Catalyzing sustainable development through digitalization: the nexus between government digital transition and green total factor productivity

Highlights

• Digital government significantly improves urban green total factor productivity (GTFP).
• GTFP is driven by technological progress rather than technical efficiency.
• Government information transparency and technological innovation are key mechanisms.
• Factor marketization and digital economy amplify the impact of digital government.

1 Introduction

Under the mounting pressures from global climate change and resource depletion, the transition to an ecologically sustainable and low-carbon economic paradigm has become an imperative strategy for nations worldwide (Jänicke, 2012; Khan et al., 2022). Traditional economic growth metrics often overlook environmental externalities, necessitating a more comprehensive benchmark that synchronizes economic expansion with environmental stewardship (Wang et al., 2022). In this context, Green Total Factor Productivity (GTFP) has emerged as a pivotal indicator, as it integrates both conventional inputs and undesirable environmental outputs, such as pollution emissions, into a unified analytical framework (Chen et al., 2023). Consequently, accelerating GTFP growth is widely recognized as a cornerstone for achieving high-quality, sustainable development, as it offers a viable pathway to decoupling economic progress from environmental degradation (Wen et al., 2025; Zhong et al., 2022).

While market mechanisms and spontaneous technological innovation are critical drivers of ecological performance, institutional frameworks and governance efficiency exert an equally indispensable influence on environmental sustainability (Dhaoui, 2022; Khan et al., 2021). As a key component of the modernizing national governance, the digital transition in public administration—commonly called digital government or e-government—has profoundly transformed the operational processes within the public sector (Janowski, 2015; Wilson and Mergel, 2022). Although modern information technologies are reshaping public administration, a critical distinction exists between “e-government” and “digital government.” The former primarily involves digitizing existing administrative interfaces (e.g., online portals), whereas the latter requires the systemic integration of back-end data infrastructures and the restructuring of cross-departmental governance (Castelnovo and Sorrentino, 2018). In this study, we frame the 2018 pilot not as a simple continuation of earlier e-government efforts, but as the formal advent of digital government in China. This classification rests on the policy’s mandate for unified platforms and real-time data sharing, which fundamentally moved beyond isolated departmental websites (Medaglia et al., 2024; Yang et al., 2024). Accordingly, we use the term “digital government” throughout this paper to describe this emerging governance paradigm.

Theoretically, the deployment of digital government architectures acts as a systemic catalyst for regional environmental performance through multiple administrative channels. First, digital governance effectively dismantles institutional information silos, fostering cross-departmental data sharing and significantly reducing information asymmetry between regulatory authorities and industrial actors (Wang and Guo, 2024; Zuo and Yu, 2026). This heightened transparency enables real-time, precision-targeted environmental monitoring, thereby suppressing opportunistic polluting behaviors. Second, the streamlining of administrative procedures via digital platforms lowers both transaction costs and institutional friction for enterprises. By mitigating compliance burdens, digital government liberates corporate resources, allowing firms to reallocate capital toward green technological innovation and ecological upgrades (Martins and Veiga, 2022; Zhang et al., 2025; Zhao et al., 2023).

Despite the growing body of research on the determinants of green growth, several empirical gaps persist within the extant literature. Prior studies have extensively explored the ecological ramifications of factors such as green financial instruments, industrial robot adoption, and the broader digital economy (Du et al., 2024; Lyu et al., 2024; Qian et al., 2024; Wang R. et al., 2025; Wen et al., 2025; Yang and Liu, 2024; Yue et al., 2024). However, distinct from market-driven digital transformation, digital government represents a top-down institutional restructuring that centers on public governance and regulatory refinement. Systematic research evaluating the direct nexus between formal digital government initiatives and urban GTFP remains relatively scarce. Specifically, whether the National E-Government Comprehensive Pilot Program implemented in China has delivered environmental dividends that foster sustainable macro-economic development require comprehensive empirical validation.

To fill these gaps, this paper exploits China’s 2018 National E-Government Comprehensive Pilot Program as a quasi-natural experiment to identify the causal effect of institutional digitalization on environmental performance. Utilizing a balanced panel dataset encompassing 285 prefecture-level cities from 2013 to 2021, we employ a rigorous time-varying difference-in-differences (DID) framework. The empirical findings reveal that the execution of the e-government pilot substantially propels urban GTFP. Further analysis demonstrates that this positive shift is predominantly powered by frontier technological progress rather than improvements in technical efficiency. These baseline findings withstand a battery of robustness checks, including parallel trends validations and placebo tests. Additionally, mechanism analyses indicate that digital governance elevates GTFP by expanding administrative transparency and accelerating regional innovative capabilities, and that these positive effects are moderated by factor marketization and the maturity of the local digital economy.

The marginal contributions of this study are threefold. First, it enriches the theoretical boundaries of environmental economics by shifting the analytical lens from market-led digital economic phenomena to formal, institutional digital governance frameworks, thereby establishing a direct link between public administration modernization and green development (Chen and Li, 2025; Ding et al., 2024; Jiang, 2025). Second, by focusing on the 2018 National E-Government Comprehensive Pilot Program through a multi-period DID methodology, this paper addresses potential endogeneity concerns, providing highly reliable causal evidence regarding the environmental efficacy of digital policy interventions. Third, this research opens the “black box” of the digital government-GTFP nexus by identifying the mediating roles of administrative informational transparency and regional innovation, as well as the boundary conditions factor market distortions and digital infrastructure readiness, thereby offering nuanced policy insights for developing economies striving for sustainable transitions.

The rest of this paper proceeds as follows. Section 2 outlines the policy background of the e-government pilot. Section 3 develops the theoretical framework and formulates related hypotheses. Section 4 elaborates on the research design, including data, variables, and methodology. Section 5 reports and interprets the empirical results, and Section 6 concludes.

