Does E-Waste Recycling Promote Environmental Quality in the EU? E-Waste Policy-Oriented Empirical Analysis for SDGs 12 and 13

Abstract

This study tackles the issue using an innovative, multidimensional approach that considers economic indicators and the e-waste problem triggered by technological developments, while evaluating environmental sustainability. The original contribution lies in being the first study to examine e-waste recycling rates through the lens of the “load capacity curve hypothesis” (LCC_h). Empirical scrutiny of the top 10 European Union (EU) countries that generated the highest e-waste from 2007 to 2022 verified the validity of the LCC_h. The “Environmental Kuznets Curve Hypothesis” (EKC_h) is confirmed for Carbon dioxide (CO₂) emissions, but not Ecological Footprint (EF). Moreover, in lieu of e-waste density, elevated e-waste e-recycling rates have been shown to enhance the ecological structure. Empirical evidence suggests that prolonged use of e-devices, along with reduced waste and increased recycling rates, is essential for improving environmental quality.

1. Introduction

The nexus between solid waste and the environment has been discussed extensively in the literature [1,2,3]. Additionally, many studies [4,5,6] that focus on e-waste and address the issue using environmental degradation (ED) indicators like CO₂ and EF for various countries and country groups; nonetheless, not as many as waste due to advancements in technology and transformation. Yet, no study has been found to address the issue through ED indicators such as CO₂ and EF [7,8] and environmental quality (EQ) indicators such as Load Factor Capacity (LCF) for e-waste and e-waste recycling rates. In this regard, this paper seeks to contribute to the literature in three significant ways.

(i) Alongside the advancement of technology, the European continent has been one of the regions that has seen the fastest shifts and alterations in consumer habits. From this standpoint, the study focuses on the top ten EU countries with the highest levels of e-waste as of 2022. To provide broader insights, the sample comprises countries that dominate Europe in terms of population and development, such as France, Germany, Spain, and Italy. The study aims to reveal the findings in countries where environmental sensitivities are at the forefront with such a sample. (ii) This study investigates the EKC_h developed by [9] over the ED indicators EF and CO₂, and the LCC_h proposed by [10] over the LCF variable suggested by [11], which serves as an EQ indicator.

No study has been found in the literature that evaluates e-waste in terms of the LCC_h. The novelty of this study lies in its being the first to address e-waste through the LCC_h. (iii) Due to the concern that including income squared in the same model along with income in testing the EKC_h and LCC_h may cause multicollinearity problems, the present study adopts the approach of [12] based on the comparison of short-run and long-run elasticities of GDP per capita. Hence, the study seeks to conduct a comprehensive sustainability analysis of e-waste management by considering the environmental impacts of e-waste volumes and e-waste recycling rates. To this end, along with CO₂ emissions and EF indicators, a relatively newer and more comprehensive indicator, LCF, which considers both the demand and supply sides of the environment [13], is utilised to achieve the study’s target. In light of this, incorporating e-waste indicators into models using an innovative approach will provide valuable information for establishing the EU e-waste policy framework and for evaluating the success of e-waste management in the context of the SDGs for EU countries.

In the remainder of the paper, Section 2 presents the study’s theoretical background, including the empirical literature on the environmental impact of e-waste quantity and recycling. This section also identifies a gap in the existing literature. Section 3 provides detailed information on the data, models, and methodology. Section 4 includes the empirical results and discussion, while Section 5 concludes with specific policy recommendations on e-waste management based on the findings.

2. Literature Review

2.1. Theoretical Framework

With the rapid development of technology, many goods and services that were not considered essential in previous years are now among the fundamental needs. Today, humans rely on numerous electrical and electronic devices that were not in use in the past [14]. Modified versions of the same products, such as mobile phones, computers, and televisions, have been purchased on multiple occasions. In particular, the global proliferation of electric vehicles in recent years is considered a significant development for environmental sustainability. However, this transformation raises e-waste management issues. Notably, the failure to properly recycle lithium-ion batteries at the end of their life cycle poses a serious potential environmental risk. E-waste is a national asset for countries because it contains rare elements, such as gold, silver, palladium, cobalt, and copper.

The same waste also contains harmful heavy metals, such as lead, mercury, barium, and beryllium, that are harmful to both human and environmental health [15]. Therefore, failure to manage these e-wastes properly not only wastes resources but also creates severe pressures on human health and well-being by destroying natural environments and resources. On this basis, the work herein draws attention to the e-waste problem caused by modern technological consumption, emphasising the importance of addressing the issue across environmental and economic dimensions. Furthermore, it aims to provide policymakers with insights into e-waste management through a multidimensional analysis of environmental sustainability.

Given the increasing rate of e-waste, the e-waste recycling rate has the potential to create a significant leverage effect to achieve the world’s net-zero emissions target by 2050 [16]. Specifically, this study directly contributes to SDGs 12 and 13 (The corresponding

Table A1. SDGs and E-Waste Relations.

SDG	SDG Title	Targets	Link	Channels of Impact	Policy-Relevant Explanation
SDG 12	Responsible Consumption and Production	12.2 12.4 12.5	Direct	Resource efficiency, safe management of hazardous waste, and recycling rates	E-waste recycling directly promotes the sustainable use of natural resources (Target 12.2), ensures environmentally proper management of hazardous waste (Target 12.4), and contributes to waste reduction through recycling and reuse (Target 12.5).
SDG 13	Climate Action	13.2	Direct	Emission reduction, energy savings, and climate policy integration.	Recycling e-waste reduces energy demand relative to primary material extraction, reducing CO2 emissions and supporting the integration of climate change mitigation into national policies and strategies.
SDG 3	Good Health and Well-being	3.9	Indirect	Reduced exposure to toxic substances, public health protection	Proper e-waste management limits exposure to hazardous materials such as heavy metals, thereby reducing health risks associated with air, water, and soil pollution.
SDG 8	Decent Work and Economic Growth	8.4 8.5	Indirect	Green employment and sustainable economic growth	The development of formal e-waste recycling systems enhances resource efficiency (Target 8.4) and creates green job opportunities, supporting inclusive and decent employment (Target 8.5).
SDG 9	Industry, Innovation and Infrastructure	9.4	Indirect	Clean technologies and sustainable industrial processes	Investment in e-waste recycling infrastructure encourages the adoption of cleaner technologies and supports the transition toward more sustainable industrial production systems.
SDG 11	Sustainable Cities and Communities	11.6	Indirect	Urban waste management and pollution reduction	Effective e-waste recycling reduces the environmental impact of cities by improving urban waste management and limiting pollution generated from improper disposal.
SDG 14	Life Below Water	14.1	Indirect	Prevention of marine pollution and ecosystem protection	Proper handling and recycling of e-waste help prevent the release of toxic substances into aquatic ecosystems, contributing to reduced marine pollution.

