Impact of science and technology fiscal expenditure on green innovation: moderating role of environmental decentralization

To address global warming, caused by climate change, the 20th National Congress of the Communist Party of China emphasized the importance of actively and prudently working toward carbon emissions peak and carbon neutrality, and accelerating the shift toward green growth for achieving high-quality development (Li et al., 2024; Yang et al., 2023). Green innovation (i.e. innovation relating to environmentally sound technologies), possessing dual effects – technological innovation and environmental protection, plays a crucial role in mitigating pollution and carbon emissions while fostering sustainable growth (Dian et al., 2024). Hence, how to effectively promote green innovation has attracted extensive attention from scholars and policymakers in the past decades (Wei et al., 2023).

Government fiscal and taxation systems play pivotal roles in incentivizing green innovation (Shao and Chen, 2022; Huang, 2023). Particularly, science and technology fiscal expenditure (STFE) is critical in advancing technological breakthroughs and fostering talents (Chen et al., 2023; Deng et al., 2023; Liu et al., 2020a) as well as pushing green innovation further. Moreover, Chinese environmental decentralization (ED) operates within the framework of “political centralization and economic decentralization,” hence, varying degrees of ED come with different political and economic incentives (Xia et al., 2021), shaping the fiscal expenditure preferences of local governments. ED allows local governments’ discretionary powers over environmental management, boosting their motivation for environmental governance (Ran et al., 2020). This avenue encourages them to increase STFE to support green innovation, and achieve both economic development and ecological conservation objectives (Ran et al., 2020; Chen et al., 2023). However, it also provides political space for local governments to emphasize the economy over environmental protection (Zhao et al., 2022), leading to a fiscal model that emphasizes economic development while neglecting public expenditure (Liu et al., 2020a; Xue et al., 2023), thus weakening the incentive effect of STFE on green innovation. Therefore, ED may influence the relationship between STFE and green innovation, yielding either enhanced or reverse effects. However, previous studies have not fully explored this, making it difficult to accurately assess the actual effects of STFE on green innovation under the Chinese-style ED system. Hence, a partially linear functional-coefficient model is developed in this study to analyze how the degree of ED could shape the impact of STFE on green innovation, using panel data for 284 Chinese cities for the period 2007–2021.

This study offers several contributions. First, it provides an in-depth analysis and examination of the impact of STFE on green innovation with an evaluation of the external constraints that could influence the outcomes of STFE. The study innovatively uses a partially linear functional-coefficient (PLFC) model to explore the nonlinear effects of STFE on green innovation as influenced by different degrees of ED, thus supporting policymakers in formulating precise fiscal incentive policies to enhance the effect of STFE on green innovation. Second, it expands the empirical analysis of the moderating effects of ED on the relationship between STFE and green innovation, including robustness tests and heterogeneity analysis to further enrich the relevant research contents on the relationship between ED, fiscal policy and green innovation.

The remainder is structured as follows. Section 2 contains the theoretical mechanism and literature review. Section 3 is related to the method and data. Section 4 reports the empirical research results and analysis. Section 5 is the conclusion and policy implications. Figure 1 shows the theoretical framework of this study.

The decentralized fiscal policies in China empower local governments to effectively manage STFE, but existing research lacks systematic discussions on the impact of STFE on green innovation. Instead, much of the related research focuses on analyzing the effects of fiscal decentralization or fiscal policies on green development (Yang et al., 2025; Abbas et al., 2024; Chang and Wu, 2024; Deng et al., 2023; Hu et al., 2023), resulting in three mainstream perspectives, including the promotion theory, the inhibition theory and the nonlinear theory. Green innovation is a crucial indicator for measuring the level of regional green development and is also influenced by fiscal decentralization or fiscal policies (Shao and Chen, 2022; Sun and Razzaq, 2022; Wei et al., 2023; Feng and Li, 2024; Yin et al., 2025). As an important means for the government to intervene in the economy, STFE plays a crucial role in guiding technological innovation, nurturing technology talent and stimulating market vitality (Wei et al., 2023). Therefore, the green innovation effect of STFE cannot be ignor