2 Background

The process of China’s administrative reform and its alignment with global governance modernization have been significantly catalyzed by the acceleration of government digitalization. Historically, the adoption of successive generations of information technologies from computers and the Internet to artificial intelligence, has played a pivotal role in reshaping the Chinese governance framework and its operational structures. However, the trajectory of this digital metamorphosis has been a protracted and iterative process. Throughout this evolution, administrative documents and academic discourses have interchangeably utilized the concepts of “e-government” and “digital government” to characterize the digital transition within the public sector. The fluctuating usage of these terms underscores the deepening comprehension of digital governance dynamics by both state agencies and the academic community (Gil-Garcia et al., 2018; Manoharan and Ingrams, 2018; Zou et al., 2023).

The 2018 National E-Government Comprehensive Pilot in China (henceforth the e-government pilot) represents a watershed moment in this institutional transformation, spearheaded by the Cyberspace Administration of China, the National Development and Reform Commission, and other ministries. Approximately 71 cities across eight province-level regions (including Beijing, Shanghai, Jiangsu, Zhejiang, Fujian, Guangdong, Shaanxi, and Ningxia) were designated as leading pilots. The initiative aims to rectify long-standing administrative inefficiencies, such as fragmented management, system redundancy, weak cross-departmental coordination, and inadequate public service delivery. To achieve this, the policy mandates five key tasks: creating a unified e-government coordination mechanism; centralizing infrastructure; enabling on-demand data sharing through exchange platforms; promoting “Internet Plus Government Services” via integrated portals; and standardizing electronic documentation for priority areas like approval and licensing. The pilot aimed to achieve stronger inter-departmental coordination, responsive on-demand data sharing, and markedly more convenient online services.

In contrast to isolated digital administrative reforms, the e-government pilot is characterized by its holistic approach to institutional restructuring. It reconfigures government-market interactions through systematic service innovation rather than through the mere digitization of paper-based processes. At the local level, these mandates went beyond the simple development of government websites, requiring a fundamental re-engineering of administrative workflows. In Shanghai, the “One-stop Online Government Services” initiative set up a municipal big-data center and a unified online portal that integrated electronic certificates, licenses, and regulatory data; this integration markedly streamlined approval processes for construction permits and other administrative matters. Similarly, in Zhejiang, an integrated government service platform enabled “single-visit” processing across a wide spectrum of administrative items. More broadly, cities in pilot provinces carried out the reform through provincial-level unified data platforms and municipal government-service portals: previously dispersed records—covering enterprise registration, project approvals, environmental permits, credit information, and compliance—were consolidated into unified databases. Approval processes, meanwhile, shifted from serial, paper-based, siloed procedures to parallel, data-driven operations, which enhanced the transparency and accessibility of government information for regulators and the public.

3 Hypotheses

3.1 Direct effect of digital government on GTFP

The comprehensive e-government pilot program serves as a critical institutional instrument for modernizing China’s governance architecture. By championing the sharing of government information resources and fostering “Internet Plus Government Services,” this initiative fundamentally reconfigures the complex interplay between the state, the market, and civil society through technological empowerment and digital enfranchisement (Yuan et al., 2023). We contend that these mandates act on both margins of urban production: they curb undesirable environmental outputs by transforming environmental regulation, while simultaneously expanding desirable outputs by easing institutional frictions and improving allocative efficiency. This dual mechanism is what distinguishes the policy’s impact on green total factor productivity (GTFP) from a conventional productivity effect: GTFP explicitly incorporates undesirable outputs, such as pollutants, alongside desirable economic outcomes, thereby offering a more precise gauge of sustainable development (Wang D. et al., 2025; Zhong et al., 2022).

The environmental governance channel operates through three interlocking dimensions. First, the pilot mandated cross-agency data sharing and infrastructure integration, consolidating government information resources into unified data platforms: enterprise registration, project approval, environmental permitting, credit records, energy consumption, and compliance data are now housed in a single system. This allows regulators to identify abnormal discharge patterns of sulfur dioxide and ammonia nitrogen in near real time. Such capability marks a departure from traditional Chinese environmental regulation, which has often been constrained by information asymmetry: with pollution sources dispersed, inspections reliant on manual sampling, and grassroots enforcement vulnerable to selective implementation and local protectionism. Digitally enabled monitoring moves regulatory oversight from sporadic, campaign-style inspections to standardized, verifiable, and continuous monitoring; it automatically detects violations and maintains auditable enforcement logs, which systematically raises the expected cost of opportunistic pollution (Wang and Guo, 2024).

Second, digital platforms open public access to environmental data, including discharge permits, violation penalties, and monitoring results. This transparency places polluting firms under continuous public scrutiny and markedly lowers the cost to citizens of identifying and reporting environmental violations. Meanwhile, reputational and financial sanctions from communities, creditors, and consumers complement formal enforcement, further narrowing the information asymmetry between the regulator and the market (Jun et al., 2014) and lending credibility to environmental regulation.

Third, the pilot moved online environmental impact assessment, pollution discharge permitting, and related approval procedures through integrated service portals. Streamlined approval workflows shorten the administrative delays faced by firms seeking to adopt cleaner production technologies, while the digital trail lets regulators scrutinize pollution-intensive projects more systematically. Together, stricter enforcement and streamlined administration alter the relative cost structure that firms face: more consistent enforcement raises the shadow price of emissions, whereas digitalized approvals and targeted incentives for green initiatives lower the sunk costs of clean technologies. These concurrent price adjustments push technical change toward cleaner production methods, speed the exit of energy-intensive and high-emission enterprises, and encourage the adoption of both end-of-pipe and process-level abatement technologies (Hao et al., 2024; Zhao et al., 2023). Through these three dimensions, digital governance curbs undesirable outputs rather than merely expanding desirable ones, thus improving directly the pollution dimension of the GTFP accounting framework.

The productivity channel complements the green channel by expanding desirable outputs. Digital government draws on modern technological platforms—including AI, big data, and the Internet—to integrate cross-level administrative structures and business workflows, thereby raising governance efficiency (Gong and Li, 2023). One-stop online platforms and shorter approval timelines lower institutional transaction costs, releasing corporate resources for productive and green investments (Cordella and Tempini, 2015; Martins and Veiga, 2022). Data-centered governance—built on data sharing, data-driven decision-making, and digital service provision—eases information asymmetry between regulators and market actors and redraws the boundary between state intervention and market-driven allocation (Zhou and Ge, 2025). With coordination costs lower and information more transparent, capital, labor, and intermediate inputs shift toward their most productive uses, and relative prices carry allocative signals toward more efficient resource utilization (Yu et al., 2025). These conventional efficiency gains enlarge the desirable-output component of GTFP and raise productivity more broadly.

