Carbon Reduction Pledges and Renewable Energy Adoption in East Asia’s Early Corporate Energy Transition

Abstract

This paper examines the relationship between corporate carbon-reduction pledges and the subsequent adoption of renewable energy by pledging firms, and whether this relationship depends on the institutional conditions in which they operate. We propose a pressure-capacity model, highlighting two different institutional dimensions, (1) environmental policy stringency (institutional pressure) and (2) renewable energy infrastructure (institutional capacity), that may shape when firms’ symbolic pledges lead to observable change in their energy procurement behavior. We estimate random-effects logistic regression models with panel data on 552 publicly listed firms in South Korea, China, and Japan from 2002 to 2017. We find that the relationship between carbon-reduction pledges and renewable energy adoption is strengthened by both the stringency of environmental policy and the availability of renewable energy infrastructure. The marginal effects analysis indicates that the pledge effect is close to zero when institutional capacity is low. However, it increases to about 13 percentage points when policy stringency is high and 9 percentage points when renewable supply is high. The country-specific subsample analysis further uncovers that the conditional effect of institutional capacity is particularly pronounced among Japanese companies. The analysis of correlated random effects shows that these patterns remain robust even after controlling for between-firm confounding. Overall, our findings indicate that the extent to which voluntary corporate climate change commitments translate into actual green implementation depends on the regulatory and infrastructural environment in which firms operate.

1. Introduction

Corporate carbon-reduction pledges have become a key feature of global climate governance. Corporations are publicly announcing plans and ambitious targets to reduce greenhouse gas emissions, align with climate science, and ultimately achieve carbon neutrality. Some pledges are externally validated, and commitments aligned with the Science-Based Targets initiative (SBTi) are often considered examples of corporate climate leadership [1]. But the key question persists: under what circumstances do these public carbon-reduction commitments lead to significant, concrete changes in companies’ energy practices?

The importance of this question lies in the fact that corporate climate pledging has become the focus of two related debates in the literature on corporate environmental strategy and sustainability performance. A body of literature examines the credibility and efficacy of voluntary environmental sustainability initiatives of firms [2,3]. For example, do self-target-setting and public commitments play a key role as antecedents of change in firms’ sustainability behavior? Or are they primarily used to rebrand stakeholder perceptions without any significant change in operational decisions?

The second debate, more directly focused on pledge–action gaps, concerns the strategic symbolic management of environmental objectives. This perspective is often referred to as “greenwashing,” whereby companies may have incentives to control outward appearances, particularly when stakeholders are more likely to reward words than actions [4]. Recent research also shows that firms can boost the credibility of projects by adjusting strategic targets, such as updating baselines, extending timelines, reducing scopes, or reclassifying activities, even though decarbonization activities themselves are constrained [5]. A growing body of critical literature on greenwashing further documents the means through which firms maintain gaps between stated commitments and actual behavior [6].

Nonetheless, these problems do not mean that corporate decarbonization promises are always hollow. Rather, they bring attention to the importance of distinguishing between commitments that are genuine and result in consequential decisions and those that are mainly deployed as selective disclosure or as symbolic compliance [2,3]. Indeed, the empirical evidence on the pledge-action relationship is inconclusive so far [7]. Research shows that externally validated or market-recognized promises may denote strong intentions and encourage adherence, especially when verification increases reputational costs and the costs of backsliding [8]. Others are not convinced that verification alone is enough to encourage costly changes in practice, particularly when firms respond with relatively cheap symbolic measures or when there are considerable barriers to implementation [9].

These mixed findings are consistent with the idea that corporate sustainability is a field in which decoupling may occur, but disproportionately so, a point stressed in the institutional literature on the conditional nature of symbolic–substantive coupling [10]. A firm’s ability to maintain symbolic compliance depends on the nature of the activity domain and the institutional environment, which affect monitoring, feasibility, and incentives [2,11]. In this light, we ask when carbon pledges are more likely to become operational practice.

This study examines the conditions under which carbon-reduction pledges are linked to subsequent renewable-energy adoption. To answer this question, we will focus on a specific, quantifiable element of corporate climate action: the use of renewable energy by companies. We test the hypothesis that publicly committed firms are more likely to subsequently adopt renewable energy, and that the relationship between pledging and adopting renewable energy varies by national institutional settings. The use of renewable energy is one behavioral outcome, as it implies a shift in how companies obtain energy toward at least partial reliance on renewable energy sources rather than fossil fuels. Moreover, renewable energy can be implemented in various ways, such as through Renewable Energy Certificates (RECs) and Power Purchase Agreements (PPAs) [12,13]. We conceptualize renewable-energy adoption as an entry into procurement behavior, a signal that a firm has initiated some form of renewable-energy sourcing, rather than a uniform measure of deep operational transformation. This measure captures the extensive margin (entry into renewable sourcing) rather than the intensive margin (depth of adoption) and thus encompasses heterogeneous procurement pathways. This view is focused on operational practices rather than disclosure, as renewable sourcing is concrete and may require additional managerial work and procurement changes, and it considers the heterogeneity in how companies adopt it [13]. By focusing on the adoption of renewable energy, we make the pledge-action relationship a key element of decarbonization: how companies acquire energy to generate and deliver their services.

We suggest that the pledge-to-adoption relationship varies across national settings. Drawing on institutional theory and the literature on organizational decoupling [12], we develop a pressure-capacity framework that identifies two complementary institutional dimensions that shape when pledges couple with observable procurement behavior. Institutional environments affect the incentives of firms to keep their public promises and the viability of energy transitions. Thus, we focus on two national-level variables that reflect distinct but potentially reinforcing institutional conditions: environmental policy stringency and renewable-energy infrastructure. The rigidity of environmental policies implies regulatory pressure, which may increase accountability, increase scrutiny, and increase the cost of failing to follow through on publicly stated commitments [14]. With more stringent rules, pledges are more likely to translate into observable action because symbolic compliance is more expensive and the utility of backsliding is lower [15,16]. Renewable-energy infrastructure captures institutional capacity and market feasibility. Renewable generation, grid integration, and supporting infrastructure increase the number of viable renewable energy sourcing options and reduce the transaction and implementation costs for companies seeking to switch their energy sources [13,17]. Incentives can be strengthened by regulatory pressure, but only if supply-side conditions are not too restrictive. Meanwhile, good infrastructure can facilitate adoption, but businesses may still underinvest without accountability or credibility incentives. We therefore expect that carbon commitments are most successfully translated into renewable energy adoption where institutional pressure and capacity co-exist.

Our study is based on a panel dataset of 552 publicly listed firms in South Korea, China, and Japan, covering the period 2002 to 2017. These three East Asian economies are major industrial economies with substantial carbon footprints and with different configurations of regulatory traditions, state-market relations, and infrastructure development [18,19], producing informative variation in the conditions of pressure and capacity that frame corporate climate action. We focus on the 2002–2017 period for both theoretical and empirical reasons. This period was a period of the establishment of corporate climate pledging in East Asia, when stakeholder expectations and pledge-based governance were institutionalized [20] as the three countries diverged in their national environmental policies and renewable energy infrastructure. These cross-country variations, as well as within-country variations over time, provide useful context for determining when pledges are more likely to be converted into future adoption of renewable energy. The period also covers the early post-Paris window (2016–2017), providing a common global policy reference point, and is consistent across countries due to similar firm-level disclosures and data coverage.

Our results indicate that companies that commit to reducing carbon emissions are more likely to adopt renewable energy in the following years. Also, the link between pledging and adopting is stronger in countries with stricter environmental regulations and more developed renewable energy infrastructure, but weaker in countries with lax environmental regulations and poorly developed renewable energy infrastructure. On the whole, these results suggest that the key mechanisms through which public commitments are translated into action are institutional structures that increase accountability and expand feasibility, making symbolic compliance more costly and actual follow-through more feasible.

This study adds to the understanding of corporate environmental pledges and the energy transition in three ways. First, it explains why previous results on the pledge-action relationship have been inconsistent by determining institutional boundary conditions that systematically affect when pledges become apparent changes in renewable-energy sourcing [2,8]. Second, the research does not focus on the different levels of implementation across the different modes of adoption but instead treats renewable-energy adoption as a procurement signal, although the implementation levels across the modes are heterogeneous [13]. Third, it integrates regulatory pressure and infrastructural capacity into an integrated pressure-capacity framework, emphasizing the role of accountability and capacity in determining the effectiveness of pledges. Drawing on institutional theory, the decoupling literature, and comparative institutional perspectives on East Asian business systems [19], tighter policies can increase the stakes of credibility [14,16]. In general, our results suggest that voluntary climate commitments are more likely to affect emissions-related changes when institutional settings not only enhance accountability but also provide the necessary infrastructure.

We note at the outset two important scope conditions. First, our dependent variable captures entry into renewable-energy sourcing, which includes heterogeneous procurement pathways that vary in operational depth and additionality. Second, our study period (2002–2017) precedes several developments that have accelerated corporate climate action, including rapid cost declines in renewable technologies, the proliferation of net-zero commitments, and major policy announcements such as China’s 2020 carbon peak and neutrality goals. We describe this period as an early transition period, a temporal context that allows for analytically useful variation but also represents a boundary condition that we address in the discussion.

