The impact of innovation-driven policies on urban resilience

Abstract

How to enhance urban resilience (UR) through innovation is an emerging issue in urban governance. The National Innovative City Pilot Policy (NICPP) has become an important driving force to enhance UR and promote sustainable urban development. Examining the panel data of 280 Chinese prefecture-level cities from 2006 to 2020, this paper uses a time-varying difference-in-differences (DID) model to identify the impact of the NICPP on UR and its impact path and investigates the spatial effect of UR by constructing a spatial Durbin model (SDM). Benchmark regression results show that NICPP can significantly improve UR. The results of the moderating effect show that both industrial diversification and market demand play a positive moderating role in the relationship between NICPP and UR. Heterogeneity analysis reveals that the effect of NICPP on UR is greater in cities east of the Hu Huanyong line, with larger city sizes and higher administrative levels. The SDM test shows that NICPP has a positive spatial spillover effect on UR in the neighboring areas. This study not only enriches the empirical evidence of the microeconomic impacts of the NICPP but also puts forward novel insights and new starting points for the construction of resilient cities.

Introduction

Rapid urban growth is accompanied by increasing uncertainty. The fragmentation of urban patterns can accelerate the agglomeration of rural populations and economic activities into cities, posing significant challenges to urban governance (Harrison and Williams 2016). In addition, factors such as climate change, economic fluctuations, public health events, and traffic congestion can similarly contribute to urban vulnerability (Joerin et al. 2012; Martin et al. 2016; Büyüközkan et al. 2022; Pelling et al. 2022), constraining sustainable urban development. UR is often associated with sustainable development as an interacting dynamic change process (Mendizabal et al. 2018). Therefore, enhancing UR to alleviate vulnerability or sensitivity, respond to disaster risks, and improve urban governance appears to be critical (Jabareen 2013). In this context, the concept of “resilient governance” has been put on the agenda, which advocates the enhancement of UR through resilient cities from the perspective of urban governance to effectively withstand and adapt to uncertain risk shocks promptly (Datola 2023).

In the wake of COVID-19, China has also paid attention to the importance of building resilient cities. The 14th Five-Year Plan historically incorporates “resilient cities” into national planning, and clearly puts forward the goal of building resilient cities. The report of the 20th National Congress of the Communist Party of China emphasized the implementation of urban renewal actions, strengthening urban infrastructure construction, and creating livable, resilient, and smart cities. This also points out the direction for promoting the construction of resilient cities.

Cities are the gathering places and important carriers of innovative elements. Shaped by a new generation of information technologies such as big data, cloud computing, and artificial intelligence, scientific and technological innovation has provided important support and guarantee for urban industrial development, overall coordination, and efficient governance. However, innovation is characterized by externalities and uncertainties, which require government-led innovation support policies to be effective (Yang et al. 2022; He et al. 2023). NICPP aims to achieve efficient resource allocation, optimize regional industrial structure, improve urban eco-efficiency, and reshape the development advantages of modern cities through high-quality scientific and technological supply and innovation resource agglomeration, thus realizing high-quality urban development (Li et al. 2021; Zheng et al. 2021). NICPP is an incremental reform policy piloted in batches, guiding cities to accelerate innovation-driven development through the experience of “demonstration leading and gradual promotion” to form innovation demonstration and leading highlands with regional and even international competitiveness. In 2008, Shenzhen was the first city to be approved as a “National Innovative City Pilot”. In 2010, the National Development and Reform Commission (NDRC) continued to expand the scope of the pilot policy and successively approved 16 cities as pilot innovative cities. Simultaneously, the MOST selected 20 cities (districts), including the Haidian District of Beijing, to become national innovative pilot cities (districts). From 2011 to 2013, Lianyungang and other cities were successively set up to carry out innovative city pilots. In 2017, the Ministry of Science and Technology (MOST) released the “National Innovation Cities Innovation Capability Testing Report” to evaluate the innovation capability of 61 pilot cities. In 2018, the MOST and the NDRC again supported 17 cities, including Shaoxing, to construct innovative cities, boost science and technology support, and lead economic and social transformation and upgrading. By 2022, seven batches of 103 cities have been established as national innovative cities (districts) (see Fig. 1).

Fig. 1

Distribution of the innovative cities.

Similarly, resilient cities are essential to modernizing the urban governance system and governing capability as a new paradigm for urban sustainable development. Resilient cities cannot be separated from the empowerment of science and technology innovation. It can be seen that there is a strong correlation exists between the establishment of innovative cities and UR. So, can the NICPP effectively improve urban resilience? Further, what are the impact mechanisms? What about its heterogeneity? Is there a spatial spillover effect of the NICPP on enhancing UR? These questions deserve further exploration. The scientific answers to the above questions offer valuable insights into resilient cities, the modernization of urban governance systems, and the enhancement of governance capacity. Furthermore, they provide practical guidance for assessing the impact of the NICPP and achieving high-quality urban development.

The following are some of the paper’s potential marginal contributions. First, in terms of research topics, the policy effects of the NICPP on UR and their spatial spillovers are discussed. Seldom has the link between the NICPP and UR been examined in the literature, and the spatial effect of NICPP on UR is often overlooked. This work broadens the research scope of UR and explores the spatial impact of NICPP on UR under the geographic-economic nested weight matrix, which presents a distinctive perspective on constructing resilient cities. Second, from a research standpoint, the paper innovatively introduces the dual perspectives of industrial diversification and market demand effect, which profoundly reveals the influencing variables affecting the resilience-enhancing role of innovative cities. Based on the analysis of industrial diversification, it further distinguishes the impact of related and unrelated diversification on the resilience-enhancing role of innovative cities, and enriches the influencing path of the resilience-enhancing role of innovative cities. Third, concerning the research methodology, we use the heterogeneity treatment effect test to correct possible biases in the two-way fixed model, which improves the accuracy and reliability of the results on innovative cities enhancing UR.

The remaining content is structured in the following manner. The second section provides an overview of the NICPP’s background, while the third section examines existing literature and presents research hypotheses. The fourth section outlines the method design and variable selection process, followed by the fifth section, which analyses the benchmark regression results and conducts robustness tests. The sixth section further discusses the heterogeneity analysis and mechanism test. Finally, the conclusion and policy recommendations are presented in the seventh section.

Literature and research hypothesis

Literature review

In physics, the term “resilience” was initially used. Then Holling (1973) pioneered applying resilience in ecological contexts, resilience is defined as a system’s capacity to absorb shocks and then return to its pre-shock condition following shock alterations. As research has progressed, the concepts of engineering, evolutionary resilience (adaptive resilience), urban resilience, and economic resilience have gradually derived and present a multidisciplinary cross-development trend. UR is characterized by dynamic changes and interactions, frequently linked to sustainable development (Mendizabal et al. 2018). The definition of UR also differs based on the perspective and disciplinary area of research inquiry (Datola 2023). Cutter et al. (2014) put forth the perspective that UR pertains to a social system, encompassing individuals, households, infrastructure, or broader systems, which can prevent, absorb, recover, and adapt to actual or potential disasters from a geoscience standpoint. Martin et al. (2015) propose from the field of economics that UR involves the vulnerability or sensitivity exhibited in response to external disturbances, achieved through adjustment, adaptation, and recovery to return to the original development trajectory or shift to a new one. UR is introduced in urban planning to design, plan, and manage cities in dynamic socio-ecological systems (Masnavi et al. 2019; Feng et al. 2020; Datola 2023). UR is defined as the ability of an urban system and all its constituent socio-ecological and socio-technical networks to maintain or rapidly restore desired functioning, adapting, changing, and rapidly changing current or future changes on spatial and temporal scales (Meerow et al. 2016; Datola 2023).

