Impact of Supply Chain Management on Energy Transition and Environmental Sustainability: The Role of Knowledge Management and Green Innovations

Abstract

This study unpacks how supply chain management, knowledge management, and green innovations act as critical levers in driving energy transition while safeguarding environmental sustainability in an era of escalating climate challenges. Focusing on the G7 nations and using data from 2000 to 2022, this study addresses two central research questions: (i) What are the key determinants of energy transition (ET)? And (ii) what are the key determinants of environmental degradation (ED)? To answer these questions, the study applied Lewbel IV-2SLS and FGLS estimators, revealing that in G7 economies, supply chain performance reduces environmental degradation but slows energy transition. Digital transformation also hinders transition in the short run, though at higher maturity it helps curb degradation. Trade openness supports transition but increases degradation, while urbanization promotes transition. Knowledge management and green innovation follow an inverted-U pattern, and control of corruption shows mixed effects. Energy transition itself strongly reduces environmental degradation, whereas economic growth generally increases it. Based on these results, the study formulates a set of policy recommendations to align economic growth with long-term sustainability goals.

1. Introduction

The G7 nations—comprising Canada, France, Germany, Italy, Japan, the UK, and the USA—have seen increasing climate-driven urgency and intensified their energy transition policies accordingly. All members have pledged to fully or predominantly decarbonize their power systems by 2035, aligning with the International Energy Agency’s 1.5 °C scenario, and collectively committed over USD 500 billion toward clean energy post-COVID-19, with nearly 20% channeled to the electricity sector alone. Further commitment was evident when G7 ministers agreed to phase out coal-fired power plants by 2035, tripling renewable capacity and doubling energy efficiency by 2030 [1]. Public R&D spending on renewable energy technologies has also been shown to improve carbon efficiency in G7 economies. Energy transition investment, therefore, is clearly rising, driven by both policy commitments and economic imperatives’ progress [2,3].

Yet, achieving carbon neutrality by 2050 remains a formidable challenge owing to the significant scale of transformation needed. While all G7 members have net-zero pledges, transitions toward fully clean electricity must accelerate to meet those commitments [4]. Structural hurdles such as continuing fossil fuel subsidies—amounting to over USD 1.1 trillion in 2023, mostly consumption-based—undermine this. Additionally, lingering reliance on unabated gas and delayed coal-phaseout plans, particularly in coal-dependent members like Japan, further hamper momentum. Energy transition offers a clear solution: scaling up renewables, increasing energy efficiency, and developing clean supply chains can bridge the gap to carbon neutrality by cutting emissions at source, replacing fossil-based systems, and stimulating new green technologies and industries [5].

Supply chain management (SCM) plays a pivotal role in shaping both energy transition and environmental degradation by influencing how resources are procured, transported, and integrated into production systems. Efficient and sustainable supply chains enable firms and nations to shift toward low-carbon energy by reducing emissions embedded in logistics, procurement, and manufacturing networks [5,6]. For example, incorporating green procurement standards [7] and renewable-powered logistics can accelerate the adoption of clean energy technologies and reduce dependence on fossil-fuel-intensive inputs [8]. Conversely, poorly managed or carbon-intensive supply chains hinder the pace of energy transition and exacerbate environmental degradation by locking firms into fossil-based energy dependencies and creating rebound effects through inefficiencies [7,9]. Moreover, supply chain resilience—through localization, circular economy practices, and low-carbon sourcing—has been shown to reduce lifecycle emissions and facilitate the integration of renewable energy systems in industries with high energy demand [10,11]. In this sense, SCM operates as both a barrier and a catalyst: when aligned with sustainability principles, it lowers carbon footprints and supports transition to renewables, but when optimized solely for cost and speed, it often aggravates CO₂ emissions and delays the global decarbonization agenda [12,13]. Thus, embedding low-carbon practices in SCM is essential for achieving long-term carbon neutrality and mitigating environmental harm.

Green innovations (GI) significantly influence both energy transition and environmental degradation by fostering technological advancements that reduce carbon intensity and accelerate the shift toward renewable energy adoption. Through eco-innovation in renewable energy technologies, energy-efficient processes, and low-carbon industrial solutions, GI enhances the capacity of economies to decouple growth from emissions, thereby facilitating cleaner production systems [14]. At the same time, GI helps mitigate environmental degradation by improving resource efficiency, promoting circular economy practices, and reducing pollutant emissions across sectors [10,15]. Empirical evidence suggests that countries with higher levels of GI not only achieve faster energy transition but also experience substantial declines in greenhouse gas emissions, as innovation fosters both cleaner energy generation and more sustainable consumption patterns [16,17]. However, the effect of GI can sometimes be nonlinear—while initial innovations accelerate decarbonization, diminishing returns may emerge if deployment and diffusion lag behind technological breakthroughs [18]. Therefore, green innovations serve as a cornerstone for advancing carbon neutrality, directly enabling energy transition while simultaneously reducing the pace of environmental degradation.

Digital transformation (DT) impacts energy transition and environmental degradation by reshaping how energy systems, production processes, and consumption patterns are managed. On one hand, DT facilitates energy transition by enabling smart grids, digital monitoring, blockchain traceability, and AI-driven optimization, all of which improve efficiency, reduce energy losses, and accelerate the integration of renewable energy sources [19]. Digital platforms also enhance demand-side management and foster decentralized energy systems, such as rooftop solar and peer-to-peer energy trading, thereby supporting low-carbon pathways [20]. At the same time, DT reduces environmental degradation through predictive analytics, IoT-enabled resource management, and digital supply chain systems that minimize waste and emissions [21]. However, DT is not without challenges: the rebound effects of increased energy demand from data centers, cloud computing, and ICT infrastructure can partially offset environmental gains [9]. Thus, while DT is a catalyst for accelerating decarbonization and advancing sustainable energy systems, its net effect on environmental degradation depends on policy alignment, renewable-powered digital infrastructure, and sustainable deployment strategies.

This study contributes to the literature by integrating supply chain management (SC), digital transformation (DT), knowledge management (KM), and green innovations (GI) into a unified analytical framework to explain the dynamics of energy transition (ET) and carbon emissions (CEM). Unlike prior studies that examine these drivers in isolation, this research advances understanding by capturing their interactive and nonlinear effects, thereby reflecting the complex ways technological, organizational, and structural factors shape sustainability outcomes. By focusing on G7 economies, the study provides empirical evidence from advanced contexts where innovation, digitalization, and institutional capacity are pivotal to global decarbonization, strengthening debates on the determinants of ET and CEM.

Another major contribution lies in the methodological and theoretical insights offered. Using advanced econometric techniques—including cross-sectional dependence tests, Westerlund cointegration, Driscoll–Kraay standard errors, and Lewbel [22] 2SLS estimator—the study addresses challenges of endogeneity, heterogeneity, and nonlinearity, ensuring robust and reliable findings. Beyond methodology, it demonstrates that sustainability drivers such as DT, KM, and GI may generate threshold effects and diminishing returns, challenging linear assumptions in the literature. This not only advances econometric applications in sustainability research but also provides policymakers with a nuanced understanding of how to balance digitalization, innovation, and knowledge diffusion to accelerate ET while reducing CEM.

The next subsections are structured as follows: Section 2 presents the literature; Section 3 discloses the data and method; Section 4 discloses the findings; and Section 5 presents the conclusion and policy.

2. Theoretical Framework and Literature

2.1. Theoretical Framework

The energy transition (ET) represents a systemic shift from fossil-based energy systems toward renewable and low-carbon alternatives, and its determinants are rooted in innovation, organizational learning, supply chain governance, and macroeconomic structures. Green innovation (GI) provides the technological foundation for decarbonization by facilitating the development and deployment of renewable energy technologies. According to innovation diffusion theory, the effectiveness of GI depends not only on invention but also on the absorptive capacity of firms and institutions to adopt and integrate new technologies [14,18]. Thus, the theoretical expectation is that GI reduces carbon emissions (CEM) through technological upgrading, but this relationship is contingent on the institutional and regulatory environment.

Complementing innovation, knowledge management (KM) serves as an enabling mechanism by embedding absorptive capacity and organizational learning within firms and industries. Grounded in the knowledge-based view of the firm, KM ensures that innovation is not confined to research but extended to effective diffusion and deployment. The literature highlights that KM promotes ET by fostering energy-efficient practices, renewable adoption, and firm-level adaptation to sustainability transitions [20,23]. Theoretically, KM reduces CEM by improving decision-making, enabling sustainability-oriented cultures, and facilitating knowledge spillovers across sectors. Unlike GI, whose impact may vary across contexts, KM is consistently associated with positive ET outcomes and reduced emissions, positioning it as a stabilizing factor in the transition toward carbon neutrality.

