THE IMAPCT OF DIGITAL INFRASTRUCTURE DEVELOPMENT ON URBAN ADVANCED SERVICE INDUSTRY ENTREPRENEURSHIP ATTRACTION IN CHINA A Thesis Presented to the Faculty of the Graduate School of Cornell University In Partial Fulfillment of the Requirements for the Degree of Master of Science in Regional Science by Qirun Jiao August 2025 © 2025 Qirun Jiao ABSTRACT This study estimates the causal impact of digital infrastructure development on attracting advanced service industry entrepreneurs, leveraging the staggered implementation of the 'Broadband China' policy as a quasi-experiment. Applying a multi- period Difference-in-Differences (DID) model, we find that the policy significantly enhanced the inflow of these entrepreneurs. Our analysis provides a nuanced perspective by identifying key mediating pathways: expanding the digital market size, fostering entrepreneurial space agglomeration, and enhancing urban innovation capacities. These findings empirically justify continued digital infrastructure investment, underscoring its critical role in regional economic upgrading and offering valuable empirical insights for policy development. Keywords: Digital Infrastructure, Broadband China Policy, Advanced Service Industry, Entrepreneurs, Talent Attraction. iv BIOGRAPHICAL SKETCH Qirun Jiao was born in Ningbo, Zhejiang Province, China. He earned a Bachelor of Science in Economics and a Bachelor of Arts in Geography (Data Science) from the University of Washington before pursuing a Master of Science in Regional Science at Cornell University. His research interests include digital economy, spatial demography and migration. v To the beginning of a journey. My first serious piece of academic work—may it not be the last vi ACKNOWLEDGEMENTS I would like to express my heartfelt gratitude to everyone who supported me throughout the writing of this thesis. I am especially thankful to Professor John Carruthers and Professor Ding Fei for their academic guidance and encouragement during my graduate studies. Their insights and support have been invaluable to my development. To my mother, thank you for your unwavering support and comforting words, especially during moments of frustration. Your steady presence has been a source of strength and reassurance. To my roommate, Moheng Ma—thank you for being there whenever I faced difficulties. Your help and companionship have meant a great deal, and I truly value the friendship we have built along the way. To Zheng Yu, thank you for your patience in explaining the concepts I found challenging. Your clarity and consistent support played an important role throughout this journey. Lastly, to my girlfriend, Qiuyi Rao—thank you for standing by me during the most stressful and monotonous stages of writing. Your companionship, emotional support, and encouragement carried me through the toughest times. With you, I’ve always felt safe to share whatever’s on my mind, and that has meant more to me than words can express. vii TABLE OF CONTENTS 1. Introduction ......................................................................................................... 1 1.1 Research Background ................................................................................ 1 1.2 Research Questions .................................................................................... 2 1.3 Structure of the Study ................................................................................ 4 2. Literature Review................................................................................................ 5 2.1 Theoretical Framework .............................................................................. 5 2.2 Definition and Measurement of DI ............................................................ 9 2.3 Determinants of Urban Entrepreneurial Attractiveness ........................... 10 2.4 Relationship Between DI and ASI Entrepreneurial Attractiveness.......... 13 3. Hypothesis Development .................................................................................. 15 3.1 Direct Effect ............................................................................................. 16 3.2 Indirect Effects ......................................................................................... 17 4. Empirical Strategy and Data ............................................................................. 23 4.1 Data .......................................................................................................... 23 4.2 Model Specification ................................................................................. 24 5. Main Results ..................................................................................................... 27 5.1 Baseline Results ....................................................................................... 27 5.2 Parallel Trends Test .................................................................................. 28 5.3 Placebo Test ............................................................................................. 30 5.4 Robustness Checks................................................................................... 33 viii 6. Endogeneity Analysis........................................................................................ 37 7. Mechanism Analysis ......................................................................................... 39 8. Conclusion ........................................................................................................ 42 8.1 Contributions............................................................................................ 43 8.2 Limitations ............................................................................................... 44 ix LIST OF FIGURES Figure 1: Pilot Cities of the “Broadband China” Policy ....................................... 26 Figure 2: Parallel Trends Test ............................................................................... 30 Figure 3: Kernel Density Distribution of the β₃ Coefficient ................................. 32 x LIST OF TABLES Table 1: Descriptive Statistics of Principal Variables ........................................... 26 Table 2: Baseline Regression Results ................................................................... 28 Table 3: Robustness Check Results ...................................................................... 33 Table 4: PSM-DID Results ................................................................................... 37 Table 5: IV Estimation Results ............................................................................. 39 Table 6: Path-Analysis Coefficients by Mediator ................................................. 42 xi LIST OF ABBREVIATIONS Abbreviation Full Term DI Digital Infrastructure ASI Advanced Service Industry UIE Urban Innovation Ecosystem NEG New Economic Geography 1 1. Introduction 1.1 Research Background Digital Infrastructure (DI) constitutes the foundational architecture of the contemporary global economy and serves as a critical driver of high-quality economic development (Nambisan, 2017). Defined as an open, dynamic socio-technical system built upon Information and Communication Technologies (ICT), DI integrates broadband networks, cloud computing, big data analytics, artificial intelligence, and the Internet of Things (IoT) to enhance efficiency, resource sharing, and innovation (Hanseth & Lyytinen, 2010). Its scope extends beyond physical connectivity to include sophisticated application platforms and supportive innovation ecosystems (Autio et al., 2018), collectively facilitating seamless information flows critical for diverse digital services and operations. Recognizing the strategic importance of DI, policymakers globally have increasingly prioritized its development as a fundamental pillar supporting national competitiveness, economic expansion, and improved public services (Sang et al., 2025). Prominent initiatives such as the United States' "National Broadband Plan" (Federal Communications Commission, 2010), the European Union’s "Digital Decade" (European Commission, 2021), Japan’s "Society 5.0" vision (Government of Japan, 2016), and Singapore’s "Smart Nation" strategy (Smart Nation Singapore, 2014) illustrate this international commitment. Similarly, China's "Broadband China" initiative, has demonstrated substantial positive impacts across domains including rural digital finance (Niu et al., 2022), employment in 2 service sectors (Ndubuisi et al., 2021), and urban innovation capabilities (Ben Arfi & Hikkerova, 2021). Amid the accelerated digital transformation driven by enhanced DI, entrepreneurship has emerged as a pivotal element for economic growth with entrepreneurial activities increasingly shaped by digital technologies and digital ecosystems (D’Angelo et al., 2024). However, as economies strive to surpass the middle-income threshold, the focus has gradually shifted toward fostering high value-added entrepreneurial opportunities, predominantly found within Advanced Service Industries (ASIs). These industries, integral to modern urban economies, are typically embodied as Knowledge-Intensive Business Services (KIBS) and encompass sectors such as finance, law, research and development (R&D), information technology (IT), and consulting (Muller & Doloreux, 2009). ASIs are characterized by their reliance on specialized knowledge, skilled human capital, customized solutions, client co-production, and their central role within broader innovation systems (Santos, 2019; Zieba, 2013). Consequently, ASIs significantly contribute to urban growth, competitiveness, and long-term economic sustainability. 