STFE may drive green innovation in the following three aspects. The first is the R&D subsidies effect. Green innovation has the dual externalities of R&D spillover and pollution reduction (Feng et al., 2023), leading to a dilemma where enterprises face insufficient incentives to engage in green innovation. As an effective policy tool to support R&D initiatives (Chen et al., 2023), STFE can alleviate the financial constraints of enterprises in clean production and pollution control technologies through R&D subsidies (Bai et al., 2019), thus incentivizing enterprises to engage in green innovation. Moreover, government R&D subsidies serve as a signal. Under the background of “Carbon Peaking and Carbon Neutrality,” the government will intentionally increase investment in the green and low-carbon field (Deng et al., 2023; Lin and Ma, 2022). This conveys a signal to external investors that the field is worth investing in, thereby enabling enterprises to obtain external resources for green innovation (Kleer, 2010; Li et al., 2022). With the influx of relevant resources, the ability of enterprises to explore cutting-edge technologies will continuously be enhanced, thus increasing the conversion rate of green innovation outcomes.

The second is the technological progress effect. Many studies have confirmed that STFE can generate technological progress (Deng et al., 2023; Wei et al., 2023). Technological progress accelerates the phase-out of backward production sectors and promotes industrial structure (IS) toward capital and technology-intensive industries. In this process, more green innovative technologies and products will continue to emerge (Li et al., 2022; Wei et al., 2023). In addition, technological progress leads to spillover effects (Yu et al., 2024), which is conducive to driving technological innovation in surrounding areas and generating collision and integration effects with technologies from different fields. This process achieves the breakthrough of the original technological trajectory and increases the probability of the emergence of new green products and technologies. As a result, this effect promotes the diversified development of regional green technologies (Lin et al., 2023) and also contributes to forming a virtuous cycle for the development of regional green innovation.

The third is the effect of talent accumulation. In the knowledge economy era, green innovation largely relies on high-quality scientific and technological talents (Hu et al., 2022). Increasing the STFE of the government can facilitate the continuous optimization and upgradation of the structure and quality of scientific and technological human capital. This provides necessary talent support for green innovation, and it also enhances regional capacity to absorb and utilize green technologies (Hu et al., 2022). Moreover, technology-oriented talents, as a unique economic factor, tend to cluster in certain spaces and industries, generating collective learning and knowledge spillover effects. This will enhance the overall level of social-technological innovation, leading the development of green innovation and achieving the dual benefits of environmental preservation and economic advancement. In summary, H1 is proposed as follows:

Existing studies have not directly discussed the relationship between ED, STFE and green innovation. Instead, studies have predominantly focused on two themes: “fiscal expenditure, fiscal policy, fiscal decentralization, and green development” (Cai et al., 2025; Abbas et al., 2024; Fang et al., 2024; Deng et al., 2023; Sun et al., 2023) and “environmental decentralization and green development” (Hong et al., 2019; Li and Yang, 2021; Ren et al., 2023; Tang and Mao, 2023; Xia et al., 2021). Nonetheless, these studies offer valuable insights for our hypothesis. The origin of ED theory provides the theoretical foundation for incorporating ED, STFE and green innovation into a unified analytical framework. Specifically, ED theory is an extension of fiscal federalism theory in the direction of environmental governance (Oates, 1999; Wu et al., 2020a), and it is closely related to STFE policies. Meanwhile, the impact of ED on green development has been widely recognized in academia (Ran et al., 2020; Wu et al., 2020a; Wu et al., 2020b), and green innovation, as an important indicator of regional green development, also influenced by ED (Feng et al., 2022). In summary, ED is inseparable from STFE policies and is also a condition that affects green innovation. Therefore, it is necessary to clarify the inherent connection among these three. However, there exists a gap in theoretical analysis and empirical research on the relationship among these three, leaving room for examination and supplementary research. Notably, ED may manifest a “enhancing effect” as well as an “distorting effect” on STFE, thereby impacting the positive effect of STFE on green innovation. Table 1 summarizes some of the research outcomes of fiscal expenditure, ED and green development.