The rest of the paper is as follows: Section 2 introduces the theoretical framework and hypotheses. We next outline the data, measures, and empirical strategy, our findings and robustness checks, and finally conclude by discussing the implications of our research and policy on corporate decarbonization.

2. Theory and Hypotheses

2.1. Theoretical Framework: Institutional Pressure and Capacity

Corporate sustainability pledges, particularly carbon-reduction pledges, are a subject of debate in the institutional literature. The decoupling literature suggests that organizations often adopt policies and pledges without implementing them fully [12,21]. Yet institutional scholars have shown that the degree of decoupling depends on the institutional environment firms face [2]. Extending this line of work, we suggest that the link between carbon-reduction pledges and renewable-energy adoption is likely to be influenced by two complementary institutional dimensions: institutional pressure and institutional capacity.

Institutional pressure arises mainly from regulatory conditions and associated norms that elevate the costs of symbolic compliance relative to substantive implementation. For example, when environmental policies are stringent, discrepancies between pledged targets and actual behavior become more consequential under stricter enforcement and scrutiny, thereby increasing the incentive for firms to fulfill their commitments [14].

In our framework, we define institutional capacity as the infrastructural and market factors that lower the barriers to translating commitments into action. For example, when renewable energy infrastructure is well developed, firms face lower transaction costs and have more procurement options, which helps to implement renewable sourcing [13,17].

These two dimensions yield four stylized institutional configurations. Conditions for tight coupling between pledges and renewable-energy adoption are high pressure and high capacity. When pressure is high and capacity is low, firms are motivated yet constrained, which can result in symbolic compliance or reliance on limited procurement methods. When pressure is low but capacity is high, firms may adopt renewables for cost or efficiency reasons, but the link to pledging is weaker. When both are low, conditions are most permissive of decoupling.

Institutional configurations differ in national business systems. The varieties of capitalism literature highlights that different state-market relations, regulatory traditions, and coordination mechanisms create distinct institutional environments for corporate behavior [18,19,22,23]. China, Japan, and Korea are good examples of this diversity. Japan’s coordinated market has established regulatory norms and coordination mechanisms that shape both institutional pressure and institutional capacity. Korea’s state-led development model works through direct regulatory mandates and industrial policy that can expand implementation capacity. China’s state-orchestrated system has a mix of ambitious state-level targets with evolving regulatory enforcement. These differences create systematic variation in the configurations of pressures and capacities that firms encounter in deciding whether and how to act on their climate commitments.

2.2. Carbon-Reduction Commitments and Renewable Energy Adoption

Corporate carbon-reduction pledges can be symbolic, but they can also create a sense of accountability that makes inaction more costly over time. Public commitments raise stakeholders’ expectations and increase the risk of reputational damage, particularly because stakeholders can more easily assess whether pledges are backed by observable actions [4,5]. This external channel of accountability operates through stakeholder monitoring. Once a firm has made a public commitment to a target, the reputational penalty for visible inaction often outweighs the cost of making no such pledge in the first place [24,25,26].

Pledges may also serve as internal commitment mechanisms by supporting budgeting, monitoring, and cross-unit coordination around decarbonization goals, thereby increasing follow-through even when firms have competing operational priorities [3]. This internal mobilization channel does not require external monitoring to function; the pledge itself alters organizational priorities and decision-making [27]. In this sense, pledges may increase the likelihood that firms go beyond talk to more externally legible forms of action.

The adoption of renewable energy represents one salient and externally legible form of implementation [12]. Shifting part of a firm’s energy sourcing to renewables is a readily observable operational step through self-generation and/or contractual procurement. Thus, it can be cited as evidence of follow-through beyond narrative disclosure. Renewable sourcing is tangible and verifiable in a way that is harder to sustain through communication alone, even when adoption occurs through various pathways, including self-generation as well as contractual procurement such as Renewable Energy Certificates (RECs) and Power Purchase Agreements (PPAs) [13]. Although decoupling is still possible and the depth of implementation may vary [28], the logic above suggests that, on average, pledge-making firms should exhibit higher subsequent rates of renewable-energy adoption than non-pledging firms [3,12].

Hypothesis 1.

Carbon-reduction pledges are associated with a higher likelihood of subsequent adoption of renewable energy.

2.3. Environmental Policy Stringency as Institutional Pressure

Within our framework of institutional pressure and capacity, environmental policy stringency (EPS) constitutes a key component of institutional pressure. Whether carbon pledges are translated into action should depend on institutional conditions that affect the costs and benefits of implementation.

Environmental policy stringency is a particularly consequential condition as it changes the payoff structure of carbon-intensive operations and, by implication, the relative attractiveness of low-carbon alternatives. More stringent policies increase the perceived and real cost of continuing with carbon-intensive practices through stringent standards, enhanced disclosure and enforcement, and the introduction of policy instruments that increase the economic cost of emissions and expand incentives towards cleaner alternatives [29]. As the stringency of policies increases, the shadow price of carbon-intensive energy can increase at the firm level, and uncertainty about returns to cleaner energy decisions can decline as policy signals become clearer, making renewable-energy adoption more financially and strategically defensible [15,16]

Although policy stringency may also affect whether firms make pledges in the first place, the theoretical interest here is in the margin of implementation among pledging firms; that is, how policy stringency affects the relative payoffs of follow-through versus symbolic compliance once a firm has already made a public commitment. For pledging firms, tighter policy regimes raise the existing and anticipated costs of maintaining carbon-intensive practices and increase reputational and regulatory risks of failing to show credible progress. Policy stringency may also enhance expectations of future regulatory tightening, incentivizing earlier action to minimize expected compliance costs and to develop implementation capabilities in changing conditions [30]. In this respect, more stringent policy environments do not just help to provide background pressure, but they increase the credibility stakes of public commitments by increasing the level of regulatory scrutiny and making backsliding more consequential [14].

Macro-level evidence is generally consistent with the notion that policy stringency is associated with cleaner energy outcomes. Studies using measures of policy stringency developed by the Organization for Economic Cooperation and Development (OECD) report that more stringent environmental policy is associated with more renewable energy use, consistent with a mechanism of substitution in which policies that increase the cost of pollution make cleaner energy more attractive [16]. Applied to the pledge-to-adoption relationship, this logic suggests that greater regulatory pressure should increase the chances that pledging firms turn declared commitments into measurable changes in energy sourcing because symbolic compliance becomes more costly during tighter enforcement and substantive follow-through becomes relatively more attractive.

Hypothesis 2.

Environmental policy stringency strengthens the positive association between carbon-reduction pledges and subsequent renewable-energy adoption.

2.4. Renewable Energy Infrastructure as Institutional Capacity

Within our framework, renewable-energy infrastructure is a key component of institutional capacity with respect to implementation feasibility. In addition to regulatory pressure, such as policy stringency, infrastructural conditions should influence whether pledges can be translated into practice. Renewable-energy infrastructure is not just a question of generation capacity but includes the market infrastructure and contracting architecture that allow firms to acquire renewables at scale, grid integration, reliable procurement channels, and instruments like Renewable Energy Certificates (RECs) and Power Purchase Agreements (PPAs). In places where this infrastructure is in place, companies encounter fewer obstacles to implementing renewable-energy sourcing, as they can shift procurement through existing markets and contracting mechanisms instead of having to develop generation capacity in-house. On the other hand, in areas where the renewable energy infrastructure is underdeveloped, pledging firms are likely to face major implementation challenges despite regulatory pressure.

In theory, policy stringency has a major impact on expected payoffs and the cost of inaction, and renewable-energy infrastructure has a major impact on feasibility and the cost of implementation. These are, thus, different enabling mechanisms: one is the incentive-side calculus of pledging firms, and the other are the feasibility-side enabling conditions for implementation, although they can co-evolve over time as policy can assist in the construction of infrastructure. From an institutional theory perspective, this logic of feasibility is consistent with the argument that external pressures can be sufficient to stimulate commitment decisions but not to substantively implement when enabling conditions are absent [31]. Applied to carbon pledges, this implies that when infrastructural limitations make implementation technically or economically difficult, public commitments may not be translated into renewable procurement.

In this regard, we expect renewable-energy infrastructure to improve the pledge-to-adoption relationship by expanding feasible procurement pathways and reducing frictions to implementation. In places where renewable infrastructure is stronger, follow-through should be more likely (and occur sooner), as companies are able to more easily operationalize commitments through available procurement mechanisms. In weaker infrastructure, the same promises can be kept at the rhetorical stage for longer, not necessarily because companies face weaker incentives to do so, but because there are fewer viable channels for implementation.

Hypothesis 3.

Renewable-energy infrastructure strengthens the positive association between carbon-reduction pledges and subsequent renewable-energy adoption.

