Smart cities and sustainable development

ABSTRACT The last two decades have witnessed a surge in interest in the smart–sustainable city, but it remains unclear how smart cities can achieve the multifaceted goals of sustainable development, such as ecological performance, human development and sustainability efficiency. This study performs a fuzzy-set qualitative comparative analysis (fsQCA) of 35 Chinese large smart cities through configurational theorizing of the smart–sustainable city nexus. The findings not only reveal the causal complexity of constructing smart cities to achieve urban sustainability, but also develop a taxonomy of smart city configurations leading to ecological performance or human development, namely, duplex-centric, eco-centric, human-centric and double-bind modes.


INTRODUCTION
As is becoming the 21st-century urban norm, the shift towards the smart-sustainable city has attracted great debate among multilateral bodies (e.g., urban developers, administrators and academia) (Appio et al., 2019;Florida et al., 2017;Kitchin & Moore-Cherry, 2020;Mora et al., 2017;Nicholds et al., 2017). While the excessive emphasis on the hardware side of efforts (internet communication and technology infrastructure) during the early stage of smart city construction is being criticized (e.g., Mora et al., 2017), a more systematic view is embraced. The integrative view advocates for technology connectivity, responsiveness to people, economic attractiveness, social responsibility and environmental stewardship in urban development (Vanolo, 2014;Zuzul, 2019).
Despite its insightfulness, the new smart urbanization confounds means and ends, with the underlying causal complexity not well-articulated. To begin, is the smartness of a city a requisite for sustainable development? The experience of smart city development often argues a large amount of economic investment and infrastructure upgrading but deliberately avoids the negative impact of such an investment on its surroundings (cf. Martin et al., 2018;Simmons et al., 2018). Second, can smart cities achieve many, even contradictory, goals of sustainable development concurrently, such as ecological performance and human development (World Commission on Environment Development (WCED), 1987)? Third, if multiple elements must work together to lead to a smart city, how do cities configure these elements to achieve sustainability or reconcile their multiple goals? To recapitulate, the fundamental theoretical question remains: How are smart city elements configured to achieve or hinder urban sustainability?
To address this overarching research question, we adopt a configurational approach. This offers advantages for untangling causal complexity where 'multiple explanatory factors combine in complex and at times contradictory ways', and where 'equifinality, that is, multiple alternative paths to an outcome' exists (Furnari et al., 2021, p. 778;Misangyi et al., 2017). To realize this configurational approach, we follow a scoping-linking-naming configurational theorizing process that Furnari et al. (2021) recommend. That is, we identify a set of smart city elements (e.g., smart people, living, economy, governance, supporting and infrastructure) that can be orchestrated together to lead to urban sustainability outcomes. We then explore how these smart city elements, as a relatively loose coupling structure, can interdependently and asymmetrically lead to an outcome and articulate how multiple configuration patterns can achieve or hinder the same outcome (cf. Fiss, 2011;Misangyi et al., 2017). We further label configurations to evoke an input-output analysis, namely ecological performance and human development as two fundamental mechanisms, for urban sustainability, and propose a taxonomy of smart-sustainable cities (duplex-centric, eco-centric, human-centric and doublebind modes).
In the context of China's smart urbanization, we employ the fuzzy-set qualitative comparative analysis (fsQCA) technique. Based on the Seventh National Population Census, 1 the urbanization rate of the population exceeded 63.89% in 2021, which indicates that China has shifted from a rural society to an urban one. To accelerate this transformation, China's central government launched a national strategy of 'New Urbanization', emphasizing five key goals for cities: they must be green, smart, innovative, human and compact (Cao et al., 2021;Li & Wu, 2018; PricewaterhouseCoopers (PwC) & China Development Research Foundation (CDRF), 2018). FsQCA is well-suited for assessing small to medium-sized samples, which is the case in studying China's smart cities. To date, only large cities, such as directly administered municipalities, provincial capitals and major coastal cities, have been able to pursue smart city attempts in China (PwC & CDRF, 2018;United Nations Development Programme (UNDP), 2016, 2019).