In terms of enhancing effect, ED amplifies the positive effect of STFE on green innovation. Under the ED system, local governments have more environmental management rights and can formulate diverse environmental protection policies tailored to local conditions, which enhances the motivation of local governments for pollution control (Jiang et al., 2023; Li et al., 2021a). This incentivizes them to augment STFE to support green technology research and development, as well as to cultivate and introduce high-level technology talents to provide intellectual support for green innovation (Feng et al., 2020; Ren et al., 2023). Consequently, ED realizes a dual enhancing effect on STFE, effectively reviving green innovation. In addition, there exists a spatial demonstration effect of government environmental governance (Feng et al., 2020). Particularly, under the advocacy of the green GDP concept, local governments will engage in a competitive race to the top in environmental governance (Ran et al., 2020), leading to a series of competitions revolving around supporting green R&D activities. This is ultimately reflected in the continuous expansion and rigid enhancement of STFE (Li et al., 2021b), strengthening the positive impact of STFE on green innovation.

From the perspective of the distorting effect, ED suppresses the positive effect of STFE on green innovation. ED may lead to distorted incentives and inadequate constraints of local governments in environmental governance (Hao et al., 2022), resulting in an impairing effect on STFE, thereby hindering green innovation. Regarding incentive distortions, excessive decentralization of environmental authorities allows local governments to gain greater resource allocation advantages in the promotion championship. The inherent incentive incompatible with economic development and environmental protection often leads local governments to prioritize economic growth (EG) over environmental well-being (Li and Yang, 2021), resulting in increasing constructive expenditure and reducing STFE (Wei et al., 2023). This directly weakens the momentum for green innovation. In addition, the human, financial and material resources generated by ED, are all dependent on local fiscal policy. As local governments are constrained by limited fiscal budgets, they tend to prioritize short-term visible production investments that yield immediate results, while cutting STFE that cannot rapidly boost GDP (Jiang et al., 2023). This further exacerbates the “distorting effect,” hindering the development of green innovation. In terms of insufficient constraints, excessive ED prevents higher-level governments from effectively constraining lower-level governments, leading to issues such as lax environmental law enforcement and free-riding in environmental governance (Hong et al., 2019; Tang and Mao, 2023; Yang et al., 2021). This prompts local governments to make environmentally opportunistic choices of emphasizing investment and underestimating environmental protection, causing them to reduce STFE to support green innovation. In conclusion, H2 is proposed as follows:

To verify H1, this article adopts the Ordinary Least-Squares (OLS) model to test the positive effect of STFE on green innovation, as shown in equation (1):

where GIN represents green innovation, STFE represents STFE, X represents a series of control variables, u and λ are the city and time fixed effects, ε is the random error and i and t represent region and time, respectively.

Meanwhile, this study preliminarily tests H2 by constructing a moderation effect model to examine the moderating effect of ED on the effect of STFE on green innovation. To thoroughly evaluate the moderating effect of ED, the study introduces ED into the model in two different forms: as a continuous variable and as a dummy variable, as shown in equations (2) to (3), respectively. Here, ED represents the continuous variable, while dummy_it represents the dummy variable. If the level of ED surpasses the average value, dummy_it takes the value of 1; otherwise, it takes the value of 0:

Nevertheless, both (2) and (3) are linear regression models, which may suffer from model misspecification problems because of the linearity assumptions (Lin and Ma, 2022). The marginal effect of ED may not strictly follow the hypothesized linear patterns. In contrast, the partially linear functional-coefficient (PLFC) model relaxes the linearity and homogeneity assumptions of linear models (Lin and Ma, 2022), adapting to nonlinear structural assumptions and capturing nonlinear relationships between variables. Therefore, this paper refers to the research of Du et al. (2021) and adopts the PLFC model to solve the potential model misspecification problem. Specifically, we let STFE_it enters the model with coefficients being functions of ED_it, aiming to identify the nonlinear moderating effect of ED. This model is denoted as equation (4):