Supply chains (SC) and economic growth (EG) introduce greater complexity into the theoretical framework. Transaction-cost and resource-dependence theories suggest that SCs can either accelerate or impede ET depending on alignment with low-carbon practices. Sustainable SC governance, supplier engagement, and life-cycle assessments are expected to reduce CEM and facilitate ET, whereas carbon-intensive dependencies may lock economies into fossil pathways [9,19]. Similarly, the relationship between EG, ET, and CEM is best explained through environmental Kuznets curve (EKC) dynamics and growth-environment trade-offs. Growth may stimulate ET by expanding resources for renewable investment [24,25], yet in fossil-dependent contexts, it increases energy demand and carbon intensity. Therefore, the framework anticipates that achieving ET and lowering CEM requires not only technological and organizational innovations but also institutional reforms, supply chain restructuring, and growth strategies aligned with sustainability.

2.2. Literature Review

2.2.1. Green Innovation and Energy Transition/Carbon Emissions

Green innovation (GI) is widely regarded as a cornerstone of the global energy transition (ET), yet the evidence reveals nuanced and sometimes contradictory dynamics. GI directly promotes ET by accelerating renewable adoption and improving clean-technology penetration, suggesting that countries investing in innovation benefit from faster decarbonization pathways [26]. However, Alexiou [27] and Bhuiyan et al. [14] reveal nonlinearities, showing that in some contexts GI initially diverts resources or reinforces incumbent fossil energy systems before yielding positive ET outcomes. These mixed findings imply that the effectiveness of GI depends heavily on absorptive capacity, regulatory frameworks, and the alignment of innovation policies with deployment incentives. From the emissions perspective, some studies [28,29,30] confirm that innovation-induced transitions substantially reduce CO₂ emissions (CEM). Thus, while GI is a necessary driver of ET and CEM mitigation, its impacts are conditional on structural and policy environments that determine whether innovation diffuses broadly or remains locked in elite technological niches.

2.2.2. Knowledge Management and Energy Transition/Carbon Emissions

The literature converges more consistently on the role of knowledge management (KM) in advancing ET and reducing CEM. Solnørdal and Thyholdt [23], Miremadi et al. [20], Reinhardt et al. [31], and Dzwigol et al. [21] show that KM fosters organizational learning, enhances absorptive capacity, and facilitates renewable technology adoption across European and OECD contexts. These findings emphasize that KM ensures innovation is not only produced but also effectively deployed and integrated within firms and industries. Parallel evidence confirms KM’s strong carbon-mitigation role: Provaty et al. [32] and Mariyakhan et al. [33] find that organizational knowledge and absorptive capacity significantly lower CO₂ intensity in U.S. and Chinese firms, while Geneidy et al. [34] and Seuring and Müller [8] show that improved knowledge-sharing practices reduce emissions from corporate operations and supply chains. These results reinforce the argument that KM serves as a strategic enabler—transforming innovation potential into tangible decarbonization outcomes by embedding sustainable practices at organizational and systemic levels.

2.2.3. Supply Chains and Energy Transition/Carbon Emissions

The impact of supply chains (SC) on ET and CEM remains contested, reflecting their dual role as both enablers and obstacles. Wen et al. [19] and Raygoza-Limón et al. [35] find that sustainable supply chain practices and life-cycle assessments of renewable energy projects contribute positively to ET, aligning supply networks with low-carbon objectives. However, [9,36] illustrate the opposite, showing that global supply chains often hinder ET through carbon-intensive dependencies, information asymmetries, and weak disclosure practices. When extended to emissions, the evidence is clearer. For instance, Liew and Cao [9], Karaman et al. [37], and Mohsin et al. [12] consistently document that robust supply chain governance and sustainable procurement reduce corporate and sectoral CO₂ emissions. Taken together, the findings suggest that supply chains can accelerate decarbonization if aligned with environmental objectives, but without strong governance and incentives, they risk reinforcing fossil-based production and slowing ET progress.

2.2.4. Economic Growth and Energy Transition/Carbon Emissions

The relationship between economic growth (EG), ET, and CEM is perhaps the most contested in the literature. On the one hand, Urom et al. [38] and Olanrewaju and Kirikkaleli [30] provide evidence that growth supports ET by expanding fiscal space for renewable investment, while Dogan [39] demonstrates a bidirectional relationship between growth and renewable energy in OECD nations, suggesting mutual reinforcement. On the other hand, Perera [40] and Bhuiyan et al. [41] highlight mixed and asymmetric effects, arguing that in some contexts growth exacerbates energy demand and slows transition progress. Evidence from emissions studies tilts more strongly toward a negative association. For instance, Işık et al. [42], Caldera et al. [43], and Zuhal and Göcen [44] all show that growth drives an increase in CO₂ emissions, particularly where fossil energy dominates the supply mix. This divergence underscores that while growth may facilitate ET in supportive policy environments, it simultaneously elevates emissions in fossil-dependent economies, pointing to the centrality of policy alignment and structural transformation in determining the net effect.

Table 1. Summary of determinants of energy transition.

Author(s)	Nation(s)	Periods	Method(s)	Findings(s)
Alexiou [27]	47 economies	1998–2017	System GMM	GI ↓↑ ET
Liu et al. [45]	Multi-country	1990s–2010s	Panel regression	GI ↑ ET
Alofaysan et al. [18]	38 emerging	2006–2021	FD-GMM	GI ↑ ET
Habiba et al. [15]	China/Global	1990s–2020s	Panel	GI ↑ ET
Albaker et al. [14]	MENA	1990–2021	Panel cointegration	GI ↓↑ ET
Solnørdal and Thyholdt [23]	Norway (firm-level)	2012–2015	Logit (752 SMEs)	KM ↑ ET
Miremadi et al. [20]	OECD/EU	1990s–2010s	Panel	KM ↑ ET
Reinhardt et al. [31]	Latin-EU cases	2010s	KM model	KM ↑ ET
Dzwigol et al. [21]	Europe	2000s–2020	Panel econometrics	KM ↑ ET
Wen et al. [19]	Multi-country	1990–2022	Panel	SC ↑ ET
Liew and Cao [9]	Global firms	2010s–2020s	Firm panel	SC ↓ ET
Zhang et al. [36]	Global (6 cases)	2022	Case study	SC ↓ ET
Raygoza-Limón et al. [35]	Global RE projects	2025	LCA review	SC ↑ ET
Urom et al. [38]	Multi-country	1990–2018	Asymmetric	EG↑ ET
Perera et al. [40]	152 nations	1990–2019	Granger causality	EG ↑↓ ET
Bhuiyan et al. [41]	Bangladesh	1990–2019	Simulation	EG ↑↓ ET
Dogan [39]	OECD	1980–2012	Panel ARDL	REC ↔ EG
Olanrewaju et al. [46]	USA	1995–2020	Wavelet Quantile	EG↑ ET

↓ decrease and ↑ increase, ↔ denotes feedback causality.

and

Table 2. Summary of determinants of carbon emissions.

Author(s)	Nation(s)	Periods	Method(s)	Findings(s)
Khaoula Aliani et al. [29]	G7	2000–2019	Panel regressions	ET ↓ CEM
Gao et al. [47]	China	2008–2020	Panel fixed-effects; mediation	ET ↓ CEM
Zeng et al. [48]	China	2001–2019	Spatial econometrics (SDM)	ET ↓ CEM
Shi et al. [49]	China	2006–2019	GTWR spatial	ET ↓ CEM
Tuo et al. [50]	71 countries	1996–2018	Panel analysis	ET ↓ CEM
Provaty et al. [32]	USA firms	2002–2019	Firm-level panel; 2SLS/PSM	KM ↓ CEM
Mariyakhan et al. [33]	USA & China	1970–2018	Nonlinear ARDL	KM ↓ CEM
Geneidy et al. [34]	Multinational org.	2019 baseline	Carbon footprinting (Scope 1–3)	KM ↓↑ CEM
Seuring and Müller [8]	Europe (4 orgs)	2019–2023	Case study	KM ↓ CEM
Liew and Cao [9]	Global firms	2010s–2020s	Firm-level analysis	SC ↓ CEM
Karaman et al. [37]	Global firms	2010s–2020s	Panel; 2SLS; PSM	SC ↓ CEM
Mohsin et al. [12]	13 EU	2000–2022	Panel ARDL; causality	SC ↓ CEM
Işık et al. [42]	27 OECD	2001–2020	Panel quantile regression	EG ↑ CEM
Zuhal and Göcen [44]	USA	1973–2022	VAR/cointegration	EG ↑ CEM

↓ decrease and ↑ increase.

present a summary of the findings.

2.3. Hypotheses Formulation

The study formulates the following null hypotheses based on its research objectives:

Supply chain management

H1: 

Supply chain management has no significant effect on energy transition.

H2: 

Supply chain management has no significant effect on environmental degradation.

Digital transformation

H3: 

Digital transformation has no significant effect on energy transition.

H4: 

Digital transformation has no significant effect on environmental degradation.

H5: 

The squared term of digital transformation has no significant nonlinear effect on energy transition.

H6: 

The squared term of digital transformation has no significant nonlinear effect on environmental degradation.

Trade Openness

H7: 

Trade openness has no significant effect on energy transition.