Distinguishing the two channels is essential for understanding why digital government affects GTFP more than conventional TFP. The environmental governance channel directly targets the pollution dimension of GTFP. By reducing industrial SO₂ and ammonia nitrogen emissions, it improves GTFP through the undesirable-output term, whereas conventional TFP, which ignores pollution externalities, would register no such gain. By contrast, the productivity channel represents a generic efficiency effect that conventional productivity measures would also capture. The distinctive “green” effect of digital government thus rests primarily on the environmental governance mechanism, while the productivity mechanism provides a complementary but less specific contribution. Given these institutional, environmental, and economic mechanisms, we postulate hypothesis H1: Digital government prompts urban GTFP.

3.2 Mechanism analysis

In this section, we posit that digital government fosters urban GTFP via four pathways: government information transparency, technological innovation, factor marketization, and digital economy. These variables play distinct roles depending on their causal position relative to the e-government pilot. Mediating variables lie on the causal path between the pilot and outcome, acting as conduits for the policy impact. Moderating variables predate the intervention and constitute the environmental context that shapes the pilot’s effectiveness ¹ . In other words, mediators respond to the pilot, whereas moderators are taken as exogenous. Government transparency and technological innovation serve as mediators because they are direct outcomes of the pilot’s mandates. Conversely, factor marketization and the digital economy are structural conditions established independently of the pilot and are therefore treated as moderators.

3.2.1 Government information transparency

Distortions such as moral hazard and adverse selection, which fundamentally stem from information asymmetry, constitute substantial barriers to the optimal distribution of resources (Bhandari and Bhuyan, 2023). The origin of the transparency effect lies in the pilot’s specific mandates: consolidated data platforms, standardized electronic documentation, and mandatory disclosure of regulatory information convert transparency from a discretionary practice into a designed output of government operations. Enterprise registration, permitting, compliance, and energy-consumption records previously scattered across agencies are cataloged, integrated, and rendered retrievable by mandate. Transparency rises not because individual agencies choose to disclose more, but because the digital architecture leaves them no practical alternative.

This government digital transition bolsters administrative transparency to propagates to GTFP through three links. First, heightened administrative transparency deters rent-seeking activities and related corrupt practices and directly elevates bureaucratic execution capabilities (Wang and Guo, 2024), thereby facilitating the efficient allocation of public goods, ecological governance resources, and infrastructural support, which collectively upgrade public service delivery and drive sustainable green development trajectories (Bearfield and Bowman, 2017). Visible and auditable regulatory records further discipline selective enforcement, raising the credibility of announced environmental standards; credible regulation lengthens firms’ planning horizons and favors investment in durable abatement equipment over short-lived evasion. Second, accessible compliance and credit information lowers the information asymmetry between regulators, financiers, and firms, so that capital can differentiate clean borrowers from polluting ones and price environmental risk accordingly. The transparency dividends generated through digital governance fortify the institutional credibility of state agencies, shielding firm-level production workflows and investment strategies from the detrimental shocks of policy uncertainty (Martins and Veiga, 2022; Wu et al., 2020). Third, public access to discharge permits, violation penalties, and monitoring results converts disclosure into quasi-enforcement: citizens obtain at near-zero cost the evidence previously monopolized by inspectors, subjecting polluting firms to continuous public scrutiny. In each link, transparency operates on the pollution dimension as well as the output dimension of GTFP, which suppresses undesirable outputs by making violations costlier to conceal, and expands desirable output by reducing the informational frictions of capital allocation. Therefore, we postulate hypothesis H2: Digital government prompts GTFP by enhancing the transparency of government information.

3.2.2 Innovation

Technological innovation functions as a fundamental engine for elevating GTFP and propelling long-term ecological sustainability (Silvestre and Ţîrcă, 2019). The logical chain from the pilot to innovation operates on two margins. On the quantity margin, the origin is administrative: unified e-government platforms compress approval times and curtail institutional friction (Tan et al., 2024), lowering the institutional transaction costs of research and productive investment; resources previously absorbed by administrative delay are released for R&D. Through comprehensive data governance, service integration, and information sharing, public sector digitalization creates profound spillover effects within regional innovation ecosystems. By exploiting the inherently non-rivalrous and non-excludable characteristics of digital knowledge, these spillovers catalyze society-wide technological advancements and trigger breakthroughs in foundational technologies (Inigo and Albareda, 2019). Precisely targeted incentive schemes accelerate eco-friendly technological evolution (Hao et al., 2024).

On the direction margin, the price signals generated by more credible enforcement consistent with the environmental governance channel, and raise the expected return to green R&D relative to conventional R&D (Zhao et al., 2023; Wang D. et al., 2025), while digitalized approval lowers the sunk costs of adopting cleaner production technologies. Innovation quantity therefore translates into GTFP specifically because its composition tilts toward green technologies; absent this directional tilt, a general innovation surge would raise conventional TFP without necessarily improving its green component. Simultaneously, the expedited elimination of highly polluting, energy-intensive sectors, systematically upgrading the industrial composition to support green economic trajectories (Tang et al., 2025), reinforcing the pollution-reduction dimension of GTFP (Tan et al., 2024). Drawing on this theoretical rationale, we postulate hypothesis H3: Digital government prompts GTFP by fostering technological innovation.

