When Digital Trade Meets Regulatory Distance: Implications for Carbon Intensity in International Trade

Abstract

Digital services trade is often viewed as a pathway to lower carbon intensity by reducing reliance on carbon-intensive physical trade. However, its environmental benefits may depend critically on the regulatory environments governing cross-border digital interactions. Integrating institutional distance theory with environmental economics, this study examines how regulatory divergence in digital services trade shapes the carbon intensity of international trade. Using bilateral trade data and country-level measures of digital services trade regulations, renewable energy capacity, and environmental policy rigor, we analyze the effects of digital regulatory gaps on carbon emissions embodied in exports. The results show that greater regulatory divergence significantly increases both total carbon emissions and export carbon intensity. The analysis further reveals that the scale effect associated with increased trade volume dominates the technique effect, such that the potential environmental benefits of digitalization are frequently offset by structural inefficiencies and compliance costs induced by regulatory fragmentation. Moreover, exporters’ renewable energy capability amplifies—rather than mitigates—the carbon-intensity-increasing effect of digital regulatory gaps, indicating that institutional misalignment imposes higher environmental opportunity costs on countries with greater low-carbon potential. By contrast, environmental policy rigor in importing countries does not significantly attenuate these effects. Overall, the findings highlight regulatory alignment as a critical condition for realizing the environmental benefits of digital trade.

1. Introduction

Rapid advances in digitalization are reshaping the global trade landscape, making digital trade an increasingly important mode of cross-border exchange [1,2]. Digital trade refers to cross-border economic activities enabled by the internet and information and communication technologies (ICTs), including internationally traded digitally deliverable services, cross-border data flows, and digitally embedded value that can be exchanged without the physical movement of goods [3,4]. Unlike conventional trade, which is organized around the cross-border shipment of tangible products, digital trade is anchored in data mobility and virtual service delivery [5]. This shift is altering how firms configure production networks, coordinate global value chains, and access international markets [6].

Digital trade is often portrayed as a pathway to more sustainable growth [7,8]. Because digitally delivered services (e.g., cloud computing, streaming, and cross-border data transfer) can substitute for some physical logistics, they may reduce transportation, warehousing, and inventory-related emissions in international exchange [9]. By lowering material intensity and improving resource efficiency, digital services trade could reduce carbon intensity across global value chains [10]. For this reason, digital trade has attracted increasing attention as a potential mechanism for reconciling globalization with environmental sustainability [11]. Yet such benefits are not assured. We conceptualize this tension as the Digital–Green Paradox: the decarbonization potential of digital trade (the technique effect) can be offset—or even reversed—when two structural barriers dominate. First, a digital rebound effect can amplify electricity demand through exponentially growing data processing and storage requirements [12]. Second, institutional friction arising from regulatory misalignment can generate carbon-intensive redundancies by forcing firms to reconfigure and duplicate digital operations across jurisdictions [13]. As a result, expanding digital trade does not necessarily translate into lower carbon intensity and may, under some conditions, increase it.

A key reason is that digital trade is governed by heterogeneous institutional and regulatory environments that shape cross-border data flows, market access, and competitive conditions [14,15]. Countries vary widely in rules on data localization, cross-border data transfers, platform governance, cybersecurity, and digital taxation [16,17]. While some jurisdictions pursue interoperable regimes that facilitate cross-border provision of digital services, others impose restrictive requirements that fragment digital infrastructures and constrain service delivery. Although recent free trade agreements (FTAs) and digital trade agreements aim to harmonize rules—often emphasizing tariff-free digital transactions, data mobility, and fair competition—substantial cross-national divergence persists, creating regulatory frictions that firms must navigate in practice.

From the perspective of institutional distance, such divergence constitutes formal institutional distance that increases coordination and compliance costs in cross-border activities [18,19]. Importantly, these costs are not purely financial. From an environmental economics perspective, regulatory frictions can distort resource allocation and generate unintended externalities by increasing energy use through duplicated data processing, parallel compliance systems, and suboptimal deployment of digital resources [20,21]. Consequently, regulatory gaps in digital services trade may raise the carbon intensity of bilateral trade relationships, offsetting—or even reversing—the environmental advantages commonly attributed to digitalization [22,23].

Despite growing interest in digital trade and sustainability, evidence on how regulatory divergence in digital services trade shapes environmental outcomes remains limited. Prior research has largely examined the environmental effects of trade openness or digitalization in isolation [24,25], paying less attention to how cross-national regulatory misalignment conditions carbon intensity. Moreover, we know relatively little about when and where these effects are amplified by country-specific contexts—particularly national energy structures and environmental policy regimes. We therefore ask: Why does digital trade fail to decarbonize in some corridors? We argue that a central explanation lies in the regulatory gap between trading partners.

To address this question, we examine how regulatory divergence in digital services trade between countries affects the carbon intensity of major trading relationships. Beyond estimating the direct impact of regulatory gaps, we assess whether digital services trade itself mitigates or exacerbates carbon intensity under conditions of regulatory divergence. Recognizing that trade-related emissions are jointly shaped by countries’ energy systems and policy environments, we further theorize and test moderation at both ends of the dyad. Specifically, we examine whether exporters’ renewable energy capability and importers’ environmental policy rigor condition the relationship between digital services trade regulatory gaps and carbon intensity. Accordingly, we address four research questions:

RQ1. Does regulatory divergence in digital services trade between countries increase the carbon intensity of major trading partners?

RQ2. Does digital services trade reduce carbon intensity, and how does regulatory divergence condition this relationship?

RQ3. Does the exporting country’s renewable energy capability strengthen or weaken the effect of digital services trade regulatory gaps on carbon intensity?

RQ4. Does the importing country’s environmental policy rigor strengthen or weaken the effect of digital services trade regulatory gaps on carbon intensity?

This study advances the literature in three ways. First, we shift attention from trade volume to institutional friction. Whereas prior work has focused on whether expanding digital trade reduces emissions via the technique effect or increases them via the scale effect, it has paid less attention to how the governance of digital trade shapes environmental performance. We provide evidence that regulatory heterogeneity (the digital services trade restrictiveness gap) operates as an independent determinant of carbon intensity, generating environmental inefficiencies distinct from those driven by trade expansion.

Second, we theorize a specific environmental mechanism—carbon-intensive redundancy—through which regulatory gaps translate into higher emissions. Extending gravity-model logic to the digital–environmental domain, we argue that misalignment does not merely impose financial compliance costs; it can compel firms to replicate infrastructure and processes (e.g., jurisdiction-specific data storage, duplicated verification, and parallel reporting and monitoring systems). This framing bridges institutional distance and environmental economics by conceptualizing regulatory friction as a source of deadweight environmental loss.

Third, we identify a green paradox in digital trade: the adverse impact of regulatory gaps may be stronger in renewable-energy-intensive exporting economies. This challenges the assumption that green energy capacity automatically buffers environmental shocks. Instead, we propose a systemic complementarity argument: the environmental returns to renewable energy investments depend on interoperability in the digital trade regime—without which digital redundancies erode potential decarbonization gains.

2. Theoretical Framework and Hypotheses Development

2.1. Integrating Institutional Distance and Environmental Mechanisms

Our framework integrates institutional distance theory with the scale–technique–composition decomposition of trade’s environmental effects [26]. Institutional distance theory argues that cross-national regulatory differences increase firms’ liability of foreignness and transaction costs by complicating coordination, compliance, and governance across borders [27]. In digital trade, we conceptualize institutional distance as a regulatory gap that constrains the seamless cross-border mobility of data and digital services [13].

From an environmental economics perspective, digital trade is expected to reduce carbon intensity primarily through the technique effect—by improving efficiency and enabling dematerialization of economic activities [28]. We argue, however, that regulatory gaps can obstruct this pathway. When misalignment is substantial, firms must adapt operations to jurisdiction-specific requirements, which can involve duplicating data infrastructures, maintaining parallel systems, and performing redundant processing to ensure compliance with divergent rules [17]. This structural friction weakens the technique effect and can intensify a digital rebound mechanism [12]: any efficiency gains from digitalization are offset by additional energy demand induced by compliance-driven redundancy under incompatible regulatory regimes.