H8: 

Trade openness (has no significant effect on environmental degradation.

Urbanization

H9: 

Urbanization has no significant effect on energy transition.

H10: 

Urbanization has no significant effect on environmental degradation.

Climate Change

H11: 

Climate change has no significant effect on energy transition.

H12: 

Climate change has no significant effect on environmental degradation.

Economic growth

H13: 

Economic growth has no significant effect on energy transition.

H14: 

Economic growth has no significant effect on environmental degradation.

Knowledge management (KM)

H15: 

Knowledge management has no significant effect on energy transition.

H16: 

Knowledge management has no significant effect on environmental degradation.

H17: 

The squared term of knowledge management has no significant nonlinear effect on energy transition.

H18: 

The squared term of knowledge management has no significant nonlinear effect on environmental degradation.

Green innovation

H19: 

Green innovation has no significant effect on energy transition.

H20: 

Green innovation has no significant effect on environmental degradation.

H21: 

The squared term of green innovation has no significant nonlinear effect on energy transition.

H22: 

The squared term of green innovation has no significant nonlinear effect on environmental degradation.

3. Data and Methods

3.1. Data

Our sample consists of the G7 economies for the period 2001–2022, selected due to their global significance in energy consumption, technological advancement, and carbon emissions (CEM). These advanced economies are ideal for examining the determinants of the energy transition (ET) and their role in reducing environmental pressures. The study period is guided by data availability for ET, CEM, and the explanatory variables, ensuring consistency across countries. The dependent variables are carbon emissions (CEM) and energy transition (ET). CEM represents the environmental outcome of economic and industrial activities, while ET captures the extent to which economies are shifting from fossil-fuel-based systems toward cleaner and more sustainable energy use. The data for ET is measured as renewable energy consumption (kWh per capita), while the data for CEM is measured as per capita CO₂ emissions. Both CEM and ET are obtained from the WDI [51].

The main explanatory variable is supply chain management (SC), which plays a pivotal role in shaping energy transition (ET) and carbon emissions (CEM) through procurement practices, logistics efficiency, and the carbon intensity of production networks. In addition, three composite indices are constructed to capture technological and organizational drivers of energy and environmental outcomes: digital transformation (DT), knowledge management (KM), and green innovation (GI).

The DT index is derived from indicators of digital penetration and connectivity, including internet users (% of population), fixed broadband subscriptions (per 100 people), mobile cellular subscriptions (per 100 people), secure internet servers (per million people), ICT goods exports (% of total goods exports), and ICT service exports (% of total service exports). The KM index reflects knowledge creation and absorptive capacity, constructed from R&D expenditure (% of GDP), researchers per million people, scientific and technical journal articles, patent applications by residents, high-tech exports (% of manufactured exports), and ICT service exports (% of service exports). The GI index captures eco-innovation and technological advancement toward sustainability, measured through R&D expenditure (% of GDP), the number of scientific and technical journal articles, population-adjusted research output (per million people), and high-tech exports as a share of manufactured exports. All the variables are sourced from WDI [52].

To ensure methodological rigor and avoid multicollinearity, these indices (DT, KM, and GI) are constructed using Principal Component Analysis (PCA), which reduces dimensionality while preserving the maximum variance in the data. This approach yields robust composite indicators that capture the latent dimensions of digitalization, knowledge management, and green innovation, thereby allowing for a comprehensive analysis of their roles in influencing ET and CEM. The study used STATA 14 for the analysis.

3.2. Methods

This study used Driscoll–Kraay (DK), Feasible Generalized Least Squares (FGLS), and Lewbel IV techniques. These methods are advantageous in this context because these methods are specifically designed to address heteroskedasticity, serial correlation, and endogeneity in relatively small to medium panel samples, which characterizes much of the sustainability and energy transition data. DK provides robust standard errors even under cross-sectional dependence, FGLS efficiently handles heteroskedastic and autocorrelated error structures, and Lewbel IV generates internal instruments to address endogeneity when external instruments are weak or unavailable. By contrast, approaches such as CCE, AMG, System-GMM, or PMG are more suited for very large panels, dynamic settings, or long time horizons, but they can lead to biased or inefficient estimates when sample sizes are moderate and instrument proliferation or cross-sectional heterogeneity becomes a concern. Thus, DK/FGLS/IV provide a more reliable and context-appropriate framework for validating results in this study.

Using panel data, we analyze the effects of supply chain management (SC) on energy transition (ET) and carbon emissions (CEM), with the estimable model defined in Equations (1) and (2).

$$E{T}_{i,t}={\gamma }_i+{\beta }_1{ \mathrm{SC}}_{i,t}+{\delta }_j{X}_{i,t}+{\mu }_t+{\epsilon }_{i,t}$$

(1)

$$CE{M}_{i,t}={\gamma }_i+{\beta }_1{ \mathrm{SC}}_{i,t}+{\delta }_j{X}_{i,t}+{\mu }_t+{\epsilon }_{i,t}$$

(2)

Let $\mathrm{i}$ index countries and $\mathrm{t}$ years. ${ET}_{\mathrm{i},\mathrm{t}}$ and ${CEM}_{\mathrm{i},\mathrm{t}}$ is the vector of energy transition and carbon emissions proxy for environmental degradation; ${\mathrm{S}C}_{\mathrm{i},\mathrm{t}}$ denotes supply chain management, and the control variable is denoted by ${\mathrm{X}}_{\mathrm{i},\mathrm{t}}$t. The model incorporates country-specific effects (γᵢ), the coefficient β₁ associated with supply chains (SC), a K × 1 vector of control-variable coefficients (δ), time fixed effects (μₜ), and the idiosyncratic error term (εᵢ,ₜ).

Panel data analysis often faces econometric challenges such as cross-sectional dependence, serial correlation, and heteroscedasticity, which, if unaddressed, may bias results and compromise inference. Cross-sectional dependence arises when unobserved shocks or spillovers simultaneously affect multiple countries or firms, a common phenomenon in globalization and environmental studies [53]. Similarly, serial correlation occurs when error terms are correlated over time within a unit, while heteroscedasticity reflects unequal error variances across observations. Correcting for these issues is essential to avoid inefficient and inconsistent estimators. Employing robust estimation techniques that account for these problems ensures valid statistical inference and more reliable policy recommendations [54].

The [55] cointegration test is widely employed in panel econometrics to examine long-run relationships among variables, particularly in the presence of cross-sectional dependence and structural breaks. Unlike traditional Engle–Granger or Pedroni tests, Westerlund’s test is based on error-correction dynamics, making it more powerful in detecting cointegration when heterogeneity exists across countries or firms [55]. This feature is particularly valuable in sustainability and macroeconomic studies, where countries may converge toward long-run equilibrium at different speeds. Thus, the Westerlund test provides a more reliable framework for validating long-run associations between economic growth, energy use, innovation, and environmental outcomes.

Driscoll–Kraay Standard Errors and Feasible Generalized Least Squares (FGLS)

To obtain robust inference in the presence of cross-sectional dependence, Driscoll–Kraay standard errors are commonly applied. These standard errors are heteroscedasticity- and autocorrelation-consistent, while also robust to general forms of spatial and temporal dependence [56]. Their use allows researchers to maintain valid inference even in unbalanced panels with nonstationary error structures. Similarly, Feasible Generalized Least Squares (FGLS) is particularly effective in addressing heteroscedasticity and contemporaneous correlation across units, thereby producing efficient parameter estimates [57]. By combining Driscoll–Kraay with FGLS, studies can ensure that results remain robust across diverse forms of dependence and error variance, making findings more credible for policy analysis.

Finally, endogeneity is a recurring concern in applied econometrics, particularly when regressors are correlated with error terms due to omitted variables, simultaneity, or measurement error. Traditional instrumental variable (IV) approaches require strong external instruments, which are often unavailable. To address this, this study used a 2SLS estimator that generates internal instruments from model residuals under specific conditions. This method has gained prominence in recent empirical research because it allows consistent estimation without relying on external instruments, thereby overcoming one of the main limitations of conventional IV techniques [22]. Its application is particularly relevant in studies on energy transition, innovation, and growth, where finding valid instruments is inherently difficult.

4. Results

4.1. Descriptive Statistics and Correlation

Table 3. Descriptive Statistics.

