Does Local Government Green Attention Promote Green Total Factor Productivity?

Abstract

Improving green total factor productivity (GTFP) is critical for balancing economic benefits and ecological constraints. While most existing studies emphasize the pivotal role of governments in GTFP enhancement, they predominantly treat governments as homogeneous entities, overlooking the fundamental premise of local government attention allocation. Analyzing 2010–2020 data from 285 Chinese cities, this study reveals that increased local government green attention significantly stimulates GTFP through three channels: fostering green technology collaboration among firms, deepening green involvement of public research institutions, and elevating green innovation quality. Heterogeneity analyses demonstrate amplified effects in cities characterized by intense intergovernmental competition, stringent intellectual property protection, robust fiscal capacity, and advanced technological infrastructure, but attenuated impacts in resource-dependent regions with heavy reliance on extractive industries.

1. Introduction

Decades of rapid industrialization and urbanization have driven significant global economic growth, particularly in emerging economies. Nevertheless, the prevalent “develop first, manage pollution later” model in developing nations [1] has not only depleted natural resources [2] but also intensified systemic tensions between economic growth and ecological sustainability [3]. Consequently, resolving the dual challenge of advancing economic development while mitigating environmental damage has become a critical global research priority [4,5]. Sustainable development rests on two strategic pillars: economic development meeting present needs, and ecological conservation preserving future generations’ consumption opportunities [6]. While draconian measures such as forced shutdowns of polluters stifle economic growth [7], laissez-faire approaches exacerbate ecological degradation. Achieving genuine sustainability, therefore, requires coordinated governance of economic and environmental systems.

Enhancing GTFP—generating greater economic benefits with reduced resource consumption and environmental impact—constitutes the fundamental pathway to achieving economic gains and ecological conservation [8]. However, due to inherent non-productivity in green investments and environmental benefit spillovers [9], most enterprises lack the initiative to improve GTFP [10], rendering governments the primary drivers of progress [11]. This is evidenced by China’s central government implementing the environmental regulations and policies to transform energy consumption structures under intensifying pollution and growing energy scarcity [6,12].

Nevertheless, substantial spatial heterogeneity in policy outcomes persists across regions despite uniform central government environmental targets. Prevailing studies often treat governments as homogeneous, emphasizing explicit institutions like environmental regulations while neglecting the fundamental role of local government allocation of attention. Although central governments set macroeconomic targets [13], local governments create and implement specific development plans [14] while balancing ecological preservation and economic growth under acute resource scarcity constraints.

This dual-constraint mechanism triggers “selective prioritization” in local governance: when ecological goals are deprioritized, even robust institutions become performative without adequate resources. Under economic constraints, ecological race-to-the-bottom intensifies [15], weakening environmental objectives in development plans. Consequently, local governments trade off between ecology and growth [16], which causes policy divergence. Conversely, over-centralized campaign-style governance risks policy instability by ignoring economic complexity. Under external constraints, this approach undermines green innovation and environmental efficiency while neglecting social welfare [17,18].

The intensity and endurance of local government attention allocation to green issues—termed local government green attention (GA) (the following full text refers to local government green attention as GA for short)—constitutes a “political signaling ecosystem” in environmental governance: this system shapes incentives and constraints through fiscal resource allocation, while conveying hierarchical priorities via policy orientations to establish a “policy context” for firms’ perception of governmental preferences. This subsequently modulates corporate perception and responsive behaviors. Compared to static institutional analyses, the dynamic perspective of government attention allocation reveals micro-mechanisms of environmental policy implementation bias and provides breakthrough pathways for ecological governance. Clarifying GA’s impact pathways upon GTFP holds practical significance for optimizing policy attention allocation; however, empirical evidence regarding its mechanisms remains scarce.

Given the urgent need to improve GTFP, we explore the GA-GTFP relationship from a policy signaling perspective to provide robust empirical evidence. This study addresses two key questions: Can GA enhance GTFP? What are the mechanisms underlying GA’s effect on GTFP?

To examine GA’s impact, we quantified GA as the frequency of ecology-related keywords in local government work reports. Results show that higher GA increases local GTFP: a one percentage point increase in GA leads to a 17.31 percentage point increase in GTFP. We used two complementary approaches to address omitted variable bias and unobserved heterogeneity—constructing a Bartik instrumental variable and modifying model settings. All robustness tests support the baseline findings.

Enhanced GA produces profound impacts through three key mechanisms. First, it reshapes firms’ regulatory pressures and institutional environments, breaking down barriers to green collaboration while fostering knowledge-sharing and resource complementarity. Second, increased GA redirects funding while stabilizing green alliances, thereby strengthening public research institutions’ involvement in green innovation and enabling firms to absorb advanced green knowledge. Third, an increased GA both stimulates substantive innovation and curbs symbolic green behaviors, thereby improving the overall quality of corporate green innovation and simultaneously securing returns while ensuring sustainability.

Our analysis contributes to three areas. First, we enrich the literature on GTFP determinants. Building on discussions of GTFP definitions and empirical approaches [8,19,20], prior literature examines the impacts of eco-development targets [21], green finance [22], technical efficiency [23], industrial upgrading [24], energy efficiency [25], industrial agglomeration [26], environmental regulation [27] and public participation [28] on GTFP. Unlike prior studies, however, we begin with the empirically observed spatial heterogeneity in regional GTFP under centrally mandated environmental constraints, focusing on the “selective prioritization” manifest in local policy enactment and implementation. We explore how GA modulates GTFP via a “political signaling system”, thereby enriching the literature on GTFP determinants.

Second, our research extends micro-mechanisms of local government green policy signaling. Unlike existing literature using green policy-regulation/subsidy-innovation frameworks to explain corporate green innovation adjustments [29,30] or applying stakeholder perspectives to analyze green practices through financial/environmental performance [31,32], we uniquely demonstrate how governmental ecological prioritization (GA) functions as a “political signaling system” in environmental governance. This shapes a policy context enabling firms and public research institutions to perceive state preferences, ultimately driving GTFP micro-transmission. Specifically, enhanced GA elevates GTFP through three channels: fostering inter-firm green collaboration, increasing green involvement by public research institutions, and improving green innovation quality—supplementing indirect mechanisms beyond regulations and subsidies.