is presented in the Appendix A to illustrate these connections) and indirectly to SDGs 3, 8, 9, 11, and 14. Regular monitoring, measurement, and setting of policy targets for e-device waste volumes are critical to emphasise their economic benefits and their potential to protect natural resources and mitigate the effects of the climate crisis. The primary objective of this paper is to stress that achieving environmental sustainability requires consideration not only of economic factors but also of current environmental risks, such as e-waste, arising from technological change and transformation [17]. Embedded in SDG 12 and the circular economy vision [18,19], accurately measuring and assessing e-waste recycling [20] and its environmental impacts will provide a better understanding of the role of e-waste management on environmental indicators and provide valuable information to policymakers in a leading supra-governmental structure in terms of environmental regulation, such as the EU.

Approximately 62 billion kilograms of e-waste were generated worldwide in 2022, corresponding to an average of 7.8 kg of e-waste per person. However, only one-fifth of this waste was recycled [21]. The low recycling rate of electronic waste can be attributed to multiple factors. The main challenges can be summarised as follows. Lack of advanced and efficient recycling technologies to safely recover valuable materials. High processing costs compared to the market value of recovered materials. Unsafe and inefficient informal-sector activities reduce material recovery rates. Low collection rates and inefficient take-back systems. Strict handling requirements increase operational costs. But, both the production and recycling of e-waste have increased over time; even so, the increase in production is approximately 5 times that in recycling. If urgent action is not taken to address this issue, a point of no return for the environment will be reached. When the numerical data for the top 10 EU countries with the highest e-waste density are presented in Figure 1, the amount of e-waste per capita has increased in all countries except Norway from 2007 to 2022. Nevertheless, it should not be overlooked that Norway still has the highest amount of e-waste among these countries. The countries with the highest increases in the same period are Poland, France, and Czechia. This situation can be explained by factors such as digitalisation, rising living standards, and the constant replacement of electronic devices with newer models in these countries, which have a GDP per capita above the world average. Although Spain and France had the lowest e-waste per capita in 2022, their levels are well above the world average. Except for France, where the most significant increase occurred, Poland and the Czech Republic are also among the top two in terms of increases in e-recycling rates. This situation shows that e-waste management systems in these countries have improved, and social awareness has increased. As shown in Figure 2, the countries in the study are well above the global average for e-recycling rates, and their e-waste volumes are similarly high. In the relevant period, e-waste recycling rates increased in all countries except Spain, Belgium, and the Netherlands. Despite the increase in e-waste density, the decrease in recycling rates in some countries, such as Spain, Belgium, and the Netherlands, is alarming for environmental sustainability under SDG 12 (responsible production and consumption) and SDG 13 (resource efficiency).

The EKC_h explains the relationship between ED and GDP per capita: a positive relationship exists up to a threshold, and a negative relationship thereafter. This hypothesis is inverse-U-shaped, stating that ED increases in the early stages of economic growth but then decreases after income reaches a certain level [9]. Unlike the EKC_h, the LCC_h posits a U-shaped relationship between EQ and income, as shown in Figure 3. Under this hypothesis, increasing income reduces EQ below a particular threshold. However, once this turning point is reached, increased income is associated with improvements in EQ [10]. Thus, the LCC_h predicts a negative relationship at low-income levels and a positive relationship at high-income levels, revealing the bidirectional dynamics between environment and economy.

LCF, the fundamental indicator for assessing the LCC_h, is calculated by dividing biocapacity by ecological footprint and serves as an indicator of environmental sustainability. This ratio indicates the extent to which a country’s environmental supply capacity is sufficient in relation to its environmental demand. If this ratio, namely LCF, is greater than one, it indicates that ecological capacity meets environmental demand; in other words, it provides an ecological surplus. In contrast, if the ratio is less than one, it indicates that ED exceeds the ecological capacity and an ecological deficit occurs [13]. As shown in Figure 4, the countries in the study that saw the most tremendous improvement in LCF environmental sustainability between 2007 and 2024 are Spain, with a 57.8% increase; Italy, with a 40.8% increase; and Czechia, with a 31.6% increase.

Interestingly, improvements were more limited for Austria and Belgium. This situation implies that environmental pressure continues in these countries. In addition, Norway, with the highest LCF value, stands out as having the strongest ecological capacity.

2.2. Empirical Literature

Although many different variables are used to explain ED, GDP per capita stands out as the most frequently used remarkable macroeconomic variable. Undoubtedly, the theoretical background for including GDP per capita as an explanatory variable in models is based on the EKC_h. First-generation studies that focused solely on CO₂ emissions as an indicator of ED, Terzi and Pata [24], were severely criticised for failing to account for individual effects. To overcome this issue, EF, a better indicator of ED, was predominantly used in second-generation studies. Another study by Siche et al. [11] noted that EF alone was insufficient to evaluate ecological sustainability and proposed the LCF variable. Based on this proposal, Dogan and Pata [10] argued that economic growth could reduce environmental pressure above a certain income level and increase nature’s self-renewal capacity. They opened the door to third-generation studies by introducing the LCC_h. In recent years, the LCF indicator has been increasingly used in third-generation studies in environmental economics as an alternative and more comprehensive measure for assessing environment-economy interactions. For example, Jin et al. [25] for Germany, Pata and Kartal [26] for South Korea, Pata and Ertugrul [27] for India, and Wang et al. [28] for 146 countries found the LCC_h to be valid. Contrary to these studies, the number of studies that did not find the LCC_h valid. Among these studies, Ulussever et al. [29] for six Middle Eastern countries, Pata et al. [30] for Portugal, and Çamkaya et al. [31] for France, found the LCC_h invalid.

Unlike the previous studies, some papers examining the EKC_h and focusing on e-waste have revealed significant findings in recent years. An inverted-U-shaped relationship between GDP per capita and e-waste production, Boubellouta and Kusch-Brandt [4] was found for the EU28+2 countries from 2000 to 2016. Extending this study to a larger global sample, Boubellouta and Kusch-Brandt [32] detected a similar inverted-U relationship at the global level, except in Asia. Later, Boubellouta and Kusch-Brandt [33] integrated the STIRPAT model with the EKC_h and confirmed the inverted-U curve once again by applying quantile regression to different e-waste components across EU28+2 countries for the period 2000–2015.