Given the significance of UR, scholars have concentrated on its measurement, albeit with varying emphases. For example, several methodologies such as GIS technology and remote sensing model (Stotten et al. 2021), Resilience Maturity Model (RMM) (Hernantes et al. 2019), resilience knowledge network (Elkhidir et al., 2023), and evaluation indicator system have been employed (Zhou et al. 2021). Currently, most studies use the evaluation index system of UR, and scholars believe that cities are a complex system with multiple dimensions such as economy, society, and ecology (Xia and Zhai 2022), while some scholars have added the infrastructure dimension based on the system again (Xun and Yuan 2020; Zhou et al. 2021), and Zhou et al. (2022) measure UR from five dimensions: economic, social, ecological, infrastructure and institutional. In short, there is no consensus on the measurement of UR, and there is a lack of a systematic framework for measuring indicators. From an urban resilience enhancement perspective, MCR2030 emphasizes the exchange of experiences among stakeholders through knowledge networks to improve UR (Campbell 2009; United Nations Office for Disaster Risk Reduction (2020)). Additionally, some academics recommend enhancing urban intelligence and industrial structure optimization to improve UR (Zhou et al. 2021; Tang et al. 2023).

Most of the research on the NICPP has focused on the evaluation of the effectiveness of the policy. Regarding economic effects, such studies have found that innovative cities have a positive impact on the level of financial development (Zhao et al. 2021), the export performance and product quality of enterprises (Yang et al. 2023), and the digital transformation of enterprises (Liu et al. 2024). Regarding ecological effects, scholars have studied the impacts of innovative cities on land green use efficiency (Zhang et al. 2025), carbon intensity (Fan et al. 2024), and environmental pollution (Gao and Yuan 2021). Moreover, other relevant literature focuses on the impact of technological innovation and infrastructure development (Schintler and McNeely 2022; Wang et al. 2023; Zhang et al. 2023, 2024; Zhang et al. 2024), as well as the impact of artificial intelligence, regional integration, and smart cities on urban resilience (Zhou et al. 2021; Jiang and Jiang 2024).

Although the existing literature has analyzed NICPP and UR from various perspectives, few studies have explored the causal relationship and its spatial effects between NICPP and UR from the micro standpoint of prefecture-level cities, and the mechanism of UR also needs to be deepened and expanded. In summary, this paper enriches the effect of NICPP and broadens the research boundary of UR.

Research hypothesis

Innovative cities and urban resilience

Resilient cities represent a new trend in adapting to the modernization of urban development, offering a fresh perspective for urban sustainable development (Ahern 2011). The innovation capacity of cities is an important driving mechanism for improving urban resilience.

The National Innovative City Pilot Policy emphasizes technological innovation as the guide and enhances the city’s ability to withstand risks and adjust to adaptation by aggregating innovative resources, cultivating innovative talents, and improving the innovation environment, to continuously improve the level of urban innovation and thus enhance the overall resilience of cities. Cities are multi-dimensional integrated systems that include economy, society, ecology, and infrastructure (Xia and Zhai 2022). Therefore, this paper will explore the impact of innovative cities on urban resilience from four dimensions: economy, society, ecology, and infrastructure.

First, the National Innovative City Pilot Policy can enhance urban economic resilience. Relying on a favorable innovation ecosystem, innovative cities can overcome the constraints of technological dependence and low-end locking of the industrial chain. Technological innovation can enhance the ability of cities to navigate uncertainty, improve their forecasting capabilities, seize economic development opportunities, and provide technical support for strengthening urban economic resilience (Martin et al. 2015; Chirumalla 2021; Zhang et al. 2024). Second, the National Innovative City Pilot Policy can enhance urban social resilience. Factors such as urban employment issues, living environment, human resources, social security, and stability are all crucial manifestations of social resilience (Cutter and Finch 2008). By leveraging intelligent technologies like AI, blockchain, big data, and cloud computing, innovative cities can improve employment structures, strengthen social governance and security capabilities, foster new strategies and ideas for urban management, and increase their resilience (Pagliacci and Russo 2020; Li et al. 2023). Third, the National Innovative City Pilot Policy can enhance urban ecological resilience. Urban environmental elements are fundamental to achieving urban resilience (Zhao et al. 2020). Ecological risks and ecological vulnerability pose serious challenges to urban ecological carrying capacity (Du et al. 2020; Zhao et al. 2021). Innovative cities are supported by scientific and technological innovation, which makes urban disaster prevention, mitigation, and emergency management more precise and effective, enabling real-time monitoring of urban disasters and improving disaster response capabilities. Additionally, innovative cities, guided by the principles of green development, have a pivotal impact on mitigating carbon emissions, improving resource utilization efficiency, and reducing environmental pollution (Li et al. 2021). Fourth, the National Innovative City Pilot Policy can enhance urban infrastructure resilience. The rapid pace of urbanization can lead to the overloading of urban infrastructure, resulting in significant vulnerabilities (Zhao et al. 2020). The abundance of innovative resources is a key advantage in constructing innovative cities. Innovative cities rely on creative ways to achieve infrastructure intelligence, promote the development of new infrastructure kinds, ensure the cities’ orderly growth, create a modernized infrastructure system, and enhance urban resilience. In summary, the hypothesis is put out in this paper:

Hypothesis 1. The National Innovative City Pilot Policy can enhance urban resilience.

Industrial diversification

Industrial diversification can help innovative cities enhance urban resilience. As mentioned above, piloting innovative cities can strengthen the city’s innovation function, integrate dispersed innovation resources in the region, and realize the transfer of resource elements to high-tech industries. In this context, the development of industrial diversification strengthens the degree of inter-industry linkage, guides the rational allocation of innovation elements among sectors, and vigorously cultivates strategic emerging industries like new energy and new materials by amplifying the knowledge or technology spillover effect of innovative cities, and promotes the cross-integration of emerging industries with traditional industries to improve the level of urban resilience (Zhang et al. 2021). In addition, industry diversification has an automatic stabilizer function. Industrial diversification is one of the crucial factors affecting urban resilience (Evans and Karecha 2014; Brown and Greenbaum 2017). Compared with cities relying on a single specific industry, the formation of industrial diversification can help innovative cities dilute uncertainty risks, optimize the urban industrial structure through industrial innovation clusters and diversified agglomeration (Yang et al. 2022), and exhibit stronger urban resilience. This also implies that industrial diversification enhances the positive impact of innovative cities on urban resilience.

Industrial diversification can be further distinguished into related and unrelated diversification (Frenken et al. 2007). The JACOBS theory of externalities suggests that diversified agglomerations of complementary industrial structures promote collaborative innovation and technology diffusion through the sharing of heterogeneous resources and capabilities (Jacobs 1969). Especially during the shock resistance period, technological spillovers from related diversification enhance the impact of innovative cities on urban resilience. Castaldi et al. (2015) argue that unrelated diversification can lead to technological breakthroughs, stimulate innovative thinking, and generate disruptive innovations. Disruptive innovation also promotes the cross-integration of different industries, giving rise to new models, new industries, and new economies. The combination of unrelated industries offers greater possibilities for innovative cities to promote urban resilience. Rather than spreading the crisis to other industry sectors, unrelated industries can more effectively disperse and reduce the risk of shocks to innovative cities. Therefore, unrelated diversification is more likely to contribute to the positive impact of innovative cities on urban resilience than related diversification. This leads to the following hypothesis:

Hypothesis 2a. Industrial diversification positively moderates the relationship between the National Innovative City Pilot Policy and urban resilience.

