The Impact of Smart City Policies on City Resilience: An Evaluation of 282 Chinese Cities

Abstract

This study uses four dimensions, namely social resilience, economic resilience, infrastructure resilience, and ecological resilience, to construct an index system for urban resilience. The subject data came from the panel data of 282 prefecture-level cities in China from 2006 to 2020. We selected a multiperiod double-difference model to study the effects of smart city pilot policies on macro-urban resilience. By conducting parallel-trend tests and selecting appropriate robustness tests, this study drew the following relevant conclusion: smart city pilot policies can have a positive effect on the urban resilience level. These policies exert their influence by facilitating industrial structure upgrading, which plays a partial mediating role. Considering different city area distributions and city scales, smart city pilot policies can have significant heterogeneity in their enhancement of urban resilience. The effect is pronounced in the “east > central > west” states and is more likely to have a significant impact on small- and medium-sized cities. Therefore, promoting the scope of smart city pilots, strengthening the intermediary role of industrial structure upgrading, and implementing differentiated policies for different regions and city sizes are important for sustainable urban development.

1. Introduction

In recent years, various natural disasters and man-made disasters have frequently occurred around the world. Urban problems such as high temperatures, earthquakes, political instability, environmental pollution, and traffic congestion have become increasingly serious. At the same time, infectious diseases, such as COVID-19, have begun to spread, threatening and damaging human survival and the sustainable development of the health of the planet. “Urban resilience” was first introduced at the United Nations Global Sustainable Development Summit in 2002 [1]. The summit aimed to explore ways to help cities cope with a variety of challenges, including natural disasters, man-made disasters, atmospheric change, and urbanization. In 2013, the “100 Resilient Cities around the World” project was proposed, with the main purpose of promoting the construction of resilient cities around the world. The project provides financial and macro-support to help cities reasonably solve various difficulties and enhance their resilience. It has played an important role in promoting urban development and calmly responding to developmental challenges, and it has become an important symbol of resilient city construction [2]. In the 20th Report of the Communist Party of China (CPC) Central Committee, specific implementation strategies were put forward for urban renewal actions, focusing on strengthening urban infrastructure construction and creating livable, benign, and smart cities.

The emphasis on urban resilience is reflected in everything from international action to the strategic requirements of China, a developing country. To sum up, urban resilience can help cities successfully cope with various challenges, pressures, and adverse factors. Specifically, these can be categorized as emergency management, adaptive smart city systems, urban sustainability, equity, and urban resilience. In the context of emergency management, urban resilience is the ability of cities to recover in the immediate aftermath of a disaster to save lives and provide shelter [3]. In the context of adaptive smart city systems, adaptability emphasizes organized resilience to external threats [4]. In terms of urban sustainability, however, there are conceptual differences between urban resilience (UR) and urban sustainability (US). Unlike US, UR emphasizes the stability and diversity of the urban economic structure and the institutional arrangements of the urban structure; therefore, it was more suitable for testing the policy effects in this study [5]. In terms of equity and urban resilience, equity has been recognized as one of the key factors influencing urban resilience, which Heather analyzed in detail through case studies [6]. Resilient cities have the ability to resist, adapt, and recover, thus promoting the sustainable and healthy development of cities. However, there are few studies on the impact of policies on urban resilience, and given the availability of data to the authors and the depth of analysis, each prefecture-level city in China was chosen as the scope of this study. China began to build smart cities in three batches of pilot projects in 2012, but the impacts of the policy on urban resilience, the specific path of implementation, and the heterogeneity remain to be verified. This study focuses on solving the abovementioned problems to provide a reliable theoretical basis and case reference for the study of urban sustainable development strategies.

This study contains an introduction, literature review, theoretical analysis, research hypotheses, data and methods, results and analysis, research findings, and policy recommendations. In the Section 1, the background and significance of the study are briefly introduced, the importance of smart city pilot policies in enhancing urban resilience is emphasized, and the research questions and objectives are presented. In Section 2, relevant studies on smart cities, urban resilience, industrial structure upgrading, and the relationships between them are reviewed, and gaps in the research on smart city pilot policies for enhancing urban resilience through industrial structure upgrading are described. In the Section 3, based on the finding that smart city pilot policies enhance urban resilience through four dimensions (economic, social, environmental, and infrastructure), the hypothesis that smart city pilot policies can enhance urban resilience through industrial structure upgrading is proposed, and the differences in smart city pilot policies across different regions and city scales are further analyzed. In the Section 5, the positive impact of smart city pilot policies on urban resilience and the mediating role of industrial structure upgrading in this process are verified by presenting the results of the empirical analysis. In the Section 6, the main findings, including the positive impact of smart city pilot policies on urban resilience, the mediating role of industrial structure upgrading, and the heterogeneity in regional and city scales, are summarized. The following relevant recommendations are also put forward: expanding the scope of smart city pilots, strengthening the mediating role of industrial structure upgrading, and formulating differentiated policies for regional- and city-scale differences.

2. Literature Review

The word “resilience” is derived from “resilio”. Originally, resilience refers to the essence and ability of material objects to recover their original state after being affected by external forces [7]. In 1973, Holling, an ecologist from Canada, applied this concept to the field of ecology for the first time and provided a detailed classification, which was divided into engineering resilience and ecological resilience [8]. In other research, “resilience” was introduced into urban studies. From an economic perspective, Hill (2008) defines urban resilience as the ability of an economy to maintain its pre-shock level of growth following external shocks [9]. According to Brown (2012), from the perspective of management, urban resilience refers to the ability of individuals, communities, or institutions to show dynamic trends and effectively cope with changing factors so as to continue operating at an acceptable level [10]. Marta (2016) mainly took provincial capitals of Spain as an example to analyze the basic meaning of urban resilience from the perspective of social ecosystems and suggested that urban resilience levels should be measured regularly. At the same time, an urban resilience index should be integrated with the theoretical framework of the resilience of urban social ecosystems [11].

The assessment of urban resilience can be divided into two aspects, namely drivers and mechanisms, with drivers including the single dimension of disaster risk management and the overall systemic aspect of sustainable urban development [12].

