Does ICT Exacerbate the Consumption-Based Material Footprint? A Re-Examination of SDG12 Challenges in the Digital Era Across G20 Countries

Abstract

Global resource depletion has intensified scrutiny on Sustainable Development Goal 12 (SDG12), where consumption-based material footprint serves as a critical sustainability metric. Despite its transformative potential, the paradoxical role of Information and Communication Technology (ICT) in resource conservation remains underexplored. This study adopts an extended STIRPAT model as the analytical framework. It employs the Method of Moments Quantile Regression to evaluate the non-linear effects of digitalization-related indicators and other influencing factors on material footprint. The analysis is conducted across different quantiles for G20 countries from 2000 to 2020. The results show that (1) ICT exhibits a substantial positive effect on consumption-based material footprint under all quantiles. This leads to an increase in the material footprint, hindering the G20’s progress toward achieving SDG12. (2) The impact of ICT varies notably, with a more pronounced adverse effect on SDG12 in countries with higher resource consumption. (3) ICT goods export trade, technological innovation, and globalization significantly mitigate ICT’s adverse impact on resource consumption. This study provides targeted recommendations for G20 countries on how to leverage ICT to achieve SDG12 more effectively.

1. Introduction

Since the United Nations launched the Sustainable Development Agenda, environmental sustainability has emerged as a critical global priority. Growing human pressures—such as resource overexploitation, environmental degradation, and climate-related disasters—pose unprecedented threats to planetary boundaries [1]. In order to mitigate the rapid depletion of resources, the international community must work together to reduce resource consumption [2]. Fortunately, the growth of global resource use could be slowed down by 25% if resource efficiency is improved, along with responsible production and consumption [3]. This imperative is clearly reflected in Sustainable Development Goal 12 (SDG12). SDG12 promotes sustainability through two core objectives: encouraging responsible consumption/production patterns and maximizing welfare gains for each unit of material used [4,5]. However, traditional production-based accounting frameworks systematically underestimate transnational environmental externalities [6]. Consequently, the adoption of consumption-based material footprint metrics is essential. Defined as the global raw material extraction required to satisfy final demand [7], consumption-based material footprint provides a comprehensive measure of anthropogenic resource pressures, positioning it as a critical indicator for SDG12 monitoring [8]. It should be noted that the above consumption-based concept is used to emphasize the material footprint. For convenience, this paper will directly use the term “material footprint”.

Information and Communication Technology (ICT) has brought about a rapid penetration of digitalization. It has promoted the optimization and upgrading of conventional industries, the global economy’s evolution towards a circular model, and the diminution of natural resource waste [9]. Scholars have initiated investigations into the correlation between ICT and natural resource pressure, yet no consensus has been reached. Some studies contended that ICT enhances resource efficiency through optimized management systems and smart manufacturing, substantially lowering energy demands in traditional sectors [10]. Thus, it will help support sustainable production and consumption patterns [11,12]. Conversely, others ascertained that ICT would inevitably produce a significant carbon footprint throughout the entire life cycle [13]. Critically, the accelerating pace of technological advancement suggests that ICT’s resource-conservation potential may be offset by rebound effects [14]. Specifically, due to the ubiquitous utilization of the Internet [15,16], emissions of electronic waste and growth in per capita electricity consumption have become the culprits for the negative impact of ICT on energy consumption.

Notwithstanding the inconclusive and contradictory nature of research regarding the influence of ICT on environmental quality and resource pressure, the overall research framework is relatively sound. Significant knowledge gaps persist concerning ICT’s complex relationship with SDG12 and its specific linkages to material footprint. Crucially, the causal mechanisms through which ICT influences material footprint remain underexplored. The majority of current research has only considered the direct impact of digital trade, renewable resource management, financial development, globalization, and technological innovation on material footprint [8,17]. Representing a consortium of the world’s principal economies, G20 countries have substantially facilitated global economic expansion and the generation of renewable energy [18]. Research has also shown that G20 countries likewise persistently confront substantial impediments in their endeavor to achieve sustainable development [19]. This dual role underscores the urgency of identifying life cycle resource reduction strategies. Consequently, investigating how G20 countries can leverage ICT to decouple economic activity from material consumption and achieve SDG12 emerges as an imperative research priority.

This study makes multifaceted contributions to the extant literature as follows: (1) Current research on digitalization’s environmental impact overlooks critical dimensions. Most studies focus on carbon emissions or ecological footprints, neglecting material footprint as a core SDG12 indicator. This gap is significant because material footprint captures transnational resource flows essential for sustainable consumption assessment. Our study specifically addresses this deficiency by establishing material footprint as the primary dependent variable. (2) The existing literature insufficiently explores mechanisms linking ICT and material footprint. To address this gap, we introduce three theoretically grounded moderators: ICT goods export trade, globalization level, and technological innovation. These variables elucidate context-specific pathways through which digitalization influences resource consumption patterns. (3) Methodologically, we advance beyond the literature’s predominant dependence on mean-based estimation approaches. The Method of Moments Quantile Regression is employed to consider the non-linear relationship and investigate how ICT affects material footprint under different quantiles, better dealing with distributional heterogeneity. It is useful in providing policymakers of G20 countries with pathways to reduce their material footprint and achieve SDG12.

The subsequent sections are organized as follows: The literature review and Hypothesis are shown in Section 2. The main research methods and data description are outlined in Section 3. The empirical results are analyzed in Section 4. The conclusions and policy implications are presented in Section 5.

2. Literature Review and Hypothesis Development

2.1. Literature Review

2.1.1. SDG12 and Material Footprint

Despite concerted global efforts to implement the SDGs, an equivocal or ambiguous association persists between the research priorities of countries and the complexity of the SDG challenge, with a clear mismatch in SDG12 in particular [20]. Nevertheless, SDG12’s critical role in enabling energy reduction was recently underscored as fundamental to achieving “de-fossilisation”, which was proposed at the 28th Conference of the Parties. Consequently, there is a pressing need to study the ways in which SDG12 can be achieved. Current assessments predominantly employ either domestic material consumption or material footprint to evaluate progress. Crucially, material footprint, as an updated indicator of SDG12, diverges from domestic material consumption. The quantification of domestic material consumption is constrained to the measurement of raw materials directly employed by a nation or region. It fails to capture upstream resource flows embedded in international trade [21]. In contrast, material footprint optimized the allocation of responsibility for environmental impacts arising from natural resource exploitation and processing. Its unique advantages in measuring SDG12 and material flow analysis have attracted a growing number of scholars to use this index to assess the factors that affect SDG12 [22,23].

2.1.2. Environmental Impacts of ICT

Empirical studies increasingly recognize digitalization’s dual role in sustainable development. In the existing research on digitalization and SDG12, digitalization has been found to have significant negative effects. Pérez-Martínez et al. [16] employed the ICT development index as a digitalization proxy, revealing significant adverse impacts on SDG12 across nations. This outcome primarily stems from accelerated technological obsolescence, where rapid device replacement cycles generate substantial e-waste streams. Moreover, the advancement of ICT has resulted in adverse environmental consequences to some extent. These encompassed a significant escalation in carbon emissions and energy consumption [2], and a detrimental impact on the improvement of environmental performance [24]. Collectively, these findings suggest that ICT’s net reduction of resource efficiency may amplify material footprint across industrial value chains.

Digital trade research reveals similarly complex dynamics. Nejati and Shah [25] demonstrated that while intra-regional ICT goods trade generates positive scale economies in both developed and developing nations, it concurrently increases energy intensity and carbon emissions. Notably, ICT exports from developed economies could improve environmental quality in importing developing countries [26]. Conversely, He and Xiang [17] concentrated on the interrelationship between digital trade and material footprint within BRICS regions, showing that digital trade considerably reduces resource consumption and material footprint.

2.1.3. Research Gaps

Most research on digitalization’s environmental effects examines its dual influence on environmental quality, carbon emissions, environmental degradation, and ecological footprints. However, the relationship between digitalization and material footprint, a key indicator of SDG12, has not been widely explored by researchers. Furthermore, prevailing studies focus predominantly on digitalization’s direct sustainability impacts, while ignoring the transmission mechanism. To address the existing research gap, this study focuses on the relationship between two indicators of digitalization (ICT development level and ICT goods export trade) and the material footprint in G20 countries. ICT development level is identified as the core explanatory variable. Moreover, the pathway through which ICT influences the material footprint is elucidated.