Studies on e-waste recycling rates in the literature are generally at the household level and focus on individuals’ knowledge, willingness, awareness, decisions, and behaviours regarding e-waste. Apart from these studies, Delcea et al. [34] addressed the issue in Romania by examining the impact of demographic and socioeconomic factors on e-waste recycling decisions, while Ylä-Mella [35] examined households’ awareness and perceptions of mobile phone recycling and reuse in Finland. Furthermore, studies examining e-waste recycling rates using macro data at the country or country-group level are limited. To the best of our knowledge, only two studies examine this relationship. Boubellouta and Kusch-Brandt [33] strongly support the notion that the relationship between economic growth and e-waste recycling rates follows an N-shaped curve for 30 European countries from 2008 to 2018. However, this study included the square and cube of income as independent variables, in addition to income, and employed a cubic methodological approach. Unlike Boubellouta and Kusch-Brandt [33], this paper adopts the linear approach of Narayan and Narayan [12], which compares short- and long-term elasticities, given concerns about multicollinearity with quadratic and/or cubic approaches. The second study, by Yılmaz and Koyuncu [5], examined the issue in the context of globalisation and found a U-shaped relationship between GDP per capita and the e-waste recycling rate. Both of these studies used e-waste recycling rates as dependent variables. Differing from these two studies, this study aims to contribute to the literature by examining the effects on environmental sustainability through a multidimensional approach, utilising the LCF variable as an innovative indicator alongside the CO₂ and EF variables. This analysis considers not only e-waste production but also e-waste recycling rates.

3. Materials and Methods

3.1. Data Set

In light of data covering the period from 2007 to 2022, the study uses Stata 17 and E-Views and focuses on the impact of e-waste density and e-waste recycling rates on environmental indicators, using an innovative approach alongside total natural resource rents and energy intensity variables commonly used in traditional environmental models. The study collects annual data and focuses on the top ten EU countries (Austria, Belgium, Czechia, France, Germany, Italy, Netherlands, Norway, Poland, Spain) with the highest e-waste density. The first three rows of

Table 1. Definitions of Variables.

Symbol	Description	Source
lcf	LCF (BIO x EF−1)	[23]
ef	EF (global hectares per capita)	[23]
co2	CO2 emissions per capita (metric tons per capita)	[36]
gdp	GDP per capita (constant 2015 US$)	[36]
natres	Total Natural Resources Rents (% of GDP)	[36]
e-recy	Recycling rate of waste of electrical and electronic equipment (WEEE) (Percentage)	[22]
e-inten	Energy intensity of GDP in chain-linked volumes (2010) (Kilograms of oil equivalent per thousand euros)	[22]
e-waste-den	WEEE by waste management operations (Tonne)/Total Population (e-waste per capita)	[22,36]

, where the definitions and data sources of the variables used in the study are shown, represent the dependent variables, and the remaining rows represent the independent variables.

3.2. Models

Consistent with [13], the study addresses the issue using the same independent variables across 3 models: 2 representing ED and 1 representing EQ. Thus, the comparability of the effects of e-waste on different environmental indicators is ensured. In addition, since the independent variables include GDP per capita, the study tests the LCC_h through Model 1, where LCF is the dependent variable, and the EKC_h through Models 2 and 3, where EF and CO₂ are the dependent variables.

Model 1: lnlcf = β₀ + β₁ lngdp + β₂ lne-recy + β₃ lne-inten + β₄ lne-waste-den + β₅ lnnatres + ε_i

(1)

Model 2: lnef = α₀ + α₁ lngdp + α₂ lne-recy + α₃ lne-inten + α₄ lne-waste-den + α₅ lnnatres + ε_i

(2)

Model 3: lnco₂ = λ₀ + λ₁ lngdp + λ₂ lne-recy + λ₃ lne-inten + λ₄ lne-waste-den + λ₅ lnnatres + ε_i

(3)

All variables in the models have undergone a natural logarithmic transformation. Thus, the effects of independent variables on the dependent variable are interpreted as elasticity. β, α, and λ used in the models show the flexibility coefficients of the independent variables.

3.3. Methodology

After determining the gap in the literature, it first identifies appropriate variables from a methodological perspective. Subsequent to these steps, it presents descriptive statistics. Then, the study tests for possible multicollinearity, especially among the independent variables, using the correlation matrix and VIF test. Thereafter, to test the cross-sectional dependency (CSD), the bias-corrected scaled LM test by Pesaran et. al. [37] and the CD test by Pesaran [38] are used; the homogeneity or heterogeneity of the slope coefficients is first suggested by Swamy [39] and then developed by Pesaran et. al. [37]. Then, the study employs the CIPS unit root test, developed by Pesaran [40], a second-generation unit root test. To estimate long- and short-term elasticities, the Panel ARDL-PMG method is used, following Hausman [41]. In addition, country-based short-term coefficients are compared. The DH panel causality approach [42] is employed to detect causality relationships between variables. In the final stage, the results are compared with empirical studies in the existing literature, interpreted, and followed by concrete policy recommendations for policymakers. The paper ends with a discussion of limitations and suggestions for future studies. The empirical methodological framework summarises all these processes and is visually structured in Figure 5 below.

This methodological process followed for panel data analysis aims to obtain consistent and unbiased estimates. In panels with numerous cross-sections and time dimensions, the relationships between variables cannot be explained by the time-series approach. These relationships are also affected by inter-country interactions and structural differences. Therefore, the analysis began by examining potential multicollinearity among the independent variables and testing the reliability of the model coefficients. Then, cross-sectional dependence tests were conducted to examine whether global shocks, such as economic crises and energy market shocks, have simultaneous effects across countries. Thus, the use of second-generation panel unit root tests was methodologically justified. Homogeneity tests of the slope were conducted to examine whether structural differences across countries affect long-term coefficients. In light of these findings, the CIPS unit root test, a second-generation test that accounts for cross-sectional dependence, was preferred. The panel ARDL-PMG approach, which allows for the simultaneous analysis of short-term and long-term relationships by enabling variables to be stationary at different orders of magnitude, was adopted as the primary estimation method for this study because it can reveal both long-term equilibrium relationships and country-specific short-term dynamics. Finally, the direction of relationships between variables was analysed using the Dumitrescu–Hurlin panel causality test, and the findings were further explored in terms of policy implications.