Regarding the influencing factors, in terms of disaster risk management, Shaw et al. (2011) selected 10 cities in Asia as research objects, established an Asian climate change-resilience research network based on actual development conditions, and adopted a dynamic management mechanism to help cities cope with regional climate change as much as possible [13]. Du et al. (2018) took Da Nang, Vietnam, as an example and established a dynamic analysis of urban water supply services to cope with the risk of urban droughts and a water shortage governance system. The above content can provide a reference for other work [14]. Herrmann et al. (2020) further analyzed a tsunami in Chile through semi-structured interviews and other means and planned four villages that were prone to disasters; they concluded that the inclusion of risk zones in urban planning can enhance tsunami resilience [15]. Hemmati et al. (2020) pointed out the relatively unexplored topic of flood risk, emphasizing that flood risk should be considered during urban growth. They suggested that building the resilience of communities is now a national and global priority [16]. Shan et al. (2023) set up an analytical framework for urban resilience in post-disaster recovery based on social media data, mainly to help define urban resilience in post-disaster recovery from the perspective of infrastructure conditions and psychological construction [17]. Zhai et al. (2024) explored the differences in perceptions of urban disaster resilience among different races, genders, and groups based on Twitter data, and they concluded that such differences should be evaluated objectively in practice to ensure the inclusiveness of disaster resilience in a targeted manner [18]. Kim et al. (2024) utilized the data envelopment analysis to assess urban disaster resilience in terms of social, economic, infrastructural, social capital, and institutional aspects, and they concluded that a strong infrastructure would enhance urban disaster resilience and recovery in terms of infrastructure and psychological conditions, concluding that strong infrastructure enhances urban disaster resilience [19]. Zhai Guofang (2024), on the other hand, elaborated on the connotation of key concepts, such as urban rain and flood disaster resilience, from the perspective of planning for rain and flood disasters and proposed the basic ideas and frameworks for coping with urban rain and flood disasters based on the five aspects of the risk governance process, elements, subjects, levels, and types of disasters from the experience of Australia, Japan, and other countries [20]. Jiang Yinghong and Shen Leihong (2021) took the outbreak of the novel coronavirus epidemic at the end of 2019 and the southern flood in 2020 as a research background, proposed the concept of resilient and healthy open space from the perspective of resilient and healthy city construction, and put forward characteristics and planning strategies of resilient and healthy open spaces based on the complexity of multiple systems superimposed on each other for integrated disaster prevention [21]. Hu Yuchen et al. attempted to apply the resilience theory in multi-hazard scenarios, differentiated resilience according to the loss and recovery rate, constructed a new resilience assessment system and method and applied them to 21 prefecture-level cities in Guangdong Province as an example [1].

In the context of sustainable development, Feofilovs et al. (2021) simulated different scenarios of urban resilience based on system dynamics modeling, which was superior to the method of assessing urban resilience on the basis of evaluation metrics and was more applicable to the development of urban resilience strategies [22]. Moraci et al. (2018) expressed, through the analysis of several European case studies in which resilience methods were successfully utilized, the approach of resilience as a new paradigm for smart planning, in addition to highlighting the important role that resilience plays in urban planning and healthy development at all levels [23]. Sharma et al. (2023) put forward the concepts of vulnerability and adaptability mainly on the basis of the concept of resilience, and they constructed a framework for urban resilience with four dimensions, vulnerability, urban planning, countermeasures and adaptability, and urban governance, and suggested that this could be an effective tool [24]. Fischer et al. (2018) proposed a resilience framework with five phases—preparation, prevention, protection, response, and recovery—to quantitatively analyze the sensitivity, vulnerability, response, and recovery behaviors of a complex system in different scenarios, and the method was able to provide a holistic assessment of urban resilience; the results could be quantitatively compared with those of other cities to achieve regional sustainable development [25]. Sajjad et al. (2021) validated the resilience level of Hong Kong’s Shatin region using ANOVA based on the construction of the Spatial Disaster Resilience Analysis (S-DReP) framework, which included 24 evaluation indicators. In the end, it was suggested that the framework could provide basic guidance and a roadmap for urban resilience knowledge systems in high-density cities, which can be beneficial for timely decision making in the face of disasters to reduce the impact on cities [26]. Tang et al. (2023) composed an urban resilience evaluation index system with five dimensions, economic resilience, social resilience, environmental resilience, infrastructure resilience, and institutional resilience, and they further studied the impact of industrial structure transformation and upgrading on urban resilience according to this content and concluded that industrial structure rationalization can promote urban resilience more than upgrading [27]. He et al. (2023) constructed an elastic bid evaluation system based on “scale–density–form” as the main criterion and discussed the influencing factors of urban resilience in view of the spatial changes in the urban resilience level. Ecological factors, which were represented by the per capita ecological land area, have gradually become the key influencing factors [28]. Zhu Zhengwei et al. (2024) responded to the need for urban safety resilience evaluation from a safety perspective, and they pioneered a resilience perspective that included structure, function, and safeguarding; then, they proposed a framework for urban safety resilience assessment [29]. Li Na et al. (2023) constructed an urban resilience evaluation index system with four dimensions—economy, society, ecology, and digitalization—in 2023, and they investigated the heterogeneity of urban resilience levels among different city clusters and earthquake zones. They concluded that there is a significant link between cities within city clusters, even in earthquake areas, and the division of resilience levels is polarized [30].

In terms of the mechanism of action, You et al. explored the mechanism of urban resilience based on panel data from Jiangsu Province from 2009 to 2018. They found that, among government factors, the ratio of general government expenditure to GDP can positively promote the level of urban resilience, while foreign trade plays an inhibiting role [31]. Based on panel data from 110 prefecture-level cities in the Yangtze River Economic Belt from 2010 to 2019, Tang et al. explored the mechanism of the role of industrial structure transformation and upgrading in urban resilience. The study concluded that both industrial structure rationalization and industrial structure upgrading can significantly promote the level of urban resilience, and the promotive effect of industrial structure rationalization is enhanced. However, there is a lag effect in the promotive effect on economic resilience [27]. Yu et al. explored the impact of the new urbanization pilot policy on urban resilience from the perspective of macro-policy by analyzing data from 281 prefectural-level cities in China from 2006 to 2020. The study found that new urbanization policies can significantly enhance the level of urban resilience; specifically, there are stronger effects on economic resilience, infrastructure resilience, and institutional resilience and weaker promotive effects on social resilience. In addition, new urbanization policies are mainly promoted through technological innovation and economic agglomeration [32]. However, in the context of the development of China’s urban agglomeration, Jiang et al. further explored the impact of regional integration policies on urban resilience. The study concluded that the regional integration policy was able to enhance the level of urban resilience by 8.6%. Specifically, the policy had a more pronounced positive effect on economic resilience, a weaker effect on social and infrastructural aspects, and a negative effect on ecological resilience [33].

