How Do Innovation-Driven Policies Affect Urban Green Land Use Efficiency? Evidence from China’s Innovative City Pilot Policy

Abstract

China has already joined the ranks of innovative nations. Accelerating technological innovation to lead a green transformation in land use is an urgent requirement for promoting ecological civilization and, in turn, driving high-quality economic development. This study examines urban data spanning from 2006 to 2021, focusing on cities classified at the prefecture level or above. Employing the Chinese Innovative City Pilot Policy (ICPP) as a quasi-natural experiment, this study utilizes a super-efficiency Slack-Based Measure (SBM) model that incorporates undesirable outputs to assess Green Land Use Efficiency (GLUE). Additionally, a multi-period Difference-in-Differences (DID) model, combined with a mediation effect model, is employed to evaluate the influence of innovation-driven policies on GLUE. The findings are as follows: (1) Although GLUE showed variability throughout the study period, it generally trended upwards, with significant improvements noted in the eastern regions and coastal city clusters. (2) Innovation-driven policies have effectively enhanced urban GLUE, a conclusion supported by extensive robustness tests. (3) The heterogeneity investigation indicates that the ICPP’s impact on GLUE is more significant in cities with advantageous geographic locations, increased environmental awareness, and strong market potential. (4) A mechanism analysis demonstrates that the ICPP positively influences GLUE by reducing urban sprawl and promoting the concentration of digital service industries. Based on these results, this study proposes policy recommendations aimed at refining innovation-driven approaches to improve urban GLUE. These recommendations are pivotal in promoting a green, low-carbon transformation in China’s economic and social development.

1. Introduction

Land constitutes the material basis for human survival and serves as a critical foundation for constructing an ecological civilization [1]. With the acceleration of urbanization worldwide, land expansion has engendered both positive and negative consequences in developed and developing countries [2]. Positive outcomes of land expansion include attracting investment, fostering entrepreneurial opportunities, and enhancing social welfare [3]. Conversely, rapid land expansion can lead to overexploitation of resources, elevated environmental pollution, and increased carbon emissions [4]. As the world’s largest developing country, China has undergone rapid urbanization since the initiation of the reform and opening-up policy [5], with the urbanization rate rising from 17.92% in 1978 to 66.16% in 2023, thereby significantly contributing to the nation’s economic development and social progress. However, China’s low-density ‘leapfrog’ spatial expansion policy has encroached upon green spaces, including arable and forest lands, resulting in adverse effects on ecological integrity and land-use efficiency [6]. In this context, improving the green utilization efficiency of land is not only essential for the optimal allocation and effective use of urban land resources but also imperative for promoting the comprehensive green transformation of economic and social development.

Green land use refers to adopting green development principles throughout the land use process to harmonize economic, social, and ecological benefits; the efficiency of green land use is pivotal for resolving production–structure contradictions, achieving ‘dual-carbon’ goals, and promoting urban sustainability [2]. Therefore, improving GLUE is crucial to mitigate the conflict between urban green development and land utilization. Studies have demonstrated that innovation constitutes the primary productive force and an effective mechanism for enhancing GLUE in urban areas [7]. However, the innovation process entails substantial uncertainties and externalities, necessitating government provision of R&D subsidies, tax incentives, and other supportive measures to mitigate associated risks. National policies play a critical safeguarding role in the innovation process [8]. Notably, the diversity of green policy instruments beyond direct innovation incentives has been extensively studied. For example, financial incentives for building energy efficiency retrofits accelerated market adoption of low-carbon technologies through subsidized structural upgrades [9]. Subsidies for electric vehicle purchases significantly increased clean transport penetration and indirectly reduced land demand for fossil fuel infrastructure [10]. Dedicated policies for the hydrogen industry enhanced spatial land use intensity [11]. These studies indicate that green incentive policies must be tailored to sector-specific characteristics, and their synergies can provide multidimensional support for green land use. Consequently, the Chinese government has proactively reformed and enhanced the innovation environment, introducing timely policies to foster innovation. The National Innovative City Pilot Policy (ICPP) exemplifies China’s innovation-driven development strategy, aiming to refine the innovation policy framework, address green and low-carbon development challenges, and position innovation as the core driver of urban development. In the context of green development, leveraging the construction of national innovative cities presents an opportunity to evaluate whether innovation-driven policies can strategically support urban land green utilization, and to elucidate the mechanisms through which they synergistically drive land green development and the economic and social transition towards a green, low-carbon economy.

The pertinent literature for this investigation is categorically divided into two primary groups. The initial category concentrates on gauging GLUE and exploring its determinants. The notion of land use efficiency has long been a focal point of scholarly study. Certain academics evaluate this efficiency by calculating the economic output per land area unit [12]. This approach, however, is considered overly reductive and fails to encapsulate the complexities inherent in land use systems. To surmount this shortcoming, other investigators have devised intricate indicator frameworks grounded in land use’s fundamental principles, employing methodologies like Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis [13]. Traditional DEA models, however, fail to consider undesirable outputs such as pollutant discharges. To address this issue, Tone (2002) [14] improved the conventional DEA model by incorporating the SBM model, which accounts for undesirable outputs like sulfur dioxide and wastewater emissions, thus expanding the framework to include GLUE [15]. Expanding on this groundwork, the super-efficiency SBM model enhances the SBM model by addressing slack and radial inaccuracies present in conventional DEA models. This advancement allows for the evaluation of decision-making units with efficiency ratings surpassing one, resulting in more accurate assessments [16]. This methodology has emerged as the predominant approach for calculating GLUE. Consequently, identifying the factors influencing GLUE has become a focal area of research. Researchers have found that exogenous policy impacts, such as big data pilot zones [17] and smart city construction [18], are beneficial for enhancing GLUE. Likewise, advancements in regional integration [19], digital finance [20], and land urbanization [21] have been demonstrated to produce comparable benefits.