4. Results and Discussion

The descriptive statistics are shown in

Table 2. Statistical description.

	lnlcf	lnef	lnco2	lngdp	lne-recy	lne-inten	lne-waste-den	lnnatres
Mean	−1.0108	1.7028	2.0823	10.4237	4.3831	4.9180	2.1378	−1.4202
Median	−0.9545	1.7134	2.1484	10.5596	4.3944	4.8161	2.1819	−1.8596
Max.	0.3030	2.1225	2.5533	11.2621	4.5486	5.6980	3.1736	2.5351
Min.	−1.8476	1.3242	1.4470	9.1861	3.6687	4.3221	−0.3385	−4.0561
Std. Dev.	0.5731	0.1859	0.2452	0.4917	0.0924	0.3485	0.5356	1.7254
Skew.	0.4582	0.0468	−0.6826	−0.6345	−3.6008	0.7348	−0.8602	0.7289
Kurto.	2.8307	2.2995	2.5022	3.0376	26.9181	2.5634	5.7557	2.4464
J-Bera	5.3558	3.0797	13.0202	9.9393	3847.6420	14.4938	65.0812	14.9937
Prob.	0.0687	0.2144	0.0015	0.0069	0.0000	0.0007	0.0000	0.0006

. These values offer important insights into the environmental, economic, and e-waste conditions in the top ten EU countries that generate the most e-waste. When the average values of the environmental indicators are assessed, it becomes clear that countries with the highest e-waste production experience significant environmental pressures. The highest average values are found in the GDP per capita and energy intensity variables. In addition, although the average e-recycling rate is high, this variable shows high skewness and excessive kurtosis, suggesting that some countries have very low e-waste recycling rates. This finding reveals that EU recycling policies are heterogeneous, and countries exhibit significant disparities in their e-waste recycling performance.

When the correlation matrix results in

Table 3. Correlation matrix.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	VIF	VIF−1
(1) lnlcf	1
(2) lnef	−0.2279 ***	1
(3) lnco2	−0.1048	0.6564 ***	1
(4) lngdp	0.1781 **	0.4208 ***	0.0468	1					4.7800	0.2094
(5) lne-recy	0.1119	−0.1375 *	0.0206	0.1612 *	1				3.2600	0.3070
(6) lne-inten	−0.2540 ***	0.1282	0.3908 ***	−0.8163 ***	−0.2266 ***	1			2.7900	0.3580
(7) lne-waste-den	0.3275 ***	0.2791 ***	0.0963	0.7377 ***	0.3882 ***	−0.5793 ***	1		1.2800	0.7794
(8) lnnatres	0.5301 ***	0.0805	0.4920 ***	−0.0114	0.0447	0.1252	0.1457 *	1	1.1000	0.9125

Note: ***, ** and * denote 1%, 5%, and 10% significance, respectively.

are examined, one striking finding is that the coefficients between e-recycling and other environmental indicators are very low. This result reveals that the effect of EU e-waste recycling policies on environmental indicators has not yet been fully reflected. The low correlation among the independent variables in the study indicates that there is no multicollinearity. VIF is applied to check the robustness of the findings. The VIF values of all variables range from 1.1 to 4.78. These values are well below the generally accepted limit of 10 [43]. Therefore, these findings again suggest that there is no multicollinearity among the independent variables, indicating that their relationships do not mask their effects on the dependent variable.

CSD offers informative guidance on the appropriate choice of generation unit root tests and estimation methods.

Table 4. CSD and Homogeneity test results.

CSD	Pesaran CD		Bias-Corrected Scaled LM
Variables	t-Stat.	p-Value	t-Stat.	p-Value
lnlcf	18.2635 ***	0.0000	32.0252 ***	0.0000
lnef	20.7068 ***	0.0000	42.4350 ***	0.0000
lnco2	20.6170 ***	0.0000	45.0490 ***	0.0000
lngdp	18.5662 ***	0.0000	37.9162 ***	0.0000
lne-recy	3.61528 ***	0.0003	9.9476 ***	0.0000
lne-inten	24.1480 ***	0.0000	56.8713 ***	0.0000
lne-waste-den	10.3172 ***	0.0000	35.2441 ***	0.0000
lnnatres	13.9033 ***	0.0000	21.7118 ***	0.0000
Slope homogeneity	Δ Statistic	p-value	Δadj Statistic	p-value
Model 1 (lnlcf)	0.2310	0.8170	0.3770	0.7060
Model 2 (lnef)	0.8300	0.4060	1.3560	0.1750
Model 3 (lnco2)	0.0340	0.9730	0.0560	0.9550

Note: *** denotes 1% significance.

reflects the CSD and homogeneity test results. The applied tests clearly support CSD. This result implies a strong interaction among the top 10 EU countries with the highest e-waste density, suggesting that shocks in one of these countries can have a significant impact on others. In addition, it requires the use of second-generation unit root tests and estimation methods that consider CSD. In addition, the slope coefficients are homogeneous.

Based on the CIPS test results from the second-generation unit root tests presented in

Table 5. Unit Root test results.

CIPS	Level	First-Difference	Decision
	C&T	C&T
lnlcf	−2.4967	−4.7177 ***	I(1)
lnef	−2.2741	−4.0179 ***	I(1)
lnco2	−1.2222	−3.7701 ***	I(1)
lngdp	−1.4913	−3.1671 **	I(1)
lne-inten	−2.5967	−3.3909 ***	I(1)
lnnatres	−1.338	−3.6574 ***	I(1)
lne-waste-den	−2.7012 *	−3.6944 **	I(1)
lne-recy	−3.4235 ***		I(0)

Note: ***, ** and * denote 1%, 5%, and 10% significance, respectively.

, all the remaining variables, except the lne-recy variable, are I(1). lne-recy is I(0). Therefore, the variables are stationary at different levels, and none is I(2). As a result, while the dependent variable is I(1), the panel ARDL procedure can be run since the independent variables are I(0) or I(1).

In this study, which focuses on the effects of e-waste density and recycling on environmental indicators, coefficient estimates were obtained using the panel ARDL method. These results cover both short- and long-term effects and are presented in

Table 6. PMG test results.