3.2.3 Factor marketization

The e-government pilot does not directly restructure factor markets. The institutions that set factor prices predate the pilot by decades and fall outside its mandates. Consequently, factor marketization serves as a predetermined environmental context within which the pilot operates rather than a channel for its effects. Its moderating influence operates through the transmission of signals. The pilot generates two distinct categories of signals relevant to pricing. These include allocative signals arising from data sharing and open information (Zhou and Ge, 2025) and shadow price signals stemming from more credible environmental enforcement. For resources to shift, these signals must be transmitted through relative factor prices. However, in environments where factor markets remain distorted due to bureaucratic monopolies and policy distortions rooted in informational frictions (Wilson and Mergel, 2022), the transmission process is weakened. Under such conditions, capital cannot flow to clean borrowers at prices reflecting their risk, land cannot be repriced away from uses that are emission intensive, and labor is unable to migrate toward green sectors. Therefore, a higher degree of factor marketization amplifies the effect of digital government on GTFP. By ensuring that the price signals generated by these channels effectively reach firm allocation decisions, marketization transforms the reduction in institutional friction into actual resource reallocation (Martins and Veiga, 2022; Wang et al., 2022). Grounded in this framework, we propose Hypothesis H4: The impact of digital government on GTFP is amplified by the improvement of factor marketization.

3.2.4 Digital economy

Digital government and the digital economy are complements rather than substitutes, and the digital economy is causally prior to the pilot: private-side digital infrastructure, data-processing capability, and application ecosystems accumulated in a region before 2018 and are not targeted by the e-government mandates. As a fundamental engine for ecological and economic sustainability, the digital economy plays an important role in propelling GTFP (Lyu et al., 2024). The digital economy supplies advanced informational infrastructure, digital management expertise, and digitized production frameworks catalyze industrial upgrading and incubate novel business prototypes (Wen et al., 2023). Concurrently, robust digital infrastructure augments the data-processing capacity of public administrations, meaning the diffusion of these technologies supplies the critical technological scaffolding required for state digitalization to drive sustainable outcomes (Li and Yue, 2025). The moderating mechanism operates primarily through absorption efficiency. In regions with a mature digital economy, firms possess the data processing capabilities to exploit disclosed information, the application ecosystems to commercialize released public data, and the managerial capacity to reorganize production around digital services. Consequently, monitoring data is actively analyzed rather than merely archived, and participation platforms are effectively utilized (Qian et al., 2024; Liu et al., 2024). Conversely, in regions where the digital economy is less developed, these governance outputs face limited absorption, causing the twin margins of emission abatement and desirable output growth to contract. Therefore, the development of the digital economy determines the conversion efficiency of digital governance into GTFP. Drawing upon this rationale, we postulate hypothesis H5: The influence of digital government on GTFP is amplified by the development of digital economy.

4 Research design

4.1 Econometric model

Initiated by a joint directive from the Cyberspace Administration of China alongside multiple national ministries in late 2017, the comprehensive e-government pilot program officially commenced in 2018. To causally identify the influence of digital public administration on green total factor productivity (GTFP) and explore its underlying mechanisms, this research constructs a Difference-in-Differences (DID) framework, which is formulated in Equation 1:

G T F P i t = α 0 + α 1 E G O V i t + α x X i + c i t y i + y e a r t + ε i t (1)

In this specification, the subscripts i and t represent the municipality and the year, respectively. The primary dependent variable is GTFP. EGOV (calculated as t r e a t i × p o s t 18 t) serves as the core DID estimator representing the digital government intervention via the pilot program. The vector X incorporates a series of control variables. To account for unobservable confounders, city _*i* and year _*t* are included to absorb city-level and time-level fixed effects, with ε i t acting as the stochastic error term. A statistically significant and positive coefficient for α ₁ would validate the stimulatory effect of digital governance on green productivity.

To evaluate the transmission channels through which state digitalization impacts GTFP, we adopt the mediation evaluation procedure proposed by Li and Luo (2020) and construct the models in Equations 2, 3: where, M _*it* denotes the mechanism variables, specifically administrative transparency and regional innovation. The mediation effect is corroborated if both γ 1 and β 1 exhibit statistical significance, implying that digital transition fosters GTFP by augmenting these two conduits.

M i t = β 0 + β 1 E G O V i t + β x X i + c i t y i + y e a r t + u i t (2)

G T F P i t = γ 0 + γ 1 M i t + γ x X i + c i t y i + y e a r t + ν i t (3)

To further explore the boundary conditions of this policy effect, we introduce moderating variables (T _*it*) including the digital economy index and the degree of factor marketization, along with their interaction terms with the core DID estimator, as depicted in Equation 4:

G T F P i t = η 0 + η 1 E G O V i t + η 2 E G O V i t × T i t + η 3 T i t + η x X i + c i t y i + y e a r t + μ i t (4)

Within this equation, T _*it* represents the respective moderators. A significantly positive η 2 would provide empirical evidence that the positive externalities of digital government on GTFP are amplified in environments characterized by advanced digital economies and highly marketized factor allocations.

4.2 Variables

4.2.1 Dependent variable

In this study, green total factor productivity (GTFP) serves as the primary explained variable. To accurately capture environmental externalities alongside economic performance, we calculate GTFP utilizing the slack-based global Malmquist-Luenberger (SBM-GML) framework. This specific model is advantageous because it simultaneously accounts for both favorable economic outputs and unfavorable ecological byproducts (Tone, 2021). Specifically, GTFP is quantified via the Malmquist-Luenberger (ML) index, which reflects the dynamic growth of productivity across consecutive periods. Consistent with the mathematical derivations in Equations 5–7, we further decouple the ML index into two core components: technical level change (GTC) and technical efficiency change (GEC), yielding the relationship GTFP = GTC × GEC. Here, GTC illustrates the inter-period shift of the production frontier (signifying technological advancement or innovation), while GEC delineates fluctuations in the relative position of a decision-making unit (DMU) to the optimal frontier (Li and Chen, 2021).

G T F P t t + 1 = 1 + D t x t + 1 , y t + 1 , d t + 1 , g t + 1 1 + D t x t , y t , d t , g t × 1 + D t + 1 x t + 1 , y t + 1 , d t + 1 , g t + 1 1 + D t + 1 x t , y t , d t , g t (5)

G T C t t + 1 = 1 + D t x t + 1 , y t + 1 , d t + 1 , g t + 1 1 + D t + 1 x t + 1 , y t + 1 , d t + 1 , g t + 1 × 1 + D t x t , y t , d t , g t 1 + D t + 1 x t , y t , d t , g t (6)

G E C t t + 1 = 1 + D t + 1 x t + 1 , y t + 1 , d t + 1 , g t + 1 1 + D t x t , y t , d t , g t (7)

Regarding the input configurations within the SBM-GML model, we utilize aggregate employment figures to proxy labor inputs. The capital stock is estimated via the perpetual inventory approach, mathematically represented as K i t = 1 − δ k K i t − 1 + I i t − 1, where I stands for gross fixed asset investment and δ indicates the capital depreciation rate, fixed at 9.6%. To eliminate inflationary distortions, fixed asset investments are deflated using municipal-level price indices anchored to a 2012 base year. For the outputs, regional real GDP is designated as the desirable economic outcome, whereas the volumes of industrial ammonia nitrogen and sulfur dioxide emissions are assigned as the undesirable environmental penalties.

