A Study on the Impact of Dual Pilot Smart Cities and Innovative Cities on City Resilience

Abstract

Smart cities and innovative cities are important strategies for enhancing city resilience. By using panel data from 285 Chinese cities from 2009 to 2022 and examining “dual pilot” cities via the concurrent implementation of smart and innovative cities as a quasi-natural experiment, this study employs the difference-in-differences model to examine the impact of the “dual pilot” initiative. Furthermore, quantitative assessments are conducted from multiple perspectives through heterogeneity analysis, mechanism analysis, and spatial spillover effect analysis. The findings are as follows: “Dual pilot” cities have the ability to enhance city resilience and have a synergistic effect. The effect on city resilience is significantly greater than that of the “single pilot” design, but policy synergy effects are sensitive to the policy’s implementation sequence. A mechanism test reveals that innovation and development levels and industrial structure upgrading are important paths for “dual pilot” cities to boost city resilience. The heterogeneity study demonstrates that the positive impact of “dual pilot” cities on city resilience is statistically significant only in ordinary cities; non-resource-dependent cities; and cities across eastern, central, and western regions. A spatial spillover analysis demonstrates that “dual pilot” cities exert positive spillover effects on both the implementing cities and their neighboring areas. The above conclusions can serve as a source of reference and inspiration for building “resilient cities”.

1. Introduction

The rapid growth of cities brings about many uncertainties. The concentration of rural economic activities in urban centers caused by the fragmentation of urban layouts leads to new urban governance problems, which makes the vulnerability and powerlessness of cities increasingly obvious [1,2,3]. Therefore, it is crucial to enhance the disaster risk management capabilities of cities and to improve urban governance [4]. In this context, China‘s 14th Five-Year Plan proposed the construction of “resilient cities”, which has emerged as a key strategic path for both theoretical innovation and practical breakthroughs. It advocates for sustainable development through resilient cities, which can rapidly withstand uncertain risk shocks [5], and its basic role in urban transformation is increasingly apparent.

The term “resilience” stems from the Latin “resilio”, historically denoting a “return to an original state” [6]. Initially emerging in physics, the concept was subsequently adopted in ecology before being extended to engineering and socio-ecological systems [7]. The theoretical underpinnings of resilience were significantly advanced by Canadian ecologist C. S. Holling, who proposed the “hierarchy, chaos, and adaptive cycle” framework [8]. This model provided a holistic interpretation of sustainable development by challenging linear and equilibrium-based perspectives, while introducing the notion of cross-scale dynamic interactions and systemic innovation. Holling’s work thus established a conceptual foundation for city resilience theory. In 2002, the International Council for Local Environmental Initiatives (ICLEI) formally introduced the concept of the “resilient city,” positioning it as a critical element in urban planning and disaster risk reduction. This initiative stimulated a growing body of research focused on city resilience. The concept gained further international recognition during the 2005 World Conference on Disaster Reduction, where resilience was emphasized as a key theme in disaster governance. Related policy frameworks began integrating measures for vulnerability reduction, risk prevention, and preparedness into sustainable development agendas [9]. Early discourse on city resilience primarily addressed disaster risk management. Facing mounting urban challenges, enhancing city resilience has emerged as a global consensus for achieving sustainable urban development. Various international organizations have launched corresponding initiatives and frameworks to advance this goal. Notable examples include the Rockefeller Foundation’s “100 Resilient Cities” program, the United Nations’ Sustainable Development Goals (SDGs) for 2030, and the “Resilient Cities” agenda adopted at the UN Habitat Conference [10]. Cities worldwide have also taken concrete steps to strengthen their resilience. For instance, the Melbourne Government published the Resilient Melbourne Strategy in June 2016, outlining objectives, key action areas, guiding principles, and specific projects to bolster city resilience [11]. Similarly, Seoul introduced the “One Less Nuclear Power Plant” policy in April 2012, aiming to enhance resilience through participatory governance [12]. Furthermore, following Hurricane Sandy, the New York City Panel on Climate Change formulated a comprehensive adaptation plan to address future climate risks in coastal areas [13].

Cities serve as critical platforms for the convergence of innovative elements. Technological innovation driven by next-generation information technologies provides critical support and safeguards for urban industrial development and government management [14]. However, innovation is related to externality characteristics, which necessitates innovation policies led by governments [15]. For this reason, China has actively carried out urban pilot work. The innovative city focuses on building a high-quality innovation ecosystem and continuously injecting new momentum into urban development, thereby building an environmental foundation to enhance urban innovation abilities. Smart cities utilize digital technology throughout their operation management structures, which strengthens their capacity to respond to threats through advanced technical means to quickly return normal order. The innovative city and smart city are two pilot policies that promote urban development in the new era. These are implemented together, with technological innovation and digital governance at the core. They can not only optimize the external environment for innovation and development but also be embedded in every aspect of innovation activities, which is beneficial for the enhancement of resilience. However, the following questions still need to be addressed: Can the mix of the two pilot policies generate super-additive effects when implemented in cities, and how do they affect city resilience?

Research on the policy impacts of smart cities or innovative cities has focused on four areas: the city, economy, enterprise, environment, and society. For example, Jiang et al. (2025) pointed out that smart cities can enhance the economy and ecology of society in the short term, but there is a negative impact on resources in the early stages of construction [16]. Gong & Shan (2023) took Chinese cities as an example and found that a smart city policy boosts coordinated development in urban and rural areas, and this impact is related to positive spatial spillover effects [17]. Based on data from 297 cities in China, Ji et al. (2024) [18] empirically found that smart city policies can significantly increase tourism revenues and tourist arrivals in pilot cities and that this effect can break down spatial barriers to affect neighboring cities. Cai et al. (2025) [19] analyze American urban data and reveal that smart city initiatives may exert significant impacts on educational attainment, income inequality, and transportation systems. Zhou et al. (2024) investigated the Chinese innovative city and found that the innovation ability of corporations can be enhanced by this policy [20]. Xu et al. (2025) [21] empirically found that the green land use efficiency of pilot cities is greater than in non-pilot cities. In addition, several studies have demonstrated that innovative cities exert varying positive effects on multiple aspects of business development, from export and product quality enhancement [22] to organizational digital transformation [23].

In summary, studies on single smart or innovative city pilots are relatively abundant, but the research on the effect of dual pilots of smart and innovative cities is still relatively scarce. At the same time, studies on the synergistic effect of policy mainly focus on the ecological environment [24], green development [25,26], and so on, and the research on city resilience is relatively insufficient. In view of this, the “dual pilot” city is taken as a quasi-natural experiment, and we use the multi-time point difference-in-differences (DID)method to theoretically analyze the impact, mechanism, and spatial spillover effect of the ”dual pilot” policy on city resilience. This study’s contributions are as follows: Firstly, rather than limiting ourselves to a single pilot study, we focus on exploring the impact and mechanism of the innovative city and smart city combination on city resilience. Secondly, we compare the differences in city resilience between the “dual pilot”, “single pilot”, and “non-dual pilot” cities to explore the net effect of policy combinations on city resilience. Thirdly, benchmark regression is used to deeply explore the mechanism, heterogeneity, and spatial spillover effect of the “dual pilot” city on city resilience, which comprehensively discusses methods to enhance city resilience through the “dual pilot” policy.

2. Policy Context and Theoretical Assumptions

2.1. Policy Context

The smart city pilot policy utilizes emerging technologies to achieve the intelligent management and operation of cities [27]. Initially, to adapt to the problems brought about by urbanization, IBM was the first to put forward smart cities. These arrived relatively late in China, but with the development of the country’s economy, the establishment of smart cities has progressed at a rapid pace [28]. The government has implemented the pilot program in phases beginning in 2010. On 5 December 2012, the government issued the Notice on the National Smart City Pilot Work and other related documents and at the same time announced China’s first national smart city pilot list, involving a total of 90 cities (districts, counties, and towns), marking the official launch of the smart city pilot. By 2022, the number of planned smart cities (districts, counties, and towns) in China exceeded 700. This means that smart cities play a very important role in China’s future urban governance and are becoming increasingly prominent.

Before the implementation of the smart cities, China conducted research on innovative cities. In 2008, Shenzhen was established as the first innovative city in China, indicating the country’s transition from the planning to implementation stage. In 2010, the government issued the Notice on Promoting the Pilot Work of National Innovative Cities, formally establishing the first 16 pilot cities (districts). Since then, the scale of the project has gradually expanded, and from 2012 to 2013, 70 pilot cities (districts) were approved. In 2018 and 2022, the government added an additional 42 pilot cities (districts), and at the 2022 mark, the number of innovative cities (districts) in China reached 103. In general, innovative cities experienced a development process from a partial pilot to comprehensive promotion and then to extensive coverage, which is an important step for China in working towards an innovative country.

