positions, whereas Norway dropped to the 16th position and Poland plummeted to the 24th hierarchical position. Geographical disparities can be spotted, as Northern and Western Europe have greater eco-efficiency than Eastern and Southern Europe. At a second level of analysis, the convergence between the 27 European countries in the period under consideration is examined using the log t regression test and club clustering. The analysis leads to three final clubs where conditional convergence dominates.

1. Introduction

Geyer et al. [1] stated that “a world without plastics… seems unimaginable today”, showing the multiple benefits they provide to our well-being as plastics are cheap and useful; however, these advantages come at a significant environmental cost. Currently, more than 9 out of 10 plastics are produced with fossil fuels, while the rest are recycled and biobased plastics. Projections from the Organisation for Economic Co-operation and Development (OECD) [2,3] show that plastic use is going to double by 2060; moreover, plastic leakage is going to double in rapidly developing countries and triple regarding the plastic stocks that are disposed of in marine ecosystems by 2060. Similarly, Inger Andersen [4] emphasized the need to shift from the current linear economy model to a circular economy (CE) to address plastics’ impact on the triple planetary crisis, i.e., climate change, biodiversity loss, and pollution and waste crisis. Plastic pollution can aggravate the triple planetary crisis; for example, the global plastic outlook reports of OECD [2,3] state that energy- and carbon-intensive plastic production emits GHGs, generates waste, and leads to air, soil, and water pollution, but most importantly, it can lead to biodiversity loss, as animals can die due to plastic pollution (e.g., turtles are the most common example of plastic pollution impact).

The endless reliance on fossil fuels, either for raw materials or for energy production (e.g., electricity), has contributed to numerous environmental problems and health harms [1,5]. Apparently, the substitution of fossil-based with biobased plastics will decrease greenhouse gas (GHG) emissions, but the production of biobased plastics also requires additional land, emitting more GHGs into the atmosphere due to deforestation activities, also contributing to climate change [2,3]. It is apparent that the prices of plastics are highly coupled with fossil fuel markets, constituting an institutional conundrum in Europe, as many European governments back fossil fuel consumption through environmentally unsustainable subsidies. For instance, Watkins and Schweitzer [6] show that such fossil-based subsidies total approximately EUR 112 billion annually, while European Union member states are far from reaching the goal of eliminating such subsidies. Given that energy use and GHGs can heavily impact the environment, the present study aims to find which parameters influence the eco-efficiency performance in the European rubber and plastics industry and to find if there is convergence in this industry towards a common steady state.

The CE can be deemed as an alternative to the linear model as it enables the circularity of resource and material flows by boosting preventive actions in material losses [7,8]. CE highlights the loss of value regarding plastic manufacture, usage, and disposal, as the current linear model not only captures economic benefit, but also harms the environment [9,10]. This is the reason why CE can be strengthened by novel technological solutions that target not only economic aspects, but environmental as well in line with digital transformation under the scope of Industry 4.0 [11,12]. Hence, having in mind the twin green and digital transition, the United Nations proposed the sustainable development goals (SDGs) [13] in an attempt to deal with different development issues, including environmental ones. The European rubber and plastics industry can be linked to the achievement of several SDGs, i.e., proper economic growth (SDG8) and inclusive and sustainable industrialization and cleaner production (SDG9 and SDG12). These industries can also be connected to climate change mitigation solutions (SDG13) and the reduction in plastics on land and in water ecosystems (SDG14 and SDG15). Stricter environmental policies, like the European Green Deal and especially the green deal industrial plan, focus on the achievement of the net-zero emissions goal in 2050 [14]. Similarly, the Circular Economy Action Plan and specifically the European strategy for plastics in the circular economy paved the way for the future re-orientation of national plastic-related policies in the European Union (EU) member states [15].

Therefore, it is imperative to create synergies as mentioned by Karali et al. [16] in order to cope with the interrelations between climate change and plastic pollution by initiating common global climate and plastic treaties. Such an attempt was made in March 2022 when all United Nations member states agreed to negotiate an internationally legally binding instrument (ILBI) on plastic pollution [17]. This initiative takes into account the impacts of plastic pollution on climate change in the UNFCCC negotiations [16].

This paper addresses two research questions (RQs) regarding the eco-efficiency of the European rubber and plastics industry and potential club convergence among the decision-making units (DMUs):

RQ1: 

What is the current eco-efficiency in the European rubber and plastics industry and how can be strengthened by circular economy strategies?

RQ2: 

Is there convergence among the DMUs?

The objective of the present research is to monitor the eco-efficiency of the European rubber and plastics industry in 27 European countries and to inspect if there is convergence among them. Hence, the present paper employs a sequential analysis. Firstly, there is the application of a hybrid window data envelopment analysis (WDEA) that takes into account both radial and non-radial parameters and also treats the undesirable output via a weak disposability approach, and secondly, a club convergence methodology is applied to show if there is convergence among the European rubber and plastics industries. The novel contribution of this paper lies in the present sequential methodological application in the European rubber and plastics industry as it has never been examined in this aspect in the specific industry; consequently, the present paper introduces a novel methodology and provides new policy insights for CE solutions in the European rubber and plastics industry. Additionally, this paper addresses the existing research gap on how CE can improve eco-efficiency in the rubber and plastics industry considering the global plastics treaty that is under negotiation. This paper is structured as follows: Section 2 covers the theoretical background on efficiency in the rubber and plastics industry, Section 3 outlines the methodology, Section 4 presents the results and discussion, and Section 5 concludes this paper and provides circular economy policy implications.

2. Literature Review

Economic performance that also considers environmental parameters is called eco-performance, and more specifically, there are two well-known categories: eco-efficiency and eco-productivity. Eco-performance has been widely applied in several sectors, aiming to find ways to improve DMUs’ structural characteristics or to provide policy implications, inter alia, in agriculture [18,19], in manufacture [20,21,22,23,24,25,26,27], and at a broader nationwide level [28,29,30,31,32,33,34,35]. The DEA multidisciplinary application enables the valuation of efficiency in different sectors and activities.