1.2 Research Questions Despite substantial attention paid to DI's economic implications, existing literature predominantly investigates traditional sectors, such as manufacturing (Xie et al., 2024), retail (Hänninen et al., 2018), agriculture (Salemink et al., 2017), and general small- and medium-sized enterprises (Müller et al., 2018). Xie (2024), for example, specifically 3 explored DI’s impact on business model innovation within manufacturing Small and Medium-sized Enterprises (SMEs), highlighting sector-specific digital transformation processes, yet neglecting considerations pertinent to ASIs. Similarly, Hänninen et al. (2018) provided insights into retail SMEs, emphasizing digital integration within conventional commerce settings. These industry-specific analyses, while insightful, inadequately reflect the unique operational dynamics and requirements inherent in advanced, knowledge-driven service sectors. In contrast, ASIs exhibit distinctive operational features, marked by high dependence on skilled labor, intricate information exchanges, extensive communication networks, and necessity for specialized knowledge and global market accessibility (Zieba, 2013), rendering them particularly sensitive to the quality and availability of DI. Thus, robust DI directly facilitates critical functions within ASIs, such as efficient information exchange, knowledge dissemination, professional networking, and global market integration (Melville & Kohli, 2021). Given the strategic importance attributed to DI in regional policy frameworks for attracting ASI entrepreneurs, an in-depth understanding of the specific mechanisms underlying this relationship is imperative. This study, therefore, aims to fill this critical research gap by examining how urban DI influences the attraction and growth of entrepreneurship within ASIs. Specifically, it integrates the innovation ecosystem perspective (Autio et al., 2018), the theory of knowledge spillovers (Acs et al., 2013), and 4 theories of economies of scale and agglomeration from the New Economic Geography (NEG) literature (Glaeser et al., 1992; Krugman, 1991). Building on these foundations, the study investigates how DI contributes to expanding market size, fostering enterprise agglomeration, and optimizing the innovation environment. Accordingly, the research seeks to address the following core questions: (1) Does DI development influence cities’ ability to attract entrepreneurial talent in ASIs? (2) Through what specific mechanisms do DI shape this attraction? 1.3 Structure of the Study The structure of this paper is organized as follows: Section 1 provides an introduction, outlining the background, research question, and overall structure of the study. Section 2 offers a comprehensive literature review, synthesizing theoretical insights from studies on DI, innovation ecosystems, and NEG, while identifying gaps in current research. Section 3 develops hypotheses regarding both the direct and indirect effects of DI on entrepreneurial attraction. Section 4 describes the research design, including data sources, variable construction, and model specification. Section 5 presents empirical analysis, reporting estimation results and conducting robustness checks. Section 6 addresses potential endogeneity concerns through instrumental variables. Section 7 investigates the mediating mechanisms through which DI shapes entrepreneurial dynamics. Finally, Section 8 concludes the paper by summarizing the key findings, offering policy implications, and admitting limitations. 5 2. Literature Review 2.1 Theoretical Framework 2.1.1 Urban Innovation Ecosystem (UIE) 2.1.1.1 UIE: Theoretical Foundations and Core Components The concept of an UIE emerges from systematic views of innovation and entrepreneurship (Stam and van de Ven, 2021), integrating insights from regional innovation systems, the Triple Helix model, and entrepreneurial ecosystems. Specifically emphasizing urban contexts, the UIE framework conceptualizes cities as geographically bounded yet dynamically evolving networks characterized by intensive interactions among diverse actors, institutions, resources, and infrastructures (Foguesatto et al., 2023). UIE thus brings together various actors and critical resources within a city to facilitate innovation. Core actors include universities and research institutions (Etzkowitz and Leydesdorff, 2000), firms such as startups and KIBS (Kumar et al., 2020), government agencies, financial institutions, support organizations like incubators (Appio et al., 2019), and skilled individuals. Additionally, citizens increasingly participate as active co-creators within the innovation process. UIEs are also shaped by enabling factors such as knowledge flows, availability of capital, infrastructure (particularly DI), and supportive local policies (Băzăvan, 2019). Cultural factors—including openness to innovation, risk-taking attitudes, and collaborative norms—further play a significant role (Arz and Kuckertz, 2019). Through continuous and 6 dynamic interactions among these components, UIEs foster entrepreneurship and sustain innovation within urban environments. 2.1.1.2 UIE Framework as an Analytical Lens To examine how DI influences the attractiveness of cities to entrepreneurs in ASIs, the UIE framework offers a particularly valuable analytical lens. It conceptualizes the city not merely as a collection of resources, but as a dynamic ecosystem in which DI interacts synergistically with talent pools, knowledge assets, financial institutions, policy frameworks, support organizations, and local cultural factors. From this systemic perspective, the impact of DI is understood as multifaceted—shaping entrepreneurial attraction not only by enhancing connectivity, but also by improving access to resources, strengthening network interactions, and fostering a supportive ecosystem environment. Applying the UIE framework thus enables a more nuanced and comprehensive exploration of the multiple pathways through which DI contributes to innovation and entrepreneurial growth in urban settings, particularly within knowledge-intensive service sectors (Stam and van de Ven, 2021). 2.1.2 New Economic Geography (NEG) NEG emerged in the early 1990s to explain the spatial concentration of economic activity and the formation of agglomerations such as cities and industrial clusters (Fujita and Krugman, 2004). Developed by Krugman and his collaborators, NEG integrates monopolistic competition, increasing returns to scale, and transport costs into general 7 equilibrium models. This marks a departure from traditional location theory, which emphasized only natural advantages (Fujita et al., 2001). Instead, NEG focuses on “second nature” geography, wherein economic activity becomes self-reinforcing through cumulative processes rather than relying solely on geographic endowments (Redding, 2010). 2.1.2.1 Economies of Scale A foundational mechanism in NEG is firm-level increasing returns to scale (IRS), which helps explain why firms tend to cluster spatially (Krugman, 1991; Fujita et al., 2001). When firms face high fixed production costs and declining average costs, as output increases, there is a strong incentive to concentrate on production in fewer locations to achieve scale efficiencies. NEG commonly employs the Dixit-Stiglitz model of monopolistic competition to formalize IRS under product differentiation, allowing multiple firms to coexist while benefiting from large-scale production (Fujita et al., 2001). 2.1.2.2 Spatial Agglomeration NEG explains spatial agglomeration as the result of centripetal forces outweighing centrifugal forces. Centripetal forces include IRS, lower input costs from supplier proximity, and greater market accessibility. In contrast, centrifugal forces include congestion, land scarcity, and intensified local competition (Fujita et al., 2001). Forward linkages, or demand-side effects, describe how firms benefit from locating near large consumer markets, where proximity lowers transport costs and increases potential sales 8 (Krugman, 1991; Fujita et al., 1999; Redding, 2010). Backward linkages, or supply-side effects, arise when firms co-locate with suppliers of intermediate goods or specialized services, thus reducing input costs and improving supply quality (Fujita et al., 2001). These linkages reinforce agglomeration through cumulative causation: firm concentration attracts labor, which in turn expands local markets and input availability, further reinforcing initial clustering. These mechanisms align with Marshallian externalities such as labor pooling, supplier specialization, and knowledge spillovers, which enhance innovation and productivity in dense urban environments (Marshall, 2013). Although early NEG models paid limited attention to knowledge dynamics, subsequent extensions recognized the importance of informal information flows and tacit knowledge exchange enabled by spatial proximity— particularly in high-tech and knowledge-intensive sectors. 