This paper identifies cities designated as both smart city and innovative city pilots as “dualpilot” cities, regardless of whether the two pilot policies were established simultaneously. Therefore, the first batch of dual-pilot cities for smart cities and innovative cities began construction in 2012, and by 2022, there were a total of 48 “dual pilot” cities. The specific list is shown in

Table 1. List of “dual pilot” cities and their establishment years.

	Innovative Cities	Smart Cities	Daul Polit Cities
2010	Beijing, Tianjin, Shijiazhuang, Tangshan, Qinhuangdao, Taiyuan, Shanghai, Wuxi, Changzhou, Ningbo, Wuhan, Yantai, Nanning, Chongqing, Baoji, Lanzhou, Yinchuan, Guiyang
2011	Hohhot, Zhenjiang
2012	Nantong, Zhengzhou, Ürümqi	Beijing, Tianjin, Shijiazhuang, Qinhuangdao, Handan, Taiyuan, Changzhi, Shanghai, Wuxi, Changzhou, Zhenjiang, Wenzhou, Jinhua, Wuhu, Tongling, Pingxiang, Dongying, Weihai, Dezhou, Zhengzhou, Wuhan, Zhuzhou, Chongqing, Taizhou	Beijing, Tianjin, Shijiazhuang, Qinhuangdao, Taiyuan, Shanghai, Wuxi, Changzhou, Zhenjiang, Dongying, Zhengzhou, Wuhan, Chongqing
2013	Taizhou, Piangxiang, Nanyang, Yichang, Xiangyang, Zunyi	Yingkou, Nantong, Ningbo, Xinyu, Yantai, Yichang, Xiangyang, Nanning, Liuzhou, Mianyang, Guiyang, Zunyi, Baoji Lanzhou, Yinchuan, Ürümqi	Nantong, Ningbo, Pingxiang, Xiangyang, Yantai, Xiangyang, Nanning, Guiyang, Zunyi, Baoji, Lanzhou, Yinchuan, Ürümqi, Yichang, Taizhou
2014		Tangshan, Hohhot Xuzhou, Chuzhou, Quanzhou, Nanyang, Yuxi, Hanzhong	Tangshan, Hohhot, Nanyang,
2018	Xuzhou, Jinhua, Wuhu, Quanzhou, Dongying, Zhuzhou, Yuxi, Hanzhong		Xuzhou, Quanzhou, Zhuzhou, Yuxi, Hanzhong, Jinhua, Wuhu, Dongying
2022	Handan, Changzhi, Yingkou, Wenzhou, Tongling, Chuzhou, Xinyu, Weihai, Dezhou, Liuzhou, Mianyang		Handan, Changzhi, Wenzhou, Tongling, Chuzhou, Xinyu, Weihai, Dezhou, Liuzhou, Mianyang

.

Overall, unlike the horizontal, networked development approaches seen in Europe or the reliance on international aid and fragmented implementation in many Global South countries, China’s pilot policy model demonstrates distinct vertical characteristics, emphasizing centralized planning and resource allocation by the central government. The Chinese approach illustrates a potential pathway for systematically fostering endogenous local development capabilities through a state-led, comprehensive policy framework. This suggests that the value of China’s experience lies not in the direct replication of its specific institutional arrangements, but rather in its role as a critical comparative case study. It offers insights into how to design adaptable and integrated pilot policy implementation pathways under varying governance capacities and resource constraints.

2.2. The Direct Impact of “Dual Pilot” Cities on City Resilience

Innovative cities and smart cities are indispensable initiatives for enhancing city resilience. For one thing, in order to establish the National Innovation City, pilot cities usually invest resources in policy areas such as industry, R&D, taxation, etc., in order to support innovation activities within the city and accelerate the aggregation and flow of innovation elements. This helps stimulate the development of emerging industries and increase the urban production capacity, which constitutes a significant contribution to the enhancement of city resilience. Additionally, by integrating urban infrastructure and social services via digital technology and intelligent means, smart cities can improve the efficiency and convenience of urban operations while promoting the green and sustainable development of cities. Through intelligent technology, sensors, the Internet of Things, and so on, smart cities also establish a digital learning environment, build intelligent medical facilities, and realize real-time monitoring and intelligent management of infrastructure, which not only improves education, medical care, and other resident issues but also improves the operation efficiency and safety of the city. In addition, smart cities can foster the emergence of innovative industries, generating expanded employment opportunities, which increases residents’ income and government tax revenues and boosting the economic scale of the entire city.

“Dual pilot” cities provide a better practical basis compared to “single pilots”. First of all, “dual pilot” cities have a more superior talent system. Through a combination of policies, such as convenient settlement, preferential housing purchases, and financial support, they can systematically solve the key pain points faced in talent flow, such as household registration and housing and financial constraints, and form a closed loop of whole chain policies, from attracting talent to retaining and cultivating talent, which lays an institutional foundation for the construction of a high-quality talent ecosystem. Secondly, “dual pilot” cities have more flexibility in their fiscal policy. The government can effectively guide the flow of venture and other capital to the innovation field by continuously increasing investments in innovation, so that the city’s innovation potential continues to increase. Finally, “dual pilot” cities can continuously strengthen the agglomeration of innovative elements through the aggregation of policies, so as to further improve city resilience.

H1: 

Innovative cities and smart cities can synergistically enhance city resilience.

2.3. The Indirect Impact of “Dual Pilot” Construction on City Resilience

City resilience can be boosted by “dual pilots” that promote innovation and development levels. From the perspective of the macro-environment, “dual pilot” cities optimize the innovation environment via both software and hardware. In terms of software, innovative cities stimulate the innovation vitality of the people through institutional reforms and cultural innovations, which effectively improves the innovation level of the cities [29]. In terms of hardware, smart cities widely use emerging digital technologies, such as smart medical care and smart transportation, to achieve communication and cooperation between urban information departments, providing broad application prospects for innovation [28]. With respect to micro-subjects, “dual pilot” cities are able to raise the urban innovation level by stimulating the development of high-tech industries. First, innovative cities adjust their policies and resources in favor of enterprise innovation, which leads to more talent, capital, and technology entering these enterprises. As a result, the level of innovation is enhanced [30]. Second, smart cities can optimize the external conditions of enterprise operations by improving the information infrastructure to create a good atmosphere for technological innovation, thereby promoting the development of high-tech industries. Many scholars have examined the relationship between innovation and city resilience: Zameer et al. [31] argue that an increase in innovation factors can drive technological progress and reduce industrial lock-in by breaking traditional energy constraints, thereby enhancing productive capacities and urban economies. Bristow & Healy [32] found that cities with higher innovation levels demonstrate a stronger resistance and resilience to crises.

City resilience also can be boosted by “dual pilots” through industrial structure upgrading. On the one hand, innovative cities can optimize the allocation efficiency of innovation factors and promote the transformation and upgrading of related industries, thereby promoting a higher level of industrial structures and value chains [29], which allows the economy to actively respond to external shocks and quickly recover from them. On the other hand, smart cities can lead to the emergence of productive service industries, such as data services and software development, which brings about new economic growth points, thereby enhancing the competitiveness of urban economies and laying the material foundation for greater city resilience. Smart cities and innovative cities complement each other in terms of environments and materials, jointly promoting industrial structure upgrading and enhancing city resilience.

H2: 

The innovation and development level and industrial structure upgrading are two important paths for “dual pilot” cities to enhance city resilience.

2.4. The Spatial Spillover Effect of “Double Pilot” Construction on City Resilience

With the support of “dual pilot” cities, the cross-regional flow of innovative elements not only has an impact on the industrial structure and talent aggregation in the region but also has a variety of impacts on “non-pilot” cities through competition effects, knowledge spillovers, and regional linkages. For one thing, smart cities promote the deep integration of the digital city and physical space by breaking through the limitations of traditional administrative divisions and relying on technical means to build a resilient smart service ecosystem. Under this framework, the cities’ resources can be efficiently circulated, and an intelligent network resource sharing platform can be effectively built, thereby giving birth to a new cross-regional collaborative governance and cooperation model based on the dynamic flow space. Additionally, innovative city pilots promote the optimal allocation of resources, such as high-skilled labor and high-quality capital, which leads to the efficient flow of innovative factors among pilot cities and provides scientific, technological, and innovative support for enhancing the resilience of neighboring cities. In addition, in order to seize the innovation high ground and establish policy support, “non-dual pilot” cities may absorb overflow resources from pilot cities, such as knowledge, technical experience, talent factors, and innovative products, enabling “non-dual pilot” cities to drive innovation through the “tournament-type competition” development trend, enhancing the resilience of “non-dual pilot“ cities.