A body of research specifically assesses the policy implications of the ICPP, with a primary emphasis on its effects on innovation and the environment. Studies have found that the ICPP significantly enhances innovation performance. At the macro level, ICPP implementation attracts increased investment in innovation resources, improves innovation outcomes, and elevates urban innovation levels [22]. However, it may exacerbate innovation disparities between cities [23]. At a micro level, the ICPP significantly boosts firms’ patent output and innovation capabilities [24]. However, government innovation policies may suppress corporate innovation incentives [25] and displace private investment [26]. The second key area of research concerns the environmental effects of ICPP. Empirical research indicates that ICPP differs from low-carbon and smart city pilot policies, as it emphasizes principles and goals centered on green development and innovation. It effectively fosters green technological progress and enhances urban environmental quality [27]. Additionally, ICPP significantly improves energy efficiency [25], facilitate the transition of urban residents toward green lifestyles, and contribute to breaking carbon lock-in [28], thus advancing green and low-carbon urban development.

In summary, extensive research has been conducted on urban GLUE, confirming the positive effects of innovation-driven policies on environmental governance, thereby providing a solid theoretical and empirical foundation for this study. However, there remain significant gaps. First, the quantitative assessment of GLUE is still underdeveloped, as most studies rely solely on the “three wastes” to capture non-desired outputs. Given that green and low-carbon development are complementary and closely integrated in China’s policy context, carbon emissions should be incorporated into the non-desired outputs to enhance the GLUE indicator system. Second, current assessments of the ICPP’s environmental effects mainly focus on haze mitigation, carbon reduction, green economic efficiency, and carbon emission efficiency, yet there is no clear conclusion as to whether this innovation-driven policy promotes urban GLUE. Moreover, existing studies fail to propose concrete pathways for achieving green land use from both the innovation-driven and land use perspectives. To address these gaps, this paper treats the ICPP as an exogenous policy shock and utilizes panel data from 282 Chinese cities from 2006 to 2021 to examine its impact on urban GLUE and the underlying mechanisms, from both theoretical and empirical perspectives. Furthermore, this study seeks to address the following questions: (1) As urban land is the spatial carrier of regional green development, can the environmental governance effects of innovation-driven policy be reflected in urban GLUE? (2) If so, through what mechanisms do these effects occur? (3) Do the impacts vary across cities with different characteristics?

Against this backdrop, this paper aims to explore the intrinsic linkage between ICPP and GLUE, and its potential marginal contributions over existing studies are as follows: (1) From a land use perspective, this study refines the GLUE indicator system and systematically investigates the role and internal mechanism of innovation-driven policies on GLUE, thereby expanding the research domain on the relationship between pilot policies and GLUE, and offering new insights into achieving green and low-carbon land development. (2) To mitigate potential biases in traditional two-way fixed-effects estimations, this paper employs multiple state-of-the-art heterogeneity-robust estimation models to evaluate the overall effect of innovation-driven policies on GLUE, thereby enhancing the robustness and credibility of the empirical findings. (3) To deepen the understanding of the causal relationship between innovation-driven strategies and GLUE, this paper pioneers the identification of internal mechanisms by analyzing two key channels: the urban sprawl effect and the digital services agglomeration effect, thereby aiming to open the “black box” of how pilot policies influence GLUE. (4) Taking into account city-level heterogeneity, this study rigorously examines how the effects of pilot innovation city policies vary across cities with distinct geographic locations, environmental priorities, and market structures, aiming to provide valuable insights for policymakers in designing more targeted innovation strategies.

2. Theoretical Examination and Research Hypotheses

2.1. The Effects of ICPP on GLUE

The cultivation of innovative cities forms a strategic component of China’s broader goal to evolve into an innovation-oriented nation, with key focuses on nurturing innovation and advancing green, low-carbon development [27]. This is consistent with the fundamental principles of GLUE. The policy influences GLUE through two main channels. First, economic incentives: the pilot policy encourages financial support for research and innovation from local governments and financial institutions, broadening funding channels for green innovation and reducing financing barriers for innovative enterprises. This initiative helps alleviate financial constraints on green innovation [25], creates a more conducive urban innovation environment, increases the efficiency of energy usage [29], and optimizes land use structures [30]. Furthermore, the policy facilitates the evolution of digital technologies within urban environments, enabling enterprises to integrate innovations like the industrial internet, big data, and cloud computing into the progression of intelligent energy systems. These technologies significantly improve the precision of monitoring, forecasting, and managing environmental pollution and carbon emissions [31], ultimately aiding in the enhancement in GLUE. Second, in terms of evaluation and regulation, the “Indicator System for Constructing Innovative Cities” specifies that energy consumption per GDP and urban environmental quality are essential indicators. These rigorous criteria force high-emission industries to innovate and upgrade [25], streamline industrial structures, and boost land use efficiency. Hence, we formulate the following research hypothesis 1:

H1. 

ICPP can effectively enhance the city’s GLUE.

2.2. Mediating Mechanism Analysis of the Impact of ICPP on GLUE

2.2.1. Urban Sprawl Effect

As previously highlighted, the relationship between urban sprawl and GLUE remains ambiguous. On one hand, controlled urban sprawl can generate “economies of scope”, fostering communication and collaboration among industries and enhancing both the scale and quality of economic agglomeration. Moreover, urban sprawl can stimulate the growth of the “circular economy” and “sharing economy”, enhance the effects of environmental governance, and boost the efficiency of public facility usage and land value per unit area [1]. On the other hand, within the context of urbanization and “land finance”, certain cities, motivated by economic growth, have implemented expansive and unregulated land use policies, ultimately reducing GLUE [2]. The development of innovative cities could exert varying impacts on urban sprawl. Sprawl typically occurs when urban expansion outpaces population growth [32]. ICPP can foster industrial agglomeration and expedite the development of critical infrastructure, such as high-speed railways, generating a strong “siphon effect” on surrounding cities that stimulates population growth [33], thereby curbing urban sprawl. On the other hand, ICPP elevates urban digital technology levels, accelerates information dissemination, reduces the significance of geographic distance, and reshapes human–land dynamics, encouraging population dispersal from urban centers and thereby accelerating urban sprawl [34]. Accordingly, we propose research hypothesis 2:

H2. 

ICPP can indirectly impact GLUE through urban sprawl, but the effect is uncertain.