2.2. Digital Services Trade Regulatory Gap and Carbon Intensity

Building on our integrated framework, we examine how regulatory divergence in digital services trade shapes carbon intensity. Institutional theory argues that cross-national differences in formal rules, regulations, and governance arrangements create institutional distance that constrains firm behavior and performance in cross-border activities [29]. When institutional environments diverge, firms must adjust strategies and operating routines to heterogeneous requirements rooted in differences in legal systems, regulatory frameworks, and governance structures [30,31]. These adjustments raise coordination and compliance costs, increase uncertainty, and erode operational efficiency [32,33]. Such frictions are particularly salient when cross-border exchange depends on regulatory compatibility and system-level interoperability—conditions that are central to digitally enabled trade.

In digital services trade, divergence in rules governing data governance, cross-border data flows, market access, and platform conduct constitutes a salient form of formal institutional distance [34,35]. When trading partners operate under different regulatory regimes, firms may need to comply with multiple—and sometimes incompatible—requirements relating to data localization, cybersecurity, and digital market conduct [3,15]. This complexity constrains firms’ ability to standardize, integrate, and scale digital operations across markets, thereby weakening the efficiency advantages often associated with digitalization [36].

A critical distinction is between regulatory stringency and regulatory heterogeneity. Stringency reflects the absolute restrictiveness of a given regime (i.e., the “height” of regulatory barriers), whereas heterogeneity captures the incompatibility between regimes, regardless of whether each regime is strict or permissive [17]. Two countries may both maintain stringent privacy protections, for example, yet still impose substantial adaptation burdens if their legal definitions, technical standards, or compliance procedures differ. In such cases, firms must incur interface and adaptation costs—such as reformatting data, maintaining jurisdiction-specific servers, or managing conflicting compliance protocols—even when the overall level of restrictiveness is comparable. Accordingly, regulatory heterogeneity can generate structural inefficiencies independent of regulatory stringency. In our empirical specification, we isolate this incompatibility effect (the DSTRI gap) while accounting for the overall level of restrictiveness across the dyad (e.g., the DSTRI sum).

We argue that the inefficiency created by regulatory gaps is not merely financial; it can translate into carbon-intensive redundancy through three channels. First, physical redundancy arises when localization or residency requirements compel firms to duplicate infrastructure across jurisdictions. When trading partners impose divergent data-storage rules, firms may need to maintain multiple data centers (or jurisdiction-specific cloud instances) rather than relying on centralized, energy-efficient architectures [17]. Such duplication increases electricity demand for server operation and cooling.

Second, operational friction arises when interoperability is limited by incompatible technical and procedural standards. Regulatory misalignment can require additional processing and verification—such as repeated encryption, protocol translation, auditing, and compliance checks—to ensure that cross-border digital services meet jurisdiction-specific requirements [13]. These extra computational steps increase energy use per unit of digitally delivered service.

Third, resource diversion occurs when compliance burdens crowd out decarbonization investments. High regulatory fragmentation can force firms to allocate managerial attention and financial resources toward compliance and monitoring rather than adopting low-carbon technologies, improving energy efficiency, or optimizing supply-chain emissions [12]. Together, these mechanisms can convert ostensibly “virtual” exchange into a more energy- and emissions-intensive activity, thereby increasing carbon intensity.

Beyond these direct channels, regulatory gaps can also prevent firms from fully capturing scale economies and efficiency gains that typically arise in integrated digital markets [18,37]. By limiting interoperability and locking firms into compliance-driven rigidity, regulatory divergence reduces the extent to which digitalization can substitute for carbon-intensive physical activities and deliver technique-effect improvements [38]. Consequently, the carbon-reducing potential of digital services trade may be weakened—or even reversed—when regulatory divergence is substantial.

This relationship may also be non-linear. Institutional theory suggests that firms can absorb modest cross-national differences through incremental adjustments, implying roughly linear adaptation costs. However, when divergence becomes large—such as when a restrictive data-control regime interacts with a free-flow regime—adaptation may require parallel systems and separate governance infrastructures, leading to disproportionately higher costs and redundancy [17,27]. While our primary objective is to establish the baseline relationship, this logic implies that the carbon impacts of regulatory gaps could intensify as institutional distance widens. Taken together, institutional theory and environmental economics suggest that regulatory divergence in digital services trade constitutes a structural constraint that increases operational inefficiency and environmental burden in cross-border digital exchange. Accordingly, we propose the following:

Hypothesis 1. 

Greater digital services trade regulatory divergence (DSTRI gap) between trading partners is positively associated with bilateral export carbon intensity.

2.3. The Dual Nature of Digital Trade: Technique Effect vs. Rebound Effect

Institutional theory suggests that lower regulatory barriers can facilitate cross-border digital services trade. However, the net environmental consequences of expanding digital trade are theoretically ambiguous because two opposing mechanisms may dominate: a technique effect and a digital rebound effect.

From a technique-effect perspective, digital services trade can contribute to decarbonization by substituting some physical activities with data flows and by improving the coordination and efficiency of global value chains [39]. Digitally delivered services may reduce emissions associated with transport, inventory holding, and other logistics-intensive processes, while also enabling more efficient allocation of resources through faster information processing and coordination. Under this mechanism, increases in digital services trade volume (X_ijt) should be associated with lower export carbon intensity.

Hypothesis 2a. 

(Technique effect dominance). Higher digital services trade volume is associated with lower export carbon intensity.

By contrast, environmental scholarship emphasizes the digital rebound effect, often linked to Jevons’ paradox [40]. Efficiency gains from digitalization can lower the implicit cost of data usage and digital service provision, stimulating greater overall demand for data processing, storage, and transmission. The resulting growth in data traffic increases reliance on energy-intensive infrastructures—such as hyperscale data centers, network equipment, and cooling systems—which can offset, neutralize, or even exceed initial efficiency gains. If the rebound mechanism dominates, expanded digital services trade may have no decarbonizing effect or may be associated with higher export carbon intensity.

Hypothesis 2b. 

(Rebound effect dominance). Higher digital services trade volume is associated with higher (or unchanged) export carbon intensity.

2.4. The Moderating Role of Renewable Energy: A Digital-Dependency Perspective

Renewable energy capability is often assumed to buffer environmental inefficiencies. We propose a more conditional view based on a digital-dependency perspective: renewable-energy-intensive systems may be more exposed to digital regulatory frictions because their efficient operation relies heavily on data-intensive coordination.

Unlike conventional fossil-fuel generation, renewable sources such as solar and wind are intermittent and geographically dispersed [41]. Maintaining reliability under high renewable penetration requires sophisticated digital infrastructures—smart grids, forecasting systems, real-time monitoring, and often cross-border data exchange—to balance supply and demand, optimize dispatch, and reduce curtailment [41]. In this sense, the operational efficiency of renewable energy systems is digitally dependent.

When digital services trade regulations diverge (i.e., when the regulatory gap is high), cross-border data mobility and interoperability can be constrained. Such frictions can disrupt data exchange and coordination, increasing the likelihood of suboptimal system operation (e.g., higher curtailment, less efficient balancing, and greater reliance on backup generation) and raising the emissions associated with economic activity, even in economies with strong renewable capacity [17]. Put differently, regulatory misalignment increases the opportunity cost of renewable capability by widening the gap between potential and realized decarbonization benefits. Fossil-fuel-intensive systems, which are comparatively less reliant on real-time data coordination for operational stability, may experience smaller marginal efficiency losses from the same regulatory frictions. Accordingly, we expect the carbon-intensifying effect of regulatory divergence to be stronger for exporters with higher renewable energy capability.

Hypothesis 3. 

The positive association between the digital services trade regulatory gap and export carbon intensity is stronger when the exporting country’s renewable energy capability is higher.

2.5. Moderating Role of Importers’ Environmental Policy Rigor

Institutional theory emphasizes that host-country policy environments shape firms’ strategic behavior and operational outcomes in cross-border activities [42,43]. Among these conditions, the environmental policy rigor of importing countries is particularly salient because it influences how firms manage operational inefficiencies and externalities associated with international exchange [44,45]. Stringent environmental policies increase the expected costs of carbon-intensive activities and strengthen incentives to adopt cleaner technologies, improve energy efficiency, and reduce emissions embodied in production and trade.