2.1. Eco-Efficiency Measurement

Most eco-performance measurement studies focus either on the parametric Stochastic Frontier Analysis (SFA) or on data envelopment analysis (DEA) methodology. The present study chooses the DEA analysis, which is based on Charnes et al. [36]. Efficiency is a typical scalar measure between zero and unity or as a percentage (0% for inefficiency and 100% for efficiency). An alternative form is WDEA, proposed by Charnes et al. [37], which utilizes panel data via the application of the moving average principle. In WDEA, each window (i.e., average based on the window width) is typically a new DMU, but also each DMU is related to the other DMUs and to its own efficiency [35].

Regarding the suitability of DEA model choice, Tone [38] stated that efficiency measurement should clarify the radial or non-radial aspects of the model’s variables. The most-known DEA models, the CCR (by Charnes et al. [36]) and the BCC (by Banker et al. [39]), are radial approaches and ignore the non-radial slacks of the variables. This omission might lead to an under-evaluation of the mathematical modeling [40]. On the contrary, the slack-based measure (SBM) as proposed by Tone [41] is a non-radial methodology that surpasses the above disadvantage of radial modeling. In the literature, there are interesting examples of non-radial applications (inter alia, [28,42,43]), but this method also lacks proportionality as it ignores the radial aspects [40]. Aiming to overcome both of these issues, Tone [38] introduced the hybrid DEA that considers both radial and non-radial aspects.

The last decades have witnessed a rising interest in dealing with undesirable outputs. In the literature, there are different ways to cope with them, inter alia, through their ignorance (For more information on how to cope with undesirable outputs, please check the following references: [44,45,46]). Another issue in DEA modeling is the choice between weak or strong disposability. The former as introduced by Shepard [47] and according to Färe and Grosskopf [48,49] is linked to the null-jointness property in order to cope with undesirable outputs. For example, if pollution can be diminished through contemporary technological ways, then weak disposability is proper; otherwise, strong disposability is applicable [35]. Nevertheless, in the literature, there is criticism regarding the choice between the most suitable disposability method (inter alia, by [50,51,52,53]). The main drawback of weak disposability might be the assumption that there are unaltered relative proportions among outputs; however, wanting to reduce the impact of undesirable outputs, the desirable outputs ought to be lowered as well, and ultimately, this process might lead to unwanted results sometimes. Recently, some publications, inter alia, by Mehdiloo and Podinovski [53], made efforts to monitor and specify the interrelations among variables, upon which the choice of the weak or strong disposability process has been made.

In the following text, the plastic and rubber industries exhibit different performances in the literature; however, the environmental impacts of these industries have attracted the attention of academicians and policymakers in recent decades. In line with the literature review, several aspects are going to be revealed such as existing gaps, novel insights, and potential policy implications relevant to the present research.

In Malaysia and Iran, the plastics and rubber industries need to improve efficiency by increasing investments in vocational training for employees; additionally, policymakers should consider catching-up effects when developing strategies to enhance industry performance. Firstly, Mahadevan [54] found that the overall total factor productivity in the rubber and plastic industries, among 28 Malaysian industries, showed progression by 0.9% and 1.1%, respectively, mainly due to catching-up effects. Additionally, in Malaysia, Sabli et al. [55] analyzed 586 plastic manufacturing firms and proposed that improving efficiency performance should focus on addressing labor-related issues; for example, this includes providing financial incentives (e.g., higher wages) and implementing proper vocational training programs. In Iran, Ghondaghsaz et al. [56] highlighted the need for protection of national plastic manufacturing against international competition as the domestic plastic firms are incapable of reaching the optimal efficiency standards, even though the mean efficiency levels were 73% and 92.2% for the CCR model and BCC model respectively; moreover, two parameters seem to play a pivotal role in efficiency, namely outsourcing and the experience of the staff. In essence, the existing gaps from the previous research are mainly due to the omission of undesirable outputs, whereas the present paper also takes into account the influence of GHGs as undesirable outputs to showcase their impact on overall eco-efficiency and provide policy insights for a circular economy.

The rubber and plastics sector plays a key role among the Chinese manufacturing sectors in the transition towards CE due to its carbon and energy intensity. Han et al. [57] inspected the productivity performance in the period 1999–2008 and found that the rubber industry had only negative performance, whereas plastics faced negative performances mainly in the periods 2002–2003 and 2004–2005. Specifically, the plastics industry was one of the top performers but lost its dynamism in the later period, whereas the rubber industry experienced only a productivity decline [57]. Emrouznejad and Yang [58] showed that the rubber and plastics sector had mainly positive eco-performance through both CSR and BCC models and drew the following policy implications to further boost the overall Chinese industrial performance: (i) strengthen research and development initiatives to lower carbon intensity, (ii) incentivize carbon neutrality, (iii) enable mimicry of other developed countries, and (iv) harmonize the institutional framework with technical standards.

Similarly, Lin et al. [59] compared a traditional DEA model with a WDEA model, showing that the plastic and rubber sector reached a 64.9% efficiency in the traditional DEA but only 50.3% in the WDEA, showing that the overall Chinese manufacturing should target the production of novel products that are innovation-led and environmentally friendly. In addition, Xie et al. [60], based on a multiple DEA model and by applying a Gini criterion through a clustering analysis, revealed that the plastic and rubber industry altered its pollution-related status as it belonged to the medium–high pollution sectors from 2006 to 2009, but from 2010 to 2015, it was part of the high pollution sectors, showing a need for more environmentally oriented industrial policies and structural manufacturing changes. Furthermore, Wang et al. [61], based on a biennial Malmquist–Luenberger DEA-based analysis followed by panel quantile regression, found that plastics belong to the most innovative and catching-up sectors among 34 Chinese industrial sectors, having one of the greatest performances regarding the green total factor productivity. Even though the previous papers covered productivity performance aspects, they did not consider the efficiency parameters especially either as hybrid DEA or WDEA, which the present paper evaluates.