H3: 

“Dual pilot” cities demonstrate significant spatial spillover effects that enhance regional city resilience.

3. Research Design

3.1. Methods

Difference-in-Differences (DID) Model

The DID model is widely used in many fields, such as macroeconomic policy and urban development [33,34,35]. Based on policy exogeneity characteristics, this method can not only effectively overcome endogenous bias and avoid the interference of reverse causation but can also further control the influence of missing variables by introducing a fixed effect model, thus substantially enhancing the accuracy of the estimation outcomes [36]. Therefore, we utilize the DID model to examine, both theoretically and empirically, how the dual pilot of an innovative city and smart city affects city resilience, as shown in Equation (1):

$$U{R}_{i,t}={\mathrm{\alpha }}_0+{\mathrm{\alpha }}_{1}Polic{y}_{i,t}+{\displaystyle \sum {\mathrm{\alpha }}_{\mathrm{k}}Control{s}_{i,t}+{\lambda }_i+{\mu }_t+{\epsilon }_{i,t}},$$

(1)

where $i$ and $t$ denote the city and year, respectively; $U{R}_{i,t}$ denotes city resilience; $Polic{y}_{i,t}$ is a core explanatory variable—“dual pilot” cities; $Control{s}_{i,t}$ are control variables; ${\mathrm{\alpha }}_1$ demonstrates the marginal contribution of “dual pilot” cities to city resilience; ${\mathrm{\alpha }}_0$ is a constant term; ${\mathrm{\alpha }}_{\mathrm{k}}$ is defined as a control variable; ${\lambda }_i$ and ${\mu }_t$ denote individual and time fixed effects, respectively; and ${\epsilon }_{i,t}$ is defined as a random perturbation term.

According to the previous transmission mechanism, “dual pilot” cities affect city resilience by enhancing the innovation and development level and improving industrial structure upgrading, as shown in Equation (2):

$$M_{i,t}={\mathrm{\beta }}_0+{\mathrm{\beta }}_{1}Smartinnovatio{n}_{i,t}+{\displaystyle \sum {\mathrm{\beta }}_{\mathrm{k}}Control{s}_{i,t}+{\lambda }_i+{\mu }_t+{\epsilon }_{i,t}},$$

(2)

where $M_{i,t}$ is defined as the mechanism variable, including the innovation and development level and industrial structure upgrading; ${\mathrm{\beta }}_0$ represents the constant term; ${\mathrm{\beta }}_1$ represents the influence of “dual pilot” cities on mediating variables; ${\mathrm{\beta }}_{\mathrm{k}}$ represents the coefficient of each control variable; and other variables are consistent with Equation (1).

Finally, to investigate the spatial spillover effect in “dual pilot” cities, referring to Zhang et al. [37], we introduce the spatial regression term into the DID model. This model not only considers the spatial lag characteristics of explanatory variables but also solves the problem of the spatial error correlation, fully capturing the explanatory variables and their comprehensive impact on the spatial spillover of dependent variables. The model is shown in Equation (3):

$$\begin{array}{l}U{R}_{i,t}=\mathrm{\rho }{\displaystyle \sum _{j=1}^n{w}_{ij}U{R}_{i,t}+}{\mathrm{\delta }}_{1}Smartinnovatio{n}_{i,t}+\mathrm{\theta }{\displaystyle \sum _{j=1}^n{w}_{ij}Smartinnovatio{n}_{i,t}}\\ +{\mathrm{\delta }}_{\mathrm{k}}{\displaystyle \sum Control{s}_{i,t}+{\lambda }_i+{\mu }_t+{\epsilon }_{i,t}}\end{array},$$

(3)

where $\mathrm{\rho }$ is the spatial lag coefficient of the dependent variable, indicating the impact of changes in city resilience in neighboring regions on the city resilience of the region; $\mathrm{\theta }$ is the spillover effect of the “dual pilot” cities; $w_{ij}$ is the weighting matrix, including the geographic and the economic distance matrix; ${\mathrm{\delta }}_1$ indicates that m is the coefficient of the direct impact of the “dual pilot” cities on city resilience; and ${\mathrm{\delta }}_{\mathrm{k}}$ is the estimator of the control variable. The other symbols are the same as in Equation (1).

3.2. Variable Description

3.2.1. Dependent Variable: City Resilience

At present, there is no universally accepted standard for city resilience assessment frameworks, though they generally begin with the characteristics or composition of city resilience. The former is based on the framework of resistance, adaptability, and resilience [38]. The latter argues that city resilience includes four components, including the economy, ecology, society, and infrastructure [39]. Considering that the three specific resilience elements of the former are reflected in different indicators, this makes the framework prone to confusion in the process of variable selection. So, we take the latter as the framework and refer to related studies [40] to construct a city resilience indicator system, as shown in

Table 2. Indicator system for evaluating city resilience.

Primary Indicators	Secondary Indicators (Unit)	Attributes	Weight
Economic resilience	Total retail sales of consumer goods (104 CNY) [42]	+	0.1021
	The ratio of tertiary industry value to GDP (%) [42]	+	0.0086
	GDP per capita (CNY) [42]	+	0.0304
	Local general public budget expenditure (104 CNY) [43]	+	0.0857
	Year-end deposit and loan balances across financial institutions (109 CNY) [44]	+	0.1656
Social resilience	Count of students enrolled in general universities per 10,000 population (104 people) [42]	+	0.0860
	Total collection of books in public libraries (pieces) [45]	+	0.1410
	Number of hospital beds per 10,000 population (pieces/people) [45]	+	0.0166
	Population density (people/km2) [43]	−	0.0008
	Average wage of employed workers (CNY) [44]	+	0.0167
Ecological resilience	Greening coverage in built-up areas (%) [42]	+	0.0050
	Rate of wastewater undergoing centralized treatment in sewage treatment plants (%) [43]	+	0.0037
	Percentage of household waste undergoing non-toxic processing (%) [43]	+	0.0039
	Industrial sulfur dioxide emissions (104 tons) [43]	−	0.0013
Infrastructure resilience	Road area per capita (km2/person) [45]	+	0.0288
	Number of public automobiles (trams) per 10,000 people (vehicles/104 people) [42]	+	0.0399
	Road freight volume (104 tons) [46]	+	0.0659
	Number of Internet broadband user accesses (104 households) [42]	+	0.0848
	Count of end-of-year mobile phone users (104 households) [46]	+	0.0638
	Drainage pipe density in built-up area (km/km2) [45]	+	0.0495

. The entropy weight-TOPSIS method effectively combines its strengths with those of the TOPSIS approach, which not only reflects the information value of each indicator objectively and comprehensively but also helps to avoid subjective interference [41]. Therefore, in this study we measure the city resilience by utilizing the entropy weight-TOPSIS method. The specific formula is as follows:

Step 1: Perform normalization processing. Standardize the matrix formed by the original cities resilience evaluation indicators. The standardization formula is as follows:

$$y_{ij}=\left\{\begin{array}{c}{\displaystyle \frac{x_{ij}-\min \left(x_{ij}\right)}{\max \left(x_{ij}\right)-\min \left(x_{ij}\right)}},x_i is a positive indicator\\ {\displaystyle \frac{\max \left(x_{ij}\right)-x_{ij}}{\max \left(x_{ij}\right)-\min \left(x_{ij}\right)}}, x_i is a \mathrm{negative} indicator\end{array}\right..$$

(4)

Here, $x_{ij}$ denotes the value of the i-th indicator in the j-th region, while $\max \left(x_{ij}\right)$ and $\min \left(x_{ij}\right)$ represents the maximum and minimum values of indicator xi, respectively.

Step 2: Determine the weight of each indicator using the entropy method. The contribution rate of each indicator is

$$p_{ij}={\displaystyle \frac{y_{ij}}{{\displaystyle \sum _i{y}_{ij}}}}.$$

(5)

Calculate the entropy value of the i-th indicator based on the definition of information entropy:

$$e_i={\displaystyle \frac{{\displaystyle \sum _j{p}_{ij}\ln \left(p_{ij}\right)}}{\ln \left(n\right)}}.$$

(6)

Here, $n$ represents the total number of regions. When $p_{ij}=0$ for a given region, it is assumed that $p_{ij}\ln \left(p_{ij}\right)=0$. The weight for the i-th indicator is calculated using the following formula:

$$w_i={\displaystyle \frac{\left(1-e_i\right)}{{\displaystyle \sum _i\left(1-e_i\right)}}}.$$

(7)

Here, $1-e_i$ denotes the utility value of the i-th indicator.