Hypothesis 2b. Both related and unrelated diversification can positively moderate the relationship between the National Innovative City Pilot Policy and urban resilience, with unrelated diversification having a stronger positive moderating effect.

Market demand effects

Theoretically, mega-market demand can motivate innovative cities to resist external risk disturbances and improve their competitiveness and resilience through innovation. Demand serves as a crucial tool for governments to encourage and stimulate technological innovation (Aschhoff and Sofka 2009). When the market demand is high, the government adopts subsidies to stimulate innovation in specific industries (Feng et al. 2022; Wang et al. 2022). The larger the market demand, the greater the market potential. Market potential is likely to positively influence the relationship between innovative cities and urban resilience. On the one hand, the market potential can amplify the benefits of technological innovation in innovative cities, and achieve resilience enhancement by leveraging the advantages of new technology applications and iterative upgrading (Han and Yuan 2023). For cities with greater market demand potential, this demand stimulates the generation of new industries, new models, and new kinetic energy through technological innovation, creates new development advantages, and then achieves the goal of reducing risks. On the other hand, the effect of market demand can also help pilot cities continuously improve their urban innovation capabilities, withstand uncertainties, provide endogenous impetus for regional sustainable development (Han and Yuan 2023), and make urban development more resilient (Christopherson et al. 2010). The new economic geography theory (Krugman 1991) argues that relying on huge market demand can enhance comparative advantage and economies of scale. In general, the larger the market size effect of a region, the more likely it is to reap the benefits of economic growth, meaning that the market potential can be more fully utilized to promote regional development (Clemente et al. 2009). Several studies have also affirmed the importance of market size advantages for regional development (Head and Mayer 2004; Brülhart et al. 2020). Particularly in the context of a “dual circulation” development pattern, the potential of domestic market demand is constantly released, which means that market demand advantages can drive innovative cities to offer various types of services, including adapting to economic changes, promoting social stability, maintaining ecological security and providing infrastructure, making cities competitive and advancing sustainable urban development. Therefore, when the market demand potential is greater, innovative cities are more likely to achieve urban resilience. As a result, research hypothesis 3 is proposed:

Hypothesis 3. The market demand effect positively moderates the relationship between the National Innovative City Pilot Policy and urban resilience.

Materials and methods

Data and sample selection

China has approved 103 innovative pilot cities in seven different batches from 2008 to 2022. It is crucial to highlight that the seventh batch, comprising 25 innovative pilot cities established in 2022, is excluded from the research scope of this paper due to its short construction period. Furthermore, cities such as Lhasa have large missing indicator data, and Changji and Shihezi are county-level cities, so these samples are excluded. Finally, the research data set consisted of 4200 observations drawn from panel data collected from 280 Chinese cities at the prefecture level between 2006 and 2020. This study examined the effect of the NICPP on UR using the program as a quasi-natural experiment. The 75 innovative pilot cities served as the treatment group, while the other non-pilot towns served as the control group. The objective was to assess the influence of the NICPP on UR. This research relies on the China City Statistical Yearbook for most city-level statistical indicators. The statistical bulletin of national economic and social development and the statistical yearbook of each prefecture-level city are consulted to fill in missing data. Patent data were sourced from the CSMAR database, while carbon emissions data were obtained from the Institute of Public and Environmental Affairs (IPE). For the remaining missing indicators, interpolation methods were employed for data completion.

Methods

Entropy weight TOPSIS method

The technique for order preference by similarity to ideal solution (TOPSIS) is used for multi-attribute decision-making. It assesses the distance between the evaluation object and the ideal solution (both positive and negative) and determines the best evaluation object by ranking them (Hwang and Yoon 1981; Chen 2019a). The entropy-weighting method (EWM) is more effective than subjective approaches like the analytic hierarchy process (AHP) in determining indicator weights. It helps avoid the bias induced by subjective preference (Chen 2019b). The entropy-weighting TOPSIS approach is utilized in this paper to objectively evaluate the UR composite index and estimate the distance between each sample and the ideal state. Here are the steps to calculate:

• 1. Data standardization. $X={({x}_{{ij}})}_{m\times n}$ is the original indicator evaluation matrix. The observed value of the j-th indicator for the i-th city is represented by x_ij (i = 1, 2,…,m; j = 1, 2,…, n). To mitigate the possible influence caused by the varying measurement units of each indicator in the UR assessment system of this study, it is imperative to standardize the original indicators. This will eliminate any potential impact and acquire the normalization matrix $Y={({y}_{{ij}})}_{m\times n}$. The higher value is preferable for positive indicators, while for negative indicators, the opposite is true. The following is the normalization equation used in the UR evaluation system for the positive and negative indicators:

  Positive indicators:

  $${y}_{{ij}}=\frac{{x}_{{ij}}-\min ({x}_{{ij}})}{\max ({x}_{{ij}})-\min ({x}_{{ij}})},{x}_{{ij}}\epsilon \left[0,1\right]$$

  (1)

  Negative indicators:

  $${y}_{{ij}}=\frac{\max ({x}_{{ij}})-{x}_{{ij}}}{\max ({x}_{{ij}})-\min ({x}_{{ij}})},{x}_{{ij}}\epsilon \left[0,1\right]$$

  (2)

• 2. Calculate the weight p_ij.

  $${p}_{{ij}}=\frac{{y}_{{ij}}}{\mathop{\sum }\nolimits_{i=1}^{m}{y}_{{ij}}}$$

  (3)

• 3. Calculate the information entropy e_j for indicator j.

  $${e}_{j}=-1/{\mathrm{ln}}\,m\mathop{\sum }\limits_{i=1}^{m}{p}_{{ij}}\,{\mathrm{ln}}\,{p}_{{ij}}$$

  (4)

• 4. The weights of the urban resilience index w_j.

  $${w}_{j}=\frac{1\!\!-\!\!{e}_{j}}{\mathop{\sum }\nolimits_{j=1}^{n}({1-e}_{j})},\mathop{\sum }\limits_{j=1}^{n}{w}_{j}=1$$

  (5)

• 5. Construct the weighted matrix ${c}_{{ij}}={w}_{j}{y}_{{ij}}$, then determine the Euclidean distance between the city and the ideal solutions, both positive and negative. For each indicator in the city, the positive ideal solution is ${c}^{+}=\left[{{c}_{1}}^{+},{{c}_{2}}^{+},\cdots ,{{c}_{n}}^{+}\right]$, whereas ${{\rm{c}}}^{-}=\left[{{{\rm{c}}}_{1}}^{-},{{{\rm{c}}}_{2}}^{-},\cdots ,{{{\rm{c}}}_{{\rm{n}}}}^{-}\right]$ is the negative ideal solution. What is the distance from the i-th city to both the positive ideal solution ${{d}_{i}}^{+}$ and the negative ideal solution ${{d}_{i}}^{-}$:

  $${{d}_{i}}^{+}=\sqrt{\mathop{\sum }\limits_{j=1}^{n}{\left({c}_{{ij}}-{{c}_{j}}^{+}\right)}^{2}}$$

  (6)
  $${{d}_{i}}^{-}=\sqrt{\mathop{\sum }\limits_{j=1}^{n}{\left({c}_{{ij}}-{{c}_{j}}^{-}\right)}^{2}}$$

  (7)

  In all cities, the j-th indicator’s maximum value is represented by ${{c}_{j}}^{+}$, while its minimum value is represented by ${{c}_{j}}^{-}$.