In summary, the current studies on urban resilience are numerous, and they mostly focus on influencing factors and mechanisms of action, where the mechanisms of action include analyses of specific variables and macro-policy analyses. However, from a policy perspective, considering the differences between smart cities and resilient cities, the exploration of the net effect of smart city pilot policies on the level of urban resilience still needs to be further supplemented. Based on this, the possible marginal contributions of this study are as follows.

(1) An urban resilience index system was constructed, which included four dimensions: social resilience, economic resilience, infrastructural resilience, and ecological resilience. Based on the panel data of 282 prefecture-level cities in China from 2006 to 2020, the level of urban elastic development was preliminarily explored. (2) A multiperiod, double-difference model was constructed to examine the net effect of smart city pilot policies on urban resilience from an ex post perspective to test whether the results are robust in the long run and to illustrate the effectiveness of policies at the regional and urban levels. (3) At present, there are few academic research results on the mediating role of industrial structure upgrading in the implementation of smart city pilot policies, and there is a certain gap in this field. The authors mainly adopted the Sobel test and Bootstrap test to verify the mediating effect of industrial structure upgrading, defined the effect proportions, and put forward realizable policy suggestions based on the results.

3. Theoretical Analysis and Research Hypotheses

Based on the theory of pilot policies and policy implementation [34], this study analyzes the four dimensions of urban resilience to analyze how smart city pilot policies can enhance the level of urban resilience. The pilot policy implementation process can play a leading role in policy promotion and technology. From a macro-perspective, industrial upgrading can reflect not only the economic base of a city at the macro-level but also the level of resource allocation at the micro-level. Thus, smart city pilot policies are further explored to enhance the level of urban resilience through industrial structure upgrading. However, in the promotion process, given the vastness of China, the gaps between the east, middle, and west are large, and there are some prefecture-level cities with varying levels of development. It is necessary to further explore the regional- and city-scale heterogeneity of smart city pilot policies to enhance the level of urban resilience.

3.1. Pilot Smart City Policies Can Enhance Urban Resilience

Policy piloting is a unique policy formulation and implementation mechanism in China. Its operational mechanism is a process of central guidance, local practice, gradual extension, optimization, and improvement. In policy piloting, policy implementation is also required. According to the theory of policy implementation involving mutual adjustment between policy implementers and those affected by the policy, the level of urban resilience is enhanced in the process. Starting from the aspects of economic resilience, social resilience, ecological resilience, and infrastructural resilience, this study presents a theoretical analysis of the ability of smart city pilot policies to improve urban resilience. In terms of economic resilience, smart cities maximize macro-resource allocation through the introduction of advanced technologies, such as the Internet of Things, digitalization, and cloud computing, and guide industries to achieve supply-side structural reform and transformation, which is conducive to improving the operational capacity of traditional industries and the germination of emerging industries, such as smart finance and smart logistics, thereby enhancing the diversity and flexibility of the city’s economy [35]. During economic downturns, a diversified industrial structure helps cities better withstand economic shocks and achieve rapid recovery and adaptation. In addition, smart cities can attract more businesses and talents, promote innovation and entrepreneurship, and further promote economic growth and resilience. In terms of social resilience, smart cities can drive development and progress in social resilience. First, smart cities strengthen infrastructure construction through smart government, medical care, education, and teaching; improve the efficiency, coverage level, and degree of public services as much as possible; and enhance social security capabilities [36]. Second, digital big data and artificial intelligence technology can be used to systematically optimize urban management, improve emergency response speed and handling capacity, and effectively reduce the impacts of natural disasters and social emergencies on cities. Finally, smart cities also promote social equity and inclusive development, reducing social inequality and enhancing overall social resilience through the provision of equalized public services [37]. In terms of ecological resilience, smart cities also have significant advantages in ecological environmental protection. Through the construction of smart environmental protection, smart energy, and other systems, smart cities are able to realize the real-time monitoring and scientific management of the urban ecological environment. This helps to discover and solve environmental problems in a timely manner and prevent damage to the ecosystem. Smart cities can also promote environmental protection concepts, such as green transportation and construction; reduce the consumption of natural resources and pollution emissions; protect ecological harmony; and improve the ecological resilience of cities. The construction of infrastructure in smart cities is also key in enhancing urban resilience. Smart cities achieve the real-time monitoring, early warning, and emergency response of urban infrastructure by building intelligent infrastructure systems, such as smart transportation, smart grids, and smart water grids [36]. These systems are able to identify and solve problems and hidden dangers in infrastructure operations in a timely manner, preventing disruptions in city operations and disasters caused by infrastructure failures. In addition, smart cities focus on the redundant design of infrastructure and the construction of backup mechanisms to ensure that urban infrastructure can still operate normally under extreme circumstances and provide continuous and stable services.

In general, smart city pilot policies promote the overall resilience of cities by improving the overall economic, social, ecological, and infrastructural resilience, which can help cities solve various internal and external challenges and help society achieve sustainable development and prosperity. This allows us to formulate the following hypothesis:

H1: 

Smart city pilot policies can improve urban resilience.

3.2. Smart City Pilot Policies Enhance Urban Resilience through Industrial Structure Upgrading

Smart city pilot policies improve the level of urban resilience through the upgrading and innovation of the industrial chain; in particular, this happens through the promotion and stimulation effects of smart city pilot policies on industrial upgrading and the promoting role of industrial structure innovation on urban resilience.