2.2.2. Agglomeration Effect

The ICPP supports the improvement in urban digital foundation, advancing the informatization of pilot cities and providing a solid informational foundation for the digital service industry [35]. This policy facilitates the concentration of talent and digital technologies within the digital service sector, ultimately enhancing GLUE [17]. On one hand, the digital service sector is highly integrative, merging technologies like artificial intelligence, big data, and the Internet of Things, which support the digital transformation of business clusters [36]. These digital tools optimize supply chains, bolster the integration of digital and traditional industries, and support industrial restructuring. Such progress leads to more efficient utilization of land, promotes intensive land use, and ultimately contributes to the improvement in GLUE in urban settings [17]. On the other hand, ICPP has promoted a green lifestyle transformation among residents, increasing the demand for green and eco-friendly products [28]. The agglomeration of the digital service industry leverages digital technologies and algorithms to enable companies to accurately capture customer demands and reduce information asymmetry [36]. This fosters a new green market environment, strengthening the demand for ecologically and economically coordinated land use through the production of green, low-carbon products. Building on this analysis, we suggest research hypothesis 3:

H3. 

ICPP indirectly fosters improvements in GLUE by stimulating the concentration and growth of the digital service industry.

3. Materials and Methods

3.1. Models

Baseline Model

Commenced in 2008, the ICPP serves as an external policy stimulus to GLUE and is examined as a quasi-natural experiment. As pilot cities under the ICPP continue to expand over time, a multi-period DID model is employed to comprehensively evaluate its impact on urban GLUE [37]:

$${GLUE}_{i,t}={\alpha }_0+{\alpha }_1{ICPP}_{i,t}+{\alpha }_2{X}_{i,t}+{\theta }_i+{\gamma }_t+{ϵ}_{i,t}$$

(1)

where ${GLUE}_{i,t}$ ignifies the GLUE of a city, and ${ICPP}_{i,t}$ represents the innovation-driven policy. $X_{i,t}$ encompasses a set of control variables. ${\theta }_i$ and ${\gamma }_t$ represent the city and year fixed effects. ε is the random error term. The estimated coefficient ${\alpha }_1$ captures the average variation in urban GLUE before and after the ICPP’s implementation, offering an evaluation of its effect on urban GLUE.

3.2. Variables

3.2.1. Explained Variable

GLUE serves as a crucial metric for evaluating land resource utilization, indicating pressures on the ecological environment, and gauging the synchronized development of the economy and society. With the existing GLUE evaluation framework and the incorporation of carbon emissions as non-desirable outputs, an enhanced GLUE evaluation index system is proposed (

Table 1. Input–output indicators for measuring urban GLUE.

Layer of Criteria	Layer of Factors	Layer of Indicators	References
Inputs	Land	Urban built-up area	[38]
	Capital	Urban capital stock	[39]
	Labor	Employees in secondary and tertiary industries	[40]
	Resource	Total energy consumption	[41]
Desirable outputs	Economic benefits	Added value of secondary and tertiary industries	[38]
	Social benefits (Government)	Local government revenue	[6]
	Social benefits (Resident)	Residents’ disposable income	[40]
	Environmental benefits	Green coverage rate of built-up areas	[42]
Undesirable outputs	Negative impact on the environment	Composite environmental pollution index	[39]
		CO2 emission	[43]

).

To assess GLUE, a super-efficiency SBM model incorporating undesirable outputs is utilized. This approach is consistent with recognized methodologies within the discipline. The super-efficiency SBM model addresses a notable constraint of the conventional SBM model, which allocates an efficiency score of 1 to multiple decision-making units, thus lacking the ability to differentiate between them. By addressing this flaw, the super-efficiency SBM model allows for the identification of efficiency differences among decision-making units, leading to more precise and detailed assessments [14]. The model is constructed as follows:

$$\min \rho ={\displaystyle \frac{\frac{1}{m}{\displaystyle \sum _{i=1}^m(\overline{x}/x_{ik})}}{\frac{1}{r_1+r_2}\left({\displaystyle \sum _{s=1}^{r_1}\overline{y^d}}/y_{sk}^d+{\displaystyle \sum _{q=1}^{r_2}\overline{y^u}}/y_{qk}^u\right)}}$$

(2)

$$\left\{\begin{array}{c}\overline{x}\ge {\displaystyle \sum _{j=1,\ne k}^n{x}_{ij}{\lambda }_j};\overline{y^d}\le {\displaystyle \sum _{j=1,\ne k}^n{y}_{sj}^d{\lambda }_j};\overline{y^d}\ge {\displaystyle \sum _{j=1,\ne k}^n{y}_{qj}^d{\lambda }_j};\\ \overline{x}\ge x_k;\overline{y^d}\le y_k^d;\overline{y^u}\ge y_k^u;{\lambda }_j\ge 0;\\ i=1,2,\cdots ,m;j=1,2,\cdots ,n;\\ s=1,2,\cdots ,r_1;q=1,2,\cdots ,r_2\end{array}\right.$$

(3)

where $\rho$ represents the GLUE value; $n$ denotes the number of decision-making units; $m,r_1\mathrm{a}\mathrm{n}\mathrm{d}{r}_2$ are the numbers of input indicators, desired output indicators, and undesirable output indicator categories, $x,y^d\mathrm{a}\mathrm{n}\mathrm{d}{y}^u$ represent the elements in the corresponding input matrix, desired output matrix, and undesirable output matrix. $\overline{x},\stackrel{\_\_}{y^{\mathrm{d}}}\mathrm{a}\mathrm{n}\mathrm{d}\stackrel{\_\_}{y^{\mathrm{u}}}$ indicate the slack amounts for inputs, desired outputs, and undesirable outputs, respectively.

3.2.2. Core Explanatory Variable

This study considers the National ICPP as a quasi-natural experiment. The analysis quantifies the policy’s effect by employing an interaction term that joins the city type dummy variable with the dummy variable representing the time of policy implementation (treat × post). In this context, the designation of “treat” is quantified as 1 for cities recognized as national innovative pilot cities, while it is assigned a value of 0 for those cities that do not participate. The “post” variable is valued at 1 from the year the policy is enacted onwards, and 0 for years before the policy’s implementation.