Towards a smart-sustainable city nexus
Sustainable cities focus on solving problems related to urban sustainability and overall liveability (Krueger & Gibbs, 2008;Satterthwaite, 1997). In some studies, the term 'smart cities' can be used interchangeably with 'sustainable cities' since sustainability and liveability are embodied in the definition of smart cities (e.g., Albino et al., 2015). Nonetheless, other studies view the smart-sustainability city nexus as a means-to-end chain, arguing that smart cities provide digital solutions that may aid in resolving critical societal problems and improve the quality of life for citizens (Appio et al., 2019;Bawany & Shamsi, 2015;Hollands, 2008). On the one hand, smart cities use information and communication technologies (ICTs) to manage transportation, energy, housing, health and the environment. This helps alleviate transportation congestion, improves public safety, optimizes and customizes services, collects data on population health, and places a premium on education and continual training (Dirks & Keeling, 2009). On the other hand, smart cities facilitate technological innovation and socio-economic change by bolstering soft infrastructures such as knowledge networks, voluntary organizations and a safe, crime-free environment (Appio et al., 2019).
However, the development of smart cities is not without costs, and it also brings new risks and challenges. Wachsmuth et al. (2016) highlight a slew of costs associated with networking and computer equipment, as well as the environmental impact of sustaining smart cities. Appio et al. (2019) warn of the risks of implementing smart cities initiatives; for example, managing data and algorithms may itself induce crimes and other social problems. Martin et al.'s (2018) comprehensive review of the smart-sustainable city in Europe and North America identifies five tensions between smart cities and sustainable development goals: (1) unsustainable economic growth with waste and resource consumption; (2) the unequal distribution of the benefits of digital innovation; (3) a weakening and marginalization of civil rights; (4) a disregard for environmental protection; and (5) the challenge to consumerism. Given the benefits and hazards associated with smart city construction, their impact on sustainable development warrants additional investigation.
2.2. Untangling the causal complexity of the smart-sustainable city through configurational theorizing To untangle the causal complexity of the smart-sustainable city, we adopt Furnari et al.'s (2021) model of configurational theorizing, which consists of a three-stage iterative process: scoping, linking and naming. Scoping identifies how relevant attributes may form configurations from both theory and insights about the phenomenon. Linking focuses on how attributes connect by thinking conjunctively, equifinally and asymmetrically. Naming involves labelling the individual configurations to identify their overall, higher level themes. Such an analytical model not only identifies multiple configurations that can produce the same outcome but also summarizes them into overarching patterns (Furnari et al., 2021).
Our study starts with the scoping stage with a comprehensive literature review and identifies 11 possible smart city frameworks in prior studies (for a summary, see Table 1). Inspired by Furnari et al.'s (2021) approach, we complexify from anchors of each smart city framework (namely one or more attributes of smart city that are important to explaining sustainability of the city), decompose them into factors and indicators, and examine the plausible coherence between them.
Overall, all the existing smart city frameworks simplify factors and indicators to higher order anchors (cf. Furnari et al., 2021); yet the properties of anchors, factors or indicators still need scrutinizing. For instance, Giffinger et al. (2007) suggest that sunshine hours are a proper indicator of natural attractiveness, but such a factor can hardly be planned or designed as an element of a smart city. 'Place', too, such as via buildings, parks, ravines, mountains and waterfronts, may not be easily separated from other anchors such as infrastructure, collaboration ecosystem and applications (Bedford et al., 2011).
Based on scoping and linking, our study employs a modified version of the Smart City Development Model proposed by the Chinese Academy of Social Sciences (CASS) (2016) as a framework to configure the smart city for three reasons. First, the elements of this framework are comprehensive, and each can be triangulated by other smart city frameworks (Bawany & Shamsi, 2015;Bedford et al., 2011;Giffinger et al., 2007;Kumar et al., 2020;Lee et al., 2014). The modified framework encompasses both the 'hard' (e.g., technology and infrastructure) and 'soft' (e.g., human capital and social innovation) smart city strategies (Angelidou, 2015;Appio et al., 2019). Second, all the Smart cities and sustainable development 723 indicators are means-oriented and focus on the planning, installation, usage, operation, and control of ICT-enabled services and devices, with the only exception of resource consumption. As a modification, we exclude this from smart economy because resource consumption is a by-product of a smart city layout. The last and most important reason is its relevance to Chinese smart city research. The development of Chinese smart cities is (1) technologyfocused, with all the elements embodying the ICT applications, and (2) government-centric with extensive online services and data management overseen by the government (e.g., Hu & Zheng, 2021). 2 In summary, our smart city framework consists of six elements: smart people, living, economy, governance, supporting and infrastructure.

Smart people
The first set of smart city elements for a city to be sustainable is the knowledge workers and managers in ICT-based industries, e-skilled consumers and digital public administrators. The aim is to improve the life quality of citizens without incurring extra ecological inputs (Andreani et al., 2019). From the supply side perspective, these 'smart people' provide the necessary human and social capital to materialize innovative ideas for building sustainable cities (Toppeta, 2010). From the demand side, it is the smart people who propose creative demands and urge a city to produce the services and products they need. In our framework, smart people entail ICT personnel, mobile internet usage and the online shopping index, all of which point to the human capital Table 1. Summary of smart city frameworks.

Macro-level framework
Smart City Characteristics (Giffinger et al., 2007) Smart economy, smart people, smart governance, smart mobility, smart environment, smart living City I-COA Model a (Bedford et al., 2011) Place, infrastructure, collaboration ecosystem, applications, life Community Smart City Dimensions (Lee et al., 2014) Urban openness, service innovation, partnership formation, urban proactiveness, smart city infrastructure innovation, smart city governance City Smart City Architecture (Bawany & Shamsi, 2015) and capability that support ICT-based industry, infrastructure and business models (CASS, 2016).

Smart living
A set of components focuses on citizen-centred smart living, including initiatives to improve housing, health, education and social inclusion (e.g., Appio et al., 2019;Giffinger et al., 2007). Two featured components of smart living in the information society are the development of third-party platform services and open data systems that make citizens' lives more convenient (CASS, 2016). Appio et al. (2019) conceive smart living as the culmination of all other elements. Smart living is critical to attracting and retaining smart people by making local life fun and engaging (Florida, 2014). Thus, smart living encompasses a range of healthy lifestyles that produce innovative ideas, influence market demand and secure long-term outcomes.

Smart economy
This element refers to the competitive business environments and innovative industries that contribute to sustainable cities. Our study constructs smart economy through ICT industry development and internet commercial applications, both of which are critical to promoting economic growth and quality of life. ICT applications upgrade traditional industrial sectors and breed strategic emergent industries, accompanied by an outbreak of business model innovation and entrepreneurship (e.g., Leydesdorff & Deakin, 2011). Smart economy not only relies on smart people from both the supply and demand sides but also ensures a convenient and enjoyable lifestyle for these participants (Luger & Goldstein, 1991). Likewise, smart economy interacts with governing bodies in regulating new business models. Thus, the effectiveness of smart economy must be interdependent with other smart city elements for urban sustainability.

