The effects and mechanisms of place-based innovation policy: Evidence from National Independent Innovation Demonstration Zones

Abstract

Using panel data from Chinese prefecture-level cities covering 2003–2023, this study employs the entropy method to construct a comprehensive evaluation system for urban innovation capability. The staggered difference-in-differences method is used to systematically identify the impact of the National Independent Innovation Demonstration Zone policy on urban innovation capability. The findings reveal that demonstration zone implementation significantly promotes urban innovation capability, and this effect remains robust after a series of robustness tests. Mechanism analysis indicates that government innovation preferences and intellectual property protection are two significant channels through which the policy enhances urban innovation capability. Heterogeneity analysis demonstrates significant regional and city-type differences in policy effects, with pronounced impacts on eastern, large, and non-resource-based cities, while effects on central and western regions, small and medium-sized cities, and resource-based areas are relatively limited. Accordingly, we propose recommendations to enhance demonstration zones’ effectiveness and promote coordinated regional innovation development. These include systematically expanding demonstration zones’ coverage, improving innovation incentive mechanisms and intellectual property protection systems, and implementing strategically targeted regional policies. This study provides empirical evidence and valuable policy insights for optimizing place-based innovation policies.

Introduction

Since its reform and opening up, China has achieved decades of high-speed economic growth by leveraging abundant labor supply, massive capital investment, and intensive resource consumption. However, with a fading demographic dividend, diminishing marginal returns on investment, and a deteriorating ecological environment, the previous growth model has become unsustainable (Wang & Shao, 2025). China’s development approach urgently needs to shift from factor-driven to innovation-driven. To address this challenge, China launched the Building an Innovative Country Strategy in 2006. In recent years, China has further emphasized the enhancement of self-reliance in science and technology, positioning innovation capacity-building as a key national objective.

Enhanced innovation capacity drives the emergence of new industries, business models, and economic paradigms, serving as a core engine for boosting total factor productivity. In today’s era of accelerating global technological and industrial transformation, innovation capabilities have become a pivotal factor for building regional competitive advantage, reshaping regional competitive landscapes, and advancing the high-quality development and sustainable prosperity of nations or regions (Acemoglu et al., 2016; Cheng et al., 2022). In this context, effectively enhancing innovation capacity is a critical concern that demands in-depth exploration.

As vital economic units, cities aggregate diverse innovation factors and resources, entrusted with the crucial mission of spearheading innovation. Urban innovation capability is a core support for driving regional economic transformation and upgrading, shaping new development momentum, and ensuring China’s long-term stable economic growth (Lan et al., 2022; Tang et al., 2025). However, compared with globally leading innovation cities, Chinese cities exhibit gaps in innovation capability, with a particularly pronounced lack of disruptive innovation. To enhance urban innovation capability, the Chinese central government established National Independent Innovation Demonstration Zones (NIIDZs) in 2009, exploring institutional and mechanism innovations alongside scientific research management reforms. As of November 2023, China has established 23 NIIDZs based on 66 National High-Tech Industrial Development Zones (NHTIDZs), covering 58 prefecture-level cities across the country. These demonstration zones have become vital place-based innovation policy instruments for refining urban innovation systems and elevating innovation capacity.

In driving rapid economic development, the spatial and industrial organization model of high-tech industrial parks has become a crucial platform for developing high-tech industries and fostering new growth drivers, leveraging robust cumulative scientific research capabilities and innovation potential (Huang et al., 2012; Albahari et al., 2018; Zhuang & Ye, 2020). To foster high-tech industries and aggregate innovation resources, China began establishing NHTIDZs in 1988, encouraging pilot cities to gain experience in technological innovation and high-tech industrial development. Existing research indicates that NHTIDZs implementation significantly boosts economic growth and reduces carbon emissions and environmental pollution, while positively enhancing urban innovation output (Wang et al., 2023; Zhong & Yao, 2024; Wang & Feng, 2021). NIIDZ represents a higher-level place-based innovation policy zone built upon the foundations of NHTIDZs, intended to exert greater innovation leadership and demonstration effects.

Existing literature has explored the effects of the NIIDZ policy from multiple dimensions, determining that it significantly enhances urban green total factor productivity (Yu et al., 2023; Zhang et al., 2024c), which is largely driven by positive environmental governance outcomes such as reduced haze pollution (Liu et al., 2022a) and carbon emissions reduction (Fang et al., 2023). Concurrently, research has indicated its catalytic effect on emerging industries such as the e-sports sector (Luo & Zhang, 2025). As core platforms for implementing the National Innovation Driven Development Strategy, the primary mission of NIIDZs is to comprehensively enhance innovation capabilities through pioneering trials and pathfinding. While existing studies have examined the impact of these pilot zones on urban innovation, most have narrowly focused on innovation outputs measured by patent grants (Aisaiti et al., 2022; Lan et al., 2022; Zhu, 2025). Innovation capacity extends beyond mere innovation output levels (Furman et al., 2002). Few studies have systematically evaluated whether this place-based innovation policy can enhance urban innovation capability or examined the channels through which it influences such capacity.

To address this gap, we propose a Policy–Preference–Protection theoretical analytical framework. Using panel data from Chinese prefecture-level cities from 2003 to 2023 and the staggered difference-in-differences approach, this study systematically evaluates the impact of NIIDZ policy implementation on urban innovation capability, with three main objectives. First, we construct an indicator system to measure Chinese urban innovation capability across innovation resource, innovation output, and innovation environment dimensions. Second, the study evaluates the impact of the NIIDZ policy on urban innovation capability. Third, we explore government innovation preferences and intellectual property protection as two significant mechanisms through which the policy affects urban innovation capabilities.

This study intersects with two strands of literature. First, it enriches the research on how place-based policies influence regional innovation. As a core driver of economic growth, innovation depends heavily on place-based government policies (Ahlstrom, 2010; Zhong & Yao, 2024). While existing research has examined the effects of such policies on regional innovation through factors such as resource allocation and talent mobility (Aisaiti et al., 2022; Sun, 2024), the influence of government innovation preferences and intellectual property protection has received relatively limited attention. This study systematically examines the roles of government innovation preferences and intellectual property protection in explaining how place-based policies influence regional innovation capabilities, providing a new theoretical perspective for understanding policy effectiveness.