In terms of the incentivizing effect of smart city pilot policies on industrial structure upgrading, the introduction of smart city pilot policies has provided a clear direction and incentive for industrial structure upgrading. Through macro-control and institutional guidance, the government encourages industries, schools, scientific research institutions, and other social levels to join in the construction of smart cities, and it strives to promote technological and industrial innovation and upgrading. The government also creates a favorable environment and conditions for industrial structure upgrading by providing financial support, tax incentives, talent introduction, and other policy measures.

From the perspective of the impact of industrial structure upgrading on urban resilience, in terms of economic resilience, industrial structure upgrading promotes the optimization and improvement of the urban economic structure, thus enhancing the stability and risk-prevention ability of the economic system. Cities with more advanced industrial structures are able to adapt more quickly to economic fluctuations and external shocks and recover more effectively, maintaining stable economic growth. In the dimension of social resilience, smart city construction enhances the social resilience of cities by upgrading the level of public services and optimizing urban management. Meanwhile, the upgrading of industrial structure provides technical support and guarantees the intelligence and efficiency of these services. In the dimensions of ecological resilience and infrastructural resilience, the structure also promotes the improvement of the urban ecological environment and the intelligent construction of infrastructure.

In summary, Hypothesis H2 is proposed:

H2: 

Smart city pilot policies can increase the level of urban resilience through industrial structure upgrading.

3.3. Smart City Pilot Policies to Enhance Urban Resilience Levels Have Regional- and City-Scale Heterogeneity

Smart city pilot policies to improve the level of urban resilience differ in their degrees of implementation due to differences between regions and cities in terms of the economic development level, scientific and technological innovation capacity, infrastructure construction, and population size.

Concerning regional heterogeneity, differences in the economic development level and resource endowment will lead to regional development gaps. In regions with better development degrees, the high-tech industry is often more active, there are stronger independent innovation capabilities, the digital transformation of traditional infrastructure is promoted, and resilience is improved. Conversely, regions with a low level of economic development may face larger gaps in terms of capital, technology, and talent, which can limit the effectiveness of smart city construction and the enhancement of resilience levels. At the same time, there are also gaps in macro-policy deployment and financial support in different regions. For example, the national government and the hierarchical (local) government have different understandings of smart city construction, which will result in different effects and levels of smart city construction in different regions. In terms of city-scale heterogeneity, whether smart city pilot policies can be effective at the level of urban resilience depends on the size and system of the city. At the initial stage of urban development, big cities, due to their dense populations and frequent economic activities, have more pressing needs for smart city construction and resilience improvement. Simultaneously, they possess greater resources and conditions to support smart city construction. In the later stages of urban development, the phenomenon of urban residents returning to their hometowns becomes pronounced, which also leads to a more urgent need for smart city construction and resilience enhancement in small- and medium-sized cities. Consequently, smart city pilot policies may yield different effects across different city sizes.

As a result, smart city pilot policies have been proven to have regional- and city-scale heterogeneity in their improvement of urban resilience.

H3: 

Smart city pilot policies to enhance urban resilience have regional- and city-scale heterogeneity.

4. Data and Methods

4.1. Indicator System Construction, Data Sources, and Sample Selection

4.1.1. Construction of the Indicator System

(1)

Explanatory variables

For the measurement of urban resilience, which was guided by the principles of scientificity, comprehensiveness, and accessibility of indicators based on connotations, this study drew on the research of Zhou Qian et al. [38], Zhang Mingdou et al. [39], and Cheng Hao et al. [40]. It began with four dimensions: social resilience, economic resilience, infrastructural resilience, and ecological resilience. The key indicators selected are presented in

Table 1. System of urban-resilience indicators.

Target Level	Standardized Layer	Indicator Layer	Causality
Urban resilience	Social resilience	Unemployment rate	−
		Number of teachers in regular institutions of higher learning	+
		Number of participants in basic medical insurance	+
		Number of hospital beds	+
	Economic resilience	GDP per capita	+
		Per capita disposable income	+
		Consumption expenditure per capita	+
		Revenue	+
	Infrastructural resilience	Road area per capita	+
		Road freight	+
		Gas penetration rate	+
		Drainage pipe length	+
	Ecological resilience	Green coverage rate of built-up area	+
		Rate of harmless treatment of domestic waste	+
		Annual urban water supply	+
		Urban sewage treatment rate	+

Notes: A positive sign in an attribute indicates that the larger the value of the indicator in the indicator layer, the better the criterion layer. A negative sign indicates that the larger the value of the indicator at the indicator level, the worse the criterion level.

. This study defined social resilience, economic resilience, infrastructural resilience, and ecological resilience by referring to the study of Jiang et al. [33]. Social resilience focuses on the ability to stabilize and promote recovery in the social-structure aspect of urban resilience. The unemployment rate is used to indicate the living standard and social stability of residents, the number of teachers in schools for general higher education indicates the status of education resources and human resources in the city, the number of participants with basic medical insurance indicates the economic support of residents when they face health risks, and the number of beds in hospitals indicates the ability to cope with public health emergencies. Economic resilience focuses on the ability of the economic system to adapt to changes and resume growth in the context of urban resilience. GDP per capita indicates the current level of economic development, disposable income per capita indicates the current consumption capacity of residents and their ability to cope with future risks, consumption expenditure per capita indicates the degree of economic activity, and fiscal revenue indicates the government’s ability to cope with shocks and provide public services. Infrastructural resilience focuses on the ability of infrastructure systems to maintain normal operation and recover in the face of shocks, such as disasters. The road area per capita indicates the smoothness and efficiency of the transportation infrastructure; the road freight volume indicates the efficiency of the urban logistics system; the gas penetration rate indicates the improvement of the living infrastructure for residents; and the length of drainage pipes indicates the drainage capacity of a city to cope with natural disasters, such as waterlogging. Ecological resilience focuses on the ability of urban ecosystems to maintain ecological balance and recover in the face of environmental degradation. The rate of greening of built-up areas indicates the capacity of a city’s ecological environment and climate regulation, the rate of safe treatment of domestic waste indicates the capacity to manage pollution at source, the annual water supply of the city indicates the capacity of the city’s ecosystems and development, and the rate of wastewater treatment in the city indicates the capacity to manage the environment. By using the entropy method to comprehensively evaluate 282 prefecture-level cities in China, we obtained a comprehensive index of urban resilience.