2.1.2.3 Agglomeration in the Digital Economy Despite the rise of DI, NEG remains relevant in explaining persistent urban clustering. Although digital technologies reduce the cost of transmitting information, agglomeration persists in advanced service sectors where face-to-face interaction, access to localized talent, and proximity to financial institutions remain critical (Koster et al., 2019; Nayyar et al., 2021). DI complements rather than replaces traditional agglomeration forces. For example, broadband access may enhance the attractiveness of peripheral areas, but it does not fully substitute for the innovation advantages offered by dense urban environments 9 (Koster and Thisse, 2024). Thus, the NEG framework remains essential for understanding the spatial dynamics of entrepreneurship and the role of DI in shaping urban economic development (Krugman, 1991). 2.2 Definition and Measurement of DI The concept of DI has evolved alongside technological advancement, becoming increasingly expansive and multidimensional in scope. Initially, DI was primarily conceived as physical communication facilities—such as broadband networks, fiber-optic cables, telecommunication towers, and data centers—highlighting its role in basic connectivity and interoperability (Henfridsson & Bygstad, 2013). With the advent of cloud computing, big data, and artificial intelligence, the definition has broadened to encompass technological platforms, data storage and processing capabilities, intelligent systems, and innovation-supporting environments, forming a comprehensive socio-technical system (Tilson et al., 2010). Contemporary scholarship now generally recognizes DI not merely as hardware-based connectivity, but as a complex ecosystem that underpins a wide range of economic and social activities through the integration and application of digital technologies (Autio et al., 2018). From a methodological standpoint, accurately identifying exogenous variation in DI development remains a persistent challenge. Given that infrastructure deployment is often closely correlated with pre-existing regional economic conditions, industrial structures, and governance capacities, using continuous measures—such as broadband penetration or 10 internet adoption rates—can introduce significant endogeneity bias. Consequently, recent studies increasingly rely on exogenous policy shocks or interventions as quasi-natural experiments to enable more credible causal identification of DI effects. 2.3 Determinants of Urban Entrepreneurial Attractiveness To better understand the mechanisms underlying urban entrepreneurial attractiveness— particularly in theoretical contexts predating the widespread diffusion of DI—this section first revisits traditional perspectives from location theory and NEG, before turning to recent transformations in the digital economy era. 2.3.1 Traditional Location Theory Perspective Traditional location theory and NEG have long argued that the spatial distribution of entrepreneurial activity is driven by the joint forces of economies of scale and agglomeration. According to scale economies theory, urban market expansion reduces transaction and service delivery costs, thereby increasing expected returns and improving the sustainability of entrepreneurial venture (Krugman, 1991). Larger local markets offer new firms a broader customer base while enabling the amortization of fixed costs across greater output. Agglomeration theory further posits that the concentration of firms and skilled labor within urban areas facilitates entrepreneurship by deepening labor market pools, allowing for the sharing of specialized inputs, and fostering knowledge spillovers (Audretsch & Feldman, 1996; Glaeser et al., 1992). Regional industrial diversity and spatial clustering 11 are thus widely regarded as critical factors for new firm formation and survival (Balland et al., 2015). Storper & Venables (2004) further emphasize that, even in an era of advanced information technology, face-to-face interaction and localized knowledge networks remain essential for knowledge-intensive entrepreneurship. From an innovation standpoint, research underscores that active knowledge creation and spillover processes—such as those emanating from research universities—are important precursors to entrepreneurial activity (Feldman, 2001). Empirical evidence from Acs & Armington (2004) further demonstrates that enhanced local innovation capabilities and elevated human capital levels substantially raise regional entrepreneurial activity and new firm birth rates. Overall, economies of scale, agglomeration effects, and regional innovation capabilities represent the three core mechanisms through which traditional theories explain urban entrepreneurial attractiveness and spatial variation in firm formation. 2.3.2 Digital Technology-Driven Transformation In the context of the digital economy, the foundational mechanisms shaping urban entrepreneurial attractiveness have undergone substantial transformation. As digital technologies advance rapidly, the influence of geographic proximity has diminished, while virtual connectivity, knowledge networks, and innovation ecosystems have gained growing importance (Nambisan, 2017). DI enhances entrepreneurs’ information accessibility and remote collaboration capabilities, creating a more flexible and efficient environment for 12 entrepreneurship. A growing body of empirical research confirms these shifts. Atasoy (2013), using county-level data from the United States, finds that broadband internet adoption significantly increases self-employment rates, particularly in knowledge-intensive industries, highlighting the critical role digital connectivity plays in fostering entrepreneurship. Oyinlola et al. (2022), analyzing Africa’s plastic value chains, show that digital innovation accelerates the development of local entrepreneurial ecosystems and improves value chain sustainability—especially in service-oriented and knowledge-driven industries. Delacroix et al. (2019), in their study of low-income female entrepreneurs in French towns, demonstrate that digital platforms such as Facebook effectively lower entrepreneurial barriers, increase participation among disadvantaged groups, and foster entrepreneurial ecosystem expansion and diversity in peripheral regions. Overall, the literature suggests that DI stimulates entrepreneurial activity through both direct and indirect mechanisms. Directly, it reduces informational and transactional friction; indirectly, it enhances knowledge diffusion, fosters innovation collaboration, and supports ecosystem development. In contrast to traditional theories that emphasize economies of scale and agglomeration, contemporary perspectives increasingly highlight the enabling role of DI and technological empowerment in shaping the spatial dynamics of entrepreneurship. 13 2.4 Relationship Between DI and ASI Entrepreneurial Attractiveness Building upon insights from traditional location theory and studies of digital transformation, this section identifies three critical mediation mechanisms related to DI, serving as theoretical foundations for subsequent empirical analysis. 2.4.1 DI as a Technological Foundation DI serves as critical foundational infrastructure within urban innovation and entrepreneurship ecosystems, directly enhancing convenience and feasibility of entrepreneurial activities in ASIs. The development of digital technologies significantly improved cross-regional collaboration and resource integration, thereby stimulating entrepreneurial dynamism in service-oriented sectors. Based on empirical evidence from innovative U.S. cities, Mack & Faggian (2013) find that broadband infrastructure adoption significantly increases regional labor productivity—particularly in cities characterized by high levels of human capital and concentrations of knowledge-intensive industries—thus creating favorable conditions for ASI development. Similarly, Malecki (2018) argues that communication infrastructure and digital connectivity are essential prerequisites for attracting highly skilled entrepreneurs and knowledge-based firms in the context of knowledge-driven urban development. As such, DI not only reduces transaction and communication costs but also functions as a technological foundation enabling innovation in ASI entrepreneurship. 14 2.4.2 Multidimensional Transmission Mechanisms Beyond direct effects, existing studies suggest DI also indirectly influences entrepreneurial attractiveness via intermediary mechanisms. Existing studies have identified multiple transmission pathways. Forman et al. (2012) show that internet diffusion significantly expands local markets and customer bases, driving regional economic vibrancy and entrepreneurial opportunities. Neumark et al. (2011), using U.S. enterprise level data, find that regional firm agglomeration enhances startup survival and growth through strengthened inter-firm interactions, resource sharing, and technological learning. Qian et al. (2013) highlight that local innovation systems improve the efficiency of knowledge production and diffusion, thereby strengthening entrepreneurial ecosystems through the dynamic integration of human capital and knowledge assets. Using provincial data from China, Du & Wang, (2024) show that DI promotes innovation indirectly—by accelerating knowledge flows and enhancing R&D collaboration—thereby amplifying entrepreneurial activity. Feldman (2001) notes that the attraction and concentration of high-skilled talent can indirectly promote entrepreneurship by increasing knowledge spillovers and improving opportunity recognition. Stam (2015) further underscores the importance of social network connectivity, institutional quality, effective governance, and public service delivery in enhancing the resilience and vibrancy of entrepreneurial ecosystems. Taken together, the literature suggests that market expansion, firm agglomeration, and innovation capacity enhancement represent the core mechanisms through which DI 15 influences urban entrepreneurial attractiveness. This study, therefore, centers on these three mediation pathways to systematically explore how DI enhances cities' appeal to ASI entrepreneurs. 