The specific estimation procedure follows the research of Du et al. (2021), and the detailed steps are outlined as follows:

Step 1: Take the first-time difference of equation (4):

Step 2: Approximate the varying coefficient function g(ED_it) by a linear combination of k base functions:

where p(ED_it) represents a k*1 vector of base functions, and θ represents a k*1 vector of unknown parameters. When k grows large, there exists a linear combination of p_i(ED_it) that can well approximate smooth function g(ED_it), and the approximation MSE will be as small as possible. Then, equation (5) can be rewritten as equation (7):

where $Δμit=Δεit+Δ(g(EDit)STFEit)−Δ(STFEitp(EDit))′θ$⁠, which denotes the series approximation error.

Step 3: Calculating the least-square estimators:

where, $ΔX∼′=[ΔX11′,Δ(STFE11p(ED11))⋮ΔXNT′,Δ(STFENTp(EDNT))]$⁠, and $ΔGIN∼=[ΔGIN12⋮ΔGINNT]$

What’s more, the coefficient function g(·) can be estimated as:

Equation (9) denotes the estimated value of the function coefficient g(ED_it) of STFE in equation (4), aiming to capture the nonlinear impact of STFE on green innovation under different levels of ED.

The dependent variable is green innovation (GIN). Different from traditional innovation, green innovation emphasizes the use of new technologies to achieve efficient utilization of resources and effective pollution reduction, while obtaining corresponding economic performance (Liu et al., 2020b). Based on the connotation of green innovation, and previous researches (Lin and Ma, 2022; Xu et al., 2021), this study uses the number of green patent applications per 10,000 people to measure green innovation. For two reasons: first, while patents may not fully reflect the connotation of green innovation, they are a key output of technological innovation and reflect enterprises’ innovation capacity, particularly their emphasis on pollution control and green development (Wei et al., 2023). Second, considering that patent granting requires detection and payment of annual fees, there is more uncertainty, instability and timeliness, whereas patent application data are more stable, reliable and timely than the number of grants. To enhance the robustness of the research conclusions, this paper also uses the number of green patents granted per 10,000 people (GIN2) as another measure of green innovation. Green patents are determined based on the IPC Green List provided by the International Patent Classification, with data sourced from the Chinese Research Data Services Platform (CNRDS).

The independent variable is STFE, represented by the share of science and technology expenditure within the total fiscal expenditure. This metric reflects the significance of science and technology expenditure relative to the overall fiscal expenditure of local governments (Wei et al., 2023). The academic community primarily employs two methods to measure STFE levels. First, the measurement approach adopted in our study, which uses the proportion of STFE to total fiscal expenditure (Wei et al., 2023), as this method avoids the biasness caused by regional differences in total fiscal expenditure. When research questions focus on government resource allocation preferences or science and technology policy priorities, scholars adopt this method to directly reflect the government’s emphasis on technological innovation. Second, the per capita STFE measurement (Bai and Zheng, 2024), which may be affected by population distribution, and can mask aggregate differences. When research questions emphasize on analyzing the social benefits of science and technology investment or regional innovation equity, scholars generally employ this method. Considering that our study leans toward policy orientation, we prefer the first measurement method.