PMG	Model 1 (lnlcf)		Model 2 (lnef)		Model 3 (lnco2)
(A) Long run	Coefficient	Prob.	Coefficient	Prob.	Coefficient	Prob.
lngdp	−0.6971 ***	0.0000	0.8646 ***	0.0000	0.6245 ***	0.0000
lne-recy	0.6130 ***	0.0000	−0.3644 ***	0.0000	−0.4889 ***	0.0000
lne-inten	−0.5474 ***	0.0000	1.6081 ***	0.0000	1.9573 ***	0.0000
lne-waste-den	0.0240	0.2210	0.1781 ***	0.0000	0.2984 ***	0.0000
lnnatres	−0.0382 ***	0.0000	−0.0365 ***	0.0000	0.0180 ***	0.0000
(B) Short run	Coefficient	Prob.	Coefficient	Prob.	Coefficient	Prob.
ECT	−0.6072 ***	0.0000	−0.5286 ***	0.0000	−0.2022 ***	0.0010
Δlngdp	−0.7179 **	0.0200	0.8318 ***	0.0070	1.0159 ***	0.0000
Δlne-recy	−0.1909	0.2310	0.4438 ***	0.0000	0.1833 **	0.0180
Δlne-inten	−0.1657	0.5950	−0.4960 *	0.0510	0.7595 ***	0.0000
Δlne-waste-den	0.1407 **	0.0230	−0.0951 ***	0.0000	−0.0394	0.1780
Δlnnatres	−0.0050	0.8050	0.0339 *	0.0650	−0.0102	0.1510
C	3.8200 ***	0.0000	−7.4320 ***	0.0000	−2.5577 ***	0.0010
(C) PMG/MG	Statistic	Prob.	Statistic	Prob.	Statistic	Prob.
Hausman Test	0.0700	0.9999	0.7700	0.9789	1.3800	0.9269
(D) Panel overall	Model 1-LCCh		Model 2-EKCh		Model 3-EKCh
Decision	✔		✖		✔

Note: ***, ** and * denote 1%, 5%, and 10% significance, respectively.

. The fact that the independent variables have different integration degrees, provided that the dependent variables are I(1), and that this method is the best-fitted in capturing long-term relationships, even in small samples [44], supports the choice of this approach. The preference between the PMG and MG estimators is evaluated using the Hausman test [41]. The results showed that the Hausman test statistics for all three models had p-values greater than 0.05. Therefore, the PMG method is more suitable than the MG method. Since the error-correction term coefficients are negative and statistically significant in all three models, a long-term relationship exists among the variables in each model. As indicated by the error-correction term coefficient in Model 1, approximately 60% of the imbalances that occur in one year are corrected in the following year, and the system returns to balance after approximately 1.5 years. The results are similar for Model 2, but the period to reach equilibrium is approximately six months shorter. In Model 3, the period to reach equilibrium is approximately five years, as only approximately 20% of the imbalances that occur in one year are corrected in the following year.

The study adopts the approach of Narayan and Narayan [12], which involves comparing the short- and long-term elasticities of the GDP per capita variable to test the EKC_h and LCF hypotheses. The dependent variable is LCF in Model 1; the short-term and long-term elasticities of the GDP per capita variable are negative and statistically significant. A 1% increase in the GDP per capita variable will decrease the LCF variable representing EQ by approximately 0.70% in the long term, and by approximately 0.72% in the short term. Therefore, the long-term coefficient of GDP per capita is greater than the short-term coefficient of GDP per capita. We interpret this result as indicating that the long-term effect of GDP per capita on EQ is greater than the short-term effect. Hence, the LCC_h is valid for Model 1. In Model 2, the dependent variable is EF, an indicator of ED; the long-term coefficient for GDP per capita is greater than the short-term coefficient. A 1% increase in the GDP per capita variable will result in an approximately 0.87% increase in the EF variable representing ED in the long term. In comparison, it will increase it by approximately 0.83% in the short term. We interpret this result as indicating that the effect of GDP per capita on ED is more dominant in the long term. Thus, the EKC_h is not valid for Model 2.

The short- and long-term elasticities of the GDP per capita variable in Model 3, where the dependent variable is CO₂ as another indicator of ED, are positive and statistically significant. The long-term coefficient of GDP per capita is smaller than the short-term coefficient of GDP per capita. A 1% increase in the GDP per capita variable will result in a 0.63% increase in the CO₂ variable in the long term, while it will result in a 1.02% increase in the short term. We interpret this result as indicating that the long-term effect of GDP per capita on ED is reduced. Accordingly, the EKC_h is valid for Model 3.

The coefficient for the e-recycling variable is statistically significant in all models and is positive in Model 1 and negative in Models 2 and 3. This result shows that as the e-recycling rate increases, EQ increases and ED decreases. For Model 1, a 1% increase in e-recycling rates results in a 0.61% increase in the LCF variable representing EQ. For Models 2 and 3, a 1% increase in e-recycling rates reduces the EF and CO₂ variables reflecting ED by approximately 0.36% and 0.49%, respectively. The variable for e-waste per capita is statistically significant in Models 2 and 3 but not in Model 1. This positive coefficient in Models 2 and 3 indicates that ED deepens as e-waste per capita increases. Because a 1% increase in e-waste density for Models 2 and 3 increases the EF and CO₂ variables by approximately 0.18% and 0.30%, respectively. The energy intensity variable has the most tremendous impact on ED indicators. A 1% increase in energy intensity increases the EF and CO₂ variables by approximately 1.61% and 1.96%, respectively. It decreases the LCF variable by 0.55%. Parallel to literature, as energy intensity increases, EQ deteriorates, and ED increases. Similarly, in the long term, a 1% increase in natural resource rents reduces LCF and EF by roughly 0.04%, while increasing CO₂ emissions by 0.02%. In the short term, however, the effect of this variable on environmental indicators is generally statistically insignificant. Therefore, it appears that the increase in natural resource rents weakens countries’ long-term ecological carrying capacity, negatively affects environmental sustainability, and puts pressure on EQ, particularly through carbon-intensive production structures. This situation poses serious threats to environmental sustainability and contradicts the SDGs.

When the country-based short-term results in

Table A2. Country-specific Short-Term Results for Model 1-lcf.