2.4.3 Research Gap Although existing studies have explored various mechanisms linking DI and entrepreneurial attractiveness, limitations remain regarding ASI entrepreneurship. First, empirical research frequently targets overall entrepreneurial rates or all-industry startup formation rather than specifically examining ASI entrepreneurs. Second, studies commonly emphasize single mechanisms, lacking parallel examination or comparative analysis across multiple mechanisms. Addressing these gaps, this paper empirically tests three major mediation pathways—economies of scale, agglomeration effects, and innovation capabilities—specifically in the ASI context, providing a nuanced understanding of DI’s multifaceted role in shaping urban entrepreneurial attractiveness and enriching existing theoretical frameworks. 3. Hypothesis Development Building on the preceding literature review and theoretical foundations, this study conceptualizes DI as a foundational, open, and dynamically evolving socio-technical system that is profoundly reshaping the urban environment for innovation and entrepreneurship. Considering this perspective, the study proposes a set of hypotheses, including a direct effect hypothesis and three mediation hypotheses, which are detailed in 16 the following section. 3.1 Direct Effect Building on Nambisan’s seminal work (Nambisan, 2017), digital technologies fundamentally reshape entrepreneurial processes by transforming the dynamics of uncertainty and opportunity recognition. Specifically, DI serves as a critical enabler of entrepreneurship by adding entrepreneurial opportunities (Afawubo and Noglo, 2022), reducing transaction costs (Niebel, 2018), enhancing operational agility (Luo and Fan et al., 2012), and facilitating product customization (Marion and Fixson, 2021; Watson IV and Weaven et al., 2018). Additionally, for entrepreneurs operating within ASIs, where rapid information exchange, data analytics, and cross-border collaboration are paramount, access to high-speed broadband and integrated digital platforms significantly bolsters their entrepreneurial capacity (Dabbous and Barakat et al., 2023), mitigates uncertainty, and fosters a more predictable and efficient business environment (Afawubo and Noglo, 2022). These mechanisms collectively reduce entry barriers and enhance the perceived viability of entrepreneurial ventures in cities with advanced DI. Empirical research underscores that cities characterized by robust digital ecosystems exhibit heightened levels of entrepreneurial activity (D Angelo and Cavallo et al., 2024), particularly within knowledge-intensive sectors such as finance, law, scientific research, and management consulting. Building on these insights, we hypothesize: 17 H1: DI development has a significant direct positive effect on urban attractiveness for Advanced Service Industry entrepreneurs 3.2 Indirect Effects 3.2.1 Economies of Scale Mechanism Building on insights from NEG, which posits that market size expansion fosters entrepreneurial dynamism through agglomeration economies and reduced transaction costs(Glaeser et al., 1992; Krugman, 1991), this study theorizes that DI enhances urban attractiveness for entrepreneurs in ASIs by redefining and expanding digital market scale. As a foundational enabler of market growth, DI strengthens connectivity and facilitates digital entrepreneurship (Olan and Troise et al., 2024). The development of digital markets enables ASI entrepreneurs to overcome geographical constraints, granting access to global markets through e-commerce, remote collaboration, and platform ecosystems (Reuber and Fischer, 2011; Dabbous and Barakat et al., 2023). Moreover, digital platforms exhibit network externalities (Karhu and Heiskala et al., 2024), wherein user growth attracts service providers, generating self-reinforcing ecosystems that expand market opportunities. Additionally, entrepreneurs in ASIs rely on knowledge spillovers and scalable client bases and are particularly responsive to digital market scale expansion. Strengthened digital connectivity allows them to deliver services remotely, leverage global talent pools, and capitalize on data-driven opportunities (Zahra and Liu et al., 2023). Furthermore, digital markets generate digital assets—scalable and non-rivalrous data resources—that serve as 18 essential inputs for high-value service innovation (Giustiziero and Kretschmer et al., 2023). Consequently, cities with robust DI provide ASI entrepreneurs with dual advantage: access to expansive digital markets and the technological foundation necessary to transform data into competitive solutions. Thus, this thesis puts forward the following hypothesis: H2a: DI development positively drives urban attractiveness for advanced service industry entrepreneurs by expanding digital market size 3.2.2 Agglomeration Effect Mechanism The spatial agglomeration of entrepreneurial activity has long been recognized as a driver of regional economic vitality, rooted in Marshall’s theory of agglomeration economies (Marshall, 2013), which emphasizes knowledge spillovers, labor market pooling, and specialized supplier networks. In the digital era, this framework is further reinforced by DI, reshaping the spatial dynamics of entrepreneurship. First, DI reduces barriers to resource access and collaboration—critical for entrepreneurs in ASIs who rely heavily on rapid information exchange, data analytics, and cross-disciplinary innovation. High-speed broadband, cloud computing, and shared data platforms diminish geographic constraints on knowledge diffusion, enabling ASI entrepreneurs to leverage digital platforms, remote collaboration, and data-driven decision- making, thereby reinforcing spatial agglomeration effects (Ferraris and Santoro et al., 2020; Zahra and Liu et al., 2023). This perspective aligns with Bouncken and Kraus's (2022) 19 ecosystem model, wherein digital tools facilitate complementary resource integration among firms, fostering dense networks of specialized ASI ventures. For instance, data centers and 5G networks enable real-time collaboration among fintech startups, legal consultancies, and R&D labs, generating a self-reinforcing cycle of agglomeration. Second, DI supports hybrid entrepreneurial ecosystems that transcend physical boundaries while simultaneously intensifying localized synergies. Traditional agglomeration economies emphasized geographic proximity, whereas DI enables "virtual agglomeration" effects, allowing entrepreneurs to participate concurrently in local clusters and global digital communities. For example, smart city platforms enable ASI entrepreneurs to engage in co-creation initiatives with public institutions and multinational corporations, blending place-based advantages (e.g., urban innovation districts) with digital scalability (Ferraris and Santoro et al., 2020). This dual structure mitigates congestion costs associated with physical clustering while enhancing urban attractiveness by embedding entrepreneurs within both local and global value chains. Third, DI fosters knowledge spillover effects that drive innovation and talent agglomeration, creating a dynamic entrepreneurial ecosystem (Giustiziero et al., 2023). ASIs traditionally thrive on tacit knowledge exchange facilitated by face-to-face interactions. However, AI-driven analytics platforms now codify and disseminate expertise on unprecedented scales, generating "digital knowledge commons" that attract entrepreneurs seeking advanced insights (Martínez-Caro et al., 2020). This aligns with 20 Spigel and Harrison’s (2018) entrepreneurial ecosystem perspective, where digitally augmented networks strengthen interdependence among entrepreneurs, investors, and institutions. For example, cloud-based open innovation platforms in Silicon Valley-like ecosystems enable ASI ventures to rapidly prototype solutions using shared datasets, appealing to entrepreneurs who prioritize iterative learning and risk mitigation (Isenberg, 2010). Empirical research further substantiates this argument. Foguesatto et al. (2023) find that cities with robust DI exhibit higher entrepreneurial retention rates due to enhanced ecosystem connectivity. Similarly, Zahra et al. (2023) demonstrate that digital interoperability among firms in entrepreneurial ecosystems correlates with increased venture survival rates, particularly in knowledge-intensive sectors. These findings indicate that spatial agglomeration, when augmented by DI, effectively attracts ASI entrepreneurs who prioritize innovation networks over traditional factors such as cost minimization. Thus, this thesis puts forward the following hypothesis: H2b: DI development drives urban attractiveness for Advanced Service Industry entrepreneurs by enhancing the spatial agglomeration of entrepreneurship. 