Does climate change respond to government digitization? Empirical evidence from greenhouse gas emissions Open Access

Amid ongoing social and economic development, the intensification of human activities, particularly the large-scale combustion of fossil fuels and alterations in land use, has led to a sustained increase in atmospheric greenhouse gas (GHG) concentrations. This accumulation has disrupted the earth’s radiative balance and become a primary driver of global climate change, manifesting in long-term temperature rise and increasingly erratic climate patterns (Bhatti et al., 2024). The consequences of climate change are multifaceted and profound, including the escalation of extreme weather events, rising sea levels and biodiversity loss. These climate-induced disruptions not only threaten natural ecosystems but also compromise food security, damage critical infrastructure and strain global supply chains, thereby exposing economies to heightened risks and volatility (Yan et al., 2023). Furthermore, governments are compelled to divert significant fiscal resources toward climate mitigation and adaptation efforts, which adds pressure to public finances. As such, addressing the root causes and impacts of climate change, most notably the persistent growth in GHG emissions, has become an urgent global challenge (Pavel et al., 2024).

As asserted by the United Nations Environment Programme, the rapid development of digital technologies plays a critical role in tackling global environmental sustainability (Wang et al., 2024). In response to the escalating climate crisis, digitization has been widely recognized as a transformative force in environmental governance (Bonsignore, 2024). It is often viewed as an important facilitator of green finance flows into environmentally responsible sectors, thereby enhancing environmental sustainability (Ye et al., 2024). A growing body of empirical literature supports the view that digitization contributes to reducing GHG emissions, a conclusion confirmed in studies on China (He et al., 2025; Zhao et al., 2023), OECD economies (Alvi et al., 2025), the European Union (Dimian et al., 2025; Kwilinski et al., 2023) and the G7 economies (Chen et al., 2023).

While the emission-mitigating role of digitization is widely acknowledged, it is essential to highlight the pivotal role that governments play in steering digitization. Acting as both regulators and public service providers, governments are instrumental in promoting digital initiatives and realizing their potential effects on GHG emissions. In this context, a growing body of research has shifted focus from broad digitization to the more specific domain of government digitization, and several recent studies explicitly emphasize this directional shift (Feng et al., 2024; Li et al., 2025; Wang et al., 2024; Yi, 2025; Zhang et al., 2025). Notably, these studies diverge in their conclusions regarding government digitization’s emission impact: most argue it curbs carbon emissions, whereas Li et al. (2025) emphasized carbon emissions from digital government operations and empirically revealed an inverted U-shaped effect.

A brief review of the existing literature reveals that most studies tend to overemphasize the environmental benefits of government digitization while neglecting the GHG emissions associated with its own operation. Although Li et al. (2025) provided a valuable supplement to this line of research, certain limitations in their research design could weaken the validity of their findings. First, Li et al. (2025) did not use a U-test to verify the existence of an inverted U-shaped relationship; instead, they inferred its presence solely based on the statistically significant negative coefficient of the squared term of government digitization. This approach is insufficiently rigorous, as convex and monotonic effects could also erroneously exhibit a U-shape with an extremum (Lind and Mehlum, 2010). Second, the digital government data released by the United Nations reflect the situation of the previous year rather than the current year. Li et al. (2025) failed to account for this temporal characteristic, and the resulting time mismatch could introduce bias into their findings. Moreover, while Li et al. (2025) discussed the emission-increase effect of government digitization at the theoretical level, they provided no direct empirical evidence to support the underlying mechanism. As such, this study empirically revisits the impact of government digitization on GHG emissions, considering both its positive and negative effects.

In addition to empirically examining the impact of government digitization on GHG emissions and uncovering mechanisms that have not been sufficiently explored in the existing literature, this paper contributes to the literature in several other ways. First, various economic and political variables are incorporated into the analytical framework to explore their moderating roles. Specifically, energy security risk could influence the logic of resource allocation associated with digital government initiatives by shaping regulatory priorities between environmental and energy considerations; policy uncertainty could alter the way government digitization affects emissions by disrupting policy coherence and market expectations; and ruling party ideology could affect the environmental consequences of government digitization by determining government priorities. Existing studies have yet to clarify how these three contextual factors shape the nexus between government digitization and GHG emissions, making it necessary to investigate their moderating effects to refine the theoretical framework. Moreover, the principle of diminishing marginal effect suggests potential heterogeneity in the impact of government digitization across different distributional stages, a critical dimension that has been largely overlooked in prior empirical investigations. To address this, this paper uses quantile regression analysis. Together, the moderation analysis and quantile regression provide a more nuanced understanding of how and under what conditions government digitization affects environmental outcomes, thereby offering more targeted and effective policy implications.