We propose a feasible pathway to enhance GTF by leveraging GA. While existing studies predominantly treat governments as homogeneous entities, focusing on explicit tools such as environmental regulations and industrial policies [33,34,35,36], they overlook GA as a fundamental precursor to decision-making. Our findings reveal that enhanced GA not only directly elevates GTFP through regulatory and subsidy-based mechanisms but also indirectly boosts it by emitting green policy signals that reshape corporate expectations and preferences, thereby serving as a signal amplifier for establishing actionable pathways.

The remainder of this paper unfolds as follows: Section 2 develops research propositions; Section 3 details data sources and constructs core variables; Section 4 presents baseline regression results with mechanism analysis; Section 5 summarizes key insights and policy implications.

2. Theoretical Mechanism and Hypothesis Presentation

2.1. Local Government Green Attention and GTFP

Improving GTFP requires firms to integrate environmental factors into their operations; however, intrinsic motivation is limited by non-productive green investments and environmental benefit spillovers [9]. Although GTFP enhances competitive advantages [37], firm decisions depend on cost–benefit trade-offs between compliance costs and efficiency gains [38,39,40], necessitating government intervention as an irreplaceable driver [11]. However, government attention allocation lacks consistency, cyclically alternating between prolonged neglect and abrupt concentration on specific issues, triggering disruptive policy shifts [41]. GA signifies environmental prioritization [11,42], reflecting decision-makers’ resource allocation to environmental issues and policy advancement [43,44]. Elevated GA directly shapes resource allocation through explicit policies, such as regulations and subsidies [11], while simultaneously influencing firms via implicit policy signals.

Explicit policy regulations under higher GA impose stricter constraints on energy consumption, emissions control, pollution management, and resource efficiency, while intensified penalties systematically restrict polluters. Under rigorous supervision (e.g., negative lists and pollution redlines), governments encourage firms to adopt green innovation [45], thereby reducing conservation/reduction costs to meet GTFP mandates [6,12]. Regarding explicit policy incentives, elevated GA signals stronger support for green projects. Given that the double externality of environmental issues discourages private green initiatives [46,47], government subsidies and fiscal incentives resolve short-term conflicts, thereby enhancing green innovation motivation [48] and catalyzing eco-investments [49]. Consequently, this process facilitates energy retrofits, clean energy transitions [50,51,52], and improvements in GTFP.

Through implicit policy signals, government green attention conveys durable guidance to firms, thereby solidifying their confidence in GTFP improvement. Elevated GA reflects shifts in policy priorities and sectoral development goals, functioning as a coordinating mechanism that reallocates fiscal resources toward green innovation—specifically eco-product design and green technology R&D. Firms consequently intensify commitments to energy conservation, pollution remediation, and renewable energy adoption, strategically aligning corporate initiatives with governmental frameworks to secure policy support. Simultaneously, heightened GA prioritizes green products in government procurement, channeling private investments toward sustainable innovators [6,53]. Such prioritization induces financial institutions to fund ecological projects —particularly those advancing green technology— thereby enabling GTFP enhancements. Therefore, we propose:

Hypothesis 1.

Increasing local government green attention significantly improves GTFP within the jurisdiction.

2.2. The Impact Mechanism of Local Government Green Attention on GTFP: Green Technology Cooperation

Green technology innovation is crucial for reconciling economic and ecological imperatives [37,54], yet the double externality dilemma constrains it. The public-good nature of mitigation weakens emission-reduction incentives, while knowledge spillovers prevent monopolization of green technology benefits—jointly creating R&D investment-return imbalances [46]. Amid intensifying global competition and shortened product lifecycles, firms face compounding pressures from incremental innovation bottlenecks, as well as prolonged high-risk R&D [55]. Inter-firm green technology cooperation thus becomes vital for knowledge sharing and resource complementarity. Empirical evidence confirms that external resource integration advances green innovation and GTFP [51,56,57,58]. However, such cooperation requires not just technology exchange but also a strategic consensus on actionable environmental pathways. Market barriers and information asymmetry inherently hinder this process [46], which underscores the necessity of government mediation.

Government emphasis on green governance overcomes corporate collaboration barriers by reshaping regulatory and institutional frameworks. First, heightened GA drives collaboration through regulatory pressure: stricter emission standards and higher green technology requirements intensify environmental constraints and concealment costs. Given the environmental complexities and the dual externality of green innovation, firms pursue cross-firm cooperation to acquire external capabilities [59,60,61], especially for non-green sector firms that must meet compliance standards to avoid penalties [62]. Second, increased GA reduces barriers via institutional support. More substantial governmental commitment enhances policy enforcement [43,44], lowers cooperation costs through subsidies, and mitigates information asymmetry with certification systems (e.g., green projects/firms/technologies) to facilitate green alliances [57]. Regulatory measures combined with credit certification also deter opportunism by imposing higher breach costs [11]. Finally, heightened GA strengthens endogenous motivation through reputational incentives. Prioritizing ecological governance restructures societal environmental awareness [11], generating public oversight pressure and state-guided market premiums that jointly incentivize cooperation. These informal institutions complement formal regulations in reinforcing firms’ internal drives. Collectively, heightened GA promotes green cooperation through regulatory pressure, institutional support, and reputational mechanisms, thereby enhancing regional GTFP. Therefore, we propose:

Hypothesis 2.

Increasing local government green attention improves GTFP within the jurisdiction by promoting green technology cooperation.

2.3. The Impact Mechanism of Local Government Green Attention on GTFP: Green R&D Involvement of Public Research Institutions

Compared to traditional innovation, green innovation is characterized by greater knowledge complexity and a technological frontier orientation [63]. It relies more heavily on breakthroughs in basic research, engaging greater participation from universities and public research institutions than does traditional innovation [46]. The science-technology innovation model [64] supports corporate environmental investments through economic stimuli and knowledge transfer [46], enabling green knowledge absorption and innovation that ultimately enhance GTFP. Government grants constitute the primary R&D funding for public research institutions [65]. Heightened GA directs funding toward green projects, thus increasing institutional involvement in green R&D [66]. Furthermore, research evaluation systems that prioritize environmental performance indicators further steer institutional focus toward green initiatives.