Second, this study offers a valuable supplement to the methods for measuring regional innovation capabilities. Existing literature has largely relied on single patent indicators (e.g., patent applications or grants) to gauge regional innovation capabilities (Aisaiti et al., 2022; Lan et al., 2022). While accessible, this approach does not fully capture the multidimensional nature of innovation capacity. This study constructs a comprehensive evaluation framework for urban innovation capability. In addition to patent data, we further incorporate diverse indicators related to innovation resources and innovation environment to provide a more complete and accurate depiction of urban innovation capability.

The remainder of this paper is organized as follows: Section 2 presents the institutional background and theoretical analysis; Section 3 introduces the research design and variable selection; Section 4 presents empirical results; and Section 5 concludes and provides policy implications.

Institutional background and theoretical analysis

Institutional background

The establishment of National Independent Innovation Demonstration Zones represents a significant initiative to advance the innovation-driven development strategy. Approved by the Chinese central government, these zones aim to maximize the conversion of scientific and technological achievements into tangible productive capacity through institutional and technological innovation, leveraging optimized soft and hard environments to lead the transformation of economic development models. In 2009, to fully leverage Beijing’s abundant scientific and educational resources, enhance the efficiency of technology transfer and commercialization, and promote closer integration between science and industry, the Chinese central government approved the establishment of the first National Independent Innovation Demonstration Zone—the Zhongguancun Science Park. The government subsequently picked up the pace in deploying National Independent Innovation Demonstration Zones. The Shenzhen National Independent Innovation Demonstration Zone, approved in 2014, became the first such zone based on a single city, while the Southern Jiangsu Independent Innovation Demonstration Zone was the first established around an urban cluster. As of November 2023, China has established 23 National Independent Innovation Demonstration Zones based on 66 National High-Tech Industrial Development Zones, covering 58 prefecture-level cities across the country. The cities covered by these zones are listed in Table 1.

Table 1.

List of cities involved in national independent innovation demonstration zones.

Year	City
2009	Beijing, Wuhan
2011	Shanghai
2014	Nanjing, Suzhou, Wuxi, Changzhou, Zhenjiang
2015	Changsha, Zhuzhou, Xiangtan, Tianjin, Chengdu, Xi’an, Hangzhou, Guangzhou, Shenzhen, Zhuhai, Foshan, Huizhou, Dongguan, Zhongshan, Jiangmen, Zhaoqing
2016	Zhengzhou, Luoyang, Xinxiang, Jinan, Qingdao, Zibo, Weifang, Yantai, Weihai, Shenyang, Dalian, Fuzhou, Xiamen, Quanzhou, Hefei, Wuhu, Bengbu, Chongqing
2018	Ningbo, Wenzhou, Lanzhou, Baiyin, Urumqi, Changji, Shihezi
2019	Nanchang, Xinyu, Jingdezhen, Yingtan, Fuzhou, Ji’an, Ganzhou
2022	Changchun, Harbin, Qiqihar, Daqing

After years of development, the NIIDZs have achieved notable progress in aggregating innovation resources, fostering high-tech industries, and piloting reforms in the science and technology system. Statistics show that in 2022, these pilot zones hosted 114,000 high-tech enterprises, 118,000 science and technology-based small and medium-sized enterprises, and 2,468 listed companies. The 23 National Independent Innovation Demonstration Zones together contributed 40.9 trillion yuan in operating revenues, 21.6 trillion yuan in industrial output, 2.9 trillion yuan in net profits, and 1.6 trillion yuan in tax revenues.

Theoretical analysis

Government innovation preferences

From the perspective of institutional change theory, the establishment of National Independent Innovation Demonstration Zones is essentially a process of stimulating institutional entrepreneurship. Within traditional bureaucratic systems, local governments face significant reform costs and uncertainty risks (North, 1990). As national-level “institutional testing grounds,” demonstration zones effectively reshape local incentive structures by granting governments the mandate for pioneering and piloting, transforming administrative constraints into institutional dividends. This preference shift is not a random policy response but a purposeful institutional evolution driven by systemic incentives—transforming governments from “production-oriented” to “innovation-service-oriented” (Cunningham et al., 2025).

As key participants in regional innovation systems and primary drivers of R&D activities, the innovation preferences of China’s local governments directly determine resource allocation and innovation orientation within these systems (Li et al., 2022). Government innovation preferences reflect the extent to which local governments support the development of regional innovation systems and the conduct of innovation activities. Existing research indicates that enhancing government innovation preferences can effectively leverage governmental functions to boost regional innovation capabilities (Wu et al., 2025a). The impact of government innovation preferences on regional innovation capabilities manifests primarily in two aspects:

First, elevating government innovation preferences helps alleviate financing constraints faced by innovation entities. As the key engine of technological progress, R&D activities are characterized by long cycles, substantial investment, and high risks. In an imperfect capital market, enterprises often face severe financing constraints, with many viewing the lack of external financial support as the primary obstacle to conducting R&D activities (Chen & Yang, 2019). Policy tools such as direct government R&D subsidies and tax incentives can effectively bridge corporate funding gaps, providing critical support for innovation activities (Arrow, 1962; Zhang et al., 2024a; Wang et al., 2022). The effectiveness and operational mechanisms of these policy tools have been substantiated by extensive research. Based on Chinese listed companies’ data, Chen & Yang (2019) found that R&D tax credits significantly enhance firms’ innovation inputs and outputs. Wang et al. (2022) further demonstrated that tax subsidies alleviate firms’ financing constraints, thereby promoting green innovation. Beyond tax incentives, direct government R&D subsidies play a distinct role by offsetting risks and compensating potential losses, thereby providing safeguards for firms investing in high-risk novel ideas (Liu & Xia, 2025). Furthermore, Zhang et al. (2025) emphasize that many government subsidy policies are often tied to firms’ innovation outcomes. This design cleverly leverages incentive mechanisms to fundamentally enhance firms’ intrinsic motivation for innovation.