2.2. Hypothesis Development

2.2.1. Direct Impact of ICT on Material Footprint

This paper examines the potential adverse effects of ICT on material footprint. Firstly, the advancement of ICT has led to a significant escalation in the manufacture of electronic devices, including smartphones, computers, base stations, network servers, and so on. The production of these devices required a considerable input of raw materials and energy [27], leading to an increased material footprint. Secondly, operational phases further compound resource demands: data centers and communication facilities exhibit high energy intensity, predominantly sourced from fossil fuels [28,29], generating significant carbon emissions and resource throughput [30]. Finally, the progress of ICT necessitates the storage and transmission of digital content. Although digital content is inherently virtual, it requires the operation of servers and the implementation of necessary maintenance procedures when transmitting data [31]. Consequently, the consumption of resources is considerable. Collectively, the comprehensive life cycle of ICT development demonstrates ICT’s inherent resource dependency, expanding material footprint despite digitalization’s dematerialization promise.

Empirical evidence remains scarce but revealing. Sun et al. [32] conducted a study into the impact of infrastructure development on the material footprint in BRICS nations, revealing that advancements in ICT infrastructure amplify the material footprint. Ceci and Razzaq [33] examined the extent to which ICT is integrated into sustainable resource management and ecological governance within G10 countries. They constructed a cumulative ICT index and found that ICT has a significant negative effect on the material footprint in G10 countries. Furthermore, some studies have contemplated the dynamic correlation or causality between ICT and other indicators that reflect resource consumption, such as ecological footprint [34,35,36] and domestic material consumption [37]. Drawing on these observations and a comprehensive review of pertinent literature, Hypothesis 1 is proposed as follows:

Hypothesis 1.

ICT development increases the material footprint and does not facilitate the achievement of goals established in SDG12.

2.2.2. Influencing Mechanism of ICT on Material Footprint

While research explicitly examining ICT’s transmission channels to material footprint remains limited, analogous studies illuminate ICT-related trade’s mechanistic role in resource consumption dynamics. Liu and Chen [38] discussed the comprehensive effect of digital trade on material footprint from the perspective of emerging MINT countries. Results showed that although financial development has increased carbon emissions and material footprint, the moderating effect of digital trade has weakened this negative environmental impact. The optimal allocation of green financial resources in the credit market was conducive to sustainable production and consumption. Furthermore, digital trade was a stimulus for sustainable resource management. Contrastingly, Wang et al. [39] documented digital trade’s detrimental footprint expansion across lower-middle quantiles in E7 nations. This divergence highlights context-dependent outcomes.

Moreover, ICT trade, effective regulation, and governance could also significantly reduce environmental pollution [40]. From a realistic perspective, ICT goods export trade promotes global trade in digital products and reduces energy consumption required for traditional commodity transactions [41]. Thus, it might ameliorate the deleterious effects of ICT development on material footprint. Given ongoing scholarly contention regarding ICT trade’s environmental externalities versus protective benefits, this paper will explore its role in ICT affecting material footprint. Hypothesis 2a and 2b are proposed as follows:

Hypothesis 2a.

ICT goods trade has a negative effect on the relationship between ICT and material footprint.

Hypothesis 2b.

ICT goods trade has a positive effect on the relationship between ICT and material footprint.

The pivotal function of globalization in the nexus between ICT and environmental quality, along with resource utilization, warrants meticulous examination. In general, the empirical results demonstrated certain inconsistencies in the impact of globalization on the environment. Energy-conserving technological advancements and compositional alterations facilitated by globalization contributed to the enhancement of environmental quality [42]. Conversely, the economic consequences of globalization encouraged the expansion of production, leading to elevated levels of raw material consumption and carbon emissions, which was not conducive to the improvement of environmental quality [43]. Using GMM estimation, Awad and Mallek [44] demonstrated that the interplay between ICT and economic globalization exerts a detrimental net effect on environmental quality. From the standpoint of sustainable development, indicators of globalization, such as trade openness and foreign direct investment, have facilitated substantial positive effects of ICT on sustainable development [45]. A body of research has empirically discussed ICT and globalization as direct factors affecting sustainable development. For instance, ICT and economic globalization have exerted a substantial, favorable influence on the sustainable development of APEC nations [46]. Drawing on the discourse of analogous topics in this literature, Hypothesis 3 is proposed as follows:

Hypothesis 3.

Globalization plays a weakening inhibitory role in the impact of ICT on the material footprint.

Technological innovation represents the tangible output of knowledge creation and R&D processes, operationalized through indicators like patents, R&D expenditure, and adoption rates of advanced technologies [47]. While ICT expansion unavoidably increases material footprint through heightened demand for critical minerals and electronic waste generation [30], technological innovation fundamentally alters this dynamic through three compensatory mechanisms: (1) Dematerialization via cloud computing and virtualization, reducing per-unit resource requirements [48,49]. (2) Circular economy enablers such as AI-driven resource recovery systems and modular device designs extending product lifespans [50]. (3) Promoting the development of more energy-efficient and environmentally friendly technologies pertaining to ICT, such as innovative technologies for production servers and large data centers [51]. Ejemeyovwi et al. [52] focused on the interaction between ICT usage and technological innovation in Africa, which can positively impact financial development and contribute to sustainable development. Collectively, these innovation-driven pathways translate into measurable reductions in material footprint by decoupling ICT-driven productivity gains from resource throughput intensity. Based on an analysis of the existing literature and the current situation, Hypothesis 4 is proposed as follows:

Hypothesis 4.

Technological innovation negatively moderates the positive association between ICT development and material footprint.

3. Materials and Methods

3.1. Theoretical Model

Material footprint captures resource depletion and environmental pressures from human production and consumption. By definition, a lower level of a region’s material footprint is more conducive to the realization of SDG12, and, consequently, to the attainment of the 28th Conference of the Parties’ goal of “de-fossilization”. The current analytical toolkit for discerning the drivers of environmental impacts comprises three primary types of models: IPAT, ImPACT, and STIRPAT. A thorough comparison undertaken by York et al. [53] indicated that the STIRPAT model facilitates analyses of a more sophisticated and refined nature than those achievable with either the IPAT or ImPACT models.

The STIRPAT model constituted an extension of the conventional IPAT model [54]. It can be expressed as Equation (1):

$$I_i=a{P}_i^b{A}_i^c{T}_i^d{e}_i$$

(1)

In Equation (1), I, P, A, and T denote environmental impact, population, affluence, and technology, respectively. The term “a” is the constant of proportionality, “i” denotes the region, and “b”, “c”, and “d” are index terms, while “e” is the error term.

STIRPAT is a multivariate non-linear model that allows for the inclusion of factors beyond population, affluence, and technology level that may influence environmental pressures [53]. The extended STIRPAT model has been shown to possess a superior degree of precision in identifying the drivers of environmental impacts, thereby overcoming the limitations inherent in the IPAT and ImPACT models. These limitations included the inability to reflect the degree of change in environmental impacts when drivers alter them and the inability to conduct hypothesis testing. In order to comprehensively consider the factors related to the environmental situation, the model needs to be extended to reflect the reality. As detailed in Section 2.2, the analysis of prior literature and the specific characteristics of material footprint motivate the extension of the core STIRPAT model. Consequently, this paper utilizes the extended STIRPAT model to examine the determinants of variations in the material footprint of G20 countries. To mitigate potential errors caused by heteroskedasticity and inconsistency in the magnitude of the data, the model is transformed into a log-linear model to derive Equation (2):

$$\ln I_{it}=\ln a+b\ln P_{it}+c\ln A_{it}+d\ln T_{it}+f\ln {{\rm K}}_{it}+\ln e$$

(2)

where P, A, and T are measured using the total population, total GDP based on 2015 constant prices, and the number of patent applications, respectively. K represents driver influences other than P, A, and T. lna and lne represent the constant term and the error term, respectively. Considering the research problem of this paper, Equation (2) is extended and deformed to obtain Equation (3):

$$\begin{array}{l}\ln M{F}_{it}={\beta }_0+{\beta }_1\ln IC{T}_{it}+{\beta }_2\ln ICTGT{R}_{it}+{\beta }_3\ln KO{F}_{it}\\ +{\beta }_4\ln GD{P}_{it}+{\beta }_5\ln PO{P}_{it}+{\beta }_6\ln TEC{H}_{it}+{\epsilon }_{it}\end{array}$$

(3)

In Equation (3), material footprint represents the material footprint, ICT denotes the development level of ICT and is quantified using the ICT Development Index. ICTGTR refers to digital trade and is measured using the total export trade of ICT products. KOF is utilized as a globalization index. GDP signifies economic development status and is quantified using GDP at constant prices in 2015. POP stands for total population, and TECH represents the degree of technological innovation and is measured using the number of patents filed. We maintain technological innovation (TECH) as the STIRPAT identity’s core ‘T’ factor. Our model then extends the framework to address critical drivers of material footprint. ICT is incorporated as a specific transformative technology vector, refining “T” to account for digitalization’s paradoxical role: its infrastructure demands, rapid obsolescence, and systemic energy/resource needs can amplify material throughput despite potential efficiency gains. ICT goods trade (ICTGTR) extends the “A” (Affluence) and “T” dimensions, reflecting how integration into global ICT value chains redistributes material burdens: export-led production embeds significant raw material extraction and processing domestically while servicing global consumption. Globalization conceptually drives material footprint by integrating economies, thereby reshaping the scale, patterns, and spatial distribution of global resource extraction, production, and consumption linked to material footprint. The terms β₁~β₆ are the elasticity coefficients of each variable concerning the material footprint, and β₀ serves as a constant term. Similarly, i and t represent regions and years, respectively.