Smart governance
Inspired by prior studies (e.g., Tokoro, 2015), smart governance can be defined as the general capacity of government entities to manage and coordinate stakeholders in the smart city configuration through service application integration. We follow CASS (2016) to capture smart governance by three indicators: online government service, public resource sharing and transparency. Ideally, city administrations implement cutting-edge ICT applications, process all the information flow between stakeholders and provide stakeholders with one-stop, integrative service solutions (Tokoro, 2015). Yet recent studies point to the hierarchical model in China's smart city initiatives that highlights the proactive role of government and the scarcity of systematic stakeholder engagement (Hu & Zheng, 2021), and this may impede sustainable urbanization.

Smart supporting
Following CASS's (2016) terminology, smart supporting refers to the planning, evaluation, and information security management in the construction and operation of smart cities. Kumar et al. (2020) propose a framework of smart city transformation, entailing planning phase, physical infrastructure, ICT infrastructure and deploying smart solutions, which thus emphasizes government planning and policy design. Smart supporting provides an overarching coordinative framework for directing and connecting various stakeholders, as well as dynamically designing and adjusting smart city configuration for sustainability.
Historically, city administrations have demonstrated their capacity for planning, coordination and transformation potentials in smart city examples, such as San Francisco (National League of Cities (NLC), 2016) and Seoul (Lee et al., 2014).

Smart infrastructure
This element stresses the determinant role of ICT-based infrastructure in fostering creative ideas, boosting industry competitiveness and innovation, and facilitating the operations of smart governance (Andreani et al., 2019). Since early work has predominantly adopted a technocentric approach, some scholars have come to recognize that the 'hard' ICT applications cannot do without the 'soft' smart city strategy (Albino et al., 2015;Mora et al., 2017). This explains why some cities are data-poor, not only due to technical problems but also through shortfalls in government data institutions, private sector involvement and citizens who embrace new technologies (Acuto & Parnell, 2016). Therefore, without collaborations and monitoring mechanisms, smart infrastructure contributes to the riddle of sustainable city solutions. In summary, through the scoping and linking of the complex causality in urban smartness-sustainability nexus, it can be observed that the theoretical foundation of the smart city framework suggests interactions among the six elements. Vanolo (2014) argues that a smart city is path-dependent because of various institutional roots and existing infrastructure diversity among cities, which implies the coexistence of multiple but equally effective patterns for building sustainable smart cities. However, this working proposition has not yet been proven. Therefore, we propose the following: Proposition 1: Multiple configurations of smart city elements can be associated with the level of urban sustainability.

Smart city configuration, ecological performance and human development
While the six smart city elements work conjunctively to reach their promise, their underlying mechanisms towards urban sustainability remain a puzzle. One major source of causal complexity is the multifaceted goals and objectives of sustainable cities. The term 'sustainable development' can be traced back to the 1987 World Commission on Economic Development (WCED) report, positing that 'present needs can be met without compromising the ability of future generations to meet their own needs' (WCED, 1987, p. 43). When the concept of sustainability is applied to cities, sustainable cities must meet the social, Smart cities and sustainable development 725 economic and environmental goals inherent in human needs without transferring environmental costs to future generations (Satterthwaite, 1997). Recent discussions of coupled human and natural systems (CHANS) (Liu et al., 2007a(Liu et al., , 2007b) also shed new light on sustainable city construction by warning of the danger of vast human disturbances in irreversible unsustainability. To analyse the complexity of CHANS in the urban context, the UNDP (2016, 2019) proposes a concept of sustainability efficiency, namely the efficiency of converting ecological inputs into human development, as a key to sustainability. The central idea is that a city pursues human development, such as a long and healthy life, being knowledgeable, and having a decent standard of living, at the ecological cost of resource consumption and environmental pollution. 3 Drawing on insights from prior studies, our study decomposes the smart-sustainable city by analysing the effects of smart city configuration on sustainability efficiency by affecting ecological performance 4 and human development. As illustrated in Figure 1, human development strives to meet the present needs of human systems by increasing environmental, social and economic outputs; ecological performance concerns the needs of future generations by reducing resource consumption and pollutant discharge in the ecological systems; and sustainability efficiency simplifies coupled human and ecological systems into an input-output analysis.
While smart city elements promote economic performance by creating new market demand and accelerating service innovation, their complex social and ecological consequences remain under-explored (Acuto & Parnell, 2016;Martin et al., 2018;Viitanen & Kingston, 2014). A disproportionate emphasis on the talent and demands of smart people enlarges social inequality. Prevalent lowcarbon living initiatives may bring even more municipal waste and pollution. For example, urban bicycle-sharing programmes can create e-waste mountains; and greening initiatives can expel the poor from their original communities (Wachsmuth et al., 2016). In the case of smart economy, the rapid growth of industrial innovation and internet marketing exacerbate the consumerism trend, resulting in increased material consumption and environmental burdens (Viitanen & Kingston, 2014). Automation and artificial intelligence (AI) applications in a smart economy may deprive ordinary people of their jobs and worsen their living conditions. The risks that pertain to smart governance and supporting have also been noticed, such as ideological manipulation, a corporation of city governance and hackable networks (e.g., Appio et al., 2019). Furthermore, smart infrastructure has unintended consequences, including the creation of e-waste and increased global greenhouse gas emissions (Bawden, 2016). Therefore, to make cities economically viable, environmentally friendly and socially responsible, smart city elements must be carefully designed and configured based on a cost-benefit analysis. Taking together, we posit the following: Proposition 2: Smart city elements can be configured to achieve or hinder urban sustainability efficiency through reconciling ecological performance and human development mechanisms.