In digital services trade, regulatory gaps between trading partners can generate operational inefficiencies—such as duplicated digital processes, fragmented infrastructures, and compliance-driven redundancies—that raise energy use and may increase export carbon intensity. Whether these inefficiencies translate into higher embodied emissions, however, depends on the importing country’s environmental policy regime. When importers enforce rigorous environmental policies, exporting firms face stronger scrutiny and higher penalties associated with emissions-intensive practices, limiting the extent to which compliance-driven inefficiency can be accommodated through carbon-intensive operational responses.

From an environmental economics perspective, stringent environmental regulation internalizes externalities by increasing the marginal cost of emissions and strengthening incentives for mitigation. In importing countries with high environmental policy rigor, exporters are therefore more likely to respond to regulatory-induced inefficiencies by investing in efficiency-enhancing processes, adopting lower-carbon energy inputs where feasible, and optimizing operations to reduce emissions per unit of output. As a result, even if regulatory divergence continues to impose operational frictions, its carbon-intensity-increasing effect should be attenuated under stringent policy regimes.

By contrast, when importing countries have weaker environmental policies, exporters face fewer constraints and weaker incentives to offset compliance-driven inefficiencies through decarbonization measures. In such contexts, the energy and process redundancies induced by digital services trade regulatory gaps are more likely to pass through directly into higher embodied emissions. Thus, importers’ environmental policy rigor functions as a moderating institutional condition that shapes the environmental consequences of regulatory divergence in digital services trade.

Taken together, these arguments suggest that stricter environmental policy in importing countries weakens the positive association between digital services trade regulatory gaps and export carbon intensity. Accordingly, we propose the following:

Hypothesis 4. 

The positive association between the digital services trade regulatory gap (DSTRI gap) and export carbon intensity is weaker when importing countries exhibit higher environmental policy rigor.

3. Methodology and Data

3.1. Empirical Model

To examine how regulatory divergence in digital services trade affects the carbon intensity of bilateral exports, we employ a structural gravity framework, which is the standard empirical workhorse in international trade research [46,47].

Micro-founded gravity theory shows that bilateral trade outcomes depend not only on economic size and bilateral trade costs but also on each country’s trade barriers relative to all other partners—captured by multilateral resistance terms (MRTs). The structural system can be written as

$$X_{ijt}={\displaystyle \frac{Y_{it}{E}_{jt}}{Y_t}}{({\displaystyle \frac{{\tau }_{ijt}}{{\mathsf{\Pi }}_{it}{P}_{jt}}})}^{1-\sigma }$$

Crucially, (${\mathsf{\Pi }}_{it}$) and ($P_{jt}$) represent the outward and inward multilateral resistance terms (MRTs), respectively. These terms capture the fact that trade between two countries depends on their trade barriers relative to all other trading partners.

Ignoring these MRTs leads to significant omitted variable bias [48]. However, since MRTs are theoretically defined price indices and unobservable in practice, we follow the established methodology in recent empirical studies [13,49] by using time-varying country fixed effects to account for them.

Specifically, the exporter–year fixed effects (${\lambda }_{it}$) absorb the outward MRTs (${\mathsf{\Pi }}_{it}^{1-\sigma }$) as well as country-specific output ($Y_{it}$), while the importer–year fixed effects (${\psi }_{jt}$) absorb the inward MRTs ($P_{jt}^{1-\sigma }$) and expenditure ($E_{jt}$). This approach allows us to consistently estimate the effect of bilateral trade costs while controlling for all time-varying country-specific shocks.

We acknowledge that the inclusion of such high-dimensional fixed effects absorbs all country-specific time-varying variables (monadic terms), such as unilateral regulatory stringency or GDP. However, this rigorous specification is necessary to prevent omitted variable bias arising from unobserved Multilateral Resistance Terms [46]. Crucially, our primary variable of interest, the DSTRI Gap, is bilateral (dyadic) in nature. Unlike unilateral regulation levels, the regulatory distance varies across each country pair (ij). Therefore, the variation in the ‘Gap’ is orthogonal to the country–year fixed effects, allowing for precise identification of the impact of regulatory heterogeneity while rigorously controlling for national-level confounders [47]. (Technically, one might consider estimating the separate coefficients for Exporter DSTRI and Importer DSTRI (beta₁ DSTRI_it + beta₂ DSTRI_jt). However, in a structural gravity model with time-varying country fixed effects, these monadic variables are perfectly collinear with the Exporter–Year (μ_it) and Importer–Year (ν_jt) fixed effects, respectively, and thus cannot be identified. The bilateral gap (Gap_ijt), being a dyadic variable, avoids this collinearity and allows for identification under strict MRT controls.

However, applying this framework to carbon intensity requires methodological justification, as the gravity model typically explains trade volumes (flows). To address this, we employ an ‘Augmented Gravity Specification’. While the traditional gravity model is designed for flows, recent environmental trade literature extends this framework to analyze the quality or intensity of these flows [50,51].

The theoretical justification is twofold. First, carbon intensity (CI_ijt) is defined as the ratio of embodied emissions (E_ijt) to trade value (V_ijt). Since both the numerator (E_ijt) and the denominator (V_ijt) are structurally determined by gravity forces—such as economic mass, geographic distance, and institutional frictions—the resultant intensity is inherently a function of these same bilateral covariates. Second, gravity variables in our context serve as proxies for ‘Bilateral Friction’ rather than just transport costs. As posited in our theoretical framework, frictions such as regulatory gaps act as a selection mechanism that alters supply chain efficiency, thereby directly shaping the carbon content per unit of value. Therefore, estimating carbon intensity using a gravity-consistent structure allows us to identify how bilateral barriers shape the environmental efficiency of digital trade.

Based on this theoretical derivation, our baseline empirical specification is formulated as follows:

$$Y_{ijt}=\exp ({\beta }_0+{\beta }_{1}Ga{p}_{ijt}+{\beta 2Z}_{ijt}+{\lambda }_{it}+{\psi }_{jt})+{ϵ}_{ijt}$$

We estimate the specified gravity equation using the Poisson Pseudo Maximum Likelihood (PPML) estimator. The choice of PPML over the traditional Ordinary Least Squares (OLS) on the log-linearized form is driven by two critical econometric concerns: heteroskedasticity-induced bias and the zero trade flow problem.

First and foremost, traditional gravity models often log-linearize the multiplicative relationship $Y_{ijt}=\exp (X_{ijt}\beta ){\eta }_{ijt}\ln Y_{ijt}=X_{ijt}\beta +\ln {\eta }_{ijt}$ to estimate parameters via OLS. However, Santos Silva and Tenreyro (2006) demonstrate that if the error term (${\eta }_{ijt}$) is heteroskedastic—meaning its variance depends on the regressors ($X_{ijt}$)—OLS estimates become inconsistent [52]. This inconsistency arises from Jensen’s Inequality, where the expected value of the logarithm of a random variable is not equal to the logarithm of its expected value:

$$E[\ln Y_{ijt}|X_{ijt}]\ne \ln E[Y_{ijt}|X_{ijt}]$$

Specifically, if the variance of the error term depends on the covariates (e.g., trade flows vary more for larger economies), the condition $E[\ln {\eta }_{ijt}|X_{ijt}]=0$ is violated, rendering OLS estimates biased and inconsistent. In contrast, the PPML estimator addresses this issue by estimating the non-linear model directly in levels, relying on the following moment condition:

$$E[Y_{ijt}-\exp (X_{ijt}\beta +{\lambda }_{it}+{\psi }_{jt})|X_{ijt}]=0$$

By targeting the conditional mean directly, PPML provides consistent estimates even in the presence of heteroskedasticity, making it the standard estimator for gravity models.

Second, the OLS log-linearization approach cannot handle observations where the dependent variable is zero ($Y_{ijt}=0$), as the logarithm of zero is undefined. Standard practices of dropping these observations or adding a small arbitrary constant (e.g., $\ln (Y_{ijt}+1)$) introduce sample selection bias and inconsistency [53]. Our dataset contains zero trade flows or negligible carbon intensities in certain dyads. The PPML estimator naturally incorporates these zero values because it models the dependent variable in levels, thereby utilizing the full information contained in the sample.