Finally, we note that pressure and capacity are conceptually distinct but not completely independent in practice: environmental policy can stimulate infrastructure development by creating demand for clean energy, and well-developed infrastructure can increase the feasibility of stringent regulation. Accordingly, for analytical clarity, we discuss their conditioning roles separately in H2 and H3 and examine their possible complementarity empirically using a three-way interaction analysis in Section 4.

3. Methods

3.1. Research Setting: Corporate Energy Transition in East Asia (2002–2017)

Corporate carbon-reduction pledges have become an increasingly prominent tool through which firms signal their commitment to climate action since the 2000s. In practice, these pledges vary widely in content and credibility. Some commitments are externally validated, most visibly through frameworks such as the Science-Based Targets initiative (SBTi). By contrast, many others remain self-declared statements embedded in broader stakeholder communication. Rather than restricting attention to a single certification regime, our study treats carbon-reduction pledges as a heterogeneous set of public commitments observed in corporate sustainability reporting during the study period, reflecting the real-world diversity of corporate climate pledging [8].

A central premise of this study is that translating pledges into operational change is not automatic. Whether pledges become practice depends on institutional conditions that shape (a) the incentives to follow through and (b) the feasibility of implementation. We therefore focus on two country-level institutional dimensions that map onto these distinct mechanisms: environmental policy stringency as institutional pressure and renewable-energy infrastructure as institutional capacity.

Environmental policy stringency captures the regulatory dimension of institutional pressure. In stricter policy environments, the expected costs of maintaining carbon-intensive practices are higher, and the credibility stakes of public commitments are more salient. Policies can raise compliance costs through multiple channels—market-based instruments that price emissions, non-market instruments that impose standards and limits, and policy supports that facilitate low-carbon technological alternatives [29]. We operationalize this regulatory dimension using the OECD Environmental Policy Stringency (EPS) Index, which captures cross-national and over-time variation in the stringency of environmental policies, aggregating policy instruments and support measures into a comparable metric [15]. Higher EPS values indicate more stringent policy environments and, correspondingly, stronger regulatory pressure on carbon-intensive activities.

Figure 1 (left panel) shows that policy stringency followed distinct trajectories across China, Japan, and Korea between 2002 and 2017. Japan exhibits relatively high policy stringency throughout much of the period, consistent with a long-established environmental policy framework. Korea shows a pronounced upward trend in policy stringency over time, indicating meaningful strengthening of the regulatory environment. China, while lower in absolute terms for much of the period, shows a clear upward trend, especially in the early 2010s, suggesting rapid tightening of policy stringency as environmental governance intensified. This divergence provides the cross-country and temporal variation needed to examine whether regulatory pressure conditions the pledge-to-adoption relationship among firms that have already made public commitments.

Renewable-energy infrastructure captures the infrastructural dimension of implementation feasibility. Even when firms face incentives to shift away from fossil fuels, adoption is constrained if renewable energy is not readily available through scalable procurement channels. This feasibility dimension includes not only renewable generation capacity, but also grid integration, market organization, and contracting mechanisms that allow firms to procure renewable power. In the context of corporate adoption, feasibility is especially salient because many firms transition through procurement choices—such as contracting and certificate-based pathways—rather than building generation capacity in-house. Accordingly, our infrastructural framing aligns with our dependent-variable interpretation of renewable adoption as entry into renewable-energy procurement rather than a uniform indicator of deep operational transformation.

Figure 1 (right panel) illustrates cross-national differences in renewable energy conditions using OECD data on renewables as a share of total primary energy supply. The three countries display sharply different levels and trajectories over the study period. China exhibits a high starting level and a pronounced decline through the 2000s, followed by a gradual rebound in the early 2010s, a pattern consistent with rapid growth in total energy supply during industrial expansion and the subsequent strengthening of renewables in the national energy mix. Japan remains at comparatively low levels with a modest upward trend over time. Korea is persistently lowest across the period, with only gradual increases, indicating that renewable sourcing remained more constrained relative to its regional peers. These differences matter for firms because they reflect the broader feasibility of sourcing renewables in the economy—whether through market availability, integration, or other procurement-relevant conditions, thereby shaping how readily pledges can be operationalized.

The core implication is that regulatory pressure and infrastructural feasibility constitute distinct, and potentially reinforcing, institutional conditions that shape whether pledges translate into action. Regulatory stringency raises the costs of symbolic compliance and strengthens incentives for follow-through. Renewable-energy infrastructure reduces implementation frictions and expands the set of feasible pathways for renewable procurement. The combination of these conditions produces analytically meaningful institutional configurations. In high-stringency and high-infrastructure environments, firms face both strong incentives and feasible routes to adoption—conditions most conducive to translating pledges into practice. In high-stringency but low-infrastructure environments, firms face pressure without equally viable pathways to respond through renewables. In low-stringency but high-infrastructure environments, feasibility exists, but incentives may be weaker, leaving more room for discretionary follow-through. Finally, in low-stringency and low-infrastructure environments, pledges are most likely to remain communicative rather than operational, given limited incentives and constrained feasibility.

Taken together, East Asia during 2002–2017 provides research setting in which corporate climate pledging was becoming institutionalized while national-level pressure and capacity diverged in ways that generate especially informative variation. Our empirical analyses leverage this divergence to test whether carbon-reduction pledges are associated with subsequent renewable-energy adoption and whether this association is systematically stronger under conditions of higher policy stringency and more developed renewable-energy infrastructure.

3.2. Data and Sample

We constructed our dataset by merging multiple data sources: the Refinitiv ESG database for corporate environmental practices and carbon-reduction pledges; Compustat Global for financial variables; the OECD Environmental Policy Stringency Index for regulatory pressure; and the OECD Renewable Energy Supply database for national renewable energy infrastructure.

Our sample consists of publicly listed firms from China, Japan, and Korea covered by Refinitiv ESG during the period 2002–2017. We focus on these three East Asian economies because they represent major industrial powers with significant carbon footprints and exhibit diverse national business systems that influence corporate sustainability disclosure [32]. Because ESG disclosure coverage and third-party data availability expand over time, our panel is unbalanced. The initial universe comprises 9502 firm-year observations across 1276 firms in the three countries. The largest source of attrition is the renewable energy supply variable (approximately 3647 observations lacking OECD coverage), followed by the forward lag of the dependent variable and missing financial controls. China’s representation declines disproportionately—from 26.5% of the initial universe to 10.8% of the final sample—primarily due to limited Refinitiv ESG coverage and the availability of OECD renewable energy data in the earlier years of the study period. After removing observations with missing values on key variables, our final sample comprises 552 firms with 5414 firm-year observations. Japan accounts for the majority of observations (75.8%), followed by Korea (13.3%) and China (10.9%). This distribution is consistent with cross-country differences in disclosure infrastructure and third-party coverage, which are relatively more established in Japan than in its neighbors [33]. We acknowledge that this sample composition may cause the pooled estimates to disproportionately reflect the Japanese experience; we address this concern through country-specific subsample analysis in Section 4.

3.3. Variables

Dependent variable. Our dependent variable is renewable energy adoption, measured using Refinitiv’s “renewable energy use” indicator. This binary variable indicates whether a firm reports in its sustainability reports or similar channels that it uses renewable energy in its operations, coded as 1 if it does and 0 otherwise. Because renewable sourcing can occur through heterogeneous channels, including on-site generation and procurement arrangements such as Renewable Energy Certificates (RECs) and Power Purchase Agreements (PPAs), we interpret this indicator as entry into renewable energy sourcing rather than as a uniform measure of deep operational transformation [13]. We use the one-year forward value (t + 1) to allow time for pledges to influence energy-related decisions and to mitigate reverse-causality concerns. Approximately 41% of firm-year observations report renewable energy use.

Independent variable. The carbon reduction pledge is measured using a binary indicator that captures whether a firm’s sustainability report contains a carbon-reduction commitment statement, coded as 1 if present and 0 otherwise. This measure captures publicly communicated pledges that are visible to external stakeholders and can therefore generate accountability and expectations. The content of pledges varies in credibility and specificity, ranging from self-declared commitments to externally validated targets (e.g., SBTi-aligned statements). Because our measure does not differentiate among these pledge types, the estimated effect should be interpreted as an average association across heterogeneous commitment types. Approximately 62% of firm-year observations report a carbon-reduction commitment statement.

Moderating variables. We examine two national-level moderation variables. Environmental policy stringency is measured using the OECD Environmental Policy Stringency (EPS) Index, a multidimensional indicator of regulatory pressure that aggregates policy instruments across market-based, non-market, and technology-support policies [15,29]. The EPS index ranges from 0 (least stringent) to 6 (most stringent). In our sample, the EPS index ranges from 0.77 to 3.52, with a mean of 2.55 (SD = 0.71). Renewable energy supply is measured as the contribution of renewables to total primary energy supply at the country level, obtained from the OECD database. This measure captures the availability and accessibility of clean energy that conditions the feasibility of corporate energy transitions [34,35,36]. The variable encompasses energy from hydro, wind, solar, geothermal, and biomass sources as a percentage of total primary energy supply. In our sample, renewable energy supply ranges from 0.4% to 12.9%, with a mean of 4.2% (SD = 1.9%). To facilitate interpretation of interaction effects, we mean-centered both moderating variables by subtracting each variable’s sample mean. This transformation allows the main effect of the carbon reduction pledge to be interpreted as the effect at average levels of environmental policy stringency and renewable energy supply.