Step 3: Construct the evaluation matrix V based on entropy weights. The specific expression for V is as follows:

$$V=\left[\begin{array}{cccc}{w}_1{y}_{11}& w_1{y}_{12}& \cdots & w_1{y}_{1n}\\ w_2{y}_{21}& w_2{y}_{22}& \cdots & w_2{y}_{2n}\\ ⋮& ⋮& ⋮& ⋮\\ w_m{y}_{m1}& w_m{y}_{m2}& \cdots & w_m{y}_{mn}\end{array}\right]=\left[\begin{array}{cccc}{v}_{11}& v_{12}& \cdots & v_{1n}\\ v_{21}& v_{22}& \cdots & v_{2n}\\ ⋮& ⋮& ⋮& ⋮\\ v_{m1}& v_{m2}& \cdots & v_{mn}\end{array}\right].$$

(8)

Here, m is the total number of indicators.

Step 4: Calculate the positive and negative ideal solutions. The positive ideal solution refers to the maximum value of the i-th indicator in the j-th region. The negative ideal solution refers to the minimum value of the i-th indicator in the j-th regions. The formulas are as follows:

$$V^{+}=\left\{\max v_{ij} | i=1,2,\cdots ,m\right\}=\left\{v_1^{+},v_2^{+},\cdots ,v_m^{+}\right\},$$

(9)

$$V^{-}=\left\{\min v_{ij} | i=1,2,\cdots ,m\right\}=\left\{v_1^{-},v_2^{-},\cdots ,v_m^{+}\right\}.$$

(10)

Step 5: Calculate the distances from each indicator to the positive and negative ideal solutions based on Euclidean distance:

$$D_j^{+}=\sqrt{{\displaystyle \sum _{i=1}^m{\left(v_i^{+}-v_{ij}\right)}^2}},$$

(11)

$$D_j^{-}=\sqrt{{\displaystyle \sum _{i=1}^m{\left(v_i^{-}-v_{ij}\right)}^2}}.$$

(12)

Step 6: Calculate the city resilience level for each region. Let $T_j$ j represent the cities resilience index of the j-th region. A higher value of $T_j$ indicates a greater level of cities resilience, while a lower value corresponds to a weaker level. The calculation formula is as follows:

$$T_j={\displaystyle \frac{D_j^{-}}{D_j^{+}+D_j^{-}}}.$$

(13)

3.2.2. Core Explanatory Variable

Policy is defined as the core explanatory variable—that is the “dual pilot” city combining a smart city and innovative city—which is obtained by multiplying ${\mathrm{treat}}_{\mathrm{i}}$ by ${\mathrm{post}}_{\mathrm{it}}$. If city i is a “dual pilot” city, this city is the sample of the treatment group and is assigned a value of 1; otherwise, it is regarded as the sample of the control group and is assigned a value of 0. On this basis, the ${\mathrm{post}}_{\mathrm{it}}$ of the treatment group is assigned to 1 from the first year of the implementation of the “dual pilot”, and the remaining years are assigned to 0. Since the “dual pilot” cities were first established in 2012, the values before 2012 are established as 0, and the value after 2012 is 1. A total of 48 “dual pilot” cities are selected.

3.2.3. Control Variables

Referencing the works of Zheng et al. [47], Chang et al. [48], and Wang et al. [49], this study selects the following control variables: Degree of Government Intervention (Gov): Excessive government intervention may crowd out market-driven resource allocation, dampen economic vitality, which potentially undermines city resilience. Consumption Level (Con): The consumption level reflects the momentum of domestic demand, and a robust domestic demand can promote economic circulation and stability, potentially exerting a positive influence on city resilience. Industrial Structure Sophistication (Ind): Industrial structure upgrading enhances a city’s risk resilience and adaptability. Urban Education Level (Edu): The concentration of high-quality educational resources enhances innovation capability, thereby strengthening city resilience. Financial Development Scale (Finc): The expansion of financial scale may promote risk dispersion and the supply of innovative capital, thereby enhancing city resilience. Human Capital Level (Hd): High-quality labor may concentrate in policy-favored regions, leading to imbalances in resource allocation structures, exacerbating resource constraints, and thereby hindering improvements in city resilience. The variable descriptions are shown in

Table 3. Explanation of control variables.

Variable Name	Variable Symbol	Variable Definition	Excepted Sign
Degree of Government Intervention	Gov	Ratio of local government fiscal expenditure to regional GDP	Negative
Consumption Level	Con	Logarithm of retail sales of consumer goods	Positive
Industrial Structure Sophistication	Ind	Ratio of tertiary industry value-added to secondary industry value-added	Positive
Urban Education Level	Edu	Logarithm of full-time faculty members in regular higher education institutions	Positive
Financial Development Scale	Finc	Ratio of year-end deposits and loans of financial institutions to regional GDP	Positive
Human Capital Level	Hd	Ratio of enrolled students in regular higher education institutions to regional GDP (person/CNY)	Negative

as follows, and specific details are presented in

Table 4. Descriptive statistics.

Symbol	Obs	Mean	Std.dev	Min	Max
UR	3990	0.0674	0.0674	0.0134	0.6623
Policy	3990	0.0877	0.2829	0	1
Gov	3990	0.1979	0.1038	0.0439	1.4852
Edu	3990	7.6720	1.3097	4.2485	11.2900
Con	3990	15.5427	1.0784	12.1556	19.0140
Ind	3990	1.0408	0.5935	0.1087	5.7848
Finc	3990	2.4688	1.2432	0.5879	21.3018
Hd	3990	35.1196	32.3939	0.3300	282.7800
Inn	3990	0.1705	0.3569	0.0006	5.0842
Struc	3990	0.7905	0.4248	0.1082	5.1996

.

3.2.4. Mechanism Variables

In this paper, the innovation and development level (Inn) and industrial structure upgrading (Struc) are taken as mechanism variables. In reference to related studies [29,48] the innovation and development level is measured by the count of patents granted per million people per capita, and industrial structure upgrading is expressed by the ratio of the tertiary industry’s added value to the combined value added of primary and secondary industries.

3.3. Data Sources

This paper adopts 285 prefecture-level cities in China as the research object, including 48 “dual pilot” cities and 237 “non-dual pilot” cities. In order to balance the panel as much as possible, the data and the newly established prefecture-level cities are converted from counties into cities. Considering the impact of the financial crisis in 2008, whether the sample fully covers the key phases of initiation, expansion, and deepening of both the smart cities and innovative cities and data availability, this paper uses data from 2009 to 2022. The data sources are as follows: the China City Statistical Yearbook, the China Regional Economic Statistical Yearbook, statistical yearbooks of various provinces and cities, and the Chinese Innovation Research Database (CIRD). The missing data is filled in using interpolation methods.

4. Results and Analysis

4.1. Baseline Regression

The benchmark model uses cluster-robust standard errors, while fixing the city and time effects. The estimators used to add the double fixed effect of the city and time without control variables are shown in column (1) of

Table 5. Benchmark regression results.

Variables	(1)	(2)	(3)	(4)	(5)
	L0.	L0.	L1.	L2.	L3.
Policy	0.0181 *** (3.7073)	0.0178 *** (3.7180)	0.0188 *** (3.7753)	0.0168 ** (3.2322)	0.0154 ** (3.3986)
_cons	0.0658 *** (153.5928)	−0.0866 ** (−2.1300)	−0.0545 (−1.5204)	−0.0156 (−0.4622)	0.0326 (1.0334)
Controls	N	Y	Y	Y	Y
Year and City	Y	Y	Y	Y	Y
Obs	3990	3990	3705	3420	3135
Adj. R2	0.9453	0.9466	0.9536	0.9579	0.9616

Note: t-statistics in parentheses. *** p < 0.01, ** p < 0.05. And, N denotes no, and Y denotes yes.

. Columns (2) to (5) present the results for variables with a 0 to 3 period lag, respectively, after fixing the city and time effects and including control variables. The results demonstrate that the estimators for explanatory variables are positive at the 1% significance level across all model specifications. Furthermore, the regression coefficients with 1-period to 3-period lags indicate that the “dual pilot” cities also effectively boost city resilience. So, Hypothesis H1 is supported.