Finally, following standard practice in the gravity literature [47], we cluster standard errors at the country-pair level to account for potential serial correlation in the error term over time.

3.2. Variable Construction

3.2.1. Dependent and Independent Variables

To assess the environmental consequences of digital trade regulation, we use two complementary outcome measures that capture distinct environmental dimensions of bilateral exchange: carbon intensity (environmental efficiency) and the level of embodied emissions (aggregate environmental burden).

Our primary dependent variable is Carbon Intensity ($Intensit{y}_{ijt}$), which captures the environmental efficiency of trade flows. It is calculated by normalizing the total greenhouse gas (GHG) emissions embedded in bilateral trade by the gross export volume, as shown in the following equation:

$$Intensit{y}_{ijt}={\displaystyle \frac{\mathrm{Emissions} \mathrm{Embodied} \mathrm{in} \mathrm{Bilateral} {\mathrm{Trade}}_{{ijt (\mathrm{TonCO}}_{2)}}}{\mathrm{Gross} {\mathrm{Exports}}_{ijt (\mathrm{Million} \mathrm{USD})}}}$$

This intensity-based approach is crucial because it allows us to distinguish between a simple increase in trade volume and a structural change in environmental performance. A higher value of carbon intensity indicates a lower level of carbon efficiency, implying that the exporting country emits more carbon dioxide to generate the same unit of economic value in its exports. Conversely, a lower intensity signifies a cleaner, more carbon-efficient trade structure.

Theoretically, changes in carbon intensity can stem from two sources: the ‘Composition Effect’ (shifts in the sectoral mix of exports) and the ‘Technique Effect’ (changes in emission efficiency within sectors). While our dependent variable aggregates both effects, our empirical strategy relies on Exporter–Year Fixed Effects to absorb the exporter’s overall industrial structure and composition changes over time. Consequently, the estimated impact of the regulatory gap in our model primarily captures the ‘Technique Effect’—specifically, the efficiency losses arising from bilateral regulatory friction (e.g., duplicated infrastructure and operational redundancy)—rather than unilateral shifts in industrial composition.

For robustness, we also utilize the Level of Embodied Emissions ($Emission{s}_{ijt}$) as an alternative dependent variable. This variable represents the absolute volume of GHG emissions generated throughout the production, transport, and delivery processes of traded goods and services. Measured in million tonnes of CO₂-equivalent (MtCO₂-eq.), this indicator reflects the total carbon footprint embedded in the bilateral trade of all product categories. While intensity measures efficiency, the absolute level provides valuable insights into the aggregate environmental burden and offers policy directions for effective climate regulation.

It is important to note that these two dependent variables capture distinct environmental dimensions. ‘Total Emissions’ reflect the ‘Scale Effect,’ representing the aggregate environmental burden driven by trade volume. In contrast, ‘Carbon Intensity’ reflects the ‘Technique Effect,’ representing the environmental efficiency per unit of economic value [26]. By analyzing both, we aim to disentangle whether the regulatory gap merely expands the scale of pollution (via increased trade) or structurally degrades environmental efficiency (via institutional friction). Our theoretical framework posits that regulatory gaps negatively affect both, but the policy implications differ: the former calls for cleaner energy scaling, while the latter calls for regulatory harmonization to eliminate structural waste.

The core independent variable of interest is the Digital Services Trade Restrictiveness Index (DSTRI) Gap, which proxies the institutional distance in digital trade regulations between trading partners. We utilize the OECD DSTRI, a composite index that measures horizontal barriers affecting trade in digitally enabled services.

The OECD DSTRI ranges from 0 to 1, where 0 represents a completely open regulatory environment and 1 indicates a completely closed regime. The index aggregates policy restrictions across five key pillars: (1) infrastructure and connectivity, (2) electronic transactions, (3) payment systems, (4) intellectual property rights, and (5) other barriers affecting trade in digitally enabled services.

Since our empirical framework is based on a bilateral gravity model, we construct the DSTRI Gap ($Ga{p}_{ijt}$) variable by calculating the absolute difference in DSTRI scores between the exporting country i and the importing country j in year t:

$$Ga{p}_{ijt}=|{DSTRI}_{it}-{DSTRI}_{jt}|$$

This variable captures the degree of regulatory heterogeneity. By using the absolute value, we assume a ‘Symmetric Friction’ effect, positing that any deviation from the domestic regulatory environment—whether stricter or more lenient—imposes adaptation costs on firms [27]. This approach aligns with the ‘Institutional Mismatch’ perspective, which argues that the primary driver of transaction costs in digital trade is the lack of interoperability and the need to maintain dual compliance systems, rather than the direction of the stringency gradient. Consequently, a larger gap implies higher regulatory friction and coordination costs, which we hypothesize will hinder the efficiency of digital operations and subsequently affect carbon intensity.

3.2.2. Moderators and Mediators

To examine the heterogeneous effects of digital trade regulations, we introduce two moderating variables representing the energy structure of the exporter and the policy environment of the importer.

First, Renewable Electricity Output (${RenRatio}_{it}$) measures the exporting country’s renewable energy capability. Defined as the share of electricity generated from renewable sources in total electricity output, this variable serves as a proxy for the carbon intensity of the energy grid powering digital operations. A higher share of renewables implies that the energy consumed by digital infrastructure—such as data centers and networks—is cleaner, thereby facilitating the decoupling of digital trade growth from carbon emissions. It is measured in percentage terms (%).

Second, the Environmental Policy Stringency Index (${EPS}_{jt}$) captures the regulatory rigor of the importing country. Developed by the OECD, this composite index assesses the stringency of environmental policies across 13 policy instruments, including market-based mechanisms (e.g., taxes), non-market regulations (e.g., emission limits), and technology support measures (e.g., R&D subsidies). The index ranges from 0 (least stringent) to 6 (most stringent). A higher score indicates that the importer imposes stricter costs on pollution and stronger incentives for eco-innovation, which acts as an institutional pressure on trading partners to improve their environmental performance.

To identify the transmission mechanism through which digital regulations affect carbon intensity, we construct a measure of Digital Trade Volume (${DigiTrade}_{ijt}$). Following the conceptual framework of UNCTAD and the U.S. Bureau of Economic Analysis (BEA), we define digital trade as “digitally delivered trade,” focusing specifically on Mode 1 (cross-border supply) exports of ICT services (telecommunications, computer, and information services).

Since direct bilateral data for Mode 1 ICT services is limited, we calculate this variable using a proportional allocation method by combining two datasets: the Balanced Trade in Services (BaTiS) dataset and the Trade in Services by Mode of Supply (TiSMOS) dataset. Specifically, we apply the share of Mode 1 exports derived from TiSMOS to the bilateral service trade flows reported in BaTiS. This proportional allocation method aligns with recent methodological approaches aimed at increasing the granularity of services trade data (see, e.g., WTO and OECD, 2023), ensuring that our measure accurately reflects the digital component of bilateral service flows. This approach allows us to isolate the value of services that are actually delivered over computer networks. The variable is measured in billions of USD.

We construct this variable following the established definition of ‘Potentially ICT-enabled services’ proposed by UNCTAD (2015) [54]. This approach aggregates service categories—such as telecommunications, computer services, insurance, finance, and other business services—that can predominantly be delivered remotely over ICT networks. We acknowledge that this broad aggregation relies on the assumption that the majority of trade in these categories is digitally delivered, which may introduce measurement error. However, we contend that this proxy is superior to narrower alternatives (e.g., restricting the measure solely to ‘Computer and Information Services’) because it captures the broader digitalization of the service economy. Furthermore, to the extent that random measurement error exists, standard econometric theory suggests it would lead to attenuation bias, biasing our estimates towards zero. Thus, our significant positive findings likely represent a conservative lower bound of the true effect of digital trade expansion.

Finally, we clarify the dual analytical role of this variable in our empirical strategy. In the baseline gravity estimation, Digital Trade Volume serves as a control variable to satisfy the ceteris paribus condition, allowing us to isolate the direct impact of regulatory friction on carbon intensity independent of trade scale. Subsequently, in the mechanism analysis, we treat it as a mediator to explicitly examine whether the regulatory gap also influences environmental outcomes through the indirect channel of trade expansion/contraction. This stepwise distinction prevents circularity and enables the decomposition of effects into ‘efficiency’ and ‘scale’ components.