The possible marginal contributions of this study are as follows. First, unlike the prevailing view in the literature that emphasizes the emission-reducing effect of government digitization, this paper empirically confirms that government digitization can intensify GHG emissions. This finding challenges the unidirectional perception of its environmental impact, provides a new empirical perspective for understanding its complex role in GHG governance and offers evidence for reassessing the environmental costs and tradeoffs of government digitization. Second, by incorporating energy security risk, policy uncertainty and ruling party ideology into the analytical framework, this study addresses the insufficient attention to moderating factors in existing research and enriches the analytical dimensions of the intersection between digital governance and environmental economics. Finally, the heterogeneous effects of government digitization across different GHG emission levels revealed in this study not only provide new empirical testing of the applicability of the diminishing marginal effect theory in the field of environmental governance but also offer practical guidance for enhancing the precision of environmental policy design based on this finding.

Section 2 develops the research hypotheses. Section 3 describes the data and methodology used. Section 4 reports and interprets the empirical findings. Section 5 summarizes the main conclusions, outlines relevant policy implications and acknowledges the study’s limitations.

Government digitization is typically accompanied by the rapid expansion of information infrastructure, such as the large-scale deployment of data centers, cloud computing platforms and smart terminal devices (Li et al., 2025). The operation of these facilities requires substantial electricity consumption. Given that the energy mix remains heavily reliant on fossil fuels, the resulting increase in energy demand inevitably leads to higher GHG emissions (Ikevuje et al., 2024). As such, without a green development pathway, the process of government digitization will inevitably exert negative pressure on environmental sustainability. Based on this, we propose H1a:

Government digitization could influence GHG emissions by affecting total factor productivity (TFP). First, digital governance can reduce information asymmetry and administrative barriers, thereby optimizing the allocation efficiency of public resources and directly improving TFP in the public sector (Ding et al., 2025a). Specifically, cross-departmental data sharing on digital government platforms helps streamline redundant procedures in policy implementation, thereby reducing energy consumption and GHG emissions of the administrative system itself. At the same time, by lowering institutional transaction costs for enterprises (Mohamed et al., 2023), it indirectly enhances the production efficiency of market entities and promotes the reallocation of production factors from high-emission industries to low-emission sectors. In addition, digital-enabled collaborative governance can amplify the synergy between TFP improvement and emission reduction. Specifically, government-led cross-regional and cross-sector digital platforms can integrate data related to low-emission industrial transitions, promoting the efficient flow and allocation of production factors within circular economy systems. On one hand, this efficient allocation directly enhances resource use efficiency, thereby driving the improvement of TFP; on the other hand, precise matching and recycling of production factors help reduce logistical redundancy and avoidable emissions in production processes. However, it is worth noting that the economic scale expansion brought by TFP improvement could also lead to emission rebounds. As such, the effect of government digitization on GHG emissions via TFP remains ambiguous. Based on this, we propose H1b:

Business digitization could serve as a transmission channel through which government digitization affects GHG emissions. On one hand, government actions have a guiding role. The digitization of high-frequency public services creates a strong demonstration effect that stimulates businesses to undertake similar transformations. Within industrial clusters, this imitation effect generates a diffusion mechanism, forming a transmission chain from governments to leading businesses and eventually to small and medium-sized businesses, thereby promoting widespread adoption of digitization across the business sector. On the other hand, governments can reduce businesses’ costs of data collection and processing by establishing standardized data interfaces and shared platforms. In addition, the provision of modular digital tools can shorten the business digitization cycle and lower initial investment costs, further enhancing businesses’ willingness to pursue digitization. The emission effects of business digitization are twofold. On the mitigation side, business digitization can reduce emissions by improving energy efficiency, enhancing financial performance and fostering green innovation (Gao et al., 2023). On the amplification side, the energy consumption of digital devices is non-negligible. Moreover, increasing investment in digitization by businesses can crowd out input in environmentally friendly production and alter production modes to accommodate large-scale output, thereby increasing emissions (Yang et al., 2024). In sum, government digitization can promote business digitization, while the emission impacts of business digitization exhibit a dual nature. Based on this, we propose H1c:

The analysis of H1a through H1c reveals that government digitization could affect GHG emissions through multiple channels. Among these pathways, only the energy consumption effect identified in H1a has a clearly defined direction of influence, while the impacts of the other two mechanisms remain theoretically ambiguous. Accordingly, we propose an overarching H1:

Energy security risk could influence the environmental effects of digital government initiatives by shaping resource allocation. In countries with lower energy security risk, stable energy supply alleviates these governments’ pressure to ensure energy availability, thereby allowing digital regulatory efforts to focus more on improving environmental quality. Governments in such contexts are more likely to prioritize digital resources toward green governance tools, fostering a virtuous interaction between digital regulation and emission reduction incentives. This, in turn, facilitates a synergistic alignment between digital government development and climate mitigation goal. In contrast, in countries facing higher energy security risk, regulatory priorities are often skewed toward ensuring energy accessibility (Cong et al., 2025). Governments in these contexts tend to concentrate digital resources on strengthening the capacity for energy supply assurance. Such a resource allocation preference could crowd out investment in digital infrastructure needed for environmental regulation, leading to underdevelopment in this area. Consequently, the effectiveness of environmental enforcement is weakened, and the effect of government digitization in increasing GHG emissions tends to be amplified. Based on this, we propose H2:

In countries with low economic policy uncertainty, policy directions are clearer and institutional arrangements tend to be more coherent. Under such conditions, market participants are more likely to make long-term green transition decisions based on stable expectations. This institutional environment provides critical support for the effective application of government digitization in the environmental domain. On one hand, coherent policy frameworks help clarify environmental regulatory targets, enabling governments to leverage digital technologies for real-time emission monitoring and data sharing, thereby enhancing the precision and transparency of environmental supervision. On the other hand, stable expectations make firms more willing to comply with digital regulatory initiatives, which lowers the resistance to their implementation. In short, in such countries, government digitization is more readily translated into concrete environmental governance capacity, thus significantly curbing GHG emissions. By contrast, in countries with high economic policy uncertainty, policy coherence is often disrupted, leading to unstable market expectations (Fu et al., 2022). To hedge against policy volatility, firms tend to shorten investment horizons and focus on short-term returns, thereby weakening their motivation for green transition and environmental compliance. At the same time, frequent policy shifts and inefficiencies in regulatory coordination hinder the stable integration of digital tools into the environmental regulatory framework. As a result, in such countries, government digitization is less likely to evolve into effective environmental governance and its emission reduction effect tends to be substantially diminished. Based on this, we propose H3:

A government’s attention to different sectors can be influenced by the ideology of the ruling party (Wen et al., 2022). As such, ruling party ideology could affect government’s attention to the environmental sector, thereby moderating the relationship between government digitization and GHG emissions. Compared with right-wing parties, left-wing parties are more willing to prioritize environmental protection (Neumayer, 2004). This is further supported by Matuszczak et al. (2020), who found that members and supporters of left-wing parties are mostly environmentalists. Driven by such ideological preferences, governments led by left-wing parties are more likely to prioritize the application of digital technologies in the field of environmental governance. Accordingly, government digitization tends to play a substantive role in environmental governance, thereby effectively curbing GHG emissions. In contrast, right-wing parties place more emphasis on market freedom and capital efficiency, and their policy priorities tend to tilt toward industrial expansion and deregulation of enterprises (Jäger, 2017). Under this policy orientation, government digitization could be used to simplify approval procedures for resource-heavy sectors and reduce environmental compliance costs, turning digital tools into catalysts for the development of energy-intensive industries, and ultimately exacerbating GHG emissions. By comparison, centrist parties tend to adopt a more pragmatic stance, seeking a balance between environmental objectives and economic concerns. In this context, government digitization is less likely to be systematically directed toward strengthening environmental governance, as is often the case under left-leaning governments, nor is it inclined to foster resource-intensive industries in the manner of right-leaning governments. Instead, government digitization’s impact on GHG emissions could be more neutral, reflecting the centrist pursuit of compromise between environmental protection and economic growth. Based on this, we propose H4:

The impact of government digitization on GHG emissions could vary significantly across countries with different emission levels. In low-emission countries, the environmental role of government digitization tends to be relatively limited, primarily reflected in two types of contexts. First, in countries that have not yet completed industrialization, weak economic foundations and a lack of capital constrain the development of digital infrastructure and human capital, thereby restricting the effectiveness of government digitization in environmental governance. Second, in countries that have already undergone green transitions and possess a high level of government digitization, although their governance capacity is relatively strong, the small pollution base and limited room for further improvement mean that government digitization plays a more auxiliary role, with low marginal effects.

In countries with moderate emission levels, the influence of government digitization could be more pronounced. These countries are typically undergoing rapid industrialization and urbanization, characterized by dynamic economic activity and strong energy demand. Government digitization can enhance administrative efficiency and resource allocation, thereby improving enterprise productivity (Ding et al., 2025b). This improvement in productivity, in turn, intensifies energy consumption and GHG emissions.

By contrast, in high-emission countries, the emission-promoting effect of government digitization tends to weaken. On one hand, the expansion of emission-intensive industries faces diminishing marginal returns, reducing incentives for governments to further support their growth. On the other hand, due to their prolonged industrialization processes, these countries have developed relatively mature digital governance systems, possessing a solid technical foundation for digital regulation while also facing greater pressure in environmental governance. As a result, government digitization is more likely to be leveraged to enhance the enforcement of environmental policies and strengthen monitoring of emission sources, thereby mitigating its contribution to emission increases. Based on this, we propose H5:

Based on data availability, this study uses data from 130 countries ^[1] over the period 2002–2021 to empirically examine the impact of government digitization on GHG emissions. The dependent variable, total GHG emissions (TGHGE), is obtained from the World Development Indicators (WDI) database. Government digitization (GD), the core explanatory variable, refers to the structured integration of advanced information and communication technologies into public sector operations, with the goal of enhancing administrative efficiency and improving the accessibility and quality of digital public services (Verma and Dawar, 2019). To quantify GD, we adopt the online service index published by the United Nations, which systematically assesses the extent and performance of digital government services (Castro and Lopes, 2022) ^[2]. Following Ding et al. (2023), this research introduces several control variables, including gross domestic product (GDP) and its squared term (GDP_squared), total population (TP), urbanization rate (UR), the share of value added by the service industry in GDP (SI), net inflows of foreign direct investment as a share of GDP (FDI) and the share of fossil fuel energy consumption in total energy use (FFEC). The control variable data are obtained from the WDI database. Table 1 summarizes the descriptive statistics for all the above variables. Notably, the variables named TGHGE, GDP and TP are transformed using natural logarithms in the regression analysis to mitigate possible heteroscedasticity. Accordingly, the descriptive statistics for these three variables are based on their natural logarithmic values.

To test H1, we will use a fixed-effect model for the benchmark regression, as outlined in equation (1). In this equation, Z denotes the control variables, while fixed effects for countries and years are captured by $μi$ and $νt$⁠, respectively. The parameters of interest are $α0$⁠, $α1$ and β, with $εit$ representing the error term.

To ensure the reliability of the benchmark finding, we will conduct robustness tests. First, a placebo test will be conducted. Following the approach of Cornaggia and Li (2019), we remove the original GD values and randomly reallocate these values across the sample before re-estimating equation (1).

To alleviate potential endogeneity, we will use system GMM estimation and two-stage least squares (2SLS) estimation. To conduct a system GMM estimation, as shown in equation (2), we added the lag term of the dependent variable (TGHGE_i,t–1) as an instrumental variable into equation (1). As for 2SLS estimation, following Ding et al. (2025a), GD is treated as an endogenous variable and the average GD of other countries in the same region ^[3], and year is selected as the instrumental variable (IV):

To conduct mechanism testing, we follow the empirical strategy used by Wen et al. (2021), which entails incorporating two additional explanatory variables into equation (1). Specifically, to investigate the channel proposed in H1a, namely, that government digitization influences GHG emissions via electric power consumption (EPC), we introduce both EPC and its interaction term with GD (GD × EPC) into the regression model, resulting in equation (3). The EPC data are from the WDI database.