Control variables. We include several firm-level control variables. Firm size is measured as the natural logarithm of total assets. Return on assets (ROA) is a key control for financial performance. R&D intensity is measured as the natural logarithm of R&D expenditure, capturing innovation capacity. Organizational slack is measured using the current ratio. Capital intensity is measured as the natural logarithm of property, plant, and equipment, reflecting the asset base required for energy-related investments. Industry carbon intensity is measured as the natural logarithm of industry-average (CO₂) emissions to account for sector-specific emission profiles. Finally, we include country indicators to capture time-invariant differences across China, Japan, and Korea.

3.4. Analytical Approach

We employ random effects logistic regression to analyze the determinants of renewable energy adoption. The model takes the following form [37]:

Log [P(Y_it+1 = 1)/P(Y_it+1 = 0)] = β₀ + β₁Pledge_it + β₂EPS _it + β₃REsupply_it + β₄(Pledge_it × EPS_it) + β₅(Pledge_it × REsupply_it) + γX_it + δCountry_i + u_it + ε_it

where u_i is a firm-specific random intercept capturing unobserved time-invariant heterogeneity, and X_it is the vector of control variables.

Random effects models are appropriate for our panel data structure because they account for unobserved, time-invariant firm-level heterogeneity while retaining both within-firm and between-firm variation. We prefer the random-effects specification over fixed effects for three reasons. First, our country-level moderators (EPS and renewable energy supply) require between-firm variation for identification; a fixed-effects specification would absorb this variation through firm-level dummies. Second, fixed-effects logistic regression drops all firms with no variation in the dependent variable over the study period, eliminating approximately 43% of observations and creating its own selection problem. Third, our theoretical interest lies in cross-national institutional conditions rather than purely within-firm dynamics. This approach also allows us to include country indicators alongside other covariates without relying on a country fixed-effects specification. The logistic specification is suitable given our binary dependent variable.

We estimate a baseline main-effects model to test Hypothesis 1 and then introduce interaction terms to test Hypotheses 2 and 3. Specifically, we interact the carbon reduction pledge with each mean-centered moderating variable—environmental policy stringency and renewable energy supply—to examine whether institutional pressure and infrastructural feasibility systematically condition the pledge–adoption relationship. Mean-centering facilitates interpretation by allowing the main effect of pledges to be interpreted at average levels of the moderators.

The random-effects model includes a firm-specific error component that captures unobserved heterogeneity. The likelihood ratio test strongly rejects the null hypothesis of no firm-level variance (χ² = 1948.40, p < 0.001), and the intraclass correlation (ρ = 0.79) indicates that approximately 79% of the total variance is attributable to between-firm differences, supporting the appropriateness of the random-effects specification. We acknowledge that the RE assumption that firm-specific effects are uncorrelated with the predictors may not hold in this context. To address this concern, we implement a Mundlak correlated random-effects approach [38] that adds firm-level means of all time-varying predictors, effectively decomposing between-firm and within-firm effects. This and other robustness checks, including fixed effects logistic regression, propensity score matching, and alternative temporal specifications, are reported in Section 4.

4. Results

4.1. Descriptive Statistics

Table 1. Sample descriptive statistics.

Variable	Mean	SD	Min	Max
Renewable energy use (t + 1)	0.413	0.492	0	1
Carbon reduction pledge	0.616	0.486	0	1
Environmental policy stringency (EPS)	2.549	0.713	0.771	3.521
Renewable energy supply	0.042	0.019	0.004	0.129
R&D intensity (ln)	7.568	4.146	0	16.637
Organizational slack	0.555	1.283	0	53.658
Firm size (ln total assets)	13.617	1.892	3.224	19.525
ROA	0.040	0.080	−0.597	4.076
Industry carbon intensity (ln)	−0.064	1.833	−7.749	4.261

Notes: N = 5414 firm-year observations from 552 firms. The sample comprises observations from Japan (75.8%), Korea (13.3%), and China (10.9%). Continuous variables are winsorized at the 1st and 99th percentiles. RE adoption is measured at t + 1. EPS is the OECD Environmental Policy Stringency index. Renewable energy supply is the share of renewable energy in total electricity generation.

reports descriptive statistics for the variables used in the analysis. Renewable energy adoption in year (t + 1) is observed in 41.3% of firm-year observations, while 61.6% of observations indicate the presence of a carbon-reduction pledge disclosed in sustainability reporting. The national institutional variables exhibit meaningful dispersion over the study period: the OECD Environmental Policy Stringency (EPS) index ranges from 0.77 to 3.52 with a mean of 2.55 (SD = 0.71), and renewable energy supply ranges from 0.4% to 12.9% of total primary energy supply with a mean of 4.2% (SD = 1.9%). These ranges are consistent with the divergent institutional trajectories across China, Japan, and Korea during the period studied and motivate our focus on institutional boundary conditions in the pledge–adoption relationship.

Table 2. Correlation matrix.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
(1) Renewable energy use (t + 1)	1.00
(2) Carbon reduction pledge	0.31 *	1.00
(3) Environmental policy stringency	0.21 *	−0.09 *	1.00
(4) Renewable energy supply	−0.03 *	−0.29 *	−0.18 *	1.00
(5) R&D intensity	0.28 *	0.40 *	0.07 *	−0.29 *	1.00
(6) Organizational slack	−0.12 *	−0.16 *	0.02	0.09 *	−0.01	1.00
(7) Firm size	0.28 *	0.32 *	0.34 *	−0.71 *	0.37 *	−0.21 *	1.00
(8) ROA	−0.09 *	−0.12 *	0.01	0.07 *	−0.06 *	0.16 *	−0.13 *	1.00
(9) Industry carbon intensity	0.06 *	0.14 *	−0.08 *	−0.09 *	0.03	−0.22 *	0.42 *	−0.12 *	1.00

Notes: N = 5414 firm-year observations. * p < 0.05. The correlation between firm size and RE supply (r = −0.71) reflects sample composition: larger Japanese firms dominate the sample, and Japan has relatively low renewable energy supply during the study period.

presents bivariate correlations. A carbon-reduction pledge is positively associated with subsequent renewable energy adoption (r = 0.31, p < 0.05), consistent with the baseline expectation that pledged firms are more likely to adopt renewable energy. Both R&D intensity and firm size are also positively correlated with renewable energy adoption (r = 0.28 for each, p < 0.05), suggesting that innovation capacity and organizational scale covary with renewable sourcing decisions. In contrast, renewable energy supply is weakly and negatively correlated with adoption (r = −0.03, p < 0.05), a pattern consistent with the cross-country structure of the sample in which national supply conditions are not simply aligned with firm-level uptake in bivariate terms. Importantly, the moderators are not collinear with the pledge measure in a manner that would mechanically generate interaction findings: pledge is modestly negatively correlated with policy stringency (r = −0.09, p < 0.05) and more strongly negatively correlated with renewable energy supply (r = −0.29, p < 0.05), indicating that pledge-making is not concentrated only in the most favorable institutional environments.

We also assessed multicollinearity using variance inflation factors (VIF). VIF values for the focal predictors and control variables remain below 2.0.

4.2. Hypothesis Testing: Random Effects Logistic Regression Results

Table 3. Random effects logistic regression results: renewable energy adoption (t + 1).

	Model 1	Model 2	Model 3	Model 4	Model 5
Focal variables
Carbon reduction pledge		0.878 ***	0.764 ***	0.905 ***	0.791 ***
		(0.151)	(0.158)	(0.153)	(0.161)
EPS (centered)	1.041 ***	0.994 ***	0.740 ***	0.993 ***	0.747 ***
	(0.187)	(0.189)	(0.207)	(0.189)	(0.207)
RE supply (centered)	87.200 ***	93.396 ***	92.953 ***	85.116 ***	84.977 ***
	(12.427)	(12.587)	(12.570)	(13.221)	(13.222)
Interactions
Pledge × EPS			0.592 **		0.617 **
			(0.197)		(0.199)
Pledge × RE supply				17.886 †	19.649 *
				(9.498)	(9.394)
Controls
R&D intensity (ln)	0.213 ***	0.170 ***	0.169 ***	0.169 ***	0.169 ***
	(0.028)	(0.029)	(0.029)	(0.029)	(0.029)
Organizational slack	−0.224	−0.330 *	−0.342 *	−0.329 *	−0.343 *
	(0.143)	(0.144)	(0.144)	(0.144)	(0.144)
Firm size (ln)	0.540 ***	0.389 **	0.382 **	0.400 **	0.395 **
	(0.129)	(0.130)	(0.130)	(0.130)	(0.130)
ROA	−3.386 **	−3.698 **	−3.770 **	−3.707 **	−3.786 **
	(1.309)	(1.317)	(1.318)	(1.317)	(1.318)
Industry CO2 (ln)	−0.079	−0.118	−0.114	−0.107	−0.103
	(0.082)	(0.083)	(0.083)	(0.083)	(0.083)
Country dummies	Yes	Yes	Yes	Yes	Yes
Model fit
Log likelihood	−1965.2	−1944.3	−1939.4	−1942.4	−1935.7
Wald χ2	543.1	568.6	578.3	573.3	584.5
σu	3.57	3.55	3.55	3.55	3.55
ρ	0.79	0.79	0.79	0.79	0.79
Observations	5414	5414	5414	5414	5414
Firms	552	552	552	552	552

Notes: Standard errors in parentheses. † p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001. LR test of ρ = 0: χ²(1) = 2012.63, p < 0.001 (Model 5). Moderators are mean-centered; the pledge coefficient in Model 5 represents the effect on the sample mean of both moderators.

presents random-effects logistic regression estimates for the prediction of renewable energy adoption in year (t + 1). Across model specifications, the results are consistent with the idea that carbon pledges are more likely to translate into observable changes in the energy sourcing of firms when national institutional conditions increase both accountability and feasibility.