Second, elevating government innovation preferences fosters collaborative partnerships among diverse innovation actors. As the core entities of innovation activities, enterprises often struggle to tackle complex innovation challenges solely with their own resources. They urgently need to connect with external entities such as universities, research institutes, and other enterprises to access complementary resources like technology, talent, and knowledge (Li et al., 2025a). Governments with an innovation-friendly orientation typically formulate strategies that foster regional innovation development, emphasizing talent cultivation, stakeholder collaboration, and resource aggregation (Ye & Zeng, 2024). By elevating the quality of public services—such as education, housing, healthcare, and social security—governments can attract more highly skilled, innovative talent. Industry-academia-research collaboration serves as the primary channel for innovation in emerging industries (Wang et al., 2024). Governments with a stronger innovation orientation often encourage enterprises to integrate resources and engage in deep collaboration with knowledge-producing institutions and intermediary service providers, driving breakthroughs in key core technologies.

The establishment of NIIDZs has further reinforced local governments’ innovation preferences. As institutional experimentation platforms at the national level, these demonstration zones empower local governments with greater autonomy for reform and space for pioneering policies, incentivizing them to proactively explore institutional mechanisms and policy environments conducive to innovation. The development of demonstration zones not only requires local governments to introduce more groundbreaking measures in areas like fiscal support, talent recruitment, and financial services, but also to strengthen institutional provision and organizational safeguards in cross-departmental coordination and deep industry-academia-research integration. This “point-to-area” policy practice significantly elevates local governments’ prioritization of innovation and resource allocation efficiency, thereby forming sustainable innovation governance models that enhance the overall effectiveness of regional innovation systems.

Intellectual property protection

From the perspective of evolutionary economic geography, enhancing regional innovation capacity is a process of path creation aimed at breaking existing industrial path dependencies. In this process, a high-standard intellectual property protection system acts as the guardian of the “selection environment.” By reducing the risks of technological imitation, it ensures the economic returns of new paths during their evolutionary development. The core of the NIIDZ policy lies in profound institutional supply-side reforms, not merely in providing fiscal and tax incentives. Their paramount focus is on establishing a high-standard intellectual property protection system to effectively strengthen intellectual property safeguards. This is explicitly reflected in the legislative safeguards of the demonstration zones. For instance, Article 35 of the Regulations of the Zhongguancun Science Park stipulates: Administrative departments responsible for patents, trademarks, copyrights, and other intellectual property shall establish and improve platforms for reporting, complaint handling, rights protection, and assistance within the demonstration zone, along with expedited administrative processing channels for related cases. They shall also refine systems for case transfer and information sharing.

As innovation serves as the primary driving force for development, protecting intellectual property is synonymous with safeguarding innovation (Zhu & Xia, 2025). The impact of intellectual property protection on urban innovation capability manifests primarily in three aspects: First, stringent intellectual property protection provides core incentives and fundamental safeguards for innovation activities. Resource-based theory posits that organizations compete based on their unique resources and capabilities. Knowledge resources possess high levels of inimitability and irreplaceability, particularly in technology-intensive industries, where the distinctiveness of knowledge resources determines a firm’s competitive position (Liu et al., 2022b). However, innovation and R&D typically require substantial resource investments. Without effective intellectual property protection, innovative outcomes are highly susceptible to replication by competitors. This not only places firms at a competitive disadvantage but also severely dampens their future R&D investment intentions (Ginarte & Park, 1997). Through its exclusive protective function, intellectual property significantly reduces the risk of innovation being imitated or used without authorization (Wang & Wang, 2025), granting enterprises a monopoly on their innovations for a specific period (Ang, 2010; Wang et al., 2025). This safeguard of economic benefits and market exclusivity constitutes a key driver incentivizing firms to sustain innovation activities (Zheng et al., 2021; Wang et al., 2025).

Second, strengthening intellectual property protection helps foster a fair competitive market environment. Strict enforcement against infringement and counterfeiting effectively upholds equitable market competition (Yang & Cai, 2025). A level playing field ensures all enterprises enjoy equal rights, opportunities, and rules in economic activities, thereby enhancing the vitality and creativity of economic entities (Jiang et al., 2020). A sound market environment not only facilitates resource flows, broadens enterprises’ access to resources, alleviates financing constraints, and promotes innovation investment (Sun et al., 2024), but also accelerates the accumulation of innovation factors, thereby enhancing urban innovation capability (He et al., 2025). Liang et al. (2025) found that optimizing the business environment enhances sustainable innovation by reducing perceived uncertainty, alleviating financial constraints, and attracting innovative talent. Zhang et al. (2024b) further revealed that a favorable market environment attracts greater foreign investment, facilitates technology transfer and knowledge dissemination, and significantly boosts urban innovation capability.

Finally, intellectual property protection drives collaborative innovation among enterprises. On the one hand, within regional industrial clusters highly dependent on knowledge sharing and collaboration, intellectual property protection defines property rights ownership and establishes trust relationships. This incentivizes firms to engage more openly in joint R&D activities, thereby enhancing overall regional innovation efficiency (Zheng et al., 2025). On the other hand, intellectual property protection increases firms’ willingness to disclose information about R&D projects, reducing information asymmetry among collaborators and investors and facilitating technological development (Ang et al., 2014). Using data from Chinese listed companies, Xie & Zhou (2025) found that intellectual property protection significantly promotes collaborative innovation among firms. Zhu & Xia (2025) further discovered that intellectual property protection enhances collaborative innovation by alleviating financing constraints and expanding strategic cooperation networks, with this positive effect being more pronounced in industries with lower environmental uncertainty and in technology-intensive firms.

Based on the above analysis, this paper argues that the NIIDZ policy is likely to enhance urban innovation capability. Further, government innovation preferences and intellectual property protection may serve as two important channels through which the policy exerts its influence. Accordingly, this paper constructs a theoretical framework of “Policy–Preference–Protection.” The NIIDZ policy reconfigures local incentive structures, shifting government “Preference” toward innovation-service to facilitate regional path creation. Simultaneously, it strengthens “Protection” to fortify the “selection environment,” securing economic returns for new technological paths. This synergy reflects the institutional complementarity between proactive governance and market rules in driving regional innovation. Consequently, this study proposes the following hypotheses:

Hypothesis 1

The NIIDZ policy exerts a positive effect on enhancing urban innovation capability.