4.2. Robustness Test

4.2.1. Parallel Trend Test

Before the DID model is utilized, the parallel trends test should be passed. This test is used to verify whether the treatment group and control group show similar time trends before the establishment of the “dual pilot” cities. As shown in Figure 1 and

Table 6. Parallel trends test regression results.

Variable Symbol	Year	Estimated Coefficient	t-Value
Pre_3	2009	−0.0052	−2.14
Pre_2	2010	−0.0016	−1.00
current	2012	0.0036 **	2.24
Post_1	2013	0.0063 ***	3.82
Post_2	2014	0.0110 ***	3.52
Post_3	2015	0.0125 ***	3.46
Post_4	2016	0.0150 ***	3.50
Post_5	2017	0.0176 ***	3.58
Post_6	2018	0.0200 ***	3.44
Post_7	2019	0.0225 ***	3.28
Post_8	2020	0.0172 **	2.48
Post_9	2021	0.0269 ***	3.71
Post_10	2022	0.0201 ***	3.81
Controls	Y
Year and City	Y
Obs	3990
Adj. R2	0.9479

Note: Since STATA automatically omits the pre_1 to avoid multicollinearity, the coefficient for this period is not reported in the table. *** p < 0.01, ** p < 0.05. And, Y denotes yes.

, 2012 is the year when the policy was established. Before 2012, the coefficient was not significant; however, after this time, the estimators became significantly positive and gradually increased, which satisfies the parallel trend assumption.

4.2.2. City Placebo Test

To avoid false regressions caused by the possible omission of unobservable variables, we followed the protocol established by Chang et al. [48]: randomly generate 1000 random “dual pilot” construction shocks to the city, select 48 treatment groups as pseudo-policy dummy variables each time, and then obtain the kernel density distribution of the coefficient estimates, as depicted in Figure 2. In the figure, the solid line represents the estimators after randomization, the scatter represents the corresponding p-value, and the horizontal dotted line is the reference line of p = 0.1. The results indicate that the majority of parameter estimates and their corresponding p-values approach zero, following a normal distribution. The majority of scatter points are distributed above the horizontal reference line, indicating that most of the coefficients are not significant. The placebo test was passed.

4.2.3. Time Placebo Test

In order to control for temporal effects that might affect the differences between the treatment and control groups, this paper assumes that the policy is introduced 1 to 4 years in advance. The estimators presented in

Table 7. Robustness test.

Variables	Time Placebo Test				PSM-DID (Kernel Matching)	Tailing Processing		Replace the Explained Variable
	One Phase	Two Phase	Three Phase	Four Phase		1~99%	5~95%
Policy	0.0143 *** (3.6218)	0.0096 *** (3.0496)	0.0012 (0.9138)	−0.0019 (−0.5396)	0.0159 *** (3.5209)	0.0126 *** (3.8634)	0.0102 *** (4.5481)	0.1159 ** (2.0230)
_cons	−0.0821 ** (−1.9947)	−0.0799 * (−1.9163)	−0.0797 * (−1.8970)	−0.0796 * (−1.8931)	−0.0573 (−1.3151)	−0.0668 ** (−2.5261)	−0.0675 * (−1.9476)	−4.6667 *** (−6.8157)
Controls	Y	Y	Y	Y	Y	Y	Y	Y
Year and City	Y	Y	Y	Y	Y	Y	Y	Y
Obs	3990	3990	3990	3990	3630	3990	3990	3990
Adj. R2	0.9358	0.9451	0.9447	0.9447	0.9483	0.9591	0.9528	0.9543

Note: t-statistics in parentheses. *** p < 0.01, ** p < 0.05, and * p < 0.1. And, Y denotes yes.

on the core explanatory variables in advance to periods 1 to 2 are positive at the 1% significance level. This may be because before a “dual pilot” policy is introduced, cities will prepare in advance and promote the construction of new infrastructure to ensure that they can be selected as “dual pilot” cities, and this process usually takes one to two years [48]. Therefore, the regression coefficients remain significant in advance for periods 1 to 2, but the results are no longer significant in advance for periods 3 to 4, implying that the differences in the time trend are not significant between both groups for the non-declaration preparation period. These estimates are robust.

4.2.4. Staggered DID

When staggered treatment timing coincides with heterogeneous treatment effects, traditional Two-Way Fixed Effects (TWFE) estimators may introduce confounding bias. To ascertain the applicability of TWFE estimators, this paper diagnoses and addresses the issue of heterogeneous treatment effects. Following the estimator selection guidance proposed by Liu et al. (2022) [50], this study adopts the imputation-based counterfactual approach introduced by Borusyak et al. (2024) [51] to address the estimation bias in TWFE models. The corresponding results from the Borusyak estimator indicate a heteroscedasticity-robust estimate of 0.0265, which shows no substantial deviation from the baseline regression and is statistically significant at the 5% level. The corresponding dynamic estimation results are presented in Figure 3. It is evident that prior to the implementation of the policy, there was almost no discernible treatment effect. However, as the policy was rolled out, the “dual pilot” initiative progressively contributed to enhancing city resilience. This reinforces confidence in the validity of the parallel trends assumption and indicates that heterogeneous treatment effects do not undermine the conclusions drawn from the baseline regression analysis in this study.

4.2.5. PSM-DID

This study further tests whether the model has a selective bias problem by using the PSM-DID method. The control variables in the benchmark model are used as covariates, and the logit model is used for kernel matching between the treatment and the control groups, with a bandwidth set at 0.06. After sample matching, it is necessary to test the balance of the matched samples. As shown in

Table 8. Balance test before and after propensity score matching.

Variables	Unmatched or Matched	Mean		Standard Deviation		t-Test
		Experimental Group	Control Group	Percentage	Reduce the Scale	t-Value	p-Value
gov	U	0.1523	0.2023	−60.0		−8.67	0.000
	M	0.1507	0.1434	8.7	85.6	1.86	0.064
ln_con	U	16.7680	15.4250	140.7		23.78	0.000
	M	16.708	16.6380	7.3	94.8	1.10	0.273
ind	U	1.2988	1.0160	42.9		8.59	0.000
	M	1.2240	1.1668	8.7	79.8	1.31	0.190
finc	U	3.1024	2.4079	51.5		10.11	0.000
	M	3.0029	2.8473	11.5	77.6	1.55	0.121
hd	U	47.8220	33.8980	37.6		7.74	0.000
	M	48.7680	48.6680	0.3	99.3	0.10	0.975
Ln_edu	U	9.1431	7.5303	137.7		23.47	0.000
	M	9.0879	8.9906	8.3	94.0	1.11	0.266

, the standardized differences between the treatment and control groups decrease substantially after matching. For most variables, the absolute value of the standardized bias after matching is less than 10%, and the p-values are greater than 0.1. This indicates that the selection of treatment variables and the matching method are unbiased, and the matching performance is satisfactory. To more intuitively and concretely demonstrate the matching effectiveness, the propensity score matching effect graph and the common support test graph, were further plotted. As can be seen from Figure 4, the absolute values of the standardized differences for all covariates are significantly reduced after matching, indicating good matching results. Figure 4 shows that there is sufficient overlap in the propensity score distributions across groups, with no complete separation, gaps, or unilateral absence observed among the groups, satisfying the common support assumption. As revealed in Table 7, the estimated value after matching is similar to the benchmark regression value. Therefore, the results are reliable.

4.2.6. Tailing Processing

Because the economic development level is different between cities, potential outliers may exist in the dataset. To prevent the bias caused by outliers, this paper shrinks the data by 1% and 5%, respectively. As shown in Table 7, the results are still positive at 1% and 5% significance, respectively. Therefore, the tail test is passed.

4.2.7. The Replacement of the Explained Variable

To minimize methodological bias in the results, this study re-evaluates city resilience using a principal component analysis. The outputs illustrate that the estimators of the explanatory variables are still significantly positive under the other measurement method, indicating that the findings are solid.

4.2.8. Eliminate Other Policy Interferences

Considering the long time span from 2009 to 2022, there may be other policies that have an impact on city resilience during this period, so it is essential to eliminate the interference caused by other policies [52]. As such, it is vital to consider the possible impact of “Broadband China” on city resilience. In addition, the construction of climate-resilient cities can effectively reduce risks posed by climate change, protect people’s lives and property, and boost the disaster warning level in cities, which also has a certain influence on city resilience. Finally, given that innovation factors significantly influence city resilience [14], and both national independent innovation demonstration zones and innovative cities are important strategic measures for building an innovative country in a holistic way, their influence on city resilience must be carefully examined.