Main effect of carbon-reduction pledges (H1). Hypothesis 1 predicted a positive relationship between carbon reduction pledges and later renewable energy adoption. In the full specification with mean-centered moderators (Model 5), the coefficient of carbon reduction pledge is positive and statistically significant (b = 0.791, p < 0.001). Because the moderators are mean-centered, this estimate is the pledge effect at average levels of environmental policy stringency and renewable energy supply; substantively, Model 5 is a re-parameterization of the full interaction model (Model 4) that moves the reference point to average institutional conditions without changing fitted values. To put it another way, Models 4 and 5 are exactly the same in terms of model fit; the only difference is the zero point of the moderators. In Model 4 (uncentered), the pledge coefficient is the effect when both EPS and renewable energy supply are equal to zero, which is outside the range of observed data and is not substantively interpretable.

The pledge coefficient in Model 5 (mean-centered) is the effect at sample-mean values of both moderators—a meaningful baseline. The interaction terms and the moderator main effects are the same for both parameterizations. The odds ratio (exp (0.791) = 2.21) suggests that, on average across all institutional conditions, pledge-making firms have a roughly 2.2 times higher probability of adopting renewable energy in the following year than non-pledging firms. Consistent with our conceptualization of the dependent variable, we interpret this pattern as evidence that public pledges are associated with entry into renewable-energy sourcing/procurement, rather than a uniform indicator of deep operational transformation. This finding is in line with recent evidence that corporate climate commitments are linked to subsequent environmental action [3,8], but our study differs in scope (broader pledge definition) and outcome (renewable procurement rather than emissions).

Conditional role of environmental policy stringency (H2). Hypothesis 2 predicted that environmental policy stringency would increase the pledge-adoption relationship. The interaction between pledge and policy stringency is positive and statistically significant in Model 5 (b = 0.617, p < 0.01). This result indicates that the relationship between pledging and later renewable adoption increases with the strictness of regulatory environments. In more stringent policy situations, pledges seem to have more follow-through implications, consistent with the argument that regulatory pressure increases the costs of symbolic compliance and increases the relative payoff to implementation. This finding is consistent with macro-level evidence that environmental policy stringency facilitates renewable energy diffusion [16] while identifying the micro-level mechanism: policy stringency conditions the firm-level relationship between stated commitments and observable behavior.

Conditional role of renewable energy supply (H3). Hypothesis 3 predicted that renewable energy supply would enhance the pledge-adoption relationship by reducing constraints of feasibility. Supporting this expectation, the interaction between pledge and renewable energy supply is positive and statistically significant in Model 5 (b = 19.649, p < 0.05). Given the scaling of the supply measure (a country-level share of total primary energy supply), the interaction coefficient is best interpreted in terms of marginal effects and predicted probabilities and not in terms of raw log-odds units. In line with this,

Table 4. Marginal effects of carbon reduction pledges on RE adoption probability.

Panel A: Across Environmental Policy Stringency
EPS Level	EPS Value	dy/dx	SE	p	95% CI
Low (−1 SD)	1.84	0.014	0.018	0.452	[−0.022, 0.050]
Below mean (−0.5 SD)	2.19	0.044	0.014	0.002	[0.016, 0.072]
Mean	2.55	0.075	0.013	<0.001	[0.049, 0.100]
Above mean (+0.5 SD)	2.90	0.105	0.016	<0.001	[0.073, 0.138]
High (+1 SD)	3.26	0.131	0.023	<0.001	[0.086, 0.176]
Panel B: Across Renewable Energy Supply
RE supply level	RE supply (%)	dy/dx	SE	p	95% CI
Low (−1 SD)	2.31	0.010	0.024	0.679	[−0.037, 0.057]
Below mean (−0.5 SD)	3.25	0.042	0.017	0.012	[0.009, 0.075]
Mean	4.20	0.070	0.013	<0.001	[0.045, 0.096]
Above mean (+0.5 SD)	5.15	0.082	0.015	<0.001	[0.053, 0.112]
High (+1 SD)	6.09	0.090	0.020	<0.001	[0.050, 0.129]

Notes: Marginal effects (dy/dx) computed from Model 5 at sample means of all other variables. EPS and RE supply values correspond to the sample mean ± 0.5 and ±1 standard deviation.

shows that the marginal effect of pledging (i.e., the change in predicted probability of renewable adoption associated with having a pledge) increases in a systematic way as renewable supply increases.

This pattern indicates that pledge-making firms are more capable of converting commitments into renewable sourcing when national supply conditions offer more accessible pathways to procurement. This result links to the burgeoning literature on corporate renewable energy procurement [13], identifying institutional boundary conditions under which firms enter renewable sourcing.

Magnitude of interaction effect (Table 4; Figure 2). To make the patterns of interaction more transparent, Table 4 reports marginal effects from Model 5. The marginal effect of pledging on the predicted probability of renewable adoption is close to zero and statistically indistinguishable from zero under low institutional support but grows rapidly as institutional conditions increase. For example, as the environmental policy stringency increases from low (EPS = 1.5) to high (EPS = 3.5), the estimated marginal effect increases from 0.014 to 0.131. A similar pattern is seen for renewable energy supply: the pledge-related increase in predicted probability is small at low supply (e.g., 1%) but becomes substantially larger at higher supply levels. Figure 2 visualizes these patterns by showing that the pledge-adoption relationship is strongest when institutional pressure and institutional capacity are both present—conditions making follow-through both more consequential and more feasible.

Several firm-level controls behave in expected ways. R&D intensity is positively associated with renewable adoption (β = 0.169, p < 0.001), consistent with the view that innovative capacity facilitates entry into new energy sourcing practices. Firm size is positive and significant (β = 0.395, p < 0.01), suggesting that larger firms are more likely to adopt renewable energy. ROA is negative and significant (β = −3.786, p < 0.01), indicating that more profitable firms in this period were less likely to shift toward renewable sourcing, potentially reflecting lower urgency to alter established energy practices. Industry carbon intensity is weakly negative (β = −0.103, p > 0.10), though not statistically significant in the full specification.

4.3. Robustness Checks

To assess the robustness of our primary results, we performed a series of robustness tests and present the findings in

Table 5. Robustness checks: renewable energy adoption (Logit models).

	(Model 6) Model 5 RE	(Model 7) FE Logit	(Model 8) f2.DV	(Model 9) L1.IV	(Model 10) Year FE	(Model 11) Melogit Cluster	(Model 12) GEE
Pledge	0.791 ***	0.289	0.556 **	0.656 ***	1.242 ***	0.788 **	0.377 **
	(0.161)	(0.193)	(0.162)	(0.166)	(0.177)	(0.266)	(0.124)
Pledge × EPS	0.617 **	0.481 *	0.636 **	0.498 *	0.210	0.618 *	0.382 ***
	(0.199)	(0.223)	(0.208)	(0.212)	(0.268)	(0.280)	(0.111)
Pledge × RE supply	19.649 *	13.259	23.119 *	31.013 **	12.781	19.686	9.568
	(9.394)	(11.174)	(9.877)	(10.231)	(15.025)	(12.830)	(6.553)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Country dummies	Yes	—	Yes	Yes	Yes	Yes	Yes
Year dummies	No	No	No	No	Yes	No	No
N	5414	3076	5408	4971	5414	5414	5414
Firms	552	314	548	547	552	552	552

Notes: Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001. Column 1: baseline random effects logit. Column 2: fixed-effects logit (drops non-switchers; country dummies absorbed). Column 3: dependent variable at t + 2. Column 4: pledge lagged to t − 1. Column 5: year fixed effects added. Column 6: melogit with cluster-robust SE at firm-level. Column 7: GEE population-averaged model with exchangeable correlation and robust SE. Controls include R&D intensity, organizational slack, firm size, ROA, and industry CO₂ intensity.