Hypothesis 2

The NIIDZ policy enhances urban innovation capability through two important channels: strengthening local government innovation preferences and intensifying intellectual property protection.

Research design and variable selection

Variable selection

Dependent variable

Urban innovation capability (UIC) is conceived as the comprehensive capability of a city’s innovation ecosystem—comprising diverse actors—to efficiently allocate resources and transform them into socioeconomic value through knowledge creation, diffusion, and spillover. In the absence of a universally accepted evaluation framework, this paper adopts the approach of Tang et al. (2025) to develop a composite index. This index is predicated on three dimensions that reflect the urban innovation capability’s systemic character: the innovation resources, innovation outputs, and innovation environment.

The innovation resources dimension, rooted in the resource endowment perspective, underscores the foundational role of key resources in driving innovation. It comprises two key aspects: labor resources and enterprise resources. In terms of labor resources, human capital serves as the core dynamic element of innovation; researcher density indicates the level of specialized human input in R&D activities; digital talent density reflects the availability of specialized professionals driving the digital economy and cutting-edge fields such as artificial intelligence; and public education investment intensity represents the government’s long-term strategic commitment to human capital accumulation. As for enterprise resources, enterprise scale reflects the industrial foundation and ecosystem supporting regional innovation, illustrating the basic conditions for knowledge transformation and application, while frontier technology enterprise scale, focusing on strategic emerging industries, measures a city’s capacity to cluster high-quality innovation agents.

The innovation output dimension measures the direct results of innovation activities and their economic transformation benefits from multiple perspectives. Specifically, the scale of frontier patent output and green innovation output assess a city’s performance in high-tech fields and sustainable development, respectively. Furthermore, the scale of collaborative innovation is introduced to capture the city’s ability to integrate into national innovation networks and generate knowledge spillovers. Finally, the scale of scientific and technological achievement transformation reflects the efficiency of market conversion of innovation outcomes and the level of economic value realization.

The innovation environment dimension, grounded in innovation ecosystem theory, highlights the enabling role of external conditions. Knowledge resources reflect a city’s cultural heritage and its capacity for knowledge dissemination; innovation infrastructure measures the city’s digital connectivity and its capacity to facilitate real-time information exchange and rapid knowledge dissemination; financial environment gauges the depth of credit support and the mitigation of financing constraints for innovative firms (Sun et al., 2025); urbanization level generally correlates with denser knowledge spillovers, stronger public services, and more diverse innovation interactions, thereby providing fertile social ground for innovation.

Based on this indicator system, this paper employs the entropy method to evaluate urban innovation capability. This method automatically assigns weights according to the information entropy differences among indicators, which can eliminate the bias in results caused by subjective weighting and improve the evaluation’s objectivity and accuracy (Ma et al., 2025). To mitigate the interference of extreme values on empirical results, the composite index is multiplied by 1,000 and subsequently log-transformed (LnUIC), which serves as the dependent variable for subsequent empirical analysis. The specific definition of indicators is shown in Table 2.

Table 2.

Urban innovation capability indicator system.

Dimension	Secondary indicator	Indicator calculation method
Innovation resources	Human capital	Number of students enrolled in regular higher education institutions / Permanent resident population
	Researcher density	(Number of employees in scientific research, technical services, and geological exploration × 10,000) / Permanent resident population
	Digital talent density	(Number of employees in information transmission, computer services, and software industries × 10,000) / Permanent resident population
	Public education investment intensity	Education expenditure / Local government general budget expenditure
	Enterprise scale	Number of industrial enterprises above designated size
	Frontier technology enterprise scale	Number of strategic emerging industry enterprises
Innovation output	Scale of frontier patent output	Number of strategic emerging industry patents granted
	Scale of collaborative innovation	Number of patents applied for in collaboration with other cities
	Scale of green innovation output	Number of green patents granted
	Scale of scientific and technological achievement transformation	Technology market transaction value at the prefecture level
Innovation environment	Knowledge resources	Number of public library books per 10,000 people
	Innovation infrastructure	Broadband Internet subscribers (in thousands of households) per 10,000 people
	Financial environment	Loans balance of financial institutions at year-end / GDP
	Urbanization level	Urban permanent resident population / Total permanent resident population

Explanatory variable

The independent variable is the interaction term Treat × Post. Treat is a treatment dummy variable that equals 1 if a city is designated as an NIIDZ pilot city, 0 otherwise. Post is a time dummy variable that equals 1 for the year of NIIDZ establishment in that city and all subsequent years, and 0 otherwise.

Control variables

Drawing on Tang et al. (2025) and Aisaiti et al. (2022), this study selects industrial structure, market size, degree of openness, educational attainment, abundance of higher education resources, and level of economic development as control variables. These variables are defined as follows: (1) Industrial structure (IS) is measured by the share of value added by the secondary and tertiary industries in GDP. (2) Market size (LnMarket) is measured by the natural logarithm of total retail sales of consumer goods. (3) Degree of openness (LnOpen) is expressed as the natural logarithm of the number of foreign-invested enterprises. (4) Educational attainment (LnEdu) is represented by the natural logarithm of the city’s average years of schooling. (5) Abundance of higher education resources (LnSchool) is measured by the natural logarithm of the number of higher education institutions. (6) Level of economic development (LnPGDP) is represented by the natural logarithm of per capita real GDP.

Model construction

The NIIDZ policy was implemented in multiple batches across Chinese cities. Its pilot coverage spans both economically and technologically advanced eastern cities and central and western cities with relatively lower levels of development. This staggered rollout across a wide range of cities provides a useful setting for empirical identification. Therefore, this study treats the NIIDZ policy as a quasi-natural experiment. According to Li et al. (2025b) and Wu et al. (2025b), this study adopts the staggered difference-in-differences method with standard errors clustered at the city level to identify the causal effects of the NIIDZ policy on urban innovation capability. Pilot cities from 2003 to 2023 serve as the treatment group, while other cities form the control group. The specific model specification is as follows:

where the subscript t denotes the year and i denotes the city. LnUIC is the dependent variable, representing urban innovation capacity; Treat × Post is the core explanatory variable, representing the policy shock of the NIIDZ. X represents a series of control variables, used to control for other factors that may affect the dependent variable. The model incorporates city fixed effects (i.city) and province-year fixed effects (i.province × i.year) to control for time-invariant city-level traits and unobservable time-varying provincial characteristics, respectively. This high-dimensional fixed effect structure provides a more robust identification of the causal impact of the NIIDZ policy. ξ is the error term, representing random disturbances that the model cannot explain. This model assesses the policy effect of NIIDZ on urban innovation capability by estimating the coefficient of Treat × Post.

