Has the Establishment of National Big Data Comprehensive Pilot Zones Promoted Tourism Economic Growth? Empirical Evidence from 284 Chinese Cities

Abstract

The strategic value of the digital economy in driving the transformation of the tourism industry has become increasingly prominent. This study investigates the impact of China’s National Big Data Comprehensive Pilot Zones (NBDCPZs) on tourism economic development. Using panel data from 284 cities (2009–2019), the study treats the policy as a quasi-natural experiment and applies multi-period DID, PSM, and spatial models. Results show that NBDCPZs significantly promote tourism growth, though effects decline over time and vary spatially—stronger in eastern regions and moderate in large cities. Industrial structure upgrading is identified as the main transmission channel, while financial support amplifies the policy effect, and human capital shows no significant moderation. The findings reveal how digital policy drives tourism transformation and provide practical insights for strengthening data—tourism integration, optimizing industrial structures, and fostering high-quality regional development.

Keywords

national big data comprehensive pilot zone tourism economy spatiotemporal heterogeneity transmission mechanism

Introduction

As a strategic pillar of China’s national economy, the tourism industry is experiencing a fundamental shift from a traditional resource-dependent model to a data-driven paradigm. The convening of the 2024 National Tourism Development Conference—China’s first central-level meeting devoted exclusively to tourism—marked a turning point in the nation’s development agenda. The conference emphasized the need to “promote high-quality tourism development and accelerate the building of a strong tourism nation,” signifying that advancing high-quality growth in the tourism economy has become both a national priority and a key component of China’s modernization process. High-quality tourism development entails improving efficiency, optimizing industrial and consumption structures, and enhancing environmental sustainability (Liu & Han, 2020). Scholars have approached this issue from multiple perspectives, including the theoretical connotations of tourism economic development (Li & Wei, 2014), the spatial correlation structure and its effects (Wang et al., 2017), and the determinants influencing tourism growth (Zhou et al., 2021). At its core lies a fundamental question: how can data-driven innovation be harnessed to achieve sustainable tourism economic growth?

Amid this transformation, the digital economy has emerged as a decisive force reshaping the production logic and growth trajectory of the tourism industry. By improving productivity, optimizing resource allocation, and promoting structural upgrading, digitalization enables tourism enterprises and destinations to achieve higher levels of operational efficiency and service innovation (Chen et al., 2025). In this context, big data policies play a crucial role by facilitating the market-oriented allocation of data elements, deepening industrial integration, and enhancing policy precision (Wei et al., 2022). Understanding how these policies influence tourism economic development and through which mechanisms they operate is therefore of significant theoretical and practical importance for achieving data-enabled improvements in tourism quality and efficiency. Since 2015, China has successively established several National Big Data Comprehensive Pilot Zones (NBDCPZs)—including the Guizhou pilot zone, two interregional zones (Beijing-Tianjin-Hebei and the Pearl River Delta), four demonstration zones (Shanghai, Henan, Chongqing, and Shenyang), and one infrastructure-oriented zone (Inner Mongolia). These zones serve as experimental platforms for promoting digital transformation, integrating data resources, and advancing cross-sector collaboration. From a broader industrial policy perspective, National Big Data Comprehensive Pilot Zones (NBDCPZs) represent a quintessential top-down digital governance tool within China’s evolving industrial strategy framework. Existing research indicates that large-scale digital and industrial policies—such as Made in China 2025—play a pivotal role in advancing sustainable development by reshaping production efficiency, innovation structures, and green transition pathways (Zhang et al., 2025). Emerging evidence further demonstrates that digitally oriented trade policies foster green growth by reducing transaction costs, enhancing cross-border service efficiency, and accelerating industrial upgrading—particularly in service-intensive sectors (Zhang & Choi, 2025). Compared to manufacturing-focused strategies, NBDCPZs prioritize data factor allocation, digital infrastructure integration, and platform-based governance—characteristics highly aligned with the service-dominant nature of the tourism economy. Therefore, analyzing digital trade pilot free trade zones within the broader framework of China’s digital industrial policy system helps clarify their structural role in promoting high-quality, sustainable tourism development.

Given the lack of standardized indicators to comprehensively measure big data development, the establishment of NBDCPZs provides an ideal quasi-natural experiment for evaluating the economic impact of data policies. Drawing on panel data from 284 prefecture-level cities in China between 2009 and 2021, this study investigates the causal effects and mechanisms through which NBDCPZs influence tourism economic development. It contributes to the literature in three main ways. First, while existing studies have examined the effects of big data policies on innovation, industrial upgrading, and macroeconomic growth few have explored their impact on tourism (Afolabi, 2023; Chen et al., 2022; Zhao et al., 2024). Second, prior studies have investigated the formation and development of the tourism economy (Wei & Hu, 2025), but most remain at a qualitative level and have paid limited attention to the causal relationship between big data policy and tourism economic performance by empirically testing the relationship between digital policy interventions and tourism performance, this study clarifies the role of data as a production factor in the tourism economy. Third, whereas prior research on NBDCPZs has focused primarily on firm-level or regional outcomes (Yue et al., 2024), this study highlights their effects on local tourism economies and spatial heterogeneity, refining the theoretical understanding of how big data initiatives can drive tourism transformation in the era of digital intelligence. This study analyzes the innovation-driven logic of data elements in shaping tourism economic development, examines the policy effects and transmission mechanisms of the pilot zones, and thereby contributes to improving the theoretical framework of coordinated development between digital technology and the tourism economy. The findings also offer practical implications for optimizing regional big data policy design and advancing the digital transformation of the tourism industry.

Literature Review

Theoretical Analysis and Research Hypotheses

As a key platform for advancing the market-oriented allocation of data elements and optimizing digital infrastructure, the National Big Data Comprehensive Pilot Zones (NBDCPZs) have injected new momentum into China’s tourism transformation. In the context of the digital economy, tourism production modes and industrial boundaries are being reshaped, profoundly influencing the future trajectory of tourism development (Wei, 2022). Empirical evidence indicates that smart tourism models significantly enhance sustainable tourism growth, improving efficiency and innovation through digital technologies (Phaosathianphan & Leelasantitham, 2021). From an international trade perspective, tourism is widely regarded as an intangible export whose competitiveness is highly dependent on information and communication technology (ICT). A growing body of trade literature indicates that ICT is fundamentally reshaping international trade patterns by reducing information asymmetry and facilitating cross-border service transactions (Park & Choi, 2025). Digital tools such as online booking platforms, intelligent recommendation systems, and data-driven marketing channels enable tourism destinations to expand market reach and enhance service matching efficiency. Further evidence indicates that ICT development significantly enhances export performance, particularly in service-oriented and information-intensive industries (Nguyen & Choi, 2025). By strengthening digital infrastructure, data integration, and platform governance, New Frontier Development Cooperation Plan Zones (NBDCPZs) have elevated tourism destinations’ capacity to participate in digital services trade. Therefore, this study proposes the following hypothesis:

H1: The establishment of NBDCPZs significantly promotes the development of the tourism economy.