Variables	Obs	Mean	Std. Dev.	Min	Max
KM	161	0	1	−2.024	1.6
GI	161	0	1	−2.437	1.264
SC	161	0	1	−1.824	2.663
DT	161	0	1	−2.526	1.7
TRD	161	52.691	17.368	19.56	89.064
UB	161	0.761	0.525	−1.602	2.283
CC	161	1.416	0.521	0.006	2.083
LnEG	161	10.589	0.178	10.291	11.07
ET	161	1.723	0.365	1.276	2.704
LnCEM	161	2.267	0.428	1.448	3.061

presents the descriptive statistics results. The results show that KM, GI, SC, and DT were standardized as z-scores (mean ≈ 0, sd ≈ 1) by design; their ranges (down to about −2.5 and up to ~2.7) indicate observations up to ~2–3 standard deviations from the mean but no extreme outliers. TRD has the largest raw dispersion (mean = 52.69, sd = 17.37, range = 19.56–89.06), pointing to strong cross-sectional or time variation. UB (mean = 0.76, sd = 0.53) spans negative to positive values (−1.60 to 2.28), consistent with an index or growth-rate–type measure. CC is strictly positive (mean = 1.42, sd = 0.52), suggesting a moderately concentrated distribution. The logged macro variables—lnEG (10.59, sd = 0.18) and lnCEM (2.27, sd = 0.43)—exhibit relatively low variability, while ET (mean = 1.72, sd = 0.37) shows moderate spread, implying noticeable but not extreme movements in the energy-transition proxy. Overall, variability is greatest for TRD and in the tails of the standardized constructs, whereas the logged series are comparatively stable.

Table 4. (a) Correlation with ET; (b) correlation with CEM.

a
Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
(1) ET	1.000
(2) SC	−0.396	1.000
(3) DT	0.353	0.206	1.000
(4) KM	−0.358	0.843	−0.023	1.000
(5) GI	−0.145	0.695	0.148	0.897	1.000
(6) TRD	0.485	−0.496	0.441	−0.606	−0.333	1.000
(7) UB	0.213	−0.063	−0.254	0.033	0.051	−0.180	1.000
(8) CC	0.082	0.202	0.098	0.316	0.567	0.255	0.159	1.000
(9) lnEG	−0.042	0.723	0.308	0.445	0.442	−0.159	0.259	0.432	1.000
b
Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
(1) CEM	1.000
(2) ET	−0.328	1.000
(3) SC	0.321	−0.396	1.000
(4) DT	−0.239	0.353	0.206	1.000
(5) KM	0.272	−0.358	0.843	−0.023	1.000
(6) GI	0.326	−0.145	0.695	0.148	0.897	1.000
(7) TRD	−0.319	0.485	−0.496	0.441	−0.606	−0.333	1.000
(8) UB	0.331	0.213	−0.063	−0.254	0.033	0.051	−0.180	1.000
(9) CC	0.398	0.082	0.202	0.098	0.316	0.567	0.255	0.159	1.000
(10) lnEG	0.560	−0.042	0.723	0.308	0.445	0.442	−0.159	0.259	0.432	1.000

presents the correlation results. The correlation matrices paint a consistent picture for the G7. Energy transition (ET) is negatively correlated with supply chain performance (SC, −0.396) and knowledge management (KM, −0.358) and only mildly with green innovation (GI, −0.145), but positively correlated with digital transformation (DT, 0.353), trade openness (TRD, 0.485) and, to a lesser extent, urbanization (UB, 0.213) and governance (CC, 0.082). Emissions (CEM) move in the opposite direction: CEM falls as ET rises (−0.328), and is lower where DT (−0.239) and TRD (−0.319) are higher, but higher where SC (0.321), KM (0.272), GI (0.326), UB (0.331), CC (0.398) and especially economic scale (lnEG, 0.560) are larger—consistent with scale effects. A striking feature is the tight clustering among capability variables—SC, KM, and GI are very strongly interrelated (0.695–0.897) and each correlates with lnEG and CC—flagging potential multicollinearity if included together.

4.2. Cross-Sectional Dependence Result

The study proceeds to test for cross-sectional dependence (CD) among the G7 countries (see

Table 5. Cross-sectional Dependence Result.

Variables	CD-Test	p-Value	Mean	Mean Abs
KM	4.765	0.000	0.22	0.45
GI	10.52	0.000	0.48	0.49
SC	20.17	0.000	0.92	0.92
DT	21.72	0.000	0.99	0.99
TRD	9.111	0.000	0.41	0.66
UB	3.676	0.000	0.17	0.26
CC	8.023	0.000	0.37	0.58
LnEG	12.94	0.000	0.59	0.75
ET	13.32	0.000	0.61	0.71
LnCEM	20.00	0.000	0.91	0.91

), given their high economic and policy integration. Checking the CD is essential because common shocks and spillovers can induce contemporaneous correlation across panels; detecting it guides the choice of second-generation panel methods and prevents biased estimates and invalid inference. For the G7 panel, all CD-tests are highly significant (p < 0.01), so cross-sectional independence is decisively rejected across variables. Dependence is very strong for DT, SC, and lnCEM (Mean ≈ 0.91–0.99), substantial for lnEG and ET (≈0.71–0.75), moderate for GI, TRD, and CC (≈0.49–0.66), and non-trivial for UB (≈0.26). This pattern is consistent with tightly integrated G7 economies that share common shocks—global energy prices, technology diffusion, policy cycles, and financial conditions. Accordingly, inference should rely on second-generation techniques.

4.3. CADF Stationarity Test Result

Table 6. CADF test.

	Level	p-Value	First Difference	p-Value
KM	−1.051	0.974	−2.561	0.014
GI	−1.360	0.863	−2.647 ***	0.008
SC	−1.990	0.265	−3.691 ***	0.000
DT	−1.688	0.578	−3.519 ***	0.000
TRD	−1.291	0.899	−2.398 **	0.041
UB	−2.406 **	0.039	−3.218 ***	0.000
CC	−2.015	0.244	−3.168 ***	0.000
LnEG	−1.063	0.971	−2.424 **	0.035
ET	−1.805	0.451	−3.249 ***	0.000
LnCEM	−1.999	0.257	−3.523 ***	0.000

Note: ** p < 0.05, *** p < 0.01.

shows CADF unit-root tests that account for cross-sectional dependence. At levels, the unit-root null cannot be rejected for all variables (p > 0.24) except UB (stat = −2.406, p = 0.039), which is stationary. After first differencing, all series become stationary: GI, SC, DT, CC, ET, and lnCEM are stationary at 1% (*), TRD and lnEG at 5% (), and KM is also significant (p = 0.014). Thus, most variables are I(1) while UB is I(0), motivating cointegration/Lewbel IV-2SLS.

4.4. Westerlund Cointegration

Table 7. Westerlund Cointegration.

	(1)	(2)	(3)	(4)	(5)	(6)
	Energy Transition			Environmental Sustainability
VR	−2.89 ***	−3.03 ***	−2.78 **	−2.53 **	−2.88 ***	−2.92 ***

Note: VR denotes the variance ratio. ** p < 0.05; *** p < 0.01.

reports Westerlund’s panel cointegration results using the variance-ratio (VR) statistic across six model specifications for both energy transition and environmental degradation. In every case, the VR statistic is negative and statistically significant—at the 1% level in columns (1), (2), (5), and (6), and at the 5% level in columns (3) and (4)—which leads to rejecting the null of no cointegration. This provides consistent evidence of a long-run equilibrium relationship among the variables in each specification for both constructs. Practically, it justifies proceeding with long-run estimators.

4.5. Baseline Regression Result

Table 8. Effects of supply chain management on energy transition and environmental degradation (Driscoll–Kraay standard error results).

	(1)	(2)	(3)	(4)	(5)	(6)
	Energy Transition			Environmental Degradation
SC	−0.268 ***	−0.686 ***	−0.414 ***	−0.295 **	−0.389	−0.561 ***
	(0.019)	(0.197)	(0.094)	(0.116)	(0.267)	(0.095)
DT	−0.546 ***			−0.025
	(0.168)			(0.589)
TRD	0.012 ***	0.011 ***	0.007 ***	−0.009	−0.008	−0.006
	(0.003)	(0.002)	(0.001)	(0.007)	(0.007)	(0.004)
UB	0.201 ***	0.083 **	0.131 ***	−0.013	−0.029	−0.079
	(0.035)	(0.036)	(0.040)	(0.101)	(0.095)	(0.066)
CC	0.118	−0.321 **	−0.276 *	0.231 ***	0.159	−0.157
	(0.069)	(0.135)	(0.152)	(0.074)	(0.202)	(0.127)
LnEG	0.945 ***	1.929 ***	1.212 ***	2.159 **	2.438 **	3.161 ***
	(0.103)	(0.647)	(0.331)	(0.809)	(0.878)	(0.477)
KM		0.453 ***			0.086
		(0.146)			(0.277)
GI			0.237 **			0.287 **
			(0.089)			(0.115)
ET				−0.422 ***	−0.460 ***	−0.587 ***
				(0.049)	(0.129)	(0.076)
Constant	−10.389 ***	−18.998 ***	−11.219 ***	−19.638 *	−22.427 **	−29.409 ***
	(1.605)	(6.695)	(3.298)	(9.970)	(9.076)	(4.758)
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Observations	161	161	161	161	161	161
R-squared	0.559	0.630	0.593	0.660	0.662	0.719

Note: * p < 0.10, ** p < 0.05, and *** p < 0.01.

presents the baseline regression result. In Model 1 (Baseline ET model), for G7 economies, supply chain management (SC) loads negatively on energy transition (ET) (−0.268 ***), while digital transformation (DT) also shows a negative association (−0.546 ***). By contrast, Trade openness (TRD) and Urbanization (UB) relate positively to ET (both significant), and economic growth (lnEG) is strongly positive (0.945 ***). A plausible mechanism is that SC improvements in mature, fossil-anchored networks first optimize incumbent chains (leaner logistics, better scheduling) before the energy mix itself switches—dampening measured “transition” in the short run even as processes become cleaner. Meanwhile, DT can raise near-term electricity demand (data centers, networks), offsetting efficiency gains (a rebound dynamic), whereas TRD facilitates diffusion of clean technologies and standards in highly regulated G7 markets; UB lowers per-capita transport and heating loads via density and electrification. These channels align with sustainable SCM reviews, ICT rebound evidence, and trade–environment theory [8,58].