After adding the three policies to the benchmark model, the estimated value is still significantly positive, as presented in

Table 9. Eliminate relevant policy interference.

Variables	(1)	(2)	(3)	(4)
	Broadband China	Climate-Resilient Cities	National Independent Innovation Demonstration Zones	All
Policy	0.0153 *** (3.3277)	0.0178 *** (3.6871)	0.0158 *** (3.3567)	0.0141 *** (3.1074)
Broadband China	0.0115 *** (4.0959)			0.0086 *** (3.7383)
Climate-resilient cities		0.0006 (0.1002)		0.0001 (0.0230)
National independent innovation demonstration zones			0.0225 *** (4.5385)	0.0206 *** (4.4142)
Controls	Y	Y	Y	Y
Year and City	Y	Y	Y	Y
Obs	3990	3990	3990	3990
Adj. R2	0.9453	0.9466	0.8740	0.9558

Note: t-statistics in parentheses. *** p < 0.01. And, Y denotes yes.

. Furthermore, by adding the three policies to the benchmark model at the same time, the coefficients are still significantly positive. Therefore, the results are robust.

4.2.9. Endogenous Test: Instrumental Variable Method

An instrumental variable estimation is employed to address endogeneity concerns. First, the lag one phase (IV1) of the explanatory variable is used as the instrumental variable. Secondly, according to Wang et al. [29], the national historical and cultural city (IV2) is used as an instrumental variable. On the one hand, the purpose of establishing smart cities and innovative cities is to optimize urban development and promote innovation, so as to form a new urban economic growth model and ultimately achieve high-quality economic development; therefore, most of the selected cities demonstrate better modern economic development, and most of the national historical and cultural cities are ancient political and economic centers. There is a certain correlation between the two. On the other hand, national historical and cultural cities do not directly affect city resilience. So IV2 satisfies the exogenous condition.

The first stage estimates in

Table 10. Endogenous test: instrumental variable method.

Variables	IV1		IV2
	First	Second	First	Second
Policy		0.0266 *** (3.83)		0.0300 *** (3.72)
IV1	0.7447 *** (57.43)
IV2			0.9491 *** (75.66)
Controls	Y	Y	Y	Y
Year and City	Y	Y	Y	Y
Kleibergen–Paap rkLMstatistic		44.459 ***		27.453 ***
Cragg–Donald Wald Fstatistic	4382.47	4462.42
Kleibergen–Paap rk Wald Fstatistic	3298.47	5724.39
Obs	3705	3990
R2		0.4146		0.4437

Note: t-statistics in parentheses. *** p < 0.01. And, Y denotes yes.

reveal that all instrumental variables yield positive coefficients with 1% statistical significance, and both F statistics exceed 10, indicating that the “dual pilot” cities are related to the national historical and cultural city. In the second stage of regression, the estimator remains positive at 1% significance, which means that after controlling for endogeneity, the “dual pilot” cities maintain a significant positive impact on city resilience. As such, the conclusion is unchanged.

4.3. Further Discussion

We further examine the synergistic effects between smart cities and innovative cities. First, we analyze the influence of the “single pilot” cities on city resilience. Referring to Zheng et al. [47], on the basis of the full sample, the “single pilot” sample and the “dual pilot” sample of the innovative city are eliminated, and only the smart city “single pilot” sample and the “non-pilot” sample are retained. Then the impact of the smart city “single pilot” on city resilience is obtained. At this time, the coefficient estimators indicate the net effect of the smart city “single pilot” on city resilience. Similarly, the net effect of “single pilot” of the innovative city on city resilience can be obtained. Column (2) in

Table 11. The synergistic effect of the dual pilot driving city resilience improvements.

Variables	Smart City		Innovative City		Total Effect
	(1)	(2)	(3)	(4)	(5)
Only	0.0016 (1.4448)		0.0108 *** (4.4286)
Dual		0.0187 *** (4.7876)		0.0029 (0.6178)	0.0123 *** (2.6334)
Sample	Single polit	Dual polit	Single polit	Dual polit	-
Controls	Y	Y	Y	Y	Y
Year and City	Y	Y	Y	Y	Y
95% CI	[−0.0006, 0.0038]	[0.0109, 0.0264]	[0.0059, 0.0156]	[−0.0065, 0.0123]	[0.0031, 0.0215]
Obs	2590	1456	2562	1386	2184
Adj. R2	0.8632	0.9559	0.9372	0.9480	0.9476

Note: t-statistics in parentheses. *** p < 0.01. And, Y denotes yes.

indicates that the innovative cities effectively boost city resilience. However, the results in column (1) demonstrate that smart cities have a non-significant effect on city resilience. The reason may lie in the fact that smart cities primarily focus on the development of digital government, intelligent transportation, IoT infrastructure, and other related areas. The enhancement of city resilience resulting from these efforts is likely long-term, requiring time to materialize and systemic integration. Within the observation period of this study, such effects may not yet be fully evident. In contrast, innovative cities centers on enterprise R&D, high-tech industries, and similar domains. It may contribute to city resilience more directly and rapidly, for instance, industrial upgrading that enhances economic stability and the clustering of high-skilled talent.

Second, we explore whether “dual pilot” cities are more likely to promote resilient urban growth than “single pilot” cities. Referring to Su et al. [53], all “non-pilot” samples are excluded from the full sample, and only the “dual pilot” samples and the samples that have become “single pilots” are retained. In this case, the coefficient estimators represent the net effect of the “single pilot” becoming a “dual pilot” on city resilience. As shown in column (5) of Table 11, “dual pilot” cities exert a stronger impact on city resilience compared to the “single pilot” policy. This may be because the two policies have formed a virtuous cycle: smart cities provide application scenarios and infrastructure for innovative cities, while innovative cities concentrate technological resources and talent that in turn feed back into the development of smart cities. Meanwhile, the dual-pilot initiative brings concentrated investment of fiscal and human resources, compelling an upgrade of the urban governance system, thereby generating systemic gains that exceed the simple sum of the individual policies.

Furthermore, we explore the differences in the impact on city resilience between “dual pilots” and “single pilots” for smart cities or innovative cities. Referring to the approach of Su et al. [53], we exclude all “non-pilot” and “single pilot” samples of innovative cities from the full sample and retain only smart city “single pilot” and “dual pilot” samples to make comparisons between the two. For the innovative city we utilized the approach. As shown in column (2) of Table 11, when the smart city “single sample” is used as the control group, the “dual pilot “can better enhance city resilience. However, column (4) shows that the estimators of the “dual pilot” cities are not significant for the sample that has become an innovative city. The reason may be that smart cities, as a digital infrastructure and governance platform, can provide strong empowerment for the construction of innovative cities, thereby significantly enhancing city resilience. However, due to the complex systemic nature of smart cities, their short-term marginal gains are limited and difficult to manifest quickly. Therefore, superimposing smart cities on top of existing innovative city frameworks does not lead to a significant additional improvement in city resilience.

Finally, this study explores whether the order of the establishment of pilot policies has different influences on city resilience. According to the practice of Su et al. [53], we retained the smart city “single pilot” and “dual pilot” samples that established smart cities first. The estimated coefficient reflects the net effect of the impact of establishing smart cities before innovative cities on city resilience. The net effect of the impact of establishing innovative cities before smart cities can be obtained in the same way. As can be seen from

Table 12. Comparison of the impact of pilot sequencing on city resilience.

Variables	Smart Cities Before Innovative Cities	Innovative Cities Before Smart Cities
Policy	0.0068 * (1.8257)	0.0122 (1.5860)
Controls	Y	Y
Year and City	Y	Y
95% CI	[−0.0006, 0.0141]	[−0.0031, 0.0274]
Obs	1078	1036
Adj. R2	0.9036	0.9500

Note: t-statistics in parentheses. * p < 0.1. And, Y denotes yes.

, the net effect of policy shocks is significantly positive for establishing smart cities before innovative cities, but the effect of establishing innovative cities before smart cities is not significant. This demonstrates that the order of the policy implementation has a certain impact on the construction effect of “dual pilot” cities, and only the establishment of smart cities before innovative cities can significantly enhance city resilience. The possible reason for this is that an innovative city pilot policy focuses on macro-strategic innovations for concepts, models, and management. However, a smart city pilot policy mainly utilizes information technology and data analysis to improve urban management efficiency, the quality of public services, and the residents’ standard of living, with an emphasis on intelligent and digitalized urban infrastructure and services, which lays the material foundation for innovative cities. As innovative cities focus on creating strategic environments and a smart city pilot policy takes time to improve infrastructure, their effects are not realized in time, which makes the policy implementation sequence of establishing innovative cities before smart cities ineffective at improving city resilience in the short term.