4.2.2 Independent variable

The core explanatory variable in our empirical design is digital government (EGOV). We operationalize this construct by leveraging the exogenous policy shock of the “National E-Government Comprehensive Pilot” program. Mathematically, it is constructed as an interaction term (t r e a t i × p o s t 18 t.). The spatial dummy, treat _*i*, takes the value of 1 if a municipality is included in the official pilot roster, and 0 otherwise. The temporal dummy, post18_t, equals 1 for the implementation year of 2018 and all subsequent periods, acting as a post-treatment indicator.

4.2.3 Mechanism variables

To empirically elucidate the underlying transmission channels and boundary conditions, we incorporate the following variables: (1) Administrative information transparency (TGI). This is operationalized using municipal fiscal transparency scores, extracted from the “Report on the Fiscal Transparency of Chinese Municipal Governments” published by Tsinghua University’s School of Public Policy and Management. (2) Regional innovation capacity (PAT). This is quantified by taking the natural logarithm of the aggregate volume of patent applications within a given city. (3) Digital economy development (DEI). We employ an entropy-weighted approach to measure the comprehensive development of the digital economy, drawing on dimensions such as broadband coverage, internet-related output and employment, and the number of mobile internet subscribers. (4) Factor marketization index (MKI). The degree of market-oriented factor allocation is gauged relying on the annual Marketization Index report.

4.2.4 Control variables

To rigorously mitigate the risk of omitted variable bias, we introduce a comprehensive suite of municipal-level controls: (1) Financial development (FIN), calculated as the proportion of outstanding credits to total deposits within regional financial institutions at year-end; (2) Foreign direct investment dependency (FDI), determined by the ratio of actually utilized foreign direct investment to the municipal GDP; (3) Industrial structure optimization (IND), computed as the output ratio of the tertiary sector relative to the secondary sector; (4) Internet penetration (INT), expressed as the share of broadband internet users out of the total regional population; (5) Economic development (GDP): measured as the natural logarithm of per capita GDP; and (6) Population density (POP), parameterized as the natural logarithmic transformation of the total population divided by the city’s administrative land area.

Table 1 systematically outlines the variable measurements and data origins. Institutionally, the comprehensive e-government pilot was officially launched in 2018, paving the way for the State Council’s subsequent release of the “Guiding Opinions on Strengthening Digital Government Construction” (Guofa [2022] No. 14). This central directive dictates the completion of a unified national public big data architecture by 2025. Aligning with this policy evolution, our empirical estimations rely on a municipal panel dataset comprising 285 Chinese cities over the 2013–2021 observation window. The baseline summary statistics for the selected variables are subsequently reported in Table 2.

TABLE 1. Variable definition and data source.

TABLE 2. Descriptive statistics.

5 Empirical results

5.1 Baseline results

Table 3 reports the baseline regression results, with standard errors clustered at the municipal level to mitigate potential serial correlation. As illustrated in Column (1), the digital public administration coefficient is significantly positive at the 5% threshold, confirming its stimulatory effect on urban GTFP. When decoupling the primary dependent variable into technical level change (GTC) and technical efficiency change (GEC), Columns (2) and (3) demonstrate that this digital transition robustly propels GTC but exerts no statistically distinguishable influence on GEC. Within the SBM-GML paradigm, GTC denotes an outward expansion of the production frontier, whereas GEC measures shifts in the proximity of decision-making units to this optimal boundary. Upon the inclusion of comprehensive municipal control variables in Columns (4) through (6), the positive externalities on both GTFP and GTC remain highly robust, while the GEC coefficient stays negligible. These findings empirically validate Hypothesis H1, implying that state digitalization augments GTFP predominantly by spurring technological advancement rather than elevating technical efficiency.

TABLE 3. Baseline regression results.

values in () are robust standard errors clustered at the city level; *P < 0.1, P < 0.05, *P < 0.01.

Under the SBM-GML accounting framework, GTC represents outward shifts in the production frontier, whereas GEC measures the extent to which decision-making units (DMUs) converge toward this frontier. The significantly positive GTC coefficient aligns with the frontier-shifting mechanism emphasized in Section 3.1. By streamlining environmental permitting and approval workflows, digital government reduces the administrative burdens facing firms that adopt cleaner production and abatement technologies. In contrast, the null effect observed for GEC highlights the constraints on the efficiency margin. Catch-up requires process reengineering and the reallocation of inputs toward existing best practices. However, local protectionism and institutional barriers systematically obstruct this mechanism in the Chinese context. This asymmetry implies a clear hypothesis that digital government enhances GTFP primarily by shifting the production frontier outward rather than moving firms toward the existing frontier. Although the GEC component remains theoretically relevant, it is unlikely to capture policy effects in the short to medium term. This is because realizing efficiency gains necessitates organizational restructuring, managerial learning, and structural reforms, the benefits of which extend well beyond the pilot period. Consequently, our subsequent heterogeneity analysis prioritizes GTFP and its dominant GTC component, while we acknowledge that the efficiency channel may only become observable over a longer horizon.