Adding knowledge management (KM) (see Model 2) yields a positive and significant effect on ET (0.453 ***), with SC remaining negative (−0.686 ***). KM strengthens firms’ abilities to capture, share, and apply codified and tacit know-how for low-carbon process redesign, supplier development, and project execution—accelerating adoption of renewables/electrification along the chain. Recent evidence links “green” KM practices to higher rates of green innovation and improved sustainability performance, consistent with the channel “KM → capabilities → faster ET.” The persisting negative SC coefficient suggests transitional lock-in: when SC metrics reward cost and on-time delivery in legacy networks, structural energy switching can lag until governance, targets, and supplier criteria are retuned around decarbonization [59,60].

Introducing green innovation (GI) (see Model 3) increases ET (0.237 **), while SC stays negative (−0.414 ***); TRD and UB remain positive and significant, and lnEG remains strongly positive. GI captures product/process inventions that reduce energy use or emissions, easing integration of renewables, storage, and electrified processes. Meta-analyses show GI has a systematically positive effect on environmental performance, which is coherent with its role in enabling transition technologies and accelerating learning curves in G7 manufacturing and power systems. The still-negative SC term again points to timing: innovation signals capability, but full supply-base switch-over (contracts, qualification, risk assurance) is slower, so short-run ET can trail GI unless purchasing and logistics KPIs explicitly internalize decarbonization [26].

In Model 4, with environmental degradation (CEM) as the dependent variable, ET enters negatively and strongly (−0.422 ***): a deeper transition is associated with lower degradation—an internal validity check for the framework. SC is also negative (−0.295 **), implying greener logistics, reverse flows, and supplier collaboration reduce emissions and wastes even before complete energy mix changes. LnEG is positive (2.159 **), reflecting scale effects of output on environmental pressure in high-income economies; DT is statistically null here, consistent with mixed near-term net effects of digitalization. The ET→ED link matches consensus assessments that ramping renewables, efficiency, and electrification reduces GHGs and co-pollutants in advanced economies [8].

In Model 5, KM is small and not significant (0.086), while ET remains strongly negative (−0.460 ***), SC is negative but imprecise (−0.389), and lnEG increases ED (2.438 **). This suggests KM’s environmental benefits are indirect and lagged—operating mainly through the acceleration of innovation and capability building rather than immediate emission cuts. In the short run, knowledge processes (codification, sharing platforms, training) raise absorptive capacity; measurable ED improvements materialize as those capabilities translate into deployed technologies, redesigned processes, and supplier transitions. Evidence on KM’s contribution to eco-innovation and performance supports this sequencing [61].

In Model 6, ET remains negative and sizable (−0.587 ***), SC is negative (−0.561 ***), lnEG is positive (3.161 ***), and GI shows a positive coefficient (0.287 **). The GI sign can look counter-intuitive; a common explanation is timing and composition: GI proxies (e.g., patents) often peak before large-scale deployment, during intensive R&D and prototyping phases that can temporarily raise energy/material use—especially in capital- and knowledge-intensive G7 sectors. Meta-evidence confirms GI improves environmental performance on average, but effects vary by GI type and implementation stage; early-stage innovation may raise measured ED until adoption diffuses and legacy assets retire. This is consistent with a “J-curve of green innovation deployment” [26].

Regarding the control variables, TRD is pro-ET in (1)–(3) but weakly negative and insignificant for ED in (4)–(6), consistent with G7 conditions where technique/composition effects (clean capital, standards) can dominate scale effects under strong regulation. UB is pro-ET (densification enables electrified mobility and efficient networks) but does not significantly reduce ED contemporaneously—again pointing to staging and infrastructure lags. Control of corruption (CC) shows mixed signs across ET and ED, but the broader literature finds that stronger institutions improve environmental policy stringency and outcomes; noisy short-run estimates can reflect collinearity with growth and concurrent reforms.

The six models paint a consistent picture. We observed that transition (ET) reduces degradation (ED); GI and KM increase ET (but GI’s direct effect on ED can be positive at early stages); SC lowers ED yet may initially slow measured ET unless procurement/logistics KPIs are aligned with decarbonization; DT risks rebound without complementary policy; TRD and UB can be allies for ET under strong rules; and growth raises both ET capacity and environmental pressure.

4.6. Robustness Check

The FGLS robustness in

Table 9. Effects of supply chain on energy transition and environmental sustainability (Robustness Check-FGLS) results.

	(1)	(2)	(3)	(4)	(5)	(6)
	Energy Transition			Environmental Degradation
SC	−0.292 ***	−0.427 ***	−0.340 ***	−0.215 ***	−0.239 ***	−0.316 ***
	(0.027)	(0.067)	(0.038)	(0.040)	(0.078)	(0.056)
DT	−0.221 **			0.006
	(0.106)			(0.113)
TRD	0.004 *	0.005 **	0.002	−0.006 ***	−0.005 ***	−0.006 ***
	(0.002)	(0.002)	(0.001)	(0.002)	(0.002)	(0.001)
UB	0.122 ***	0.086 **	0.081 **	0.077 ***	0.072 **	0.045
	(0.028)	(0.037)	(0.034)	(0.026)	(0.030)	(0.029)
CC	0.085 **	−0.104 *	−0.090 *	0.224 ***	0.206 ***	0.152 **
	(0.033)	(0.056)	(0.050)	(0.033)	(0.057)	(0.066)
LnEG	1.175 ***	1.421 ***	1.227 ***	1.973 ***	2.056 ***	2.357 ***
	(0.152)	(0.229)	(0.160)	(0.175)	(0.259)	(0.211)
KM		0.186 ***			0.021
		(0.057)			(0.059)
GI			0.108 ***			0.065
			(0.031)			(0.042)
ET				−0.606 ***	−0.617 ***	−0.640 ***
				(0.047)	(0.053)	(0.051)
Constant	−11.723 ***	−13.667 ***	−11.483 ***	−17.511 ***	−18.375 ***	−21.389 ***
	(1.772)	(2.377)	(1.625)	(2.036)	(2.665)	(2.124)
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Observations	161	161	161	161	161	161
R-squared	0.80	0.73	0.67	0.74	0.84	0.63

Note: * p < 0.10, ** p < 0.05, *** p < 0.01. Models (1)–(3) focus on the determinants of energy transition (ET), while Models (4)–(6) focus on environmental degradation (ED). Within each set, additional sustainability drivers are introduced stepwise: Model (1) includes the baseline variables (SC, DT, TRD, UB, CC, and EG); Model (2) adds knowledge management (KM); Model (3) further incorporates green innovation (GI); Model (4) begins with baseline variables plus ET; Model (5) adds KM; and Model (6) adds GI.

broadly confirms the baseline patterns for the G7 while tightening standard errors and lifting fit (R² up to 0.80 for ET and 0.84 for ED). In the ET equations (Models 1–3), SC remains negative and highly significant (−0.292 *** to −0.340 ***), consistent with a short-run “optimization of legacy chains” story: greener procurement and logistics initially squeeze waste in fossil-anchored networks but do not immediately flip the energy mix, so measured transition can dip even as processes get cleaner. DT stays negative for ET (−0.221 **), consistent with rebound and the electricity footprint of data centers and connectivity offsetting efficiency gains without complementary clean power. TRD is pro-ET but with smaller elasticities than in the baseline (0.002–0.005, weak-to-moderate significance), reflecting that technology diffusion and standards help, yet most transition progress in the G7 is policy- and investment-driven. Urbanization (UB) is positive for ET (0.081 **–0.122 ***), in line with density enabling electrified transport, district systems, and smart grids in advanced economies. CC shows mixed signs across ET models (positive in Model 1; negative in Models 2–3), which is plausible given collinearity with growth episodes and the timing of regulatory upgrades; nonetheless, the literature’s long-run view is that stronger institutions facilitate transition by reducing transaction costs and uncertainty. KM (0.186 ***) and GI (0.108 ***) each strengthen ET, capturing capability building and invention that ease the adoption of low-carbon technologies.