The moderating variable is ED, which originates from the decentralization system and environmental federalism and is specifically manifested in the central and local division of environmental management rights (Hao et al., 2021). Given the remarkable differences in environmental regulatory systems across nations, it is difficult to choose a single indicator of ED that is consistent with its theory and practice (Wu et al., 2020a), and existing research has not yet formed a universal indicator to measure the degree of ED. Currently, the mainstream method for measuring the Chinese regional ED index is using the number of personnel in environmental protection departments (Hao et al., 2021; Wu et al., 2020a; Wu et al., 2020b). The reasons are as follows: the essence of ED lies in the allocation of personnel in environmental governance. Using personnel distribution to measure may better embody the essence of ED, and using personnel distribution to measure decentralization is also an internationally accepted practice. In addition, the scale of environmental protection personnel is relatively stable, which provides us with a convenient assessment of the level of ED (Hao et al., 2021). Therefore, this paper analogizes this methodology to depict the degree of ED in Chinese cities. Considering that environmental protection departments need to collaborate with public infrastructure departments such as water conservancy to effectively play their role, there are some limitations to the accessibility of data in Chinese cities. This paper uses the distribution of employees in water conservancy, environmental and public facility management industries to measure the degree of ED. The specific calculation formula is shown in equation (9):

where CEP_it denotes the number of employees in the water conservancy, environmental protection and public facilities management industry for city i in year t, and NEP_t be the national aggregate for the same industry and year. POP_it represents the population size of city i in year t, and GDP_it refers to the gross domestic product of city i in year t. POP_t represents the total population size of the country in year t, and GDP_t represents the gross domestic product at the national level in year t. [1−(GDP_it/GDP_t)] is the economic scale scaling factor, used to minimize the impact of relative economic scale on the degree of ED, and the corresponding endogenous problem could be to some extent avoided (Wu et al., 2020a).

To control the influence of other factors on green innovation, the following control variables are selected based on the relevant determinants theory and literature on green innovation. Foreign direct investment (FDI) is measured by the ratio of actual FDI to GDP. EG is represented by per capita actual GDP. Infrastructure (IN) is gauged by per capita urban road area (square meters). IS is represented by the proportion of the output value of the secondary industry to GDP. The level of industrial development (IND) is assessed by the number of industrial enterprises above a designated size.

To ensure data set integrity, firstly we have excluded 9 out of 293 prefecture-level cities in China showing significant missing data across multiple years. Therefore, based on the availability of data, this study takes 284 prefecture-level cities as research sample for the period 2007–2021. Data were primarily extracted from the China Urban Statistical Yearbook, China Statistical Yearbook, EPS database and the Chinese Research Data Services Platform (CNRDS). In case of the sampled cities, missing values for the key variables accounted for about 5% of the total sample and these were first supplemented by the data mined from the official municipal statistical bureau websites, while remaining gaps were filled through linear interpolation. As one of the most fundamental and widely applied interpolation methods (Zhao et al., 2022), this approach estimates unknown values by establishing linear relationships between adjacent known data points. Its core advantages include computational efficiency, satisfactory precision and effective avoidance of numerical oscillations associated with higher-order interpolation. The descriptive statistics for all variables are presented in Table 2.

Table 3 reports the baseline regression results for STFE’s impact on green innovation based on equation (1). In column (1), the regression coefficient of STFE is significantly positive (α₁ = 33.2463, p₁ < 0.01). Specifically, increasing STFE by 1 unit will increase green innovation by 33.2 units. This indicates that STFE has a positive effect on green innovation, thereby validating H1. This is mainly attributed to STFE bringing significant R&D subsidy effect and technological progress effect to green innovation (Wei et al., 2023), helping catalyze the implementation of green innovation outcomes.

For robustness, three methods are used to test the empirical results. First, we address potential endogeneity by using the two-stage least square (2SLS) method. Referring to the study by Wei et al. (2023), provincial-level STFE is chosen as the instrumental variable for the independent variable. As shown in column (2), the STFE coefficient remains significantly positive (α₁ = 163.8634, p₁ < 0.01). Second, we include lagged independent variables to account for the delayed effects of STFE on green innovation. In column (3), both current and lagged STFE coefficients are significantly positive, suggesting that the impact of STFE on green innovation is continuous. Third, we replace the dependent variable with GIN2, measured by green patents per 10,000 people, in column (4). The result shows that the STFE coefficient remains significantly positive (α₁ = 14.9912, p₁ < 0.01). These results confirm the robustness of the baseline regression.