In Europe, some studies showed that parameters such as the survival rate and profitability can play a significant role in a sector’s performance. Tsekouras et al. [62,63], based on technical and scale efficiency, found in two similar publications that among 440 and 359 firms, the firm size and the industry’s business and life cycle are critical parameters for a firm’s survival, with the former having a positive effect and the latter a negative effect. Additionally, technical efficiency can reveal the strengths of a firm to continue, whereas scale efficiency shows how profitable a firm can be in this sector. Next in order, Robaina et al. [9] employed a different methodology, i.e., the multi-directional efficiency analysis (MEA), instead of using DEA due to the inability of DEA to measure inefficiency patterns that are taken into account in MEA. The results of MEA showed that the Czech Republic, Germany, Denmark, Ireland, Luxembourg, and Malta are the most efficient DMUs, followed by Austria, France, the Netherlands, Norway, and the UK. On the other hand, some DMUs showed improvements but at a lower extent, inter alia, Bulgaria, Spain, Greece, Finland, Romania, and Portugal, mainly due to input improvements. The gaps in the literature from the previous research can be directed at the level of performance monitoring; for instance, the previous studies were applied either at the firm level or the national level, while the present paper applies an eco-efficiency measurement at the sectoral level. The importance of highlighting the sectoral-level analysis is because one might understand the core system dynamics by measuring sectors such as the rubber and plastics industry.

Next in order, the proposed methodology that combines hybrid WDEA and club clustering in 27 European countries is presented. The proposed methodology considers parameters of the variables that the literature has presented as shown in Appendix A in

Table A1. Literature review on studies that include the performance monitoring in plastics and rubber industries.

DMUs (Period)	Techniques	Inputs	Outputs	Ref.
28 industries in Malaysia including the plastics and rubber industries (1981–1996)	MPI-DEA (output-based)	(i) Fixed capital stock (ii) Labor as number of workers	Value Added	[54]
440 firms in the Greek plastics and rubber industry (1989–1997)	technical efficiency and scale efficiency	(i) Net value of total assets as proxy of capital (ii) Labor as number of employees	Total value of shipments	[62]
359 firms in the Greek plastics and rubber industry (1989–1997)	technical efficiency and scale efficiency	(i) Net value of total assets as proxy of capital (ii) Labor as number of employees	Total value of shipments	[63]
14 Chinese industrial sectors (1999–2008)	MPI-DEA	(i) Coal consumption (ii) Electricity consumption	Gross industry output value	[57]
30 Chinese manufacturing industries (2004–2012)	global Malmquist–Luenberger productivity index	(i) Labor (ii) Asset (iii) Energy	(i) Gross Industrial Output Value (GIOV) (desirable) (ii) CO2 emissions (undesirable).	[58]
28 Manufacturing industries in China (2006–2014)	Traditional DEA, WDEA (window width 5), and absolute β convergence analysis	(i) R&D Personnel Full-time Equivalent (ii) Technical Renovation Expenditure (iii) Internal R&D investment (iv) Energy Consumption	(i) Accepted Invention Patent Applications (ii) Sales Revenue from New Products (iii) Total Wastewater Discharged (iv) Total Waste Gas Emissions	[59]
17 Iranian plastic production firms (2015–2016)	CCR and BCC DEA and Bootstrapped Tobit regression model	(i) Cost of goods sold (ii) Fixed assets (iii) Products spoiled	(i) Sales revenues (ii) Market share	[56]
586 firms inthe plastic products manufacturing industry in Malaysia (2015)	DEA (output-oriented) and Tobit regression analysis	(i) Capital (ii) Labor as total employees (iii) Intermediate input	(i) Sales	[55]
36 Chinese industries (2006–2015)	Multiple DEA model with a Gini criterion with a clustering analysis	(i) Labor as number of employees (ii) Resources (e.g., energy, assets, scale, environmental governance input)	(i) Technology (new product and patents) (ii) Economic output (desirable) (iii) Environmental pollution output (undesirable)	[60]
26 European countries (2006–2016)	MEA	(i) Labor (ii) Capital invested (iii) Energy consumed	(i) GDP (desirable) (ii) CO2 emissions (undesirable) (iii) Plastic waste (undesirable) (iv) Plastic recycling (desirable) (v) Plastic recovery (desirable)	[9]
34 Chinese sectors (2005–2015)	Biennial Malmquist–Luenberger and fixed-effect panel quantile regression	(i) Capital stock (ii) Labor (iii) Energy consumption	(i) Gross Industrial Output Value (desirable) (ii) Industrial wastewater (undesirable) (iii) Industrial waste gas (undesirable) (iv) Industrial solid waste (undesirable)	[61]

, aiming to introduce an interesting and applicable methodology not only for measuring eco-efficiency but also for combining it with club clustering technique to create groups of DMUs with common trajectory paths. It should also be mentioned that the results were also underpinned by a sensitivity analysis.

2.2. Circular Technologies for Higher Industrial Performance

The OECD [64] categorized sustainable design goals to improve value chain performance, emphasizing the importance of reducing emissions, avoiding hazardous chemicals, and minimizing material use at each stage of production. In this context, innovative circular technological solutions should focus on redesigning and innovating the value chain. In the literature, innovative circular technological solutions should target the redesign and innovation of the value chain for abating environmental issues such as waste generation of plastics or other waste flows [65,66]. For instance, advanced digitalization tools, such as Radio Frequency Identification (RFID) and the Internet of Things (IoT), enable real-time monitoring and data collection to trace plastic-containing products, facilitating their recollection for reuse or recycling [67,68]. Similarly, the Ellen MacArthur Foundation [69] presented the material innovation that targets the recyclability of primary goods in the production process as a potential solution; additionally, the promotion of chemical recycling or alternatives can lead to a reprocessing of “currently unrecyclable plastic packaging into new plastics feedstocks”. Furthermore, integrating eco-efficiency measurement into process optimization can benefit significantly from novel predictive maintenance enabled by digitalization.

All of the above technological solutions and strategies are widely applied in Industry 4.0 and play a crucial role in advancing extended producer responsibility alongside the digital transformation toward a circular economy [11,12]. To enhance the impact of these efforts, it is essential to prioritize inter-regional collaborations, which can help tailor circular economy solutions to the needs of low-performing DMUs [33,70]. For example, voluntary commitments or policy instruments can pave the way for minimizing the negative externalities (e.g., emission abatement) in plastic pollution, especially through the extensive use of recycled plastics [2,3]. Overall, digital transformation under the scope of a circular economy can align with global initiatives against plastic pollution and climate change.