• 6. Calculate the urban resilience index for the i-th city.

$${f}_{i}={{d}_{i}}^{-}/({{d}_{i}}^{+}+{{d}_{i}}^{-}),{f}_{i}\in \left[0,1\right]$$

(8)

Econometric model

To assess the NICPP effect, this study establishes treatment and control groups using the NICPP as a “quasi-natural experiment” and the DID method. The DID method is effective in addressing the endogeneity problem in policy evaluation. The typical DID method is inappropriate since the national innovative pilot cities are developed in batches, and there are time differences in policy construction among cities. Thus, in line with Beck et al. (2010), the effect of the NICPP on UR is evaluated in this paper through the use of a time-varying DID model. Here are the model’s specifications:

$${{UR}}_{{it}}=\alpha +\beta {{NICCP}}_{{it}}+\gamma {{control}}_{{it}}+{\mu }_{i}+{\lambda }_{t}+{\varepsilon }_{{it}}$$

(9)

Among them, the explanatory variable ${{UR}}_{{it}}$ stands for the UR level of the i-th city in the t-th year. The variable ${{NICCP}}_{{it}}$ represents a dummy variable for the NICPP. ${{control}}_{{it}}$ refers to variables that influence the UR, such as the level of economic development, government intervention, financial development, external dependence degree, technological innovation, and human capital. The variables ${\mu }_{i}$ and ${\lambda }_{t}$ correspond to the fixed effects for cities and years, respectively, whereas ${\varepsilon }_{{it}}$ represents the random disturbance term. $\beta$ is the core estimation parameter. A considerably positive value of $\beta$ indicates that the UR level may be greatly improved by the construction of innovative cities, which also verifies hypothesis 1. On the contrary, it means that it does not achieve the policy effect.

Variable definition

Explained variable: urban resilience (UR)

Drawing on the research conducted by Cutter et al. (2014), this paper considers the concept of UR with the current state of urban development. It adheres to principles of systematic, scientific, representativeness, and accessibility. It deconstructs UR into four sub-dimensions: economic resilience, social resilience, ecological resilience, and infrastructure resilience. Table 1 shows the 28 indicators refined to build the UR evaluation index system, and the UR index is measured using the entropy weight TOPSIS method.

• (1) Economic resilience: the capacity of urban economic systems to withstand shocks, recover rapidly, adapt, and modify is known as economic resilience (Martin et al. 2015). Indicators were selected to measure economic resilience in terms of economic strength, industrial structure, and foreign investment (Zhou et al. 2021; Xia and Zhai 2022; Ye et al. 2022; Tang et al. 2023; Jiang and Jiang 2024).

• (2) Social resilience: urban social systems’ affordability to withstand external risks, provide social security, and maintain smooth operation is measured by their social resilience (Pagliacci and Russo 2020; Sajjad et al. 2021). This paper measures social resilience from four aspects: population livability, talent pool, social security, and employment status, covering seven indicators (Liu et al. 2021; Zhou et al. 2021; Xia and Zhai 2022; Ye et al. 2022).

• (3) Ecological resilience: the ability of urban ecosystems to coordinate, manage uncertainty, and preserve ecological security and stability in the face of ecological pressure is emphasized by ecological resilience. Ecological resilience is important for achieving urban resilience and sustainable development (Zhao et al. 2021; Li and Wang 2023). Therefore, this study focuses on ecological pressure, governance capability, and environmental endowment (Zhou et al. 2021; Ye et al. 2022; Tang et al. 2023; Jiang and Jiang 2024).

• (4) Infrastructure resilience: infrastructure resilience carries the urban problems brought about by the urbanization and industrialization process (Tang et al. 2023), providing the safeguard capacity that cities need. We have selected seven indicators from the aspects of municipal engineering construction and public service supply capacity perspectives (Liu et al. 2021; Zhou et al. 2021; Xia and Zhai 2022; Ye et al. 2022).

Table 1 Evaluation indicator system of the UR.

Explanatory variable: the National Innovative City Pilot Policy (NICPP)

In this study, the NICPP dummy variables are the main explanatory variables. Considering that innovative cities are established in batches with differences in approval times, we construct the NICPP dummy variable by setting it to 1 for the year of approval as an innovative city and 0 for all other years.

Control variables

Drawing upon the existing literature, this paper selects several key control variables that may affect the explanatory variable UR. These variables encompass: (1) Economic development level (lnpergdp): calculated using the per capita GDP natural logarithm. The level of economic development is an important factor affecting UR, and economic growth can promote social stability, which in turn enhances UR (Jiang and Jiang 2024). (2) Government intervention degree (gov): measured by taking the logarithm of the local government’s budgeted expenditures. The operation of cities is inseparable from financial investment, and government financial support can mobilize social resources, provide financial support and guarantee for urban economic and social development, ecological construction, and infrastructure construction, which in turn affects UR (Zhou et al. 2021). (3) Financial development level (finance): defined as the proportion of the loan balance held by financial institutions to the GDP at year’s end. Expanding finance can revitalize idle social capital, provide financial support for innovative entrepreneurial activities and infrastructure development, and affect the resilience of urban economies (Zhou et al. 2022; Jiang and Jiang 2024). (4) External dependence degree (open): the share of GDP attributable to imports and exports. Import and export trade reflects the degree of openness of cities and affects their ability to withstand potential external risks (Zhou et al. 2022). (5) Human capital (human): quantified as the proportion of students enrolled in general higher education institutions relative to the urban registered population. The development and transformation of cities require the support of talent, which is also a key factor for cities to remain competitive and resilient (Zhang and Wang 2022; He et al. 2023). (6) Consumption level (consume): measured by the logarithm of the aggregate retail sales of consumer products in society. To a certain extent, the level of consumption represents the market demand and scale, and the increase in market demand can enhance UR (Zhou et al. 2022).

Mechanism variables

• (1) Industrial diversification: external uncertainties will impact UR, while inter-industry diversification of risks can effectively diversify risks, guide the rational allocation of innovation factors, optimize industrial structure, and enhance resilience. According to Frenken et al. (2007), this paper breaks down the industrial diversification index into two components, related and unrelated diversification, by using the entropy index as a building block. Related variety (RV) refers to the level of diversification among subcategories with strong input-output linkages within major industry categories. Unrelated Variety (UV) refers to the level of diversification among major industry categories, reflecting the relatively weak industrial linkages among major categories. Assuming that the region can be divided into S major industry categories and n sub-industries, the percentage of city employees employed in the subcategory industry $i(i=\mathrm{1,2},\ldots ,n)$ is denoted as ${P}_{i}$; ${P}_{S}$ is the employment share of each major industry, which is obtained by the sum of the proportion of employment in several sub-industries, i.e., ${P}_{S}=\sum _{i\in S}{P}_{i}$. According to the entropy index method, the overall diversification (of the city can be expressed as: ${DIV}=\mathop{\sum }\nolimits_{i=1}^{n}{P}_{i}\mathrm{ln}\left(1/{P}_{i}\right)$. A greater value of the industrial diversification index signifies a heightened degree of industrial decentralization, resulting in enhanced resilience to shocks and increased autonomous control. The specific calculations are as follows:

  $$\begin{array}{lll}{DIV}=\mathop{\sum }\limits_{i=1}^{n}{P}_{i}{\mathrm{ln}}\left(\frac{1}{{P}_{i}}\right)=\mathop{\sum }\limits_{S=1}^{S}\mathop{\sum}\limits_{i\in S}{P}_{i}{\mathrm{ln}}\left(\frac{1}{{P}_{i}}\right)=\mathop{\sum }\limits_{S=1}^{S}\mathop{\sum}\limits_{i\in S}{P}_{i}\left[{\mathrm{ln}}\left(\frac{{P}_{S}}{{P}_{i}}\right)+{\mathrm{ln}}\left(\frac{1}{{P}_{S}}\right)\right]\\\qquad\;=\mathop{\sum }\limits_{S=1}^{S}\left[\mathop{\sum}_{i\in S}{P}_{S}\left(\frac{{P}_{i}}{{P}_{S}}\right){\mathrm{ln}}\left(\frac{{P}_{S}}{{P}_{i}}\right)\right]+\mathop{\sum }\limits_{S=1}^{S}\left[\mathop{\sum}\limits _{i\in S}{P}_{i}{\mathrm{ln}}\left(\frac{1}{{P}_{S}}\right)\right]\\\qquad\,\,=\mathop{\sum }\limits_{S=1}^{S}{P}_{S}\left[\mathop{\sum}\limits_{i\in S}\left(\frac{{P}_{i}}{{P}_{S}}\right){\mathrm{ln}}\left(\frac{{P}_{S}}{{P}_{i}}\right)\right]+\mathop{\sum }\limits_{S=1}^{S}{P}_{S}{\mathrm{ln}}\left(\frac{1}{{P}_{S}}\right)\\\qquad\,\,={RV}+{UV}\end{array}$$

  (10)

  Concerning Sun and Chai (2012), this paper categorizes the service industry into production, consumption, distribution, and social services. Built upon the foundation of the three industries, the sector is segmented into 6 major categories and 19 sub-categories, encompassing: primary industry (agriculture, forestry, animal husbandry and fishery), secondary industry (extractive industry, electricity, gas and water production and supply, manufacturing industry, construction industry), production services, consumer services, distribution services, and social services. Quantitative approach: the city-industry-year level employment data are summarized based on the data of each prefecture-level city in the China Urban Statistical Yearbook, and the employment numbers of 6 major categories and 19 sub-categories of industries in 280 cities are extracted, and the overall diversification, related diversification and unrelated diversification indices of cities are step-by-step calculated by the formula of industrial diversification. Employment data for all industries are from the China City Statistical Yearbook.

• (2) Market demand effect: when demand is high in the internal market of a city, it tends to accelerate the flow of production factors to that city, attracting enterprises to locate there to gain a larger market share. In addition, the market potential of cities provides the necessary conditions for urban development. The high demand for cities in the internal market gives the internal circulation more space and flexibility to cope with external shocks caused by unexpected factors. Increased market potential promotes urban development, and this demand stimulates the breeding of new development models through technological innovation, creating new development advantages, which in turn lead to the goal of reducing risk and increasing resilience. Therefore, this paper refers to Harris (1954) and uses the market potential indicator for each city to measure the market demand effect. Moreover, this is how the equation is configured:

$${{MP}}_{it}={\sum }_{j\ne i}\left(({{GDP}}_{jt}/{d}_{{ij}})+({{GDP}}_{{it}}/{d}_{ii})\right)$$

(11)

Where ${{GDP}}_{jt}$ and ${{GDP}}_{{it}}$ represent the real GDP of city j and city i in the sample period, with 2006 as the base period. ${d}_{{ij}}$ represents the geographical distance from city i to city j; ${d}_{ii}$ it is the intra-city distance, i.e., ${d}_{{ii}}=0.66\times \sqrt{{{area}}_{i}/\pi }$, where ${{area}}_{i}$ denotes the city’s land area. The data are also from the China City Statistical Yearbook.

Descriptive statistics

The descriptive statistics for the variables are summarized in this section (Table 2).

Table 2 Descriptive statistics of variables.

Empirical results and analysis

Benchmark regression

Column (1) of Table 3 illustrates the impact of innovative cities on urban resilience. Columns (2–5) examine the impact of innovative cities on the four dimensions of urban economic, social, ecological, and infrastructure resilience, respectively. The results find that the NICPP can significantly enhance urban resilience (UR). This verifies Hypothesis 1. In terms of sub-dimensional resilience levels, the regression coefficients of the NICPP on economic, social, and infrastructure resilience are significantly positive, while the regression coefficients on ecological resilience are significantly negative. This indicates that the NICPP is mainly to enhance the overall urban resilience level by enhancing the city’s economic resilience, social resilience, and infrastructure resilience. The enhancement effect of these three dimensions is much greater than the negative impact of urban ecological resilience.

Table 3 Benchmark regression results of the NICPP affecting the UR.

Parallel trend test

The fundamental requirement for employing the DID method to discern policy effects is fulfilling the parallel trends assumption. This assumption posits that the treatment group and control group should exhibit consistent trends in urban resilience before implementing the policy. Given the varied establishment times of innovative cities, the traditional DID method, which sets time dummy variables based on a specific year of policy impact, is not suitable. Instead, it is essential to introduce a dummy variable related to the chronological order of each inventive city’s establishment. Therefore, a parallel trend test is carried out in this article using the event research approach proposed by Beck et al. (2010) to conduct a parallel trend test. Here is the set for the specific model:

$${{UR}}_{{it}}=\alpha +\mathop{\sum }\limits_{k=-4}^{+5}{\beta }^{k}{{{NICCP}}^{k}}_{{it}}+\gamma {{control}}_{{it}}+{\mu }_{i}+{\lambda }_{t}+{\varepsilon }_{{it}}$$

(12)

In this setup, ${{{NICCP}}^{k}}_{{it}}$ there is a series of dummy variables. The ${{{NICCP}}^{-4}}_{{it}}$ indicates that the treatment group consists of the 4th year and above before the innovative pilot city, and the ${{{NICCP}}^{5}}_{{it}}$ indicates that the treatment group consists of the 5th year and above after the innovative pilot city, that is to say, the relative time before the period of −4 is categorized as the period of −4, and those after the period of 5 is categorized as the period of 5. Simultaneously, to avoid multicollinearity, this paper uses the sample of the 5th year before the innovative city pilot as the reference, i.e., the dummy variable of k = −5 period is deleted. Other variable definitions are consistent with the benchmark regression model setting. The NICPP is an important embodiment of China’s construction of an innovative country and the implementation of an innovation-driven development strategy, and can be used as a quasi-natural experiment. Essentially, the pilot policy is selected among all cities, and the list of pilot cities is selected based on the policy documents of the MOST and the NDRC. The data in this paper include a sample of 280 prefecture-level cities in China, of which 75 are innovative pilot cities in the treatment group and 205 other non-pilot cities are in the control group.

Figure 2 shows that the coefficients of the relative time dummy variables before the implementation of the policy are insignificant. This suggests there was no significant difference in UR between innovative cities (treatment group) and non-innovative cities (control group) before the implementation of the NICPP. Therefore, the NICPP satisfies the assumption of parallel trends. After the implementation of the innovation policy, there is a significant difference between the treatment and control groups, and the impact coefficient of the NICPP is significantly positive and continues to increase, indicating that the impact of the NICPP on UR is consistently effective.

Fig. 2

Parallel trend test.