(2)

Mediating variable

Using the transformation and upgrading of industrial structure (struc) as the mediating variable and drawing on the research of Chang X. [41], the ratio of the added value of tertiary industry to the added value of secondary industry was employed for measurement. This indicator directly reflects the relative level of development and trend of change in the service sector within the industrial structure.

(3)

Control variables

In order to ensure that the impact of smart city pilot policies on urban resilience was measured without interference from other key factors, we controlled for five aspects: human capital level, government intervention level, industrialization level, financial development level, and openness to the outside world. Among these, the level of human capital was measured by using the study by Shi Yufang et al. [42] on the ratio of university graduates to permanent residents. The level of government intervention was measured using the ratio of general budget expenditure to GDP. The level of industrialization was measured using the ratio of industrial added value to GDP. The level of financial development was measured using the ratio of the regional deposit and loan balance to GDP. The level of openness to the outside world was measured using the total import and export trade volume/GDP.

4.1.2. Data Sources and Sample Selection

Taking the scientificity, comprehensiveness, and accessibility of indicators into account, this study utilized panel data from 282 prefecture-level cities in China, spanning from 2006 to 2020, as research objects. Central governments engage in strategic planning to identify a list of pilot smart city policies, taking into account the technological development capacity, government governance capacity, and local human capital base capacity. All data were sourced from the China Urban Statistical Yearbook; the China Science and Technology Statistical Yearbook; and the statistical yearbooks of various provinces, autonomous regions, and municipalities in China. For some missing data, linear interpolation was employed to address the issue. Additionally, to mitigate the effect of heteroscedasticity, the variables of human-capital level and government-intervention level were logarithmically processed. Descriptive statistics for each variable are presented in

Table 2. Descriptive statistics.

Variables	Number of Observations	Average Value	Standard Deviation	Minimum Value	Maximum Values
UR	4230	0.0663	0.0663	0.0087	0.6345
lnhuman	4230	−0.0431	1.0632	−5.5558	2.5467
lngover	4230	−1.8097	0.4643	−3.1554	0.3955
lnindus	4230	−0.9573	0.3626	−3.5906	−0.1144
finan	4230	2.2865	1.1849	0.5879	21.3015
open	4230	0.9539	0.5384	0.0940	5.3500
lnstruc	4230	−0.1578	0.4773	−2.4220	1.6771

.

4.2. Research Methodology

4.2.1. Entropy Method

The entropy-value method is a kind of objective weighting method that can effectively avoid the interference of human factors and provide objective evaluation results. The specific calculation process is as follows.

(1)

Selection of data

θ years, m indicators, and n samples are selected; then, $x_{\theta ij}$ is the jth indicator of the ith sample in the θth year; i = 1, 2, 3, … n; j = 1, 2, 3, … m; and θ = 1, 2, 3, … r.

(2)

Standardized treatment

Among the selected indicators, some are positive indicators, meaning that a larger value indicates better results. Some indicators are negative, meaning that the smaller the value, the better the effect. When the directions of the above indicators are inconsistent, it is necessary to use Equations (1) and (2) to standardize positive and negative indicators, respectively.

$$x_{\theta ij}^{\prime }={\displaystyle \frac{x_{\theta ij}-x_{min}}{x_{max}-x_{min}}}$$

(1)

$$x_{\theta ij}^{\prime }={\displaystyle \frac{x_{max}-x_{\theta ij}}{x_{max}-x_{min}}}$$

(2)

In order to avoid the problem of logarithmic meaninglessness when seeking the entropy value, this study standardized the processing of the data after shifting: a very small value of 0.0001 that did not affect the original data was added to all.

(3)

Calculation of the sample weights

After standardizing the data, the weight of the sample, $P_{\theta ij}$, was further calculated, i.e., the weight of the standardized value of the jth indicator of the ith sample in year θ over the sum of all data under the selected indicator, as shown in Equation (3).

$$P_{\theta ij}={\displaystyle \frac{x_{\theta ij}^{\prime }}{{\sum }_{\theta }{\sum }_i{x}_{\theta ij}^{\prime }}}$$

(3)

(4)

Calculation of the entropy value of the jth indicator

The formula is shown in Equation (4).

$$E_j=-k\sum _{\theta =1}^r\sum _{i=1}^n{P}_{\theta ij}\ln P_{\theta ij}$$

(4)

(5)

Calculation of the coefficient of variation

The information utility value, ${\mathrm{g}}_{\mathrm{j}}$, of the indicator is dependent on the difference between the information entropy, $E_j$, and 1, and it has a direct impact on the magnitude of the weights. Specifically, a larger information utility value indicates greater importance and correspondingly larger weights. The formula for this relationship is presented in Equation (5).

$$g_j=1-E_j$$

(5)

(6)

Calculation of indicator weights

With the help of the entropy method to measure the weight of each indicator, ${\mathrm{W}}_{\mathrm{j}}$, the essence is calculated by finding out the coefficient of variance; the higher the coefficient of variance, the greater the weight and the greater the contribution to the results. The formula is shown in Equation (6).

$$W_j={\displaystyle \frac{g_j}{{\sum }_{j=1}^m{g}_j}}$$

(6)

(7)

Calculation of the sample composite index

The composite index of the sample is further obtained, and the formula for this calculation is presented in Equation (7).

$$Z_{\theta i}=W_j{x}_{\theta ij}^{\prime }$$

(7)

4.2.2. Multiperiod Double-Difference Models

From 2012 to 2014, China established pilot cities for smart cities in three batches. This study considers pilot policies as quasi-natural experiments. By treating the smart cities established by the country as a treatment group and other cities as a control group, we were able to employ a multiperiod difference-in-differences model to assess the net effect of the smart city pilot policies on urban resilience [43]. The basic model is established as follows:

$${UR}_{it}={\alpha +{\beta }_1{did}_{it}+{\beta }_2{controls}_{it}+\mu }_i+{\delta }_t+{\epsilon }_{it}$$

(8)

where i denotes the city, and t denotes the time. ${UR}_{it}$ denotes the urban resilience level of city i in year t. ${did}_{it}$ denotes the state of city i in year t. If city i was set up as a pilot smart city in year t, it is assigned a value of 1; otherwise, it is 0. ${controls}_{it}$ is the selected control variable; ${\mu }_i$ and ${\delta }_t$ denote the fixed effect of the city and the fixed effect of time, respectively; $\alpha$ is a constant term; and ${\epsilon }_{it}$ is a random perturbation term.