3.2.3. Control Variables

To address potential biases arising from omitted variables, the analysis includes various control variables that have been previously recognized as significant to GLUE [40,44]. The variables under consideration include: (1) the economic development level, shown as the log of per capita GDP; (2) the urbanization rate, defined as the percentage of people living in cities compared to the total population; (3) the financial development level, measured by the ratio of financial loans to deposits in a city; (4) the level of openness, calculated as the ratio of foreign direct investment to GDP, adjusted to the current year’s exchange rate and converted to constant 2006 prices; (5) human capital, indicated by the logarithm of students enrolled in higher education within the city; and (6) the informatization level, represented by the logarithm of mobile phone users per 100 people.

3.2.4. Mechanism Variables

The measurement of urban sprawl is a key area of investigation in urban studies, typically using either single-indicator or multi-indicator approaches. Since the single-indicator method is more suitable for econometric modeling, this study calculates the urban sprawl index based on the methodology proposed by Wang and Wang (2023) [45], as illustrated in the following formula:

$$Sprawl={\displaystyle \frac{\frac{A_{it}}{A_{il}}}{\frac{P_{it}}{P_{il}}}}$$

(4)

where $Sprawl$ represents the urban sprawl index for a city. The variables $A_{it}$ and $A_{il}$ correspond to the built-up areas of city i in year t and the base year, respectively. $P_{it}$ and $P_{il}$ denote the population of city $i$ in year $t$ and the base year, respectively. This study designates 2006 as the base year. A sprawl index of 1 implies no urban sprawl, while values greater than 1 indicate increasing levels of sprawl.

This study utilizes the ‘Statistical Classification of Digital Economy and Its Core Industries (2021)’ published by the National Bureau of Statistics of China, alongside the ‘White Paper on China’s Digital Economy Development and Employment (2020)’ from the China Academy of Information and Communications Technology. The analysis utilizes the location quotient index to assess employment in information transmission, software, and information technology services, serving as an indicator of the agglomeration level within the digital service industry. This methodology is based on the approach outlined by Yu and Hu (2024) [36]. The specific calculation method is as follows:

$${Dsia}_{\mathrm{i}\mathrm{t}}={\displaystyle \frac{\frac{p_{it}}{q_{it}}}{\frac{P_t}{Q_t}}}$$

(5)

where $p_{it}$ and $q_{it}$ denote the number of individuals employed in digital services and the total employment in region $i$ during year $t$, respectively. Similarly, $P_t$ and $Q_t$ represent the number of individuals employed in digital services and the total employment at the national level in year $t$, respectively.

3.3. Data Resource

The data’s completeness and accessibility allow for an observation period extending from 2006 to 2021. This study includes 282 cities in China that are at the prefecture level or higher, which includes 75 national innovative pilot cities. Cities like Bijie and Tongren in Guizhou Province, which were formed following administrative changes after 2011, as well as Lhasa are omitted from this analysis because of significant data deficiencies. The essential sources for the required data comprise the China City Statistical Yearbook, the China Urban Construction Statistical Yearbook and the Economy Prediction System (EPS) data platform. In cases of missing specific indicators, gaps are addressed through city statistical bulletins or linear interpolation. A comprehensive description of the data is provided in

Table 2. Descriptive statistics of key variables.

Variables	Obs	Mean	Std.Dev	Min	Max
GLUE	4512	0.481	0.189	0.0579	1.201
ICPP	4512	0.163	0.370	0	1
lnpgdp	4512	10.49	0.725	4.595	13.06
urban	4512	0.527	0.165	0.115	1
fin	4512	0.675	0.238	0.0598	7.076
openfdi	4512	0.017	0.018	0.001	0.198
lnhum	4512	10.440	1.423	4.234	13.96
lninfo	4512	3.287	0.440	1.665	4.324
Sprawl	4512	0.896	0.210	0.503	2.152
Agg	4512	1.411	0.469	0.032	3.486

.

4. Results

4.1. The Spatial–Temporal Evolution Description of GLUE

Figure 1 illustrates a 3D kernel density plot of GLUE for 282 Chinese cities from 2006 to 2021. It reveals three main characteristics. First, based on the trend observed in the kernel density curves, the center of the distribution curve gradually shifts from left to right, indicating that the overall GLUE in China has been steadily increasing. Second, regarding the distribution pattern, the height of the curve’s peak exhibits a fluctuating pattern characterized by a ‘decrease-rise-decrease’ sequence, while the width of the main peak narrows year by year. This suggests that the disparity in GLUE among different cities is gradually decreasing, indicating a more balanced level of urban development. Finally, in terms of polarization, from 2006 to 2021, the distribution takes the form of either a unimodal or bimodal structure, which implies no significant trend of multilevel divergence and reflects the simultaneous improvement in GLUE across various regions.

Overall, China’s regional GLUE has exhibited a consistent year-on-year increase from 2006 to 2021, although the rate of growth remains modest. This upward trend has been accompanied by a gradual reduction in inter-regional disparities. In the long term, although measurable progress has been made, how to further enhance GLUE in less-developed regions and mitigate regional disparities remains a central policy challenge requiring targeted interventions.

To investigate the spatial evolution characteristics of GLUE, data from four time points—2006, 2011, 2016, and 2021—were selected. Using ArcGIS 10.8 software, a visual analysis of GLUE was conducted for these years. Based on the specific measurement results, GLUE values were classified into five levels, as illustrated in Figure 2.

As illustrated in Figure 3, the overall level of urban GLUE in China remained relatively low in 2006. Cities exhibiting higher levels of GLUE were primarily concentrated in the Southwest, South China, and Northwest regions. Over time, the number of cities with elevated GLUE levels increased by 2011, with notable growth in the Northwest, South China, and North China. However, certain areas in the Southwest experienced a downward trend. By 2016, the upward trajectory persisted, albeit at a decelerated pace, with higher-level cities increasingly concentrated in East and Central China, while some cities in the North and Northwest exhibited signs of regression. By 2021, the number of cities maintaining high GLUE levels showed an overall decline, though a modest improvement was observed in parts of East China.