3.2.3. Control Variable

To isolate the specific impact of digital trade regulatory gaps on carbon intensity, we control for a set of standard gravity variables that proxy for bilateral trade costs, cultural proximity, and preferential trade policies. First, we draw standard geographic and cultural variables from the CEPII Gravity Database.

Geographic Distance (${\ln distcap}_{ij}$): We include the natural logarithm of the simple distance between the capital cities of the trading partners (measured in kilometers). This variable serves as a primary proxy for transportation costs and information friction between the two countries.

Contiguity ($Conti{g}_{ij}$): A dummy variable that takes a value of 1 if the origin and destination countries share a common land border, and 0 otherwise. This captures the trade-facilitating effect of geographic adjacency.

Common Language ($\mathrm{com}lan{g\_off}_{ij}$): A dummy variable equal to 1 if the trading partners share a common official or primary language, and 0 otherwise. This variable accounts for cultural proximity and reduced communication costs, which are known to facilitate cross-border economic activities.

Second, to control for the effect of traditional trade liberalization, we utilize the Free Trade Agreement (${fta}_{ijt}$) dummy variable sourced from the World Bank Deep Trade Agreements (WB-DTA) database. This variable equals 1 if a free trade agreement is in force between country i and country j in year t, and 0 otherwise. Controlling for FTA status is essential to distinguish the marginal effect of digital trade regulations from the broader trade-promoting effects of tariff reductions and market access provisions included in traditional trade pacts.

3.3. Data Description

Our final dataset is a panel of 10,776 bilateral (country-pair–year) observations covering major trading relationships. We construct the dataset by merging bilateral trade and embodied-emissions measures with digital services trade regulation indices and standard gravity covariates.

Our empirical analysis focuses on OECD member countries for three reasons. First, OECD economies account for a substantial share of global digitally enabled services trade, ensuring the economic relevance of the setting. Second, the OECD provides comparatively harmonized and high-quality data on both services trade regulations (DSTRI) and environmental indicators, reducing concerns about cross-country measurement inconsistency. Third, restricting the sample to a relatively institutionally similar set of countries offers a conservative test of our theory: if regulatory divergence is associated with meaningful carbon inefficiencies even among OECD partners, the environmental costs of regulatory fragmentation may plausibly be larger in more institutionally heterogeneous global settings. Accordingly, our estimates should be interpreted as conservative benchmarks rather than upper-bound effects.

Table 1. Variable Construction and Data Source.

Variables	Definition and Measurement	Data Source
ghgfp_intensity	(Intensity) Emissions embodied in bilateral trade (TonCO2/Million USD)	OECD Data Explorer
ghgfp_millionton	(Level) Emissions embodied in bilateral trade (million tonnes of CO-equivalent)	OECD Data Explorer
oecd_dstri_gap	Digital Services Trade Restrictiveness Index (0~1)	OECD Going Digital Toolkit
fta	Dummy variable taking 1 if a Free Trade Agreement exists, 0 otherwise	World Bank Deep Trade Agreement
distcap	Weighted distance between trading partners (km)	CEPII
comlang_off	Common official or primary language (Dummy 0/1)	CEPII
contig	Origin and destination are contiguous (Dummy 0/1)	CEPII
eps_index	Environmental Policy Stringency Index (index 0~6)	OECD Data Explorer
renewable_energy_percent	Renewable electricity output (% of total electricity output)	World Bank
digitaltrade_mode1_million	Digital trade (Mode 1 ICT exports) (Million USD)	BaTiS, TiSMOS

reports variable definitions and data sources.

Table 2. Descriptive Statistics.

Variable	Mean	Std. Dev.	Max	Min	Obs
ghgfp_intensity	0.001	0.000	0.007	0.000	10,776
ghgfp_millionton	3589.285	16,012.173	400,025.00	1.000	10,776
oecd_dstri_gap	0.095	0.079	0.350	0.000	10,776
oecd_dstri_gap	0.222	0.122	0.650	0.000	10,776
eps_index	2.788	1.010	4.889	0.556	10,776
renewable_energy_percent	40.277	27.506	99.994	1.586	10,776
digitaltrade_mode1_million	162.745	545.815	12,496.755	0.000	10,776
fta	0.689	0.463	1.000	0.000	10,776
distcap	8.274	1.126	9.896	4.007	10,776
contig	0.046	0.210	1.000	0.000	10,776
comlang_off	0.079	0.270	1.000	0.000	10,776

Notes: The variable distcap is log-transformed. DSTRI Sum refers to the sum of the exporter’s and importer’s DSTRI scores. Total Emissions are measured in Gigagrams (CO_2e).

summarizes the main descriptive statistics for the balanced panel of 10,776 observations. The dependent variables exhibit substantial cross-dyad variation. Carbon intensity (export emissions per unit of export value) has a low mean (0.001) with non-trivial dispersion, indicating meaningful differences in emissions efficiency across trading relationships. Embodied emissions levels vary widely across dyads, consistent with heterogeneity in trade scale and the carbon content of production across exporters and sectors.

Our main explanatory variable, the DSTRI gap, has a mean of 0.095 and ranges from 0 to 0.350, indicating moderate average regulatory divergence but substantial heterogeneity across pairs. The DSTRI sum—included to capture the overall level of restrictiveness across the dyad—has a mean of 0.222, suggesting that regulatory distance and regulatory level are empirically distinct dimensions in the sample.

The moderators also display meaningful variation. The environmental policy stringency (EPS) index averages 2.788 on the 0–6 scale, while exporters’ renewable electricity share ranges from 1.6% to nearly 100% (mean = 40.28%), reflecting large differences in grid composition across OECD members. Our bilateral digital trade measure (Mode 1 digitally delivered services proxy) also varies substantially, with a mean of approximately 162.7 million USD and a long right tail consistent with the presence of major digital-trade hubs.

Among gravity-style controls, 69% of dyads are covered by an FTA in a given year. Only 4.6% of pairs share a land border (contiguity), and 7.9% share a common official language. The mean log distance is 8.274. Overall, the descriptive statistics indicate sufficient cross-sectional and temporal variation to identify associations between regulatory divergence and the environmental efficiency and burden of bilateral trade.

4. Empirical Results

4.1. Baseline Results

Table 3. Baseline Results: The Impact of Digital Regulatory Gap on Carbon Emissions.

Variable	Intensity(OECD) (1)	Total Emissions(OECD) (2)
oecd_dstri_gap	0.390 ***	3.76 ***
	(0.121)	(1.032)
fta	0.063 ** (0.032)	0.687 *** (0.103)
distcap	−0.084 ***	−0.234 ***
	(0.013)	(0.078)
contig	0.084 ** (0.033)	1.155 *** (0.193)
comlang_off	0.003 (0.036)	0.375 ** (0.185)
Constant	−6.790 ***	3.690 ***
	(0.124)	(0.762)
Observations	10,925	11,189
Exporter–Year FE	Yes	Yes
Importer–Year FE	Yes	Yes

Notes: Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05.

reports the baseline estimates of the relationship between digital services trade regulatory divergence and embodied carbon outcomes in bilateral trade. We estimate PPML models with exporter–year and importer–year fixed effects to account for multilateral resistance terms and absorb time-varying country-level heterogeneity.

Column (1) uses export carbon intensity as the dependent variable. The coefficient on the DSTRI gap is positive and statistically significant at the 1% level (β = 0.390, p < 0.01). This finding indicates that greater regulatory divergence between trading partners is associated with higher emissions per unit of export value. Substantively, the result is consistent with the view that regulatory heterogeneity functions as a dyadic friction that reduces interoperability and induces compliance-related redundancy, thereby lowering the emissions efficiency of bilateral exchange.