To further verify the credibility of the parallel trend, this paper applies two methods to examine the UR development trends of the treatment and control groups before the implementation of NICPP. (1) The first method: investigating the UR trajectories of the treatment group (selected innovative cities) and the control group (not selected innovative cities) in the first 1–4 years of NICPP by relative years, as shown in Fig. 3. It is found that although there is a certain difference between the UR trajectories of the treatment group and the control group, the difference is essentially the same, which justifies the rationality of the parallel trend. (2) The second method: drawing on Du and Li (2022), the UR development trend of the treatment and control groups before the implementation of NICPP is examined. Given that only Shenzhen was selected as the NICPP in 2008 and more cities were approved as NICPP in the second batch in 2010, 2010 was used as an illustration to examine whether the UR development trend of the treatment group and the control group before 2010 was consistent. The results are shown in Fig. 4. Figure 4 intuitively reflects that the UR trajectories of the treatment and control group cities were essentially parallel before the establishment in 2010. Therefore, the necessary prerequisites for using DID are satisfied.

Fig. 3

Parallel trend test for method 1.

Fig. 4

Parallel trend test for method 2.

Overall, Figs. 3 and 4 show that in the absence of NICPP, the treatment and control cities follow similar UR trends. The parallel trend test has high credibility.

Placebo test

The control variables can effectively capture urban characteristics in the benchmark regression model (Eq. 9), but it is still possible that other random factors affect the evaluation effect of the NICPP. This study conducts a placebo test following the methodology described in (Chetty et al. 2009) to verify the robustness of the benchmark regression results. Since innovative cities are subject to different policy shock time points, we randomly generate two dummy variables: “pseudo-innovative city” and “pseudo-policy shock time”, i.e., we randomly select the policy shock time for each sample (Zeng et al. 2023). As an illustration, the treatment group consists of 75 cities randomly chosen from a pool of 280 samples each time, with the timing of policy shocks randomly assigned. This simulation process is iterated 500 times to enhance the reliability of the placebo test results.

Figure 5 plots the kernel density of the estimated coefficients and their p-value distribution after 500 random samples. The results show that the estimated coefficients of the sampled regressions approximate a normal distribution with a mean of 0, with most p-values exceeding 0.1 and significantly different from the benchmark regression coefficient of 0.0247. These results suggest that the conclusion that the NICPP improves UR is unlikely to be affected by non-random factors, and the conclusions of this paper pass the placebo test.

Fig. 5

Placebo test.

Robustness tests

Multi-point PSM-DID

To alleviate the potential sample selection effect caused by the non-random selection of national innovative city pilots, this research employs the PSM-DID model to conduct robustness testing and ensure the reliability of the results. We utilize both a period-by-period matching method and a constructed cross-sectional PSM method for propensity score matching (Böckerman and Ilmakunnas 2009). The approach follows: First, we choose control variables as matching variables, and then we process the matching samples by using a kernel-matching approach. Second, to construct the necessary panel data, we employ a strategy that involves matching city samples on a year-by-year basis using a period-by-period matching approach. After that, we vertically combine the matched data for each year. The cross-sectional PSM is built using the nearest-neighbor matching approach to determine which control group is closest to the innovative pilot cities in the treatment group. Following these matching procedures, a balance test is conducted on the generated matching data from both methods. Ultimately, we employ the time-varying DID approach to re-estimate the impact of the NICPP on UR.

The PSM-DID regression outcomes are presented in Table 4, Columns (1 and 2). The estimated coefficients related to the NICPP are notably positive and statistically significant. This suggests that the implementation of NICPP is beneficial for enhancing urban resilience. Importantly, these findings affirm the robustness of the results observed in the benchmark regression analysis.

Table 4 Results of the robustness test.

Heterogeneity treatment effects test

The estimation of time-varying DID is essentially a weighted average of the treatment effects of all individual treatments. Time-varying DID model for policy evaluation may be subject to heterogeneous treatment effects, and in this context even if the parallel trend assumption is satisfied, estimation results based on two-way fixed effects can still suffer from estimation bias (de Chaisemartin and D’Haultfoeuille, 2020). To this end, this paper examines the possible heterogeneity treatment effect of the benchmark regression model.

First, the two-stage difference-in-difference (did2s) test proposed by Gardner (2022) is used. In the first stage, the heterogeneity of the group treatment effect and period treatment effect was identified, and the treatment variables were regressed after the heterogeneity treatment effect was eliminated in the second stage. The results are shown in column (3) of Table 4, and the estimated coefficient of the robust estimator is found to be 0.029, which is significant at the 1% level, indicating that the impact of heterogeneous treatment on the estimation results in this paper is relatively limited.

Second, this paper further uses the bacondecomp command to decompose the multi-period DID estimators. The results in column (4) of Table 4 show that the estimated coefficients remain significantly positive. From the decomposition results, it can be seen that the estimated results mainly come from the treatment group and the effect of the results with the never-received treatment group, which is 90.18%. Heterogeneous treatment effects due to time variance are generally not observed in this group. The above conclusions indicate that the heterogeneous treatment effect has a more limited impact on the estimation results in this paper.

Other robustness test

• (1) Excluding municipalities: compared to other cities, municipalities directly under the central government, such as Beijing, Tianjin, Shanghai, and Chongqing, have a higher administrative hierarchy than other cities and have significant advantages in politics, economy, culture, and other aspects (Du and Li 2022; He et al. 2024). These cities possess resource endowments and economic volume unmatched by other cities, they are at the forefront of economic development, scientific and technological innovation, infrastructure development, and policy endowment conditions. This also makes it possible that the increase in urban resilience may be a result of innovative cities or due to these cities’ advantages. Therefore, to exclude the potential bias in the benchmark model, this paper excludes these city samples and then re-estimates the benchmark model. The estimates are shown in column (5) of Table 4.

• (2) Excluding other policy effects: considering that NICPP occurred in the same period, accompanied by the implementation of other policies, these related policies may have an impact on the policy effects of the NICPP. Policies such as the smart city pilot policy, the “Broadband China” pilot policy, and the sponge city pilot policy are closely related to this paper (Yang et al. 2022; Jiang and Jiang 2024). Therefore, to avoid the bias of the benchmark regression results caused by the interference of other similar policies during the sample period, this paper separately includes the above policy dummy variables in the benchmark regression model to control their potential impact. The results of the study are shown in columns (1–3) of Table 5, which represent the results of controlling the above three policies individually, and column (4), which represents the results of controlling the three policies at the same time. It can be found that controlling for the above disturbing policies, the NICPP still has a significant positive effect on urban resilience. This indicates that the pilot policy of innovative cities has not been affected by the above policies, and the benchmark conclusion is robust.

  Table 5 Excluding other policy effects.

(3) Replacing indicator weights: differences in the indicators' weighting methods may lead to differences in the results of the benchmark regression. If the weights of the key indicators are underestimated or overestimated, it is easy to ignore the impact of other potential factors on the results, thus reducing the accuracy of the results. To avoid the bias of the benchmark regression results caused by the changes in indicator weights, this paper adds four methods of determining the indicator weights, such as the objective weighting method such as the informative weighting method, and the CRITIC weighting method, the subjective weighting method such as the AHP hierarchical analysis method, and the combined weighting method combining the entropy weighting method (EWM) and the AHP hierarchical analysis method, and re-calculates the UR index to ensure the robustness of the results. The results show (as shown in Table 6) that the positive effect of NICPP on UR does not vary with different weights, which indicates that the change of weights is insensitive to the results and verifies the robustness of the benchmark regression results.

Table 6 Results using different weighting methods.