To test the mechanisms proposed in H1b and H1c, we replace EPC in equation (3) with TFP (TFP) and business digitization (BD), respectively, as shown in equations (4) and (5). The TFP data are sourced from the Penn World Table. As for BD, we select five indicators from the United Nations Conference on Trade and Development Digital Economy Database: proportion of businesses placing orders over the Internet (PBPOI), proportion of businesses receiving orders over the Internet (PBROI), proportion of businesses using computers (PBUC), proportion of businesses using the Internet (PBUI) and proportion of businesses with a web presence (PBWP), as its proxies. Principal component analysis (PCA) is then applied to these five indicators to construct the BD index ^[4]. Although this approach could not capture the full complexity of business digitization, it remains appropriate given the constraints of data availability. Specifically, PBUC and PBUI represent the digital infrastructure dimension, reflecting firms’ basic capacity to access digital tools and serving as a prerequisite for subsequent digital activities. PBWP captures firms’ ability to establish an online presence and display information, which is a core feature of digital existence. PBPOI and PBROI focus on the digitization of business processes, directly reflecting the extent to which digital technologies have penetrated transaction activities:

To examine H2 and H3, which pertain to the moderating roles of energy security risk (ESR) and economic policy uncertainty (EPU), we adopt the approach used by Jadiyappa et al. (2021). Concretely, the full sample is split into four subgroups based on the median values of the two moderators ^[5], and equation (1) is estimated independently for each subgroup. Following Cong et al. (2025), we used the national energy security risk index from the Global Energy Institute as ESR’s proxy. Notably, a higher score on this index reflects a weaker state of energy security. In line with Ryu and Yu (2022), the economic policy uncertainty index is selected as EPU’s proxy. Because the index is reported at a monthly frequency, we take its annual average to more accurately capture the overall policy uncertainty. As EPU data are only available for 22 economies, the scope of this analysis is accordingly limited.

To examine H4, which concerns the moderating effect of ruling party ideology (RPI), the full sample is divided into left-wing, right-wing and centrist governments, and equation (1) is estimated separately for each subgroup. Information on RPI can be obtained from the Database of Political Institutions.

To test H5, which posits that the effect of government digitization on GHG emissions varies across emission levels, we apply quantile regression. This method provides a more refined understanding of distributional heterogeneity that cannot be captured by mean-based estimates (Islam et al., 2025). We estimate equation (1) at the 0.1, 0.25, 0.5, 0.75 and 0.9 quantiles of the emission distribution.

Column 1 of Table 2 reports the benchmark regression results. The coefficient of GD is statistically significant at the 1% level, indicating that government digitization has a significant impact on GHG emissions, thereby confirming H1^[6]. Moreover, the positive sign of the coefficient (0.023) suggests that government digitization exacerbates GHG emissions. This finding challenges the conventional view of government digitization as a driver of green transformation and, to some extent, corroborates the dominant role of the electric power consumption effect proposed in H1a. Specifically, in the absence of a well-developed green development pathway, the expansion of information infrastructure driven by government digitization could increase electric power consumption and thus lead to higher GHG emissions. Although government digitization could also reduce emissions indirectly by enhancing TFP (H1b) and promoting business digitization (H1c), as discussed in subsection 2.1, both channels could theoretically also increase emissions. Even if the mechanisms proposed in H1b and H1c indeed contribute to emission reduction in practice, the significantly positive coefficient of GD suggests that their potential mitigating effects are not sufficient to offset the emission increases driven by higher electric power consumption in this context. Overall, this result highlights the complexity of the environmental consequences of government digitization and underscores the need to integrate green development strategies into digital governance reform to ensure alignment between technological advancement and environmental sustainability.