3. Materials and Methods

The current paper retrieved data regarding the variables of labor, Gross Fixed Capital Formation (GFCF), and Value Added from the INDSTAT 2 2023 edition ISIC revision 3 at the 2-digit level from the United Nations Industrial Development Organization (UNIDO) database [71] and the variables of energy use, GHG emissions (Greenhouse gases are CO₂, N₂O in CO₂ equivalent, CH₄ in CO₂ equivalent, HFC in CO₂ equivalent, PFC in CO₂ equivalent, SF₆ in CO₂ equivalent, and NF₃ in CO₂ equivalent. CO₂ emissions have been utilized for Türkiye due to a lack of data availability.) from the EUROSTAT database [72,73]. In the rare instances of missing values, appropriate interpolations were performed using a moving average as well as single and double exponential smoothing techniques to forecast the missing data. The selection of the most suitable method was determined based on accuracy measures such as Mean Absolute Percentage Error (MAPE), Mean Absolute Deviation (MAD), and Mean Squared Deviation (MSD). Since GFCF data were unavailable in 2022, the central tendency was calculated, and the missing values were filled accordingly; the same procedure was also performed for Türkiye for the CO₂ emissions in 2022.

This study covers the period 2014–2022. It should also be mentioned that the countries were selected based on data availability for the plastics and rubber industry code (This refers to the manufacture of rubber and plastics products in the division 25 covering the manufacture of rubber products with the code 251 and the manufacture of plastics with the code 252. For more information, please see UNDESA [74] (pages 88–89)), and a sequential analysis was utilized. The reporting period, which spans from 2014 to 2022, is characterized by the vulnerability of the plastics recycling sector due to macroeconomic factors such as the drop in oil prices in 2014 and the COVID-19 crisis with the resulting decline in oil prices. Due to this decline, it is becoming increasingly difficult for recycled plastic to compete with virgin plastic as input costs (e.g., for energy) are lower and the gap between them is widening, which poses a major threat to the plastic recycling industry [75]. In this context, mapping the eco-efficiency profile of the European rubber and plastics industry in the period 2014–2022 is crucial. More specifically, the analysis conducted using an appropriate model provides specific information on the level of eco-efficiency that can contribute to the transformative change needed in the European rubber and plastics industry to achieve green recovery and shape the appropriate roadmap for achieving the Sustainable Development Goals in the post-pandemic period. In the first stage, a hybrid WDEA analysis is employed with window width 3 applied in 27 European countries as DMUs (The 27 DMUs are as follows: Austria (AUS); Belgium (BEL); Croatia (CRO); Cyprus (CYP); Czechia (CZE); Denmark (DEN); Estonia (EST); Finland (FIN); France (FRA); Germany (DEU); Greece (GRC); Hungary (HUN); Ireland (IRE); Italy (ITA); Latvia (LTV); Lithuania (LTH); Netherlands (NTH); Norway (NOR); Poland (POL); Portugal (POR); Romania (ROM); Slovakia (SLK); Slovenia (SLN); Spain (ESP); Sweden (SWE); Switzerland (SWT); Türkiye (TRC).). Characteristics of the DMUs are presented in Appendix B in

Table A2. Characteristics of the DMUs. Data retrieved from the World Bank, and reference is the last available year.

Country	Surface Area (1000 sq. km)	Population (Million Persons)	Life Expectancy (Years)	GDP Growth (Annual %)
Austria	83.88	9.13	81.09	4.81
Belgium	30.53	11.82	81.70	3.01
Croatia	88.07	3.85	77.58	7.03
Cyprus	9.25	1.26	81.89	5.06
Czechia	78.87	10.87	79.03	2.35
Denmark	42.92	5.95	81.30	2.73
Estonia	45.34	1.37	77.94	−0.46
Finland	338.47	5.58	81.19	1.34
France	549.09	68.17	82.23	2.45
Germany	357.59	84.48	80.71	1.81
Greece	131.96	10.36	80.64	5.56
Hungary	93.03	9.59	76.02	4.58
Ireland	70.28	5.26	83.06	9.43
Italy	302.07	58.76	82.90	3.99
Latvia	64.59	1.88	74.58	2.95
Lithuania	65.29	2.87	75.79	2.44
Netherlands	41.54	17.88	81.71	4.33
Norway	624.50	5.52	82.56	3.01
Poland	312.71	36.69	77.30	5.64
Portugal	92.23	10.53	81.58	6.83
Romania	238.40	19.06	75.30	4.11
Slovakia	49.03	5.43	77.07	1.87
Slovenia	20.48	2.12	81.28	2.46
Spain	505.96	48.37	83.08	5.77
Sweden	528.86	10.54	83.11	2.66
Switzerland	41.29	8.85	83.45	2.57
Türkiye	785.35	85.33	78.48	5.53
Average	207.10	20.06	80.10	3.85

.

The hybrid WDEA modeling, as presented in Figure 1, has the inputs of (i) labor as number of employees, (ii) GFCF in USD, and (iii) energy use in terajoules. Furthermore, the desirable output is Value Added in current USD, whereas the undesirable output is GHGs in tonnes. The production of the desirable output is utilized in a separable way, whereas the production process also creates the undesirable output, which negatively affects the overall eco-efficiency; hence, it is expressed in a non-separable way. The descriptive statistics are presented in

Table 1. Descriptive statistics.

	INPUTS			OUTPUTS
	GFCF	Energy Use	Labor	Value Added	GHGs/CO2
	(Current USD)	(Terajoules)	(Number of Employees)	(Current USD)	(Tonnes)
Mean	690,700,815.65	14,153.70	65,674.00	4,150,932,116.64	316,791.42
Min	2,077,775.77	61.60	840.00	27,734,506.75	3632.44
Max	4,897,864,010.77	99,336.50	471,527.00	41,925,714,434.40	3,142,295.41
STDEV	948,814,254.64	21,727.37	97,433.00	6,992,657,417.51	532,583.64

, and the correlation matrix in

Table 2. Correlation matrix.

	GFCF	Labor	Value Added	GHGs/CO2	Energy Use
GFCF	1
Labor	0.971 **	1
Value Added	0.957 **	0.924 **	1
GHGs/CO2	0.862 **	0.863 **	0.893 **	1
Energy Use	0.952 **	0.932 **	0.928 **	0.838 **	1

** Correlation is significant at the 0.01 level (2-tailed).

; regarding the correlations between the variables, there is a strong positive statistically significant correlation.