Data sources

This study utilizes Chinese city panel data for analysis. Considering data availability, the sample period spans 2003–2023, yielding 5,922 observations across 282 cities over 21 years. Patent data originates from the Chinese Research Data Services Platform. Other data primarily comes from the National Bureau of Statistics official website, the China Statistical Yearbook, the China Urban Statistical Yearbook, and statistical annual reports of selected cities. Cities with substantial missing data (e.g., all cities in Tibet) were excluded from the sample. In addition, missing values in some variables were imputed using the linear interpolation method. Descriptive statistics for the main variables are presented in Table 3.

Table 3.

Descriptive statistics.

Variable	Obs	Mean	Standard deviation	Minimum	Maximum
LnUIC	5,922	3.093	0.767	1.377	6.666
Treat × Post	5,922	0.077	0.266	0.000	1.000
IS	5,922	0.864	0.088	0.501	1.000
LnMarket	5,922	14.921	1.162	11.236	18.588
LnOpen	5,922	3.065	1.583	0.000	8.471
LnEdu	5,922	2.188	0.097	1.785	2.514
LnSchool	5,922	1.672	0.910	0.405	4.691
LnPGDP	5,922	10.104	0.755	7.545	12.172

Empirical analysis

Benchmark regression

Table 4 presents the benchmark regression results. Column (1) only controls for city and year fixed effects; Column (2) further incorporates all control variables; Column (3) controls for province-year fixed effects on top of the previous specifications. All coefficients of Treat × Post are positive and significant at the 1% level, indicating that the NIIDZ policy significantly enhances urban innovation capability. Column (3) shows that, on average, establishing such a zone increases urban innovation capability by 6.7%. The regression results confirm that establishing NIIDZs promotes urban innovation capability enhancement, thereby supporting Hypothesis 1.

Table 4.

Benchmark regression.

	(1)	(2)	(3)
	LnUIC	LnUIC	LnUIC
Treat × Post	0.084***	0.097***	0.067***
	(0.027)	(0.023)	(0.024)
IS		0.995***	0.647***
		(0.157)	(0.185)
LnMarket		0.084***	0.054
		(0.026)	(0.035)
LnOpen		0.057***	0.038***
		(0.012)	(0.011)
LnEdu		1.461***	1.613***
		(0.289)	(0.357)
LnSchool		0.111***	0.098***
		(0.017)	(0.016)
LnPGDP		0.031	0.027
		(0.035)	(0.044)
Constant	3.086***	-2.905***	-2.368***
	(0.002)	(0.636)	(0.809)
City fixed effects	YES	YES	YES
Year fixed effects	YES	YES	YES
Province-year fixed effects	NO	NO	YES
Obs	5922	5922	5922
adj. R^2	0.969	0.975	0.980

Note: *, **, *** denote significance at the 10%, 5%, and 1% levels, respectively. Standard errors in parentheses are clustered at the city level. The variation of year fixed effects is fully absorbed by the province-year fixed effects. This note applies to all subsequent tables in this paper.

Robustness tests

Parallel trends test

The difference-in-differences approach requires the parallel trend assumption. Following Lu et al. (2025), this study employs an event-study specification to test the parallel trends assumption, and the specific model specification is as follows:

Here, Treat denotes the individual dummy variable, and dumyear represents the relative time indicators. To rigorously assess the trends over an extended horizon and according to Liu et al. (2022a), this paper applies front-end merging and back-end merging to the policy. Specifically, dumyear_-6 is defined as the sixth year prior to the policy and all earlier years, and dumyear₁₀ is defined as the tenth year following the policy and all subsequent years. To avoid collinearity issues, this paper respectively selects the period immediately preceding the treatment period (-1 period) and the earliest period within the event window (-6 period) as the baseline period, and the results are shown in Fig. 1. The regression coefficients prior to policy implementation were all insignificant, indicating that the innovation capability of pilot cities and non-pilot cities shared a common trend before the policy, satisfying the parallel trends assumption. After policy implementation, coefficients become significantly positive and gradually increase, indicating that the NIIDZ policy progressively enhances the innovation capability of pilot cities.

Fig. 1.

Parallel trend test.

Placebo test

Although this study controls for numerous city-level characteristics and various fixed effects in the baseline model, unobservable factors may still influence the evaluation results. To further ensure the robustness of the estimation results, we employ a placebo test. Specifically, for each year in which a new batch of NIIDZs was established, this paper randomly selects a corresponding number of cities from the sample and assigns them as pseudo-treated cities, thereby maintaining the original staggered policy rollout structure. To exclude interference from low-probability events, this process is repeated 500 times to obtain the kernel density distribution of the coefficient estimates. The test results are shown in Fig. 2. The dashed line on the right represents the true regression coefficient. It is evident that the distribution of the sampled estimated coefficients clusters around zero and approximates a normal distribution, showing significant divergence from the actual estimated value of 0.067. This indicates that the policy effect observed in the benchmark regression is not driven by chance or unobservable variables, thereby confirming the robustness of the research conclusion regarding the promotional effect of the NIIDZ policy on urban innovation capability.

Fig. 2.

Placebo test.

EBM-DID

The selection of pilot cities for the NIIDZs may introduce selection bias in the benchmark regression. To address this potential bias, this study further employs the entropy balancing method (EBM) proposed by Hainmueller (2012) to construct a more comparable control group. Compared to the propensity score matching method, the EBM achieves precise balancing of all moments of control variables through weight optimization algorithms while preserving the full sample. This effectively overcomes the sample attrition bias and model specification dependency inherent in traditional matching methods. The EBM-DID results are presented in Table 5. Columns (1)-(2) and (3)-(4) present the results under the first-order (mean) and second-order (variance) moment balancing, respectively. All coefficients of Treat × Post remain positive and statistically significant, showing that NIIDZs significantly enhance the innovation capacity of pilot cities.