5. Impact Mechanism and Heterogeneity Test

5.1. Impact Mechanism Analysis

As shown in column (1) of

Table 13. The impact mechanism analysis.

Variables	(1)	(2)	(3)
	Ur	Inn	Struc
Policy	0.0178 *** (3.7180)	0.0696 ** (2.4315)	0.0495 ** (2.5317)
Controls	Y	Y	Y
Year and City	Y	Y	Y
Obs	3990	3990	3990
Adj. R2	0.9466	0.8218	0.9628

Note: t-statistics in parentheses. *** p < 0.01, ** p < 0.05. And, Y denotes yes.

, “dual pilot” cities can significantly enhance city resilience, but through what path does this occur? According to Equation (2), we conduct the mechanism test shown in Table 13 to examine the impact of “dual pilot” cities on the mechanism variables, and the impact of the mechanism variables on the dependent variables is mainly derived from the logical analysis in the previous section.

As shown in Table 13, the regression result in column (2) shows that the impact coefficient of “dual pilot” cities on the level of innovation and development is 0.0696 at 5% significance, demonstrating that the innovation and development level plays an intermediary role in the impact effect of “dual pilot” construction on the resilience of the city. Column (3) demonstrates that “dual pilot” cities significantly promote industrial structure upgrading, suggesting this upgrading serves as an important mediating channel for enhancing city resilience.

In summary, hypothesis H2 is supported.

5.2. Heterogeneity Test

5.2.1. Urban Grade Heterogeneity

Generally speaking, municipalities, provincial capitals, and cities with separate plans are the political and economic centers of the region, where innovation factors are concentrated, which may lead to city-level differences in the impact of “dual pilot” cities on city resilience. In this paper, we refer to the study of Zheng et al. [47], which divides all cities into core cities and general cities, among which municipalities, provincial capitals, and planned cities are core cities, and the rest are general cities. As shown from column (1) to (2) in

Table 14. Heterogeneity analysis.

Variable	Urban Hierarchy		Geographic Regions				Resource Endowment
	Core Cities	General Cities	Eastern Zone	Central Zone	Western Zone	Northeast Zone	Resource- Dependent City	Non- Resource-Dependent City
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Policy	0.0085 (0.7167)	0.0080 *** (3.4035)	0.0187 ** (2.4589)	0.0101 ** (2.0593)	0.0148 * (1.7702)	−0.0007 (−0.2421)	0.0055 (1.2279)	0.0181 *** (3.1499)
Controls	Y	Y	Y	Y	Y	Y	Y	Y
Year and City	Y	Y	Y	Y	Y	Y	Y	Y
Obs	490	3500	1204	1120	1190	476	1596	2394
Adj. R2	0.9432	0.8912	0.9533	0.9556	0.9224	0.9593	0.8232	0.9499

Note: t-statistics in parentheses. *** p < 0.01, ** p < 0.05, and * p < 0.1. And, Y denotes yes.

, the impact of “dual pilot” cities on core cities is not significant, while there is a significant positive impact on city resilience in general cities. The reason for this is that the core cities have already reached a high economic development level, and they themselves already have richer technological resources and better institutional environments, indicating that an innovation environment has already been established. According to the law of diminishing marginal utility, the impact of the “dual pilot” policy on core cities is relatively weak. General cities lack innovation resources, and their economic management authority is limited, making it difficult to attract innovation factors through preferential policies, such as fiscal, taxation, and finance policies, so the innovation level is relatively low. “Dual pilot” cities help to upgrade the information infrastructure of general cities, creating a more comfortable innovative environment. After the implementation of the policy, its marginal effect is more obvious, which leads to a significant increase in city resilience in the general cities.

5.2.2. Geographical Regional Heterogeneity

As the geographical location, climate, and other external conditions vary from region to region, their levels of economic development, innovation capacity, and other aspects are also different, so the promotion of “dual pilot” construction and its effectiveness may also change. Based on the latest regional division criteria of the National Bureau of Statistics, this study segments the sample into the eastern, central, western, and northeastern regions. As shown in columns (3)~(6) in Table 14, the influence of the “dual pilot” cities on city resilience is significantly positive in the east, central, and west areas, but the degree of the effect is different. The enhancement of city resilience in the western region is better than in the central region. This may result from the relatively underdeveloped economy in the western regions, and there are less applications of the emerging technologies proposed in the “dual pilot” construction. However, in the process of building “dual pilot” cities, the western region can effectively utilize the new generation of information technologies to upgrade the original backward industrial sectors, which will inject vitality into the economic development of the western region. As seen in column (6), the estimator for the northeast is not significant. This may be due to the relatively greater difficulty associated with applying new-generation technologies in traditional industries in the northeast.

5.2.3. Resource Endowment Heterogeneity

Resource endowment plays a key role in the economic development of a region. For example, abundant mineral resources lay the foundation for industrialization, which promotes economic growth. For separate regressions, the sample is segmented into resource-dependent cities and non-resource-dependent cities based on the Notice of the State Council on Printing and Distributing the National Sustainable Development Plan for Resource-Based Cities (2013–2020). As illustrated in columns (7) to (8) in Table 14, the coefficients for resource-dependent cities are not significant, while the estimators for non-resource-dependent cities are significantly positive at 1% significance. Resource-dependent cities are affected by the “resource curse”, and their excessive dependence on natural resources limits the industrial structure, technological progress, and the development speed and direction of the regional economy. Moreover, resource-dependent industries take up a large amount of capital, limiting the space for the survival and development of other industries.

6. Spatial Spillover Effect

6.1. Global Moran’s Index

The global Moran index is utilized to test whether the dependent variable, city resilience, is characterized by spatial aggregation. As shown in

Table 15. Global Moran index of city resilience.

Year	Geographical Matrix		Economic Matrix
	Moran’s I	z-Value	Moran’s I	z-Value
2009	0.0148 ***	3.6832	0.3306 ***	11.2171
2010	0.0157 ***	3.8890	0.3216 ***	11.0203
2011	0.0144 ***	3.6084	0.3352 ***	11.3916
2012	0.0133 ***	3.3834	0.3090 ***	10.5017
2013	0.0151 ***	3.7322	0.3259 ***	11.0707
2014	0.0123 ***	3.1808	0.3135 ***	10.6605
2015	0.0138 ***	3.4825	0.3018 ***	10.2866
2016	0.0134 ***	3.4049	0.3137 ***	10.6833
2017	0.0163 ***	3.9961	0.2974 ***	10.1625
2018	0.0166 ***	4.0688	0.2935 ***	10.0395
2019	0.0157 ***	3.8905	0.2873 ***	9.8247
2020	0.0152 ***	3.8149	0.2616 ***	9.0327
2021	0.0189 ***	4.5091	0.2795 ***	9.5327
2022	0.0187 ***	4.4724	0.2678 ***	9.1154

*** p < 0.01.

from 2009 to 2022 for 285 prefecture-level cities, the Moran index of city resilience values are all positive with at least 5% significance, indicating that there is a significant positive spatial autocorrelation of city resilience with spatial aggregation characteristics. Therefore, it is vital to analyze spatial spillover effects.

Based on the Moran’s Index, a scatter plot is constructed, as shown in Figure 5. Under both the geographical distance matrix and the economic distance matrix, the scatter points are clustered around the origin, predominantly in the first and third quadrants. This indicates that city resilience exhibits positive spatial clustering characteristics, although this clustering trend is relatively weak in terms of geographical proximity.

6.2. Model Selection

In this study we conducted a series of tests to select the appropriate spatial econometric model.

Table 16. Spatial econometric model selection.