3.1. Hybrid DEA Method

Initially, the present paper formulates the WDEA methodology [35,38,76] for all 27 DMUs (n = 27) with $\mathcal{k}=3$ inputs and $\mathcal{l}=2$ outputs in the ${\mathbb{R}}_{+}^{\mathcal{k}+\mathcal{l}}$ the Euclidean space, creating the according input ($X\in {\mathbb{R}}_{+}^{\mathcal{k}\times n}$) and output matrices ($Y\in {\mathbb{R}}_{+}^{\mathcal{l}\times n}$). Moreover, the modeling should include the radial and non-radial parameters, so the model has the radial input ($X^R\in {\mathbb{R}}_{+}^{{\mathcal{k}}_1\times n}$) and output ($Y^R\in {\mathbb{R}}_{+}^{{\mathcal{l}}_1\times n}$), as well as the non-radial input ($X^{NR}\in {\mathbb{R}}_{+}^{{\mathcal{k}}_2\times n}$) and output ($Y^{NR}\in {\mathbb{R}}_{+}^{{\mathcal{l}}_2\times n}$), given that $\mathcal{k}={\mathcal{k}}_1+{\mathcal{k}}_1$ and $\mathcal{l}={\mathcal{l}}_1+{\mathcal{l}}_2$. Then, the matrices are presented below in relation (1):

$$X=\left(\begin{array}{c}{X}^R\\ X^{NR}\end{array}\right) \mathit{and}Y=\left(\begin{array}{c}{Y}^R\\ Y^{NR}\end{array}\right)$$

(1)

where “R” is the radial and “NR” is the non-radial parameters, and the production possibility set (P) is determined under the constant-returns-to-scale (CRS) assumption in relation (2):

$$P=\left\{\left(x,y\right)|x\ge X\lambda , y\le Y\lambda , \lambda \ge 0\right\}$$

(2)

where λ acts as a non-negative vector in ${\mathbb{R}}^{\mathrm{n}}$, and relation (2) can take into account the variable returns to scale (VRS) by also including the constraint of λ (i.e., ${\sum }_{\mathrm{j}=1}^{\mathrm{n}}{\mathsf{\lambda }}_{\mathrm{j}}=1$). The modeling also considers the slacks through input overflows and output shortages as in the work of Cooper et al. [76] and Halkos and Polemis [35].

From Cooper et al. [76], a hybrid efficient index ρ might be produced via relation (3):

$$\rho ={\displaystyle \frac{1-\frac{{\mathcal{k}}_1}{\mathcal{k}}\left(1-\sigma \right)-\frac{1}{\mathcal{k}}{\sum }_{i=1}^{{\mathcal{k}}_2}\raisebox{1ex}{$s_i^{NR-}$}\!\left/ \!\raisebox{-1ex}{$x_{io}^{NR}$}\right.}{1+\frac{{\mathcal{l}}_1}{\mathcal{l}}\left(\tau -1\right)+\frac{1}{\mathcal{l}}{\sum }_{r=1}^{{\mathcal{l}}_2}\raisebox{1ex}{$s_r^{NR+}$}\!\left/ \!\raisebox{-1ex}{$y_{ro}^{NR}$}\right.}}$$

(3)

where σ and τ are the coefficients of radial parameters for inputs and outputs, respectively; $s_i^{NR-}$ is an input overflow, and $s_r^{NR+}$ is an output shortage. A ${\mathrm{D}\mathrm{M}\mathrm{U}}_{\mathrm{O}}\left({\mathrm{x}}_{\mathrm{o}},{\mathrm{y}}_{\mathrm{o}}\right)$ is deemed as hybrid efficient when $\rho =1$ if $\sigma =1, \tau =1, {\mathcal{s}}^{\mathsf{{\rm N}}\mathrm{R}-}=0, and {\mathcal{s}}^{\mathrm{N}\mathrm{R}+}=0$. One should consider the assumption that the 27 DMUs are based on a production function that has $\mathcal{k}$ inputs and produces a desirable ($Y^D$) and an undesirable output ($Y^{UD}$) with vectors $x\in {\mathbb{R}}^{\mathcal{k}},Y^D\in {\mathbb{R}}^{{\mathcal{l}}_1}, and{ Y}^{UD}\in {\mathbb{R}}^{{\mathcal{l}}_2}$. Hence, under the assumption that $X,Y^D,$ and $Y^{UD}>0$, the production possibility set is presented in relation (4):

$$P=\left\{\left(x,y^D,y^{UD}\right)| x \ge X\lambda , y^D\le Y^D\lambda , {y^{UD}\ge Y}^{UD}\lambda , \lambda \ge 0 \right\}$$

(4)

The efficiency for a ${\mathrm{D}\mathrm{M}\mathrm{U}}_{\mathrm{O}}\left({\mathrm{x}}_{\mathrm{o}},{\mathrm{y}}_{\mathrm{o}}^{\mathrm{D}},{\mathrm{y}}_{\mathrm{o}}^{\mathrm{U}\mathrm{D}}\right)$ that treats undesirable outputs is achieved only if there is no vector $\left(x,y^D,y^{UD}\right)\in P$ with at least one strict inequality and ${\mathrm{x}}_{\mathrm{o}}\ge x, {\mathrm{y}}_{\mathrm{o}}^{\mathrm{D}}\le y^D,{\mathrm{y}}_{\mathrm{o}}^{\mathrm{U}\mathrm{D}}\ge y^{UD}$ [35].