3.2.3 Urban Innovation Capability Mechanism The knowledge spillover theory of entrepreneurship (Audretsch, 1995; Acs et al., 2013) and the New Schumpeterian growth framework both identify knowledge recombination as a fundamental driver of innovation. ASIs are inherently knowledge-intensive, requiring 21 continuous access to specialized expertise, interdisciplinary collaboration, and rapid commercialization of novel ideas (Santos, 2019). By transforming the spatial and temporal dimensions of knowledge production and dissemination, DI significantly reshapes the innovation ecosystems in which ASI entrepreneurs operate. First, DI reduces the marginal cost of knowledge recombination—an essential mechanism underlying innovation (Breschi and Lissoni, 2001). High-speed broadband and cloud platforms facilitate real-time data exchange among ASI startups, research institutions, and established firms, accelerating the recombination of scientific, technical, and market knowledge (Ben Arfi and Hikkerova, 2021). Empirical studies suggests that cities with robust DI exhibit heightened entrepreneurial activity, driven by enhanced knowledge- sharing mechanisms, talent concentration, and ecosystem- oriented business models that foster innovation and collaboration (Hajli et al., 2025). Furthermore, responsive IT infrastructure directly strengthens technological innovation capabilities by promoting effective knowledge sharing (Cassia et al., 2024). Second, DI amplifies localized knowledge spillovers by overcoming traditional constraints on knowledge transmission. While conventional knowledge spillovers tend to be spatially bounded (Audretsch and Belitski, 2013), digital platforms create hybrid spaces where tacit knowledge is increasingly codified through iterative interactions (Marion and Fixson, 2021). Proximity to innovation districts equipped with advanced facilities such as 5G networks and data centers enables ASI entrepreneurs to leverage localized knowledge 22 flows—including tacit expertise from universities or collaborative pilot projects with municipal governments—while maintaining global connectivity (Caragliu and Del Bo, 2019). This hybrid localization-globalization dynamic enhances entrepreneurs' capacity to transform abstract knowledge into commercially viable innovations, a critical determinant of urban innovation capability (Acs et al., 2013). Third, DI facilitates data-driven innovation, offering a crucial competitive advantage for ASIs. The integration of IoT sensors, AI analytics, and interoperable data systems generates granular insights into market trends and consumer behavior, enabling entrepreneurs to identify niche opportunities (Grimaldi et al., 2025). For example, smart city platforms aggregate real-time urban data, empowering legal-tech or management consulting startups to deliver highly localized services. This aligns with findings that digital transformation enhances firms' innovation performance by enabling rapid experimentation and scalable growth (Wu et al., 2023). Notably, due to the public-good nature of DI, even smaller firms gain access to advanced technological tools, democratizing innovation resources (Ferraris and Santoro et al., 2020). Thus, the following hypothesis: H2c: DI development drives the urban attractiveness of advanced service industry entrepreneurs by upgrading their urban innovation capability. 23 4. Empirical Strategy and Data 4.1 Data This study utilizes the "Broadband China" initiative as a quasi-natural experiment. Officially launched by the Chinese central government, this national policy aims to foster the coordinated development of regional broadband infrastructure, accelerate the optimization of existing networks, and enhance the overall level of broadband application across the country. The policy was implemented through the selection of designated pilot cities in three distinct batches, announced in 2014, 2015, and 2016, respectively. In total, 117 cities at the prefecture level or above were designated as "Broadband China" pilot zones. The official list identifying these selected cities was obtained from the website of the Ministry of Industry and Information Technology of the People's Republic of China. To investigate the policy's impact, this research utilizes individual-level data drawn from the China Migrants Dynamic Survey (CMDS) for the period spanning 2011 to 2018. The CMDS is a large-scale, nationally representative annual survey that collects comprehensive information on China's internal migrant population. The survey data relevant to this study includes detailed records on the basic demographic characteristics of individual migrants and their family members, their migration scope and trajectory, education attainment as well as employment status. Complementary city-level control variables employed were sourced from the China City Statistical Yearbook for the corresponding years (2011-2018). These yearbooks 24 compile authoritative and major statistical data concerning the socioeconomic development of Chinese cities across various administrative levels. 4.2 Model Specification This study employs a multi-period Difference-in-Differences (DID) method to examine the impact of DI development on urban advanced service industry entrepreneurs. The regression model is specified as follows: Migran𝑡𝑖,𝑡 = α + βPolicyTreat𝑖,𝑡 + γ𝑋𝑖,𝑡   + μ𝑡 + λ𝑖 + ε𝑖,𝑡 (1) The dependent variable, 𝑀𝑖𝑔𝑟𝑎𝑛𝑡𝑖,𝑡 , represents the cumulative count of incoming advanced service industry entrepreneurs attracted to city i in year t. This variable was constructed using CMDS. Specifically, we identified respondents whose employment status was recorded as "employer" or "self-employed worker" and who worked in the following advanced service sectors: information transmission, software, and IT services; finance; real estate; leasing and business services; scientific research and technical services; education; or culture, sports, and entertainment. These selected individual samples were then aggregated at the city-year level, forming a panel dataset reflecting the annual inflow of these entrepreneurs to each city. Given that our data spans eight years (2011-2018), addressing missing values was necessary to ensure data continuity and sample representativeness. Our imputation procedure involved two steps: First, cities with missing data for four or more years within 25 this period were excluded from the analysis, as the substantial data gap was deemed too large for reliable extrapolation. Second, for cities missing data for fewer than four years, linear extrapolation was employed to impute the missing values. We acknowledge that this imputation choice using only linear extrapolation could potentially influence the results. Therefore, the potential impact of this data processing step will be further investigated in the subsequent robustness check section, where additional tests are performed to verify the stability of our main findings. The core explanatory variable, 𝑃𝑜𝑙𝑖𝑐𝑦𝑇𝑟𝑒𝑎𝑡𝑖,𝑡, equals 1 if city i is a designated pilot city and year t is a post-treatment year, and 0 otherwise. 𝜆𝑖 represents city fixed effects, 𝜇𝑡 represents year fixed effects, 𝜀𝑖,𝑡 is the error term. 𝑋𝑖,𝑡 is a vector of control variables for city i in year t. These controls encompass several dimensions: digitalization level, industrial structure, economic strength, population density, education, government input, labor market conditions, and public medical services. Table 1 presents the descriptive statistics for the main variables used. To account for inflation, all monetary variables (e.g., GDP, budget expenditure, salary) were deflated using city-specific annual Consumer Price Index (CPI) data for their respective years. After constructing the entrepreneur panel data and matching it with the "Broadband China" pilot city designations, our final sample comprises 176 cities, consisting of 80 pilot cities (treatment group) and 96 non-pilot cities (control group). Figure 1 provides a visual representation of the spatial distribution of these pilot and non-pilot cities. 26 Table 1: Descriptive Statistics of Principal Variables Figure 1: Pilot Cities of the “Broadband China” Policy 27 5. Main Results 5.1 Baseline Results Table 2 displays the baseline regression results. Model 1 introduces only the treatment variable (𝑃𝑜𝑙𝑖𝑐𝑦𝑇𝑟𝑒𝑎𝑡𝑖,𝑡) alongside city and year fixed effects to control for time-invariant city heterogeneity and common time trends. The results show that the policy variable significantly increased the inflow of advanced service industry entrepreneurs. Model 2 expands on this by incorporating the set of city-level control variables to account for other observable factors that could influence entrepreneurial activity. The findings remain consistent, confirming that the 'Broadband China' pilot policy exerts a statistically significant positive influence on attracting advanced service industry entrepreneurs. 