Turning to the ED equations (cols. 4–6), ET remains strongly negative and even larger in magnitude than in Table 6 (−0.606 *** to −0.640 ***), reinforcing the core identification that deeper transition reduces environmental degradation—fully consistent with mitigation assessments for advanced economies. SC is also negative and significant for ED (−0.215 *** to −0.316 ***), indicating that greener logistics, supplier collaboration, and reverse flows lower emissions and waste independently of the energy-mix shift. DT is essentially null for ED, echoing the mixed near-term net impacts of digitalization. Notably, TRD turns clearly negative and significant for ED (−0.005 *** to −0.006 ***), a sharper result than in Table 6; in the G7 context, stringent domestic regulation and imported clean capital mean technique/composition effects can dominate scale effects, so openness can coincide with lower environmental pressure. UB is positive for ED and significant in cols. 4–5 (0.072 **–0.077 ***), consistent with construction intensity and higher consumption densities.

Across both sets of models, economic growth (lnEG) is robustly positive (ET: 1.175 ***–1.421 ***; ED: 1.973 ***–2.357 ***), reflecting the scale–composition–technique decomposition: growth expands fiscal space and demand for clean technologies (pro-ET) while raising total activity and material throughput (pro-ED) unless policy accelerates the composition/technique shift. CC is positive for ED (0.152 **–0.224 ***), which is counter-intuitive if higher values mean stronger control of corruption; two explanations are common in applied panels: (i) co-movement with growth-and-reform spurts that also scale activity, and (ii) measurement/reporting improvements that lift observed ED measures as monitoring tightens. Importantly, the large, stable ET→ED reductions and the consistently negative SC and ED link indicate that once decarbonization deployment occurs and supply chains internalize environmental KPIs, environmental pressures fall.

4.7. Results of Lewbel Two-Stage Least Squares Method

Table 10. Effects of supply chain on energy transition and environmental sustainability (Lewbel IV-2SLS results).

	(1)	(2)	(3)	(4)	(5)	(6)
	Energy Transition			Environmental Degradation
SC	−0.363 ***	−0.553 ***	−0.288 ***	−0.457 ***	−0.365 ***	−0.639 ***
	(0.041)	(0.113)	(0.065)	(0.064)	(0.141)	(0.065)
DT	−0.641 ***			−0.218
	(0.150)			(0.243)
TRD	0.010 ***	0.011 ***	0.008 ***	−0.010 ***	−0.008 ***	−0.006 ***
	(0.003)	(0.002)	(0.002)	(0.004)	(0.003)	(0.001)
UB	0.150 ***	0.130 ***	0.195 ***	−0.065	−0.024	−0.103 **
	(0.039)	(0.050)	(0.053)	(0.056)	(0.054)	(0.045)
CC	0.156 ***	−0.252 ***	−0.203 ***	0.296 ***	0.171 *	−0.204 ***
	(0.043)	(0.075)	(0.065)	(0.051)	(0.103)	(0.078)
LnEG	1.325 ***	1.492 ***	0.748 ***	2.797 ***	2.363 ***	3.433 ***
	(0.217)	(0.375)	(0.265)	(0.348)	(0.457)	(0.252)
KM		0.356 ***			0.069
		(0.091)			(0.115)
GI			0.169 ***			0.330 ***
			(0.045)			(0.050)
ET				−0.526 ***	−0.449 ***	−0.633 ***
				(0.057)	(0.080)	(0.063)
Constant	−12.195 ***	−14.251 ***	−6.305 **	−26.025 ***	−21.847 ***	−32.386 ***
	(2.199)	(3.893)	(2.704)	(3.532)	(4.691)	(2.475)
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Observations	161	161	161	161	161	161
R-squared	0.547	0.624	0.580	0.639	0.662	0.717

Note: * p < 0.10, ** p < 0.05, *** p < 0.01. Models (1)–(3) focus on the determinants of energy transition (ET), while Models (4)–(6) focus on environmental degradation (ED). Within each set, additional sustainability drivers are introduced stepwise: Model (1) includes the baseline variables (SC, DT, TRD, UB, CC, and EG); Model (2) adds knowledge management (KM); Model (3) further incorporates green innovation (GI); Model (4) begins with baseline variables plus ET; Model (5) adds KM; and Model (6) adds GI.

addresses potential endogeneity in supply chain management (SC) using [22] heteroskedasticity-generated instruments. This strategy identifies causal effects from within-sample variance when valid external instruments are scarce, under mild conditions on heteroskedastic residuals. With time-fixed effects held constant and the same G7 sample (N = 161), the IV estimates keep the signs seen earlier but often show sharper magnitudes (e.g., more negative SC effects), suggesting that simultaneity (e.g., greener firms adopting SCM because they are already transitioning) or measurement error were biasing conventional estimates toward zero.

In Models (1)–(3), the dependent variable is energy transition (ET). SC is negative and highly significant (−0.363 *** to −0.553 ***), and DT is also negative when included (−0.641 ***). A coherent mechanism for the G7 is short-run optimization of legacy, fossil-anchored supply chains: greener procurement, lean logistics, and tighter scheduling reduce waste and emissions within existing energy systems but do not instantly flip the electricity/fuels mix captured by ET indices—so measured transition can dip while processes get cleaner [8]. DT adds computing and connectivity loads (data centers, networks) that can offset efficiency gains without parallel clean-power procurement—classic rebound [58]. By contrast, trade openness (TRD) is consistently pro-ET (0.008 ***–0.011 ***), reflecting technique/composition channels in tightly regulated G7 markets that import cleaner capital and diffuse standards [62]. Urbanization (UB) is also pro-ET (0.130 ***–0.195 ***), consistent with dense systems enabling electrified transport, district energy, and smart grids. Economic growth (lnEG) remains strongly positive (0.748 ***–1.492 ***): growth expands fiscal space for clean investment even as it raises total energy demand.

When KM is added (Model 2), it is positive and significant (0.356 ***), consistent with the role of codification, sharing, and absorptive capacity in lowering adoption costs for low-carbon technologies along the chain and speeding organizational learning (innovation-capability literature). Notably, CC turns negative for ET in cols. (2)–(3) despite being positive in (1). With KM (and later GI) in the model, CC may proxy concurrent reform/growth spurts that prioritize compliance, monitoring, and risk control before large-scale energy substitution shows up in the transition index—producing a short-run negative correlation with ET even if the long-run effect is supportive [63]. This underscores why IV matters: it changes not the qualitative story but the timing interpretation—capabilities (KM) lift ET directly, while governance improvements can arrive in phases that first tighten procedures, then unlock deployment.

Adding GI (Model 3) yields a positive, significant effect on ET (0.169 ***), capturing the invention that eases electrification, storage, and efficient process redesign. Still, SC remains negative (−0.288 ***) and DT is not present in this specification, indicating that capability and invention help, but deployment is the binding margin: unless procurement criteria, supplier qualification, and logistics KPIs are explicitly decarbonized, supply chains can continue optimizing incumbent technologies and slow the measured transition. This sequencing (invention → adoption → diffusion) is emphasized in mitigation assessments: benefits materialize when innovation is paired with demand-pull and standards that force deployment at scale.

With environmental degradation (ED) as the outcome, ET enters large and negative (−0.526 *** to −0.633 ***), confirming internal validity: when the G7’s energy mix moves toward renewables, electrification, and efficiency, observed degradation falls. SC is also negative and significant (−0.457 *** to −0.639 ***), consistent with greener logistics, reverse flows, and supplier collaboration directly lowering emissions and waste, independent of full energy-mix substitution [8]. These IV results strengthen the causal reading: even after purging endogeneity, transition, and supply chain greening reduce ED.

Across ED equations, TRD is consistently negative (−0.010 *** to −0.006 ***), sharper than in baseline estimates. In the G7, stringent regulation and high standards mean that openness tends to import cleaner capital and embed best practices, so technique/composition effects dominate scale, yielding lower ED. UB is small/insignificant in (4)–(5) and negative in (6) (−0.103) **, indicating that densification can reduce per-capita emissions via transit and building efficiencies once deployment (captured by strong ET and GI in col. 6) is in place; before that point, construction and consumption can mask benefits in the short run [64].

For ET, CC is positive in (1) but negative in (2)–(3); for ED, CC is positive in (4)–(5) but negative in (6). Given CC measures control of corruption, a negative coefficient in ED (as in col. 6) aligns with expectations—better governance reduces environmental harm—while positive coefficients can reflect timing/composition: governance improvements co-move with growth and regulatory tightening that initially reveal or accompany higher measured activity [63]. LnEG is positive in all six models (ET: 0.748 ***–1.492 ***; ED: 2.363 ***–3.433 ***), consistent with the scale–composition–technique decomposition: growth both funds transition (pro-ET) and raises throughput (pro-ED) unless policy accelerates composition/technique shifts.