Additionally, when measuring eco-efficiency, the influence of separability is important; Halkos and Polemis [35] concluded that carbon pollutants treated as undesirable output were not separable from the other desirable outputs, meaning that an increase in the undesirable output can lead to a rise in the desirable output and vice versa. Henceforth, the outputs can be further categorized into separable desirable ${(Y}^{S_D}$), non-separable desirable ($Y^{{NS}_D}$), and non-separable undesirable ($Y^{{NS}_{UD}}$) outputs, whereas inputs are categorized into separable ($X^S$) and non-separable ($X^{NS}$) [26]. Relation (2), by taking into consideration the separable and non-separable parameters, takes the form of relation (5):

$$P_{NS}=\left\{\left(X^S,X^{NS},Y^{S_D},Y^{{NS}_D},Y^{{NS}_{UD}}\right)|x^S\ge X^S\lambda ,x^{NS}\ge X^{NS}\lambda ,y^{S_D}\le Y^{S_D}\lambda ,y^{{NS}_D}\le Y^{{NS}_D}\lambda ,y^{{NS}_{UD}}\ge Y^{{NS}_{UD}}\lambda ,\lambda \ge 0\right\}$$

(5)

A ${DMU}_O\left(x_o^S,x_o^{NS},y_o^{S_D},y_o^{{NS}_D},y_o^{{NS}_{UD}}\right)$ is deemed as non-separable efficient if for any $\psi (0\le \psi \le 1)\left(x_o^S,x_o^{NS},y_o^{S_D},\psi y_o^{{NS}_D},\psi y_o^{{NS}_{UD}}\right)\notin P_{NS}$ and together with at least one strict inequality, i.e., $x_o^S\ge x^S,x_o^{NS}\ge x^{NS},y_o^{S_D}\le y^{S_D},y_o^{{NS}_D}=y^{{NS}_D},y_o^{{NS}_{UD}}=y^{{NS}_{UD}}$. Below, the hybrid modeling is presented:

$${\rho }^{*}=min{\displaystyle \frac{1-\frac{1}{\mathcal{k}}{\sum }_{i=1}^{{\mathcal{k}}_1}\frac{s_i^{S-}}{x_{io}}-\frac{{\mathcal{k}}_2}{\mathcal{k}}(1-\psi )}{1+\frac{1}{\mathcal{l}}\left({\displaystyle {\sum }_{k=1}^{{\mathcal{l}}_1}}\frac{s_r^{S_D}}{y_{k_o}^{S_D}}+({\mathcal{l}}_1+{\mathcal{l}}_2)(1-\psi )\right)}}$$

(6)

$$\begin{gathered} \mathrm{Subject} \mathrm{to} {{\rm X}}_O^S=X_S\lambda +s^{S-} \\ \psi {{\rm X}}_O^{{\rm N}S}=X_{{\rm N}S}\lambda \\ {{\rm Y}}_O^{S_D}=Y_{S_D}\lambda +s^{S_D} \\ {\psi {\rm Y}}_O^{{{\rm N}S}_D}\le {{\rm Y}}^{{{\rm N}S}_D}\lambda \\ {\psi {\rm Y}}_O^{{{\rm N}S}_{UD}}\le {{\rm Y}}^{{{\rm N}S}_{UD}}\lambda \\ {s}^{S-}\ge 0,s^{S_D}\ge 0, \lambda \ge 0, 0\le \psi \le 1 \end{gathered}$$

(7)

3.2. Window DEA Method

The WDEA is based on the moving average method that allows for comparisons of a DMU with other DMUs but with itself also [37], where the input and output matrices can be expressed as

$$x_n^t=\left[\begin{array}{c}\begin{array}{c}{x}_n^{1t}\\ ⋮\end{array}\\ x_n^{\mathcal{k}t}\end{array}\right] y_n^t=\left[\begin{array}{c}\begin{array}{c}{y}_n^{1t}\\ ⋮\end{array}\\ y_n^{\mathcal{l}t}\end{array}\right]$$

(8)

After all, our model utilizes windows with a width of 3, aiming to grasp the impact of micro-term circular policies in the rubber and plastics industry; thus, the present research observes the period 2014–2022 (t = 9). Henceforth, there are 7 windows (nw = t – w + 1), and our analysis calculates the eco-efficiency performance of 567 DMUs (DMUs = N*w*nw).

Bootstrapping Analysis

The non-parametric nature of the DEA method makes its results susceptible to measurement error, so there is a risk that the efficiency gap will be under- or overestimated. To address this problem, a sensitivity analysis is considered appropriate for analyzing the causes of fluctuations in ecological efficiency in detail.

Accordingly, the non-parametric test of independence is applied to all scale efficiency assumptions (i.e., technology with constant (CRS), non-increasing (NIRS), and variable returns to scale (VRS)) in models that focus on both reducing the input for a given output level and maximizing the output for a given input. The results show that the p-value is 0.002 at a 5% significance level, rejecting the null hypothesis of independence of the radial (Debreu–Farrell) output-based measure of technical efficiency for the CRS technology and the mix of outputs.

This means that the smoothed homogeneous bootstrap can be used in the context of applying output-oriented efficiency measures under the assumption of CRS technology. In this case, the sample under consideration consists of DMUs with similar characteristics and technology levels. Therefore, the factors that could influence the level of eco-efficiency and the degree of convergence of the countries studied have to do with the development strategies and policies adopted at sectoral, national, and, of course, European levels to promote and implement sustainable industrialisation that contributes to a climate-neutral, resource-efficient, and circular economy. Next, this study focuses on the potential problems of radial efficiency measures and tries to solve them using the smoothed homogeneous bootstrap with 999 replications. As it can be seen from the results, the efficiency measures are biased upwards if the problem of bias is not considered (Figure 2). Next, we perform a scale analysis for each data point, which shows that 145 out of 243 data points are statistically scale efficient using the homogeneous bootstrap (Supplementary Material, Table S1).

3.3. Inspecting for Convergence

The next stage of the analysis is the club clustering based on the eco-efficiency measurement from the first stage. Phillips and Sul [77] proposed the “log t” regression test which is a novel form of convergence through a non-linear time-varying factor model that allows for merging variables with common characteristics into broader clubs. It utilizes the results from hybrid WDEA to monitor how DMUs’ eco-efficiency can be decomposed into two parameters as in the following expression:

$${Eff}_{it}=g_{it}+a_{it}=\left({\displaystyle \frac{g_{it}+a_{it}}{u_t}}\right)u_t={\delta }_{it}{u}_t$$

(9)

where the following definitions hold:

-

Eff_it represents eco-efficiency performance respecting efforts toward a green transition for a DMU i (i = 1,…, 27) at time t (t = 2014,…, 2022).

-

g_it represents the systematic common components, and a_it embodies transitory components.

-

u_t denotes a single common component, and δ_it is a time-varying idiosyncratic element that captures the idiosyncratic distance between the common factor u_t and the systematic part of Eff_it.

The time-varying term δ_it is in a semi-parametric structure as follows:

$${\delta }_{it}={\delta }_i+{\displaystyle \frac{{\sigma }_i{\xi }_{it}}{{L\left(t\right)t}^{\alpha }}}$$

(10)

where the following definitions hold:

-

δ_i and a scale parameter σ_i are fixed, across the panels.