Overall, China’s national GLUE level showed a clear upward trend during the study period. Cities in East China, the Beijing–Tianjin–Hebei metropolitan region in North China, and the Pearl River Delta region in South China experienced more rapid growth. The growth rate remained relatively stable in Northeast, Southwest, Central, and Northwest China, along with the majority of cities in South China. This suggests that significant regional heterogeneity exists in the pace of GLUE development across cities.

This phenomenon may be attributed to the following factors: ① Cities in East China, the Pearl River Delta in South China, and the Beijing–Tianjin–Hebei metropolitan cluster in North China possess significant locational advantages. The core cities within these regions are progressively positioned at the production frontier of GLUE development, enabling them to exert strong demonstration and agglomeration effects, with spillover benefits becoming increasingly evident. ② These regions also benefit from pronounced political and economic advantages. Through policy guidance—such as preferential resource allocation and spatial functional zoning—the central government has fostered socio-economic development that outpaces that of other regions. As a result, cities in these areas are able to generate higher levels of green land economic output with relatively lower land resource inputs and reduced ecological environmental costs.

From the above analysis of spatiotemporal evolution, it is evident that the GLUE levels of cities nationwide exhibit a fluctuating upward trend. The ICPP, which prioritizes green and low-carbon innovation as a core principle and goal for building innovative cities, likely contributes to this improvement in GLUE. However, the extent of the ICPP’s influence and the mechanisms through which it impacts GLUE require further empirical investigation to be thoroughly understood.

4.2. Results of the Baseline Model

4.2.1. Parallel Trend Test

The justification for utilizing a multi-period DID model stems from the observation that both the experimental and control groups display comparable trends before the implementation of the policy, thereby satisfying the parallel trends assumption. Due to the varied timing of policy adoption across pilot cities, assigning a single year as the cutoff for the time dummy variable proves unfeasible. Consequently, a relative time dummy variable specific to each pilot city is introduced, indicating the year when the ICPP was implemented. A methodology akin to event studies is employed to validate this assumption, with the year just before the policy’s implementation identified as the reference year. Data collected over a span of four years prior to the policy implementation are categorized as period −4, while data extending beyond six years post-implementation are classified as period 6.

Figure 3 displays the dynamic impacts of the ICPP on GLUE. The analysis indicates that during the two to four years preceding the policy’s implementation, the confidence interval encompasses zero, which suggests the absence of significant trend differences in GLUE between pilot and non-pilot cities. This finding supports the parallel trend assumption. Regarding the dynamic effects of the policy, there is no notable enhancement in GLUE observed during the year of implementation. Nonetheless, two years following the policy’s enactment, the impact coefficients become significantly positive and progressively increase, illustrating that the ICPP positively influences urban GLUE, though with a delayed effect.

In order to better understand the dynamic trends of the experimental and control groups during the study period, this paper presents a graph illustrating the evolution of GLUE in pilot and non-pilot cities (Figure 4). As shown in the figure, both groups of cities exhibited similar GLUE trends before policy implementation, indicating comparable development patterns prior to the intervention. However, following policy implementation, the GLUE growth rate in pilot cities exceeded that of non-pilot cities, suggesting that the policy contributed positively to improving GLUE in the pilot regions. Nevertheless, due to a higher initial GLUE level in non-pilot cities, the effect of ICPP on pilot cities’ GLUE was initially less pronounced. This finding is consistent with the results depicted in Figure 3. The overall GLUE level in pilot cities only surpassed that of non-pilot cities after 2015. This suggests that although the initial impact was limited, pilot cities eventually achieved a notable increase in GLUE, driven by innovation-oriented policies. This further supports the effectiveness and long-term sustainability of innovation-driven policies.

4.2.2. Benchmark Regression

Table 3. Baseline regression results.

	(1)	(2)
	GLUE	GLUE
did	0.046 ***	0.040 ***
	(0.015)	(0.015)
lnpgdp		−0.005
		(0.018)
urban		−0.128 *
		(0.066)
fin		0.006
		(0.015)
openfdi		−0.296
		(0.215)
lnhum		−0.021 **
		(0.009)
lninfo		−0.009
		(0.019)
Constant	0.473 ***	0.840 ***
	(0.002)	(0.227)
City effects	Yes	Yes
Year effects	Yes	Yes
N	4512	4512
R2	0.688	0.690

Notes: *, p < 0.1; **, p < 0.05; ***, p < 0.001.

illustrates the influence of the ICPP on GLUE. Column (1) presents results devoid of control variables, whereas Column (2) incorporates them. The table reveals that the ICPP consistently exhibits a positive and statistically significant impact at the 1% level across all specifications, affirming that the ICPP substantially boosts urban GLUE. In Column (2), the ICPP coefficient is recorded at 0.040, suggesting that the ICPP contributes to a 4% enhancement in GLUE for pilot cities relative to non-pilot cities. In comparison to the sample mean, the GLUE metric in pilot cities exhibits an increase of 8.32% (4/0.481). Since the pilot policy began in 2008, the baseline regression model captures the average treatment effect over 14 years, suggesting an annual contribution of approximately 0.29 (4/14) units to GLUE. This underscores the substantial impact of the ICPP on GLUE, providing preliminary support for H1.

4.3. Robustness Test

4.3.1. Placebo Test

To counter potential bias from unobservable omitted variables in the baseline regression, a placebo test method, as described by Wang (2023) [46], was utilized. This involved conducting a randomized simulation that generated 75 cities as placebo treatment groups. Subsequent regressions produced estimated coefficients for the pseudo-policy dummy variable DID_r, a process repeated 1000 times. As shown in Figure 5, the estimated coefficients for the placebo DID variable are primarily centered around zero, showing a normal distribution markedly different from the true estimated value of 0.040. Moreover, the majority of these results are statistically insignificant. This distribution implies that the observed policy effect cannot be ascribed to unobserved omitted variables or non-random city characteristics, thereby strengthening the robustness of the primary findings.