Sample and data
We constructed a sample of 35 Chinese large cities by matching two data sources. The first is the 2017-2018 Annual Greenbook for Sustainability of 35 Big Cities in China, which is a project of the UNDP of China (UNDP-China) (UNDP, 2019). The report calculates the China sustainable city index by applying two indices: Figure 1. A configurational theoretical framework of the smart-sustainable city nexus. the urban human development index (UHDI) and the urban ecological input index (UEII), for major 35 cities in 2017-18. It includes all the Chinese directly administered municipalities, sub-provincial cities and provincial capitals, except for Lhasa and Chinese Taipei due to the missing data. 5 Compared with prefecture-level non-capital cities, the sample cities face more population and economic growth pressures, have a more innovative economy with human capital, and provide better living conditions at the expense of resource consumption and environmental pollution. 6 They also display more strength and efforts to become sustainable by being smart.
The second data source is the China smart cities index (CSCI), which is released annually. 7 The series of reports collect smart city data for Chinese cities from three main sources: (1) China City Statistical Yearbooks, government bulletins and other archival data; (2) public information and data from websites of the Bureau of Statistics, Development and Reform Commission, Science and Technology Bureau, and Commission of Economy and Information Technology; and (3) telephone interviews with city governmental departments. However, to match our first data source to test the linkage between smart city configuration and their economic and sustainability consequences, we selected the CSCI 2016 because it has a one-year time lag.

Research design
Given the theoretical novelty of exploring smart city configuration and urban sustainability, this study exhibits a 'middle-way' research design between a purely deductive variable-oriented design and a purely inductive case-based design (Crilly, 2011;Fiss, 2011). QCA is best suited for this type of research as a synthetic strategy that integrates the strengths of variable-and case-based approaches (Ragin, 2008(Ragin, , 2013. This enables the exploration of causal configurations based on empirical cases. When designing fsQCA, the issue of limited diversity is taken into account (i.e., the logically possible causal combination -2 k possibilitiesexceeds the sample size). Given our sample of 35 Chinese comparable large cities, our smart city framework has six key elements, which exceeds the sample size (2 6 ¼ 64 possible causal combinations). Thus, to mitigate potentially violating the rule of limited diversity, we supplemented a truth table analysis.
The truth table analyses six outcomes (see Appendix A in the supplemental data online), showing 29 cases that are included, which is less than our sample size. In other words, our sample size covers all variations in forming configurations (a maximum of 29 empirical configurations, i.e., counting all case numbers that are equal to 1). Thus, we can safely conclude that our sample is free of the limited diversity issue (Ragin, 2008).

Measures and calibration
Our six outcomes include high urban sustainability, ecological input, human development and their negations, respectively. As stated, we measure urban sustainability by sustainability efficiency and collect data from UNDP (2019). UNDP (2019) adopts a slacks-based measure (SBM) of efficiency in the data envelope analysis (DEA) to calculate sustainability efficiency (Charnes et al., 1978;Tone, 2001). This method takes the UEII as inputs and the UHDI as outputs with input and output matrices (Charnes et al., 1978;Tone, 2001). The production possibility set, P, is defined as P = {(x, y)|x ≥ X l, y ≥ Y l, l ≥ 0}, where l is a non-negative vector in R n ; an expression is introduced for describing a sample city (x 0 , y 0 ) as x 0 = X l + s − and y 0 = Y l + s + , with l ≥ 0, s − ≥ 0 and s + ≥ 0. The vector s − [ R m and s + [ R s indicate the input excess and output shortfall of the expression and are called slacks. For estimation, the fraction is formulated as follows: Here the UEII is also our measure of urban ecological inputs, which is composed of two sets of sub-indices. The first set is the urban resource consumption index (URCI), comprising water consumption (per capita building area), land consumption (per capita standard coal) and energy consumption (per capita water supply). The second set is the urban pollutant discharge index (UPDI), which includes water pollutants (per capita chemical oxygen emission demand and ammonia nitrogen emissions), air pollutants (sulfur dioxide and nitrogen oxide emissions), and solid waste (industrial solid waste production and domestic garbage disposal).
Urban human development, measured by the UHDI, is an average of (1) the environmental output index (ENOI) (per capita green areas); (2) the social output index (SOI) based on a comprehensive education index (mean years of schooling, expected years of schooling and educational public service), medical index (life expectancy and medical public service), and an urban-rural income gap index (reverse-coded); and (3) the economic output index (ECOI) (per capita gross domestic production).
All six smart city elements were adapted from CASS (2016): . Smart people is measured by the capability of the city inhabitants (1) to work for ICT industries (the percentage of employees working in ICT industries); (2) to make use of ICT facilities (the usage of mobile internet); and (3) to have consumption over the internet (an online shopping index). . Smart living is based on two indicators. One is the extent to which the third-party platforms (e.g., Alibaba, WeChat) are allowed to provide services and make citizens' lives more convenient. This is measured by the Tecent Research Institute's internet smart city and Alipay's city service index. The other is the development levels of the open data system that make available the information for the public. . Smart economy is made up of ICT industry development and the internet in commercial applications. ICT industry development is proxied by the percentage of Smart cities and sustainable development 727 ICT industries' production in civic gross domestic product (GDP). Internet commercial application is calculated by integrating AliResearch's e-commerce index and Tecent Research Institute's China internet + index. . Smart governance is calculated based on three indicators: (1) the service quality of online government (the construction and application of the city government's online service platforms that deal with examination, approvals and other handling issues); (2) the accessibility and sharing level of public resources (the construction of the shared public resources trading platform); and (3) the transparency of government affairs information and public opinion monitoring (the government's usage of social media for information disclosure and public communication). . Smart supporting points to the capability of the city government in planning, evaluation, and information security management in the construction and operation of smart cities. Planning capability is indicated by the completeness and feasibility of overall smart city planning, as well as the instructiveness and operationality of smart city construction guidelines. Evaluation capability is represented in the completeness and effectiveness of organizational structure and performance appraisal in smart city construction. Through information security, we investigate whether and to what extent a focal city formulates policies, rules, and legislation to meet security and privacy requirements, and whether a major information security accident has occurred. . Smart infrastructure is represented in network construction, information resource sharing and i-cloud platform application. Network construction is measured by internet broadband connectivity, public wireless local area network (WLAN) coverage and internet of things (IoT) development. Information resource-sharing concerns the degree to which the city government shares the public information of four basic databasespopulation, legal person, spatial geography and macroeconomicsand other government affairs. The i-cloud platform application refers to the capacity of i-cloud platform construction at the city level to realize information-sharing across industries and domains.
Calibration is the unique feature in applying fsQCA. In contrast to regression methods, where variables are measured on either raw values or sample-specific scales, fuzzy sets in fsQCA are calibrated using external criteria or a data distribution approach, and not all variations are important (Crilly, 2011;Fiss, 2011;Ragin, 2008). Our datasets are constructed according to archival data, which generally lack external standards to facilitate the calibration. We thus mainly adopt the data distribution (75th-50th-25th percentiles) approach in calibrating to the three key breakpoints of full membership, the crossover point and full non-membership (Fiss, 2011;Ragin, 2008). Given this calibration rationale, we adopted the direct method to calibrate outcomes and causal conditions by fsQCA3.1b. Details of each measure and calibration are summarized in Table 2. 8