However, regional heterogeneity shapes the effectiveness of big data policies. Differences in resource endowments, economic foundations, and policy environments have created marked spatial disparities in China’s digital development (Lü et al., 2023; Song et al., 2020). Regions with weaker initial conditions often experience stronger path dependence, limiting policy effectiveness (Jedwab et al., 2017). Eastern China—characterized by advanced digital infrastructure, higher marketization, and stronger policy adaptability—translates data resources into tourism value more efficiently than central and western regions, which face deficits in infrastructure, human capital, and financial input (Zhi et al., 2021). Consequently, the policy effects of NBDCPZs are expected to be strongest in eastern cities. Therefore, this study proposes the following hypothesis:

H2: The positive effect of NBDCPZs on tourism economic development exhibits regional heterogeneity, being stronger in eastern than in central and western regions.

The NBDCPZs also promote tourism growth through industrial restructuring. Industrial structure optimization is a key determinant of high-quality regional development. Big data policies accelerate the transformation of traditional industries toward high value-added, knowledge-intensive sectors, fostering integration between tourism, culture, and technology (Guo et al., 2024). Empirical research indicates that tourism-related foreign direct investment and green technological innovation jointly promote sustainable tourism development by fostering technology spillover effects and enhancing environmental performance (Zhou & Choi, 2025). Furthermore, by deepening the integration of manufacturing and modern service industries, pilot regions have established digital service systems that support sustainable tourism upgrades, thereby strengthening their competitiveness in attracting high-end tourism projects (Chen & Zhang, 2023). The resulting “structural dividend” expands the tertiary sector and enhances environmental and allocative efficiency, improving both the quality and spillover effects of tourism growth (Wang et al., 2023).Therefore, this study proposes the following hypothesis:

H3: The establishment of NBDCPZs promotes tourism economic development through the optimization of industrial structure.

Information infrastructure plays a pivotal role in transmitting benefits. Research indicates that digitalization reduces information barriers and expands digital service coverage, thereby mitigating socioeconomic inequality (Yin & Choi, 2023). Simultaneously, data-driven systems significantly enhance audience engagement by enabling personalized interactions, real-time feedback, and behavioral response mechanisms. Empirical research indicates that digital platforms substantially boost audience participation and behavioral responses through personalized interaction mechanisms (Yun et al., 2025). Innovative initiatives such as smart scenic area systems, crowd flow monitoring, and location-based services vividly demonstrate how informatization enhances tourism industry performance (Gannon & Tahri, 2020). Collectively, these mechanisms foster the emergence of “new productive forces,” characterized by the organic integration of technological innovation, institutional efficiency, and inclusive participation. Therefore, this study proposes the following hypothesis:

H4: NBDCPZs positively influence tourism economic development by enhancing informatization.

Human capital is often regarded as a key driver of innovation, sustainable trade, and green economic growth. Existing research indicates that human capital plays a crucial role in promoting sustainable trade and facilitating the absorption of digital technologies (Choi et al., 2023). However, the effectiveness of human capital depends on its alignment with industry-specific skill requirements. Within the tourism sector, digital transformation increasingly relies on interdisciplinary competencies integrating data analytics, platform operations, creative content production, and service innovation. Furthermore, cultural and creative innovation has been demonstrated to enhance emission reductions and environmental performance in tourism-related activities (Park et al., 2024). The potential mismatch between educational structures and digital tourism demands provides a theoretical explanation for the heterogeneous or weak moderating effects of human capital observed in policy impact assessments. Therefore, this study proposes the following hypothesis:

H5: Human capital positively moderates the relationship between NBDCPZs and tourism economic development.

Finally, financial support amplifies the policy effect of NBDCPZs. Digital inclusive finance and financial technology alleviate financing constraints, promote investment in smart tourism, and facilitate digital transformation. Robust financial systems enhance policy transmission, industrial integration, and consumer responsiveness, jointly driving tourism upgrading. Empirical findings confirm that regions with higher financial support experience stronger NBDCPZ policy effects, with notable spatial spillovers (Guo & Ming, 2025; Xu et al., 2024).Therefore, this study proposes the following hypothesis:

H6: Financial support positively moderates the relationship between NBDCPZs and tourism economic development.

Research Design

Research Area

This study selects 284 prefecture-level and provincial capital cities across China as the research sample, ensuring representativeness with respect to the locations of the National Big Data Comprehensive Pilot Zones (NBDCPZs). The selected cities cover a wide geographical range, including economically developed areas such as the eastern coastal and southern riverside regions, as well as the central, western, and northeastern parts of China. In terms of urban scale, the sample encompasses mega and large cities in addition to medium- and small-sized cities, thereby providing a comprehensive reflection of the characteristics of cities across different regions and administrative levels. The study period spans from 2009 to 2021, covering the establishment cycles of multiple batches of pilot zones. This time frame provides a complete temporal series for evaluating the dynamic effects of the policy implementation. The post-2019 period is deliberately excluded from the baseline analysis. The outbreak of COVID-19 constitutes a nationwide exogenous shock that led to an unprecedented collapse of tourism demand and supply, generating a structural break that is orthogonal to the implementation of the NBDCPZ policy. Including the pandemic years would contaminate the identification of long-term policy effects with short-term crisis-induced fluctuations. Restricting the sample to the pre-pandemic period is therefore necessary to ensure internal validity.

Research Data

Data were sourced from the China Urban Statistical Yearbook, provincial statistical yearbooks, and the EPS database. A tourism economic indicator system was constructed following principles of systematicity, scientific rigor, representativeness, and data availability, using sub-indicators proposed by Huang et al. (2024) suitable for the Chinese context (Table 1). Drawing on previous research (Gao & Li, 2024; Zhang & Guo, 2023), the core explanatory variable is a dummy for national-level big data pilot zones (0/1), with designated cities as the treatment group and others as the control group. To reduce bias in estimating tourism economic development, several control variables were included: per capita GDP (log), foreign investment ratio, government expenditure ratio, urban education spending, social consumption ratio, and scientific expenditure ratio (Table 2), capturing key economic, social, and technological factors influencing tourism development.

Table l.

Index System for Tourism Economy.

Target layer	Indicator layer	Indicator attribute
Tourism economy	R (Share of tourism revenue in GDP; %)	Positive
	R (Share of total tourism revenue in tertiary industry output; %)	Positive
	R (Inbound tourism revenue; CNY)	Positive
	R (Domestic tourism revenue; CNY)	Positive
	R (Total tourism revenue; CNY)	Positive
	R (Number of inbound tourists; 10,000 persons)	Positive
	R (Number of domestic tourists; 10,000 persons)	Positive
	R (Number of star-rated hotels; units)	Positive
	R (Highway passenger traffic; 10,000 persons)	Positive
	R (Length of highways; km)	Positive
	$R (Length of railways;$ km)	Positive

Table 2.

Variable Definitions and Indicator Descriptions (Including Fixed Effects).