Regarding the control variables, two results warrant particular attention. First, the coefficient of GDP_squared is positive and statistically significant, which contradicts the inverted U-shaped relationship predicted by the Environmental Kuznets Curve (EKC). A similar empirical pattern was reported by Wang and Kim (2024) in a US context, who attribute this result to increases in emissions from the transportation sector, strong electricity demand during cold winter seasons, the reversal of eco-friendly energy policies and the reshoring of manufacturing activities. Second, although conventional wisdom suggests that urbanization tends to increase environmental pressure through greater transportation demand and energy consumption (Zhou et al., 2019), UR is negatively signed and statistically significant, indicating that urbanization reduces GHG emissions. This counterintuitive finding could be explained by the efficiency-enhancing effects of urbanization, such as intensified infrastructure use and improved resource allocation, both of which contribute to lower emission intensity (Pata, 2018).

The results of the placebo test are reported in Column 2 of Table 2. The coefficient of GD is statistically insignificant, ruling out the possibility that the benchmark finding is a placebo. Moreover, after replacing the estimation method with system GMM (Column 3, Table 2) and 2SLS (Table 3), GD consistently obtains a statistically significant positive coefficient. These results collectively confirm the robustness of the benchmark finding.

Table 4 presents the mechanism analysis results. The EPC in Column 1, the TFP in Column 3, and the BD in Column 5 all yielded statistically significant positive coefficients, suggesting that these factors contribute to higher GHG emissions. Furthermore, the interaction terms GD × EPC in Column 2, GD × TFP in Column 4 and GD × BD in Column 6 are also significantly positive, indicating that government digitization increases GHG emissions by raising electric power consumption, enhancing TFP and promoting business digitization, thereby supporting H1a, H1b and H1c^[7].

Further discussion is warranted regarding H1b and H1c. Both TFP improvement and business digitization can theoretically have dual effects on GHG emissions: on one hand, they could lower emissions by enhancing efficiency and fostering green innovation; on the other hand, they could raise emissions through the energy demands of digital infrastructure and the expansion of production capacity. Our findings show that the latter effect currently dominates, as both higher TFP and deeper business digitization are associated with increased GHG emissions. This suggests that, at the present stage, advances in productivity and digital adoption are more closely tied to output expansion than to emission reduction, highlighting the need for policies that guide TFP growth and digitization toward greener development pathways. This policy orientation is consistent with the insights from Riaz et al. (2025), who warned that without aligning digital advancement with green energy systems, the association between digitization and output expansion will overshadow emission reduction goals. Similarly, Ridwan et al. (2025) emphasized that circular economy practices are essential to counteract the emission-intensive effects of productivity growth, making the integration of circular principles into industrial systems a crucial step for achieving low-carbon development.

However, the robustness and generalizability of the finding on the business digitization mechanism encounter two limitations. First, due to limitations in BD data availability, the test of H1c relies on only 110 observations, with missing data undermining statistical reliability. Second, the sample composition further constrains generalizability: most of these observations are from high-income economies, where mature digital infrastructure could amplify the emission effects of BD, thereby biasing the overall regression results. Hence, the applicability of this conclusion to low-income economies could be substantially weakened.

Table 5 presents the estimation results for seven subsamples. In countries characterized by higher energy security risk (Column 2), higher economic policy uncertainty (Column 4) and right-wing governments (Column 6), GD exhibits a significantly positive coefficient at least at the 5% level, indicating that government digitization intensifies GHG emissions. In contrast, in countries with lower energy security risk (Column 1), lower economic policy uncertainty (Column 3) and left-wing governments (Column 5), government digitization significantly reduces GHG emissions. In addition, Column 7 reports the results for centrist governments. The coefficient of GD is positive and statistically significant at the 10% level (0.015), but its magnitude is smaller than that of right-wing governments (0.039, Column 6). This aligns with the theoretical expectation in H4, with centrist parties prioritizing a balance between environmental protection and economic concerns. Consequently, government digitization under centrist governance neither amplifies GHG emissions as strongly as under right-wing rule nor curbs GHG emissions as under left-wing rule, resulting in a weaker emission-enhancing effect. These findings confirm the moderating roles of energy security risk ^[8], economic policy uncertainty and ruling party ideology, thus supporting H2, H3 and H4.