Model 1		ECT	Δlngdp	Δlne-recy	Δlne-inten	Δlne-waste-den	Δlnnatres	C
Austria	Coef.	−0.5563 **	−0.4310	−0.0288	−0.1083	−0.0440	−0.0419	3.1634 **
	Prob.	0.0130	0.2250	0.9430	0.6880	0.4480	0.2340	0.0430
Belgium	Coef.	−1.4624 ***	1.0151 **	0.6450 **	−0.0638	0.4713 **	0.0574 **	7.0045 **
	Prob.	0.0000	0.0270	0.0160	0.7510	0.0140	0.0370	0.0160
Czechia	Coef.	−0.9875 ***	−0.0868	−0.4911 ***	0.9765 **	0.2652 ***	−0.0222	5.4809 **
	Prob.	0.0000	0.8240	0.0000	0.0380	0.0000	0.1620	0.0170
France	Coef.	−0.5398 ***	−0.2999	−0.4305 ***	0.2230	0.0120	−0.1032 *	3.0776 *
	Prob.	0.0060	0.2060	0.0050	0.4790	0.7870	0.0740	0.0520
Germany	Coef.	−0.5036 ***	−0.4083	−0.1771	−0.3293	0.2822 *	−0.0524	2.6180 *
	Prob.	0.0040 *	0.4240	0.4160 ***	0.2680	0.0550 *	0.1120 ***	0.0500
Italy	Coef.	−0.0604	−0.9088 ***	−0.0017	0.3568	−0.0725	−0.0008	0.2972
	Prob.	0.7580	0.0030	0.9880	0.4570	0.3660	0.9750	0.7420
Netherlands	Coef.	−1.0708 ***	−1.9025 ***	−0.6908 **	0.8228	0.2342	0.0641 *	5.1881 **
	Prob.	0.0000	0.0040	0.0290	0.1300	0.2360	0.0610	0.0180
Norway	Coef.	−1.0402 ***	−1.7091 ***	−0.9815 ***	0.2990 **	0.3643 **	−0.0683 ***	7.1095 ***
	Prob.	0.0000	0.0010	0.0000	0.0100	0.0180	0.0060	0.0000
Poland	Coef.	0.0199	−1.8366 ***	0.0429	−0.9015 ***	0.0159	−0.0051	−0.0597
	Prob.	0.8330	0.0000	0.2320	0.0010	0.5890	0.6940	0.9060
Spain	Coef.	−0.1006	−1.2773 ***	0.3826 *	−2.6631 ***	−0.0651	0.1114 ***	0.4932
	Prob.	0.6100	0.0010	0.0550	0.0000	0.3750	0.0020	0.6310

Note: ***, ** and * denote 1%, 5%, and 10% significance, respectively.

,

Table A3. Country-specific Short-Term Results for Model 2-ef.

Model 2		ECT	Δlngdp	Δlne-recy	Δlne-inten	Δlne-waste-den	Δlnnatres	C
Austria	Coef.	−0.3664 ***	−0.1025	1.2117 ***	−0.2432	−0.0327	0.1198 ***	−4.9895 ***
	Prob.	0.0010	0.7610	0.0020	0.2840	0.5940	0.0000	0.0020
Belgium	Coef.	−0.0476	1.3013 ***	0.4109 ***	−0.0387	−0.0715	0.0229 *	−0.6819
	Prob.	0.6030	0.0000	0.0090	0.7430	0.5400	0.0860	0.5930
Czechia	Coef.	−1.1335 ***	−0.5245 ***	0.3837 ***	−1.8716 ***	−0.1644 ***	0.0652 ***	−15.8017 ***
	Prob.	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000	0.0000
France	Coef.	−0.2620	1.0982 **	0.3660	0.3800	0.0499	−0.0678	−3.6388
	Prob.	0.2960	0.0480	0.1610	0.4970	0.5250	0.5170	0.2970
Germany	Coef.	−0.3918 ***	0.7291 ***	−0.0149	0.3175 **	−0.0905 **	0.0212 **	−5.3951 ***
	Prob.	0.0000	0.0000	0.8030	0.0100	0.0260	0.0250	0.0000
Italy	Coef.	−0.4639 **	0.7149 ***	0.0733	−0.3151	−0.0680	−0.0182	−6.2528 **
	Prob.	0.0210	0.0020	0.3500	0.3750	0.3710	0.3620	0.0250
Netherlands	Coef.	−0.7607 ***	1.5777 ***	0.5372 ***	−0.9055 **	−0.0773	−0.0171	−10.5325 ***
	Prob.	0.0000	0.0000	0.0010	0.0110	0.5140	0.3040	0.0000
Norway	Coef.	−0.7338 ***	2.4301 ***	0.8775 ***	−0.2758 ***	−0.1835	0.0682 ***	−10.1089 ***
	Prob.	0.0000	0.0000	0.0000	0.0210	0.2670	0.0080	0.0000
Poland	Coef.	−0.3798 ***	0.0486	0.0500 **	−0.6057 ***	−0.0662 ***	0.0603 ***	−5.2185 ***
	Prob.	0.0000	0.7800	0.0170	0.0020	0.0010	0.0000	0.0000
Spain	Coef.	−1.2693 ***	−0.1647 ***	0.4750 ***	−1.9471 ***	−0.1869 ***	0.0806 ***	−17.2442 ***
	Prob.	0.0000	0.4200	0.0000	0.0000	0.0000	0.0000	0.0000

Note: ***, ** and * denote 1%, 5%, and 10% significance, respectively.

and

Table A4. Country-specific Short-Term Results for Model 3-CO₂.