The Lewbel IV–2SLS estimates are causally stronger and point to a clear sequence: (i) convert SCM from a process-efficiency tool into a deployment lever by hard-wiring decarbonization targets into supplier criteria, contracts and logistics KPIs; (ii) pair DT with clean-power procurement and strict efficiency standards to neutralize rebound; (iii) turn KM and GI into diffusion via demand-pull instruments (standards, carbon pricing, and green public procurement), so their pro-ET effects translate swiftly into ED reductions; and (iv) keep leveraging trade openness and compact, electrified urban design to lock in technique/composition gains. The IV estimates show that energy transition consistently reduces environmental degradation and that supply chain management likewise lowers environmental degradation—patterns that appear structural in the G7—whereas the short-run negative effects of supply chain management on energy transition and of digital transformation on energy transition seem transitional and amenable to policy intervention.

4.8. Examining Nonlinear Effect of Knowledge Management, Digital Transformation, and Green Innovations

Table 11. Nonlinear effect of knowledge management, digital transformation, and green innovations.

	(1)	(2)	(3)	(4)	(5)	(6)
	Energy Transition			Environmental Degradation
SC	−0.373 ***	−0.735 ***	−0.347 ***	−0.371 ***	−0.127	−0.373 ***
	(0.044)	(0.107)	(0.055)	(0.063)	(0.146)	(0.044)
DT	−0.683 ***			−0.201		−0.683 ***
	(0.152)			(0.224)		(0.152)
DT2	−0.027			−0.067 *		−0.027
	(0.027)			(0.039)		(0.027)
TRD	0.011 ***	0.003	0.002 *	−0.009 **	0.001	0.011 ***
	(0.003)	(0.002)	(0.001)	(0.004)	(0.002)	(0.003)
UB	0.159 ***	0.026	0.013	0.011	0.012	0.159 ***
	(0.049)	(0.048)	(0.044)	(0.064)	(0.050)	(0.049)
CC	0.169 ***	−0.481 ***	−0.483 ***	0.285 ***	0.476 ***	0.169 ***
	(0.043)	(0.080)	(0.071)	(0.046)	(0.108)	(0.043)
lnEG	1.356 ***	2.333 ***	0.952 ***	2.428 ***	1.333 ***	1.356 ***
	(0.231)	(0.363)	(0.223)	(0.322)	(0.489)	(0.231)
KM		0.386 ***			0.056
		(0.083)			(0.099)
KM2		−0.207 ***			0.294 ***
		(0.030)			(0.047)
GI			0.070 *
			(0.040)
GI2			−0.210 ***
			(0.026)
ET				−0.501 ***	−0.130
				(0.055)	(0.099)
Constant	−12.451 ***	−22.062 ***	−7.300 ***	−22.215 ***	−12.854 ***	−12.451 ***
	(2.359)	(3.729)	(2.296)	(3.287)	(4.848)	(2.359)
Time-fixed	Yes	Yes	Yes	Yes	Yes	Yes
Observations	161	161	161	161	161	161
R-squared	0.547	0.724	0.732	0.662	0.771	0.547

Note: * p < 0.10, ** p < 0.05, *** p < 0.01. Models (1)–(3) focus on the determinants of energy transition (ET), while Models (4)–(6) focus on environmental degradation (ED). Within each set, additional sustainability drivers are introduced stepwise: Model (1) includes the baseline variables (SC, DT, TRD, UB, CC, and EG); Model (2) adds knowledge management (KM); Model (3) further incorporates green innovation (GI); Model (4) begins with baseline variables plus ET; Model (5) adds KM; and Model (6) adds GI.

presents the nonlinear results. In Model 1, for G7 economies, SC is negative and highly significant for energy transition (ET) (−0.373 ***), and digital transformation (DT) also enters negatively (−0.683 ***), while the DT-squared term is statistically indistinguishable from zero. This pattern is consistent with a short-run “optimization of legacy chains” mechanism: greener logistics and procurement reduce waste within existing (often fossil-anchored) systems before the electricity and fuel mix measurably shifts, so ET can appear to slow even as processes get cleaner. The DT coefficient likely reflects rebound—digital infrastructure adds computation and connectivity loads that can offset efficiency gains unless coupled with clean power and stringent efficiency standards [65]. Trade openness (TRD) is pro-ET (0.011 ***), and urbanization (UB) and economic growth (lnEG) are also positive, echoing G7 advantages in dense, electrifiable urban systems and deeper fiscal capacity for clean investment [64].

When KM and KM² enter (see Model 2), KM is positive (0.386 ***) and KM² is negative (−0.207 ***), indicating diminishing marginal returns: building codified knowledge, absorptive capacity, and learning routines boosts ET up to a turning point, after which additional KM yields smaller gains [66]. In this specification, SC becomes even more negative (−0.735 ***), suggesting that unless procurement and supplier KPIs explicitly target decarbonization, better-run supply chains continue to optimize incumbent technologies rather than accelerate substitution. Control of corruption (CC) turns negative for ET (−0.481 ***), which can arise when governance upgrades initially tighten compliance and risk controls—raising short-run costs and slowing visible transition—before deployment benefits materialize; the long-run literature still finds that stronger institutions facilitate cleaner technology diffusion [63].

In Model 3, GI is positive (0.070 *) while GI² is negative (−0.210 ***), implying an inverted-U: early increases in green invention lift ET, but marginal gains taper as patenting saturates or as bottlenecks shift from invention to deployment and integration [67]. SC remains negative (−0.347 ***), reinforcing that the bottleneck in advanced economies is often deployment rather than invention. TRD remains small but positive (0.002 *), consistent with technique/composition channels—openness imports cleaner capital and standards in the G7 [62]. UB and CC are statistically weak here, likely because GI absorbs variance related to capability and policy quality.

With environmental degradation (ED) as the outcome, ET is strongly negative (−0.501 ***), confirming that deeper transition reduces environmental pressure—an internal validity check aligned with mitigation assessments for advanced economies [3,68]. SC is also negative (−0.371 ***), indicating that greener logistics, reverse flows, and supplier collaboration directly lower emissions and waste even before full energy substitution. DT’s linear term is small and not significant (−0.201), but DT² is negative (−0.067 *), suggesting threshold effects: once digitalization reaches higher levels and is paired with clean electricity and efficiency standards, the net impact on ED turns more clearly downward [69].

Here, ET’s coefficient attenuates (see Model 5) and becomes statistically weak (−0.130), while KM is small and insignificant (0.056), and KM² is positive (0.294 ***). A plausible reading is staging: investments in knowledge systems can initially raise measured activity (audits, reporting, data platforms, training), with environmental gains showing up only after those capabilities are translated into technology deployment and supplier switching. Meanwhile, CC is positive (0.476 ***), which is counter-intuitive if higher values mean stronger control; this often reflects timing and composition—governance improvements may coincide with growth and regulatory tightening that reveal previously unmeasured impacts before enforcement reduces them [63]. As expected, lnEG remains positive (1.333 ***), consistent with scale effects unless composition/technique shifts accelerate.

When the model pares back to core drivers (see Model 6), SC remains protective (−0.373 ***). DT is negative and significant (−0.683 ***), while DT² is small and not significant, implying an approximately linear reduction in ED when digitalization is not evaluated jointly with ET or KM/GI—consistent with the idea that digital tools improve monitoring, routing, and process control in mature G7 systems. TRD appears positive and significant (0.011 ***), suggesting that, absent explicit transition/capability controls, scale effects of openness can dominate in the short run; this contrasts with specifications where ET or strong policy variables are present and technique/composition effects show through [62]. UB is positive (0.159 ***), indicating that without targeted design (e.g., transit-oriented development, and building codes), densification can raise local pressures via construction and consumption [64].

Across specifications, SC consistently reduces ED and, in the short run, is associated with lower measured ET—pointing to a deployment gap that policy can close by embedding decarbonization targets in procurement, supplier qualification, and logistics KPIs. DT tends to hinder ET unless paired with clean power (rebound), but can reduce ED at higher digital maturity or in stripped-down models through better information and control [65]. TRD generally supports ET and has mixed effects on ED depending on whether composition/technique channels dominate scale; UB helps ET in dense, electrifiable G7 contexts but can raise ED without supportive urban policy. CC’s sign flips reflect timing: stronger institutions eventually support transition and environmental performance, even if early enforcement phases coincide with higher recorded activity.

The nonlinear terms point to thresholds and diminishing returns: KM and GI show concavity for ET (benefits taper without deployment), and DT shows threshold benefits for ED. Priorities are therefore as follows: (i) convert SC from process efficiency to transition deployment via green contracting and supplier KPIs; (ii) pair DT with clean-power procurement and efficiency standards to neutralize rebound; (iii) push KM and GI past invention toward diffusion using demand-pull (standards, carbon pricing, public procurement), so their ET gains translate into ED reductions; and (iv) harness TRD and UB with standards-aligned trade and compact, electrified urban design. These moves align with the innovation-deployment pathway emphasized in IPCC and IEA roadmaps.