28 Table 2: Baseline Regression Results 5.2 Parallel Trends Test A key identifying assumption for the Difference-in-Differences (DID) methodology is that the treatment and control groups must satisfy the "parallel trends" assumption prior to policy implementation. This means that, without policy intervention, both groups should exhibit similar evolutionary trajectories in attracting advanced service industry entrepreneurs. To formally test this assumption, this study draws on the methodologies of Jacobson et al. (1993) and Beck et al. (2010) and employs an event study approach to 29 examine the dynamic effects surrounding the treatment. The corresponding model is specified as: Migrant𝑖,𝑡 = α + ∑ β𝑘𝑃𝑜𝑙𝑖𝑐𝑦𝑇𝑟𝑒𝑎𝑡𝑖,𝑡 3 𝑘=−3,𝑘≠−1 + γ𝑋𝑖,𝑡 + μ𝑡 + λ𝑖 + ε𝑖,𝑡 (2) All notations have the same meaning as those specified in equation 1, except for β𝑘, which estimates how the average difference in the inflow of advanced service industry entrepreneurs between pilot and non-pilot cities in year k compares to the same difference observed in the base period. k is a relative time indicator and is calculated as the difference between the calendar year and the policy implementation year. To simplify the model specification, years corresponding to three or more periods prior to policy implementation (k ≤ -3) were consolidated into the endpoint category k=-3. Similarly, years corresponding to three or more periods post-implementation (k ≥ +3) were combined into k=+3. The base period is set as the year immediately before policy implementation (k=-1) and is omitted from the regression model to avoid perfect multicollinearity. Figure 2 graphically presents the estimated βk coefficients and their confidence intervals. As depicted, the coefficient estimates for the pre-treatment periods (k=-2 and k=- 3) are statistically insignificant and hover close to zero. This indicates no significant pre- existing differences in trends between the treatment and control groups before the policy intervention, thereby supporting the validity of the parallel trends assumption. In contrast, the coefficients for the year of policy implementation (k=0) and for the subsequent years 30 shown (k=2 and k=3) are positive and statistically significant. This pattern demonstrates a positive impact emerging from the policy's inauguration, confirming that the 'Broadband China' policy effectively promoted the attraction of advanced service industry entrepreneurs after its implementation." Figure 2: Parallel Trends Test 5.3 Placebo Test 5.3.1 Time Placebo Test To verify that the observed policy effect does not merely stem from potential underlying time trends, we conducted a time placebo test. We artificially shift the policy implementation year forward by three years for each treated city. We then re-estimated the 31 event study model specified in Equation (2) using this counterfactual treatment timing, maintaining the same regression framework. Comparing the estimation results from this placebo model with the actual policy model, we found that, except for the coefficient for the third year relative to the placebo treatment (β3), the coefficients for other periods were consistently close to zero and statistically insignificant. This outcome demonstrates the absence of a "spurious effect" appearing before the actual policy implementation, thereby lending strong support to the causal identification of the "Broadband China" policy's impact on attracting advanced service industry entrepreneurs. 5.3.2 City Placebo Test To check that our estimated policy effect wasn't due to particular assignment of cities into treatment and control groups, we also performed a city-level placebo test. Specifically, we repeated the following process 500 times: First, using the original data, we randomly selected 80 cities to act as a 'pseudo-treatment' group, matching the actual treatment group size. Second, we assigned these randomly chosen cities a policy start year, picked randomly and uniformly from the actual start years (2014, 2015, or 2016). Third, using this random assignment, we re-ran our event study model (Equation 2). From each of these 500 runs, we collected the estimated β3 coefficient its p-value. We focus on 𝛃𝟑 as it captures the policy effect at a slightly later stage, arguably representing a more established impact to validate against randomness in treatment assignment. Figure 3 shows the kernel density plot of these 500 estimated 𝛃𝟑 coefficients. The 32 distribution of these placebo coefficients is centered around zero and closely resembles a normal distribution. The frequency of p-values (6% at p<0.05) was consistent with the rate expected purely by random chance. This shows that randomly assigning the policy treatment and timing does not produce a systematic positive or negative effect for the third post-treatment year. Therefore, this test provides strong support that the positive effect of the actual 'Broadband China' policy is not a spurious result driven by specific way the sample was constructed. Figure 3: Kernel Density Distribution of the β₃ Coefficient 33 5.4 Robustness Checks To further ensure the robustness and reliability of our baseline estimates, we conducted four additional robustness tests based on the main regression framework. These checks specifically aimed to rule out potential biases arising from: (1) the exclusion or inclusion of particular samples, (2) the influence of data outliers, (3) unmodeled differential time trends, and (4) confounding effects of other contemporaneous policies. The results of these tests are displayed in Table 3. Table 3: Robustness Check Results 34 5.4.1 Exclusion of Direct-Controlled Municipalities First, we consider the unique status of the four direct-controlled municipalities (Beijing, Shanghai, Tianjin, and Chongqing). Given that these cities possess significantly higher levels of economic development, resource agglomeration capacity, and policy implementation strength compared to ordinary cities, their distinctive characteristics might introduce systematic differences in attracting advanced service industry entrepreneurs, potentially confounding the precise identification of the policy effect. Therefore, we re- estimated the model after excluding these four municipalities from the sample (Model 1). 5.4.2 Exclusion of Outliers Second, to address concerns that abnormal fluctuations or extreme values in the dependent variable might disproportionately influence the estimation results and potentially mask the underlying trend, we capped the dependent variable at the 5th and 95th percentiles to mitigate the impact of outliers before re-estimating the baseline model (Model 2). 5.4.3 Controlling for City-Specific Trends Third, we address the concern that pilot city selection could be non-random and correlated with provincial capital status. Since provincial capitals typically possess superior policy resources, infrastructure, and talent attraction capabilities, they might follow distinct long- term development trends compared to other cities. To ensure such potentially different baseline trends do not confound the identification of the policy effect, we incorporated an interaction term (Capital) between a provincial capital indicator and a linear time trend into 35 the baseline model. (Model 3). 5.4.4 Controlling for Confounding Policy Interventions Additionally, we considered the potential confounding influence of the 'Smart City' pilot policy, which was implemented concurrently with 'Broadband China' during the sample period. The Smart City initiative, aimed at enhancing urban governance and public services using next-generation information technologies (such as IoT, big data, and cloud computing), could potentially exert an independent influence on cities' DI and entrepreneurial environments. To ensure that the effects attributed to 'Broadband China' are not conflated with those of the Smart City policy, we explicitly controlled for participation in the latter by adding a dummy variable (SmartCity) to our model. Given that SmartCity is a dummy variable and time fixed effects are already included, we also incorporated a linear time trend (defined for each city as the current year minus its 'Broadband China' policy implementation year) to help absorb the influence of time-varying factors like macroeconomic growth (Model 4). For clarity in the presentation table, the coefficient for this trend variable is not reported. 5.4.5 Conclusion Taken together, the results in Table 3 confirm the robustness of our main findings. The coefficient for our key variable, PolicyTreat, stays consistently positive and statistically significant across all four robustness checks. This reinforces the conclusion that the 'Broadband China' policy positively impacts the attraction of advanced service industry 36 entrepreneurs. Furthermore, the significant interaction term (Capital) in Model 3 validates the existence of long-term differential trends between capital and non-capital cities, and the significant SmartCity coefficient in Model 4 confirms that the Smart City pilot policy exerts an independent, positive influence on entrepreneur inflow. After accounting for these various factors, our baseline estimation results remain robust and credible. 5.4.6 Propensity Score Matching Difference-in-Differences (PSM-DID) To further address potential selection bias that might remain in the baseline DID estimates, we employed the Propensity Score Matching-Difference-in-Differences (PSM-DID) method. First, we estimated propensity scores predicting pilot city designations based on pre-policy city characteristics. Second, within the common support region, we matched pilot cities to comparable non-pilot cities using both 1:1 Nearest Neighbor (NN) and Kernel matching algorithms based on these scores. Finally, we applied a two-way (city and year) fixed effects DID model to these matched samples. This PSM-DID framework helps control for selection bias related to observable characteristics through the matching process. Table 4 presents the results, demonstrating that the key PolicyTreat coefficient remains positive and significant (at the 5% level) under both NN and Kernel matching approaches. These findings reaffirm that the positive impact of 'Broadband China' on attracting advanced service industry entrepreneurs holds even after explicitly balancing observable pre-treatment characteristics between groups. 