Variables	Indicator
Dependent variable	Comprehensive tourism economic indicators
Independent variable	National big data comprehensive pilot zone
Control variables	$R (GDP)$
	$\ln (Ratio of general public budget expenditure to GDP + 1)$
	$\ln (Ratio of education expenditure to GDP + 1)$
	ln (Ratio of total retail sales of consumer goods to GDP + 1)
	$\ln (Ratio of actual utilized foreign direct investment to GDP + 1)$
	ln (Ratio of science and technology expenditure to GDP + 1)
Fixed effects	City
	Year

Data Processing Methods

Tourism economic development is measured using multiple sub-indicators reflecting tourism scale and urban development, normalized and weighted via the entropy method to ensure comparability (Liang et al., 2024). Entropy assigns higher weights to indicators with greater variation (Zhi-Hong et al., 2006). Ratio-type controls are log-transformed with a constant added to avoid undefined values. After merging into a panel dataset, missing values are interpolated, and multicollinearity is checked to ensure robust model estimation.

Research Methods

This study evaluates the impact of National Big Data Comprehensive Pilot Zones (NBDCPZs) on tourism economic development using a multi-level framework. Baseline regressions are conducted with mixed, fixed, and random effects models, selecting the optimal specification via the Hausman test. Time-lag models assess immediate, cumulative, and long-term policy effects. Propensity score matching addresses potential endogeneity, followed by regression on matched samples to ensure robustness. Spatial econometric models, including SLM and SDM, examine spatial spillovers, while heterogeneity and spatial quantile regressions capture variations across regions, city sizes, and tourism development levels. Mechanism analysis explores mediation pathways, and interaction terms test moderating effects of education and finance. All baseline, dynamic, spatial, heterogeneity, and mechanism analyses are consistently conducted using the pre-pandemic period (2009–2019) to ensure identification consistency and internal validity. This multi-method design comprehensively examines policy impacts from dynamic, spatial, distributional, and mechanism perspectives.

Benchmark Regression Model Specification

Benchmark regression is conducted using Pooled OLS, Fixed Effects (FE), and Random Effects (RE) models, with the Hausman test guiding model selection. The FE model, controlling for time-invariant city characteristics and common time effects, is preferred if the Hausman test is significant (Baltagi & Liu, 2016). Considering that the COVID-19 pandemic represents a nationwide structural shock that fundamentally altered tourism demand and supply conditions, the post-2019 period is excluded from the baseline analysis rather than being controlled for using a simple dummy variable. Robust standard errors are applied to address heteroskedasticity, and the regression equation is specified to estimate the average effect of big data pilot zones on tourism economic development.

Y_{it} = α_{0} + β {Treat}_{it} + \sum_{k} γ_{k} X_{it, k} + μ_{i} + λ_{t} + ε_{it}

(1)

$Y_{it}$ denotes the comprehensive tourism economic indicator for city i in year t. Treat_it is the dummy variable for the big data comprehensive pilot zone. $X_{it}$ represents the vector of control variables k. $μ_{i}$ indicates the city fixed effect, $λ_{t}$ represents the year fixed effect, and $ε_{it}$ is the random error term. If the Hausman test is not significant, a random effects model may be considered.

Parallel Trend Test and Event-Study Design

The validity of the multi-period DID approach relies on the parallel trend assumption, which requires that treated and control cities exhibit similar tourism development trajectories prior to policy implementation. To formally test this assumption, an event-study specification is employed. Specifically, the treatment indicator is interacted with a series of lead and lag dummy variables constructed based on the relative time distance between the observation year and the policy implementation year. The year immediately preceding policy implementation is set as the reference period and omitted from the regression. If the coefficients of the pre-treatment periods (e.g., t − 3 and t − 2) are statistically insignificant and fluctuate around zero, the parallel trend assumption can be reasonably satisfied, thereby supporting the causal interpretation of subsequent dynamic treatment effects.

Time-Lag Effect Model Specification

A distributed lag model is used to assess the dynamic effects of National Big Data Comprehensive Pilot Zones (NBDCPZs) on tourism economic development. Lagged terms capture immediate (lag1), cumulative (sum of lags), and long-term effects. The ARDL framework distinguishes short-term impacts within the same period from long-term equilibrium effects, enabling a comprehensive analysis of the policy’s sustained influence across different time horizons (Cho et al., 2023). Within the ARDL framework, short-term and long-term effects are distinguished: short-term effects capture the impact of changes in explanatory variables on the dependent variable within the same period, whereas long-term effects reflect the equilibrium relationship over an extended horizon (Brauer, 2013). Accordingly, this study calculates the immediate effect (lag1), cumulative effect (lag2), and long-term effect (lag3) to reveal the sustained influence of NBDCPZ policies on tourism economic development.

Endogeneity Treatment: Propensity Score Matching (PSM)

Since the designation of National Big Data Comprehensive Pilot Zones (NBDCPZs) is non-random and may be correlated with pre-existing city-level economic and institutional characteristics, baseline DID estimates could be subject to selection bias. To mitigate bias arising from observable factors, this study applies propensity score matching (PSM), using a logit model to estimate each city’s probability of being designated as a pilot zone based on pre-treatment covariates. One-to-one nearest neighbor matching with a 0.2 caliper ensures covariate balance, assessed via standardized mean differences, t-tests, and bias reduction rates. After matching, two-way fixed effects panel regressions are conducted on the matched sample to re-estimate the policy effect. Importantly, PSM-DID is treated as a robustness check rather than the primary identification strategy, which relies on the parallel trend assumption validated through event-study analysis. Besides, the matched sample is subsequently used for all baseline estimations and robustness checks.

Placebo Test and Anticipatory Effect

Although PSM-DID mitigates selection bias from observable characteristics, unobserved time-varying factors may still bias the estimates. To further strengthen causal identification, this study conducts placebo tests by assigning fictitious policy implementation years that precede the actual policy adoption. If the estimated effects are driven by spurious correlations or pre-existing trends, significant placebo effects would be observed. Conversely, statistically insignificant placebo coefficients would provide supportive evidence that the baseline results are not driven by anticipation or reverse causality.

Spatial Econometric Model Specification

Spatial autocorrelation of tourism economic indicators is tested using Moran’s I. If significant, a spatial weight matrix is constructed, and Spatial Lag (SLM; Lam & Souza, 2020) and Spatial Durbin (SDM) models (Wang et al., 2021) are estimated. SLM captures spatial dependence of the dependent variable, while SDM accounts for both explanatory variables and their spatial lags.

Y_{i t} = α_{0} + β {Treat}_{i t} + \sum_{k} γ_{k} X_{i t, k} + ρ W Y_{i t} + μ_{i} + λ_{t} + ε_{i t}

(2)

\begin{matrix} Y_{it} = α_{0} + β Trea t_{it} + τ W Trea t_{it} + \sum_{k} γ_{k} X_{it, k} \\ + ρ W Y_{it} + \sum_{k} γ_{k} θ W X_{it, k} + μ_{i} + λ_{t} + ε_{it} \end{matrix}

(3)

Where W denotes the spatial weight matrix, represents the spatial lag coefficient for policy variables, ρ denotes the spatial lag coefficient, and θ signifies the spatial lag coefficient for independent variables. By comparing the estimation results of SLM and SDM, we can test whether policy effects are spatially robust and identify the channels of spatial spillovers. During estimation, the aforementioned conventional control variables are similarly controlled, and robust standard errors are employed to address issues such as spatial heteroskedasticity.