5.2 Robust checks

5.2.1 Parallel trend test

The parallel-trends assumption is vital for causal validity for the difference-in-differences framework, which stipulates that the baseline trajectories of the treatment and control cohorts must not systematically diverge prior to the policy shock. To empirically validate this prerequisite, we adopt the event-study approach proposed by Beck et al. (2010). This involves incorporating interaction terms between the municipal treatment indicator and temporal dummy variables, as formalized in Equation 8:

y i t = φ 0 + ∑ j = − 4 3 φ j t r e a t i × y e a r 2018 + j + φ X X i t + c i t y i + y e a r t + ε i t (8)

In this specification, the y e a r 2018 + j dummy equals 1 if the observational period aligns with the year 2018 + j, and takes the value of 0 otherwise. The validity of the assumption is confirmed if the pre-treatment estimated coefficients φ j fail to achieve statistical significance, signifying that pre-intervention comparability between the two groups. The dynamic coefficient estimates, alongside their 95% confidence intervals for both GTFP and GTC, are graphically depicted in Figures 1, 2. Prior to the 2018 rollout, the estimated parameter φ j consistently hover around zero and remain statistically insignificant relative to zero. The joint F-test yields a statistic of 0.95 (p = 0.437) for GTFP and 0.78 (p = 0.536) for GTC. This absence of divergent pre-trends confirms that the foundational assumption of our DID estimation is robustly met. After the pilot’s launch, the coefficients turn significantly positive, demonstrating an escalating trajectory over time. These patterns provide compelling evidence that the digital government transition exerts a sustained and progressively amplifying stimulus to both GTFP and GTC.

5.2.2 Placebo test

To ensure our baseline findings are not driven by unobserved random shocks, we perform a rigorous placebo test by randomly reassigning both the pilot municipalities and the policy initiation years. These randomized spatial and temporal indicators interact to generate a pseudo difference-in-differences estimator. We simulate this random assignment process 500 times, extracting the artificial coefficients and corresponding p-values from each iteration. As illustrated in Figures 3, 4, the probability distributions of these pseudo-coefficients for both GTFP and GTC are strictly centered around zero. Furthermore, only a negligible fraction of the simulated estimates exceeds the true baseline coefficients, which are indicated by the dashed vertical lines. Consequently, this falsification exercise robustly confirms that the observed positive impacts are not artifacts of chance, thereby validating the specificity of our estimations.

5.2.3 Selection bias in e-government pilot

• Reverse causation. Endogeneity issues may emerge if a municipality’s prior GTFP performance dictates its probability of being chosen for the pilot program. To verify whether pilot designation is predetermined by green productivity, we construct a Probit regression utilizing pre-intervention data. In this estimation, the pilot city dummy variable (treat) serves as the explained variable, with GTFP acting as the predictor. As reported in Column (1) of Table 4, historical GTFP exerts no statistically distinguishable impact on the likelihood of a city becoming a pilot zone, thereby ruling out reverse causality.
• Pre-pilot sample. To eliminate the possibility that local governments altered their ecological behaviors in expectation of the policy, we conduct estimations restricted strictly to the pre-pilot sample. The empirical results shown in Columns (2) and (3) of Table 4 suggest that the treatment indicator coefficients are not statistically significant. This absence of pre-intervention divergence between the treated and control cohorts confirms that our baseline estimations are not confounded by anticipatory actions, further solidifying the robustness of our conclusions.
• Sample selection bias. To systematically correct for potential non-random assignment in pilot selection, we implement the Heckman two-stage procedure. Within this framework, the chronological age of the municipal party secretary is adopted as the exclusive instrumental variable for the initial selection equation. This choice is grounded in the selection mechanism: the 2018 e-government pilot relied on applications and coordination, implying that a city’s inclusion depended largely on the leadership’s willingness to reform. Since promotion prospects for city-level leaders are non-linearly related to age, older secretaries have fewer incentives to launch visible institutional innovations (Li and Zhou, 2005). The first-stage estimates in Table A1 reveal that the secretary’s age is a significant negative predictor of selection. The second-stage outcomes are displayed in Columns (4) and (5) of Table 4. Even after incorporating the Inverse Mills Ratio (IMR) to adjust for selection probabilities, the estimated coefficients for the digital governance intervention continue to be significantly positive for both GTFP and GTC. This aligns perfectly with our primary findings.

TABLE 4. Regression results of selection bias.

The dependent variable in column (1) is a pilot city dummy. PSM-DID, denotes the Propensity Score Matching Difference-in-Differences estimator, reporting the average treatment effect on the treated (ATT). Bootstrapped standard errors are based on 100 replications.

5.2.4 PSM-DID

To further account for potential selection bias stemming from systematic disparities in city-level attributes, we augment our baseline analysis by adopting the Propensity Score Matching Difference-in-Differences (PSM-DID) methodology. This dual-stage technique refines our estimation by constructing a counterfactual group that closely approximates the observable characteristics of the pilot cohort. Initially, we derive propensity scores via a Probit specification. Subsequently, we execute a nearest-neighbor matching procedure (k = 2) within the common support region, ensuring that the balancing conditions are rigorously met. The ATT estimates are presented in Columns (6) and (7) of Table 4. These results indicate that the e-government pilot maintains a significantly positive effect at the 5% threshold, thereby providing robust evidence consistent with our primary findings.

5.2.5 Controlling for related policy influences

To mitigate potential identification bias stemming from concurrent institutional interventions, we conduct a series of robustness checks targeting other major digitalization policies. First, to isolate the effects of digital governance from the “Broadband China” initiative, which may similarly influence regional GTFP (Zhou et al., 2022), we incorporate an interaction term representing this pilot policy (BORRD) into our baseline Equation 1. As demonstrated in Columns (1) and (2) of Table 5, digital government consistently retains a significant and positive influence on GTFP and GTC. Furthermore, existing scholarship frequently utilizes the “Information Benefiting the People” pilot as an alternative proxy for digital public administration (Yan and Lyu, 2023). To account for any confounding influence attributable to this policy, we introduce a separate difference-in-differences indicator (INF) into Equation 1 to control for its presence. The estimation results, presented in Columns (3) and (4), show that the coefficient of our primary digital government variable maintains its positive statistical significance for both GTFP and GTC. Recent academic inquiries have extensively explored the institutional outcomes of digital government by scrutinizing various pilot schemes, including the restructuring of big data governance frameworks (Y. Chen and Li, 2025) and the strategic launch of public data sharing infrastructures (He et al., 2026). To effectively isolate our core results from potential confounding shocks linked to these concurrent initiatives, we incorporate additional interaction terms into our primary Equation 1. Columns (5)–(8) documents the corresponding estimation outputs, where PUB and BIG denote the DID estimators for municipalities participating in the public data opening pilot and the big data management reform, respectively. Empirical evidence demonstrates that our primary digital government indicator maintains its positive statistical significance regarding GTFP and GTC, while the impact on GEC consistently fails to reach conventional levels of statistical significance. These observations align with our baseline findings, confirming that the identified developmental dividends are robust even after controlling for overlapping institutional policy regimes.