4.3.2. Heterogeneous Treatment Effects

TWFE is the most commonly used method for estimating treatment effects and is the estimation method used in Equation (1). However, if the treatment effects exhibit heterogeneity across time and groups, this may lead to negative weights in the TWFE estimator, thereby resulting in biased estimates [47]. To address this, we employed the decomposition method proposed by Goodman-Bacon (2021) [47] to diagnose the bias in the TWFE estimator. The results are shown in

Table 4. Results of Bacon decomposition.

Decomposition	Coefficients	Weights
Treated groups vs. Never-treated groups	0.056	0.904
Earlier-treated groups vs. Later-treated groups	−0.010	0.044
Later-treated groups vs. Earlier-treated groups	−0.078	0.052

. It can be observed that the estimate for inappropriate treatment effects between later-treated groups and earlier-treated groups is −0.078, with a weight of only 5.2%. This indicates that the estimation bias of using the TWFE method in this study is small, and the research conclusions are reliable.

To further examine the robustness of the benchmark regression estimation results, we utilize heterogeneity-robust estimation models based on the latest developments in staggered DID methods to re-estimate the ICPP coefficients. These models, formulated by Sun and Abraham (2021) [48], Borusyak et al. (2024) [49], de Chaisemartin and D’Haultfœuille (2020) [50], Callaway and Sant’Anna (2021) [51], and Cengiz et al. (2019) [52], facilitate more precise evaluations. As indicated in

Table 5. The results of heterogeneous treatment effects.

Method	(1)	Method	(2)
TWFE	0.040 *** (0.015)	De Chaisemartin and d’Haultfoeuille [50]	0.035 *** (0.013)
Sun and Abraham [48]	0.041 *** (0.013)	Callaway and Sant’Anna [51]	0.072 *** (0.020)
Borusyak et al. [49]	0.061 *** (0.017)	Cengiz et al. [52]	0.045 *** (0.012)

Notes: ***, p < 0.001.

, all six heterogeneity-robust estimators produce positive values that are consistent with the TWFE estimates, reinforcing the conclusion that the ICPP enhances urban GLUE. All estimated coefficients demonstrate statistical significance at the 1% level, suggesting that the influence of treatment heterogeneity on staggered DID estimations is negligible, thereby reinforcing the reliability of the benchmark regression.

4.3.3. Accounting for Omitted Variable Bias

In addition to the variables observed and collected in this research, unobservable omitted variables may affect the estimation results. To address endogeneity concerns stemming from omitted variables, we adopt the method proposed by Oster (2019) [53] to perform a robustness test, using the estimator β* = β*(R_max, δ) to achieve a consistent estimate of the true coefficient. Here, R_max is the maximum goodness of fit in the regression model assuming all unobservable omitted variables were included. The parameter δ quantifies the selection ratio, which reflects the strength of the correlation between observable and unobservable variables with the endogenous variable. We employ two specific approaches for this test: (1) When δ = 1 and R_max = 1.3R, we derive the range for β* as the estimated coefficient of the endogenous variable. If 0 is not within this range, the robustness test is considered passed. (2) When β = 0 and R_max = 1.3R, if the absolute value of δ exceeds 1, the robustness test is also considered passed. The findings in

Table 6. Robustness to omitted variable bias.

Method	Standard of Judgement	Results
(1)	Treatment effect excludes 0	(0.040, 0.089)
(2)	|δ(β = 0, Rmax = 1.3*R)| > 1	β = 2.952

validate that both methods passed the test, substantiating the absence of regression bias from omitted variables and supporting the robustness of the main conclusions of this study.

4.3.4. Other Robustness Tests

To enhance the robustness of the estimation results, we employed several strategies. Initially, to mitigate potential systematic differences between innovative pilot cities (the treatment group) and non-pilot cities (the control group), we followed Liu et al. (2023) [54] in constructing a propensity score matching-DID (PSM-DID) model. We utilized control variables as matching criteria and apply the nearest neighbor matching method for PSM-DID estimation, with the results presented in Column 1 of

Table 7. Other robustness tests.

	PSM-DID	Exclude Central Cities	Exclude Potential Policies
	(1)	(2)	(3)	(4)
ICPP	0.028 ***	0.039 **	0.038 **	0.040 ***
	(0.008)	(0.018)	(0.015)	(0.015)
Smart City			0.016
			(0.013)
Low-Carbon City				0.005
				(0.013)
Controls	Y	Y	Y	Y
City effects	Y	Y	Y	Y
Year effects	Y	Y	Y	Y
Constant	0.820 ***	0.740 ***	0.842 ***	0.841 ***
	(0.158)	(0.231)	(0.227)	(0.227)
N	4072	3952	4512	4512
R2	0.712	0.689	0.691	0.690

Notes: **, p < 0.05; ***, p < 0.001.

. Additionally, to prevent potential interference, central cities were excluded. These cities inherently have advantages in policy, economy, resources, talent, and infrastructure. This study, following the approach by Yu and Hu (2024) [36], focuses solely on ordinary prefecture-level cities to test the policy’s generalizability, with the findings shown in Column 2 of Table 7. Lastly, in light of potential interference from other policies, such as low-carbon and smart city pilot initiatives, dummy variables for these policies were introduced to ensure the accuracy of the estimations. The results presented in Columns 3 and 4 of Table 7 provide evidence supporting the consistency of the baseline regression findings following the implementation of these robustness checks.

4.4. Heterogeneity Analysis

4.4.1. Geographical Location Heterogeneity

Cities across varying locations demonstrate unique market environments, economic foundations, and levels of innovation, which can influence the conditions and impacts of policy implementation differently. Following the regional classification standards by the National Bureau of Statistics of China, the 282 prefecture-level cities are divided into three regions: eastern, central, and western. Interaction terms for ICPP are introduced to address these regional disparities. As depicted in A of Figure 6, the ICPP shows a strong positive effect on GLUE exclusively in eastern cities. However, there is no significant effect in the central and western regions. This disparity is likely attributable to the eastern region’s advantageous geographical location, abundant human resources, and well-established market mechanisms, which enables the ICPP to drive technological and managerial innovations more effectively, enhancing production efficiency and reducing pollution emissions. In comparison, the central region lags in overall development, with some industries still transitioning from traditional to emerging sectors, presenting considerable challenges. Under such circumstances, changing the current industrial impact patterns on green land use in the short term remains challenging, even with the support of innovative pilot city policies. The western region is less developed overall, exhibiting a notable gap in development relative to the eastern and central regions. Here, urban land use primarily focuses on ecological conservation and resource development, characterized by sensitivity, fragility, and irreversibility. This may cause adverse interactions and synergy effects with ICPP, limiting its effectiveness.