37 Table 4: PSM-DID Results 6. Endogeneity Analysis The "Broadband China" pilot policy may be subject to potential endogeneity issues. While the PSM-DID approach can balance observable characteristics, estimates might still suffer from bias if unobservable factors varying over time influenced both policy selection and outcomes prior to the intervention. Specifically, unobserved city characteristics could simultaneously affect a city's likelihood of policy adoption and the inflow of entrepreneurial talent. To mitigate this challenge, we employ an Instrumental Variable (IV) strategy. The results are presented in Table 5. Specifically, we select the minimum geographic distance from each city's centroid to the nearest optical fiber backbone city as the instrumental variable. The validity of this instrument rests on two key conditions: First, relevance; cities geographically closer to the optical fiber backbone network possess a distinct advantage for policy implementation due to potentially lower broadband construction costs, making them more likely to be included 38 in the pilot program. This is empirically supported by our first-stage regression results, which yield a significant Kleibergen-Paap Wald F-statistic of 6.22. Second, a city's distance to the backbone network is determined by long-standing historical spatial patterns, making it relatively stable and unlikely to be influenced by recent economic developments or flows of entrepreneurial talent, limiting concerns about reverse causality. Furthermore, this geographic distance itself is unlikely to directly affect the migration decisions of advanced service industry entrepreneurs, satisfying the exogeneity condition. We employed Two-Stage Least Squares (2SLS) for the estimation. After confirming the instrument's relevance in the first stage, the second-stage results indicate that the 'Broadband China' policy significantly increased the inflow of advanced service industry entrepreneurs in pilot cities (Coefficient = 29.58, p = 0.025). It is crucial, however, to interpret this estimate as a Local Average Treatment Effect (LATE). This represents the causal effect specifically for the subset of cities whose likelihood of participating in the pilot program was influenced by the instrument. The potentially larger magnitude of this IV coefficient might reflect this focus on a more responsive marginal group. Furthermore, while the instrument meets the relevance criterion, its strength is somewhat modest, as indicated by the Kleibergen-Paap Wald F-statistic of 6.22 falling below conventional thresholds for strong instruments (e.g., Stock-Yogo critical values). Weak instruments can potentially lead to bias in the estimated coefficient's magnitude, but they generally do not undermine the conclusions regarding the direction or statistical 39 significance of the causal effect. In summary, this approach provides more robust empirical evidence for evaluating the causal effect of the policy. Table 5: IV Estimation Results 7. Mechanism Analysis Our hypothesis development section suggested that the "Broadband China" policy could enhance the attraction of advanced service industry entrepreneurs by expanding the digital market size, fostering the agglomeration of entrepreneurial activity, and elevating the urban innovation level. We now empirically test these proposed channels using a path analysis framework. The general models are specified as follows: 40 Mediator𝑖𝑡 = α + a PolicyTreatit + γ𝑿𝒊𝒕 + μ𝑡 + λ𝑖 + ε𝑖𝑡 (3) Migrant𝑖𝑡 = α + c PolicyTreat𝑖𝑡 + b Mediator𝑖𝑡 + γ𝑿𝒊𝒕 + μ𝑡 + λ𝑖 + ε𝑖𝑡 (4) 𝑀𝑒𝑑𝑖𝑎𝑡𝑜𝑟𝑖,𝑡 represents one of the three mediating variables for city i in year t. We operate these mediators as follows: Digital Market Size (ecom) is measured by the per capita e- commerce transaction value in each city. The rationale is that this metric indirectly reflects the scale of the online customer base accessible to businesses; entrepreneurs can leverage DI to reach broader consumer markets, thereby benefiting from digital economies of scale. Entrepreneurial Space Agglomeration (newfirm) is measured by the number of newly established firms per million inhabitants in each city. The reason is that higher startup density is a direct manifestation of a vibrant and concentrated entrepreneurial environment where key resources and activities are geographically clustered. Urban Innovation Level (innoindex) is measured by an urban innovation index, calculated based on patent data from the China National Intellectual Property Administration and data on firms' registered capital from the State Administration for Market Regulation. Methodologically, we employ a two-stage estimation approach. Equation (3) assesses the impact of 𝑷𝒐𝒍𝒊𝒄𝒚𝑻𝒓𝒆𝒂𝒕 on the mediating variable 𝑀𝑒𝑑𝑖𝑎𝑡𝑜𝑟 . Equation (4) examines the effect of the mediator on the outcome variable 𝑀𝑖𝑔𝑟𝑎𝑛𝑡, controlling for the direct effect of the policy variable. Subsequently, we estimate the indirect effect (𝑎 × 𝑏), representing the influence of the Broadband China policy transmitted through each specific 41 mediator. Confidence intervals for this indirect effect are obtained using a cluster bootstrap procedure with 1000 repetitions. The empirical results for the path analysis are presented in Table 6. The findings reveal positive indirect effects for all three proposed channels. The most substantial pathway operates via Urban Innovation Level (innoindex). We estimated a statistically significant indirect effect of 1.079. This suggests that a key mechanism by which 'Broadband China' attracts talent is through enhancing urban innovation, indirectly contributing an estimated average increase of approximately 1.08 advanced service industry entrepreneurs per treated city-year. Digital Market Size (ecom) also yields a statistically significant positive indirect effect. This finding supports the role of DI in expanding the accessible online market, thereby indirectly influencing entrepreneur inflow by an estimated 0.056 individuals per city-year. Finally, the analysis confirms a significant mediating role for Entrepreneurial Space Agglomeration (newfirm). By fostering a more vibrant startup ecosystem, the policy indirectly accounted for attracting approximately 0.03 additional entrepreneurs per city- year. 42 Table 6: Path-Analysis Coefficients by Mediator 8. Conclusion This study utilized data from the 2011–2018 China Migrants Dynamic Survey (CMDS) combined with panel data for 117 prefecture-level Chinese cities to empirically examine the effect of the "Broadband China" pilot policy on attracting advanced service industry entrepreneurs, employing a multi-period Difference-in-Differences (DID) model. The research findings clearly indicate two main points. First, the implementation of the "Broadband China" policy significantly enhanced the attractiveness of pilot cities to advanced service industry entrepreneurs. This core conclusion remained robust after undergoing a series of rigorous checks, including parallel trends tests, placebo tests, and PSM-DID, thereby strengthening the internal validity and credibility of the results. 43 Second, this study further explored the mechanisms through which the "Broadband China" policy exerted its attractive effect. We found that the policy primarily operated through three channels: (1) expanding the local digital market size, which provided a broader market base for related entrepreneurial activities; (2) promoting the emergence of new firms, leading to a more vibrant entrepreneurial ecosystem; and (3) boosting the city's overall innovation level, which optimized the macro-environment for high-caliber talent. 8.1 Contributions This study offers several contributions. First, it empirically clarifies the specific role of DI in attracting advanced service entrepreneurs, refining our understanding within economic geography and innovation ecosystem theories. Second, by identifying key mediating pathways, our research provides a nuanced, micro-level perspective on how DI reshapes regional economies, going beyond simple policy effect estimation. Our findings also yield important practical insights. First, they provide strong empirical justification for continued investment in DI, positioning high-quality broadband as a strategic asset for attracting valuable entrepreneurs and upgrading the service sector. Second, the mechanism analysis highlights the need for a systematic policy approach: complementing 'hard' infrastructure with supportive 'soft' environments, including fostering digital markets, implementing pro- startup policies, and investing in urban innovation capacity. 