Column (2) uses the level of embodied emissions as the dependent variable. The coefficient on the DSTRI gap is again positive and statistically significant (β = 3.759, p < 0.01), implying that regulatory divergence is associated with a larger aggregate embodied-emissions footprint in bilateral trade within the OECD sample. This pattern suggests that regulatory fragmentation is linked not only to lower environmental efficiency but also to higher overall embodied emissions. One plausible interpretation is that, among advanced economies, regulatory heterogeneity may coexist with (or reflect) specialization and complementarities that sustain bilateral exchange even when compliance frictions increase—thereby expanding the emissions scale alongside efficiency losses.

The estimated coefficients for gravity-style controls are broadly consistent with expectations. Log distance is statistically significant in both specifications, indicating that geographic separation remains relevant for embodied carbon outcomes in bilateral trade. The FTA indicator is positive and significant, consistent with preferential agreements being associated with higher trade activity and, correspondingly, higher embodied emissions. Contiguity and common official language also show positive associations, in line with their trade-facilitating role.

4.2. Additional Analysis

To assess the stability of our baseline findings, we conduct a set of robustness checks using alternative specifications, estimators, sample splits, and shock exclusions.

Table 4. Robustness Checks.

Variable	Lagged IV	OLS	Divide the Sample Interval		Exclude Major Event
	(1)	(2)	(3)	(4)	(5)
Lagged oecd_dstri_gap (t-1)	0.365 *** (0.124)
fta	0.068 ** (0.034)	0.124 *** (0.026)	0.027 (0.035)	0.092 *** (0.035)	0.059 * (0.033)
distcap	−0.083 *** (0.013)	−0.086 *** (0.012)	−0.098 *** (0.016)	−0.073 *** (0.014)	−0.087 *** (0.014)
contig	0.084 ** (0.034)	0.087 *** (0.030)	0.076 ** (0.032)	0.090 ** (0.035)	0.081 ** (0.032)
comlang_off	0.003 (0.037)	0.011 (0.027)	-0.015 (0.042)	0.018 (0.035)	0.004 (0.037)
oecd_dstri_gap		0.417 *** (0.116)	0.378 ** (0.148)	0.420 *** (0.130)	0.368 *** (0.125)
Constant	−6.794 *** (0.126)	−7.186 *** (0.105)	−6.588 *** (0.150)	−6.955 *** (0.127)	−6.755 *** (0.127)
Observations	9361	10,925	4687	6238	9364
Exporter–Year FE	Yes	Yes	Yes	Yes	Yes
Importer–Year FE	Yes	Yes	Yes	Yes	Yes

Notes: Robust standard errors in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1. Notes: (3) 2014–2016, (4) 2017–2020, (5) COVID-19.

summarizes the results.

Lag structure to mitigate simultaneity: Column (1) replaces the contemporaneous DSTRI gap with its one-year lag (Lagged Gap t-1) to reduce concerns that regulatory divergence and carbon intensity are jointly determined within the same year. The coefficient remains positive and statistically significant (β = 0.365, p < 0.01), indicating that the association is not driven solely by contemporaneous co-movement.

Alternative estimator: Column (2) re-estimates the baseline relationship using OLS with a log-transformed dependent variable as a sensitivity check against the PPML specification. The estimated effect of the DSTRI gap remains positive and statistically significant (β = 0.417, p < 0.01), suggesting that the main result does not hinge on the choice of estimator.

Subsample stability: Columns (3) and (4) split the sample into 2014–2016 and 2017–2020 to test whether the estimated relationship changes across periods characterized by different global regulatory conditions. The DSTRI gap remains positive in both subsamples (β = 0.378 and β = 0.420, respectively), indicating temporal stability. The slightly larger coefficient in the later period is consistent with the idea that, as digital regulation becomes more salient and complex, interoperability frictions may translate more strongly into efficiency losses—though we interpret this difference cautiously.

Excluding COVID-19: Column (5) excludes observations from 2020 to ensure that the results are not driven by the exceptional trade disruptions associated with the COVID-19 pandemic. The coefficient remains positive and significant (β = 0.368, p < 0.01), supporting the view that the observed relationship is not an artifact of a temporary global shock.

Overall, the robustness checks in Table 4 consistently support our baseline conclusion: regulatory divergence in digital services trade is systematically associated with higher export carbon intensity, consistent with the argument that regulatory heterogeneity generates interoperability frictions and compliance-related redundancy that erode environmental efficiency in cross-border exchange.

4.3. Exploratory Analysis of Transmission Channels

To further probe how regulatory divergence relates to environmental outcomes, we conduct an exploratory channel analysis that considers the role of digital services trade volume. The objective is to assess whether the empirical patterns are more consistent with a pathway operating through trade scale (changes in the magnitude of exchange) or through structural inefficiency (changes in emissions efficiency conditional on scale).

Table 5. Exploratory Channel Analysis: Associations with Digital Trade Volume.

Variable	Total Emission			Intensity
	Step 1 (1)	Step 2 (2)	Step 3 (3)	Step 1 (4)	Step 2 (5)	Step 3 (6)
oecd_dstri_gap	3.758 *** (1.033)	1.861 ** (0.778)	−0.198 (0.585)	0.375 *** (0.121)	1.861 ** (0.778)	0.381 *** (0.122)
Ln(digitaltrade_mode1)			0.407 *** (0.046)			−0.018 (0.012)
fta	0.687 *** (0.103)	0.284 ** (0.124)	0.164 * (0.086)	0.058 * (0.032)	0.284 ** (0.124)	0.062 ** (0.032)
Log(distcap)	−0.234 *** (0.078)	−0.604 *** (0.039)	−0.339 *** (0.048)	−0.085 *** (0.013)	−0.604 *** (0.039)	−0.097 *** (0.017)
contig	1.156 *** (0.193)	−0.099 (0.135)	0.526 *** (0.099)	0.092 *** (0.033)	−0.099 (0.135)	0.095 *** (0.034)
comlang_off	0.375 ** (0.185)	0.273 ** (0.115)	0.008 (0.121)	−0.005 (0.036)	0.273 ** (0.115)	0.003 (0.036)
Constant	10.599 *** (0.762)	10.914 *** (0.362)	10.207 *** (0.565)	−6.776 *** (0.123)	10.914 *** (0.362)	−6.628 *** (0.169)
Observations	11,036	10,776	10,776	10,776	10,776	10,776
Exporter–Year FE		Yes	Yes	Yes	Yes	Yes
Importer–Year FE		Yes	Yes	Yes	Yes	Yes

Notes: Columns (1)–(3) and (4)–(6) report sequential PPML regressions used to explore potential transmission patterns. Given the non-linear nature of PPML with high-dimensional fixed effects, the estimates should be interpreted as descriptive associations rather than formal causal mediation effects. Robust standard errors clustered at the country-pair level are reported in parentheses. Significance levels: *** p < 0.01, ** p < 0.05, * p < 0.1.

reports the results. Columns (1)–(3) use embodied emissions levels as the dependent variable, whereas Columns (4)–(6) use export carbon intensity.

A key methodological caveat applies. Because the models are estimated using non-linear PPML with high-dimensional fixed effects, coefficients do not admit the simple additive decomposition underlying the classic Baron–Kenny mediation logic used in linear models. Accordingly, the results should be interpreted as pattern-based evidence—that is, whether the estimated associations are consistent with the hypothesized channels—rather than as formal causal mediation tests.

(1) Patterns consistent with a scale channel for embodied emissions:

Columns (1)–(3) display patterns consistent with a scale-related channel for total embodied emissions. Column (1) reproduces the baseline result that the DSTRI gap is positively associated with embodied emissions levels (β = 3.758, p < 0.01). Column (2) shows that the DSTRI gap is also positively associated with digital services trade volume (β = 1.861, p < 0.05). This is notable in an OECD setting: regulatory heterogeneity may not operate primarily as an entry barrier, as is sometimes observed in broader global samples, but may coexist with continued cross-border exchange among advanced economies that demand specialized digital services and capabilities. Under such conditions, trade may persist (or even expand) despite higher compliance frictions, consistent with a pattern of interdependence rather than deterrence. When both the DSTRI gap and digital trade volume are included in Column (3), digital trade volume remains positive and statistically significant (β = 0.407, p < 0.01), while the coefficient on the DSTRI gap becomes statistically insignificant (β = −0.198). Interpreted descriptively, this pattern suggests that the positive association between regulatory divergence and embodied emissions levels is largely consistent with a trade-scale pathway—i.e., dyads with greater regulatory divergence also exhibit higher digital services trade volumes, which are in turn associated with higher aggregate embodied emissions.