4.4.2. Environmental Concerns Heterogeneity

Innovation-driven policies are designed to address public environmental demands while promoting a balance between economic growth and ecological improvement. In order to analyze the varying effects of these policy pilots across cities with differing levels of environmental awareness, this study categorizes 282 prefecture-level cities into primary and secondary environmental protection classifications, following the guidelines established in the “11th Five-Year Plan for National Environmental Protection”. Interaction terms with the ICPP are then incorporated into the regression model to account for these group differences. As shown in B of Figure 6, innovation-driven policies significantly enhance GLUE in key environmental protection cities, whereas their effects in non-key environmental protection cities are not statistically significant. Key environmental protection cities, driven by pressures such as significant regional environmental governance responsibilities, the urgent need to improve environmental quality, and sustained public environmental concern actively seize the policy opportunities presented by the ICPP. These cities actively develop green, low-carbon industries, optimize energy consumption patterns, and refine land use structures, effectively enhancing GLUE. In contrast, non-key environmental protection cities face weaker environmental regulatory constraints, limited motivation to reduce pollution in green land use, and insufficient investments in manpower and funds. Consequently, the ICPP lacks the necessary support to promote GLUE in non-key environmental protection cities, thereby limiting the policy’s effectiveness.

4.4.3. Market Characteristics Heterogeneity

The 19th National Congress of the Communist Party of China highlighted the necessity of cultivating an ecosystem for green technology innovation that is driven by market dynamics, wherein market mechanisms are crucial for the allocation of innovative resources. Technological advancements have the potential to significantly decrease dependence on energy and labor in land utilization processes, improve the efficiency of production factors, and reduce pollution and resource wastage. Thus, we aim to examine the differential impacts of ICPP on urban GLUE from the vantage point of the domestic large market. The domestic large market variable is represented by the Harris (1954) [55] market potential index, defined as $M{P}_i={\displaystyle \sum _{j\ne i}\left(GD{P}_j/d_{ij}\right)}+GD{P}_i/d_{ii}$, where GDP_i represents the GDP of city i, d_ij is the distance between cities i and j, and d_ii is the internal distance of city i. In this analysis, a dummy variable was established utilizing the average market potential of each city before the policy implementation as the benchmark (MP). This allowed for the classification of cities above the average as possessing strong market advantages, while those below the average were categorized as having weak market advantages. Subsequently, interaction terms with the ICPP were integrated into the group regression model. As illustrated in C of Figure 6, the ICPP significantly enhances GLUE in urban areas that exhibit strong domestic large market advantages. Conversely, the policy fails to demonstrate a statistically significant effect on urban areas characterized by weaker market dynamics. The scale advantage and resilience of the domestic large market are crucial in spurring technological innovation, reducing innovation costs, and attracting innovative talent and resources, thus forming a synergy with the positive impacts of the ICPP. This dynamic is more conducive to improving land use efficiency and minimizing pollution emissions. Conversely, regions without market-scale advantages often face resource shortages and insufficient market demand, hindering their ability to capitalize on the “policy dividends” despite policy support.

5. Mechanism Analysis

Beyond the direct effects of the ICPP on GLUE discussed above, the ICPP may also influence GLUE indirectly through alternative pathways. To investigate the mechanisms through which the ICPP impacts GLUE and to address research hypotheses 2 and 3, this study utilizes the framework proposed by Zhang et al. (2022) [38] to develop a mediation effect model:

$$M_{i,t}={\beta }_0+{\beta }_1{ICPP}_{i,t}+{\beta }_2{Control}_{i,t}+CityFE+YearFE+{ϵ}_{i,t}$$

(6)

$$Y_{i,t}={\rho }_0+{\rho }_1{ICPP}_{i,t}+{\rho }_2{M}_{i,t}+{\beta }_2{Control}_{i,t}+CityFE+YearFE+{ϵ}_{i,t}$$

(7)

where M_i,t is the mediating variable, which is sequentially replaced with two variables reflecting the sprawl effect and the agglomeration effect. The variable definitions are maintained as in Equation (1). If both coefficients ${\beta }_1$ and ${\rho }_2$ are significant, this suggests a mediation effect. Furthermore, if ${\rho }_1$ is significant and its sign is consistent with ${\beta }_1\times {\rho }_2$, it indicates that M has a partial mediation effect.

Column (1) of

Table 8. Mediating effect test results.

	(1)	(2)	(3)	(4)	(5)
	GLUE	Sprawl	GLUE	Agg	GLUE
ICPP	0.040 *** (0.015)	−0.031 ** (0.017)	0.031 ** (0.035)	0.083 ** (0.036)	0.035 ** (0.015)
Sprawl			−0.311 *** (0.035)
Agg					0.063 *** (0.012)
Bootstrap		[0.006, 0.016]		[0.003, 0.008]
Controls	Y	Y	Y	Y	Y
City effects	Y	Y	Y	Y	Y
Year effects	Y	Y	Y	Y	Y
Constant	0.840 *** (0.227)	−1.343 *** (0.337)	0.423 *** (0.218)	3.703 *** (0.539)	0.609 *** (0.215)
N	4512	4512	4512	4512	4512
R2	0.690	0.810	0.715	0.724	0.697

Notes: **, p < 0.05; ***, p < 0.001.