Government–public research institution–firm green alliances constitute a crucial mechanism for public research institutions to engage in ecological governance and enhance green innovation capabilities [67]. However, the inherent complexity of green technologies, high risks in the innovation process, and collaborative barriers—including strategic priority shifts, core knowledge leakage, and free-riding behavior among government, industry, and public research institutions—impede improvements in alliance stability and output efficiency [68]. As policy architects, governments can optimize the incentives and regulatory oversight for green technology innovation projects, creating enabling environments for industry–university collaboration [69,70] that mitigate knowledge leakage and free-riding while deepening institution–firm engagement in green initiatives [68,71]. Serving as the central actor in the government–public research institution–firm tripartite helix framework [72], governments steer R&D collaboration through funding and institutional support [73]. Within this collaborative green innovation model, heightened government subsidies [40] and strengthened penalty clauses directly impact the profits of industrial entities and public research institutions, thereby enhancing the latter’s motivation to engage in green innovation [68]. Based on this analysis, we propose:

Hypothesis 3.

Increasing local government green attention improves GTFP within the jurisdiction by enhancing the involvement of public research institutions in green R&D.

2.4. The Impact Mechanism of Local Government Green Attention on GTFP: Quality Leap in Green Innovation

As global emphasis on green transition intensifies, environmental performance has become crucial for firms to gain support and subsidies [74,75]. Since governments often require observable innovation outputs for subsidies, firms under institutional pressure may prioritize green innovation quantity over quality or engage in token innovation after obtaining subsidies, leading to strategic greenwashing [12]. This poses particular challenges for developing countries with weaker innovation capabilities. With growing environmental awareness, environmental governance has gained prominence in the agendas of local governments. The incorporation of mandatory emission standards into political evaluations, combined with high-pressure deterrence, not only stimulates innovation but also suppresses strategic green innovation behaviors.

A stringent regulatory environment drives significant innovation effects [45]. Facing binding regulatory pressures from intensified GA, firms with sufficient capital or technical capacity may proactively reallocate resources through high-quality green innovations to achieve technological transformation or upgrading, reducing environmental compliance costs [6,12]. Sustained governmental focus on green issues converts mandatory environmental regulations from short-term shocks into normalized constraints, making quality innovation integral throughout firm life-cycles. Furthermore, heightened local GA suppresses strategic green innovation. Increased GA prioritizes the evaluation of green projects and directs subsidies toward high-quality green initiatives, thereby enhancing the overall quality and sustainability of green innovation. Additionally, intensified government scrutiny and strict environmental regulations also increase concealment costs for firms attempting to implement low-quality green innovations. Data falsification and expedient green projects risk litigation and penalties, deterring greenwashing and strategic green innovation [76]. Based on this analysis, we propose:

Hypothesis 4.

Increasing local government green attention improves GTFP within the jurisdiction by enhancing the quality of green innovation.

3. Research Design

3.1. Data

Given data availability, this study selects 285 prefecture-level cities in China from 2010 to 2020 (excluding Hong Kong, Macau, Taiwan, Tibet, and cities with significantly incomplete data) as research samples to evaluate the impact of local government green attention on GTFP. Core datasets include city-level economic indicators, term frequencies from government reports, and green patent statistics. Economic data originate from the China City Statistical Yearbook, EPS Database, and local economic-social development bulletins. Term frequency data are derived from key policy documents—including the Action Plan for Green Development in National High-Tech Zones, the White Paper on Sustainable Urban Development, China’s Green Development in the New Era, the CPC 20th National Congress Report, and annual government work reports—through textual analysis to build a GA keyword lexicon. Green patent data originate from authorized invention patents published by the National Intellectual Property Administration, filtered using the International Patent Classification Green List (WIPO, 2010).

3.2. Variable Selection

3.2.1. Dependent Variable: Regional GTFP (GTFP)

DEA stands as the predominant method for measuring Green Total Factor Productivity [77]. Contemporary GTFP measurement practice most widely implements the non-radial Slacks-Based Measure (SBM) directional distance function. However, the SBM model fails to preserve the original proportional information in the projection values of efficiency frontiers. When the optimal slack variables are zero or positive, the results consequently exhibit significant differences. Given this limitation, the EBM model—which integrates radial distance functions and non-radial SBM components—delivers superior estimation precision [78]. Although the Global Malmquist–Luenberger (GML) index resolves directional distance function infeasibility and enables temporal comparability, it cannot eliminate measurement biases caused by radial and angular orientation constraints [79]. Thus, adopting the EBM-GML index model proposed by Wang and Wang (2023) [80], this study ensures enhanced precision in addressing radial/orientation biases while maintaining global comparability across production frontiers.

Drawing on the construction methodology of the GTFP indicator system from Zhou et al. (2024) [20] and Du et al. (2024) [26], this study employs an EBM-GML model covering 285 decision-making units over the 12-year period from 2009 to 2020. The framework incorporates three input indicators, one desirable output, and four undesirable output indicators (see

Table 1. Regional GTFP Measurement Indicators.

Category	Component	Indicator
Inputs	Labor	Annual total employed population (people)
	Capital	Capital stock (ten thousand yuan)
	Energy	Electricity consumption (ten thousand kilowatt-hours)
Desirable Output	Economic Benefit	Real GDP (ten thousand yuan)
Undesirable Outputs	Environmental Impact	Industrial Particulate Emissions (metric ton)
		Wastewater discharge (ten thousand metric tons)
		Sulfur dioxide (SO2) emissions (metric ton)
		PM2.5 (micrograms per cubic meter)

Note: Prior to measuring GTFP, all data underwent standardization to eliminate disparities across measurement units.

). Following Yu and Shen’s (2020) [81] approach, we estimated the regional capital stock using the perpetual inventory method (base year: 1952) and adjusted real GDP figures to constant 2000 prices.

3.2.2. Explanatory Variable: Local Government Green Attention (GA)

Drawing on the methodology of Guo and Qiao (2024) [82], this study constructs its core explanatory variable―City-Level Local Government Green Attention (GA)―using annual government work reports released by municipal authorities. Firstly, feature words related to urban green development focus were extracted from domains encompassing ecological environmental protection, pollution control, energy utilization, and governance, with foundational documents―including the Action Plan for Green Development in National High-Tech Zones, the White Paper on Sustainable Urban Development, China’s Green Development in the New Era, and the CPC 20th National Congress Report―serving as benchmarks. Secondly, the feature lexicon for Local Government Green Attention was refined by integrating established repositories from the existing literature. Thirdly, employing deep neural networks, we segmented annual government work reports of prefecture-level cities into ecological, pollution, energy, and governance dimensions, and then excluded statements containing negations such as “not”, “without”, or “no” preceding feature words. Fourthly, text analysis quantified the frequency of feature words, with higher occurrences indicating greater governmental emphasis. Ultimately, standardization of these frequencies yielded the Local Government Green Attention metric.