TABLE 5. Regression results of controlling for related policies.

BOARD, INF, PUB, and BIG, denote the treatment indicators (DID, terms) for the “Broadband China”, “Information Huimin Project”, “Opening of Public Data Platforms”, and “Reform of Big Data Management Institutions” pilot policies in China, respectively.

In summary, the Broadband China pilot prioritizes the expansion of physical network infrastructure, while the Information Benefiting the People program aims to improve citizen-facing public services. In contrast, public data opening and big data management reforms target specific aspects of data governance. The e-government pilot represents a comprehensive administrative reform that reconfigures cross-departmental data flows, approval procedures, and environmental governance processes. While complementarity among these policies is theoretically plausible, such synergy would likely manifest as a second-order interaction effect, amplifying or dampening the pilot’s impact in specific cities rather than overturning the first-order average effect. Our additive control specification effectively removes the independent effects of these policies. The stability of the coefficient on EGOV in both magnitude and significance across the specifications in Tables 5, 6 demonstrates that these additive controls adequately capture the systematic variation contributed by related policies. If strong synergy were the primary driver of our results, we expect the coefficient on EGOV to exhibit high sensitivity to the inclusion of these controls. The observed stability suggests this is not the case.

TABLE 6. City administrative heterogeneity.

cities with higher administrative levels include direct-controlled municipalities, provincial capitals, and sub-provincial cities.

5.3 Heterogeneity

5.3.1 City administrative heterogeneity

Municipalities with superior administrative standing typically enjoy expanded fiscal sovereignty and privileged access to capital markets. In this study, we categorize direct-controlled municipalities, provincial capitals, and sub-provincial cities as higher administrative-tier urban centers. As demonstrated in Columns (1) and (2) of Table 6, the digital government initiative fails to yield a statistically significant effect on either GTFP or GTC within these high-tier cities. A plausible explanation is that such urban centers often rely on alternative catalysts for green growth, such as concentrated human capital, favorable policy mandates, and structural financial advantages (Jiang et al., 2018). Conversely, the estimations for ordinary prefecture-level cities, reported in Columns (3) and (4), reveal that the pilot program significantly accelerates both GTFP and GTC. These outcomes provide compelling evidence that the digital governance transition serves as an effective lever for promoting green development primarily within ordinary prefecture-level jurisdictions.

5.3.2 Geographic heterogeneity

To further investigate whether the policy efficacy varies across regional landscapes, we segment our sample into coastal and inland municipalities. Table 7 delineates the heterogeneous effects across these geographic strata, with Columns (1) and (2) reporting results for coastal regions, while Columns (3) and (4) provide estimates for inland counterparts. The regression results show that digital government has a statistically significant and positive effect on both GTC and GTFP in both regional subsamples, supporting our initial findings. Notably, a coefficient comparison indicates that the magnitude of this positive impact is more pronounced in inland areas. This observation suggests that by institutionalizing data interoperability and optimizing the exchange of administrative information, the digital government pilot compensates for inherent geographic disparities, more effectively refining governance quality and state-market dynamics in inland jurisdictions.

TABLE 7. Geographic heterogeneity.

5.3.3 Economic development heterogeneity

To explore whether the economic maturity of a municipality modulates the policy impact, we partition our sample into two cohorts based on the median real GDP per capita in 2017, the year immediately preceding the pilot’s rollout. Table 8 provides the empirical estimates for this development-based heterogeneity. Our analysis shows that in both sub-samples, cities below and above the median economic level, the e-government pilot has a statistically significant positive impact on both GTFP and GTC. Conversely, the impact on technical efficiency change (GEC) consistently fails to attain statistical significance. These empirical patterns corroborate our baseline findings, underscoring that digital government consistently drives green productivity growth via the channel of frontier technological progress, regardless of a city’s prevailing economic development status.

TABLE 8. Economic development heterogeneity.

5.4 Mechanisms analysis

5.4.1 Mediating effects

This study postulates that digital government architectures function as a catalyst for Green Total Factor Productivity (GTFP) by enhancing administrative information transparency and incubating regional innovation. The empirical outcomes derived from the mechanism tests are synthesized in Table 9. As evidenced by Column (1), the e-government pilot (EGOV) exerts a statistically significant positive effect on fiscal transparency, implying that digital government initiatives such as the implementation of “One-Stop Online Government Services”, effectively dismantle information barriers and facilitate granular information disclosure. Furthermore, Columns (2) and (3) verify that such transparency gains substantially bolster both GTFP and GTC, thereby empirically confirming the validity of this transmission conduit and supporting Hypothesis H2. Regarding the innovation channel, Column (4) demonstrates that digital government deployments substantially ignite regional innovative activity. Subsequently, as illustrated in Columns (5) and (6), these innovation outputs are positively correlated with superior GTFP and GTC performance. These results collectively substantiate that digital government elevates green productivity by acting as a primary stimulus for technological creativity, which rigorously validates Hypothesis H3.

TABLE 9. Regression results of mediating effect.

5.4.2 Moderating effect

To investigate the boundary conditions of the digital government-GTFP nexus, we examine the moderating roles of factor marketization and digital economy development. The empirical estimations for these interaction models are summarized in Table 10. As shown in Columns (1) and (2), the coefficients of the interaction term between the pilot indicator and the factor marketization index (EGOV × FMI) are statistically significant and positive at the 1% level. Similarly, Columns (3) and (4) demonstrate that the interaction terms involving the digital economy development index (EGOV × DEI) also yield positive coefficients, which are significant at the 1% threshold. These empirical patterns suggest that both factor market integration and a mature digital economy function as critical enhancers, amplifying the catalytic influence of digital public administration on urban GTFP and GTC. In essence, the positive policy dividends of the e-government pilot become more pronounced as regional factor markets evolve and the digital infrastructure matures. Consequently, these findings provide empirical support for both Hypothesis H4 and Hypothesis H5.