Further analysis

Heterogeneity analysis

The benchmark regression results show that innovative pilot cities can significantly enhance urban resilience. However, the enhancement effect of innovative pilot cities on urban resilience may show different effects depending on the location, scale, and administrative level of the city.

City location

Considering the differences in the economic foundations, population density, resource endowment, and industrial agglomeration among Chinese cities, the enhancement effect of innovative cities on UR may also vary depending on the geographic location. Therefore, we examined the resilience-enhancing effects of cities on the east and west sides of the Hu Huanyong Line and measured the coefficient differences between groups based on Fisher’s Permutation test. Table 7 presents the regression findings for Hu Huanyong’s east and west sides in columns (1) through (2). The p-value for group coefficient differences is 0.018, implying the existence of geographic heterogeneity in the urban resilience enhancement effect. Specifically, the estimated coefficient of NICPP for cities on the eastern side of the Hu Huanyong Line is 0.026, while the estimated coefficient for cities on the western side is 0.005. This implies that cities on the east side have a bigger enhancement in resilience. The possible explanation is that these eastern cities have higher population density, better economic foundations, greater industrial agglomeration, and relatively favorable innovative ecosystems. This environment can attract the innovative elements needed for pilot city construction. UR can be enhanced by improving the innovation capacity, giving full play to the NICPP effect.

Table 7 Heterogeneity test results.

City scale

The size of cities reflects differences in technological innovation resources, market capacity, and resilience governance capabilities. Therefore, it is essential to examine whether the enhancement effects of the NICPP on UR present heterogeneity in city size. Referring to the “Notice of the State Council on Adjusting the Standards for Urban Scale Division,” cities with a permanent population of more than 3 million are classified as large cities, and cities with a permanent population of less than 3 million are classified as small and medium-sized cities. Similarly, the bootstrap-based Fisher’s Permutation test is set to sample 1000 times for between-group coefficient difference tests. Columns (3 and 4) of Table 7 show the regression results for large cities and small and medium-sized cities, respectively. The p-value for group coefficient differences is 0.000, indicating the existence of urban scale heterogeneity in the enhancement effect of UR. The analysis reveals a notable positive coefficient for large cities, and the resilience-enhancing effect is greater than that of small and medium-sized cities. Possible reasons include, according to the siphon law of urban scale expansion, large cities have a strong attraction, especially represented by cities such as Beijing, Shenzhen, and Shanghai. These megacities exhibit more prominent innovation resource aggregation, resource endowment, and resilience governance capabilities, resulting in a stronger urban carrying capacity. Furthermore, megacities are at the forefront of innovative construction, with complete innovation-supporting equipment, which can transform innovation resources on a larger scale.

City administrative

The urban resilience-enhancing effects of the NICPP may also be heterogeneous across cities in terms of policy incentives, infrastructure, and comparative advantages. Based on the differences in administrative levels, this paper classifies municipalities, provincial capitals, cities specifically designated in the state plan, and special economic zones as high administrative level cities, while other cities are classified as low administrative level cities. The outcomes of the heterogeneity analysis according to urban administrative levels are reported in columns (5 and 6) of Table 7. The p-value for the group difference coefficient is 0.010, indicating the existence of heterogeneity in the urban resilience enhancement effects across administrative levels. Specifically, the resilience-enhancing effect of the NICPP is more pronounced in cities with lower administrative levels. The possible explanation is that cities with high administrative levels are oversupplied with innovative resources, leading to underground resource allocation efficiency and showing diminishing marginal returns effect (Au and Henderson 2006). In contrast, under the same policy conditions, the substantive innovations brought by innovative cities are more beneficial to cities with low administrative levels and generate greater marginal benefits, which also implies that the resilience-enhancing effect of the NICPP is more pronounced in cities with low administrative levels.

Analysis of moderating effects

Based on the previous analysis, this paper confirms the causal relationship between the NICPP and UR. To verify the moderating effect of industrial diversification and market demand effects on innovative cities to enhance urban resilience, this paper introduces the cross-multiplication treatment between the moderating variable M and the policy variable NICCP based on the benchmark regression model Eq. (9) and re-incorporates the interaction term with the moderating variable into the model, and the econometric model is constructed as follows:

$$\begin{array}{l}{{UR}}_{{it}}=\alpha +\beta {{NICCP}}_{{it}}+\delta {{UR}}_{{it}}\times {M}_{{it}}+\theta {M}_{{it}}+\gamma {{control}}_{{it}}\\\qquad\quad\,\,+\,{\mu }_{i}+{\lambda }_{t}+{\varepsilon }_{{it}}\end{array}$$

(13)

Where $M$ represents the moderating variables for industrial diversification and market demand effects, respectively; $\delta$ is the interaction term coefficient between the moderating variable and the NICPP. The meaning of the remaining variables is consistent with the benchmark model. If the interaction term coefficient $\delta$ is significantly positive, it indicates that the moderating variable positively affects the relationship between the NICPP and UR. The regression results of the moderating effect are displayed in Table 7.

Industrial diversification

Columns (1–3) of Table 8 report the moderating effects of overall diversification, related diversification, and unrelated diversification on the relationship between innovative cities and urban resilience, respectively. The regression findings show that the interaction coefficients are all significantly positive, indicating that the variables of overall diversification, related diversification, and unrelated diversification present a positive moderating effect on the promotion of urban resilience by the NICPP. At the same time, the significance level of unrelated diversification is greater than that of related diversification, which means that unrelated diversification has a stronger positive moderating effect on innovative cities to promote urban resilience. In other words, industrial diversification can positively moderate the relationship between the NICCP and urban resilience, and the positive moderating effect of unrelated diversification is stronger.

Table 8 Results of the mechanism test.

The accelerated development of urban industrial diversification is conducive to getting rid of single-industry path dependence, enabling innovative cities to effectively deal with risky fluctuations and enhancing regional competitiveness and resilience. Related industrial agglomeration is conducive to guiding the collaborative innovation of resource elements and technology diffusion (Jacobs 1969; Boschma and Iammarino 2009), strengthening the role of innovative cities in enhancing urban resilience. Of course, according to portfolio theory, the closer the inter-industry linkage, the more likely it is to lead to multi-industry fluctuations and chain reactions when perturbed by external shocks (Sun and Chai 2012). Unrelated diversification, however, has no obvious techno-economic linkages, and can better play the innovative role of innovative cities, promote the cross-integration of new knowledge and new technologies among different industries, form industrial complementarities, mitigate the scope and intensity of the diffusion of external shocks among industries, and enhance urban resilience. Therefore, the moderating effect of unrelated diversification on the relationship between innovative cities and urban resilience is more significant than that of related diversification. In summary, overall diversification, related diversification, and unrelated diversification all play have a positive moderating effect on the resilience-driving effect of innovative cities, and the moderating effect played by unrelated diversification is more significant. This also validates the hypotheses H2a and H2b.

Market demand effect

Column (4) of Table 8 provides the regression results of the moderating effect of the market demand effect. As can be seen from the results, the coefficient of the interaction term between the market demand effect and the NICPP is 0.024, which is significant at the level of 5%, indicating that the market demand effect has a significant positive impact on the relationship between innovative cities and urban resilience. The continuous release of market demand potential will inject vitality into the constant creation of new products and services for cities, generating positive externalities and amplifying the role of innovative cities in promoting the optimization of industrial change, accelerating the transformation of innovation outcomes, and thus enhancing the city’s competitiveness. In addition, the synergistic effect generated by market demand and innovation can maximize the advantages of economies of scale, satisfy the domestic and international double-cycle development space, and show strong urban resilience. Overall, the market demand effect contributes to the positive relationship between the NICCP and UR, which also verifies hypothesis H3.