3.2.3. Mechanism Variables

This study posits that local government green attention enhances GTFP by fostering green technological collaboration among firms (GC), accelerating green R&D involvement by public research institutions (PI), and improving green innovation quality (GQ). Methodologically, GC is measured by the natural logarithm of interfirm collaborative green patent counts, PI by the natural logarithm of public research institutions’ green patents, and GQ by the natural logarithm of green technology complexity indices.

3.2.4. Control Variables

Following Chen et al. (2022) [33] and Irfan et al. (2022) [83], this study selects the following control variables: urban scale (Pop), economic development level (Pergdp), financial development level (Findev), foreign openness (Fdi), industrial structure level (Industry), higher education level (Hedu), technological support intensity (Tec), fixed-asset investment level (Ass), and fiscal self-sufficiency ratio (Gov). The definitions and calculation methods of these variables are detailed in

Table 2. Variable Description.

Symbol	Variable Name	Variable Description
GTFP	green total factor productivity	The estimation approach follows that described in the preceding section
GA	Local government green attention	The estimation approach follows that described in the preceding section
Pop	Urban scale	ln (the year-end permanent population in cities (ten thousand people))
Pergdp	Economic development level	ln (per capita gross regional product in cities (ten thousand yuan))
Findev	Financial development level	Financial institutions’ year-end total loan balance as a percentage of city-level GDP (%)
Fdi	Foreign openness	Current-year actually utilized FDI in the city as a percentage of city-level GDP, converted at the annual average exchange rate (%)
Industry	Industrial structure level	Share of Secondary Industry Value Added in city-level GDP (%)
Hedu	Higher education level	Education expenditure as a percentage of general public budgetary expenditure at the city level (%)
Tec	technological support intensity	Science and technology expenditure as a percentage of general public budgetary expenditure at the city level (%)
Ass	fixed-asset investment level	ln (urban fixed asset investment (ten thousand yuan))
Gov	fiscal self-sufficiency ratio	Ratio of city-level general public budget revenue to general public budget expenditure (%)
GC	Level of green technological collaboration	ln (the number of green patents jointly developed by firms (patent))
PI	Green R&D involvement level of public research institutions	ln (the number of green patents applied for by public research institutions (patent))
GQ	Quality of green innovation	ln (the green technology complexity)

, and their descriptive statistics are presented in

Table 3. Descriptive statistics.

Symbol	Observations	Mean	Standard deviation	Min	Max
GTFP	2935	0.9512187	0.2202188	0	1.5014
GA	2935	0.053912	0.0707287	0	0.879376
Peo	2935	5.251333	1.919523	0	8.136518
Pergdp	2614	10.70171	0.574675	8.77292	12.57928
Findev	2935	16.84085	1.115457	9.682467	21.31105
Fdi	2935	2.514774	1.911325	0	9.604745
Industry	2935	0.4112207	0.1769475	0	0.8934
Hedu	2935	0.1761123	0.0411583	0	0.3496347
Tec	2935	0.0163514	0.0168524	0	0.2118352
Ass	2935	16.0951	2.080919	0	19.33697
Gov	2935	0.448128	0.2253365	0	1.545454
GC	2935	1.667126	1.614025	0	8.37586
PI	2935	1.448779	1.535521	0	7.32251
GQ	2935	1.554704	1.464911	0	6.866568

.

3.3. Empirical Model

To empirically identify the impact of local government green attention on GTFP, this study adopts a two-way fixed-effects model using ordinary least squares (OLS) estimation, specified as follows:

$${GTFP}_{it}=\alpha +\beta {GA}_{it}+\gamma X_{it}+u_i+v_t+{\epsilon }_{it}$$

(1)

where i denotes city and t denotes year. ${GA}_{it}$ represents the local government green attention level of city i in year t, and ${GTFP}_{it}$ stands for the GTFP of city i in year t. ß is our core coefficient of interest, capturing the impact of local government green attention on GTFP. u_i and v_t represent city fixed effects and year fixed effects, respectively. X_it denotes a set of control variables affecting urban GTFP.

4. Empirical Results

4.1. Local Government Green Attention and GTFP

Based on the theoretical analysis, this study examines how local government green attention (GA) affects GTFP using Model (1) (see

Table 4. Basic regression results.

	(1)	(2)	(3)	(4)
GA	0.1183 ** (0.0592)	0.1731 *** (0.0670)	0.1505 ** (0.0647)	0.1509 ** (0.0731)
Pop		−0.0004 (0.0053)	−0.0002 (0.0054)	−0.0004 (0.0062)
Pergdp		−0.0083 (0.0135)	−0.0349 *** (0.0111)	−0.0346 * (0.0192)
Findev		−0.0213 ** (0.0100)	−0.0199 (0.0129)	−0.0227 ** (0.0104)
Fdi		−0.0009 (0.0015)	0.0001 (0.0015)	0.0001 (0.0017)
Industry		0.1386 *** (0.0339)	0.0806 ** (0.0337)	0.0807 * (0.0420)
Hedu		−0.1171 * (0.0611)	−0.1917 *** (0.0632)	−0.1845 ** (0.0743)
Tec		0.1093 (0.1204)	0.0173 (0.1235)	0.0218 (0.1343)
Ass		−0.0032 (0.0032)	−0.0042 (0.0036)	−0.0041 (0.0037)
Gov		0.0122 (0.0178)	0.0095 (0.0228)	0.0113 (0.0206)
Constant	0.9452 *** (0.0033)	1.4133 *** (0.1826)	1.7321 *** (0.2301)	1.7746 *** (0.2394)
City fixed effect	√	√	√	√
Year fixed effect	√	√	√	√
Province-time joint fixed effect			√	√
Cluster				City
Observations	2934	2613	2570	2570
Adj.R2	0.9415	0.9464	0.9575	0.9575

Note: The standard error is in brackets. * p < 0.10, ** p < 0.05, *** p < 0.01. Round to four decimal places.