Model 3		ECT	Δlngdp	Δlne-recy	Δlne-inten	Δlne-waste-den	Δlnnatres	C
Austria	Coef.	−0.1127 *	1.0424 ***	0.5470 **	1.0040 ***	0.0532	0.0160	−1.3989 *
	Prob.	0.0730	0.0000	0.0160	0.0000	0.1260	0.4130	0.0720
Belgium	Coef.	0.1115	1.1438 ***	0.6141 **	0.8181 ***	0.1453	−0.0152	1.4284
	Prob.	0.5510	0.0080	0.0140	0.0010	0.4480	0.4350	0.5550
Czechia	Coef.	−0.5783 ***	0.5703 ***	0.0323 *	−0.4176 ***	−0.0648 ***	−0.0099 ***	−7.5152 ***
	Prob.	0.0000	0.0000	0.0910	0.0000	0.0000	0.0000	0.0000
France	Coef.	−0.2853 ***	0.6063 ***	0.0550	1.2410 ***	0.0381	−0.0457	−3.6558 **
	Prob.	0.0090	0.0060	0.7080	0.0000	0.2930	0.3720	0.0120
Germany	Coef.	−0.0666 **	0.8438 ***	0.0806	0.9714 ***	−0.1690 ***	−0.0036	−0.8061 **
	Prob.	0.0270	0.0000	0.3870	0.0000	0.0070	0.8010	0.0280
Italy	Coef.	−0.1662 **	1.3222 ***	−0.0085	0.6326 ***	−0.0529	−0.0138	−2.0195 **
	Prob.	0.0430	0.0000	0.8690	0.0070	0.2340	0.3680	0.0410
Netherlands	Coef.	−0.2205	1.4835 ***	−0.0776	0.9423 ***	−0.1061	−0.0188	−2.7585
	Prob.	0.1370	0.0000	0.6020	0.0020	0.2330	0.3370	0.1380
Norway	Coef.	−0.0989 **	1.0025 *	0.4012 **	0.0023	−0.1060	0.0262	−1.2300 **
	Prob.	0.0190	0.0600	0.0210	0.9780	0.4940	0.2800	0.0180
Poland	Coef.	−0.3077 ***	0.9166 ***	0.0731 **	0.4158 **	−0.0528 **	0.0035	−3.9815 ***
	Prob.	0.0080	0.0000	0.0140	0.0410	0.0350	0.6920	0.0050
Spain	Coef.	−0.2970 ***	1.2280 ***	0.1163 **	1.9856 ***	−0.0786 ***	−0.0408 ***	−3.6400 ***
	Prob.	0.0000	0.0000	0.0300	0.0000	0.0020	0.0000	0.0000

Note: ***, ** and * denote 1%, 5%, and 10% significance, respectively.

are assessed overall, the countries with a statistically significant coefficient on the error correction term across all models are Austria, Czechia, Norway, and Germany. These results indicate a long-term balance relationship, suggesting that short-term imbalances will be corrected over time. For Model 3, which considers ED over CO₂, the coefficient on the error correction term is significant for all countries except Belgium and the Netherlands. The same coefficients are significant for all countries except Belgium and France for Model 2, where we consider ED over EF. For Model 1, the coefficients are predominantly significant. The shortest period for the imbalances to reach balance is Belgium in Model 1, Spain in Model 2, and Czechia in Model 3. The short-term e-waste recycling rate coefficients are positive for Belgium and Spain, indicating that increases in this variable are associated with higher EQ in these countries. For the remaining countries, the coefficient is either negative or statistically insignificant. Therefore, we construe that e-waste recycling policies in these countries are not yet sufficiently developed or have not yet fully realised their results. Short-term e-waste density coefficients are negative for Poland and Spain, indicating that increases in this variable are associated with reductions in EQ in these countries. For the rest countries, this coefficient is either positive or statistically insignificant. Norway is where all coefficients are statistically significant in the short term, while Czechia and Spain are the countries for Models 2 and 3. The overall findings are visually summarised in Figure 6 below.

The PMG estimator does not provide inferences regarding the causality relationship between variables. According to the D-H panel causality test results presented in

Table 7. D-H panel causality test.

Model 1 (lnlcf) H0	W-Stat.	Zbar-Stat.	Prob.	Decision:
lngdp ⇏ lnlcf	3.6283 ***	3.8588	0.0001	lngdp ⇔ lnlcf
lnlcf ⇏ lngdp	2.6191 **	2.2552	0.0241
lne-recy ⇏ lnlcf	1.3278	0.2030	0.8391	None
lnlcf ⇏ lne-recy	1.1041	−0.1523	0.8789	None
lne-inten ⇏ lnlcf	4.2605 ***	4.8634	0.0000	lne-inten ⇒ lnlcf
lnlcf ⇏ lne-inten	1.1790	−0.0333	0.9734	None
lne-waste-den ⇏ lnlcf	1.6174	0.6512	0.5149	None
lnlcf ⇏ lne-waste-den	2.3732 *	1.8429	0.0653	lnlcf ⇒ lne-waste-den
lnnatres ⇏ lnlcf	0.2497	−1.4886	0.1366	None
lnlcf ⇏ lnnatres	1.3006	0.1200	0.9045	None
Model 2 (lnef) H0	W-Stat.	Zbar-Stat.	Prob.	Decision:
lngdp ⇏ lnef	5.5818 ***	6.9631	0.0000	lngdp ⇔ lnef
lnef ⇏ lngdp	5.1193 ***	6.2282	0.0000
lne-recy ⇏ lnef	1.7115	0.8128	0.4164	None
lnef ⇏ lne-recy	3.3450 ***	3.4087	0.0007	lnef ⇒lne-recy
lne-inten ⇏ lnef	6.7866 ***	8.8776	0.0000	lne-inten ⇒ lnef
lnef ⇏ lne-inten	0.9451	−0.4050	0.6855	None
lne-waste-den ⇏ lnef	2.6505 **	2.2802	0.0226	lne-waste-den ⇔ lnef
lnef ⇏ lne-waste-den	2.6392 **	2.2624	0.0237
lnnatres ⇏ lnef	0.9822	−0.3675	0.7133	None
lnef ⇏ lnnatres	1.1165	−0.1619	0.8714	None
Model 3 (lnco2) H0	W-Stat.	Zbar-Stat.	Prob.	Decision:
lngdp ⇏ lnco2	5.4035 ***	2.5959	0.0094	lngdp ⇔ lnco2
lnco2 ⇏ lngdp	8.6612 ***	5.5820	0.0000
lne-recy ⇏ lnco2	1.8852	−0.6290	0.5293	None
lnco2 ⇏ lne-recy	4.9278 **	2.1599	0.0308	lnco2 ⇒ lne-recy
lne-inten ⇏ lnco2	2.5877	0.0149	0.9881	None
lnco2 ⇏ lne-inten	8.3154 ***	5.2650	0.0000	lnco2 ⇒ lne-inten
lne-waste-den ⇏ lnco2	3.1520	0.5049	0.6136	None
lnco2 ⇏ lne-waste-den	4.6135 *	1.8192	0.0689	lnco2 ⇒ lne-waste-den
lnnatres ⇏ lnco2	3.0423	0.3149	0.7528	None
lnco2 ⇏ lnnatres	1.9602	−0.5924	0.5536	None

Note: ***, ** and * denote 1%, 5%, and 10% significance, respectively.