4.2.3. Models of the Mediating Effect

In this study, we drew on the approach of Shi Yufang et al. [42] and considered the Sobel test and Bootstrap test together to obtain reliable mediating effect results with the following model:

$$Y=cX+e_1$$

(9)

$$M=aX+e_2$$

(10)

$$Y=c^{\prime }X+bM+e_3$$

(11)

Coefficient c in Equation (9) represents the total effect of the independent variable, X, on the dependent variable, Y, while coefficient a in Equation (10) indicates the effect of the independent variable, X, on the mediating variable, M. Coefficient b in Equation (11) denotes the effect of the mediating variable on the dependent variable under the influence of the independent variables on the dependent variables. Coefficient $c^{\prime }$ represents the direct effect of the independent variables on the dependent variables, and $e_1$, $e_2$, and $e_3$ are all residuals.

5. Results and Analysis

5.1. Benchmark Regression

The application of the multiperiod double-difference model in the regression analysis yielded the basic regression results presented in

Table 3. Benchmark regression results.

Variables	UR	UR	UR	UR	UR	UR
Did	0.0067 ** (2.12)	0.0062 ** (2.05)	0.0058 * (1.95)	0.0057 * (1.95)	0.0057 * (1.95)	0.0058 ** (2.00)
Lnhuman		−0.0072 *** (−3.26)	−0.0070 *** (−3.31)	−0.0066 *** (−3.27)	−0.0066 *** (−3.27)	−0.0066 *** (−3.36)
Lngover			−0.0178 *** (−4.67)	−0.0195 *** (−4.68)	−0.0201 *** (−4.71)	−0.0210 *** (−4.63)
Lnindus				−0.0100 *** (−3.08)	−0.0097 *** (−3.02)	0.0030 (0.41)
Finan					0.0005 (0.65)	−0.0001 (−0.13)
Open						0.0113 (1.51)
_cons	0.0403 *** (33.84)	0.0367 *** (18.04)	−0.0024 (−0.25)	−0.0154 (−1.23)	−0.0173 (−1.33)	−0.0158 (−1.28)
Time fixed effect	Y	Y	Y	Y	Y	Y
City fixed effect	Y	Y	Y	Y	Y	Y
N	4230	4230	4230	4230	4230	4230
R2	0.5875	0.5959	0.6094	0.6132	0.6133	0.6187

Notes: ***, **, * denote significance levels of 1%, 5% and 10% respectively.

. The sequential inclusion of control variables consistently yielded significantly positive coefficients for the smart city pilot policy (did) variable at a minimum significance level of 10%. The robust nature of these coefficient values underscored the pivotal role of the smart city pilot policies in enhancing urban resilience, thereby confirming Hypothesis H1.

Regarding the remaining control variables, the negative and significant coefficient of the human capital level (lnhuman) suggested that an augmentation in human capital might adversely affect urban resilience, potentially due to factors such as rising costs and substitution effects. Similarly, the negative and significant coefficient of the government intervention level (lngover) implied that excessive governmental involvement can, to some extent, diminish urban resilience. Conversely, the coefficient of the industrialization level (lnindus) displayed instability within the model and was potentially influenced by other control variables. Lastly, the coefficients of the financial development level (finan) and openness level (open) were insignificant, indicating a limited influence on urban resilience.

5.2. Parallel-Trend Test

While Hypothesis H1 was validated by the baseline regression results presented in Table 3, employing the multiperiod double-difference method necessitated adherence to a crucial precondition: the parallel-trend hypothesis test. This test ensured that both the treatment and control groups exhibited similar temporal trends prior to the implementation of the smart city pilot policies by the state. As a result, the base period for the parallel-trend test in this study is the year before the policy was introduced, and the results are shown in Figure 1. The figure reveals that the estimated coefficients of the policy dummy variables across all pre-implementation periods were statistically insignificant, thereby confirming the validity of the parallel-trend hypothesis. After implementation, the estimated coefficients of these policy dummy variables progressively escalated and unanimously surpassed the significance threshold, conclusively demonstrating that the surge in urban resilience levels was indeed attributed to the enactment of the smart city pilot policies. This finding further solidified the conclusion drawn from Hypothesis H1.

5.3. Robustness Tests

Given the validity of the parallel-trend test, this study endeavored to mitigate the potential estimation bias stemming from other confounding factors by conducting a comprehensive suite of robustness checks. These tests were designed to fortify the accuracy and reliability of the conclusions drawn.

(1)

PSM-DID test

Since there may have been non-randomness when the country determined the list of smart city pilots, this could potentially cause selective bias in the estimation results. For this reason, this study further adopted the PSM-DID estimation method to eliminate such problems. A 1:2 caliper nearest-neighbor matching was used for the test; based on this, a kernel density plot could be used to check whether there was a difference between the two groups of trend scores before and after the matching. If the kernel density curves between the two groups were relatively far apart before matching and if the kernel density curves were relatively close after matching, this indicated that the matching effect was good. The results of the matching are shown in Figure 2.

As depicted in Figure 2, prior to matching, the deviation between the two kernel density curves was notably pronounced. However, subsequent to matching, the distance between their respective mean lines diminished, and the curves converged closer together. This observation provided substantial evidence that the application of cross-sectional propensity score matching (PSM) effectively mitigated the sample selectivity bias by generating a discernible treatment effect.

The estimation results obtained after applying the PSM-DID methodology are presented in column (3) of

Table 4. Benchmark regression and cross-section PSM-DID results.