). Column (1) includes only city and year fixed effects. The significantly positive coefficient on GA (0.1183) indicates that GA boosts GTFP. Column (2) adds control variables and confirms this positive effect. Given the differences in ecological policies, development priorities, and resource constraints across provinces, Column (3) further incorporates province-time joint fixed effect to account for province-specific temporal trends. The result remains strongly significant. To address potential heteroscedasticity in city-level panel data that may lead to biased estimates, Column (4) clusters standard errors at the city level based on the specification in Column (3). The consistently positive and significant coefficients reaffirm the robustness of our findings.

Results demonstrate that increased GA significantly enhances GTFP. GA functions as an environmental governance “political signaling system”, creating fiscal incentives/constraints while conveying policy priorities. This shapes corporate perception of government preferences, influencing firms’ signal response. To secure support, firms proactively adopt emission reduction, pollution control, and renewable energy technologies. Simultaneously, enhanced GA increases green procurement, channeling social investment toward eco-innovative firms. These mechanisms collectively enhance GTFP. Using Column (2) as the benchmark result, the coefficient suggests that a 1-percentage-point increase in green attention leads to a 17.31-percentage-point rise in local GTFP. This conclusively validates Hypothesis 1 of this study.

4.2. Robustness Tests

To ensure the reliability of the conclusion that local government green attention promotes GTFP improvement, this study conducts robustness checks through the following methods.

4.2.1. Lagged Effects and Placebo Tests

Due to the long-term complexities inherent in GTFP improvement, observable outcomes may lag behind the implementation of policy. Consequently, the current GA might not exert an immediate effect on the current GTFP. To address this concern and mitigate potential endogeneity, we re-estimate the model using one-, two-, and three-period lagged GA values as explanatory variables. Results presented in Table 5, columns (1)–(3), show consistently positive and statistically significant coefficients for the lagged GA variables (lag1–lag3) on GTFP. Notably, the coefficient on GA with a three-period lag remains significant at the 10% level, indicating that while the effect of GA on GTFP exhibits persistence, its magnitude or statistical significance weakens over time. These findings confirm the robustness of our baseline finding, which supports Hypothesis 1, even after accounting for the dynamic lagged effects of GA.

To address potential omitted variable bias stemming from the concern that future GA (in the next period) could influence current GTFP, we introduce one-period-ahead GA (F1.GA) as the core explanatory variable (Table 5, column 4). The estimation results show that F1. GA has no statistically significant effect on current GTFP, which reduces concerns about omitted variable bias.

4.2.2. Re-Examination of Mean Reversion

GTFP shows a continuous increase, and its mean-reversion property can bias the baseline estimates. To address potential biases arising from mean-reversion effects, we control for the one-period lagged GTFP in Table 5, column (5). The regression results show that even after controlling for lagged GTFP, the core explanatory variable (GA) still has a significantly positive effect on current GTFP, confirming the baseline conclusions and thus establishing robustness to mean reversion.

Table 5. Robust Tests 1.

			GTFP
	(1)	(2)	(3)	(4)	(5)
L1. GA	0.1142 ** (0.0465)
L2. GA		0.1599 *** (0.0497)
L3. GA			0.1034 * (0.0529)
F1. GA				0.0601 (0.0556)
GA					0.3834 *** (0.0862)
Controls	√	√	√	√	√
City FE	√	√	√	√	√
Year FE	√	√	√	√	√
Observations	2322	2031	1731	2636	2031
Adjust_R2	0.9435	0.9426	0.9373	0.9432	0.9443

Note: The standard error is in brackets. * p < 0.10, ** p < 0.05, *** p < 0.01. Round to four decimal places.

Table 5. Robust Tests 1.

			GTFP
	(1)	(2)	(3)	(4)	(5)
L1. GA	0.1142 ** (0.0465)
L2. GA		0.1599 *** (0.0497)
L3. GA			0.1034 * (0.0529)
F1. GA				0.0601 (0.0556)
GA					0.3834 *** (0.0862)
Controls	√	√	√	√	√
City FE	√	√	√	√	√
Year FE	√	√	√	√	√
Observations	2322	2031	1731	2636	2031
Adjust_R2	0.9435	0.9426	0.9373	0.9432	0.9443

Note: The standard error is in brackets. * p < 0.10, ** p < 0.05, *** p < 0.01. Round to four decimal places.

4.2.3. Excluding Interference from Other Environmental Policies

Since environmental policies may influence GA, cities in policy pilot areas could exhibit higher GA and greater GTFP improvements than non-pilot cities, potentially biasing the results. To mitigate the confounding effects of other environmental policies, this study controls for four major additional policies using dummy variables: China’s Energy Saving and Emission Reduction Fiscal Policy Demonstration (ESER, 2011), Low-Carbon City Pilot (LCCP, 2010), Green Finance Reform Pilot (GFRP, 2017), and National Ecological Civilization Demonstration Zones (NECZ, 2016). As shown in columns (1) through (4) of Table 6, adding these dummy variables sequentially to the baseline regressions reveals that the coefficient on the core explanatory variable (GA) remains statistically significant and positive, varying only minimally. This confirms the robustness of GA’s positive effect on GTFP.

4.2.4. Accounting for Location Factors

China’s municipalities directly under the central government and provincial capitals differ markedly from other prefecture-level cities in terms of economic development, policy support, industrial structure, and technological innovation. To address potential biases related to city size and administrative status, we conduct additional estimations by excluding these cities, as presented in Column (5) of Table 6. The results reveal that GA continues to exert a significantly positive impact on GTFP, consistent with the baseline findings.

4.2.5. Revised Re-Testing with Alternative Estimation Methods

Given the potential presence of heteroscedasticity in urban samples, which may bias OLS estimates, we re-estimate the model using Generalized Least Squares (GLS) in place of Ordinary Least Squares (OLS), as reported in column (6) of Table 6. The coefficients remain statistically significant and show positive signs, further reinforcing the robustness of the statistical inferences.