-

ξ_it is an i.i.d. random variable with a mean equal to zero and variance equal to unity across i, but weakly dependent over t.

-

L(t) is a slowly varying function; an example of the function L(t) is log(t), which becomes infinite as t approaches infinity.

-

α captures the decay rate of cross-sectional variations, that is, the rate of convergence of Eff_it toward δ_i.

In this circumstance, Phillips and Sul [78] developed a regression t test for the null hypothesis of convergence. The null hypothesis and its alternative (i.e., divergence) are then obtained as follows:

$$\begin{gathered} {H_0: \delta }_{it}=\delta and a\ge 0 against \\ {H_1: \delta }_{it}\ne \delta and a<0 \end{gathered}$$

(11)

By adopting this methodology, a one-sided t-test for heteroskedasticity and autocorrelation is applied to the β-coefficient. Hence, the null hypothesis of convergence can be rejected if the t-statistic of the test does not exceed −1.65 as a critical value at a 5% significance level. Phillips and Sul [78] developed a method that takes into consideration both convergent and divergent clubs.

4. Results and Discussion

Firstly, regarding RQ1, the monitoring of eco-efficiency is based on the hybrid WDEA technique and applied to the European rubber and plastics industry, showing that there is a significant drop in average eco-efficiency, as shown in Figure 3. More specifically, the average eco-efficiency declined by 6% from 2014 to 2022. Regarding the reason for such a decline, on the one hand, most countries seem to have reduced their GHG emissions in the European rubber and plastics industries, but on the other hand, energy use has escalated due to the industry’s needs to meet the global demand. In the Supplementary Material, Table S2 presents the eco-efficiency score of each DMU in each year during the period 2014–2022.

The average eco-efficiency in the whole period is 70.33%, as presented in Figure 4a, where the overall eco-efficiency performance of the rubber and plastics industry experienced a decline during the period 2014–2022 as can be seen in Figure 3 (also in Appendix C in Figure A1). Among the top performers based on average eco-efficiency, there are mostly Northern and Western European countries such as Ireland, Switzerland, Finland, and Denmark. On the contrary, the laggards seem to be countries in Eastern Europe and Southern Europe such as Romania, Hungary, Slovakia, Lithuania, and Greece.

The average eco-efficiency shows that most of the Balkan countries expose feeble performance. Additionally, France, Spain, Portugal, and Norway have lower eco-efficiency in 2022 than in 2014. Furthermore, to understand the average eco-efficiency performance, one might inspect the fluctuations during the period 2014–2022 and find that the most impactful fluctuations in eco-efficiency are in Poland, Latvia, Estonia, and Portugal, as can be seen in Figure 4b. The greatest fluctuation in eco-efficiency is shown in Poland, Latvia, Estonia, and Portugal, showing that these countries might have significant changes in technological innovations as the European rubber and plastics industry is highly linked to the adoption of greener production forms. Additionally, stricter environmental policies and strategies might have an impact on the European rubber and plastics industry’s eco-efficiency, as for instance, the European Green Deal might have resulted in performance changes in accordance with the stricter industrial policies.

DEA is a benchmarking method; therefore, the hierarchy plays a pivotal role, as can be seen in

Table 3. Change in hierarchy based on eco-efficiency from 2014 to 2022.

Country	2014	2022	Change in Hierarchy
Latvia	27	11	16
Croatia	15	5	10
Finland	8	1	7
Sweden	10	4	6
Estonia	16	10	6
Belgium	11	6	5
Czechia	18	15	3
France	20	17	3
Spain	21	18	3
Greece	23	20	3
Germany	13	12	1
Ireland	1	1	=
Switzerland	1	1	=
Italy	7	7	=
Hungary	25	25	=
Netherlands	12	13	−1
Lithuania	22	23	−1
Romania	26	27	−1
Slovakia	24	26	−2
Denmark	5	8	−3
Austria	6	9	−3
Slovenia	19	22	−3
Cyprus	17	21	−4
Portugal	9	14	−5
Türkiye	14	19	−5
Norway	1	16	−15
Poland	1	24	−23

. The DMUs that expose the most significant positive changes are Latvia (+16 places), Croatia (+10 places), and Finland (+7 places). Oppositely, the DMUs with the most significant negative changes are Portugal and Türkiye (−5 places), Norway (−15 places), and Poland, which dropped by 23 places.

Next in order, regarding RQ2, there is the observation of the convergence hypothesis in which firstly there is the examination of whether the convergence hypothesis holds for the entire sample. Then there is the investigation of the possibility of convergence of the clubs using the clustering algorithm proposed by Phillips and Sul. After these steps, the application of the log t regression takes place to test for convergence and the test’s output gives the coefficient, standard error, and t-statistic for log(t). Lastly, since the value of the t-statistic calculated as –108.851 is less than −1.65 in

Table 4. The log t test regression results for the entire sample.

Eco-Efficiency
B	−1.122
standard error	0.010
t-statistic	−108.851

, the null hypothesis of convergence is rejected at the 5% level. Now we have to identify subgroups converging their steady states in eco-efficiency since the whole sample convergence is rejected.

Based on the results of

Table 5. Club convergence results and a possible club merging.

Initial Classification (No. of Countries)	Club Members	b	t-Statistic	Test the Merging	b	SE	t-Statistic
Club 1 (4)	Croatia, Finland, Ireland, Latvia	0.523	2.620	Club 1 + 2	0.541	0.211	2.564
Club 2 (2)	Sweden, Switzerland	0.040	0.290	Club 2 + 3	−0.444	0.033	−13.521
Club 3 (6)	Austria, Belgium, Denmark, Estonia, Germany, Italy	0.807	5.060	Club 3 + 4	−0.709	0.034	−20.740
Club 4 (9)	Czechia, France, Greece, Lithuania, Netherlands, Norway, Portugal, Spain, Türkiye	0.057	0.471	Club 4 + 5	−0.642	0.040	−15.964
Club 5 (6)	Cyprus, Hungary, Poland, Romania, Slovakia, Slovenia	0.060	0.611

, the initial Clubs 1–2 and 2–3 can be fused into larger convergent clubs. It should be noted that even though the t-stat of the Club 2+3 merging is below −1.65, the algorithm shows that they can be merged in an intermediate loop.