3.2. Estimated Technology

The utilization of panel data models may potentially engender issues of cross-sectional dependency, wherein the correlations between sample countries have the potential to distort the estimated parameters, consequently yielding invalid estimates [55]. It is submitted that consideration of inter-country dependence, and analysis of the factors contributing to it, will assist in the enhancement of the accuracy of the model’s estimates. Initially, the cross-sectional dependence test [56] is employed, which assumes cross-sectional independence as the null hypothesis. The test statistic of the CD-test is presented in Equation (4):

$$C{D}_{test}={\displaystyle \frac{\sqrt{2T}}{{\left[N(N-1)\right]}^{1/2}}}\left({\displaystyle \sum _{i=1}^{N-1}{\displaystyle \sum _{j=i+1}^N{\widehat{\rho }}_{ij}}}\right)$$

(4)

where $\widehat{{\rho }_{\mathrm{i}\mathrm{j}}}$ is the sample estimate of the pair-wise correlation of the residuals, whereas the “N” refers to cross-sections, and the “T” states the time.

Secondly, G20 countries, as emerging economies, display certain regional similarities in terms of economics, trade, energy policies, and developmental status. As a result, the joint actions of these economies may exacerbate the issue of slope heterogeneity in the panel data. To deal with this issue, the slope homogeneity test [57] is employed to ascertain whether the slope coefficients adhere to a chi-square distribution.

$${\stackrel{⌢}{\Delta }}_{SH}=\sqrt{{\displaystyle \frac{N}{2k}}}\left({\displaystyle \frac{1}{N}}S-K\right)$$

(5)

$${\stackrel{⌢}{\stackrel{⌢}{\Delta }}}_{ASH}=\sqrt{N}\left({\displaystyle \frac{1}{N}}S-2K\right){\left({\displaystyle \frac{2K\left(T-K-1\right)}{T+1}}\right)}^{-1/2}$$

(6)

In Equations (5) and (6), ${\stackrel{⌢}{\Delta }}_{SH}$ stands for the slope coefficient homogeneity, and ${\stackrel{⌢}{\stackrel{⌢}{\Delta }}}_{ASH}$ reflects the slope coefficient homogeneity after adjustment. S, K, T, and N denote Swamy’s statistics, regressors, time, and cross-section, respectively.

The subsequent stage in panel data analysis entails the implementation of unit root tests. The initial generation of these tests imposes an obligation on cross-sections to demonstrate independence. Nevertheless, this stipulation is unfulfilled in instances of cross-sectional data dependence, potentially leading to skewed outcomes. In response to this quandary, researchers introduced a subsequent generation of unit root tests, which are responsible for cross-sectional correlation. The objective of these tests is to evaluate the serial smoothness of the system under investigation. The Cross-Sectionally Augmented IPS (CIPS) test from Pesaran [58] is employed, as outlined in Equations (7) and (8). Here, G and CDF represent the test statistic for the cross-section mean and the cross-section unit, respectively.

$$\Delta G_{i,t}={\omega }_i+{\omega }_i{G}_{i,t-1}+{\omega }_i{\overline{H}}_{t-1}+{\displaystyle {\sum }_{l=0}^p{\omega }_{il}\Delta \overline{G_{t-l}}}+{\displaystyle {\sum }_{l=1}^p{\omega }_{il}\Delta G_{i,t-l}}+{\epsilon }_{it}$$

(7)

$$\widehat{CIPS}=N^{-1}{\displaystyle {\sum }_{l=1}^{n}CD{F}_i}$$

(8)

Following the implementation of the unit root test, a Pedroni cointegration test is employed to ascertain the presence of a long-term cointegrating association among all variables. This test allows for the inclusion of different panel units, each with their unique cointegration vectors and residual autoregressive coefficients.

While the traditional OLS method or non-linear approaches (e.g., threshold/polynomial regressions) provide an estimate of the average effect of explanatory variables, they fail to capture potential heterogeneity in effects across the conditional distribution of the dependent variable [2]. Specifically, they are unable to reveal how the effect of ICT on resource demand varies across environmental pressure regimes—whether in high-footprint industrial economies or low-footprint service-oriented systems. Endogeneity is an important issue in traditional quantile regression, and it often leads to invalid estimations [59]. The Method of Moments Quantile Regression (MMQR) directly addresses the core empirical challenges of our study. It is capable of accurately capturing the distributional characteristics of the data across different quantiles [60]. This capability is essential for rigorously investigating our core research question: whether the environmental impact of ICT exhibits heterogeneous pathways across different quantiles of the material footprint distribution. When applied to panel data, this method can effectively address endogeneity while accounting for individual and temporal heterogeneity. By integrating these features, MMQR generates the granular insights required to derive targeted environmental governance strategies aligned with specific material footprint regimes. The quantile can be simplified as Equation (9):

$$Q_Y\left(\tau |X_{it}\right)=\left({\alpha }_i+{\delta }_{i}q\left(\tau \right)\right)+X_{it}^{\prime }\beta +Z_{it}^{\prime }\gamma q\left(\tau \right)$$

(9)

where the scalar coefficient is denoted by $\left({\alpha }_i+{\delta }_{i}q\left(\tau \right)\right)$, and $q\left(\tau \right)$ refers to a single sample quantile.

The regression coefficient of marginal effect of explanatory variable X_l on the τ-th quantile is shown in Equation (10):

$${\beta }_l\left(\tau ,X\right)={\beta }_l+q\left(\tau \right)D_{X_l}^{\sigma }$$

(10)

Upon completion of the main effects estimation, the specific mechanisms of influence between the core explanatory variable, ICT, and the dependent variable, material footprint, are explored. This paper employs hierarchical regression to conduct moderation tests, with the objective of ascertaining through which variables ICT primarily amplifies or mitigates its effect on material footprint. The primary moderating variables in this paper are the level of technological innovation (TECH), ICT trade (ICTGTR), and globalization (KOF). Ultimately, to guarantee the dependability of the empirical findings, robustness tests are undertaken. This is achieved by replacing the explained variables, as well as substituting the core explanatory variable ICT with its first-phase differential corresponding term and its respective one-period lagged independent counterpart.

3.3. Data Sources and Description

3.3.1. Data Source and Explanation

The subject of this paper is G20 countries. Due to constraints in data availability and the time frame, the period under consideration spans from 2000 to 2020. The European region within G20 countries is excluded from the evaluation due to data limitations. All data utilized in this study are sourced from publicly available, official databases. The core dependent variable, material footprint, captures consumption-based resource demand. It quantifies the raw materials extracted globally and attributed to final consumption within an economy using high-resolution multi-regional input–output tables (MRIOTs). This indicator employs the GLORIA MRIOT system developed by Sydney University. It features 367 material categories and smoothed OECD trade data to allocate domestic extraction across global supply chains with enhanced sectoral precision. The core explanatory variable, ICT development level, quantifies national infrastructure capacity for frontier technology adoption through a composite index. This index integrates two standardized components: internet user penetration (% of population) and broadband performance metrics. The KOF index, as calculated by the KOF Swiss Economic Institute, measures the economic, social, and political dimensions of globalization. The ICT trade is measured by the export trade volume of ICT goods. Gross Domestic Product (GDP) is measured in constant 2015 U.S. dollars. Technological innovation and population are measured, respectively, by the number of patent applications and the total population. The specific variables, units of measurement, and data sources employed in the research are presented in Appendix A,

Table A1. Data description and data sources.