GA may nonlinearly affect GTFP. To test whether this influences our core findings, we extend the specification in Column (7) of Table 6 by adding a squared term of GA. The results show that the coefficient on the linear term remains significantly positive, while the coefficient on the squared term is statistically insignificant. This suggests no empirical support for a nonlinear association between GA and GTFP. Although the quadratic term test rules out U-shaped or inverted U-shaped relationships between GA and GTFP, it does not preclude heterogeneous regression coefficients across distinct regimes. To mitigate potential estimation bias from residual nonlinearity, this study further implements a panel threshold model to detect threshold effects. The threshold existence test yields p > 0.05, demonstrating no statistically significant threshold, which implies no evidence of nonlinear effects between GA and GTFP.

Table 6. Robust Tests 2.

GTFP
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
GA	0.1714 *** (0.0436)	0.1635 *** (0.0435)	0.1729 *** (0.0433)	0.1719 *** (0.0433)	0.2940 *** (0.0922)	0.5160 *** (0.1195)	0.2291 ** (0.9613)
GA2							−0.0783 (0.0889)
Controls	√	√	√	√	√	√	√
City FE	√	√	√	√	√	√	√
Year FE	√	√	√	√	√	√	√
Obs.	2613	2613	2613	2613	2335	2628	2613
Adj.R2	0.9464	0.9465	0.9464	0.9464	0.9518	0.9826	0.9391

Note: The standard error is in brackets. ** p < 0.05, *** p < 0.01. Round to four decimal places.

Table 6. Robust Tests 2.

GTFP
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
GA	0.1714 *** (0.0436)	0.1635 *** (0.0435)	0.1729 *** (0.0433)	0.1719 *** (0.0433)	0.2940 *** (0.0922)	0.5160 *** (0.1195)	0.2291 ** (0.9613)
GA2							−0.0783 (0.0889)
Controls	√	√	√	√	√	√	√
City FE	√	√	√	√	√	√	√
Year FE	√	√	√	√	√	√	√
Obs.	2613	2613	2613	2613	2335	2628	2613
Adj.R2	0.9464	0.9465	0.9464	0.9464	0.9518	0.9826	0.9391

Note: The standard error is in brackets. ** p < 0.05, *** p < 0.01. Round to four decimal places.

4.3. Endogeneity Issues

Although the regression results confirm baseline robustness, heightened policy emphasis on GTFP may prompt governments to prioritize environmental protection, suggesting bidirectional causality with GA. To mitigate potential endogeneity concerns, this paper constructs an instrumental variable (IV) for GA by hybridizing shift-share methodology with Lewbel’s (1997) [84] IV approach. First, the annual growth rate of mean GA across all sampled local governments is computed as the aggregate shift rate (shift). Next, each city’s prior-year mean GA of other same-province cities is calculated as its share component (share). The simulated GA increment for each city-year is then derived as shift × share. Finally, applying Lewbel IV methodology, the instrumental variables are constructed from both the annual GA level per city and the cubed residuals of these increments. From a relevance perspective, the overall growth rate of GA (shift) reflects macro-level trend changes, broadly influencing individual cities’ green attention. Regarding exogeneity, the prior-year average GA of other provincial cities (share)—calculated excluding the focal city—represents aggregate variation independent of specific city characteristics. This province-wide dynamic, unaffected by any single city, affects neither GTFP nor suffers from reverse causation.

Based on this approach, the study re-estimated the regression using 2SLS (see

Table 7. Re-examination of Core Hypotheses Using Instrumental Variables.

Variable	(1) The First Stage	(2) The Second Stage
	GA	GTFP
IV	0.8802 *** (0.2226)
GA		0.2348 *** (0.0840)
Controls	√	√
City FE	√	√
Year FE	√	√
observed value	2233	2233
First-stage F-statistic	1563.74 *** (0.0000)
Adjusted R2		0.0181
Kleibergen–Paap rk LM	997.739 (0.000)
Kleibergen–Paap rk Wald F	1563.735

Note: The standard error is in brackets. *** p < 0.01. Round to four decimal places.

). The first-stage estimation yields an instrument coefficient of 0.8802, significant at the 1% level, confirming the IV’s relevance to GA. With a first-stage F-statistic exceeding 10, the significant rejection of underidentification by the Kleibergen–Paap rk LM test, and the Kleibergen–Paap rk Wald F-statistic surpassing the Stock-Yogo weak IV critical value, weak instrument concerns are eliminated. Second-stage results show a statistically significant positive coefficient of 0.2348 (1% level) of the IV on GTFP, demonstrating that enhanced GA effectively promotes GTFP after controlling for endogeneity. The instrumental variable coefficient slightly exceeds the baseline regression estimate, due to the construction of the IV by approximating overall growth with average annual growth rates. This approach compressed inter-city growth disparities and imposed a more stable dynamic trend on the instrument, marginally amplifying its estimated impact on GTFP. Combined evidence of statistical significance, relevance, exogeneity, and rejection of the underidentification test reinforces the instrument’s validity. Endogeneity-controlled results confirm that GA maintains a statistically significant positive effect on GTFP, validating Hypothesis 1 of this study.

4.4. Mechanisms

The previous sections have established, through extensive identification strategies and robustness analyses, that heightened green attention (GA) from local governments enhances GTFP. Now this section examines potential channels through which GA affects GTFP.

4.4.1. Green Technology Collaboration

Theoretical analysis in preceding sections posits that heightened GA fosters green technology collaboration among intra-city firms, leading to improved local GTFP. To validate this transmission mechanism, this study employs intra-city collaborative green patents among real-sector firms as the dependent variable. Column (1) of Table 8 shows a statistically significant positive coefficient for GA at the 1% level, indicating that increased GA strengthens green technology collaborations among real-sector firms. Column (2) demonstrates that such inter-firm cooperation enhances GTFP. The economic logic is that heightened local government green attention (GA) increases regulatory pressure, compelling firms toward green collaboration. As governments prioritize environmental issues, stricter emission standards and green technology barriers raise compliance costs. Confronted with environmental complexity and the inherent dual externalities of green innovation, firms actively seek partners to access environmental knowledge and address capability gaps [59,60,61]. These regression results demonstrate that increased GA enhances GTFP by stimulating green technology collaborative efforts across the real sector, thus providing empirical support for Hypothesis 2.