displays baseline estimates for reference. The results in columns (2) and (4) demonstrate that the ICPP significantly curtails urban sprawl and promotes the agglomeration of urban service industries, both at the 5% significance level. In columns (3) and (5), the coefficients for urban sprawl are significantly negative, suggesting that urban sprawl obstructs the enhancement in GLUE. Conversely, the coefficients for urban service industry agglomeration are significantly positive, demonstrating that this agglomeration aids in the improvement in GLUE. The coefficients of the ICPP are significantly positive, though they possess smaller absolute values relative to the benchmark regression, implying that both urban sprawl and urban digital service industry agglomeration act as partial mediators in the ICPP’s effect on urban GLUE. To authenticate these mediating variables, the Bootstrap method is applied, following Peng et al. (2021) [56]. Based on 1000 random resampling estimations, the confidence intervals exclude zero. These findings, according to the principles of mediation effect testing, suggest that the ICPP enhances GLUE by reducing urban sprawl and fostering digital service industry agglomeration. Accordingly, the conclusion responds to H2, demonstrating that the ICPP can indeed influence GLUE indirectly through its impact on urban sprawl, while also identifying the specific direction of influence. It supports H3, showing that the ICPP can enhance GLUE by promoting industrial agglomeration within the digital services sector.

6. Discussion and Conclusions

6.1. Conclusions

In the context of China’s “dual carbon” objectives and the initiative aimed at creating a Beautiful China, a comprehensive analysis of innovation-driven policies is essential for improving urban innovation systems and tackling the challenges associated with green and low-carbon development. This examination is essential for enhancing GLUE and fostering robust economic development. Accordingly, we conduct a quasi-natural experiment using the Innovation-Centric Policy Package (ICPP) and analyze panel data from 282 prefecture-level and higher cities from 2006–2021. We develop a multi-period DID model to methodically assess the effects of these policies on GLUE.

Our findings substantiate that the ICPP markedly improved GLUE during the observation period, as corroborated by a series of robustness tests. These results underscore the efficacy of innovation-driven policies in advancing sustainable urban development and environmental sustainability in China. Heterogeneity analysis indicates that the ICPP’s impact on GLUE differs based on a city’s geographical location, environmental concern, and market potential. Specifically, the ICPP’s enhancement of GLUE is stronger in cities with advantageous geographical locations, greater environmental concern, and stronger market potential. Mediation effect analysis shows that the pilot policy improves GLUE primarily by reducing urban sprawl and increasing digital service industry agglomeration.

6.2. Policy Recommendations

The ICPP represents a crucial framework in China aimed at advancing urban innovation systems while addressing the complexities linked to sustainable and low-carbon development. The modifications and recalibrations of economic activities prompted by this policy are intricately linked to land, which serves as an essential physical resource. Improving GLUE is essential for optimal land resource distribution and fostering a holistic green transformation within both the economy and society. In the context of the “dual carbon” objectives and the initiatives aimed at creating a Beautiful China, this study synthesizes these elements within a comprehensive framework. The insights gleaned are invaluable for refining the ICPP and advancing urban GLUE:

(1) The government should intensify efforts to advance innovative city pilot initiatives by promptly summarizing, publicizing, and disseminating the success stories of exemplary pilot cities. Leveraging exemplary cases as a model for gradual advancement can accelerate the expansion of innovative city construction, guiding cities to embrace innovation-driven development and fully utilize innovation’s role in enhancing GLUE. Additionally, it is essential to enhance policy support by prioritizing eligible research projects, innovation initiatives, and reform pilots. Strengthening inter-departmental collaboration mechanisms will further ensure the effective implementation of ICPP.

(2) Innovation-driven policies should be tailored to align with local conditions. For cities with favorable geographical advantages, high environmental concern, and substantial market potential, policies should intensify innovation support, promote the growth of green industries, and leverage green technologies to transform and upgrade traditional industries while optimizing land use structures. In regions with less favorable location advantages and limited market potential, efforts should focus on improving infrastructure and public services, complemented by increased financial and tax subsidies to enhance market potential and fully capitalize on “policy dividends”, thereby guiding the green transformation of land use. For areas with low environmental concern, technologies like the Internet of Things, big data, and cloud computing should be utilized to establish environmental information platforms, increase the residents’ environmental awareness, reduce pollution emissions, and enhance GLUE.

(3) The mechanisms through which the ICPP achieves its policy objectives should be enhanced. First, a comprehensive and scientifically informed urban planning system should be developed and refined, including detailed land use plans that emphasize protecting natural resources and the ecological environment during urban expansion to prevent overdevelopment. Additionally, the application of green building techniques and sustainable development technologies should be encouraged to boost total factor productivity and support sustainable urban growth. Second, the government should actively promote the digital service industry, encouraging the clustering of high-tech and service industries. This can be achieved through preferential policies, enhanced talent training, and targeted financial support to attract high-tech enterprises and professional service industries to cities.

6.3. Shortcomings and Prospects

In addition, there are some shortcomings in this study that can be addressed in future research. Firstly, the concept of GLUE is inherently multidimensional, dynamic, and context-specific. While advancing green, low-carbon development and promoting high-quality economic growth have become shared imperatives across Chinese cities in the current stage of development, substantial heterogeneity persists in their developmental trajectories and resource endowments. From a theoretical standpoint, such heterogeneity suggests that city-level assessments of GLUE should ideally be tailored through differentiated indicator systems. However, due to the limited operational feasibility of such context-sensitive approaches, this study adopts a uniform set of indicators, consistent with prevailing practices in the literature. While this choice facilitates comparability, it may inevitably introduce measurement bias. Developing more precise and context-aware evaluation frameworks for urban GLUE remains a crucial direction for future research; secondly, the strength and efficacy of the ICPP vary across cities, driven by heterogeneity in resource endowments, geographic conditions, and economic capacity. A binary classification based merely on policy adoption risks masking these differences and may yield biased estimates. A key priority for future research lies in developing more granular and systematic measures of policy implementation intensity across diverse urban contexts. Finally, this study primarily relies on city-level panel data. However, in November 2018, the Chinese government announced the first batch of designated innovative counties and county-level cities. This institutional development opens the door for future research to employ more granular, county-level data to explore the relationship between the ICPP and GLUE in greater depth.