	(1)	(2)	(3)	(4)
	ols	Fe	on_support	sever the tail
did	0.0109 ***	0.0058 **	0.0051 *	0.0041 *
	(5.6773)	(2.0044)	(1.8460)	(1.84)
lnhuman	0.0146 ***	−0.0066 ***	−0.0068 ***	−0.0042 ***
	(16.4367)	(−3.3597)	(−3.0936)	(−2.72)
lngover	−0.0267 ***	−0.0210 ***	−0.0214 ***	−0.0152 ***
	(−12.6822)	(−4.6256)	(−4.8626)	(−4.99)
lnindus	0.0654 ***	0.0030	−0.0002	−0.0073 **
	(14.5997)	(0.4114)	(−0.0360)	(−2.39)
finan	0.0162 ***	−0.0001	−0.0002	0.0003
	(18.4126)	(−0.1267)	(−0.3540)	(0.50)
open	0.0620 ***	0.0113	0.0089	−0.0007
	(20.1150)	(1.5106)	(1.4709)	(−0.34)
N	4230	4230	4151	4146
Adj. R2	0.4472	0.6169	0.6212	0.6855

Notes: ***, **, * denote significance levels of 1%, 5% and 10% respectively.

. A careful examination of these results reveals that the magnitude of the coefficient associated with the core explanatory variable, namely the smart city pilot policy DID, remained largely unchanged from that reported in column (2). Additionally, the coefficients of the remaining control variables aligned well with our initial expectations. This concordance emphasized the robustness of the baseline regression findings, even when considering potential issues related to selection bias.

(2)

Tail-break test

To reduce the potential impact of outliers on the benchmark regression, this study applied a 1% winsorization treatment to the sample data and then re-estimated the model. The results of this analysis are shown in column (4) of Table 4. Despite a slight decrease in the estimated coefficients of the policy variables, the results continue to indicate that the smart city pilot policies significantly increased the level of urban resilience. This consistency underscored the robustness of the findings related to Hypothesis H1.

5.4. Analysis of Impact Mechanisms

Based on the aforementioned results, it was evident that the smart city pilot policy exerted a notable positive impact on the enhancement of urban resilience. However, to gain a deeper understanding of the underlying mechanisms and pathways through which this policy influenced urban resilience, a further analysis was required. To this end, the present study adopted industrial structure upgrading as a mediating variable and employed a mediating effect model. This approach aimed to ascertain whether the smart city pilot policies fostered urban resilience by facilitating industrial structure upgrading. The outcomes of this analysis are presented in

Table 5. Results 1 of Sobel’s mediating effect test.

	UR	Lnstruc	UR
did	0.0108 *** (5.60)	0.0139 * (1.92)	0.0101 *** (5.35)
lnhuman	0.0146 *** (16.44)	0.0150 *** (4.49)	0.0140 *** (15.88)
lngover	−0.0265 *** (−12.53)	0.0209 *** (2.62)	−0.0274 *** (−13.14)
lnindus	0.0658 *** (14.60)	−0.4072 *** (−24.00)	0.0839 *** (17.72)
finan	0.0162 *** (18.36)	0.0323 *** (9.73)	0.0147 *** (16.78)
open	0.0622 *** (20.11)	0.5147 *** (44.21)	0.0393 *** (10.65)
lnstruc			0.0445 *** (11.03)
_cons	−0.0166 *** (−2.69)	−1.0764 *** (−46.37)	0.0313 *** (4.19)
Observations	4224	4224	4224
R-squared	0.4472	0.8487	0.4627
Adjusted R-squared	0.4464	0.8485	0.4618

Notes: *** and * denote significance levels of 1% and 10% respectively.

,

Table 6. Results 2 of Sobel’s mediating effect test.

Term	Total Effect c	a	b	Intermediary Effect a × b	C′ Direct Effect	Test Conclusion
X => M => Y	0.0108 ***	0.0139 ***	0.0445 ***	0.0006	0.0101 ***	intermediary

Notes: *** denotes significance levels of 1% respectively.

and

Table 7. Results of the Bootstrap mediating effect test.

Effects	Ratio	Standard Error	Z	95% Confidence Interval
Direct effect	0.0101 ***	0.0022	4.56	[0.0058, 0.0145]
Indirect effect	0.0006 *	0.0003	1.94	[0.0000, 0.0012]

Notes: *** and * denote significance levels of 1% and 10% respectively.

.

Based on the test results, the mediation model (9) revealed that regression coefficient c was significantly positive at the 1% level, affirming that the smart city pilot policies effectively promoted urban resilience. In the mediation model (10), regression coefficient a was also significantly positive at the 1% level, suggesting that the policies drove industrial structure upgrading. Furthermore, in the model (11), regression coefficients b and c′ were both significantly positive at the 1% level. Here, c′ represents the direct effect of the smart city pilot policies on the enhancement of urban resilience, while a×b denotes the indirect effect, which was valued at 0.0006. Notably, the direct effect accounted for 93.52% of the total effect, with the mediating effect that took place through industrial structure upgrading contributing 6.48% of the total impact.

The Bootstrap test results presented in Table 7 corroborate the findings of the Sobel test, with the values of both direct and indirect effects aligning. Consequently, it was validated that Hypothesis H2 held true, signifying that the smart city pilot policies contributed to the enhancement of urban resilience by fostering industrial structure upgrading.

5.5. Heterogeneity Analysis

Based on the above analysis, it was shown that smart city pilot policies can enhance urban resilience through industrial structure upgrading. However, China is a vast country with great interregional variability and heterogeneity in the promotive effect. For this reason, this study analyzed heterogeneity from two perspectives: city distribution and city size.

Regarding urban distribution, the grouping was based on the geographical regions of the east, center, and west. As for city size, the classification followed the guidelines outlined in the “Circular on Adjusting the Standard for the Division of City Size” issued by the State Council. Specifically, cities were categorized as follows: megacities with a permanent urban population of 10 million inhabitants or more, large cities with 5-to-10-million inhabitants, Type I large cities with 3-to-5-million inhabitants, and Type II large cities with 1-to-3-million inhabitants, with the remainder being classified as small- and medium-sized cities. Based on this classification, a grouping regression analysis was conducted, yielding the heterogeneity test results presented in

Table 8. Heterogeneity test results.