Table 4 presents the results of the moderating effect of ED. In column (1), the estimated coefficient of ED*STFE is not significant (α₂ = 4.0492, p₁ > 0.01), indicating that ED has no moderating effect. In column (2), the coefficient of dummy*STFE is significantly positive (α₂ = 36.9369, p₁ < 0.01), indicating a positive moderating effect of ED above the mean level. Meanwhile, to ensure the robustness of the results, the dependent variable is replaced with GIN2 in columns (3) to (4) with no substantive changes in the direction and significance of the interaction term coefficients. The analysis suggests that introducing ED into the moderating effect model as both a continuous and a dummy variable leads to significant differences in results. This inconsistency reflects the traditional linear model may not fully reflect the real impact of STFE on green innovation under different ED degrees. Therefore, the PLFC model is used to replace the traditional linear model in subsequent research.

Figure 2(a) illustrates the function coefficient of g(ED_it) in the estimation of the PLFC model, showing the nonlinear impact of STFE on green innovation under different ED levels. The analysis reveals three intervals: when ED is below 0.5, the 95% confidence interval of the g(ED_it) function coefficient always contains zero, indicating that the impact of STFE on green innovation is not significant; when ED is between 0.5 and 6.5, the positive impact of STFE on green innovation strengthens with the increase of ED; and beyond 6.5, the positive impact weakens. The overall impact exhibits an inverted U-shaped pattern. To enhance the robustness of the findings, this section re-plots the graph with the dependent variable replaced by GIN2, resulting in Figure 2(b). Comparing Figures 2(a) and 2(b), shows that, while the fluctuation amplitude differs, the fundamental trend of STFE’s impact on green innovation remains consistent, characterized by “insignificant-promotion strengthening-promotion weakening,” confirming the robustness of the results.

The above results indicate that the positive effect of STFE on green innovation exhibits nonlinear characteristics influenced by ED. When ED is at an appropriate level, STFE has the greatest positive impact on green innovation, and this is mainly attributed to the background of the Chinese ED system. The Chinese-style ED is a microcosm of the political centralization and economic decentralization system between the central and local governments. As ED levels increase, local governments gain environmental governance rights, which helps them fully utilize their information advantages to formulate environmental policies tailored to local conditions. Especially in the context of high-quality economic development, environmental quality assessment is gradually linked to the promotion of local officials. Out of a dual consideration of political positioning and sustainable economic development, local governments will increase STFE to support enterprises in conducting green innovation, achieving dual benefits for the economy and environment. However, as ED increases further, it creates an “distorting effect” on STFE, weakening STFE’s positive impact on green innovation. Excessive ED allows local governments to prioritize economic development over environmental protection. It induces local governments to make environmentally opportunistic choices of emphasizing investment and underestimating environmental protection, leading to reduced STFE and weakened green innovation.

Resource endowment differences lead to disparities in urban IND, fiscal expenditure structure, and environmental governance. This may cause ED to play a different moderating role between STFE and green innovation. Therefore, based on the classification in the National Sustainable Development Plan for Resource-based Cities (2013–2020), this study categorizes the sample cities into resource-based and nonresource-based groups. Then, analyze the changes and discrepancies in the coefficients of the g(ED_it) function for the two groups of cities. Figures 3(a1) and 3(a2) respectively depict the impact of STFE on green innovation in resource-based and nonresource-based cities under different ED levels. Comparing the above two figures, it is evident that as ED increases, the impact of STFE on green innovation in resource-based cities is significantly positive only within the interval of 1.5 < ED < 2.8, while the majority of other intervals remain insignificant. Conversely, nonresource-based cities display a positive inverted U-shaped pattern. To enhance the robustness of the conclusions, this section re-plots the graphs with the dependent variable replaced by GIN2, resulting in Figures 3(b1) and 3(b2). Comparing them with Figures 3(a1) and 3(a2), the moderating effect of ED remains unchanged across both resource-based and nonresource-based cities, confirming the robustness of the graphical results.