TABLE 10. Regression results of moderating effect.

6 Conclusions and implications

6.1 Conclusions

This study empirically explores the nexus between government digital transition and urban Green Total Factor Productivity (GTFP), utilizing the 2018 National E-government Comprehensive Pilot as an exogenous policy shock. By applying an SBM-GML framework to quantify GTFP and a difference-in-differences (DID) estimator for causal identification, we analyze a longitudinal panel dataset spanning 285 Chinese municipalities from 2013 to 2021. Our empirical findings demonstrate that digital government architecture functions as a robust catalyst for GTFP enhancement, with decomposition diagnostics attributing this trajectory primarily to technological progress. These baseline conclusions remain resilient across a comprehensive battery of sensitivity and validity assessments. Heterogeneity analyses reveal that the policy’s environmental dividends are particularly significant in inland and ordinary prefecture-level jurisdictions, pointing to a vital compensatory institutional function. Mechanism investigations further elucidate that digital governance elevates GTFP by augmenting administrative informational transparency and stimulating regional innovation ecosystems. Moreover, this positive policy effect is strategically amplified by matured factor marketization and a robust digital economy. Consequently, this investigation establishes a compelling theoretical link and derives actionable policy blueprints for synchronizing digital government modernization with sustainable development trajectories.

6.2 Policy implications

The findings of this study offer a distinctive perspective for developing and emerging economies that are considering digital government reforms as a path to sustainable development. China’s pilot experience illustrates that digitalizing the public sector can act as more than an administrative modernizing tool; it can serve as an institutional lever for advancing green productivity. However, the transferability of these lessons depends on several enabling conditions such as baseline data infrastructure, cross-departmental coordination capacity, and institutional stability, which should be carefully assessed in each national context.

First, government digital transition should be designed as systemic institutional interventions rather than stand-alone technological upgrades. For developing economies, this means explicitly linking e-government investments to environmental governance, industrial upgrading, and broader governance modernization goals. One central lesson from China is that the green dividends of digitalization materialize when digital tools are embedded in administrative processes that directly shape pollution monitoring, emission permitting, and environmental information disclosure. Countries with weaker administrative data systems should therefore prioritize foundational data collection and integration before expecting advanced governance outcomes. Without this sequencing, digital initiatives risk being seen as technology for its own sake, with limited environmental benefits.

Second, optimize information flows to reduce system friction. Fragmented information and organizational silos constrain environmental governance in nearly all developing countries, though the severity of these constraints varies markedly. The Chinese experience shows that unified platforms consolidating firm registration, environmental permitting, and compliance records can substantially reduce information asymmetries between regulators and firms, making environmental enforcement more credible. Advancing innovation-driven strategies requires dismantling organizational silos. Establishing integrated information platforms enhances inter-agency coordination via cross-level data sharing and real-time feedback loops. Simultaneously, deploying digital portals and mobile applications accelerates policy dissemination while bolstering institutional transparency.

Third, balancing standardization with local adaptation is critical when scaling up digital government reforms. China’s pilot combined centrally defined requirements including unified technical standards, data-sharing protocols, and common service expectations, with considerable discretion for provincial and municipal governments to adjust implementation to local socioeconomic conditions. As successful digital governance practices expand from localized contexts to broader jurisdictions, cross-regional coordination becomes crucial. To ensure interoperability, authorities should standardize digital protocols and integrate data architectures, while avoiding the creation of fragmented, isolated systems. This establishes a polycentric governance framework that harmonizes system-wide standardization for consistency with localized adaptation for responsiveness.

Fourth, foster subsystem co-evolution to sustain technological progress. Because digital government drives GTFP primarily via technological advancement, governance models must co-evolve dynamically with market and technological subsystems. By aligning digital governance with the broader digital economy and efficient factor allocation, authorities can cultivate a synergistic administrative model. For developing and emerging economies, digital reforms should be sequenced or coordinated with complementary policies that support the reallocation of capital, labor, and land toward cleaner and more productive activities. Where complementary reforms are absent, the environmental gains from digital government may be limited to improvements in compliance, without the sustained technological upgrading observed in China.

6.3 Limitations and future research

First, the research boundary treats cities as relatively closed systems. The DID model effectively isolates the local treatment effect of digital governance interventions; however, it inherently assumes spatial independence among observations. In reality, urban systems are open and interconnected. The implementation of digital governance in one city likely generates spatial spillovers such as knowledge diffusion affecting GTFP in neighboring cities. Future studies should adopt a network systems perspective, employing social network analysis or spatial models to map the topological structure of inter-city digital interactions and evaluate the systemic ripple effects of digital governance across regional networks.

Second, the assumption of linear causal pathways oversimplifies complex system dynamics. While the mediation and moderation models successfully identify the specific roles of technological progress and institutional environments, they essentially approximate linear and additive causal chains. In a complex socio-technical system, causal relationships are often non-linear, characterized by threshold effects and path dependence. Future inquiries could leverage methodologies attuned to system complexity to uncover the configurational pathways and non-linear feedback loops through which digital governance, market conditions, and innovation ecosystems synergistically drive sustainable development.

Third, the evolutionary stage of digital governance warrants longitudinal tracking. The current study evaluates digital government as a relatively static institutional shock. Yet, digital government itself is an evolving subsystem transitioning from simple digitization (e-government) to intelligent algorithmic governance (AI-driven). The systemic mechanisms identified here may shift as the technology matures. Future research should conduct longitudinal system tracking, examining how the co-evolutionary dynamics between governance algorithms and sustainable development outcomes transform across different phases of digital maturity.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Tsun Se Cheong, Hang Seng University of Hong Kong, China

Reviewed by: Hao-kuang Li, Shanghai University, China

Kexing WU, Guangxi University of Finance and Economics, China

Footnotes

This distinction guides our empirical analysis method in Section 4.1.