Table 6 reports the quantile regression results. At the 0.1 quantile, the coefficient of GD is statistically insignificant. In contrast, at the remaining four quantiles, the coefficients are positive and statistically significant, with their magnitudes gradually increasing across higher quantiles. This suggests that in countries with lower levels of GHG emissions, the impact of government digitization on emissions is negligible; however, as emission levels rise, the positive effect of government digitization on GHG emissions becomes increasingly pronounced. This finding is inconsistent with the expectation of H5, which anticipated a weaker positive or even negative relationship in high-emission countries.

Nonetheless, this deviation from the hypothesis could be explainable. H5 rests on the assumption that high-emission countries have largely completed the phase in which government digitization enhances the efficiency of traditional industries and are now reallocating digital resources toward green governance. In practice, however, this assumption faces two important challenges. First, in most high-emission countries, government digitization is still in its developmental phase (Profiroiu et al., 2024); the empowerment it provides to improve the efficiency of traditional industries has not yet reached the stage of diminishing marginal returns, meaning that ongoing digitization efforts can continue to stimulate industrial capacity expansion. Second, compared to the pressure for economic growth, environmental constraints in these countries tend to be relatively weak (Ding et al., 2023), making it difficult for digital resources to shift toward green governance in the short term. Crucially, the timing of diminishing marginal effects varies substantially across countries, development stages and research domains, making its specific occurrence empirically uncertain. In the context of this study, we have not found clear evidence that the impact of government digitization on GHG emissions has begun to exhibit diminishing marginal effects.

The empirical results of this study present notable deviations from the mainstream literature and offer meaningful extensions to the current understanding of the environmental consequences of government digitization. While a large body of existing research tends to emphasize the emission-reducing potential of government digitization (e.g. Feng et al., 2024; Wang et al., 2024; Yi, 2025; Zhang et al., 2025), our findings suggest that government digitization can, in fact, intensify GHG emissions under certain conditions. This challenges the prevailing belief that digital governance is inherently aligned with environmental sustainability and calls for a more nuanced assessment of its environmental implications. Moreover, our results diverge from those of Li et al. (2025), who argued for an inverted U-shaped relationship between government digitization and carbon emissions. Although their study highlights the dual environmental effects of digital governance, it lacks rigorous empirical testing for the turning point and does not account for the temporal lag in GD data. In contrast, our quantile regression analysis does not support the existence of a U-shaped pattern; rather, it shows that the emission-promoting effect of government digitization becomes more pronounced at higher levels of emissions, implying that digitization in high-emission countries has yet to be redirected toward green governance. This finding is further reinforced by our group-testing analysis, which reveals that energy security risk, policy uncertainty and ruling party ideology, substantially moderate the impact of government digitization on GHG emissions. These results help explain the heterogeneity in previous empirical findings and underscore the importance of accounting for institutional and contextual factors. In sum, this study enriches the existing literature by shifting the focus from a predominantly optimistic view to a more balanced understanding of the complex and conditional effects of government digitization on climate outcomes.

Based on panel data from 130 countries over the period 2002–2021, this study conducts a series of empirical analyses and draws the following conclusions:

• Government digitization significantly increases GHG emissions; more specifically, it contributes to higher emissions by increasing electric power consumption, enhancing TFP and promoting business digitization.

• The impact of government digitization on GHG emissions is moderated by energy security risk, economic policy uncertainty and ruling party ideology. Specifically, in countries facing severe energy security risk, experiencing elevated policy uncertainty, or governed by right-wing parties, government digitization significantly increases GHG emissions; in contrast, under conditions of low energy security risk, stable policy environment or left-wing governance, it significantly reduces emissions; for countries governed by centrist parties, government digitization also exerts an emission-increasing effect, though the effect is weaker than under right-wing governments.

• The effect of government digitization on GHG emissions varies by countries’ GHG emission levels, with the effect being insignificant in low-emission countries but increasingly contributing to higher emissions as emission levels rise.

