Smart city construction empowers tourism: mechanism analysis and spatial spillover effects

Rapid global urbanization has intensified urban challenges such as congestion, environmental degradation, and resource misallocation, motivating governments to explore smart city initiatives. China began smart city exploration in 2008 and designated national pilot cities in three batches (2012–2014), framing smart cities as a policy instrument to upgrade urban governance and services through IoT, cloud computing, big data, and integrated information systems. Tourism, a strategic pillar of China’s economy, has faced shocks from COVID-19 and shifting, diverse demand patterns. Smart city construction offers a promising pathway to modernize tourism via enhanced infrastructure, digital services, data sharing, and improved urban livability. The paper poses key research questions: whether smart city designation promotes tourism development, the magnitude of its effects, the underlying mechanisms, and whether spatial spillovers exist. Using panel data for 297 prefecture-level cities (2000–2021), the study employs a time-varying DID framework, explores mechanisms linking smart city construction to tourism (investment, entrepreneurship, operational efficiency, talent attraction), and integrates a spatial Durbin DID model to test for spillovers, thereby contributing quantitative evidence to a largely qualitative literature.

Tourism development is shaped by industry structure, economic development, policy support, transport accessibility, resource endowments, product characteristics, and market scale. With widespread ICT and the digital economy, tourism increasingly integrates digital technologies, reducing transaction costs, transforming production and management in firms, and enabling product and format innovation. Smart city theory evolved from a technology-centered view to emphasize people, social learning, and the interplay of human capital, transport, and ICT in resource allocation and growth. While proponents argue smart cities can foster sustainable tourism through data-driven destination management and service innovation, concerns include privacy risks, digital divides, and usability barriers for some groups. In China, smart city pilots were launched in 2012 with subsequent batches in 2013 and 2014, supported by frameworks and funding from MOHURD and guided by NDRC policies. Parallel national initiatives advanced smart tourism via standards, smart scenic spots, enterprise informatization, and a national tourism informatization plan. Despite extensive conceptual work, empirical evidence on how smart city construction affects tourism remains scarce, motivating this study.

Design: The staggered rollout of China’s smart city pilots (2012–2014) is treated as a quasi-natural experiment. A time-varying DID model estimates the average treatment effect of smart city designation on urban tourism development, with city and year fixed effects and robust standard errors clustered at the city level. An event-study specification tests the parallel trends assumption and captures dynamic effects post-designation. Model: TourDev_it = β0 + β1 SmartCity_it + γ Z_it + μ_i + ν_t + ε_it. Event-study augments with leads and lags of SmartCity_it to test pre-trends and dynamic impacts. Data: Panel of 297 mainland Chinese prefecture-level cities, 2000–2021. Tourism revenue and tourist arrivals from CEIC; controls from China City Statistical Yearbook, Yearbook of China Tourism Statistics, and city statistical bureaus. Smart city pilot lists from MOHURD. Monetary values are deflated to 2000 base year. Some observations are missing; actual estimation samples vary accordingly. Variables:

• Outcomes: Tourism development measured by (i) tourism revenue (domestic + converted foreign exchange in RMB), and (ii) tourist arrivals (domestic + inbound). Both are log-transformed.
• Treatment: SmartCity_it = 1 from the year a city is designated (2012–2014) onward; 0 otherwise.
• Controls: Human capital (higher-education enrollments/population), log per capita GDP, government intervention (fiscal expenditure/GDP), openness (utilized foreign capital/GDP), industrial structure (tertiary value added/GDP), transportation (highway mileage per unit area), and public health events (dummy=1 for 2003, 2020, 2021). Robustness:
• PSM-DID to address selection bias: logit model to estimate propensity scores; nearest-neighbor within caliper and kernel matching to construct comparable samples.
• Sample change: exclusion of four municipalities (Beijing, Tianjin, Shanghai, Chongqing) to test sensitivity due to their distinct status.
• Placebo: 500 random assignments mimicking the 2012–2014 batch sizes to test whether spurious factors drive results. Mechanism analysis: Using city-level tourism enterprise data for 23 cities (2000–2016; N≈391 before cleaning; 326 in models), four mediators are examined via DID: fixed-asset investment of tourism enterprises (investment), number of tourism enterprises (entrepreneurship/competition), overall labor productivity of tourism enterprises (operational efficiency), and number of employees in tourism enterprises (talent inflow). Spatial analysis: A spatial Durbin DID model (SDM-DID) assesses spillovers: TourDev_it = β0 + ρ W TourDev_it + β1 SmartCity_it + β2 W SmartCity_it + γ W Z_it + μ_i + ν_t + ε_it. Three spatial weight matrices are used: (W1) inverse geographic distance, (W2) economic distance (per capita GDP ratios), and (W3) geo-economic nested matrix (combination of W1 and W2). Due to spatial model requirements, data from 2002–2020 are used; missing observations are completed via interpolation. Spatial autocorrelation is verified by Moran’s I (positive and significant across years and matrices). Wald and LR tests favor SDM over SLM/SEM; Hausman supports two-way fixed SDM. Effects are decomposed into direct, indirect (spillover), and total effects.