Table 5.

EBM-DID.

	First-order moment balancing		Second-order moment balancing
	(1)	(2)	(3)	(4)
	LnUIC	LnUIC	LnUIC	LnUIC
Treat × Post	0.111***	0.077**	0.104**	0.113***
	(0.039)	(0.034)	(0.042)	(0.042)
Control variables	YES	YES	YES	YES
City fixed effects	YES	YES	YES	YES
Year fixed effects	YES	YES	YES	YES
Province-year fixed effects	NO	YES	NO	YES
Obs	5922	5922	5922	5922
adj. R^2	0.973	0.986	0.980	0.989

Robust estimators for staggered DID

Standard two-way fixed effects (TWFE) models may yield biased estimates in a staggered DID design if treatment effects are heterogeneous across cohorts or over time, often referred to as the “negative weighting” problem. Goodman-Bacon (2021) notes that using treated units as controls in a staggered DID method leads to potential bias; however, this has a limited effect on the estimates if the weight assigned to these bad comparisons is small. The bacon decomposition result (Fig. 3) indicates that the weight distribution of the TWFE estimator in this paper is highly ideal. The vast majority of weights (94.6%) stem from the comparison between the treatment group and the never-treated group.

Fig. 3.

Bacon decomposition.

To further ensure the robustness of the baseline results against such biases, this paper employs several recently developed robust estimators that are immune to the pitfalls of TWFE models under differential timing. First, this study employs the improved doubly robust DID (DR-DID) estimator proposed by Sant’Anna and Zhao (2020). This method combines outcome regression with inverse probability weighting, which demonstrates outstanding accuracy and consistency under multiple conditions. Second, we employ the two-stage DID (2SDID) robust estimator developed by Gardner (2022): The first stage identifies fixed effects from untreated units; the second stage regresses their residuals on treatment to obtain the treatment effect. Table 6 reports the robustness tests using DR-DID and 2SDID estimators. Across all specifications, the coefficients remain positive and statistically significant, confirming that the baseline results are robust to potential biases arising from heterogeneous treatment effects and staggered timing.

Table 6.

DR-DID and 2SDID.

	DR-DID		2SDID
	(1)	(2)	(3)	(4)	(5)
	Simple aggregation	Dynamic effects	LnUIC	LnUIC	LnUIC
Treat × Post	0.081***		0.100**	0.120***	0.122**
	(0.026)		(0.028)	(0. 027)	(0.032)
Pre_avg		0.001
		(0.005)
Post_avg		0.100***
		(0.027)
Control variables	YES	YES	NO	YES	YES
City fixed effects	YES	YES	YES	YES	YES
Year fixed effects	YES	YES	YES	YES	YES
Province-year fixed effects	NO	NO	NO	NO	YES
Obs	5922	5922	5922	5922	5922

Changing time window

To examine whether the results are influenced by the selection of the sample time period, this study employs rolling truncation for the 2003–2023 sample range: the sample window is alternatively narrowed by moving the starting year forward and the ending year backward by 1 to 4 years in increments, with separate regressions conducted for each scenario. The results are summarized in Table 7. The size, sign, and significance of the estimated coefficients of Treat × Post remain consistent with the benchmark regression results, indicating robust validity of the study’s conclusions. Furthermore, as the time window narrowed, the estimated coefficients gradually decreased from 0.067 to 0.050, suggesting that more extended observation periods capture a more complete release of policy dividends.

Table 7.

Changing time window.

	(1)	(2)	(3)	(4)
	2004–2022	2005–2021	2006–2020	2007–2019
Treat × Post	0.066***	0.063***	0.053***	0.050***
	(0.023)	(0.022)	(0.020)	(0.019)
Control variables	YES	YES	YES	YES
City fixed effects	YES	YES	YES	YES
Province-year fixed effects	YES	YES	YES	YES
Obs	5358	4794	4230	3666
adj. R^2	0.981	0.982	0.983	0.984

Alternative measures of the dependent variable

To ensure that the results are not sensitive to the specific index construction method, we re-measure urban innovation capability using Principal Component Analysis (PCA) and Entropy-Weighted TOPSIS (EW-TOPSIS). PCA effectively reduces dimensionality while preserving the core information of indicators, whereas the EW-TOPSIS evaluates innovation capacity by calculating each city’s relative proximity to the ideal innovation capacity frontier. The re-estimation results, reported in Table 8, show that all coefficients of Treat × Post remain positive and statistically significant, which confirms that baseline results are robust to different indicator synthesis techniques and weighting schemes.

Table 8.

The re-measuring results.

	PCA		EW-TOPSIS
	(1)	(2)	(3)	(4)
	LnUIC	LnUIC	LnUIC	LnUIC
Treat × Post	1.236***	0.939***	0.031***	0.026***
	(0.189)	(0.150)	(0.006)	(0.006)
Control variables	NO	YES	NO	YES
City fixed effects	YES	YES	YES	YES
Year fixed effects	YES	YES	YES	YES
Province-year fixed effects	NO	YES	NO	YES
Obs	5922	5922	5922	5922
adj. R^2	0.906	0.925	0.831	0.837

Other robustness tests

This study further conducts three robustness checks. First, considering potential lag effects in policy implementation, this paper re-estimated the model using one-year and two-year lags of the core explanatory variable, as shown in Columns (1) and (2) of Table 9. Second, to mitigate the interference of extreme values on estimation results, all continuous variables were winsorized at the 1% and 99% levels, as shown in Column (3) of Table 9. Third, to exclude the potential impact of municipalities directly under the central government on estimation results, this study conducted regressions after excluding their samples, as shown in Column (4) of Table 9. The estimated coefficients of Treat × Post are consistent with the benchmark regression results.

Table 9.

Other robustness tests.