, a bidirectional causality relationship is detected between GDP per capita and environmental indicators. This finding demonstrates that, just as economic growth impacts environmental indicators. Environmental indicators also influence the sustainability of economic activities. A one-way causality relationship exists between the energy intensity variable and environmental indicators. The direction of the relationship is from energy intensity to LCF and EF in Models 1 and 2, whereas it is from CO₂ to energy intensity in Model 3. The bidirectional causal relationship between e-waste intensity and EF indicates that these variables mutually influence one another. The causal relationship between ED indicators and the e-waste recycling rate indicates that society or policymakers respond to e-waste recycling policies in the face of ED. The absence of any causal relationship between e-waste recycling rates and ED indicators may indicate that e-waste recycling rates are insufficient.

When the Dumitrescu–Hurlin panel causality results are reviewed, it is generally observed that environmental indicators are primarily shaped by economic structure, energy use, and waste management policies. The strong, unidirectional causal effect of energy intensity on LCF demonstrates that energy efficiency policies play a critical role in EQ. In contrast, the lack of a significant causal relationship between e-recycling rate and e-waste intensity and LCF suggests that the intensity and recycling rates of electronic waste are insufficient to improve EQ on a scale or with sufficient effectiveness. The bidirectional causality between EF and e-waste intensity shows that environmental pressures shape e-consumption behaviour, while increasing e-waste further expands EF. Causal relationships from CO₂ emissions to e-waste intensity and energy intensity reveal that carbon-intensive production structures both increase energy use and accelerate e-waste generation. Thereby, it is evident that environmental sustainability cannot be achieved solely through policies based on increasing e-recycling rates. The study concludes that structural transformations to reduce energy intensity, a shift away from carbon-based production, and policies to prevent e-waste must be implemented simultaneously.

5. Conclusions

Today, rapid technological development and transformation make our lives easier and help prevent time loss by accelerating processes. The emergence of a new technological product every day reduces the use of other products, leading some to be withdrawn from circulation. As technology continues to develop rapidly, the amount of worldwide e-waste also increases. These wastes are also one of the factors contributing to environmental degradation. This issue, which falls within the scope of the SDGs, forms the backbone of the study. In furtherance of this objective, this paper is the first to evaluate the impact of e-waste consumption and recycling on EQ across the top 10 EU countries with the highest e-waste density within the LCC_h. The lack of a study specifically examining the impact of e-waste consumption and recycling on LCF highlights an important research gap in the literature. In this regard, the study aims to fill this research gap by testing the LCC_h alongside the EKC_h using panel data analysis techniques, with a focus on e-waste variables. The overall findings and inferences obtained from the Panel ARDL/PMG method are as follows: The LCC_h is valid for EU countries. In contrast, the EKC_h varies depending on the dependent variable. While the EKC_h is invalid for Model 2, where the dependent variable is EF, the EKC_h is valid for Model 3, where the dependent variable is CO₂. In the long term, the e-waste recycling rate has encouraging effects on EQ and improving effects on ED. Similar results hold for Models 2 and 3 regarding e-waste per capita, but not for Model 1. As e-waste density increases, ED increases. In line with the literature, as energy density increases, EQ decreases, while ED increases. The most significant adverse effect on ED is attributed to this variable. Since the effects and magnitudes of the short-term country-specific coefficients differ across countries, this implies that the effects of EU general environmental policies vary across countries.

The findings clearly demonstrate that e-waste management is a critical policy area for achieving environmental sustainability in EU countries. The analysis results show that increases in e-waste volumes lead to higher CO₂ emissions and EF indicators. Conversely, e-waste recycling rates play a limiting role in environmental pressures and support the LCF indicator, especially in the long term. These findings indicate that e-waste policies should focus not only on reducing waste volumes but also on increasing the efficiency of recycling infrastructure. In this vein, reassessing EU electronic waste policy sets, taking into account the LCF indicator, which reflects EQ more holistically, will make valuable contributions to achieving the SDGs.

Digitalisation and increasing reliance on technology harm the environment by increasing energy consumption and posing environmental risks, such as e-waste. Given the dependence of human beings on electrical and electronic devices, suggestions to reduce their use are unlikely to be very realistic at this point. Thus, specific policy suggestions for EU countries could be summarised as follows: Based on the analyses, it has been concluded that the long-term use of e-devices and their recycling into the economy are both economically and environmentally crucial. Proper recycling of lithium-ion batteries and electronic devices such as mobile phones, white goods, and computers will improve countries’ welfare. Additionally, it will undertake valuable tasks to support the SDG-3 health and quality of life target by isolating harmful substances that should not be released into the air, water, or soil. From this standpoint, within the scope of SDG-13 climate action, countries should enhance the capacities of existing recycling facilities that collect, separate, and process e-waste, and establish new ones. Furthermore, easy financing tools could be developed for e-waste recycling investments, and tax advantages might be provided. As outlined in SDG-12’s responsible production and consumption target, when the life of an electronic device ends for any reason, the manufacturer might be obliged to take it back and recycle it at scrap value. On the consumption front, applications could be developed to encourage the return of old products, such as mobile phones, computers, televisions, and white goods, which are frequently replaced. To illustrate, discounts or coupons might be offered to consumers who return old products. In addition, consumers who do not return old products and want to buy the new model, except those buying a product for the first time, might be penalised with an additional e-waste tax. Furthermore, to reduce negative externalities, these policy recommendations will make e-waste management a strategic tool for achieving sustainable development goals, thereby securing a fair and sustainable future for future generations. Consequently, we consider the effective management of e-waste not only an economic or environmental necessity, but also a development strategy directly related to many SDGs targets.

Limitations

Notwithstanding that it is the first paper to address the relationship between e-waste and LCF, it has several important limitations. We see the time dimension as a remarkable limitation. Specifically, the lack of recent data on “Total Natural Resources Rents” weakens the study’s timeliness. Additionally, since the analyses are applied to a relatively homogeneous country group, such as the EU, results from countries with different levels of development should be compared. Complementarily, the asymmetric effects of e-waste variables were not considered, and it was assumed that the effects of increases and decreases in these variables were equal. However, the effects of increases and decreases in these variables on environmental indicators may differ. From this point forward, future studies could provide meaningful insights into the relationships between LCF and e-waste by accounting for the asymmetric effects of e-waste-related variables across a broader, more heterogeneous country group.