(2) Patterns consistent with structural friction for carbon intensity

Columns (4)–(6) present a different pattern for carbon intensity, which captures emissions efficiency per unit of export value. Column (4) confirms a positive association between the DSTRI gap and carbon intensity (β = 0.375, p < 0.01). Importantly, in Column (6), adding digital services trade volume does not attenuate the estimated association: the DSTRI gap remains positive and statistically significant (β = 0.381, p < 0.01), while digital trade volume is statistically insignificant (β = −0.018). This pattern suggests that, conditional on trade scale, regulatory divergence remains closely linked to lower emissions efficiency, consistent with the view that regulatory misalignment embeds compliance and interoperability frictions into the trade process that raise emissions per unit value.

Overall, the exploratory results point to a dual dynamic. For embodied emissions levels, the data patterns are consistent with a scale-related channel in which regulatory divergence is associated with higher digital trade volumes and, therefore, a larger aggregate emissions footprint. For carbon intensity, the patterns are more consistent with a structural-friction channel in which regulatory divergence is associated with higher emissions per unit value, largely independent of digital trade volume. Substantively, these findings imply that expanding digital trade alone is unlikely to improve carbon efficiency unless interoperability and institutional fragmentation in digital trade governance are addressed.

4.4. Heterogeneity Analysis

To examine whether the environmental consequences of digital services trade regulatory divergence vary across institutional and technological contexts, we estimate specifications that interact the DSTRI gap with (i) importers’ environmental policy stringency (EPS) and (ii) exporters’ renewable electricity share.

Table 6. Heterogeneity Analysis: The Moderating Roles of Environmental Regulation and Energy Structure.

Variable	Total Emissions (1)	Intensity (2)
oecd_dstri_gap	5.710 *** (1.734)	0.150 (0.352)
oecd_dstri_gap x eps_index	−0.197 (0.481)	−0.047 (0.118)
oecd_dstri_gap x renewable_energy	−4.962 ** (2.255)	0.901 ** (0.417)
Constant	10.589 *** (0.752)	−6.792 *** (0.121)
Observations	11,036	10,776

Notes: All specifications include the full set of control variables (log distance, FTA, contiguity, common official language, and DSTRI sum), which are omitted from the table for brevity. Robust standard errors clustered at the country-pair level are reported in parentheses. Significance levels: *** p < 0.01, ** p < 0.05.

reports the results.

(1) Exporter renewable electricity share and embodied emissions (levels)

Column (1) reports the results for embodied emissions levels. The main effect of the DSTRI gap remains positive and statistically significant (β = 5.710, p < 0.01), consistent with the baseline finding that regulatory divergence is associated with a larger aggregate embodied-emissions footprint. Importantly, the interaction between the DSTRI gap and exporters’ renewable electricity share is negative and statistically significant (β = −4.962, p < 0.05). This pattern suggests that a cleaner electricity mix in the exporting country attenuates the emissions-increasing association of regulatory divergence at the aggregate level. In substantive terms, when digitally mediated exchange (and the associated data-processing demand) expands under regulatory fragmentation, the resulting emissions burden is smaller when the underlying electricity used to power production and digital infrastructure is less carbon-intensive.

(2) Exporter renewable electricity share and carbon intensity

Column (2) reports results for export carbon intensity and reveals a contrasting pattern. The interaction between the DSTRI gap and exporters’ renewable electricity share is positive and statistically significant (β = 0.901, p < 0.05), indicating that the carbon-intensity-increasing association of regulatory divergence is stronger for exporters with higher renewable penetration. While this may appear counterintuitive, it is consistent with our digital-dependency logic: renewable-heavy systems rely more on data-intensive coordination and interoperability (e.g., forecasting, balancing, and smart-grid management). When regulatory misalignment constrains cross-border data mobility and interoperability, coordination frictions may translate into greater operational redundancy and efficiency losses, increasing emissions per unit of export value even in relatively “green” electricity contexts.

(3) Limited evidence for moderation by importer environmental policy stringency

Across both outcome specifications, the interactions involving importer EPS are statistically insignificant (β = −0.197 and β = −0.047). Interpreted cautiously, this suggests that stricter environmental regulation in destination markets may not, by itself, offset the efficiency losses associated with digital regulatory fragmentation. One interpretation is that environmental policy stringency primarily raises incentives to reduce emissions at the margin, whereas the frictions induced by regulatory misalignment operate through interoperability constraints and compliance-related redundancy that are not directly addressed by environmental policy instruments.

Overall, the heterogeneity analysis suggests that the environmental implications of digital regulatory divergence depend importantly on the exporter’s energy-system characteristics. Renewable electricity share appears to mitigate aggregate embodied emissions associated with regulatory divergence, while simultaneously amplifying the carbon-intensity penalty linked to regulatory fragmentation. These results underscore a broader complementarity: progress in energy-system decarbonization does not automatically translate into efficiency gains in digitally mediated trade if digital governance remains fragmented. Aligning digital interoperability with sustainability and energy transitions may therefore be necessary to avoid unintended efficiency losses.

5. Discussion and Implications

5.1. Discussion

Our findings provide consistent evidence that regulatory divergence in digital services trade is associated with both higher embodied emissions and higher export carbon intensity, supporting Hypothesis 1. By contrast, digital services trade volume is not significantly related to carbon intensity. This pattern is inconsistent with Hypothesis 2a and instead suggests that any technique-related efficiency gains from digitalization are not strong enough to reduce emissions per unit of export value in this setting. The result aligns with rebound-effect arguments in environmental economics: efficiency improvements can lower the effective cost of data use and stimulate additional demand for processing, storage, and transmission that offsets potential efficiency gains [40]. Together, these results are consistent with the “dual nature” of digital trade articulated in Section 2.3—digitalization may enable efficiency improvements, but these benefits can be neutralized by induced demand and by institutional frictions.

The estimated effects also appear economically meaningful. Using the semi-elasticity for the DSTRI gap, a one-standard-deviation increase in regulatory divergence is associated with a non-trivial proportional increase in export carbon intensity. Substantively, this implies that regulatory incompatibility can operate as a hidden efficiency penalty embedded in digitally mediated exchange. Unlike financial charges that generate revenue, this penalty represents deadweight environmental loss—energy consumed to manage interoperability and duplicative compliance rather than to create productive value.

This interpretation clarifies how digital regulatory divergence differs from conventional trade barriers. Whereas tariffs primarily distort prices and redistribute welfare, misaligned digital regimes can distort processes by inducing fragmented architectures and compliance-driven duplication. Our exploratory channel patterns are consistent with this distinction. For embodied emissions levels, the estimates are consistent with a trade-scale pathway in which regulatory divergence coexists with (and may accompany) continued digital trade exchange among advanced economies. For carbon intensity, the DSTRI gap remains robust even when accounting for trade volume, indicating that regulatory divergence is associated with a persistent emissions-efficiency penalty that is not reducible to scale alone. In this sense, digital regulatory gaps function as a technique-related shifter, degrading emissions efficiency conditional on trade magnitude.

A particularly notable finding is that the carbon-intensity penalty associated with regulatory divergence is stronger in exporters with higher renewable electricity shares. This runs counter to the expectation that cleaner electricity buffers environmental inefficiency. However, the pattern is interpretable through an opportunity-cost and complementarity lens. Renewable-intensive exporters have greater latent potential to translate digitalization into lower carbon intensity via data-driven coordination, interoperability, and dematerialized exchange. When regulatory divergence constrains cross-border data mobility and digital interoperability, these exporters may be disproportionately prevented from realizing such gains, widening the gap between potential and realized efficiency. Moreover, where production and coordination are more digitally integrated, fragmented digital governance can impose larger coordination losses—through duplicated processing, parallel infrastructures, and redesigned workflows across jurisdictions—offsetting a larger share of the technique benefits that digitalization could otherwise deliver. The implication is not that renewable energy is ineffective, but that its environmental returns in trade contexts may be conditional on interoperable digital governance.