	(1)	(2)	(3)	(4)
	LnUIC	LnUIC	LnUIC	LnUIC
Treat × Post	0.079^⁎⁎⁎	0.093^⁎⁎⁎	0.055^⁎⁎⁎	0.061^⁎⁎⁎
	(0.025)	(0.026)	(0.021)	(0.023)
Control variables	YES	YES	YES	YES
City fixed effects	YES	YES	YES	YES
Province-year fixed effects	YES	YES	YES	YES
Obs	5922	5922	5922	5838
adj. R^2	0.980	0.980	0.980	0.979

Mechanism analysis

Government innovation preferences

This paper measures government innovation preferences from two perspectives: resource allocation and strategic concern. First, the proportion of government fiscal expenditure on science and technology relative to total local fiscal expenditure is used as a metric for government innovation preferences (Perf1), representing the government’s “hard investment.” A higher expenditure ratio indicates a stronger tendency for local governments to prioritize limited fiscal resources toward supporting R&D activities and building regional innovation systems. Second, the level of governments’ emphasis on innovation (Perf2) is assessed by scraping municipal government work reports using Python and conducting text analysis.¹ This method calculates the frequency distribution of innovation-related terms to capture the government’s “soft concern.”

Empirical results in Columns (1) and (2) of Table 10 provide supporting evidence that the NIIDZ policy significantly enhances local government innovation preferences, as evidenced by the significant increases in the proportion of fiscal science and technology expenditure (Perf1) and the frequency of innovation-related terms in government reports (Perf2). Existing research suggests that enhanced government innovation preferences promote urban innovation capability by alleviating corporate financing constraints and fostering the agglomeration of innovative talent (Arrow, 1962; Wang et al., 2022; Ye & Zeng, 2024). These findings suggest that government innovation preferences are an important channel through which the NIIDZ policy enhances urban innovation capability.

Table 10.

Mechanism analysis.

	Government innovation preference		Intellectual property protection
	(1)	(2)	(3)	(4)
	Perf1	Perf2	IPP1	IPP2
Treat × Post	0.015***	0.110***	0.082***	0.147***
	(0.002)	(0.040)	(0.020)	(0.043)
Control variables	YES	YES	YES	YES
City fixed effects	YES	YES	YES	YES
Province-year fixed effects	YES	YES	YES	YES
Obs	5922	5922	5922	5922
adj. R^2	0.772	0.454	0.401	0.452

Intellectual property protection

Theoretically, the number of concluded cases is fundamentally constrained by judicial supply factors, such as court staffing and trial efficiency. In institutional environments with weak protection, the volume of concluded cases often remains low due to systemic backlogs; thus, a significant increase in concluded cases reflects a proactive improvement in the judicial environment and enforcement capacity. Following Yu et al. (2023), this study employs the number of intellectual property cases concluded by local courts (IPPCourt_it) to proxy for judicial protection intensity. Specifically, this paper constructs two variables. First, the intellectual property protection intensity (IPP1) is calculated by deflating the total number of concluded cases by the city’s GDP, thereby mitigating the confounding effects of economic scale on case volume. Second, this paper develops a relative protection index (IPP2) by benchmarking the city-level intellectual property intensity against the national baseline, following the logic of the revealed comparative advantage index. This relative measure enhances cross-city comparability and captures the city’s standing in intellectual property protection within a national context. IPP1 and IPP2 are calculated as follows:

Here, IPPCourt_it and GDP_it represent the number of concluded intellectual property cases and regional GDP of city i in year t, respectively; while IPPCourt_ct and GDP_ct denote the national total of concluded intellectual property trials and GDP in year t.

The results in Columns (3) and (4) of Table 10 provide supporting evidence that the NIIDZ policy effectively improves the level of intellectual property protection, showing significant positive impacts on both judicial protection intensity (IPP1) and the relative protection index (IPP2). Existing research suggests that stringent intellectual property protection promotes regional innovation capacity by securing innovation returns, fostering collaborative behavior, and enhancing cluster innovation efficiency (Wang et al., 2025; Xie & Zhou, 2025; Zheng et al., 2025; Zhu & Xia, 2025). The results in Table 10, along with the existing literature, suggest that government innovation preferences and intellectual property protection are two important channels through which the NIIDZ policy enhances urban innovation capability. Therefore, Hypothesis 2 is supported.

Heterogeneity analysis

Given significant variations among Chinese cities in geographic location, development stage, and resource endowments, the impact of policy on urban innovation capability may exhibit heterogeneity. This study conducts heterogeneity analysis from three dimensions: regional heterogeneity, city size heterogeneity, and city resource dependency heterogeneity.

Regional heterogeneity

China’s vast territory exhibits significant heterogeneity across regions in economic foundations, social structures, and innovation resources. These differences profoundly influence the formation and enhancement pathways of regional innovation capabilities. To examine the potential differential impacts of the National Independent Innovation Demonstration Zone policy on urban innovation capability, this study divides the 282 cities in the sample into three major regions—eastern, central, and western groups—based on the National Bureau of Statistics’ standard regional classification.² Separate regression analyses are conducted for each region to identify and compare the policy’s implementation effects under different geographical conditions.

The results in Table 11 indicate that the policy effect is significant only in the eastern region. This may suggest a complementary effect between pilot policies and regional institutional environments: the higher levels of marketization and rule of law in eastern regions may have reduced the friction costs associated with policy implementation. Conversely, the lack of significance in central and western regions may reflect a mismatch between factor endowments and institutional environments: in settings with relatively insufficient financial support and talent reserves, policy dividends struggle to overcome initial barriers and translate into innovation outputs.

Table 11.

Regional heterogeneity.

	(1)	(2)	(3)
	Eastern	Central	Western
Treat × Post	0.104***	-0.036	0.075
	(0.026)	(0.033)	(0.088)
Control variables	YES	YES	YES
City fixed effects	YES	YES	YES
Province-year fixed effects	YES	YES	YES
Obs	2100	2100	1722
adj. R^2	0.984	0.976	0.973

Heterogeneity in city size

Significant interaction effects exist between city size and regional innovation behavior (Cai et al., 2021). The impact of establishing NIIDZs on urban innovation capability may exhibit heterogeneous effects across different city sizes. This paper categorizes prefecture-level cities into large, medium, and small cities based on their permanent resident population. Specifically, cities with a permanent resident population of 5 million or more are classified as large cities, those with fewer than 3 million permanent residents are classified as small cities, and the rest are classified as medium-sized cities.