Consequently, the final club classification is presented in

Table 6. Final club classifications.

Final Classification (No. of Countries)	Club Members	b	t-Statistic
Club 1 (12)	Austria, Belgium, Croatia, Denmark, Estonia, Finland, Germany, Ireland, Italy, Latvia, Sweden, Switzerland	0.190	2.139
Club 2 (9)	Czechia, France, Greece, Lithuania, Netherlands, Norway, Portugal, Spain, Türkiye	0.057	0.471
Club 3 (6)	Cyprus, Hungary, Poland, Romania, Slovakia, Slovenia	0.060	0.611

. Initially, there were five groups, but they have been consolidated into three final clubs. The first final club combines the first three initial clubs and includes a total of 12 countries. The t-stat is above the value –1.65 in all cases; thus, there is no further merging, and the three clubs are the final possible classification of the countries based on their eco-efficiency. Moreover, there is no absolute convergence. In the plastics and rubber industry, there is only conditional convergence (0 ≤ b ≤ 2), as depicted in Appendix D in Figure A2.

The exact convergence of initial and final clubs is provided in Supplementary Material in Figures S1 and S2. Essentially, by observing the final club classification, one might determine that the first final club comprises the countries in the axis Sweden–Italy by containing all the Central European countries from North to South. Furthermore, the second club consists of the Western European countries as well as the Czech Republic, Lithuania, Greece, and Türkiye, while the third final club contains mainly Eastern European countries and Cyprus. The analysis shows that the 27 European countries are far from fully coordinating their regulatory measures to achieve a common action plan for the CE. On the contrary, the existence of the three final clubs shows the need for individual and targeted action plans with the prospect of gradual convergence in the long term and the formulation of a single policy.

Bearing in mind that there is no absolute convergence in the European rubber and plastics industry, an issue that might be raised is how it is possible to create these prerequisites for such an achievement. In recent years, Karali et al. [16] signaled the need for a global plastics treaty that is under negotiation in international forums; accordingly, it is possible that a future agreement on a plastics treaty might be a robust foundation for a common international legally binding instrument that will converge climate-related and plastic-oriented policies towards a more sustainable future. Should each country, especially in Europe, converge to a common institutional framework, then a better convergence in eco-efficiency might be at play.

5. Conclusions and Policy Implications

The present paper employs a sequential analysis of the European rubber and plastics industries. RQ1 evaluates the eco-efficiency performance of 27 European industries as DMUs, revealing varying system dynamics as some countries (e.g., Poland and Norway) exhibit significant fluctuations, highlighting the need for tailor-made policies. Most countries studied are in line with the EU’s circular economy targets, as confirmed by the non-parametric test of independence. On this account, the homogeneity of policy at the EU level provides a common framework for action and impetus, as confirmed by the non-parametric test of independence. It follows from this empirical evidence that the sample under consideration consists of DMUs with similar characteristics and similar technology levels, and therefore, output-oriented efficiency measures are used in the analysis. RQ2 pinpoints that the convergence analysis shows three final clubs. Firstly, Club 1 comprises DMUs of Central and Northern Europe (axis of Sweden to Italy); secondly, Club 2 consists of mainly DMUs of Western Europe as well as the Czech Republic, Lithuania, Greece, and Türkiye, while Club 3 contains mainly Eastern European countries and Cyprus. Conditional convergence indicates that countries with similar eco-efficiency levels converge within their groups but not toward a common steady state across all groups, as national characteristics like institutional frameworks and macroeconomic policies influence their trajectories.

Actions for common convergence should target these different system dynamics and convergence trajectories in the European rubber and plastics industry; therefore, the policy implications should focus on the production factors that can lead to a dual change: (i) cleaner production based on CE principles (e.g., digitalization based on Industry 4.0, ICT, and IoT monitoring) and (ii) improved production that leads to better eco-efficiency performance. Policymakers should invest in research and development to deal with energy intensity by targeting eco-efficiency in capital and labor factors through inter-regional collaboration as noted before. In terms of capital, most of the Central and Western European countries seem to invest thoroughly in novel technologies; however, Eastern and Southern European countries have feeble performances, possibly also due to factors such as the financially turbulent era in the previous decade (e.g., COVID-19). Labor productivity can rise through extensive vocational training of the employees in the rubber and plastics industries. Both public and private initiatives have to collaborate to educate employees either on cleaner production factors (e.g., novel recycling standards) or novel technologies that can boost their productivity (e.g., robotics and emerging technologies). Other examples include the co-development of projects that are aligned with SDGs and the promotion of knowledge sharing between stakeholders. Overall, the cornerstone of public and private sector collaboration is to strengthen policy and regulatory support to incentivize stakeholders’ engagement and minimize the regulatory risks, for example, to alleviate industrial performance from stricter environmental policies and from extensive competition with other regions (e.g., Americas and Asia).

The research limitation of this study lies in the lack of waste-related data at the sectoral level for the rubber and plastic industry, as a significant part of plastic pollution is the waste generated, but there is no disaggregation for such data to the best of our knowledge. Nevertheless, regarding the air pollution aspect of plastic pollution, the present paper considers not only CO₂ emissions but also other important GHGsW. For this reason, the inclusion of data that are at a national level (e.g., plastic waste generation) was omitted because it would deviate from the focus of the present analysis, so the present analysis incorporated only industry-relevant data. Future research might focus on other regions and other sectors, especially if waste-related data become available on the ISIC level; then, a more inclusive model might also deal with plastic waste.

To recapitulate, the European rubber and plastics industry is one of the core industrial sectors in terms of profitability and in creating secure jobs; hence, the industrial development of this sector is pivotal for Europe in competing with other regions. Moreover, a discussion on having a global plastics treaty is currently underway, and Europe has paved the way for cleaner production based on circular economy policies against the plastic pollution that threatens the environment. Therefore, the European rubber and plastics industry can not only produce high-quality products but also provide environmentally friendly services to its customers. Lastly, the European rubber and plastics industry, through capital, energy, and labor-oriented productivity measures, can cover the prerequisites of SDGs related to sustainable industrialization that contributes to a climate-neutral, resource-efficient, and circular future.