(Models 6–12) and in Appendices A–E. In these specifications, we are interested in the consistency of the key patterns of interaction, i.e., the strengthening of the pledge-adoption relationship in the presence of stronger institutional conditions, in terms of sign and statistical significance.

Model 7 re-estimates the entire specification with a conditional fixed-effects logistic regression, which is based on within-firm variation and thus eliminates firms whose change in the dependent variable is zero (as in FE logit). The carbon-reduction pledge and environmental policy stringency interaction is positive and statistically significant (b = 0.481, p < 0.05) in this more demanding specification, which is consistent with H2 and the overall findings of the random-effects model (Table 3, Model 5). The relationship between pledge and renewable energy supply is positive but not statistically significant (b = 13.259, n.s.), indicating that the capacity-related interaction effect in H3 is directionally consistent but not robust under within-firm identification, probably because the sample size is smaller and the country-level variation among switcher firms is less. In this FE specification, the main effect of pledging is positive but not statistically significant compared to zero (b = 0.289, n.s.), which is consistent with the fact that FE logit identifies effects of firms that switch pledge status and adoption status over time and, therefore, gives a conservative test compared to the baseline RE specification.

Model 8 moves the dependent variable by one more year (f2. renewable adoption) to determine whether the pattern of the baseline is time-sensitive in terms of the timing of outcomes. The main effect of the pledge is positive and significant (b = 0.556, p < 0.01), and the interaction terms are positive and significant—policy stringency (b = 0.636, p < 0.01) and renewable supply (b = 23.119, p < 0.05). This trend suggests that the pledge-adoption association and its institutional contingencies do not have a one-year horizon and can be adjusted with a somewhat longer adjustment window.

Model 9 employs a lagged measure of pledging (L1. pledge) to predict adoption next year to make sure that the primary findings are not due to contemporaneous measurement or short-run simultaneity. The lagged pledge effect is positive and significant (b = 0.656, p < 0.001). The renewable energy supply interaction is positive and significant (b = 31.013, p < 0.01), whereas the interaction between the pledge × policy stringency is positive and significant (b = 0.498, p < 0.05). These findings, combined with the previous ones, support the conclusion that pledges are more predictive of future adoption in the case of greater institutional capacity, as measured by renewable supply, and that regulatory stringency still enhances the pledge-adoption relationship in this lag structure.

Model 10 includes year fixed-effects in the baseline random-effects logit to capture common temporal shocks that might jointly affect pledge-making and renewable adoption (e.g., diffusion of renewable procurement practices in the region). The pledge main effect is still positive and significant (b = 1.242, p < 0.001) in this specification, whereas the two interaction terms are no longer significant and become imprecisely estimated (pledge × policy: b = 0.210, n.s.; pledge × supply: b = 12.781, n.s.). Having 48 country-year values (3 countries × 16 years), 16 year dummies capture a large portion of the time variation that the country-level moderators take advantage of. We describe the interaction effect results as indicative of institutional contingency, and not the estimated effects that are invariant to the time trends modeling.

Inference checks are also offered in Models 11 and 12. Model 11 re-estimates the entire specification with melogit with cluster-robust firm-level standard. The pledge main effect (b = 0.788, p < 0.01) and the EPS interaction (b = 0.618, p < 0.05) are significant, whereas the renewable energy supply interaction is attenuated to non-significance (b = 19.686, p > 0.10), which is due to the broader standard errors under clustering. Model 12 is a GEE population-averaged model with exchangeable correlation and robust standard errors. The main effect of the pledge (b = 0.377, p < 0.01) and the interaction between the pledge and the EPS (b = 0.382, p < 0.001) are significant, but the interaction between the pledge and the renewable supply is not significant (b = 9.568, n.s.). Such differences are not surprising since the GEE estimator is designed to estimate population-averaged and not subject-specific effects and can act differently when the key predictors change mainly at the country-year level.

To address certain methodological issues, we performed several other analyses in the following manner. Capital intensity was not included in the primary specification because it was highly multicollinear with firm size (r = 0.94). We evaluated variance inflation factors (VIF) to confirm this decision. Capital intensity and firm size have high VIF values (11.74 and 11.22, respectively), as is anticipated since both of them reflect scale-related factors of the asset base. In comparison, the VIFs of the focal predictors, which include carbon reduction pledge, environmental policy stringency, and renewable energy supply, are less than 2.0. Since high collinearity makes the estimates of both variables inaccurate and may influence the stability of other estimates in the model, we kept firm size as the theoretically more basic and commonly used control of organizational scale. To make sure that this omission does not influence our substantive findings, we re-estimated the entire model with capital intensity (Appendix A). The interaction between pledge and policy stringency is positive and statistically significant (b = 0.587, p = 0.003), and the interaction between pledge and renewable energy supply is also significant (b = 21.11, p = 0.026), indicating that the focal coefficients are virtually the same.

Second, we included the number of country-year ESG funds (natural log) as an indicator of investor pressure (Appendix B). Under this control, the main effect of the pledge (b = 0.809, p < 0.001) and the interaction effect of the pledge and EPS (b = 0.524, p = 0.014) are robust. The interaction effect of renewable energy supply is reduced to p = 0.112, which can be explained by the common country-year variation between the prevalence of ESG funds and renewable energy infrastructure.

Third, we used propensity score matching [39] to deal with selection into pledging (Appendix C). We nearest-neighbor matched pledging and non-pledging firms on lagged firm characteristics and country-year peer pledging rates (k = 3, caliper = 0.05). The post-matching balance improved from 52.7% to 4.2% mean bias. The pledge main effect strengthens (b = 0.969, p < 0.001), and the renewable energy supply interaction effect is preserved (b = 28.302, p = 0.021) on the matched sample (N = 4087; 492 firms). The interaction between pledge × EPS is reduced to non-significance, which is probably because the country-level variation in the matched sample is smaller.

Fourth, we estimated a Mundlak correlated random-effects model to account for the possibility that the RE assumption may not hold (Appendix D). The between-firm and within-firm effects are decomposed by adding firm-level means of all time-varying predictors. The firm-level mean pledge status is significant and large (b = 4.197, p < 0.001), indicating that between-firm variation in sustainability orientation plays a significant role in the observed association. The within-firm pledge effect is reduced to marginal significance (b = 0.362, p = 0.055). Importantly, both interaction terms are significant in this specification—pledge × EPS (b = 0.509, p = 0.016) and pledge × renewable energy supply (b = 20.671, p = 0.029), which means that the institutional interaction effect patterns are not driven by between-firm confounding.

Overall, the robustness tests show that (i) the pledge–adoption association is always positive in alternative timing specifications (Models 8 and 9), with the propensity score matching (Appendix C), and the Mundlak decomposition (Appendix D); (ii) the policy–stringency interaction effect is supported in within-firm identification and longer-horizon timing (Models 7 and 8) and significant in the Mundlak CRE model, but it becomes statistically insignificant when year fixed effects are added (Model 10); and (iii) the renewable-supply interaction effect is stable across timing checks (Models 8 and 9) and the Mundlak CRE model, though it weakens under the FE estimator (Model 7), the cluster-robust and GEE specifications (Appendix E), and becomes statistically indistinguishable from zero once year fixed effects are introduced (Model 10). These findings combined indicate that our overall inferences are usually robust, and that the robustness of the interaction effects is sensitive to the absorption of common time shocks and to other inference procedures—a weakness of the small number of country-year values that can be determined.

4.4. Country-Specific Subsamples

Given that Japan accounts for 75.8% of the sample, we estimated separate models for each country to examine whether the pooled results mask meaningful cross-national heterogeneity (

Table 6. Result of country-specific subsample analysis.

	(1) Japan	(2) Korea	(3) China	(4) Non-Japan
Carbon reduction pledge	0.378 †	−2.406	10.507 **	2.063 ***
	(0.202)	(2.137)	(3.804)	(0.452)
Pledge × EPS	−0.091	−1.399	1.240	−0.706
	(0.261)	(1.889)	(2.473)	(0.655)
Pledge × RE supply	105.039 ***	−159.757	−135.616 *	−13.915
	(27.831)	(163.225)	(64.817)	(23.610)
EPS (centered)	0.861 **	2.459	−7.115 **	0.216
	(0.278)	(1.728)	(2.486)	(0.648)
RE supply (centered)	75.174 ***	−115.006	−92.762 †	−8.614
	(14.684)	(153.843)	(48.919)	(22.775)
R&D intensity (ln)	0.189 ***	0.239 †	−0.039	0.163 *
	(0.032)	(0.132)	(0.096)	(0.065)
Organizational slack	−0.252	−4.361 *	0.263	−1.188 *
	(0.158)	(2.136)	(0.764)	(0.504)
Firm size (ln)	0.374 **	0.484	1.128 *	0.562 *
	(0.142)	(0.668)	(0.487)	(0.283)
ROA	−3.530 *	−6.285	−1.461	−3.505
	(1.464)	(8.204)	(6.019)	(3.071)
Industry CO2 (ln)	−0.101	0.282	−0.280	0.019
	(0.090)	(0.464)	(0.273)	(0.193)
Country dummies	—	—	—	Yes
σu	3.45	6.19	2.52	3.67
N	4054	775	585	1360
Firms	396	82	74	156

Notes: Standard errors in parentheses. † p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001. All models use random-effects logistic regression. Korea and China estimates should be interpreted with caution given the limited sample sizes. The China full-interaction model did not converge in preliminary runs; results shown are from a constrained specification.