The results above can be attributed to disparities in resource endowments. Resource-based cities benefit from abundant natural resources and ED brings them greater resource allocation advantages, deepening their resource dependence. The booming resource industry has lower demands for technology and labor quality and tends to neglect technological innovation and talent investment. Consequently, there is minimal STFE growth and insufficient support for green innovation, resulting in stagnation and limited progress in green innovation. Furthermore, the economic development of resource-based cities is driven by secondary industry, which faces severe environmental issues. ED affords them a pollution shelter, reinforcing the industrial path dependence of resource-based cities and further diminishing the motivation to increase STFE to support green innovation, leading to stagnation in green innovation. Conversely, nonresource-based cities, with lower resource dependence, experience weaker crowding-out effects of ED on STFE. Their IS, with a smaller secondary sector and larger tertiary sector, makes green technological innovation crucial for EG. ED thus has a “compensatory effect,” motivating local governments to increase STFE for green innovation. However, excessive ED may distort incentives and lead to fre-riding behavior, reducing its effectiveness.

In this section, a more comprehensive set of graphics is drawn based on the PLFC model to illustrate the impact of STFE on heterogeneous green innovation under different ED levels. To achieve this, heterogeneous green innovation is divided into green innovation quality (GINF) and quantity (GINS), measured by the number of green invention and utility model patent applications per 10,000 people, respectively. Furthermore, the sample cities are grouped into lower, medium and higher levels of ED based on the tertiles of ED level from 2007 to 2021. Given the initial implementation of environmental objectives accountability in the 11th Five-Year Plan. The increasing emphasis on green development in the 12th and 13th Five-Year Plans is expected to influence the moderating effect of ED. Hence, this section focuses on the 12th (2011–2015) and 13th (2016–2020) Five-Year Plans as research intervals, referred to as L1 and L2.

The horizontal and vertical axes of Figures 4(a1) and 4(a2), respectively, reflect the influence of STFE on the quantity and quality of green innovation under different ED levels. Specifically, the changing trend of colors in periods L1 and L2 in the figures indicate that, in the low and medium ED groups, STFE has little significant impact on both the quantity and quality of green innovation. However, in groups with higher ED levels, while some cities show that STFE only has a positive impact on the quantity of green innovation, the majority exhibit positive effects on both the quality and quantity of green innovation. To strengthen the robustness of the conclusion, this section replaces the dependent variable measurement methods, using the number of green invention and green utility model patents granted per 10,000 people to represent green innovation quality (GINF2) and quantity (GINS2), respectively. The results are shown in Figures 4(b1) and 4(b2). Compared to Figures 4(a1) and 4(a2), it is evident that in the overwhelming majority of cities with high levels of ED, STFE still has a positive impact on both the quality and quantity of green innovation. The above facts prove that the graphical results in this section are robust.

During the study period, cities with higher ED saw STFE’s dual effect of improvement of quantity and quality on green innovation. This may be attributed to the 12th and 13th Five-Year Plans, where environmental indicators became mandatory performance targets for officials, shifting Chinese environmental governance from soft to hard constraints. This shift prompted local governments to prioritize high-quality economic development, fully utilizing the advantages of ED to promote STFE’s support for green innovation.

In this study, a partially linear functional-coefficient model is developed to explore how the degree of ED could shape the impact of STFE on green innovation using the panel data of 284 Chinese cities spanning from 2007 to 2021. The findings reveal that when ED increases, the impact of STFE on green innovation exhibits an inverted-U trend specifying a moderate degree of ED is more conducive to harnessing the beneficial effect of such expenditure. In addition, the moderating effect of ED on the relationship between STFE and green innovation varies with the heterogeneity of urban resource endowment. For resource-based cities, ED has little impact on the green innovation effect resulting from STFE. However, for nonresource-based cities, as the level of ED increases, the positive impact of STFE on green innovation also exhibits an inverted U-shaped look. Moreover, during the 12th and 13th Five-Year Plans periods, cities in China characterized by higher levels of ED have effectively enhanced the quantity and quality of green innovation (patent outputs) through STFE.