The empirical findings of this study provide important insights for policy formulation aimed at aligning government digitization with climate mitigation goals. First, given the demonstrated positive association between government digitization and GHG emissions, as well as the identification of electric power consumption as a key transmission mechanism, it is imperative to improve the energy structure underpinning digital infrastructure. Governments should embed green standards throughout the entire lifecycle of government digitization, particularly by prioritizing the use of clean energy to power data centers and communication networks, so as to reduce the GHG footprint and mitigate the environmental consequences of digitization. Second, the finding that government digitization increases GHG emissions through business digitization underscores the importance of policy efforts that steer private-sector digitization in an environmentally responsible direction. Measures such as green tax incentives could encourage the adoption of low-emission digital technologies and discourage digital investments that lead to excessive GHG emissions. Third, the results regarding energy security risk, economic policy uncertainty and ruling party ideology highlight the critical role of national conditions in shaping the environmental consequences of government digitization. In countries facing energy supply instability or frequent policy shifts, governments should strengthen the resilience of their energy systems and enhance policy consistency to prevent digital government initiatives from exacerbating GHG emissions. In addition, fostering political consensus on the environmental objectives of digital government initiatives, regardless of political ideology, could help ensure long-term commitment to environmental sustainability in the context of digital government. Finally, because the emission effects of government digitization vary across countries with different emission levels, digital governance reforms should adopt a differentiated approach. In high-emission countries, efforts should focus not only on enhancing digital tools for environmental monitoring and enforcement, but also on applying digital technologies to improve energy efficiency and reduce overall GHG emissions in emission-intensive sectors. In contrast, in countries with lower emission levels and relatively underdeveloped digital infrastructure, priority should be given to building environmentally sustainable digital capacity from the outset, ensuring that the expansion of digital government does not lead to unintended environmental burdens.

Despite the rigorous efforts devoted to ensuring the accuracy of this study, certain unresolved limitations persist. First, due to data availability constraints, this study used online service index as GD’s proxy, which primarily captures digitization of public services rather than the broader digital transformation of the public sector. If more comprehensive indicators become available in the future, researchers could extend the analysis to encompass wider dimensions of digital governance and thereby achieve a more comprehensive assessment of government digitalization’s environmental consequences. Second, although we used 2SLS estimation to mitigate potential endogeneity, the validity of the exclusion restriction remains debatable because regional adoption initiatives, such as the European Green Deal, could be correlated with unobserved shocks that also influence GHG emissions. Future research could incorporate falsification tests to further strengthen causal identification.

Albania, Algeria, Angola, Argentina, Armenia, Australia, Austria, Azerbaijan, Bahrain, Bangladesh, Belarus, Belgium, Belize, Benin, Bhutan, Bolivia, Bosnia and Herzegovina, Botswana, Brazil, Brunei Darussalam, Bulgaria, Burkina Faso, Cabo Verde, Cambodia, Cameroon, Canada, Chad, Chile, Colombia, Comoros, Costa Rica, Croatia, Cyprus, Denmark, Dominican Republic, Ecuador, El Salvador, Equatorial Guinea, Estonia, Eswatini, Ethiopia, Fiji, Finland, France, Gabon, Germany, Ghana, Greece, Guatemala, Guinea-Bissau, Guyana, Haiti, Honduras, Hungary, Iceland, India, Indonesia, Ireland, Israel, Italy, Jamaica, Japan, Jordan, Kazakhstan, Kenya, Kuwait, Latvia, Lebanon, Lesotho, Lithuania, Luxembourg, Madagascar, Malaysia, Malta, Mauritius, Mexico, Mongolia, Morocco, Mozambique, Namibia, Nepal, the Netherlands, New Zealand, Nicaragua, Niger, North Macedonia, Norway, Pakistan, Panama, Paraguay, Peru, Philippines, Poland, Portugal, Romania, Russian Federation, Rwanda, Samoa, Saudi Arabia, Senegal, Seychelles, Singapore, Slovenia, Solomon Islands, South Africa, Spain, Sri Lanka, Sudan, Suriname, Sweden, Switzerland, Syrian Arab Republic, Tajikistan, Thailand, Timor-Leste, Togo, Tonga, Tunisia, Turkmenistan, Uganda, Ukraine, United Arab Emirates, UK, USA, Uruguay, Uzbekistan, Vanuatu, Viet Nam, Zambia, Zimbabwe.