Necessity analysis
We began by analysing the necessary conditions listed in Table 3. The purpose of this analysis is to determine whether any of the causal conditions met the definition of a necessary requirement for high sustainability efficiency, ecological performance or human development.
The results indicate that none of the individual circumstances exceeded the required consistency criterion of 0.90 (Cui et al., 2020). As a unique advantage, fsQCA enables us to investigate the factors that contribute to the negation of the outcome separately (Campbell et al., 2016;Rihoux & Ragin, 2009). We also generated negation results for three outcomes using the fsQCA3.1b negation function, in order to investigate configurations that result in low sustainability efficiency, ecological performance or human development. Again, none of the individual requirements is deemed required (i.e., > 0.90).

Sufficiency analyses
The Quine-McCluskey algorithm (the method of prime implicants) provides a deterministic way to check whether the minimal form of a Boolean function has been reached. Tables 4-6 show the results of the fuzzy-set analysis of the linkages between the configurations of smart city elements and high sustainability efficiency, ecological performance and human development, respectively. Configurations for the negation outcomes of sustainability efficiency, ecological performance and human development 9 are also displayed. The configurational solutions are presented in the style following the recommendation of Fiss (2011) (see also Douglas et al., 2020;Huang & Fan, 2021), that is, reporting both parsimonious (i.e., core conditions) and intermediate solutions (i.e., peripheral conditions). In notation, black circles ( ) indicate the presence of a condition, and circled crosses ( ) indicate its absence. Large circles indicate core conditions, while small circles mean peripheral conditions. Blank spaces indicate ambiguous situations in which the corresponding causal condition may be either present or absent, and therefore this plays an unimportant role in the configurational solution (Douglas et al., 2020;Huang & Fan, 2021). Grouped by core conditions, fsQCA returns 18 solutions for all six outcomes (high sustainability efficiency, ecological performance, human development and their negations).
Following three analytical criteria, we adopted a consistency threshold of 0.88 for high sustainability efficiency solutions and 0.86 for high sustainability efficiency's negation results, a threshold of 0.87 for high ecological performance solutions, 0.84 for its negation and a threshold of 0.89 for both high human development solutions, and its negation. First, we used the truth table algorithm to identify attribute combinations consistently linked to an outcome above the 728 Yiyi Su and Di Fan acceptable consistency benchmark of 0.80 (Douglas et al., 2020;Ragin, 2008). Cui et al. (2020) suggest choosing a threshold that corresponds to a gap observed in the distribution of raw consistency scores. However, across all six outcomes, we only observed a clear gap between 0.81 and 0.87 for high ecological performance. Second, we adopted alternative consistency measurements to make a further judgmentthe proportional reduction in inconsistency (PRI) valueas a measure of fit, to compute the degree to which a solution is sufficient for the outcome rather than a negation of the outcome (Ragin, 2008). We adopted a more stringent PRI cut-off score (0.75) to rule out