Analysis of spatial spillover effects

This paper further analyses the spillover effects of the NICPP and explores whether it can lead to the enhancement of UR in the surrounding areas. The nested weight matrix is usually used to describe situations with hierarchical or multi-scale spatial dependence. Traditional spatial weight matrices typically capture only one level of spatial dependence (e.g., economic or geographic linkages in adjacent regions). In contrast, geo-economic nested weight matrices can capture multiple levels of dependencies simultaneously. For example, in economic geography, there may be different levels of dependence between regions: interactions of localized economic activities, spillover effects of regional economic policies, etc. Referring to Dian et al. (2024), this paper selects the geographic-economic nested weight matrix and finally uses the spatial Durbin model (SDM) to study the spatial effects after passing the LR and LM tests. Under the setting of the geographic-economic nested weight matrix, the global Moran’s I index is greater than 0, which indicates that the NICPP and UR show a positive autocorrelation in space, that is, there is spatial agglomeration.

The regression results of the SDM in column (1) of Table 9 show that NICCP is significantly positive at the 1% significance level and its spatial lag term ρ is also significantly positive, suggesting that the NICPP will have a positive spatial spillover effect on UR. Since the spatial lag term of the explanatory variable NICPP is considered in the SDM, its explanatory variable NICPP has an impact on the explanatory variable UR of both local and neighboring cities. Therefore, the partial differential effects decomposition method is used to decompose the total effects in the SDM into direct effects and indirect effects (LeSage 2008), and the decomposition results are shown in columns (2–4) of Table 9. The direct effects in column (2) show that NICPP has a positive impact on the UR in the region, including feedback effects, that is, the spatial feedback effect that NICPP in the region affects UR in neighboring regions, which in turn affects UR in the region. The indirect effect result in column (3) is the spatial spillover effect, and the coefficient of NICPP is significantly positive, indicating that NICPP has a positive spillover effect on the UR of neighboring regions. Column (4) is the total effect regression result, which is the sum of direct and indirect effects, including both the direct promotion effect of NICPP on the resilience of local cities and the positive spillover effect of NICPP on the resilience of neighboring cities. Overall, the NICPP can enhance UR as a whole. This also confirms that the NICPP has a radiation-driven effect on the resilience improvement of surrounding cities.

Table 9 Spatial Dubin model decomposition results.

To ensure the reliability of the spatial analysis, we additionally add the results of three matrix tests, such as the economic distance matrix, the economic geography matrix, and the economic geography weighting matrix. Drawing on the idea of Zhou et al. (2023), this paper further selects other alternative spatial weight matrices for the robustness test to improve the credibility of the spatial analysis. W1, W2, W3, and W4 are the economic geographic nested matrix, the geospatial distance matrix, the economic distance matrix, and the economic geospatial weight matrix, respectively. The above results (shown in Table 10) show that the SDM produces essentially consistent results under the four spatial matrices, which proves the robustness of the empirical results.

Table 10 Direct and indirect effects under four weighting matrices.

Conclusions and policy recommendations

Conclusions

Against the imperative for innovation-driven development, establishing innovative cities holds significant importance in enhancing the UR. Employing the NICPP as a quasi-natural experiment, this study utilizes panel data from 2006 to 2020, encompassing 280 prefecture-level cities. A comprehensive urban resilience index is constructed and measured, then employing a time-varying DID model to examine the influence of the NICPP on UR and its mechanism of action, and explores the spatial spillover effect of the NICPP on UR. This paper improves the theoretical logic of the NICPP on urban resilience and enriches the index system and measurement method of the UR.

We have derived three research conclusions: (1) benchmark regression results indicate that the NICPP significantly enhances the UR. Following a battery of robustness checks, the findings of the benchmark regression remain robust. These checks include cross-sectional PSM and period-by-period matching in the PSM-DID approach, placebo tests, and excluding other policies in the same period. (2) Heterogeneity analysis shows variations in the impact of the NICPP on UR, depending on factors such as city location, size, and administrative level. Specifically, the resilience-enhancing effect of innovative city construction on resilience is particularly pronounced for cities located east of the Hu Huanyong line, as well as for large cities and lower administrative level cities. (3) Mechanism test results demonstrate that industrial diversification (both related and unrelated diversification) and market demand effects positively moderate the resilience-enhancing effect of the NICPP. (4) The results of the SDM reveal that the NICPP not only enhances UR of the region but also has a significant spatial spillover effect and positively promotes UR of the neighboring regions.

Policy recommendations

This paper contributes theoretical insights and empirical evidence, advancing our understanding of the influence of the NICPP on UR. It offers practical implications for cities to implement innovation development strategies and enhance urban resilience. Derived from the research findings, three key insights are suggested:

• (1) Centered on the innovation-driven development strategy, persistently advancing the construction of innovative cities serves as a focal point, fully utilizing their spatial spillover effect, and offering new impetus for UR. On the one hand, cities need to give full play to the innovation advantages, fully utilize the new generation of information technology, and systematically lay out key tasks such as urban emergency management and sustainable development, to provide effective technical support for the construction of resilient cities. For example, cities such as Hefei and Nanjing have built risk prediction platforms based on intelligent technology, successfully early warning pipeline leakage, and other types of emergencies, and effectively improve disaster prevention and emergency response capabilities for constructing resilient cities, providing useful experience. On the other hand, Shenzhen, Guangzhou, Hangzhou, and other cities rich in innovative resources play a regional driving role in promoting the construction of resilient cities to the radiation of neighboring cities to achieve synergistic development.

• (2) The construction of innovative cities should consider various factors such as geographic location, scale, and administrative level, implementing a differentiated support strategy. The government should design reasonable innovation development paths, combine urban development advantages, advocate diversified and differentiated development strategies, and leverage the full potential of innovative cities to enhance resilience. For example, relying on its location advantages, Suzhou has gathered innovative resources through international scientific and technological cooperation, such as the China-Singapore Cooperation Suzhou Industrial Park, and has achieved remarkable results. Western cities such as Nanning and Yinchuan with relatively poor innovation resources can enhance innovation momentum through east-west linkage and industry-university-research cooperation, integrate scientific and technological innovation into urban strategic planning, and enhance urban innovation capacity and UR.

• (3) Emphasizing the positive synergistic effects of transmission mechanisms such as industrial diversification and market demand effects in innovative cities. Cities with single leading industries, especially resource-based cities, need to promote industrial transformation through scientific and technological innovation, accelerate industrial diversification, and cooperate with other cities in the urban agglomeration around the traditional leading industries of the city to share the benefits of innovation. At the same time, cities need to adhere to demand-oriented, transform traditional advantageous industries through technological innovation, develop new economies, and improve the region’s ability to resist risk fluctuations. For example, Huzhou adheres to the guidance of scientific and technological innovation, transforms traditional advantageous industries such as textiles, cultivates and expands emerging industries such as information technology and biomedicine, promotes the transformation of the city to be driven by innovative factors, and enhances the city’s sustainability and resilience.

Research limitations

This paper uses a time-varying DID model to explore the impact of innovation policy on urban resilience and its mechanism and spatial measurement to explore its spillover effects. However, in the future, we can try to analyze the possible transmission paths from more micro-channels to enrich and deepen the understanding of the mechanism. In terms of theoretical framework, we can try to construct a theoretical link between innovation policy and urban resilience by deductive reasoning.