Heterogeneity Analysis

Heterogeneity tests examine how policy effects vary by region, city size, and tourism development level. Following Yu et al. (2023), Cities are grouped by geographic region and population (mega to small), and region- or size-specific regressions are estimated. Spatial quantile regression, incorporating a spatial weight matrix, assesses differential policy impacts across the conditional distribution of tourism economic development, capturing effects on cities with low, medium, and high tourism revenues.

Q_{τ} Y_{it} = α_{0} + β_{τ} {Treat}_{it} + \sum_{k} γ_{k τ} X_{it, k} + ρ_{τ} W Y_{it} + μ_{i} + λ_{t} + ε_{it}

(4)

In the formula, τ denotes the τth quantile. By conducting the aforementioned heterogeneity test, regional, scale, and level differences in policy effects can be revealed, providing reference for policy optimization.

Mechanism Testing Model Specification

To examine how big data pilot zones influence tourism economic development, mediation analysis is conducted using industrial structure (tertiary-to-secondary output ratio) and digital infrastructure (postal and telecom service volume) as mediators. Stepwise regressions assess whether policy effects operate through these channels, with the Sobel test evaluating significance. A significant policy–mediator and mediator–tourism relationship indicates partial or full mediation, while insignificance suggests the path is not primary. To test moderation, interaction terms between the policy dummy and educational or financial variables—such as higher education enrollment and the ratio of financial deposits and loans to GDP—are included. Significant interaction coefficients reveal how regional education and financial capacity strengthen or weaken the policy’s impact on tourism, capturing variations in policy effectiveness across different institutional and economic environments.

Empirical Findings and Analysis

Benchmark Processing Effects of National Big Data Comprehensive Pilot Zones

Baseline regressions using POLS, FE, and RE models (Table 3) show the Hausman test favors RE, with robust standard errors applied. Column (3) of Table 3 indicates NBDCPZs significantly boost tourism development (coef. = .0074, 10%). GDP per capita and government expenditure also positively affect tourism, supporting H1. Although the Hausman test suggests the random-effects estimator is efficient in the baseline setting, fixed-effects specifications are consistently adopted in subsequent analyses to ensure conservative identification and comparability across robustness, spatial, and heterogeneity exercises.

Table 3.

Results of the Benchmark Regression Analysis.

Variables	Pooled ordinary least squares model	Fixed effects model	Random effects model
Variables	(1)	(2)	(3)
National big data comprehensive pilot zone	.0051 (.5935)	.0156*** (2.6825)	.0074* (2.0529)
GDP	1.0651*** (19.6203)	.5781*** (6.0667)	.7350*** (11.4248)
Ratio of general public budget expenditure to GDP	.0483* (1.8804)	.1087*** (3.1981)	.0916*** (4.3722)
Ratio of education expenditure to GDP	.0814*** (2.7553)	−.0233 (−.4317)	−.0056 (−.1842)
Ratio of total retail sales of consumer goods to GDP	.0004 (.0782)	.0056 (1.6089)	.0037 (1.3156)
Ratio of actual utilized foreign direct investment to GDP	.0929 (1.4100)	.0068 (.1245)	.0578 (1.3121)
Ratio of science and technology expenditure to GDP	−.0023* (−1.9345)	−.0019 (−.8815)	−.0020* (−1.6502)
Constant term	.0126** (2.2527)	.0268*** (4.1745)	.0231*** (4.1344)
City fixed effects	No	Yes	No
Year fixed effects	No	Yes	No
R ²	.7182	.2871	.3872

Note. Values in parentheses are z-statistics.

***

, **, and * indicates significance at the 1%, 5%, and 10% levels, respectively.

Parallel Trend Test and Event-Study Results

Figure 1 reports the event-study estimates examining the dynamic effects of the National Big Data Comprehensive Pilot Zones on tourism economic development. The coefficients for the pre-treatment periods (t − 3 and t − 2) are statistically insignificant and fluctuate closely around zero, indicating no systematic differences in tourism development trends between treated and control cities prior to policy implementation. This result supports the validity of the parallel trend assumption underlying the multi-period DID framework. The absence of significant pre-treatment coefficients indicates no detectable anticipation effects.

Figure 1.

Parallel trend test results.

Following policy implementation, the estimated coefficients exhibit short-term fluctuations, with a modest negative effect observed in the contemporaneous period (t = 0), and heterogeneous responses in subsequent years. These patterns suggest that the policy impact does not materialize instantaneously and may involve adjustment costs and delayed responses.

Dynamic Effects of the Policy

A distributed lag model assesses the dynamic effects of NBDCPZs on tourism economic development. Results show the first-lag coefficient is .0060 (5%), indicating a significant 1.28% increase in the first year, while the second-lag coefficient is .0062 (5%), suggesting a persistent but smaller effect. Notably, the third-lag coefficient further increases to .0178 and is statistically significant at the 1% level, indicating a cumulative and strengthening policy effect over time. Among control variables, GDP per capita exhibits a strong and positive effect, together with government expenditure and retail sales, highlighting the importance of overall economic scale and fiscal capacity in supporting tourism development. Education expenditure shows a weakly negative effect at the 10% significance level, while foreign direct investment and science and technology expenditure remain statistically insignificant. The model’s R² = .2421, indicating that the model’s R² equals .2421, indicating that the distributed lag specification explains a substantial proportion of the variation in tourism economic development (Table 4).

Table 4.

Analysis Results of the Time-Effect Model.

Variables	Distributed lag model
Variables	(4)
National big data comprehensive pilot zone _lag1	.0060** (2.2064)
National big data comprehensive pilot zone _lag2	.0062** (2.2807)
National big data comprehensive pilot zone _lag3	.0178*** (2.6963)
GDP	.5454*** (5.5355)
Ratio of general public budget expenditure to GDP	.0798*** (3.3467)
Ratio of education expenditure to GDP	−.1201* (−1.8234)
Ratio of total retail sales of consumer goods to GDP	.0090*** (2.7560)
Ratio of actual utilized foreign direct investment to GDP	.0617 (1.1015)
Ratio of science and technology expenditure to GDP	.0008 (.2950)
Constant term	.0441*** (5.5865)
R ²	.2421

Note. Values in parentheses are z-statistics.

***

, **, and * indicates significance at the 1%, 5%, and 10% levels, respectively.