5. Conclusions and Policy Recommendations

5.1. Conclusions

In this study, we have examined the impact of the supply chain on energy transition and environmental sustainability while also considering the nonlinear nature of knowledge management, digital transformation, and green innovations. In doing so, this study used panel data spanning from 2000 to 2022 for the G7 nations. The study also employed Lewbel IV-2SLS and FGLS estimators to examine this connection. The results showed that across G7 models, supply chain performance consistently links to lower environmental degradation but is negatively associated with energy transition. Digital transformation also enters negatively for energy transition—consistent with rebound from added compute loads—while its squared term is mostly insignificant; however, at higher digital maturity, digital transformation helps curb environmental degradation. Trade openness is pro-energy transition and sometimes raises environmental degradation when scale effects dominate, whereas urbanization tends to support energy transition. When added, knowledge management is positive with a negative quadratic term, and green innovation shows the same inverted-U. Control of corruption is mixed—often negative for energy transition. Using environmental degradation as the dependent variable, we found that energy transition strongly reduces environmental degradation, and the supply chain directly lowers impacts; economic growth generally increases environmental degradation.

5.2. Policy Recommendations

Supply Chain Efficiency as a Lever for Deployment: Supply chain efficiency is central to achieving Paris-aligned decarbonization (SDG-12 on sustainable production and SDG-13 on climate action), yet empirical evidence shows a friction: while efficiency consistently reduces environmental degradation, it is often associated with slower measured energy transition (ET) in the short run. This paradox arises because supply chains initially optimize legacy, carbon-intensive processes before shifting to low-carbon alternatives. Such “lock-ins” reflect lifecycle data silos that undermine supplier emissions traceability and legacy KPI misalignments that continue to prioritize cost and delivery speed over sustainability outcomes. To close this gap between G7 pledges and practice, procurement strategies must embed explicit decarbonization objectives—supplier alignment with science-based targets, time-bound Scope 3 milestones, and transparent disclosure of embodied emissions. Institutional mechanisms such as green public procurement quotas (e.g., minimum thresholds for low-carbon steel and cement by 2030) and contracts-for-difference to guarantee clean material demand would align supply chain efficiency with Paris goals. Similarly, tying vendor qualifications to renewable freight shares and extending transition-linked working capital to SMEs would accelerate retrofits, mitigating short-term slowdowns in ET progress.

Advancing knowledge management and green innovation: Knowledge management (KM) and green innovation (GI) map directly to SDG-9 (industry, innovation, and infrastructure) and SDG-17 (partnerships), yet they too face operational frictions. Our estimates show KM has a positive effect on environmental outcomes but diminishing returns at scale, while GI follows an inverted-U relationship: initial invention boosts transition, but benefits fade unless innovations are diffused and integrated across supply chains. These diminishing returns reflect governance frictions such as PCA-built indices that obscure heterogeneity across G7 states and weak monitoring systems that track patents rather than diffusion outcomes. To bridge this gap with Paris goals, policy design must rebalance R&D incentives toward early deployment, streamline permitting and grid interconnection (SDG-7), and strengthen demand-pull mechanisms (e.g., carbon pricing, performance standards, green procurement). Complementary measures such as extension services and adoption vouchers can break knowledge silos, translating technical playbooks into supplier-level retrofits. By directly addressing diffusion bottlenecks, KM and GI can move beyond invention plateaus to generate Paris-consistent deployment at scale.

Trade openness as a Catalyst for Clean Technologies: Trade openness contributes to SDG-17 and Paris alignment through the diffusion of clean technologies, yet scale effects often generate rebound environmental pressures. The friction lies in composition: trade without carbon standards risks amplifying fossil-intensive flows rather than accelerating clean-tech diffusion. To align trade with Paris pledges, policy must privilege composition and technique effects—through mutual recognition of carbon-intensity standards, CBAM-compatible rules, and “green lanes” for clean technology imports. Export credit and development finance should be directed toward cross-border clean supply chains, while embedding carbon-intensity clauses in FTAs and harmonizing MRV frameworks across G7 nations would enhance compliance portability. Addressing these frictions ensures that openness accelerates renewable deployment rather than entrenching fossil-based dependencies.

Urbanization as a Driver of Decoupling: Urbanization directly engages SDG-11 (sustainable cities) but carries dual outcomes: compact, electrifiable systems accelerate ET, whereas poorly managed growth deepens environmental degradation. Frictions arise when urban growth is tied to car-centric design and weak municipal finance, producing rebound effects in energy intensity. To close this gap, policies should enforce zero-emission zones, transit-oriented development, and whole-life carbon building codes, while municipal finance should be tied to retrofit rates and per-capita energy intensity. Investments in electrified transport, freight corridors, and district heating from industrial and data-center waste heat can turn urbanization into a lever for Paris-aligned decoupling. By embedding climate-sensitive governance into municipal growth, cities can become accelerators rather than obstacles in the G7’s pathway to net zero.

Governance and Finance for Sequenced Impacts: Governance reforms underpin SDG-16 (institutions) but face friction in practice: short-term compliance costs discourage firms, while fragmented metrics obscure comparability. These gaps weaken the credibility of Paris-aligned pathways. Bridging instruments such as transition-linked bonds, capital expenditure grants, and carbon price floors with revenue recycling can soften short-term burdens while maintaining credibility. G7-wide standardization of impact metrics (e.g., transition intensity and carbon intensity per unit output) and independent verification would counter data silos and KPI misalignments. Adaptive governance, with annual policy reviews recalibrating procurement quotas and digital efficiency rules, would sustain momentum toward the 1.5 °C trajectory.

Digitalization and energy transition: Digitalization engages SDG-9 and SDG-13 but introduces rebound frictions: increased computational demand and connectivity elevate short-term energy consumption, slowing observed ET despite long-term efficiency potential. This explains the gap between G7 digitalization pledges and transition outcomes. To mitigate rebound, digitalization must be paired with clean power through renewable PPAs for data centers, strict PUE benchmarks, and carbon-aware scheduling. Conditional tax credits and grants linked to renewable contracting and reduced intensity would realign incentives. A proposed G7 Digital Energy Label (measuring kWh per inference or CO₂e per transaction) would make efficiency transparent, steering digitalization in line with Paris-aligned decarbonization.

5.3. Managerial Implications

The findings suggest that supply chains must serve as a central lever for the energy transition, but the evidence highlights a dual dynamic: in the short run, SCM optimizes legacy, carbon-intensive chains and slows measured ET, even as it directly reduces environmental degradation through efficiency gains. To flip this short-run SC–ET relationship into a positive driver, procurement strategies should embed explicit decarbonization levers into supplier qualification, green public procurement quotas, and reverse-logistics KPIs. This requires linking vendor and logistics standards to zero-emission freight and renewable power sourcing, while enabling SMEs to access transition-linked financing for retrofits. Knowledge management and green innovation must similarly pivot from invention toward deployment, supported by demonstration projects, streamlined permitting, and practical implementation playbooks. Trade openness can accelerate diffusion when anchored to carbon-intensity standards, clean-technology “green lanes,” and cross-border skills mobility. Urban policy should decouple growth from emissions through low-carbon transport, whole-life building codes, and district energy systems, while digitalization must be tied to clean power procurement and strict efficiency thresholds. Finally, robust governance—via carbon price floors, green finance, and standardized G7 metrics—ensures that SCM, innovation, and trade operate as deployment-aligned levers for structural transition, rather than reinforcing incremental legacy optimization.

5.4. Limitations and Future Directions

This study has several limitations that warrant consideration. First, the composite transition measures (EET/IET), constructed from OWID indicators, interpolation, and PCA, may be sensitive to variable selection, scaling techniques, and missing data adjustments. Likewise, the proxies for patents, publications, and digital transformation primarily capture invention, infrastructure, and research activity rather than deployment quality, commercialization, or operational outcomes. This limitation may partly explain why GI sometimes appears to increase environmental degradation, since invention does not always translate into adoption or market penetration. Second, although advanced estimators such as Driscoll–Kraay, FGLS, and Lewbel IV were employed to address heteroskedasticity, cross-sectional dependence, and endogeneity, residual concerns remain. These include the possibility of weak instruments, model misspecification, and reverse causality—for example, higher ET could itself encourage greener SCM adoption rather than the reverse. Third, the use of annual, country-level data inevitably obscures regional, sectoral, and short-run dynamics, where rebound effects from digitalization, legacy supply chains, or urban growth may unfold in nonlinear ways. Finally, the analysis does not explicitly account for government decisions and regulatory interventions, which are central to shaping supply chain alignment, procurement standards, and climate policy certainty, and the finding on skilled labor mobility emerges as a novel but as yet underexplored dimension in prior empirical work.

Future research should therefore consider alternative index constructions, explicitly address commercialization pathways of innovations, apply designs that better handle reverse causality (e.g., natural experiments, difference-in-differences, or dynamic panel estimators), and explore causal evaluation of policy instruments. In addition, incorporating nonlinear and distribution-sensitive frameworks and drawing on micro-level procurement, emissions, and labor mobility data would enrich understanding of timing, heterogeneity, and equity in the energy transition, while also linking structural drivers more explicitly to governance, adoption quality, and implementation mechanisms.