Variables	Symbol	Measurement	Source
Material Footprint	MF	Material Footprint per capita(t/cap).	GMFD
Information and Communications Technology Development	ICT	Fixed Line and Mobile Phones Users, Internet Accessibility, Server Security	UNCTAD
ICT Trade	ICTGTR	Exports of Total ICT Goods (US dollars at current prices in millions)	UNCTAD
globalization Level	KOF	The Economic, Social, and Political Dimensions of globalization	KOFSEI
Economic Affluence	GDP	Gross Domestic Product (constant 2015 USD, USD)	GMFD
Population	POP	Total population	GMFD
Technological Innovation	TECH	Patent Application, residents	WDI

.

3.3.2. Descriptive Statistics

To improve model estimation accuracy and minimize the effects of heteroscedasticity and outliers, all variables are transformed using base-10 logarithms. This logarithmic form also meets the specification requirements of the extended STIRPAT model.

Table 1. Descriptive statistics.

Variable	MF	ICT	ICTGTR	KOF	GDP	POP	TECH
Mean	1.201	1.668	3.679	1.831	12.19	7.995	3.922
Std.Dev	0.269	0.199	1.094	0.066	0.421	0.497	0.972
Minimum	0.504	0.968	0.301	1.657	11.35	7.279	1.663
Maximum	1.640	1.899	5.599	1.940	13.30	9.150	6.144
Jarque–Bera	32.37	154.4	19.85	27.70	13.83	55.07	10.79
Prob.	0.000	0.000	0.000	0.000	0.000	0.000	0.005

reveals the descriptive statiatics.

The results show that all variables have positive values, indicating an overall upward trend in the data. Among the variables, the KOF index shows the smallest standard deviation and range, indicating relatively low variability across G20 countries. It implies that the degree of globalization among G20 countries is relatively uniform. However, the standard deviation of ICT goods export trade and technological innovation is comparatively substantial, signifying considerable fluctuations among G20 countries for these variables. This paper employs the Jarque–Bera test for data normality. Results indicate that all variables rejected the original hypothesis that the data follow a normal distribution at the 1% significance level, thereby confirming that they are non-normal.

3.3.3. Data Distribution of the Core Dependent Variable

Figure 1 illustrates both the growth trajectory of resource consumption in G20 countries and the global distribution of material footprint in 2023.

Figure 1a–f, respectively, represent the growth trends of the material footprint from 1990 to 2023 for G20 countries located in North America, Europe, Asia, South America, Africa, and Oceania. When considered alongside the composite graph, it becomes clear that most G20 countries had a significantly higher material footprint than other nations in 2023. Additionally, since 1990, G20 countries in Asia, North America, and South America have shown a consistent upward trend in their material footprint. This highlights the urgent need for G20 countries to reduce their material footprint and improve resource management strategies.

Figure 2 depicts the specific data distribution of material footprint, with the aim of observing the distribution characteristics of the per capita material footprint of G20 countries from 2000 to 2020. Overall, the concentration trend of the per capita material footprint of all countries in Oceania and Europe is similar, with all countries at a high level. It illustrates that developed countries had higher per capita material footprint and greater resource consumption than most developing countries. This pattern is consistent with findings that high-income nations consume disproportionately more resources than lower-income ones [61], reflecting their resource-intensive infrastructures and consumption-driven lifestyles. This likely reflects not only their high consumption level but also structural factors like their extensive industrial infrastructure, large-scale construction and transportation networks, and high living standards, all of which elevate raw material demand. Among them, Canada has the highest per capita material footprint, and the annual data are scattered. India, Indonesia, South Africa, and Mexico have relatively low per capita material footprint and relatively uniform distribution. Interestingly, the distribution of China’s material footprint is similar to that of Italy. It has a higher material footprint per capita among developing countries and is more dispersed. Both nations function as global manufacturing hubs with dense industrial clusters. China’s export-oriented production system and Italy’s specialized manufacturing belts require intensive intermediate material inputs. Simultaneously, rapid urbanization in China and infrastructure modernization in Italy drive construction-material demand. Critically, each exhibits fragmented resource efficiency: advanced industrial zones coexist with less efficient sectors. This creates “medium-material-intensity” profiles distinct from resource-exporting developing nations or service-dominated advanced economies.

3.4. Preliminary Tests

3.4.1. Slope Heterogeneity and Cross-Sectional Dependence

Panel data may introduce specific challenges, including potential slope heterogeneity and cross-sectional dependence [62]. To address these concerns, we conducted a slope coefficient homogeneity test and a CD-test. The outcomes of the slope heterogeneity test are shown in Appendix A,

Table A2. Testing for slope heterogeneity.

Statistics	Delta	p-Value	Statistics	Delta	p-Value
$\stackrel{⏜}{\Delta }$	12.371 ***	0.000	$\stackrel{⏜}{\stackrel{⏜}{\Delta }}$adjusted	15.723 ***	0.000

Note: Significance level is denoted by *** for 1%

. The p-value corresponding to the statistical value of the slope coefficient is 0, signifying that the null hypothesis is rejected at the 1% level of significance. This confirms slope heterogeneity in the model, necessitating further analysis of cross-sectional dependence.

The results of the cross-sectional dependence test are documented in Appendix A,

Table A3. Testing for cross-sectional dependency.

Variable	CD-Test	p-Value	Variable	CD-Test	p-Value
MF	4.580 ***	0.000	GDP	49.590 ***	0.000
ICT	58.054 ***	0.000	POP	36.656 ***	0.000
ICTGTR	4.046 ***	0.000	TECH	5.766 ***	0.000
KOF	51.418 ***	0.000

Note: Significance level is denoted by *** for 1%

. It is discernible that the CD statistic for each variable is statistically significant at the level of 1%, suggesting that cross-sectional independence is not present. Consequently, all variables MF, ICT, ICTGTR, KOF, GDP, POP, and TECH of G20 countries are interrelated and cross-sectionally related. Consequently, changes in these variables within any G20 country propagate to others.

3.4.2. Unit Root Analysis and Cointegration Tests

Within the framework of cross-sectional dependence in panel data, the CIPS test is utilized for unit root analysis. The findings are outlined in Appendix A,

Table A4. Testing for unit root.

Variable	I(0)	I(1)	I(2)	Level of Integration
MF	−1.950	−2.505 ***	/	I(1)
ICT	−1.381	−2.184 **	/	I(1)
ICTGTR	−1.654	−2.084 *	/	I(1)
KOF	−1.834	−2.725 ***	/	I(1)
GDP	−1.526	−2.135 **	/	I(1)
POP	−1.717	−1.566	−2.532 ***	I(2)
TECH	−1.818	−2.610 ***	/	I(1)

Note: Significance level is denoted by *** for 1%, ** for 5%, and * for 10%.

. Among the outcomes, material footprint, ICT, GDP, ICTGTR, KOF, and TECH demonstrated significance at first difference. In particular, the POP plateaus after the second derivative. The comprehensive results suggest a heterogeneous order of integration among the variables, with all of them except POP attaining stationarity after incorporating their first differences. Hence, a cointegration test is necessary.

This study employs the more flexible Pedroni test to assess panel cointegration. The outcomes of the test, as shown in Appendix A,

Table A5. Testing for cointegration.

Pedroni (2004)	Statistic	p-Value
Modified Phillips–Perron t	4.7047 ***	0.0000
Phillips–Perron t	−2.3206 **	0.0102
Augmented Dickey–Fuller t	−2.1395 **	0.0162

Note: Significance level is denoted by *** for 1%, ** for 5%.

, demonstrate that the p-values of all three statistics are less than 0.01, thereby robustly refuting the preliminary hypothesis of “no cointegration” at the 1% significance level. It should be noted that since the POP becomes stable after the second difference, it was not included in the cointegration test. Consequently, it is posited that there exists a long-term association and stable equilibrium between ICT, ICTGTR, KOF, GDP, TECH, and material footprint in G20 countries.

4. Results

4.1. Direct Effects of ICT on Material Footprint

Based on the non-normality, slope heterogeneity, cross-sectional dependence, second-order smoothness, and cointegration of the data, this paper employs the MMQR approach to estimate how ICT affects material footprint.

Table 2. Results of MMQR.