Endogeneity Correction: PSM Estimation Results

Cities designated as NBDCPZs differ from non-designated cities in per capita GDP and financial development, potentially biasing estimates. Following the approaches of Zhi et al. (2021), and Shi et al. (2018), propensity score matching (PSM) is applied to mitigate selection bias. Table 5 shows baseline disparities are largely eliminated post-matching, with remaining standardized biases below 15%, validating the matched sample (Caliendo & Kopeinig, 2008). Regressions on matched samples (Table 6) show the NBDCPZ remains positive and statistically significant, though its magnitude slightly decreases from .0156 to .0133 after matching, confirming the robustness of the positive policy effect on tourism economic development. GDP per capita continues to exhibit a significant positive association, underscoring the persistent role of economic fundamentals in shaping urban tourism growth.

Table 5.

PSM Balance Test Results.

Variables	Stage	Mean		Standardized bias	t-Statistic	p-Value	Bias reduction rate (%)
Variables	Stage	Treatment group	Control group	Standardized bias	t-Statistic	p-Value	Bias reduction rate (%)
GDP	Before matching	.4903	−.0594	.3563	9.6262	.0000	83.52
	After matching	.1650	.0957	.0587	.7439	.4572	83.52
Ratio of general public budget expenditure to GDP	Before matching	.1708	−.0207	.2037	3.3101	.0009	84.40
	After matching	.1638	.1926	.0318	−.4025	.6874	84.40
Ratio of education expenditure to GDP	Before matching	−.1039	.0126	−.1281	−2.0109	.0444	−16.17
	After matching	−.0761	−.1889	.1488	1.8852	.0599	−16.17
Ratio of total retail sales of consumer goods to GDP	Before matching	−.0304	.0037	−.0385	−.5884	.5563	−6.79
	After matching	.0028	−.0353	.0412	.5215	.6022	−6.79
Ratio of actual utilized foreign direct investment to GDP	Before matching	.0451	−.0055	.0484	.8724	.3831	−29.04
	After matching	.0268	.0983	.0625	−.7952	.4289	−29.04
Ratio of science and technology expenditure to GDP	Before matching	−.0800	.0097	−.1196	−1.5474	.1219	−25.01
	After matching	−.0797	−.0874	.1496	1.8935	.0587	−25.01

Table 6.

Results of PSM-Based Regression Analysis.

Variables	(5)	(6)
National big data comprehensive pilot zone	.0156*** (2.6825)	.0133** (2.3557)
GDP	.5781*** (6.0667)	.0327** (2.2658)
Ratio of general public budget expenditure to GDP	.1087*** (3.1981)	.0120* (1.8108)
Ratio of education expenditure to GDP	−.0233 (−.4317)	−.0004 (−.0259)
Ratio of total retail sales of consumer goods to GDP	.0056 (1.6089)	−.0051 (−1.1924)
Ratio of actual utilized foreign direct investment to GDP	.0068 (.1245)	.0027 (1.3707)
Ratio of science and technology expenditure to GDP	−.0019 (−.8815)	.2652 (1.3615)
Constant term	.0268*** (4.1745)	.0880*** (5.9484)
City fixed effects	Yes	Yes
Year fixed effects	Yes	Yes
PSM	No	Yes
R ²	.2871	.3511

Note. Values in parentheses are z-statistics.

***

, **, and * indicates significance at the 1%, 5%, and 10% levels, respectively.

Placebo Test Results

To further assess whether the estimated effects are driven by spurious correlations or anticipation behavior, a placebo test is conducted by assigning a fictitious policy implementation year 2 years prior to the actual designation of National Big Data Comprehensive Pilot Zones. The regression results indicate that the placebo treatment coefficient is statistically insignificant and close to zero (Table 7), suggesting that the baseline estimates are unlikely to be driven by pre-existing trends or reverse causality.

Table 7.

Results of PSM-Based Regression Analysis.

Variables	(7)
Fake treat	−.0044 (.8548)
GDP	.4154*** (4.5440)
Ratio of general public budget expenditure to GDP	.2881 (1.5053)
Ratio of education expenditure to GDP	−.02195 (−.3935)
Ratio of total retail sales of consumer goods to GDP	−.0011 (−.7102)
Ratio of actual utilized foreign direct investment to GDP	.0538 (1.0932)
Ratio of science and technology expenditure to GDP	.0039* (1.8612)
Constant term	.0507*** (8.7805)
City fixed effects	Yes
Year fixed effects	Yes
PSM	Yes
R ²	.1298

Note. Values in parentheses are z-statistics.

***

, **, and * indicates significance at the 1%, 5%, and 10% levels, respectively.

Verification of Spatial Spillover Effects

Moran’s I statistics (.2077, significant at the 1% level) confirm the presence of significant spatial autocorrelation in tourism economic development, justifying the use of spatial econometric models. As shown in Table 8, the Spatial Lag Model (SLM, Column 8) indicates that the designation of National Big Data Comprehensive Pilot Zones (NBDCPZs) exerts a positive and statistically significant effect on local tourism economic development (coef. = .0113, 5%). Several control variables, including fiscal expenditure intensity and education expenditure, also exhibit marginal significance. The spatial lag term of the dependent variable (W_Y) is negative and significant, suggesting competitive or substitution effects among neighboring cities in tourism development.

Table 8.

Spatial Regression Model Analysis Results.

Variables	SLM	SDM
	(8)	(9)
National big data comprehensive pilot zone	.0113** (2.0786)	.0273* (1.9517)
GDP	−.0070 (−1.0051)	.0467* (1.7519)
Ratio of general public budget expenditure to GDP	.0058** (2.4221)	.0259 (1.4757)
Ratio of education expenditure to GDP	.0220* (1.6629)	−.0572 (−1.0511)
Ratio of total retail sales of consumer goods to GDP	−.0197* (−1.9684)	−.0001 (−.0060)
Ratio of actual utilized foreign direct investment to GDP	.0013 (1.0354)	.0035 (1.1731)
Ratio of science and technology expenditure to GDP	.0764 (1.2982)	.3975 (1.1208)
W_Y	−.0330** (−2.0885)	.1556** (2.1547)
W _NBDCPZ		.0067 (.7449)
W _GDP		−.0102** (−2.3171)
W _{Ratio of general public budget expenditure to GDP}		−.0046 (−1.1326)
W _{Ratio of education expenditure to GDP}		.01881 (1.5760)
W _{Ratio of total retail sales of consumer goods to GDP}		.0003 (.0313)
W _{Ratio of actual utilized foreign direct investment to GDP}		.0059(1.5971)
W _{Ratio of science and technology expenditure to GDP}		−1.2841 (−1.2693)
Constant term	.0893*** (15.3910)	−.0571 (−.6320)
City fixed effects	Yes	Yes
Year fixed effects	Yes	Yes
PSM	Yes	Yes
R ²	.2561	.2985

Note. Values in parentheses are z-statistics.

***

, **, and * indicates significance at the 1%, 5%, and 10% levels, respectively.