).

In the Japan subsample (N = 4054; 396 firms), the pledge main effect is marginally significant (β = 0.378, p = 0.062); the pledge × EPS interaction is not significant, but the renewable energy supply moderating effect is strikingly strong (β = 105.039, p < 0.001). This pattern is consistent with Japan’s institutional configuration during the study period: a well-established regulatory tradition in which additional regulatory variation yields less marginal information, and a relatively constrained renewable energy infrastructure, in which the supply dimension is the binding constraint on pledge follow-through.

In the non-Japan pooled subsample (N = 1360; 156 firms), the pledge main effect is strong (β = 2.063, p < 0.001) but neither interaction effect term is significant, which is consistent with the limited moderator variation when only two countries remain. The Korea (N = 775) and China (N = 585) subsamples yield imprecise estimates due to limited sample sizes; we report them as reflecting insufficient statistical power rather than the absence of effects.

4.5. Exploratory Three-Way Interaction

To explore whether institutional pressure and capacity operate synergistically, we estimated a three-way interaction model (pledge × EPS × renewable energy supply) and computed marginal effects across four institutional configurations (

Table 7. Marginal effects of pledging across institutional configurations (exploratory three-way interaction).

Panel A: Three-Way Interaction Model
	Coefficient	SE	p
Pledge	0.797 ***	0.161	<0.001
Pledge × EPS	0.609 **	0.200	0.002
Pledge × RE supply	19.872 *	9.437	0.035
Pledge × EPS × RE supply	−16.528	10.077	0.100
Panel B: Marginal Effects (dy/dx) across 2 × 2 Configurations
	Low capacity (−1 SD)	High capacity (+1 SD)	ΔCapacity
Low pressure (−1 SD)	−0.017 (p = 0.507)	0.075 * (p = 0.018)	+0.092
High pressure (+1 SD)	0.108 *** (p < 0.001)	0.123 *** (p < 0.001)	+0.015
ΔPressure	+0.125	+0.048

Note. Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001. Based on the three-way interaction model estimated via random-effects logistic regression (N = 5414; 552 firms). High/Low, defined as sample mean ± 1 SD of each moderator. The three-way interaction term (β = −16.528, p = 0.100) is marginally significant. The pattern shows that the incremental contribution of institutional capacity is larger under low-pressure conditions (+0.092) than under high-pressure conditions (+0.015), suggesting that regulatory pressure partially compensates for infrastructure constraints.

). The three-way interaction term is marginally significant (β = −16.528, p = 0.100). The 2 × 2 marginal effects show that the pledge effect is near zero when both pressure and capacity are low (dy/dx = −0.017, n.s.) and largest when both are high (dy/dx = 0.123, p < 0.001). The incremental contribution of capacity is larger under low-pressure conditions (+0.092) than under high-pressure conditions (+0.015), suggesting that regulatory pressure partially compensates for infrastructure constraints. We present this as an exploratory analysis given the marginal significance of the three-way term and the limited country-year variation.

5. Discussion

5.1. Summary of Key Findings

This study investigated whether corporate carbon reduction pledges are related to the next steps for adopting renewable energy sources and whether this relationship is stronger in institutional conditions that increase the salience of the follow-through or the feasibility of implementation. Using panel data on firms located in Korea, Japan, and China and estimating random-effects logistic models, we find that pledge-making is positively related to renewable energy adoption in the following year. Importantly, this main effect is estimated in the full specification with mean-centered moderators, so the coefficient on the pledge indicator represents the pledge effect when environmental policy stringency (EPS) and renewable energy supply are at their sample average levels. Substantively, the size of the effect suggests that, on average, at mean institutional conditions, pledge-making firms exhibit higher odds of adopting renewable energy in the subsequent year than otherwise-similar non-pledging firms. In line with how our dependent variable is operationalized, we interpret this pattern as indicating an increased likelihood of entry into renewable energy sourcing or procurement, rather than as direct evidence of a deep operational transformation across the firm’s entire footprint.

Beyond the average pledge effect, the results also suggest that the pledge-adoption association is conditional on institutional context, in ways that are consistent with the conceptual distinction between conditions of pressure and feasibility. First, we note that the interaction between pledge-making and EPS is positive and statistically distinguishable from zero in the mean-centered full model. This pattern is consistent with the possibility that, as policy environments get more stringent, pledges may have stronger practical implications and hence are more tightly coupled with observable renewable energy adoption. Put differently, increased EPS seems to be related to a steeper pledge-adoption slope, implying that regulatory stringency can serve as a contextual condition under which the costs of purely symbolic alignment increase and the expected benefits of demonstrable follow-through increase. At the same time, we take this interpretation to be suggestive rather than definitive, because EPS is a measure of national-level policy stringency and cannot be used to fully define the specific channels through which firms experience and respond to policy pressure.

Second, we find that renewable energy supply, measured as a country-level share and entered into the model as a moderator, also provides a positive conditionality to the pledge-adoption relationship. The interaction term shows that pledge-making firms seem to be more likely to turn commitments into adoption when national supply conditions are conducive to more accessible pathways for procurement and sourcing. Given the scale of the supply measure, the interpretation of the interaction is best conveyed through marginal effects and predicted probabilities rather than raw log-odds. In this spirit, the pattern that is summarized in the marginal effects table and visualized in Figure 2 shows that the pledge-related increase in predicted probability becomes larger as renewable energy supply increases. Taken together, these results are consistent with a feasibility account: Where constraints on supply are less binding, pledge-making more readily maps onto adoption, whereas under conditions of low supply, the mapping from pledge to adoption seems to be more limited.

An interesting consequence of the combined patterns of interaction effect is that pledge-making is not uniformly found to be associated with adoption across institutional environments: rather, the pledge-adoption association is found to be strongest when institutional conditions simultaneously increase the consequences of failure to implement (through stronger EPS) as well as the viability of implementation (through higher renewable energy supply). This joint pattern is, to some extent, in line with an institutional decoupling perspective. Specifically, our findings are in line with the notion that, in contexts where institutional support is weaker—either due to lower policy stringency or greater constraints on supply-side capacity—pledges may be more likely to remain weakly coupled with short-term adoption outcomes. Conversely, as institutional pressure and capacity strengthen, pledge-making is more closely linked to adoption, suggesting a tighter coupling between symbolic pledges and observable sourcing/procurement behaviors. The exploratory three-way interaction analysis (Table 7) provides further texture: the pledge effect is near zero when both pressure and capacity are low, and largest when both are high. The incremental contribution of capacity is greater under low-pressure conditions than under high-pressure conditions, implying regulatory pressure may be partially offsetting infrastructure constraints.

The country-specific subsample analysis (Table 6) indicates informative cross-national heterogeneity. In Japan, the interaction effect of renewable energy supply is strikingly strong, whereas the EPS interaction effect is not significant, consistent with the institutional configuration during the study period, which has a well-established regulatory tradition in which infrastructure is the binding constraint. In the non-Japan subsample, the pledge’s main effect is strong, but neither interaction term is significant, reflecting little moderator variation across only two countries.

The Mundlak correlated random-effects analysis (Appendix D) gives an important decomposition. The between-firm pledge mean is large and highly significant, confirming that sustainability-oriented firms pledge and adopt more. The within-firm pledge effect is reduced to marginal significance. Critically, both interaction terms are significant in this specification, suggesting that the institutional interaction effect is not attributable to between-firm confounding.

We emphasize, however, that our dependent variable measures adoption events rather than the depth, scope, or quality of implementation, and so the evidence is most directly relevant to whether and when pledges correlate with observable adoption, rather than to the more general question of whether firms fully operationalize decarbonization throughout their operations.

5.2. Theoretical Implications

The empirical patterns documented above are broadly consistent with a contextualized view of pledge follow-through: pledges are associated with subsequent renewable energy adoption on average, and the strength of this association varies systemically with national-level institutional conditions. Interpreting these findings, it is instructive to distinguish between those to which the analysis can speak with relative clarity and those which it cannot adjudicate decisively. At the very least, the findings indicate that pledges are not just rhetorical artifacts that are orthogonal to subsequent observable behavior. Even when holding a set of firm characteristics fixed and taking advantage of variation within the firm through panel structure, pledge-making is linked to an increased probability of subsequent adoption, and this is linked to a significant extent to the institutional environment. This conditionality is important because it suggests that the pledge-adoption relationship is not a single constant “effect” but is better conceived as reflecting a pledge process that is affected by the institutional costs of inaction and the practical feasibility of action.