Toward a taxonomy of the smart-sustainable city
In Tables 4-6, each column represents a distinct smart city configuration for sustainability efficiency, ecological inputs and human development. To facilitate interpretation, we identify representative cases associated with each solution from the truth tables 10 and automatically generated by fsQCA3.1b, where cases with a membership > 0.5 imply substantive case knowledge (Cui et al., 2020). All information on the case examples was gathered from publicly accessible archival sources (e.g., UNDP, 2019; CSCI, local government websites, media interviews with mayors, news, and investment and industrial reports). The primary analysis was involved examining the sufficiency of the identified smart city elements for observing sample cities that achieve high or low sustainability efficiency. As panels 1 and 2 of Table 4 show, three smart city configurations achieve high sustainable efficiency, and four configurations for its negation. The finding confirms Proposition 1. One key conclusion drawn from the data analysis is that the smartness of a city does not guarantee urban sustainability. Admittedly, smart city configurations that achieve high urban sustainability (i.e., S1-S3) present more smart city elements than some less sustainable counterparts (i.e., S4 and S5), but do less than others (S6 and S7). Proposition 2 identifies two fundamental underlying mechanisms for the combination of smart city elements for urban sustainability. The first mechanism is to reduce ecological inputs via smart city configurations for facilitating future generations to meet their own demands. The second mechanism is to mobilize smart city elements for comprehensive human development to meet the needs of the present, namely environmental, social and economic outputs. Our input-output analysis in Tables 5 and 6 not only supports Proposition 2 but also reveals the complex and varied interactions between the two mechanisms meeting the present needs and preserving the ecological environments for the future. As panels 1, 3 and 5 indicate, Changsha belongs to the high sustainability efficiency solutions (S1) and is also an exemplar city of both high ecological performance (S10) and human development (S14). Therefore, it displays a reinforcing effect of the two mechanisms as to how a smart-sustainable city reduces ecological inputs, on the one hand, and realizes high human development, on the other. As for other representative cities in panel 1 for high sustainability efficiency (e.g., Dalian, Tianjin and Harbin), they are also evident in panel 3 for high ecological performance, suggesting how high ecological performance facilitates urban sustainability.
Concerning low urban sustainability, although Xiamen shares a configurational solution with Changsha (S14) for high human development, it is featured with low ecological performance (S13) and classified in low sustainability efficiency configuration (S7). In this sense, its human development is neutralized by a large number of ecological costs, leading to a smart but unsustainable city. In contrast, Ningbo's low sustainability efficiency (S6) is largely attributable to its level of human development, as evidenced by its low human development (S16b). Other causes of low sustainability efficiency (S4 and S5) do not appear in Tables 5 and 6, indicating that they reach neither ecological performance nor human development; thus, their conversion of ecological inputs into human development is inefficient.
Beyond propositions testing, we follow Furnari et al.'s (2021) naming stage heuristics to articulate with simplicity by verbalizing the linkages among elements of configurations, crafting an overarching narrative across configurations, and labelling each configuration to evoke the essence. In particular, we summarize the configurational recipes, features and cases of a taxonomy of the smart-sustainable city in Table 7 and propose a taxonomy of four smart-city modes based on their purpose(s), namely, duplex-centric, eco-centric, human-centric and double bind modes.
The duplex-centric mode represents smart city configurations that balance ecological performance and human development, as indicated by common configurational solutions of panels 3 and 5. Based on S10 and S14, this mode is featured with (1) the absence of smart infrastructure as a core condition for high ecological performance; (2) smart economy, governance and support as core conditions for high human development; and (3) smart people without smart living as periphery conditions for both high ecological performance and human development. Thus, it falls into a high-geared mode with more than three smart city elements present for sustaining human development, but is equipped with light smart infrastructure for desirable ecological performance.
A representative case is Changsha. Changsha is devoted to smart city construction at all levels of government, but its smart infrastructure is relatively weak, ranking it 28th out of 35 sample cities (CASS, 2016). Although Changsha made efforts in building up an ecocity and increase the quality of living between 2011 and 2017, for example, life expectancy increased from 76.26 to 77.59 years, and the number of hospital beds per 10,000 persons increased from 60.58 to 74.39, the overall smart living was still ranked 17th of all 35 cities (CASS, 2016;China City Statistical Yearbooks). Changsha has recently transformed from a model of 'low ecological inputs, low human development' to a 'low ecological inputs, high human development' model (UNDP, 2019). On the one hand, the local government places a premium on luring high-tech enterprises to Changsha, which not only increases ICT-related jobs but also promotes a smart economy. Local government, on the other hand, emphasizes the integration of all e-service platforms and the overall   , 2018). While Changsha lacks smart living and infrastructure in comparison with other cities of comparable size, it has developed a more conventional approach urban sustainability that is highly relied on smart people and supporting.
Eco-centric mode refers to configurational solutions in which smart city elements are arranged to achieve high ecological performance at the expense of human development. Panel 3 displays three major paths of smart cities to pursue high ecological performance solely. The first path is people-oriented, as indicated by S8a and S8b, with the presence of smart people (a core condition) and the absence of smart living, economy and support. The second path is living-based, relying heavily on smart living without smart people as two core conditions (S9a-c). The first two paths can further be summarized as a low-geared smart city with the presence of either smart people or living as a single core condition. Differently, the third path is economy-supported and medium-geared, that is, smart economy with high-quality institutions as two core conditions (represented by smart governance in S10 or smart supporting in S11) but light investment in smart infrastructure.
To illustrate, we detail S8b with Tianjin as the representative city. Tianjin was accorded recognition for 'low ecological inputs' by the UNDP (2019). Archival data and news reports have attributed Tianjin's sustainability achievements to its unique practices in smart city construction. 12 First, Tianjin took a pre-emptive move in smart city initiatives. It proposed 'creating a smart city' in early 2015, and since 2017 it has been hosting the World Intelligence Conferences (WICs) annually for the purpose of aggregating advanced technology, capital and talent. For instance, during the first WIC, Tianjin struck a strategic partnership with Ant Financial Services Group to establish Northern China's first 'cashless society'. Second, Tianjin leverages intelligent manufacturing capabilities to modernize its conventional manufacturing base and transportation hub. Its initiatives have been incorporated into a 'smart' Belt and Road Initiative. Third, Tianjin partners with technological giants (e.g., Huawei, Bosch) to provide smart e-government platforms. In so doing, government officials are equipped with ICT-based analytical tools to deal with social issues.
Human-centric mode refers to smart city elements that work together to maximize human performance while minimizing ecological costs imposed on future generations. This mode corresponds to the configurational solutions in panel 5 but not panel 3. The characteristics of this mode are twofold: (1) the existence of three smart city elements as core conditions and one or two smart city elements as peripheral conditions for advancing human development; and (2) the presence of smart infrastructure for low ecological performance or high human development. This pattern is exemplified by two sets of representative cities. One set is represented by Xiamen, whose heavy investment in smart infrastructure significantly differs from that of Changsha in the same configurational solution of S14. As a consequence, Xiamen and Changsha are listed in S10 and S13, respectively, the two extremes of ecological performance. Therefore, Xiamen falls into the human-centric mode, that is, to promote human development at considerable ecological costs.
The other set of representative cities, indicated by S15a and S15b, exhibits milder smart city patterns, pursuing development goals with ecological goals unattained. Our Table 6. Configurational solutions of human development.