The Spatial Durbin Model (SDM, Column 9) further corroborates the positive policy effect of NBDCPZs, with a larger estimated coefficient (.0273, significant at the 10% level), indicating that accounting for spatially lagged explanatory variables amplifies the estimated local impact. In addition, the spatial lag of GDP per capita (WGDP) is significantly negative, implying that economic development in neighboring cities may crowd out local tourism growth. Although most spatially lagged covariates are statistically insignificant, the significant spatial autoregressive term (W_Y) highlights the existence of spatial dependence in tourism outcomes. Comparatively, the SDM achieves a higher explanatory power (R² = .2985 vs. .2561 in the SLM), suggesting that explicitly modeling spatial spillover channels provides a better representation of regional interdependencies in tourism economic development.

Heterogeneity Analysis of Policy Effects

Grouped regressions reveal pronounced spatial heterogeneity in the effects of NBDCPZs (Table 9). The policy effect is significantly positive in Northern China (coef. = .0243, t = 4.0731) and remains robust in Southern China (coef. = .0277, t = 2.3324), indicating that regions with stronger economic foundations and digital adoption capacity benefit more from the policy intervention. By contrast, the estimated effects in Central and Northwestern China are statistically insignificant, suggesting limited short-term tourism responses in these regions. Regarding control variables, GDP per capita exerts a consistently positive and significant effect in Central, Northern, and Southern regions, underscoring the fundamental role of economic development in shaping regional tourism performance. In Southern and Northwestern China, government budget expenditure shows a significant promoting effect, while education and science-related expenditures exhibit heterogeneous impacts across regions, reflecting regional differences in factor endowments and policy transmission pathways.

Table 9.

Analysis Results of Urban Samples Across Different Regions.

Variables	Full sample	Central	Northern	Northwestern	Southern
Variables	(10)	(11)	(12)	(13)	(14)
National big data comprehensive pilot zone	.0113** (2.0786)	.0049 (.8133)	.0243*** (4.0731)	.0011 (.1864)	.0277** (2.3324)
GDP	−.0070 (−1.0051)	.0382*** (8.9989)	.0198*** (4.1496)	.0058 (.7091)	.0420*** (3.4041)
Ratio of general public budget expenditure to GDP	.0058** (2.4221)	−.0090** (−2.1661)	.0062 (1.4346)	.0229*** (2.7530)	.0267*** (2.8355)
Ratio of education expenditure to GDP	.0220* (1.6629)	.0151 (1.4298)	.0181 (1.4528)	−.0270* (−1.8512)	−.1838*** (−6.6867)
Ratio of total retail sales of consumer goods to GDP	−.0197* (−1.9684)	—	—	—	—
Ratio of actual utilized foreign direct investment to GDP	.0013 (1.0354)	.0025 (.6211)	−.0055* (−1.7139)	−.0043 (−.9572)	.0025 (.6429)
Ratio of science and technology expenditure to GDP	.0764 (1.2982)	.0112 (1.6243)	−.0153 (−1.0692)	−.0158 (−1.6709)	.0127 (1.0714)
W_Y	−.0330** (−2.0885)	−.1845 (−.6399)	−.0045 (−.2399)	.0014 (.2159)	−.0454 (−.906)
Constant term	.0893*** (15.3910)	.0735*** (3.7007)	.0679*** (12.4948)	.0289*** (4.3655)	.0491*** (4.8001)
City fixed effects	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes
PSM	Yes	Yes	Yes	Yes	Yes
R ²	.2561	.5960	.4179	.1627	.3762

Note. Values in parentheses are z-statistics. “–” indicates that the corresponding variable is excluded to mitigate multicollinearity in regressions.

***

, **, and * indicates significance at the 1%, 5%, and 10% levels, respectively.

When grouped by urban resident population size (Table 10), the results reveal clear heterogeneity in policy effectiveness across city scales. In mega and extra-large cities (Columns 16 and 17), the estimated coefficients of the NBDCPZ policy are negative but statistically insignificant, suggesting that the tourism-enhancing effect of the policy is limited in highly developed metropolitan areas. By contrast, large cities (Column 18) exhibit a significantly positive policy effect (coefficient = .0171, t = 2.3447), indicating that cities at this scale are more responsive to big data policy interventions.

Table 10.

Analysis Results for Samples at Different City Levels.

Variables	Full sample	Mega	Extra-large	Large
Variables	(15)	(16)	(17)	(18)
National big data comprehensive pilot zone	.0113** (2.0786)	−.0012 (−.0455)	−.0015 (−.2325)	.0171** (2.3447)
GDP	−.0070 (−1.0051)	.1174*** (4.9423)	.0152*** (3.2425)	.0054 (.9928)
Ratio of general public budget expenditure to GDP	.0058** (2.4221)	.0188 (.9723)	.0144*** (2.9777)	.0095** (2.2194)
Ratio of education expenditure to GDP	.0220* (1.6629)	.0393 (1.0700)	−.0007 (−.0831)	.0128 (.7914)
Ratio of total retail sales of consumer goods to GDP	−.0197* (−1.9684)	—	−.0021 (−.2669)	−.0124 (−1.3916)
Ratio of actual utilized foreign direct investment to GDP	.0013 (1.0354)	.0192** (2.6110)	−.0006 (−.2672)	.0011 (.2905)
Ratio of science and technology expenditure to GDP	.0764 (1.2982)	−.1937** (−2.7806)	.0133* (1.8696)	.6136*** (2.6629)
W_Y	−.0330** (−2.0885)	.0454 (.9878)	−.0334 (−.4452)	.0828 (.9378)
Constant term	.0893*** (15.3910)	.2008*** (7.8315)	.0723*** (10.8465)	.1055*** (4.3537)
City fixed effects	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes
PSM	Yes	Yes	Yes	Yes
R ²	.2561	.7862	.1985	.1223

Note. Values in parentheses are z-statistics. “–” indicates that the corresponding variable is excluded to mitigate multicollinearity in regressions.

***

, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively.

Regarding control variables, GDP per capita is significantly positive in mega and extra-large cities, reflecting the dominant role of overall economic development in shaping tourism outcomes at higher urban tiers. Fiscal expenditure intensity remains positively significant in extra-large and large cities, while science and technology expenditure shows heterogeneous effects across city groups, highlighting scale-dependent differences in policy transmission mechanisms.

Quantile regression results reveal pronounced nonlinear effects of NBDCPZs across the distribution of tourism economic development. The policy effect is insignificant at the lower quantiles of .1 and .3 and at the upper quantile of 0.9, while it becomes positive and statistically significant at the median and higher middle quantiles, reaching its maximum at the .7 quantile with a coefficient of .0079 and a t-value of 2.7613. This pattern indicates that NBDCPZs exert the strongest promotional effect on cities with relatively high levels of tourism development, rather than on cities at the lower or extreme upper ends of the distribution. The spatial lag term WY remains highly significant across all quantiles and increases monotonically, indicating that spatial spillover effects strengthen as tourism development intensifies. In addition, science and technology expenditure exhibits evident threshold characteristics, remaining negative and statistically significant at the middle quantiles of .5 and .7, but turning strongly positive at the .9 quantile, which highlights the differentiated role of technological investment across stages of tourism development (Table 11).