	(1) Q = 0.1	(2) Q = 0.2	(3) Q = 0.3	(4) Q = 0.4	(5) Q = 0.5	(6) Q = 0.6	(7) Q = 0.7	(8) Q = 0.8	(9) Q = 0.9
ICT	0.4206 *** (3.92)	0.4381 *** (4.97)	0.4488 *** (5.67)	0.4601 *** (6.31)	0.4723 *** (6.65)	0.4846 *** (6.50)	0.5130 *** (5.19)	0.5455 *** (3.89)	0.6054 *** (2.66)
ICTGTR	−0.0428 *** (−4.47)	−0.0422 *** (−5.36)	−0.0418 *** (−5.91)	−0.0413 *** (−6.35)	−0.0409 *** (−6.44)	−0.0404 *** (−6.07)	−0.0393 *** (−4.46)	−0.0381 *** (−3.04)	−0.0358 * (−1.76)
KOF	−1.1017 *** (−5.83)	−1.0669 *** (−6.86)	−1.0455 *** (−7.49)	−1.0229 *** (−7.94)	−0.9986 *** (−7.95)	−0.9741 *** (−7.38)	−0.9174 *** (−5.24)	−0.8525 *** (−3.43)	−0.7346 * (−1.83)
GDP	0.6149 *** (12.01)	0.6262 *** (14.86)	0.6332 *** (16.73)	0.6406 *** (18.36)	0.6485 *** (19.04)	0.6565 *** (18.31)	0.6750 *** (14.19)	0.6961 *** (10.30)	0.7346 *** (6.76)
POP	−0.4565 *** (−10.96)	−0.4513 *** (−13.18)	−0.4481 *** (−14.58)	−0.4448 *** (−15.71)	−0.4412 *** (−15.99)	−0.4375 *** (−15.11)	−0.4291 *** (−11.18)	−0.4194 *** (−7.69)	−0.4019 *** (−4.55)
TECH	−0.0177 (−1.34)	−0.0229 ** (−2.11)	−0.0261 *** (−2.67)	−0.0295 *** (−3.27)	−0.0331 *** (−3.75)	−0.0368 *** (−3.94)	−0.0453 *** (−3.64)	−0.0549 *** (−3.11)	−0.0725 *** (−2.59)

Note: t statistics in parentheses; significance level is denoted by *** for 1%, ** for 5%, and * for 10%.

shows the principal outcomes of the MMQR. The graphical depiction in all quantiles is displayed in Figure 3.

Table 2 and Figure 3 show that the parameters of the core explanatory variable ICT exhibit a markedly positive effect across the 10th to 90th quantiles at the 1% level. It confirms Hypothesis 1, which posits that ICT development in G20 countries leads to an increase in material footprint, thereby undermining progress toward SDG12. ICT infrastructure (e.g., encompassing network deployment, device manufacturing, and data center operations) demands vast quantities of critical minerals, rare earth elements, energy, and water, directly amplifying resource extraction. This coincides with the research results of Sun et al. [32] and Balsalobre-Lorente et al. [42]. Simultaneously, the rapid obsolescence cycles of digital hardware accelerate electronic waste generation, further straining waste management systems. Crucially, as emphasized by Lenzen et al. [8], this rise in aggregate material throughput fundamentally conflicts with SDG12’s mandate for absolute decoupling of economic activity from environmental degradation. SDG12.2 explicitly calls for sustainable natural resource management, while SDG12.5 demands substantial waste reduction—both undermined by heightened virgin resource extraction and disposal linked to ICT’s life cycle. For high-consumption G20 countries, ICT-driven efficiency gains are thus offset by scale effects, resulting in a net increase in material pressure that erodes progress toward “doing more and better with less”, a core ethos of SDG12. Notably, our results contrast with findings of Huang et al. [34] for G7 countries, which reported ICT-driven dematerialization effects. This implies that, compared to G7 countries, the structural differences (such as a higher proportion of manufacturing or a shorter hardware update cycle) among G20 countries may amplify the physical impact of ICT. Additionally, the detrimental impact of ICT is observed to increase in a linear fashion with the quantile level of material footprint, reaching its maximum at the highest quantile level elasticity value. It suggests that countries with higher quantile levels of material footprint may experience more adverse effects when developing ICT.

Descriptive statistics confirm that high material footprint nations are predominantly developed or rapidly developing economies. Crucially, this pattern stems not merely from elevated consumption but from entrenched structural drivers. These societies operate within high-intensity material systems. Their industrial foundations rely on resource-heavy infrastructures. Consumer markets prioritize cutting-edge electronics with accelerated replacement cycles. This generates compound pressures: surging e-waste volumes and energy-intensive ICT manufacturing [61]. Concurrently, expansive digitalization initiatives demand nationwide communication infrastructure. Dense networks of base stations proliferate to ensure coverage. Such physical expansion consumes bulk raw materials. Critically, ICT deployment activates resource-intensive supply chains. Each technological layer adds material throughput. Entire production ecosystems thus amplify resource demand [13,63]. Ultimately, the interdependence of locked-in consumption paradigms, infrastructural inertia, and cascading industrial effects systematically drives the escalation of material footprint. Environmental degradation becomes an embedded outcome of this material-intensive development trajectory.

Beyond the primary explanatory variable, ICT goods export trade made a considerable negative contribution to material footprint, thereby facilitating the realization of SDG12. Nevertheless, this negative effect attenuated at higher material footprint quantiles. In countries with a low quantile of material footprint, the increase in export trade of ICT goods could improve material efficiency, lowering environmental pressure. India exemplifies this dynamic, exhibiting a relatively low material footprint baseline. Critically, the expansion of ICT goods exports could potentially reduce traditional trade costs and associated transportation energy consumption. By providing G20 nations with cost-effective digital trade platforms, ICT exports enhanced value chain participation [64], ultimately promoting resource-efficient growth [17].

The substantial beneficial effect of globalization that reduces the material footprint of G20 countries is apparent. Furthermore, the absolute value of the elasticity coefficient associated with this positive influence is maximized at the lowest quantile level of material footprint. It implies that the lower the original material footprint level of a country, the more beneficial it is to achieve SDG12 and improve its environmental quality by promoting the globalization of its economy, trade, and value chain. This aligns with Nathaniel [65] and Sun et al. [66] on globalization’s role in reducing ecological footprints and advancing sustainability. Consequently, low-MF G20 economies—particularly India, Indonesia, and Mexico—should prioritize globalization enhancement to maximize environmental improvements.

The impact of other control variables on the material footprint of G20 countries is predominantly significant at the 1% level. The coefficients of the control variable GDP display a statistically significant positive correlation from the 10th to the 90th quantile at the 1% significance level. Its positive impact on the material footprint exhibits a significant augmentation as the conditional quantile of the material footprint increases. This pattern indicates that economic prosperity has not yet reached the environmental Kuznets curve (EKC) inflection point; economic growth continues to stimulate resource consumption throughout product life cycles, potentially triggering energy crises [67]. The negative impact on raw material consumption and SDG12 has been confirmed in a substantial body of literature [17]. Combined with our earlier CD-test findings, over-expansion not only diminishes resource efficiency within the home country but also influences the material footprint of other countries within the G20 countries. Economic development strategies must therefore account for non-renewable energy costs and sustainability trade-offs.

Whereas, POP has significantly negative elasticity values at the 1% significance level for all quartiles. Its favorable effect of slowing down the material footprint is most pronounced at the lower quartile level of the material footprint. However, as the material footprint increased, the slowing-down effect of population growth became weaker. This phenomenon may be attributed to the expansion of population size, which has led to an improvement in population quality and intellectual resources. The enhancement of population quality can then reduce the intensity of energy consumption, thereby effectively diminishing the material footprint [68]. The alleviating influence of TECH on material footprint at the 20th to 90th quantile levels is statistically significant at the level of 1%, with absolute elasticity peaking at the highest material footprint quantile. It suggests that promoting technological innovation is a key initiative to mitigate material footprint intensity, especially in high-income countries facing severe energy constraints. The empirical conclusions align with the existing research [69]. According to the EKC hypothesis, technological innovation leads producers to utilize low-energy consuming and cleaner technologies, thereby improving environmental quality [70,71].

4.2. The Impact Pathways of ICT on Material Footprint

Firstly, the moderating effect of ICT goods export trade on the impact of ICT on material footprint is assessed, with results being presented in

Table 3. MMQR estimations with the interaction term between ICT and ICTGTR.