4.4.2. Green R&D Involvement of Public Research Institutions

Next, we further test the mechanism whereby GA affects GTFP by enhancing green R&D involvement within public research institutions (PRIs). We quantify PRIs’ green R&D involvement using patent counts of green inventions filed by such institutions—including universities, research institutes, and laboratories—with empirical evidence shown in Column (3) of Table 8. The results indicate that the coefficient for GA is positive and statistically significant at the 1% level, demonstrating that increased GA significantly promotes green R&D involvement in PRIs. As evidenced in Column (4) of Table 8, this enhanced involvement subsequently elevates GTFP. Green innovation entails higher knowledge complexity and more advanced technological frontiers than traditional innovation [63]. It relies more heavily on breakthroughs in fundamental research and greater involvement from universities and PRIs [46]. The research institution-led “Science-Technology” innovation model [64] facilitates corporate environmental investment through economic incentives and knowledge transfer, enabling firms to absorb green knowledge, drive innovation, and thus enhance GTFP. Moreover, as demonstrated in Columns (3)-(4) of Table 8, increased GA elevates GTFP through this enhanced green R&D involvement by public research institutes. This conclusively validates Hypothesis 3 of this study.

4.4.3. Quality Leap in Green Innovation

As GA intensifies, urban innovation entities strategically reorient their approaches—shifting focus from scaling subsidy-driven low-value green patents toward pursuing high-quality green innovations. This resultant elevation in green technology quality enhances corporate resource allocation efficiency, which in turn enables agile technological trajectory realignment and production paradigm adaptation when confronting emergent technical challenges or environmental regulatory shifts. Such dynamic responsiveness constitutes the core mechanism through which GA enhances GTFP. Regression results in column (5) of Table 8, which uses green technology quality as the dependent variable, confirm that GA exerts a statistically significant positive effect on green innovation quality as hypothesized. Empirical evidence from column (6) of Table 8 further indicates that such quality upgrading substantially elevates GTFP. Collectively, the findings from columns (5) and (6) establish a causal pathway whereby intensified GA enhances local GTFP through promoting the upgrading of green innovation quality. Therefore, these findings provide robust empirical validation for Hypothesis 4 of this study.

Table 8. Mechanisms.

	(1) GC	(2) GTFP	(3) PI	(4) GTFP	(5) GQ	(6) GTFP
GA	2.3212 *** (0.5020)		2.0713 *** (0.4292)		1.5577 *** (0.4545)
GC		0.0096 *** (0.0034)
PI				0.0078 ** (0.0035)
GQ						0.0094 ** (0.0043)
Controls	√	√	√	√	√	√
City fixed effect	√	√	√	√	√	√
Year fixed effect	√	√	√	√	√	√
Observations	2613	2613	2613	2613	2613	2613
Adj.R2	0.8755	0.4110	0.8944	0.4012	0.9435	0.3991

Note: The standard error is in brackets. ** p < 0.05, *** p < 0.01. Round to four decimal places.

Table 8. Mechanisms.

	(1) GC	(2) GTFP	(3) PI	(4) GTFP	(5) GQ	(6) GTFP
GA	2.3212 *** (0.5020)		2.0713 *** (0.4292)		1.5577 *** (0.4545)
GC		0.0096 *** (0.0034)
PI				0.0078 ** (0.0035)
GQ						0.0094 ** (0.0043)
Controls	√	√	√	√	√	√
City fixed effect	√	√	√	√	√	√
Year fixed effect	√	√	√	√	√	√
Observations	2613	2613	2613	2613	2613	2613
Adj.R2	0.8755	0.4110	0.8944	0.4012	0.9435	0.3991

Note: The standard error is in brackets. ** p < 0.05, *** p < 0.01. Round to four decimal places.

4.5. Heterogeneity Analysis

4.5.1. Disparities in Natural Resource Endowments

Based on natural resource endowments driving regional growth, local governments’ green emphasis yields varied impacts. To analyze GA effects under diverse resource conditions, this study classifies 285 Chinese cities into 115 resource-dependent and 170 non-resource-dependent groups per China’s State Council 2013 Sustainable Development Plan for Resource-Dependent Cities. We construct a “Resource” dummy (1 = resource-dependent cities) and its interaction with GA. Table 9 Column (1) shows that GA’s positive effect is significantly weaker in resource-dependent cities. These cities rely heavily on polluting traditional industries [6], where growth depends on resource extraction/processing, fundamentally undermining green governance initiatives.

4.5.2. Divergent Competition Among Local Governments

As China shifted priorities from economic growth to high-quality development, ecological performance gained greater weighting in officials’ evaluation systems. Under the political promotion tournament framework, officials face dual pressures: sustaining economic growth while complying with green development mandates. This dynamic fuels strategic environmental governance competition—local governments intensify policies to signal competence to superiors. While quantifying competition intensity directly remains challenging, variations in regional GDP rankings reflect both promotion eligibility and policy commitment. We thus measure interjurisdictional competition using per capita GDP rankings.

Following Zhou et al. (2023) [5], we rank cities within provinces by annual per capita GDP. The variable rankd is defined as city i annual ranking change in per capita GDP (year t vs. t − 1). We calculate provincial averages of Rankd, assign Com = 1 to provinces above the national average (indicating intense competition) and Com = 0 to those below (milder competition), and construct the GA × Com interaction term.

The significantly positive GA × Com coefficient (Table 9, Column 2) indicates that heightened GA boosts GTFP gains more substantially in high-competition regions. This occurs because interjurisdictional rivalry reduces the marginal returns of extensive development while enhancing the political visibility of environmental governance. To establish comparative advantages, governments deploy stringent environmental regulations and ecological investments, transforming sustainability efforts into political capital to cultivate an “ecological exemplar” profile.

4.5.3. Variations in the Degree of Intellectual Property Protection

Robust local intellectual property protection (IPP) fosters green technology collaboration and elevates innovation quality [85], triggering regional enthusiasm for green innovation. Consequently, stronger IPP amplifies the diffusion effect of government green attention (GA). Measured per Cui et al. (2023) [86] and Luo et al. (2025) [6], cities with above-average IPP are classified as high (IPP = 1), others as low (IPP = 0). Table 9 Column (3) shows that the positive IPP × GA interaction term and insignificant GA coefficient indicate effects arise only in high-IPP regions, while GA fails to drive outcomes in low-IPP areas. This validates that GA enhances GTFP more effectively under rigorous IPP.