At the same time, the findings suggest an interpretation that is cautious, taking into account the theory of decoupling. In institutional theory, the concept of decoupling is frequently used to refer to situations where symbolic commitments and substantive practices become uncoupled, whether because symbolic compliance is less costly than substantive change, because evaluation is noisy, or because there are constraints that preclude implementing certain things. Our evidence does not directly capture the depth or completeness of decarbonization efforts, and it also does not track internal operational changes beyond the indicator on adoption used here. Nonetheless, the patterns are consistent with a qualified decoupling argument in the following limited sense: where institutional support is weaker—either due to lower EPS or more constrained renewable energy supply—pledges seem less tightly coupled with near-term adoption outcomes. On the other hand, the opposite holds, leading to an association between pledges and observable adoption levels being more pronounced in places where EPS is higher, and the supply of renewable energy is more copious, hinting at tighter coupling under conditions where consequence and feasibility barriers are simultaneously elevated. This reading takes decoupling not to be an all-or-nothing condition but one of contextual tendency in the degree of the relationship between symbolic commitments and a particular observable behavior.

The two moderators we examine, EPS and renewable energy supply, also help us understand a working mechanism that does not require firm claims made about “intent” and “sincerity”. A more stringent policy environment plausibly has the effect of increasing the salience of climate-related commitments by increasing the expected costs of non-implementation, and strengthening monitoring and enforcement infrastructures, imperfectly. In parallel, increased renewable energy supply is a plausible solution to procurement constraints by creating more routes to source, contract, or certify. In combination, these moderators point to the fact that pledge follow-through is influenced by both pressure and feasibility conditions. Under stronger EPS, pledge-making may be more likely to be treated as consequential, while under a higher renewable energy supply, pledge-making may be more likely to translate into feasible procurement or sourcing actions. This framing is willfully humble: it is not claiming that there is a definitive causal pathway that operates uniformly across all firms, but rather that there is a shifting institutional environment in which pledges are more likely to coincide with adoption. We note that these mechanisms are inferred from observed interaction patterns rather than directly tested through mediation analysis; the data do not allow us to observe the specific organizational processes through which institutional conditions translate into procurement decisions.

Our findings contribute to several ongoing debates in the literature. First, the positive average association between pledges and adoption is consistent with recent evidence that corporate climate commitments carry behavioral consequences [3,8] but our study extends this by identifying systematic institutional contingencies. Second, the conditional nature of our findings speaks to the greenwashing literature, which documents the means through which firms maintain gaps between commitments and behavior [2,6]. Our results suggest that greenwashing—understood as persistent decoupling—is not uniformly distributed but is more likely to occur under institutional configurations that lower the costs of symbolic compliance and constrain the feasibility of substantive action. This is consistent with previous finding that firms strategically adjust targets to maintain credibility [5], and with the argument that the degree of decoupling depends on the coherence of institutional pressures firms face [21]. Third, the infrastructure interaction effect connects to the growing literature on corporate renewable energy procurement [13], showing that supply-side conditions are not merely background context but active boundary conditions that shape when corporate commitments translate into observable sourcing decisions.

From an international and comparative standpoint, the results also highlight the importance of researching pledges in terms of explicit consideration of cross-national institutional heterogeneity. Even in a relatively circumscribed regional setting, the magnitude and character of the pledge-adoption linkage also seem to vary according to macro-level characteristics. This suggests that it is important not to draw conclusions from single-country evidence without specifying the institutional conditions under which pledge signals become more (or less) behaviorally consequential. A related implication is methodological: when researchers are interpreting pledge effects, mean-centering of moderators helps to make clear that the estimated main effect is at average institutional conditions, and that departures from those conditions are captured by interaction terms. In this sense, the model specification is in good conceptual agreement with the argument suggested above—that pledges are not considered in isolation but rather in an institutional context.

Finally, the patterns have practical implications in the way that pledges are interpreted by stakeholders. The results suggest that the meaning of a pledge should not be taken to be equally informative across contexts. In environments that have greater policy stringency and a greater renewable energy supply, the pledge signal seems more closely aligned with the subsequent adoption. In contrast, in settings where there is less policy pressure or more supply constraints, pledges may be less predictive of near-term adoption, not necessarily because strategic misrepresentation of intent is by firms more likely, but because implementation pathways are more constrained or less rewarded. This interpretation also reinforces the value of putting pledge credibility judgments into the context of the institutional conditions that affect follow-through and not judging follow-through and pledge credibility on pledge presence alone.

5.3. Limitations and Future Research

Several limitations qualify the inferences that can be drawn from this study and also point to productive directions for future work.

First, our key variables are measured at a level of aggregation that does not capture important underlying heterogeneity. The dependent variable captures entry into renewable-energy sourcing but does not distinguish the intensity, coverage, or quality of adoption, nor does it differentiate among procurement pathways that vary in additionality or integrity (e.g., on-site generation versus certificate purchases). The independent variable is a binary pledge indicator that does not differentiate among commitment types varying in ambition, specificity, time horizon, or external validation. Self-declared statements and externally validated SBTi-aligned targets are likely to differ in the strength of the accountability channel and the specificity of internal mobilization. The estimated effects are thus averages across heterogeneous pledge types and adoption pathways, and the evidence speaks most directly to whether pledges are associated with the probability of adoption events rather than to the depth of operational transformation.

Second, the findings of interaction effects are sensitive to specification choices and estimator assumptions. With year fixed-effects (Model 10), the estimates of the interaction terms are imprecisely estimated, plausibly because the dummies for years absorb common shocks that co-move with the country-level moderators, leaving little independent variation to identify the estimates. Fixed-effects logit (Model 7) removes firms for which there is no within-panel variation in the dependent variable, reducing the estimation sample considerably and reducing the magnitude of the main pledge effect—either reflecting the elimination of between-firm differences or the selective composition of the switcher subsample. Alternative approaches to inference like cluster-robust standard errors and GEE models imply positive pledge coefficients but weakened interaction terms, especially for renewable energy supply (Appendix E). These differences result from the fact that the different estimators are targeting different quantities, are based on different assumptions, and may behave differently in a situation where key predictors vary predominantly at the country-year level. Rather than taking any single specification as definitive, we are concerned with the directionality and broad pattern across approaches. The Mundlak decomposition (Appendix D) provides some comfort as both interaction terms are still significant after controlling for between-firm confounding, and the propensity score matching analysis (Appendix C) is helpful in addressing observable selection into pledging, but neither is able to completely eliminate unobserved confounding.

Third, the institutional moderators are measured at the national level and cannot fully reflect within-country heterogeneity in the experience of policy pressure and feasibility constraints by individual firms. EPS is a composite indicator that does not encode firm-specific exposure and enforcement intensity or sectoral structure of regulations. Renewable energy supply is an aggregate national situation rather than firm-level access to specific procurement arrangements, contract markets, or grid structures. These features of the measurements have led to the conclusion that the interaction effect results should be seen not in terms of specific micro-level mechanisms, but in terms of the broad conditional context. We point out that these mechanisms are derived from observed patterns of interaction effects rather than being tested directly through mediation analysis.

Fourth, our sample is limited to publicly listed firms covered by Refinitiv ESG, which tend to be larger and internationally more visible than the rest of the population of firms in these economies. Moreover, the composition of the sample is the result of data availability and not balanced cross-national representation: China’s share declines from 26.5% of the initial universe to 10.8% of the final sample because of the limited Refinitiv coverage and the availability of data from the Organisation for Economic Cooperation and Development, which means that the pooled estimates reflect the Japanese experience disproportionately. Country-specific subsample analysis (Table 6) addresses this concern to some extent but cannot replace a more balanced design.

Fifth, our study period (2002–2017) represents the formative stage of corporate climate pledging in East Asia, which is conducive to analytical benefits in terms of variation in institutional conditions. However, this period comes before several developments that have transformed the corporate climate landscape: rapid reductions in the cost of renewable energy, the spread of net-zero commitments, and the adoption of SBTi, major policy announcements including China’s 2020 carbon peak and neutrality targets, and the increase in mandatory climate disclosure requirements. The dynamics of pledge adoption documented here may be quite different in the post-2017 environment.

Finally, the research is observational and cannot totally take care of endogeneity. Firms may self-select into pledging depending on unobserved strategic orientation, anticipated future investments, or pressure from stakeholders that also influences adoption. Our use of controls, lag structures, propensity score matching, and the Mundlak decomposition lessens but does not entirely eliminate this concern. The evidence is most defensibly argued as documenting a systematic association, conditioned by institutional context, between pledges and subsequent renewable energy adoption, rather than a singular causal effect.

Several directions for future research arise from these limitations. Researchers may be able to use the post-2017 growth in SBTi adoption and mandatory disclosure to investigate whether externally verified pledges generate stronger coupling with action than do self-declared commitments. Firm-level measures of procurement channel access and regional grid characteristics could help unpack the aggregate infrastructure moderator into more specific feasibility mechanisms. Comparative designs with a more extensive set of countries would aid in establishing the generalizability of the pressure-capacity framework. Longitudinal designs that follow the intensive margin of renewable energy use, rather than merely entry, would address the question of the depth of implementation that our binary measure cannot capture.