44 8.2 Limitations It is important to acknowledge certain considerations regarding data processing and model selection in this study. Due to annual fluctuations in the total sample size of the CMDS data (e.g., approximately 128,000 in 2011 versus 206,000 in 2015) and the unavailability of precise city-level sampling quotas, our analysis used the absolute count of cumulative advanced service entrepreneurs within target cities as the dependent variable, rather than their proportion relative to the city-level sampling quotas. This approach necessitates caution when interpreting the results, as the observed increase in entrepreneur numbers could potentially be influenced partly by variations in sample size, although the DID model design helps control for time-invariant city characteristics and common time trends. Furthermore, considering that the dependent variable exhibits overdispersion (sample mean 8.19, variance 14.45), employing a negative binomial regression model might represent a more suitable econometric strategy from a technical standpoint. Compared to Poisson regression, the negative binomial model better accommodates overdispersion. Compared to transforming the dependent variable using logarithms (e.g., log(y+1)), it avoids potential information loss when log counts are small and often offers greater interpretability. Despite not ultimately adopting the negative binomial approach in this study, we maintain that, based on the current DID model framework and the comprehensive set of robustness checks performed, the core conclusion regarding the positive effect of the "Broadband China" policy remains valid and convincing. 45 REFERENCE Acs, Z. J., & Armington, C. (2004). The impact of geographic differences in human capital on service firm formation rates. Journal of Urban Economics, 56(2), 244–278. https://doi.org/10.1016/j.jue.2004.03.008 Acs, Z. J., Audretsch, D. B., & Lehmann, E. E. (2013). The knowledge spillover theory of entrepreneurship. Small Business Economics, 41(4), 757–774. https://doi.org/10.1007/s11187-013-9505-9 Atasoy, H. (2013). The Effects of Broadband Internet Expansion on Labor Market Outcomes. ILR Review, 66(2), 315–345. https://doi.org/10.1177/001979391306600202 Audretsch, D., & Feldman, M. (1996). R&D spillovers and the geography of innovation and production. The American Economic Review. Autio, E., Nambisan, S., Thomas, L. D. W., & Wright, M. (2018). Digital affordances, spatial affordances, and the genesis of entrepreneurial ecosystems. Strategic Entrepreneurship Journal, 12(1), 72–95. https://doi.org/10.1002/sej.1266 Balland, P.-A., Boschma, R., & Frenken, K. (2015). Proximity and Innovation: From Statics to Dynamics. REGIONAL STUDIES, 49(6), 907–920. https://doi.org/10.1080/00343404.2014.883598 Ben Arfi, W., & Hikkerova, L. (2021). Corporate entrepreneurship, product innovation, and knowledge conversion: The role of digital platforms. Small Business Economics, 56(3), 1191–1204. https://doi.org/10.1007/s11187-019-00262-6 46 D'Angelo, S., Cavallo, A., Ghezzi, A., & Di Lorenzo, F. (2024). Understanding corporate entrepreneurship in the digital age: A review and research agenda. Review of Managerial Science, 18(12), 3719–3774. https://doi.org/10.1007/s11846-024-00730-8 Delacroix, E., Parguel, B., & Benoit-Moreau, F. (2019). Digital subsistence entrepreneurs on Facebook. TECHNOLOGICAL FORECASTING AND SOCIAL CHANGE, 146, 887–899. https://doi.org/10.1016/j.techfore.2018.06.018 Du, Z.-Y., & Wang, Q. (2024). Digital infrastructure and innovation: Digital divide or digital dividend? JOURNAL OF INNOVATION & KNOWLEDGE, 9(3), 100542. https://doi.org/10.1016/j.jik.2024.100542 European Commission. (2021). 2030 Digital Compass: The European Way for the Digital Decade. Federal Communications Commission. (2010). Connecting America: The National Broadband Plan. Feldman, M. P. (2001). The Entrepreneurial Event Revisited: Firm Formation in a Regional Context. Industrial and Corporate Change, 10(4), 861–891. https://doi.org/10.1093/icc/10.4.861 Forman, C., Goldfarb, A., & Greenstein, S. (2012). The Internet and Local Wages: A Puzzle. AMERICAN ECONOMIC REVIEW, 102(1), 556–575. https://doi.org/10.1257/aer.102.1.556 Glaeser, E. L., Kallal, H. D., Scheinkman, J. A., & Shleifer, A. (1992). Growth in Cities. 47 Journal of Political Economy, 100(6), 1126–1152. https://doi.org/10.1086/261856 Government of Japan. (2016). Society 5.0 Aiming for a New Human-Centered Society. Hänninen, M., Smedlund, A., & Mitronen, L. (2018). Digitalization in retailing: Multi- sided platforms as drivers of industry transformation. Baltic Journal of Management, 13(2), 152–168. https://doi.org/10.1108/BJM-04-2017-0109 Hanseth, O., & Lyytinen, K. (2010). Design theory for dynamic complexity in information infrastructures: The case of building internet. JOURNAL OF INFORMATION TECHNOLOGY, 25(1), 1–19. https://doi.org/10.1057/jit.2009.19 Henfridsson, O., & Bygstad, B. (2013). The Generative Mechanisms of Digital Infrastructure Evolution. MIS QUARTERLY, 37(3), 907-+. https://doi.org/10.25300/MISQ/2013/37.3.11 Krugman, P. (1991). Increasing Returns and Economic Geography. Journal of Political Economy, 99(3), 483–499. https://doi.org/10.1086/261763 Mack, E., & Faggian, A. (2013). Productivity and Broadband: The Human Factor. INTERNATIONAL REGIONAL SCIENCE REVIEW, 36(3), 392–423. https://doi.org/10.1177/0160017612471191 Malecki, E. J. (2018). Entrepreneurship and entrepreneurial ecosystems. GEOGRAPHY COMPASS, 12(3), e12359. https://doi.org/10.1111/gec3.12359 Manyika, J., Ramaswamy, S., Khanna, S., Sarrazin, H., Pinkus, G., Sethupathy, G., & Yaffe, A. (2015, December). Digital America: A tale of the haves and have-mores. 48 Melville, N., & Kohli, R. (2021). Roadblocks to Implementing Modern Digital Infrastructure: Exploratory Study of API Deployment in Large Organizations. Hawaii International Conference on System Sciences. https://doi.org/10.24251/HICSS.2021.723 Muller, E., & Doloreux, D. (2009). What we should know about knowledge-intensive business services. Technology in Society, 31(1), 64–72. https://doi.org/10.1016/j.techsoc.2008.10.001 Müller, J. M., Buliga, O., & Voigt, K.-I. (2018). Fortune favors the prepared: How SMEs approach business model innovations in Industry 4.0. Technological Forecasting and Social Change, 132, 2–17. https://doi.org/10.1016/j.techfore.2017.12.019 Nambisan, S. (2017). Digital Entrepreneurship: Toward a Digital Technology Perspective of Entrepreneurship. Entrepreneurship Theory and Practice, 41(6), 1029–1055. https://doi.org/10.1111/etap.12254 Nambisan, S., Wright, M., & Feldman, M. (2019). The digital transformation of innovation and entrepreneurship: Progress, challenges, and key themes. RESEARCH POLICY, 48(8), 103773. https://doi.org/10.1016/j.respol.2019.03.018 Ndubuisi, G., Otioma, C., & Tetteh, G. K. (2021). Digital infrastructure and employment in services: Evidence from Sub-Saharan African countries. Telecommunications Policy, 45(8), 102153. https://doi.org/10.1016/j.telpol.2021.102153 Neumark, D., Wall, B., & Zhang, J. (2011). Do Small Businesses Create More Jobs? New Evidence for the United States from the National Establishment Time Series. The 49 Review of Economics and Statistics, 93(1), 16–29. https://doi.org/10.1162/REST_a_00060 Niu, G., Jin, X., Wang, Q., & Zhou, Y. (2022). Broadband infrastructure and digital financial inclusion in rural China. China Economic Review, 76, 101853. https://doi.org/10.1016/j.chieco.2022.101853 Oyinlola, M., Schröder, Patrick, Whitehead, Timothy, Kolade, Oluwaseun, Wakunuma, Kutoma, Sharifi, Soroosh, Rawn, Barry, Odumuyiwa, Victor, Lendelvo, Selma, Brighty, Geoff, Tijani, Bosun, Jaiyeola, Tomi, Lindunda, Lukonga, Mtonga, Radhia, & and Abolfathi, S. (2022). Digital innovations for transitioning to circular plastic value chains in Africa. Africa Journal of Management, 8(1), 83–108. https://doi.org/10.1080/23322373.2021.1999750 Qian, H., Acs, Z. J., & Stough, R. R. (2013). Regional systems of entrepreneurship: The nexus of human capital, knowledge, and new firm formation. JOURNAL OF ECONOMIC GEOGRAPHY, 13(4), 559–587. https://doi.org/10.1093/jeg/lbs009 Salemink, K., Strijker, D., & Bosworth, G. (2017). Rural development in the digital age: A systematic literature review on unequal ICT availability, adoption, and use in rural areas. Journal of Rural Studies, 54, 360–371. https://doi.org/10.1016/j.jrurstud.2015.09.001 Sang, D., Munga, J., & Sambuli, N. (2025, February 27). Digital Public Infrastructure: A Practical Approach for Africa. Carnegie Endowment for International Peace. 50 https://carnegieendowment.org/research/2025/02/digital-public-infrastructure-a- practical-approach-for-africa?lang=en Santos, J. B. (2019). Knowledge-intensive business services and innovation performance in Brazil. Innovation & Management Review, 17(1), 58–74. https://doi.org/10.1108/INMR-03-2019-0025 Smart Nation Singapore. (2014). Smart Nation: The Way Forward. Stam, E. (2015). Entrepreneurial Ecosystems and Regional Policy: A Sympathetic Critique. EUROPEAN PLANNING STUDIES, 23(9), 1759–1769. https://doi.org/10.1080/09654313.2015.1061484 Storper, M., & Venables, A. J. (2004). Buzz: Face-to-face contact and the urban economy. JOURNAL OF ECONOMIC GEOGRAPHY, 4(4), 351–370. DRUID Summer Conference. https://doi.org/10.1093/jnlecg/lbh027 Tilson, D., Lyytinen, K., & Sørensen, C. (2010). Research Commentary —Digital Infrastructures: The Missing IS Research Agenda. Information Systems Research, 21(4), 748–759. https://doi.org/10.1287/isre.1100.0318 Xie, W., Li, Zhongshun, Wang, Z., Zheng, D., & Wang, Y. (2024). How Does Digital Infrastructure Affect Manufacturing SMEs Business Model Innovation? An Empirical Study in Guangdong Province. Emerging Markets Finance and Trade, 60(10), 2300– 2312. https://doi.org/10.1080/1540496X.2023.2293975 Zieba, M. (2013). Knowledge-intensive business services (KIBS) and their role in the 51 knowledge-based economy (Working Paper 7/2013 (7)). GUT FME Working Paper Series A. https://www.econstor.eu/handle/10419/173303