4.5.4. Disparities in Local Government Fiscal Capacity

Local governments’ prioritization of green ecology requires building incentive and constraint mechanisms through fiscal resource allocation. Regions with sufficient fiscal capacity enhance GTFP via targeted subsidies and green technology incubation, while constrained fiscal capacity under hard budgets restricts investments to end-of-pipe treatments. To verify this, local general budget expenditure serves to measure discretionary fiscal resources, thereby categorizing cities above the median as fiscally resourced-adequate (FR = 1) and others as resourced-inadequate (FR = 0). Table 9 Column (4) results show a significantly positive FR × GA interaction coefficient alongside an insignificant standalone GA term, indicating that effects originate exclusively from fiscally resourced-adequate regions while GA remains ineffective in constrained areas. This empirically supports the inference that GA exerts a stronger effect where financial resources are robust.

4.5.5. Disparities in Technological Foundations

GA enhances GTFP through institutional signals aligned with market responses. Environmental governance’s “political signaling system” drives green technology collaboration and elevates innovation quality, but efficacy depends on regional knowledge endowments: robust technological foundations allow accumulated traditional assets to provide interfaces for green knowledge recombination, fostering efficiency-enhancing policy-driven knowledge spillovers. Consequently, GA may amplify its GTFP impact in cities with mature foundations. To test this, we construct a technological foundation dummy (TB = 1 where a city’s non-green invention patents exceed the annual national average). Table 9 Column (5) results show a significantly positive TB × GA coefficient with an insignificant GA coefficient, confirming that the effects originate exclusively from technologically advanced cities. This demonstrates that GA enhances GTFP more effectively in technologically advanced regions.

Table 9. Heterogeneity analysis.

	(1) Resource	(2) Com	(3) IPP	(4) FR	(5) TB
GA	0.1813 *** (0.0434)	0.1344 *** (0.0470)	−0.2089 (0.1765)	−0.1723 (0.1972)	−0.0565 (0.1038)
GA × Resource	−0.4035 *** (0.1515)
GA × Com		0.1344 *** (0.0470)
GA × IPP			0.2078 *** (0.0423)
GA × FR				0.1805 *** (0.0420)
GA × TB					0.2023 *** (0.0435)
Controls	√	√	√	√	√
City fixed effect	√	√	√	√	√
Year fixed effect	√	√	√	√	√
Observations	2613	2613	2613	2613	2613
Adj.R2	0.9465	0.9465	0.9466	0.9464	0.9465

Note: The standard error is in brackets. *** p < 0.01. Round to four decimal places.

Table 9. Heterogeneity analysis.

	(1) Resource	(2) Com	(3) IPP	(4) FR	(5) TB
GA	0.1813 *** (0.0434)	0.1344 *** (0.0470)	−0.2089 (0.1765)	−0.1723 (0.1972)	−0.0565 (0.1038)
GA × Resource	−0.4035 *** (0.1515)
GA × Com		0.1344 *** (0.0470)
GA × IPP			0.2078 *** (0.0423)
GA × FR				0.1805 *** (0.0420)
GA × TB					0.2023 *** (0.0435)
Controls	√	√	√	√	√
City fixed effect	√	√	√	√	√
Year fixed effect	√	√	√	√	√
Observations	2613	2613	2613	2613	2613
Adj.R2	0.9465	0.9465	0.9466	0.9464	0.9465

Note: The standard error is in brackets. *** p < 0.01. Round to four decimal places.

5. Conclusions

Amid dual pressures for economic development and environmental protection, accelerating GTFP improvement has become urgent. Analyzing 285 Chinese cities (2010–2020), this study finds local government green attention (GA) significantly boosts GTFP—a 1% GA increase raises it by 17.31%. Mechanistically, GA elevates GTFP through strengthened green technology cooperation, enhanced research institutional involvement, and improved innovation quality. Heterogeneity analyses reveal stronger effects in cities with intense governmental competition, robust intellectual property protection, sound fiscal capacity, or advanced technology, but diminished impacts in resource-dependent regions.

The findings of this study not only elucidate the role of GA in promoting GTFP but also offer tailored policy implications for its enhancement. To strengthen institutional foundations for GTFP, governmental green governance mechanisms should be rationalized. The central government must establish a dynamic GA incentive mechanism by integrating local GTFP performance and green attention into evaluating officials’ political achievements—fostering sustained ecological prioritization and catalyzing a governance transition from reactive environmental responses to proactive agenda-setting. Concurrently, real-time GTFP monitoring systems should employ digital technologies to track resource inputs, pollutant emissions, and green outputs, enhancing policy precision while preventing campaign-style governance and documentary greenwashing. Furthermore, environmental protection funds and project resources should be differentially allocated based on comprehensive assessments of regional economic development levels and ecological carrying capacities, prioritizing ecologically fragile zones and green-transition bottlenecks to resolve the paradox of equitable resource allocation yielding inequitable outcomes.

Simultaneously, green technology innovation ecosystems require synergistic development to unlock endogenous transition drivers. Governments should spearhead regional green industrial alliances or innovation platforms to accelerate cross-chain collaboration on critical technologies like energy efficiency systems, clean production standards, and integrated carbon management. Complementary institutional support should bolster universities and research institutes participating in local green planning, enhancing their roles in green technology incubation, commercialization, and standardization. Comprehensive quality monitoring systems for green innovation must also be implemented, utilizing integrated ecological–economic benefit assessments to deter firms’ strategic green innovation while maximizing marginal resource efficiency.

Finally, social supervision and public engagement mechanisms should enhance green policy implementation. Environmental information disclosure systems require refinement through standardized platforms publishing data on emissions, policy execution, and enforcement—augmenting public perception of GA while strengthening policy transparency and credibility. Standing mechanisms for public hearings on environmental issues should incorporate insights from enterprises, industry associations, and citizens, thereby improving democratic legitimacy and regulatory stability while mitigating corporate transition uncertainties stemming from policy volatility. Governments must simultaneously reinforce policy continuity to establish a predictable institutional environment enabling firms to develop long-term green strategies.

Our research acknowledges limitations inherent in the GTFP measurement framework. Current data constraints—particularly insufficient disaggregation of labor inputs, capital assets, and pollution outputs—may introduce estimation errors. Future methodological refinements will become feasible with access to granular firm-level data. Although this study utilizes urban-level Chinese statistics, cross-national and cross-cultural extensions would enhance the generalizability of the findings. Subsequent research should further leverage higher-resolution datasets to examine the transmission mechanisms through which local governments influence GTFP dynamics.