Table 11.

Results of Quantile Regression Analysis.

Variables	Q0.1	Q0.3	Q0.5	Q0.7	Q0.9
Variables	(19)	(20)	(21)	(22)	(23)
National big data comprehensive pilot zone	−.0000 (−.0065)	.0021 (.8595)	.0045* (1.8793)	.0079*** (2.7613)	.0002 (.0436)
GDP	.0433*** (55.7722)	.0559*** (52.4185)	.0673*** (60.7991)	.0778*** (57.60)	.0864*** (49.4630)
Ratio of general public budget expenditure to GDP	.0038*** (2.9515)	.044*** (3.1306)	.0046*** (3.3854)	.0062*** (3.9211)	.0067** (2.4140)
Ratio of education expenditure to GDP	−.0003 (−.0950)	.0029 (1.1063)	.0004 (.1556)	.0023 (.7256)	.0146*** (2.9351)
Ratio of total retail sales of consumer goods to GDP	.0008 (.5189)	.0038 (1.3736)	.0109*** (4.9579)	.0102*** (4.3895)	−.0018 (−.4863)
Ratio of actual utilized foreign direct investment to GDP	.0006 (.5189)	−.0000 (−.0559)	−.0005 (−.3766)	−.0009 (−.6073)	.0001 (.0568)
Ratio of science and technology expenditure to GDP	.0213 (.5503)	−.0272 (−.8594)	−.0750*** (−2.8402)	−.0882*** (−2.6919)	.2602*** (7.3889)
W_Y	.1251*** (5.3006)	.2053*** (8.5752)	.2703*** (11.9292)	.3877*** (15.0812)	.8945*** (22.1133)
Constant term	.0198*** (4.9949)	.0265*** (7.3234)	.0294*** (9.1513)	.0340*** (8.8736)	.0701*** (14.2908)
City fixed effects	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes
PSM	Yes	Yes	Yes	Yes	Yes
Pseudo R²	.2949	.3732	.4115	.4002	.4048

Note. Values in parentheses are z-statistics.

***

, **, and * indicates significance at the 1%, 5%, and 10% levels, respectively.

Testing the Transmission Mechanisms of Policy Effects

Mediation analysis shows NBDCPZs partially affect tourism via industrial structure, significantly increasing the tertiary-to-secondary sector ratio (coef. = .1230, t = 2.8577) with the Sobel test confirming mediation (Z = 2.4091). When included jointly, both policy and industrial structure remain significant. In contrast, postal and telecommunications volume shows no mediating effect, with insignificant coefficients (coef. = .0287) and Sobel Z = .6820, indicating this channel does not transmit policy impacts (Table 12).

Table 12.

Results of Mediating Effect Tests.

Variables	Ratio of tertiary to secondary industry output		Postal and telecommunication services volume
	Ratio of tertiary to secondary industry output	Comprehensive index of tourism economy	Postal and telecommunication services volume	Comprehensive index of tourism economy
	(24)	(25)	(26)	(27)
National big data comprehensive pilot zone	.0245*** (4.4791)	.0222*** (4.4925)	.0089 (1.3550)	.0250*** (5.1134)
Ratio of tertiary to secondary industry output		.1230*** (2.8577)
Postal and telecommunication services volume				.0287 (.7893)
Constant term	.1168*** (37.7118)	.0534*** (9.3399)	.0341*** (9.2043)	.0669*** (22.1163)
City fixed effects	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes
PSM	Yes	Yes	Yes	Yes
Sobel test	Z = 2.4091		Z = .6820
R ²	.0456	.0780	.0044	.0614

Note. Values in parentheses are z-statistics.

***

, **, and * indicates significance at the 1%, 5%, and 10% levels, respectively.

Moderation analysis reveals heterogeneous conditioning effects of human capital and financial development on NBDCPZ outcomes. As shown in Table 13, the interaction between NBDCPZs and higher education enrollment is positive and statistically significant (coef. = .3495, t = 2.1922), indicating that stronger human capital endowment enhances the tourism effects of the policy. In contrast, although financial development exhibits a positive direct association with tourism growth, its interaction with NBDCPZs is statistically insignificant (coef. = −.2472, t = −.8771), suggesting that financial depth does not significantly amplify policy impacts. Overall, the results indicate that human capital plays a more effective moderating role than financial development in shaping the tourism effects of NBDCPZs.

Table 13.

Results of Moderation Effect Tests.

Variables	(28)	(29)	(30)	(31)
National big data comprehensive pilot zone	.0252*** (5.1624)	.0130* (1.7478)	.0226*** (4.4737)	.0272*** (3.7306)
Number of students enrolled in regular higher education institutions	.0005 (.0052)	−.2822* (−1.8133)
National big data comprehensive pilot zone × number of students enrolled in regular higher education institutions		.3495** (2.1922)
Ratio of total deposits and loans of financial institutions to regional GDP			.7217* (1.9134)	.9368** (2.0819)
National big data comprehensive pilot zone × ratio of total deposits and loans of financial institutions to regional GDP				−.2472 (−.8771)
Constant term	.0678*** (16.0633)	.0777*** (12.5671)	.0685*** (10.4673)	.0489*** (5.1724)
City fixed effects	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes
PSM	Yes	Yes	Yes	Yes
R ²	.0600	.0707	.0682	.0699

Note. Values in parentheses are z-statistics.

***

, **, and * indicates significance at the 1%, 5%, and 10% levels, respectively.

Further Robustness Checks and Accounting for the COVID-19 Shock

Although the baseline specification already controls for common time shocks through year fixed effects, the COVID-19 pandemic represents an unprecedented external disruption that may have altered both tourism demand and the effectiveness of digital governance policies. To ensure that the estimated impact of the National Big Data Comprehensive Pilot Zone (NBDCPZ) policy is not confounded by pandemic-related shocks, we conduct an additional robustness check by explicitly modeling the differential policy response during the COVID-19 period.

Specifically, we introduce an interaction term between the NBDCPZ policy indicator and a COVID-19 dummy variable that equals one for the years 2020 to 2021 and zero otherwise. The empirical specification is given by:

\begin{matrix} Y_{it} = β_{1} {NBDCPZ}_{it} + β_{2} ({NBDCPZ}_{it} \times {COVID}_{t}) \\ + X_{it} + μ_{i} + λ_{t} + ε_{it} \end{matrix}

(5)

where $Y_{it}$ denotes tourism economic development in city i and year t; ${NBDCPZ}_{it}$ is the policy treatment indicator; COVID_t is a time dummy for the pandemic period; $X_{it}$ is a vector of control variables; and $μ_{i}$ and $λ_{t}$ represent city and year fixed effects, respectively.

It is important to note that the COVID-19 dummy is not included as a standalone regressor, as its average effect across cities is fully absorbed by year fixed effects. Instead, the interaction term captures whether the tourism impact of the NBDCPZ policy differs between pilot and non-pilot cities during the pandemic period. This specification follows the logic of a difference-in-differences interaction embedded within a fixed-effects framework and allows us to assess the resilience or vulnerability of data-driven tourism governance under extreme external shocks. This approach does not alter the baseline identification strategy but serves as a robustness check to verify that the estimated policy effect is not driven by pandemic-induced structural breaks.