We acknowledge that this interpretation is necessarily indirect. Our moderation results operate at the country level and do not directly observe the technical coupling between power systems, digital infrastructures, and firm-level compliance architectures. While the patterns are consistent with complementarity, validating specific bottlenecks (e.g., redundancy induced by localization requirements, data-center energy sourcing, or grid-balancing constraints) would require micro-level evidence. Future work drawing on firm-level compliance strategies or facility-level energy-use data could strengthen mechanistic inference.

We also find limited evidence that importers’ environmental policy stringency (EPS) moderates the relationship between regulatory divergence and embodied carbon outcomes. Rather than implying that environmental regulation is unimportant, this null result is consistent with a policy–problem mismatch: EPS largely reflects instruments designed to regulate emissions from physical production and fuel use, whereas part of the digital trade footprint arises from electricity demand in distributed digital infrastructures that may be weakly targeted by conventional instruments. This suggests that environmental governance may require more digital-specific tools—such as digital carbon accounting standards, disclosure of data-center energy sourcing, and interoperable reporting and compliance requirements—to address the channels through which regulatory fragmentation generates emissions inefficiency.

Importantly, digital regulatory divergence often reflects legitimate sovereign objectives, including privacy protection, cybersecurity, consumer welfare, and national security. Accordingly, our findings should not be interpreted as advocating indiscriminate deregulation. Instead, they highlight a trade-off: when countries pursue valid objectives in non-aligned ways, the resulting lack of interoperability can generate unintended environmental costs by inducing duplication and inefficiency. A constructive policy direction is therefore regulatory interoperability rather than a race to the bottom—for example, mutual recognition arrangements, interoperable technical standards, and cross-border compliance mechanisms that preserve policy goals while reducing emissions-intensive redundancy.

Finally, limitations warrant caution. Although we employ high-dimensional fixed effects and lag structures to mitigate confounding and simultaneity, the analysis remains observational and does not claim strict causality in the absence of exogenous regulatory variation. Unobserved bilateral factors could jointly shape regulatory choices, trade patterns, and emissions outcomes. External validity may also be bounded: within the OECD, regulatory divergence may coexist with specialization and interdependence, whereas in lower-capacity contexts, divergence may function more strongly as an entry barrier with different scale–efficiency implications. Future research should explicitly examine these asymmetries across advanced and developing economies.

5.2. Theoretical Contributions

This study contributes to the digital trade–environment literature in three ways. First, we shift the analytical emphasis from trade volume to institutional friction. Whereas prior research has largely asked whether digitalization reduces emissions via technique effects or increases emissions via scale effects, it has devoted less attention to how the governance of digital trade shapes environmental performance [17]. Our results show that regulatory heterogeneity (the DSTRI gap) is systematically associated with higher carbon intensity, indicating that institutional misalignment can generate environmental inefficiencies distinct from those attributable to trade expansion alone.

Second, we articulate and provide evidence consistent with a specific mechanism—carbon-intensive redundancy—through which regulatory divergence can translate into higher embodied emissions and lower emissions efficiency. Extending gravity-based reasoning to the digital–environmental domain, we argue that regulatory gaps can induce duplicative infrastructures and processes (e.g., jurisdiction-specific storage, parallel verification and monitoring routines, repeated processing), implying that regulatory fragmentation can generate deadweight environmental loss rather than simply administrative costs.

Third, we identify a paradoxical pattern in which the carbon-intensity penalty associated with regulatory divergence is stronger in renewable-intensive exporting economies. This challenges the assumption that green energy capacity automatically buffers environmental inefficiency and motivates a systemic complementarity perspective: the environmental returns to renewable energy investments may depend on interoperable digital trade governance that enables cross-border coordination and reduces redundancy.

5.3. Practical Implications

Our findings imply three policy-relevant lessons for managing the trade–environment nexus in the digital economy. First, the appropriate response to regulatory fragmentation is not necessarily deregulation, but interoperability. Because the emissions-efficiency penalty appears linked to regulatory incompatibility rather than regulatory level per se, policymakers should prioritize instruments that reduce compliance duplication while preserving legitimate policy objectives. Examples include mutual recognition arrangements and interoperable certification/audit procedures for privacy and security standards, which can allow firms to operate with unified compliance architectures across jurisdictions and reduce redundancy-driven emissions.

Second, the results suggest the need for policy sequencing in sustainability transitions. The finding that regulatory divergence imposes a larger carbon-intensity penalty in renewable-intensive exporters implies that green energy investments may not fully translate into lower trade-related carbon intensity if digital governance remains fragmented. Governments seeking to decarbonize digitally mediated economic activity should treat digital regulatory alignment and interoperability as a complementary condition for realizing the full environmental returns to renewable energy capacity.

Third, environmental governance may need to be updated to target digital carbon more directly. The limited moderating role of EPS suggests that conventional policy instruments may not adequately address emissions that arise through electricity demand in data centers and network infrastructures. Policymakers could consider digital-economy-specific tools such as standardized disclosure of data-center energy sourcing, interoperable accounting for digitally embodied emissions, and procurement or reporting rules that incentivize low-carbon digital infrastructure.

For multinational firms, the estimated magnitude of the DSTRI-gap association implies that regulatory divergence is not merely a compliance nuisance but a potential environmental performance risk. Firms may benefit from designing modular digital architectures with adaptable compliance layers that reduce the need for duplicative infrastructures across jurisdictions. Incorporating regulatory-distance exposure into sustainability ROI and market-entry decisions may also help firms anticipate and manage emissions-efficiency penalties in digitally distant markets.

5.4. Limitations and Future Directions

This study has several limitations that motivate future work. First, our country-level measures of regulation, renewable electricity share, and policy stringency may mask within-country heterogeneity in enforcement, energy sourcing, and firm compliance strategies. Micro-level data (firm-, sector-, or facility-level) could clarify how compliance architectures and energy procurement decisions shape embodied emissions.

Second, we focus on formal regulatory divergence, but informal institutions and private governance (e.g., voluntary standards, industry codes, transnational norms) may also shape interoperability and environmental performance. Future studies could incorporate informal institutional distance and examine how private governance interacts with formal regulation.

Third, our models primarily identify average relationships. Future research could examine dynamic adjustment and non-linearities more explicitly, including threshold effects in regulatory divergence and adaptation costs, and how these evolve across stages of digitalization and energy transition.

Finally, while carbon intensity is a central sustainability outcome, digitalization may affect other environmental dimensions (e.g., material use, electronic waste, biodiversity impacts). Broadening outcome measurement would provide a more comprehensive assessment of the sustainability implications of digital services trade.

6. Conclusions

This study examines how regulatory divergence in digital services trade relates to the carbon efficiency of international trade. Integrating institutional distance theory with environmental economics, we argue that the environmental implications of digital trade depend not only on the scale of digitally delivered exchange but also on the degree of regulatory alignment that governs cross-border data and service interoperability.

Empirically, we find that wider regulatory gaps in digital services trade are systematically associated with increases in both the level of embodied emissions and export carbon intensity. Although digitally delivered trade can, in principle, reduce carbon intensity through technique-related efficiency gains, our results suggest that such gains are often offset by the inefficiencies linked to regulatory fragmentation. In this sense, the environmental consequences of digital trade are not automatic; they appear contingent on the extent to which trading partners’ digital regimes are interoperable.

We also document meaningful heterogeneity. The carbon-intensity penalty associated with regulatory divergence is, counterintuitively, stronger for exporters with higher renewable electricity shares, consistent with an opportunity-cost logic: regulatory misalignment may impose disproportionately large efficiency losses where the potential for low-carbon coordination and production is greatest. By contrast, we find limited evidence that importers’ environmental policy stringency moderates the relationship, pointing to a possible mismatch between traditional environmental policy instruments and the mechanisms through which digital trade governance shapes emissions efficiency.

Taken together, these findings identify digital regulatory alignment as a relevant—yet underexplored—determinant of trade-related emissions efficiency. While the analysis is observational and does not claim definitive causality, it provides robust evidence that fragmented digital governance is associated with higher carbon intensity. Aligning digital trade rules with sustainability objectives—through interoperability and coordinated governance—may therefore be important for realizing the decarbonization potential of digitally mediated globalization.