Table 12 indicates that policy effects are only significant in large cities. This suggests that NIIDZ policies may face an agglomeration threshold for effectiveness: large cities, with their high-density factor pooling and knowledge spillovers, may find it easier to overcome the initial cost barriers in the innovation chain. In contrast, small and medium-sized cities, constrained by insufficient factor density and facing the siphoning effect from larger cities, struggle to generate significant policy dividends in their sparse factor environments.

Table 12.

Heterogeneity in city size.

	(1)	(2)	(3)
	Large	Medium	Small
Treat × Post	0.076***	-0.026	0.007
	(0.027)	(0.032)	(0.067)
Control variables	YES	YES	YES
City fixed effects	YES	YES	YES
Province-year fixed effects	YES	YES	YES
Obs	1767	1626	2361
adj. R^2	0.991	0.973	0.959

Heterogeneity in urban resource dependency characteristics

Resource-based cities serve as vital strategic bases for China’s energy and resource security, providing crucial support for the sustained and healthy development of the national economy. However, due to historical legacy issues such as long-term extensive development models, these cities face significant challenges in enhancing their innovation capabilities. To examine the heterogeneous impact of National Independent Innovation Demonstration Zone policy on different city types, this study categorizes the sample into resource-based cities and non-resource-based cities based on the Classification Standards in the National Sustainable Development Plan for Resource-Based Cities (2013-2020), and conducts grouped regression analysis.

Table 13 indicates that the policy has generated significant incentives only for non-resource-based cities. The underlying mechanism may lie in path dependence and the lock-in effects on industrial structure: non-resource-based cities possess more flexible factor allocation mechanisms, facilitating resource flows toward high-tech sectors, whereas resource-based cities may be constrained by the lock-in effects of traditional energy and mining industries. Their rigid industrial structures produce a certain crowding-out effect, thereby weakening the implementation effectiveness of the pilot policy.

Table 13.

Heterogeneity in urban resource dependency characteristics.

	(1)	(2)
	Non-resource-based cities	Resource-based cities
Treat × Post	0.062^⁎⁎	-0.037
	(0.024)	(0.055)
Control variables	YES	YES
City fixed effects	YES	YES
Province-year fixed effects	YES	YES
Obs	3528	2331
adj. R^2	0.983	0.962

Conclusions and policy recommendations

As an important place-based innovation policy within China’s national innovation system, the NIIDZ policy plays a pivotal role in promoting regional innovation-driven development and fostering high-level innovative cities. Scientifically evaluating the actual impact of these demonstration zones on urban innovation capability is crucial not only for optimizing and promoting innovation policies but also for providing vital decision-making references for achieving self-reliance in science and technology and high-quality economic development. Therefore, this paper utilizes panel data from China’s prefecture-level cities and employs the entropy method to comprehensively measure urban innovation capability across three dimensions: innovation resources, innovation outputs, and innovation environment. By applying the staggered DID method, this study identifies the policy effects of NIIDZ on urban innovation capability. Furthermore, this study analyzes the operational channels through which NIIDZ influences urban innovation capability from two perspectives: government innovation preferences and intellectual property protection. The main conclusions of this paper include: (1) The demonstration zone policy significantly enhances overall urban innovation capability, a finding that remains robust across placebo tests, EBM-DID estimation, and a series of additional robustness checks; (2) The mechanism analysis further shows that government innovation preferences and intellectual property protection play important roles in explaining the positive effect of the NIIDZ policy on urban innovation capability. (3) Heterogeneity analysis reveals that policy effects vary by city type: the policy significantly boosts innovation in eastern, large, and non-resource-based cities, but its impact is insignificant for central and western cities, small and medium-sized cities, and resource-based cities. These findings are broadly consistent with previous studies showing that place-based policies can promote regional innovation performance, while further extending the literature by highlighting the importance of government innovation preferences and intellectual property protection as key channels.

Based on these findings, the following policy recommendations are proposed to enhance the effectiveness of NIIDZ and promote coordinated regional innovation development: (1) Adopt a prudent and differentiated expansion strategy for demonstration zones and optimize implementation mechanisms. The coverage of demonstration zones can be gradually expanded to cities with suitable institutional conditions and innovation foundations. Meanwhile, the supporting policy system should be further improved to ensure the full realization of policy benefits. (2) Strengthen government innovation incentives and intellectual property protection. On the one hand, innovation performance should be incorporated into local government evaluation systems to strengthen government innovation preferences. On the other hand, intellectual property system reform should be deepened by improving fast-track examination and rights protection mechanisms, strengthening punitive damages for infringement, and enhancing institutional safeguards for innovation. (3) Implement differentiated regional policies. For central regions, small and medium-sized cities, and resource-based cities where the policy effect is less pronounced, increase central fiscal support and resource allocation to support infrastructure and talent development, facilitate the inflow of innovation resources, and drive traditional industry transformation and upgrading, thereby fostering sustained innovation momentum.

This study retains certain limitations, and future research could deepen in the following areas: First, despite employing the DID method, policy evaluation may still be subject to unobserved factors. Subsequent studies could combine stronger causal identification strategies with more advanced data-driven evaluation tools to further validate the findings (Xiao & Qu, 2025). Next, this study primarily analyzes policy effects and mechanisms at the macro-urban level, failing to reveal the differentiated impacts of policies on micro-level innovation entities. Future research could utilize micro-level data to explore policy transmission pathways from a multi-level perspective, providing more detailed references for targeted policy implementation.

CRediT authorship contribution statement

Bo Zhou: Writing – review & editing, Supervision, Funding acquisition, Formal analysis, Conceptualization. Wenhui Jiang: Writing – original draft, Methodology, Investigation, Data curation. Yunge Han: Software, Project administration, Funding acquisition. Yaode Jian: Writing – original draft, Methodology.

Acknowledgement

This work was supported by the National Natural Science Foundation of China (grant numbers 72303094, 72474095, and 72503096), the Major Project of National Social Science Fund of China (grant number 24ZDA097), the Research Project at Nanjing University of Finance and Economics (grant number XKYC2202403), and the Research Project of the Northern Hubei Regional Development Research Center at Hubei University of Arts and Science (grant number 2026JD12).