Smart city elements
Panel 5: High human development Panel 6: Low human development S14 S15a S15b S16a S16b S17 S18 results indicate that these cities prioritize human development over ecological sustainability when configuring smart city elements. Take Beijing for example: this megacity is making strenuous efforts to establish a worldwide renowned smart city. Although attracting highly skilled global talents is not an issue, its rapid urban development is accompanied by a series of ecological challenges, such as resource depletion, air pollution, water shortages, among others. Similar patterns are also evident in international megacities, such as San Francisco and London, where consumption-based carbon counts double per capita emissions of standard calculations (Wachsmuth et al., 2016). Double bind mode describes the category of configurational solutions that fail in either ecological performance or human development, or do not outperform in both (i.e., a 'lose-lose' situation). Although all sampled cities have undertaken projects aimed at changing them into smart cities, several have failed to meet either ecological (S12) or development (S16a and S16b) goals or inefficiently converting ecological inputs into human development (S4 and S5). (3) Economy-supported: (a) smart economy with high-quality institutions as core conditions; and (b) light investment in smart infrastructure as a core condition Dalian (S1, S10) HSE Fuzhou (S11, S18)

Humancentric
Smart city elements combine to achieve high human performance while disregarding the ecological costs imposed on future generations Full-geared smart city with heavy smart infrastructure: (a) three smart city elements as core conditions and one or two smart city elements as peripheral conditions for HHD; and (b) the presence of smart infrastructure for LEP or HHD Xiamen (S7, S13, S14) LSE Shenzhen/ Guangzhou (S15a) Beijing/Shanghai (S15b) Double bind A configuration of smart city elements achieves neither human development nor ecological performance I. Low-geared smart city without a human focus: (a) the presence of one smart element as a core condition; and (b) the absence of smart people or smart living Xining/Taiyuan (S4) LSE Lanzhou/Hohhot (S5) LSE Guiyang/Kunming (S12) II. Medium-geared smart city with both human and economy foci: (a) the presence of smart people and the economy as core conditions; and (b) the absence of smart supporting as a core condition Haikou (S16a) Ningbo (S6, S16b) LSE Note: HSE, high sustainability efficiency; LSE, low sustainability efficiency; HHD, high human development; LHD, low human development; HEP, high ecological performance; and LEP, low ecological performance. Superscript acronyms (HSE and LSE) are based on the results shown in Table 4. A full-geared smart city means at least the presence of three smart city elements, a medium-geared one indicates two elements and a low-geared one suggests less than one element as core conditions. We detect some salient distinctions between the two types of solutions. Indicated by Xining and other cities, S4, S5 and S12 constitute a low-geared smart city configuration in which the presence of a single smart city element as the core condition and the absence of smart people or smart living. These solutions not only lack a human focus as lowgeared eco-centric mode but also have insufficient smart city elements to satisfy the needs of human development.
Meanwhile, S16a and S16b develop a medium-geared smart city with the presence of smart people and economy and the absence of smart supporting as three core conditions. Compared with the human-centric mode, the two solutions lack a full-geared and carefully designed system of smart city elements. Two representative case cities are Haikou and Ningbo. Taking Ningbo as an example to reflect on its experience of constructing smart city projects, one local government official pointed out that lacking a thorough design and planning at earlier stages of a smart city project caused serious financial and environmental losses after the project was completed; operating and maintaining a smart city project also requires substantive investment and detailed management (e.g., Wu, 2013). Thus, despite the fact that investing in technology infrastructure is a major driver of urban economic development in China, insufficient coordination between multiple government agencies and a lack of long-term planning jeopardizes the city's future (Deloitte, 2018).