	(1) Q = 0.1	(2) Q = 0.2	(3) Q = 0.3	(4) Q = 0.4	(5) Q = 0.5	(6) Q = 0.6	(7) Q = 0.7	(8) Q = 0.8	(9) Q = 0.9
ICT	0.7842 *** (6.19)	0.8044 *** (7.28)	0.8305 *** (8.85)	0.8494 *** (9.80)	0.8701 *** (10.25)	0.9013 *** (9.54)	0.9423 *** (7.66)	0.9858 *** (6.04)	1.0534 *** (4.51)
ICTGTR	−0.0486 *** (−4.84)	−0.0487 *** (−5.57)	−0.0489 *** (−6.57)	−0.0490 *** (−7.14)	−0.0491 *** (−7.31)	−0.0492 *** (−6.60)	−0.0494 *** (−5.08)	−0.0497 *** (−3.84)	−0.0500 *** (−2.70)
KOF	−0.9374 *** (−5.08)	−0.9168 *** (−5.70)	−0.8902 *** (−6.51)	−0.8710 *** (−6.90)	−0.8499 *** (−6.87)	−0.8181 *** (−5.95)	−0.7763 *** (−4.33)	−0.7320 *** (−3.08)	−0.6631 * (−1.95)
GDP	0.6526 *** (12.77)	0.6658 *** (14.89)	0.6829 *** (17.92)	0.6953 *** (19.74)	0.7088 *** (20.41)	0.7293 *** (18.82)	0.7561 *** (15.03)	0.7846 *** (11.79)	0.8288 *** (8.75)
POP	−0.5052 *** (−11.02)	−0.5083 *** (−12.72)	−0.5122 *** (−15.08)	−0.5151 *** (−16.43)	−0.5183 *** (−16.89)	−0.5230 *** (−15.33)	−0.5293 *** (−11.90)	−0.5359 *** (−9.08)	−0.5462 *** (−6.46)
TECH	0.0011 (0.08)	−0.0020 (−0.17)	−0.0061 (−0.59)	−0.0090 (−0.94)	−0.0122 (−1.30)	−0.0171 (−1.63)	−0.0234 * (−1.72)	−0.0302 * (−1.67)	−0.0407 (−1.58)
ICT * ICTGTR	−0.1568 *** (−3.92)	−0.1637 *** (−4.69)	−0.1727 *** (−5.82)	−0.1792 *** (−6.54)	−0.1863 *** (−6.94)	−0.1971 *** (−6.59)	−0.2112 *** (−5.43)	−0.2261 *** (−4.38)	−0.2494 *** (−3.38)

Note: t statistics in parentheses; significance level is denoted by *** for 1% and * for 10%.

. These results indicate that the interaction term between ICT and ICTGTR is significantly negative across all quantile levels. It suggests that the expansion of ICT goods export trade attenuates ICT’s positive impact on material footprint, exhibiting a significant negative moderating effect. This result rejects Hypothesis 2b and confirms Hypothesis 2a.

These findings partially align with Liu and Chen [38], while this study further delineates the digital trade dimensions examined in their research. It argues that the characteristics of ICT goods export trade, such as convenience and low energy consumption, can contribute to the green sustainability effect of ICT while offsetting the negative environmental externalities it generates. Furthermore, the more significant the material footprint quantile, the more pronounced the negative moderating effect of ICT goods export trade. The signs of the elasticity coefficients of the other control variables are also essentially the same as those in Table 2.

Secondly, we explore the moderating role of globalization. As shown in

Table 4. MMQR estimations with the interaction term between ICT and KOF.

	(1) Q = 0.1	(2) Q = 0.2	(3) Q = 0.3	(4) Q = 0.4	(5) Q = 0.5	(6) Q = 0.6	(7) Q = 0.7	(8) Q = 0.8	(9) Q = 0.9
ICT	0.3515 ** (2.50)	0.3505 *** (2.96)	0.3498 *** (3.30)	0.3491 *** (3.55)	0.3483 *** (3.65)	0.3474 *** (3.45)	0.3455 ** (2.56)	0.3433 * (1.76)	0.3401 (1.17)
ICTGTR	−0.0410 *** (−4.24)	−0.0401 *** (−4.92)	−0.0394 *** (−5.42)	−0.0388 *** (−5.74)	−0.0381 *** (−5.82)	−0.0373 *** (−5.39)	−0.0355 *** (−3.83)	−0.0355 ** (−2.51)	−0.0307 (−1.54)
KOF	−1.0980 *** (−6.13)	−1.0534 *** (−6.98)	−1.0208 *** (−7.56)	−0.9917 *** (−7.90)	−0.9578 *** (−7.84)	−0.9158 *** (−7.08)	−0.8308 *** (−4.78)	−0.7309 *** (−2.93)	−0.5903 (−1.60)
GDP	0.6366 *** (11.77)	0.6488 *** (14.23)	0.6577 *** (16.13)	0.6657 *** (17.54)	0.6749 *** (18.29)	0.6864 *** (17.56)	0.7097 *** (13.53)	0.7370 *** (9.77)	0.7754 *** (6.95)
POP	−0.4723 *** (−10.81)	−0.4699 *** (−12.77)	−0.4681 *** (−14.23)	−0.4666 *** (−15.26)	−0.4647 *** (−15.69)	−0.4624 *** (−14.80)	−0.4578 *** (−10.93)	−0.4524 *** (−7.49)	−0.4448 *** (−4.94)
TECH	−0.0199 (−1.54)	−0.0246 ** (−2.25)	−0.0280 *** (−2.86)	−0.0311 *** (−3.41)	−0.0346 *** (−3.89)	−0.0390 *** (−4.12)	−0.0479 *** (−3.76)	−0.0584 *** (−3.21)	−0.0731 *** (−2.74)
ICT ∗ KOF	−0.6564 (−1.22)	−0.7178 (−1.58)	−0.7626 * (−1.88)	−0.8026 ** (−2.13)	−0.8492 ** (−2.32)	−0.9069 ** (−2.35)	−1.0239 ** (−1.98)	−1.1613 (−1.56)	−1.3547 (−1.22)

Note: t statistics in parentheses; significance level is denoted by *** for 1%, ** for 5%, and * for 10%.

, the negative moderating effect of KOF is more pronounced in the 30th to 70th conditional quantile. This suggests that an accelerated process of globalization is beneficial in enabling ICT to reduce raw material consumption. In countries with a medium or high material footprint, Hypothesis 3 is therefore validated. This finding aligns with the conclusions of Awad and Mallek [44], who demonstrated that the interaction term of ICT and globalization slowed down environmental degradation and improved environmental quality. Globalization facilitates resource sharing and sustainable technology transfer in ICT [17], subsequently stimulating cross-border ICT capital flows and competition. This dynamic incentivizes developers to create high-efficiency, low-carbon ICT solutions.

Finally,

Table 5. MMQR estimations with the interaction term between ICT and TECH.

	(1) Q = 0.1	(2) Q = 0.2	(3) Q = 0.3	(4) Q = 0.4	(5) Q = 0.5	(6) Q = 0.6	(7) Q = 0.7	(8) Q = 0.8	(9) Q = 0.9
ICT	0.8649 *** (6.76)	0.9235 *** (8.72)	0.9622 *** (10.15)	1.0037 *** (11.46)	1.0418 *** (11.97)	1.1180 *** (11.16)	1.2192 *** (8.92)	1.2838 *** (7.70)	1.4163 *** (6.08)
ICTGTR	−0.0408 *** (−4.07)	−0.0378 *** (−4.56)	−0.0358 *** (−4.83)	−0.0337 *** (−4.92)	−0.0317 *** (−4.69)	−0.0278 *** (−3.58)	−0.0226 ** (−2.12)	−0.0193 (−1.48)	−0.0125 (−0.68)
KOF	−1.2862 *** (−6.94)	−1.2484 *** (−8.15)	−1.2234 *** (−8.94)	−1.1966 *** (−9.49)	−1.1720 *** (−9.43)	−1.1227 *** (−7.89)	−1.0574 *** (−5.39)	−1.0157 *** (−4.24)	−0.9301 *** (−2.76)
GDP	0.7169 *** (14.22)	0.7312 *** (17.55)	0.7406 *** (16.73)	0.7507 *** (21.83)	0.7599 *** (22.38)	0.7785 *** (19.97)	0.8030 *** (14.99)	0.8187 *** (12.52)	0.8510 *** (9.28)
POP	−0.5580 *** (−11.47)	−0.5635 *** (−14.03)	−0.5671 *** (−15.81)	−0.5710 *** (−17.28)	−0.5746 *** (−17.69)	−0.5817 *** (−15.64)	−0.5911 *** (−11.50)	−0.5971 *** (−9.51)	−0.6095 *** (−6.90)
TECH	−0.0115 (−0.89)	−0.0169 (−1.58)	−0.0204 ** (−2.14)	−0.0243 *** (−2.75)	−0.0278 *** (−3.17)	−0.0348 *** (−3.45)	−0.0441 *** (−3.21)	−0.0501 *** (−2.98)	−0.0623 *** (−2.66)
ICT ∗ KOF	−0.1783 *** (−4.13)	−0.1992 *** (−5.57)	−0.2129 *** (−6.64)	−0.2277 *** (−7.68)	−0.2412 *** (−8.17)	−0.2684 *** (−7.88)	−0.3043 *** (−6.57)	−0.3273 *** (−5.80)	−0.3744 *** (−4.76)

Note: t statistics in parentheses; significance level is denoted by *** for 1% and ** for 5%.

presents the moderating effect of TECH. The interaction term’s elasticity coefficients are significantly negative across quantiles (10th–90th), with most statistically significant at the 1% level. This supports Hypothesis 4. The negative moderating effect of technological innovation is most pronounced in countries with the highest quantile levels of material footprint conditions, more than compensating for ICT’s adverse impact on SDG12. The results are consistent with the findings of Gyamfi et al. [72], who suggested a two-way interaction between ICT and technological innovation. It implies that technological innovation can exert its own positive effects on environmental quality; at the same time, ICT has leveraged technological innovation to improve its own negative impact on energy consumption. This is also corroborated by the findings of Ejemeyovwi et al. [52].