• Main DID effects: Smart city designation increases urban tourism revenue by approximately 24.3% and tourist arrivals by 17.2%. Event-study confirms parallel pre-trends and shows that positive impacts grow over time as smart city construction matures.
• Control variables behave as expected: higher human capital, per capita GDP, openness, tertiary industry share, and transportation are associated with stronger tourism; major public health events suppress tourism.
• Robustness:
  • PSM-DID yields similar significant effects (e.g., coefficients around 0.286 for revenue and 0.157–0.182 for arrivals), indicating selection bias does not drive results.
  • Excluding four municipalities strengthens estimated effects (e.g., revenue ~0.346; arrivals ~0.259).
  • Placebo tests (500 random assignments) produce coefficient distributions centered near zero with non-significant p-values, distinct from actual estimates, supporting causal interpretation.
• Mechanisms: Smart city designation significantly raises (i) fixed-asset investment of tourism enterprises (~0.286), (ii) number of tourism enterprises (~0.082), (iii) overall labor productivity (~0.211), and (iv) number of employees (~0.167), supporting channels of investment attraction, entrepreneurship, efficiency gains, and talent inflow.
• Spatial spillovers: SDM-DID shows SmartCity and spatially lagged SmartCity are significantly positive across geographic, economic, and geo-economic weight matrices; the spatial lag of the dependent variable is also positive. Decomposition indicates significant direct effects on pilot cities and positive indirect (spillover) effects on neighboring or economically similar cities, implying that smart city construction enhances regional tourism beyond pilot boundaries.

The findings strongly support H1: smart city construction promotes tourism development, evidenced by significant increases in tourism revenue and arrivals and by dynamic effects that strengthen over time. They also support H2: multiple mechanisms—heightened tourism investment, stimulated entrepreneurship, improved operational efficiency, and increased tourism-related employment—mediate these gains. The robust results across matching strategies, sample alterations, and placebo tests mitigate concerns about selection or unobserved confounding. The positive spatial spillovers indicate that smart city benefits extend to non-pilot cities, particularly those proximate or economically similar, encouraging coordinated regional strategies. The observed dynamic amplification likely reflects learning, network externalities, and broader adoption of digital services that improve traveler information, experience, and destination management. Collectively, the results underscore that integrating ICT infrastructure, data platforms, and smart governance can create conducive environments for tourism sector transformation and growth, both locally and regionally.

This study provides rigorous empirical evidence that China’s smart city construction significantly boosts urban tourism performance, increasing tourism revenue and arrivals, with effects that intensify over time. It opens the mechanism “black box,” showing that smart city designation attracts tourism investment, fosters entrepreneurship, enhances operational efficiency, and draws talent into tourism. Furthermore, smart city policies generate positive spatial spillovers, elevating tourism in neighboring and economically similar cities. Policy implications include: prioritizing and supporting smart city initiatives; strengthening R&D and deployment of smart technologies in tourism; leveraging resource allocation and scale economies to raise tourism efficiency; aligning tourism policies with smart city goals (e.g., information integration, lower entry barriers, streamlined entrepreneurship approvals, skills training); and implementing inter-city joint marketing and multi-destination product design to share cross-regional tourism dividends. These insights can inform urban and tourism strategies in other countries planning smart upgrades and tourism transformation.

• Data constraints: Some observations are missing in the main panel, reducing effective sample sizes. For spatial models, missing values in 2000, 2001, and 2021 necessitated restricting to 2002–2020 and using interpolation for gaps.
• Mechanism data coverage: Tourism enterprise-level city data were available only for 23 cities (2000–2016), limiting mechanism tests’ spatial and temporal breadth.
• Potential selection concerns: Although addressed via PSM-DID and placebo tests, the non-random nature of pilot designation could still pose residual bias risks.
• Heterogeneity of municipalities: Distinct characteristics of four municipalities warranted separate robustness checks; broader heterogeneity across cities may remain.
• Data access limitations: Underlying datasets are subject to provider agreements and cannot be publicly shared, which may affect replicability routes for some researchers.