Theoretical contributions
Our study makes three theoretical contributions. First and foremost, it contributes to the smart city literature (e.g., Martin et al., 2018;Simmons et al., 2018) by incorporating an input-output analysis in recent sustainability research (Liu et al., 2007a(Liu et al., , 2007bUNDP, 2016UNDP, , 2019. Our results demonstrate the numerous ways in which cities configure smart city elements to realize or fail urban sustainability via ecological performance and/or human development. That is, some successfully exploit two mechanisms (namely duplex-centric), some heavily rely on one while ignoring the other (i.e., eco-or human-centric), and yet others abandon both, resulting in unsustainable development (i.e., double-bind). Moreover, departing from previous studies that consider sustainability as the inner nature of smart cities (Albino et al., 2015), our results infer that sustainability is a choice. In other words, we cast doubt on the tendency toward boosting smart city construction by placing an excessive emphasis on human development outcomes rather than on ecological performance.
Second, while prior smart city studies propose several influential smart city frameworks (e.g., Giffinger et al., 2007), the linkages among the elements remain unclear. By closely adhering to Furnari et al.'s (2021) configurational theorizing approach, our study elucidates the conjunctive, equifinal and asymmetric features of smart city elements in a detailed and delicate way. Concerning conjunction, we do observe the co-occurrence of smart city elements for multifaceted sustainability goals, suggesting the importance of combinational thinking to match one smart city element with another. As summarized in Table 7, more smart city elements are configured for human development than ecological performance; smart economy needs to be complemented by well-functioning institutions and light infrastructure investment for realizing high ecological performance (S10 and S11); and at least three smart city elements work together as core conditions to promote human development (e.g., the presence of smart economy, governance and supporting as core conditions in S14, and the presence of smart economy, supporting and smart infrastructure as core conditions in S15a and S15b).
For the sake of equifinality, our study reveals that different configurations of smart city elements can equally and effectively induce or hinder sustainability efficiency, ecological performance and human development. We find distinguished medium-geared smart city configurations leading to high sustainability efficiency (S1 and S3) from both low-(S4 and S5) and medium-(S6 and S7) geared ones, resulting in low sustainability efficiency. We also document three substitutive paths to high ecological performance (people-oriented, living-based and economy-supported) along with variant high-geared smart city configurations to high human development (S14, S15a and S15b). All these findings point to the patterned multiplicity of the smart city configurations that bring about desirable sustainable outcomes (e.g., Appio et al., 2019).
Noteworthy is the asymmetric role of smart city elements in achieving urban sustainability. One prominent element is smart infrastructure that hinders ecological performance: lacking infrastructure in less developed cities aggregates ecological problems (S12), but its presence coupled with other smart city elements imposes too high a burden on our ecological environments (S13). We also observe the absence of smart city elements for achieving ecological performance (S8a, S8b and S9a-c), which challenges the extant literature that considers numerous elements in smart city design (e.g., Giffinger et al., 2007).
Last, we find the presence of smart supporting for high human development and the lack of it for the negation outcome (all as core conditions), which echoes the recent paradigm change in both academic and policy arenas (Appio et al., 2019;Trencher, 2019). Smart city 2.0 has been argued to supplant the previous generation of smart city practices (Appio et al., 2019). Rather than focusing exclusively on economic benefits, this modern version of the smart city emphasizes a systematic approach that includes all inhabitants 'as co-creators or contributors to innovation, problem-solving and planning' (Trencher, 2019, p. 118). The academic enquiry has shifted from a technocentric and universalist approach to a design-driven and human-centric approach (Andreani et al., 2019). The central importance of smart supporting, as our results show, signifies that a fundamental change is underway.

Policy implications
Our findings also have several implications for practice. First, our smart city framework can be used to analyse the present development of the smart-sustainable city. A Smart cities and sustainable development policymaker can use the assessment tool to make decisions relevant to smart cities. Second, the key insight of the study is that smart city configurations can be built up and, more importantly, configured across several key causal conditions. Multiple configurations achieve high sustainability efficiency by low ecological inputs and/or high human development. In other words, there are multiple 'optimal' smart city modes. Third, government officials need to be wary of three modes (eco-centric, human-centric and double bind modes) that all have room for improvement. A careful trade-off between ecological performance and human development is of central importance to implement a successful smart city strategy.

Limitations and future research
The context of this study necessarily imposes limitations on the generalizability of its findings. Future studies are encouraged to address these limitations. First, while we frame smart city configuration primarily from the means-end view of the smart-sustainable city, prior studies have also developed the entrepreneurial and smart city modes that emphasize interactions among a multilateral set of partners or stakeholders (Adner, 2017;Ooms et al., 2020;Stam, 2015). Future studies can study smart city modes by analysing different actors.
Second, due to the limited availability of matching data, the study investigates 35 sampled cities only. Such a design strengthens internal validity by ruling out the external confounding factors other than smart city configuration (e.g., population pressure and economic growth rate), but limits its external validity, that is, the applicability of the framework among the smaller scaled prefectural-and county-level cities need further exploration. Also, our truth table analysis (see Appendix A in the supplemental data online) suggests that the limited diversity issue is not a concern, and fsQCA is particularly suitable for small sample-sized studies (10-50 cases) (Ragin, 2008). Future studies can extend the sample size to test the generalizability of findings.
Third, we only have one wave of a multiple sourced dataset with a lagged structure. When more cities engage in smart city initiatives and more years of sustainability assessments become available, future research can extend the length of the periodical variance to validate the findings of our study.
Lastly, our study is limited by a single-country research context based on CSCI. The advantage of such a design is that culture and institutions are inherently controlled, but we are unable to explore the exogenous impact on smart city configuration and its consequences. Future studies can explore how smart city configuration affects urban sustainability in urban areas across countries or a global context.

CONCLUSIONS
By leveraging the fsQCA technique, our study disentangles causal complexity of the smart-sustainable city nexus by a configurational approach. Reconceptualizing a smart city framework (including smart people, living, economy, governance, supporting and infrastructure), we investigate how smart city configurations affect sustainability efficiency by reconciling ecological performance and human development in 35 cities in China. We confirm the coexistence of varied but equally effective configurations for a smart-sustainable city. Based on an input-output analysis of sustainability efficiency, a taxonomy of smart city configuration is proposed: duplex-centric, eco-centric, human-centric and double bind modes. The study concludes with a discussion of research and practical implications. detailed analysis is available from the authors upon request. 7. See https://www.govmade.cn/. 8. More details of measures and calibrations are available from the authors upon request. 9. To ease the interpretation, we mark the negation results of 'high' sustainability efficiency, ecological performance and human development as 'low' outcomes, rather than 'non-high', as a more accurate term (Ragin, 2008). 10. The tables are available from the authors upon request. 11. See the Changsha Municipal Government website (http://zwfw-new.hunan.gov.cn/portal/index). 12. For news and reports, see http://www.echinagov.com and http://www.most.gov.cn/.