	(1)	(2)	(3)	(4)	(5)
	Eastern part	Middle part	Western part	Large cities	Small- or medium-sized cities
Did	0.0163 ***	0.0028	0.0020	0.0048	0.0019 **
	(2.66)	(0.80)	(0.43)	(0.82)	(2.02)
lnhuman	−0.0156 **	−0.0024	−0.0020	−0.0126	0.0005
	(−2.64)	(−1.29)	(−0.83)	(−1.56)	(1.00)
lngover	−0.0314 ***	−0.0101 **	−0.0192 ***	−0.0324 ***	−0.0042 ***
	(−2.72)	(−2.55)	(−2.72)	(−3.11)	(−3.70)
lnindus	0.0130	−0.0061	−0.0111 *	0.0217	−0.0005
	(0.95)	(−1.32)	(−1.77)	(1.17)	(−0.33)
finan	−0.0025	0.0002	−0.0000	−0.0010	−0.0003
	(−0.89)	(0.65)	(−0.02)	(−0.33)	(−1.31)
open	0.0275 *	−0.0031	−0.0026	0.0365 **	−0.0015
	(1.93)	(−1.25)	(−0.67)	(2.06)	(−1.18)
N	1500	1485	1245	1485	2745
Adj. R2	0.6665	0.7130	0.6083	0.6861	0.8836

Notes: ***, **, * denote significance levels of 1%, 5% and 10% respectively.

.

The analysis revealed regional disparities in the impacts of smart city pilot policies on urban resilience in terms of city distribution. Specifically, the estimated coefficient for the eastern region was significantly positive, while those for the central and western regions, though positive, were not statistically significant. This pattern suggested a promotive effect that followed the gradient of “east > central > west”, which was potentially attributable to the eastern region’s higher economic development and superior infrastructure, enabling more direct financial support and infrastructural safeguards for the implementation of smart city policies. A detailed examination of the underlying causes of the relative backwardness of the central and western regions is required. With regard to the issue of governance, it is possible that the central and western regions are experiencing difficulties in the planning and implementation of smart city policies. This may be associated with the orientation of policy, the distribution of resources, and the capacity for implementation at the local government level. In regard to infrastructure, the construction of smart cities is contingent upon the enhancement of network infrastructure. However, the central and western regions may exhibit deficiencies in network infrastructure development, including gaps in broadband penetration, network speed, and network coverage when compared to the eastern regions. In terms of resource constraints, the construction of smart cities requires a substantial capital investment, including infrastructure construction, technology research and development, and the introduction of skilled personnel. However, the central and western regions may exhibit relatively low levels of economic development and financial income, which constrains the financial investment in smart city construction. Furthermore, central and western cities confront additional challenges. The construction of smart cities necessitates cross-sector and cross-industry data sharing and utilization. Nevertheless, the central and western regions may encounter heightened difficulties in data sharing and utilization, including a lack of uniformity in data standards, low data quality, and imperfect data-sharing mechanisms.

Regarding city size, the estimated coefficients indicated distinct effects. Large cities exhibited a coefficient of 0.0048, while small- and medium-sized cities had a coefficient of 0.0019, with the latter being significant at the 5% level. This suggested that smart city pilot policies may have had a more pronounced impact in small- and medium-sized cities, and this impact was potentially due to their greater potential for improvement and the visibility of progress. However, the promotive effect in larger cities, though not as statistically significant, still indicated a positive contribution.

In summary, it could be verified that Hypothesis H3 held: there was regional- and city-scale heterogeneity in the level of enhancement of urban resilience by smart city pilot policies.

6. Conclusions and Recommendations

Drawing upon panel data spanning from 2006 to 2020, encompassing 282 prefecture-level cities in China, this study delved into the influence and underlying mechanisms of smart city pilot policies on urban resilience. The research findings obtained by employing a multiperiod double-difference model, mediating effect model, and group regression analysis can be summarized as follows.

(1)

Smart city pilot policies positively enhance urban resilience. Our regression analysis utilizing the multiperiod double-difference model revealed significant and stable estimated coefficients for the policies, which remained robust even after the inclusion of additional control variables, parallel-trend tests, and a series of robustness checks.

(2)

Industrial structure upgrading serves as a crucial pathway for smart city pilot policies to bolster urban resilience. The mediating effect model, which was corroborated by both Sobel and Bootstrap tests, underscored the significant and partially mediating role of industrial structure upgrading. Specifically, this mediating effect accounted for 6.48% of the total impact of smart city policies on urban resilience.

(3)

The effectiveness of smart city pilot policies in elevating urban resilience exhibited regional and city-scale heterogeneity. Geographically, the policies’ impact on urban resilience followed a gradient of “east > central > west”, highlighting regional disparities. As for city size, the policies tended to have a more pronounced and statistically significant effect in small- and medium-sized cities.

Drawing upon the aforementioned conclusions, the following recommendations are proffered.

(1)

Expanding the reach of smart city pilots: Given the substantial enhancement of urban resilience through smart city pilot policies, the government ought to augment its support for these initiatives and consider broadening their scope to encompass more cities. This can be achieved through strategic policy guidance, increased financial investments, and technical assistance, thereby facilitating the widespread adoption and implementation of smart city infrastructure and practices. At the same time, we should take the policy guidance role of the government in the Chinese context, the leading role of the market, and the characteristics of voluntary public services of non-governmental organizations into account [44]. They work together to promote urban resilience.

(2)

Reinforcing the mediating role of industrial structure upgrading: The government should devise policies aimed at fostering the transformation and upgrading of traditional industries toward high-tech, high-value-added sectors. Additionally, it should strategically plan industrial layouts, foster synergy between upstream and downstream industries within the value chain, cultivate industrial clusters, and, thereby, strengthen the pivotal role that industrial structure upgrading plays in enhancing urban resilience.

(3)

Implementing differentiated policies for different regions and city sizes: Given that the eastern region already has a better foundation for smart city construction and urban resilience enhancement, it can play a leading role in promoting coordinated development among regions. We should consider giving more autonomy to the market, and while the government is decentralizing, it can further increase innovation and explore more applications and demonstrations of cutting-edge technologies. Conversely, the central and western regions, which may lag behind in infrastructure and other aspects, should receive increased government support and be encouraged to collaborate with the eastern regions for joint development. As for city sizes, given the more pronounced effects and greater development potential of smart city initiatives in small- and medium-sized cities, the government should prioritize their needs and provide flexible policy measures tailored to their unique circumstances. The government can also build a platform for inter-city information interaction and resource-allocation optimization.