Test	Primary Hypothesis	Statistics		Results
		Geographical Matrix	Economic Matrix
LM test	LM (error) test	336.565 ***	275.051 ***	SDM model
	Robust LM (error) test	445.741 ***	27.634 ***
	LM (lag) test	103.124 ***	379.102 ***
	Robust LM (lag) test	212.300 ***	131.685 ***
Wald test	SDM can be degraded to SEM	38.020 ***	269.790 ***	Rejecting degradation
	SDM can be degraded to SAR	33.040 ***	100.950 ***
Hausman test	Random effect model	−405.770	−4452.010	Fixed effect model
LR test	Fixed individual is better than two-way fixed	101.140 ***	66.840 ***	Two-way fixed effect of individual and time
	Fixed time is better than two-way fixed	7637.680 ***	7308.900 ***

*** p < 0.01.

presents the results, which demonstrate that the spatial Durbin model is statistically preferred. The Wald test demonstrates rejecting degradation, and the Hausmann test results are negative. Referring to the study of Lian et al. [54], it can be assumed that Hausman test has become invalid. Therefore, based on this, Arellano’s proposed Wald statistic is used to replace the Hausman test [55]. The test results indicate both results were 284.01, indicating that the model with fixed effects should be selected. Finally, combined with the LR test, the spatial Durbin model with both individual and time fixed effects was employed to systematically examine the spatial spillover effects of “dual pilot” cities on city resilience.

6.3. Spatial Spillover Analysis

Based on these tests,

Table 17. Results of spatial Dubin model.

	Geographical Matrix	Economic Matrix
Policy	0.0171 *** (11.62)	0.0126 *** (8.64)
W [Policy]	0.1074 *** (4.53)	0.0219 *** (6.23)
Direct effect	0.1760 *** (11.70)	0.0137 *** (9.21)
Indirect effect	0.1966 *** (3.71)	0.0314 *** (7.77)
Total effect	0.2142 *** (4.04)	0.0451 *** (10.51)
Spatial-rho	0.4137 *** (3.69)	0.2370 *** (8.18)
Sigma2_e	0.0002 *** (44.62)	0.0002 *** (44.40)
Controls	Y	Y
Year and city	Y	Y
Obs	3990	3990
R2	0.2515	0.3404

Note: t-statistics in parentheses. *** p < 0.01. And, Y denotes yes.

illustrates the spatial spillover effects of the “dual pilot” cities on city resilience under the geographic and economic distance weighting matrices. The results demonstrate that the spatial autoregressive coefficients are significantly positive under the two weight matrices, indicating that city resilience has a significant spatial correlation effect. That is to say, the city resilience level in a region will be positively affected by areas with close economic ties or areas that are within a close geographical distance. This is because these regions can promote the development of city resilience through various mechanisms, such as industrial undertaking and transfer, human capital flow, and regional economic cooperation. The estimators of the spatial interaction term of the “dual pilot” cities are significantly positive, suggesting that the sample cities have both the exogenous interaction effect of the “dual pilot” cities and the endogenous interaction effect of city resilience. That is, the “dual pilot” citiesexerta spatial spillover effect on city resilience. Therefore, assumption H3 is supported.

The impact effect is further decomposed into direct, indirect, and total effects. The result demonstrates that the direct effect of the “dual pilot” cities on city resilience is significantly positive, indicating that the “dual pilot” cities can significantly and directly contribute to the improvement of local city resilience which may be attributed to the dual overlay of factors such as fiscal resources and technological innovation brought about by the policy, creating a synergistic effect where “1 + 1 > 2”. The indirect effect of the “dual pilot” cities on city resilience is also significantly positive, indicating that the local “dual pilot” cities can enhance the city resilience of geographically adjacent areas and economically connected areas of the pilot cities. The possible reason for this is that “dual pilot” cities can synergistically accelerate the cross-regional flow of urban innovation factors, enabling them to maximize their scale and network effects, thereby enhancing city resilience in surrounding areas.

7. Conclusions and Policy Recommendations

7.1. Conclusions

We explore the effect, mechanism, heterogeneity, and spatial spillover effect of “dual pilot” cities on city resilience. based on panel data from 285 prefecture-level cities from 2009 to 2022. This study first employs the Difference-in-Differences method to identify the causal effect of the “dual-pilot” policy on city resilience, examining whether the dual pilot initiative enhances city resilience. Building on this, the mediation analysis is conducted to explore the mechanisms through which this effect occurs. Furthermore, considering the significant heterogeneity among Chinese cities, the heterogeneity tests are performed to reveal for “whom” or under “what conditions” the policy is more effective, thereby supporting more targeted policy implementation. Finally, a Spatial Durbin Model is introduced to extend the perspective from isolated cities to urban networks, examining whether the policy effects exhibit spatial spillovers or competition. The findings are as follows:

(1)

“Dual pilot” cities can promote the rapid improvement of city resilience. These findings are still valid after various robustness tests (parallel trend test, placebo test, staggered DID, PSM-DID, Winsorization, replacement of explained variables, exclusion of other policy interferences).

(2)

“Dual pilot” cities have a synergistic effect on city resilience, and the policy effect of the “dual pilot” is significantly better than that of the “single pilot”. In addition, the policy effect of “dual pilot” cities is affected by the order of the policy implementation. Specifically, establishing smart cities before innovative cities is more conducive to improving city resilience.

(3)

Through the mediating effect test, it is found that the “dual pilot” cities enhance city resilience by improving innovation and development levels and industrial structure upgrading.

(4)

A heterogeneity analysis found that the policy effect of “dual pilot” cities is better in general, non-resource-dependent, eastern, central, and western cities. The “dual pilot” cities have positive spatial spillover effects on city resilience.

7.2. Policy Recommendations and Limitations

According to above conclusions, we offer the following policy suggestions:

(1)

We must accelerate digital infrastructure development in the “dual pilot” cities and achieve the greatest complementary advantages through the coordinated development of digital policies and innovation policies. First, we should focus on the policy linkage effect between digital infrastructure construction and “dual pilot” cities and explore the synergistic effect between different policies to realize the synergistic development of China’s city resilience on the basis of digital integration and innovation. Second, the implementation plan and sequence of the “dual pilot” policy should be rationally planned to fully contribute to the synergistically empowering effects of multiple policies. And we should fully consider the relevance and coordination of various policies to avoid policy conflicts and resource waste, thereby optimizing policy performance as much as possible.

(2)

We should emphasize the role of the “dual pilot” city in boosting innovative development and industrial structure upgrading. For one thing, the local government fully considers the driving force created by the integration of the digital and physical. We should also empower the real economy through digital technology to break the traditional market boundaries, to expand the market size, to stimulate new economic developments, to promote market prosperity, and to lay a solid economic foundation for improving city resilience. In addition, we should focus on enhancing the independent innovation capability of enterprises and guide innovative elements such as talent, capital, and technology in high-tech industries, so as to realize the optimal allocation of talents. It isa also necessary to fully integrate the industrial chain and external innovation elements, increase the basic research on digital technology, and use digital technology to crack the innovation bottleneck in the industrial chain, comprehensively enhancing the support capacity of science and technology innovation for resilient cities.

(3)

We should promote smart cities and innovative cities according to local conditions and needs. First, the implementation of the “Northeast Revitalization” strategy should be further pursued to promote the modernization and transformation of traditional industries in the northeast and continuously improve the level of their infrastructure, thereby creating a favorable environment. Secondly, the “late development advantages” of the central and western regions and general cities still need to be explored, so pilot policies can focus on selecting cities with characteristic industries or comparative advantages to promote coordinated development between regions in the future. Finally, the innovation environment and digital construction of non-resource-dependent cities should be further optimized to strengthen the diversified development of their industries. Meanwhile, we should accelerate transformation, actively cultivate new industries, and gradually change resource-dependent development paths in resource-dependent cities.

(4)

We should maximize the effects of pilot cities and boost the coordinated regional development of “dual pilot” construction. Taking city clusters and metropolitan areas as units, we should accelerate the exploration and formation of inter-city consultative mechanisms to enhance city resilience, encourage the exchange of experience and technology transfer between cities, and popularize the successful experiences of advanced cities to enhance the level of city resilience in the entire region.

Although this study has comprehensively explored the impact of the smart and innovative city dual pilots on city resilience, there are still some limitations in this research. First of all, the construction of city resilience indicators may involve selection bias and measurement errors. City resilience is a multidimensional and complex system. Although this study has selected as many indicators as possible, constraints such as data availability have still prevented the inclusion of all relevant factors. In the future, when constructing indicator systems, factors such as technological innovation and natural conditions could be incorporated into the framework, while due consideration should be given to addressing potential measurement errors. Second, the intermediary effect analysis only considers the influence of the innovation development level and industrial structure upgrading, ignoring the influence of factors such as enterprise investment and government intervention. Third, the analysis of the spatial spillover effect only proves its existence without providing more in-depth research. In the future, the distance between cities can be divided into several intervals, and the specific characteristics of the spatial spillover effect within different intervals can be studied in depth. Finally, this research is mainly carried out at the macro-level. However, the impact of the “dual pilot” policy on microenterprises is also worth exploring.