4.3. Robustness Checks

In order to ensure the reliability and robustness of the aforementioned estimation results, this study employs three separate methodologies for conducting a robustness test. The first methodology involves substituting the explained variable material footprint. Prior to the update of the SDG framework, the primary indicator of SDG12 is Domestic Material Consumption (DMC), ensuring our findings remain grounded in established frameworks. While material footprint measures the global material extraction embedded in a nation’s final consumption, DMC quantifies the physical material inputs directly used within the domestic economy. That is, material footprint is consumption-based and DMC is production-based. This fundamental difference in accounting perspective makes DMC a valuable complementary metric for robustness testing. Specifically, it assesses whether the identified relationships (particularly the adverse effect of ICT on material sustainability) are sensitive to this shift in how material resource use is defined and measured. As such, this paper initially replaces the dependent variable material footprint with DMC to conduct a robustness test, and the results are presented in Appendix A,

Table A6. Robustness checks 1: Replace the dependent variable with DMC.

	(1) Q = 0.1	(2) Q = 0.2	(3) Q = 0.3	(4) Q = 0.4	(5) Q = 0.5	(6) Q = 0.6	(7) Q = 0.7	(8) Q = 0.8	(9) Q = 0.9
ICT	0.8096 *** (3.50)	0.7777 *** (4.07)	0.7496 *** (4.61)	0.7195 *** (5.00)	0.6990 *** (4.96)	0.6707 *** (4.45)	0.6465 *** (3.80)	0.5736 ** (2.19)	0.5103 −1.43
ICTGTR	−0.1050 *** (−5.70)	−0.0990 *** (−6.50)	−0.0936 *** (−7.21)	−0.0879 *** (−7.64)	−0.0840 *** (−7.44)	−0.0786 *** (−6.52)	−0.0740 *** (−5.36)	−0.0601 *** (−2.86)	−0.0480 * (−1.69)
KOF	−1.6792 *** (−4.31)	−1.4251 *** (−4.42)	−1.2015 *** (−4.32)	−0.9616 *** (−3.89)	−0.7981 *** (−3.27)	−0.5729 ** (−2.20)	−0.3802 (−1.23)	0.2005 (0.43)	0.7053 −1.18
GDP	0.3093 *** (3.01)	0.3227 *** (3.80)	0.3345 *** (4.63)	0.3472 *** (5.43)	0.3559 *** (5.69)	0.3678 *** (5.50)	0.3780 *** (4.99)	0.4087 *** (3.52)	0.4353 *** −2.75
POP	−0.2246 ** (−2.56)	−0.2241 *** (−3.10)	−0.2237 *** (−3.64)	−0.2233 *** (−4.10)	−0.2230 *** (−4.19)	−0.2226 *** (−3.91)	−0.2222 *** (−3.46)	−0.2211 ** (−2.24)	−0.2202 (−1.63)
TECH	−0.0046 (−0.16)	−0.0118 (−0.50)	−0.0181 (−0.90)	−0.0249 (−1.40)	−0.0295 * (−1.70)	−0.0359 * (−1.93)	−0.0413 * (−1.95)	−0.0577 * (−1.78)	−0.0719 (−1.64)

Note: t statistics in parentheses; significance level is denoted by *** for 1%, ** for 5%, and * for 10%.

. These results largely correspond with the previously observed adverse effect of ICT on SDG12 and the sign of the elasticity coefficients of the other variables, thereby affirming the reliability of the aforementioned results.

On another scale, given the identification of non-stationarity in our panel data through Fisher-ADF tests, we implement first-difference transformation on the core ICT variable (ΔICT) to mitigate spurious regression risks. As illustrated in Appendix A,

Table A7. Robustness checks 2: The first-difference of the core explanatory variable.

	(1) Q = 0.1	(2) Q = 0.2	(3) Q = 0.3	(4) Q = 0.4	(5) Q = 0.5	(6) Q = 0.6	(7) Q = 0.7	(8) Q = 0.8	(9) Q = 0.9
ΔICT	0.3892 *** (3.52)	0.4026 *** (4.45)	0.4099 *** (4.99)	0.4174 *** (5.47)	0.4271 *** (5.78)	0.4360 *** (5.65)	0.4576 *** (4.48)	0.4832 *** (3.27)	0.5267 ** (2.22)
ICTGTR	−0.0443 *** (−4.44)	−0.0434 *** (−5.31)	−0.0429 *** (−5.79)	−0.0424 *** (−6.16)	−0.0418 *** (−6.26)	−0.0412 *** (−5.91)	−0.0398 *** (−4.31)	−0.0381 *** (−2.85)	−0.0352 * (−1.65)
KOF	−1.0942 *** (−5.45)	−1.0671 *** (−6.49)	−1.0524 *** (−7.05)	−1.0372 *** (−7.48)	−1.0177 *** (−7.58)	−0.9997 *** (−7.12)	−0.9560 *** (−5.13)	−0.9042 *** (−3.36)	−0.8163 * (−1.90)
GDP	0.6128 *** (11.43)	0.6273 *** (14.27)	0.6352 *** (15.91)	0.6434 *** (17.32)	0.6538 *** (18.14)	0.6635 *** (17.53)	0.6870 *** (13.65)	0.7148 *** (9.85)	0.7620 *** (6.63)
POP	−0.4591 *** (−10.53)	−0.4560 *** (−12.78)	−0.4544 *** (−14.02)	−0.4526 *** (−15.05)	−0.4504 *** (−15.46)	−0.4483 *** (−14.74)	−0.4433 *** (−11.00)	−0.4374 *** (−7.51)	−0.4274 *** (−4.57)
TECH	−0.0159 (−1.13)	−0.0212 * (−1.84)	−0.0240 ** (−2.30)	−0.0270 *** (−2.77)	−0.0308 *** (−3.25)	−0.0343 *** (−3.44)	−0.0427 *** (−3.22)	−0.0528 *** (−2.76)	−0.0698 ** (−2.32)

Note: t statistics in parentheses; significance level is denoted by *** for 1%, ** for 5%, and * for 10%.

of the test results, the value of the elasticity coefficient of ΔICT remains significantly positive across the full quantile level. Furthermore, the positive and negative effects of the other control variables on material footprint align with those previously described, thereby affirming the robustness of the results.

We further account for the potential for reverse causality between ICT development and material footprint. It is plausible that not only does ICT influence material consumption, but current material footprint levels might also affect current ICT adoption. For instance, countries characterized by high resource-intensive economic activity might possess greater financial capacity to invest simultaneously in ICT infrastructure. To mitigate this risk of bidirectional causality biasing our estimates, we re-estimate the model using the core ICT variable lagged by one period (LICT). This approach helps isolate the direction of influence from past ICT to current material footprint [73]. As presented in Appendix A,

Table A8. Robustness checks 3: The lagged one-year period of the core explanatory variable.