Based on the empirical findings, some policy recommendations are proposed as follows. First, it is necessary to establish a comprehensive STFE system to support green innovation. Specifically, the government can provide financial support to enterprises through initiatives such as establishing green innovation guidance funds and green technology support projects. In addition, the government could establish incentive mechanisms for STFE, provide subsidies or tax incentives for green innovation by enterprises, and effectively leverage its clean effect on pollution control. Finally, the government may strengthen the supervision and management system of STFE and prevent enterprises from manipulating green R&D funds or rent-seeking behaviors. Continuous monitoring and evaluation as well as tracking the research on science and technology supported by fiscal expenditure can ensure its rationality, security and efficiency.

Second, it is required to continuously optimize the division of environmental authority between the central government and local governments. The central government should grant reasonable environmental management authority to local governments and strengthen supervision over local environmental protection efforts. Specifically, the environmental administrative power of local governments is expected to be increased, including authority in areas such as environmental planning and investment, to fully mobilize local motivation for environmental governance. Simultaneously, the central government should implement regular supervision and accountability of environmental protection at the municipal levels. This practice can avoid problems such as incentive distortion and different constraints in local government environmental governance due to excessive ED.

Third, tailored environmental management policies should be taken for cities with different resource endowments. For resource-based cities, it is advisable to partially centralize environmental management authority to address local protectionism in environmental governance. Meanwhile, it is essential to promote the transformation of ISs and develop green industries in resource-based cities. For nonresource-based cities, it is necessary to appropriately delegate environmental management authority, and strengthen regular supervision and management of regional environmental governance. Then, the regional green innovation capabilities can be effectively enhanced. In addition, nonresource-based cities should further drive the development of regional green and clean industries to foster a virtuous cycle of green innovation.

Fourth, it is crucial to establish a diversified local government performance evaluation system that combines both constraints and incentives. The central government should increase the weight of indicators such as technology innovation investment and environmental performance in the official assessment process, which can amplify the “enhancing effect” of ED while mitigating the “distorting effect.” This can help effectively encourage local governments to employ fiscal support for the improvement of green innovation. Simultaneously, a tailored performance evaluation system should be established to avoid reverse incentives in the evaluation system. The city differences should be fully considered when designing performance evaluation indicators tailored to the fiscal capacity, resources and environment of each city.

Though the study critically explores the nexus between STFE and green innovation guided by ED, it is not free from limitations. Due to data limitations, the study uses single indicators for green innovation and ED, which may not fully capture these concepts. Future research could employ composite indicators for better measurement. In addition, while green fiscal expenditure is closely related to this study, there is currently no systematic record of green fiscal expenditure data at the prefecture-level cities in China. Hence, the collection and application of this data might be introduced in future studies. Moreover, though the study considers FDI, EG, infrastructural and INDs in the model, future studies could add more variables like human capital development, financial development and others to study the influence of STFE on green innovation. Finally, but not the least, while this research establishes that STFE enhances green innovation, moderated by ED, it is crucial to acknowledge that STFE supports both green and nongreen innovation. The focus of this study is on green innovation, which is related to environmentally favorable outcomes, such as decreased emissions or sustainable technologies. However, STFE may encourage nongreen innovation such as improvements of nonenvironmentally focused sectors that may have adverse environmental impacts. Therefore, future studies could investigate the comparative impacts of STFE on these two types of innovation to offer a more widespread understanding of its net environmental effects. Such valuable understandings would assist policymakers in boosting fiscal expenditures to prioritize sustainable outcomes.

Funding: This paper is supported by the Key Project of the National Social Science Fund of China titled ‘The Chinese Path of Modern State Building: Tentative Answers from Local Initiatives’: No: 22WZZAOO1.