	(1) Q = 0.1	(2) Q = 0.2	(3) Q = 0.3	(4) Q = 0.4	(5) Q = 0.5	(6) Q = 0.6	(7) Q = 0.7	(8) Q = 0.8	(9) Q = 0.9
L.ICT	0.3892 *** (3.52)	0.4026 *** (4.45)	0.4099 *** (4.99)	0.4174 *** (5.47)	0.4271 *** (5.78)	0.4360 *** (5.65)	0.4576 *** (4.48)	0.4832 *** (3.27)	0.5267 ** (2.22)
ICTGTR	−0.0443 *** (−4.44)	−0.0434 *** (−5.31)	−0.0429 *** (−5.79)	−0.0424 *** (−6.16)	−0.0418 *** (−6.26)	−0.0412 *** (−5.91)	−0.0398 *** (−4.31)	−0.0381 *** (−2.85)	−0.0352 * (−1.65)
KOF	−1.0942 *** (−5.45)	−1.0671 *** (−6.49)	−1.0524 *** (−7.05)	−1.0372 *** (−7.48)	−1.0177 *** (−7.58)	−0.9997 *** (−7.12)	−0.9560 *** (−5.13)	−0.9042 *** (−3.36)	−0.8163 * (−1.90)
GDP	0.6128 *** (11.43)	0.6273 *** (14.27)	0.6352 *** (15.91)	0.6434 *** (17.32)	0.6438 *** (18.14)	0.6635 *** (17.53)	0.6870 *** (13.65)	0.7148 *** (9.85)	0.7620 ** (6.63)
POP	−0.4591 *** (−10.53)	−0.4560 *** (−12.78)	−0.4544 *** (−14.02)	−0.4526 *** (−15.05)	−0.4504 *** (−15.46)	−0.4483 *** (−14.74)	−0.4433 *** (−11.00)	−0.4194 *** (−7.69)	−0.4019 ** (−4.55)
TECH	−0.0159 (−1.13)	−0.0212 * (−1.84)	−0.0240 ** (−2.30)	−0.0270 *** (−2.77)	−0.0308 *** (−3.25)	−0.0343 *** (−3.44)	−0.0427 *** (−3.22)	−0.0528 *** (−2.76)	−0.0698 ** (−2.32)

Note: t statistics in parentheses; significance level is denoted by *** for 1%, ** for 5%, and * for 10%.

, the regression coefficients of the lagged ICT variable remain statistically significant at the 5% level and are close in magnitude to those from the baseline regression. This consistency, alongside the stable effects of other control variables, indicates that controlling for potential reverse causality through temporal ordering does not alter our core finding of ICT’s adverse impact on material sustainability. These results significantly strengthen the robustness of our causal interpretation.

5. Conclusions and Policy Recommendations

5.1. Conclusions

This paper utilizes the expanded STIRPAT model to explore the non-linear effect of ICT on material footprint across G20 countries from 2000 to 2020, leveraging the Method of Moments Quantile Regression. Additionally, this paper examines how ICT goods export trade, globalization, and technological innovation moderate this relationship. To ensure the empirical results’ reliability, three robustness tests are conducted during the estimation process. The empirical results presented here offer valuable insights for policymakers in G20 countries and those involved in SDG12 research. It will facilitate the achievement of SDG12 and the “de-fossilization” target and enrich the existing literature on the study of realizing SDG12. The primary conclusions are as follows:

Firstly, there are clear regional differences in the material footprint of the G20 countries. Most developed countries, such as Canada, the United States, and Australia, perform poorly, with their material footprint at high levels. This pattern likely reflects deeply embedded structural factors such as high per-capita consumption levels, resource-intensive industrial bases, extensive built infrastructure requiring substantial material maintenance, and historically established, energy- and material-lifestyle patterns. Developing countries, however, especially India and Indonesia, which have a low full life cycle of raw materials, perform well in terms of material footprint. While classified as developing, China’s material footprint distribution aligns more closely with industrialized nations like Italy, exhibiting significant scale. This reflects China’s unique structural position as the world’s primary manufacturing hub over recent decades. Its exceptionally rapid industrialization and urbanization, fueled by massive investments in infrastructure and export-oriented production, have driven unprecedented aggregate demand for raw materials.

Secondly, the results reveal a long-term cointegrating relationship between ICT, ICTGTR, KOF, GDP, TECH, and material footprint in G20 countries. The primary explanatory variable, ICT, is found to significantly augment the material footprint of G20 countries across all quantile levels. Moreover, the higher the initial material footprint distribution, the more pronounced the ICT’s negative environmental externalities, primarily reflected in increased raw material consumption, resource inefficiency, and diminished environmental quality. This reflects a self-reinforcing “digital resource trap”, where ICT expansion paradoxically accelerates raw material depletion, undermines circular transitions, and ultimately impedes SDG12 progress—particularly in materially intensive systems. Our results strongly suggest that without deliberate policy intervention addressing the full life cycle material impacts and systemic drivers of consumption, the proliferation of ICT technologies risks locking economies into unsustainable resource pathways.

Thirdly, ICT goods export trade and technological innovation influence the effect of ICT on material footprint in all quantiles. More advanced trade structures demonstrably enhance ICT’s potential contribution to reducing material footprint, likely through knowledge spillovers and access to efficient global value chains. Meanwhile, globalization mainly mitigated the negative effects of ICT significantly at quantiles ranging from 30% to 70%. The more developed the ICT goods export trade, the better it can enhance ICT’s contribution to reducing material footprint. Further, the moderating effect of technological innovation is most pronounced at the 70% quantile level, with the strength of its effect increasing with the quantile level in other quantiles. It implies that the benefits of globalization for offsetting ICT’s footprint are primarily accessible to mid-level consumers, while high-footprint nations derive the greatest advantage from targeted technological innovation.

5.2. Policy Implications

Firstly, G20 countries require tailored strategies based on their national material footprint baselines. For high-MF developed countries (e.g., the United States, Canada, Australia), leadership in SDG12 implementation is imperative through accelerated low-carbon digital innovation and enforceable sectoral frameworks, particularly, mandating recycled-content thresholds in electronics/construction to curb excessive resource use. Concurrently, developing nations with low full-life-cycle material footprint (e.g., India, Indonesia) should leverage the Clean Development Mechanism to secure green financing and technological transfers, thereby preserving their circularity advantages while leapfrogging carbon-intensive trajectories. For developing countries with an elevated per-capita material footprint such as China, dual-track interventions are essential: stringent circularity policies targeting resource-intensive industrial zones to curb overconsumption, coupled with CDM-driven capacity building in underdeveloped regions to rectify spatial imbalances. Collective action across this spectrum will catalyze convergence toward global resource sustainability.

Secondly, future policies of G20 countries working towards SDG12 should take into account the unfavorable influence of ICT on material footprint. Given that developed countries experience a more pronounced material footprint increase with ICT development—especially at higher quantiles—their policies must focus on breaking the link between digitalization and material intensity. It can be argued that high-consumption countries ought to dedicate greater scrutiny to the consumption of materials used in ICT-generated equipment, to e-waste, and to reliance on non-renewable fossil energy sources. G20 countries are therefore urged to control resource depletion by optimizing ICT-related resource efficiency. For example, server virtualization in data centers can consolidate hardware, reducing both equipment needs and energy consumption. Additionally, strict limits should be imposed on e-waste. These limits should include caps on e-waste generation and controls on e-waste exports, consistent with the Basel Convention. Furthermore, the introduction of carbon taxes and resource taxes is necessary. These fiscal measures, however, require international coordination to avoid competitiveness concerns and emissions leakage, for instance, by harmonizing tax rates or using border carbon adjustments. This international coordination helps ensure that one country’s carbon/resource taxes do not simply shift production to another country.

Thirdly, promoting ICT export trade and facilitating globalization are required to be advocated. In particular, it is more beneficial for medium- and high-consumption countries to concentrate on these measures because of the stronger moderating effects. However, expanding ICT exports stimulates domestic production, and it may inadvertently increase the exporting nations’ material footprint without proper environmental management strategies. To mitigate this, G20 countries could emphasize cleaner, more resource-efficient ICT manufacturing (for example, by sharing advanced production technologies and promoting circular supply chains) as part of their export strategy. ICT goods export trade among G20 countries can be improved by reducing trade and political economy barriers. To enhance the process of globalization, it is imperative that G20 countries adhere to the principles of mutual benefit and win–win situations. Furthermore, it is of paramount importance to provide certain financial subsidies for ICT innovation and create conducive conditions for the exchange of ICT talents, thereby promoting technological progress.

However, due to data limitations, the study spans only from 2000 to 2020, focusing on the G20 as a whole without micro-level data from ICT enterprises in different countries. While this study utilizes material footprint as a core indicator for SDG12 progress, we acknowledge its limitation in capturing the goal’s full multidimensionality, such as circular economy performance and supply chain resilience. Future research should therefore prioritize developing integrated assessment frameworks that synergize material footprint with these complementary metrics to provide a holistic view of sustainable resource governance. Moreover, we will also extend the timeframe and employ machine learning techniques such as Quantile Regression Forests (QRFs) to explore whether micro-level data align with macro conclusions.