The results are reported in Table 14. After accounting for the COVID-19 period, the coefficient of the National Big Data Comprehensive Pilot Zone (NBDCPZ) policy remains positive and statistically significant (coef. = .0023, p < .05), indicating that the baseline tourism-promoting effect persists in pilot cities. The interaction term between NBDCPZ and the COVID-19 dummy is significantly negative, suggesting that the policy effect was temporarily attenuated during the pandemic years. This finding reflects the strong external shock imposed by COVID-19 on tourism activities rather than a reversal of the underlying policy effect. Overall, the results confirm that the main conclusions are robust to the inclusion of pandemic-related disturbances. Accordingly, the baseline analysis focuses on the pre-pandemic period to ensure internal validity, while the COVID interaction specification is implemented solely as an auxiliary robustness check rather than an alternative baseline.

Table 14.

Results of COVID-19 Robustness Test.

Variables	(5)
National big data comprehensive pilot zone	.0023** (2.4987)
GDP	.0292* (1.6570)
Ratio of general public budget expenditure to GDP	.0067** (2.3347)
Ratio of education expenditure to GDP	−.0270 (−.9006)
Ratio of total retail sales of consumer goods to GDP	−.0129 (−1.3043)
Ratio of actual utilized foreign direct investment to GDP	.0028 (1.2299)
Ratio of science and technology expenditure to GDP	.1744 (1.4566)
National big data comprehensive pilot zone × COVID	−.3708(−13.9950)
Constant term	.3776*** (8.1562)
City fixed effects	Yes
Year fixed effects	Yes
PSM	Yes
R ²	.3712

Note. Values in parentheses are z-statistics.

***

, **, and * indicates significance at the 1%, 5%, and 10% levels, respectively.

Alternative Interpretations of Scale-Related Effects

The baseline heterogeneity analysis reveals that the tourism-promoting effect of the NBDCPZ policy is insignificant or even attenuated in super-large cities. While this pattern is consistent with the notion of diminishing marginal returns and coordination costs in highly developed metropolitan areas, it may also reflect a nonlinear relationship between city scale and the effectiveness of digital policy interventions. To explore this possibility and to rule out alternative interpretations based on discrete city-size classifications, we conduct an additional robustness check by introducing a nonlinear scale term into the baseline regression. Specifically, we augment the main specification by including both the logarithm of city population and its squared term, allowing the marginal effect of city size to vary along the scale distribution:

\begin{matrix} Y_{it} = β_{1} {NBDCPZ}_{it} + β_{2} \ln ({Population}_{it}) \\ + β_{3} [\ln ({Population}_{it})]^{2} + X_{it} + μ_{i} + λ_{t} + ε_{it} \end{matrix}

(6)

In this framework, a statistically significant negative coefficient on the squared term ( $β_{3} < 0$ ) would indicate a concave relationship between city size and tourism economic development, consistent with scale-related diseconomies. Such a pattern would suggest the existence of a threshold beyond which further urban expansion weakens the marginal contribution of digital policy initiatives, potentially due to rising coordination costs, institutional complexity, or saturation of data-driven tourism services. Importantly, this specification does not aim to redefine the core treatment effect of the NBDCPZ policy, nor does it require re-estimation of all heterogeneity analyses. Rather, it provides an alternative and continuous interpretation of the scale-related findings, complementing the discrete city-size subgroup results reported earlier. By demonstrating that the attenuation of policy effects in super-large cities is consistent with an underlying nonlinear scale mechanism, this robustness check strengthens the theoretical plausibility of the observed heterogeneity patterns.

The estimation results indicate that neither the linear population term nor its squared term is statistically significant, although their coefficients display opposite signs. Specifically, the coefficient on the logarithm of population is negative (β = −2.059, p = .216), while the squared term is positive (β = 2.074, p = .214), suggesting no empirically identifiable nonlinear scale relationship within the sample period. Although the direction of the estimates is consistent with a potential concave association, the lack of statistical significance implies that no clear population threshold or turning point can be established. Importantly, core economic fundamentals such as GDP remain statistically significant (β = .048, p < .01), indicating that the insignificance of the scale terms is unlikely to stem from model misspecification. Taken together, these findings suggest that the attenuated policy effects observed in super-large cities are better interpreted as reflecting structural maturity and absorptive capacity constraints rather than a mechanically nonlinear population effect. Therefore, the absence of statistical significance indicates that scale-related attenuation is more plausibly driven by institutional saturation rather than a mechanical population-size nonlinearity.

Conclusions and Implications

Conclusions

Based on panel data from 284 prefecture-level cities in China during 2009 to 2019, this study employs multi-period difference-in-differences (DID), propensity score matching (PSM), and spatial Durbin models (SDM) to systematically examine the effects and mechanisms of the National Big Data Comprehensive Pilot Zone (NBDCPZ) policy on tourism economic development. The empirical results demonstrate that the NBDCPZ policy exerts a significant and robust positive effect on urban tourism development, which remains stable after a series of endogeneity corrections, spatial dependence tests, and robustness checks. Additional robustness analyses accounting for the COVID-19 shock further confirm that the baseline policy effect is not driven by pandemic-induced disruptions.

First, the dynamic analysis reveals that the tourism effects of NBDCPZs exhibit a lagged and cumulative pattern over time. The policy impact becomes statistically significant after implementation and continues to strengthen in subsequent periods, indicating that the effectiveness of big data–driven governance in tourism relies on gradual processes of institutional adaptation, data integration, and industrial adjustment rather than instantaneous responses. This finding highlights the importance of sustained policy support in unlocking the long-term dividends of digital transformation. The spatial econometric results suggest that spatial dependence in tourism development mainly operates through market competition rather than policy diffusion. While the significant spatial autoregressive term indicates that tourism outcomes in one city are affected by neighboring cities, the insignificant spatial lag of the NBDCPZ variable implies that policy designation itself does not generate cross-city spillovers. Combined with the negative spatial effect of neighboring GDP, these findings indicate that adjacent cities primarily interact through competitive substitution in tourism markets, rather than through cooperative or emulative policy channels.

Second, the policy effect displays pronounced multidimensional heterogeneity. From a regional perspective, the impact is significantly positive in Northern and Southern China, while remaining statistically insignificant in Central and Northwestern regions. These differences reflect disparities in economic foundations, digital infrastructure conditions, and the capacity to absorb data-driven policy instruments. Regarding city size, the results show that large cities benefit significantly from the NBDCPZ policy, whereas the estimated effects in mega and extra-large cities are insignificant. This pattern reflects the fact that tourism development and data infrastructure in super-large cities are already highly mature and stable, such that the designation of an additional national data pilot zone does not constitute a binding or transformative institutional shock. In this context, the policy effect is diluted not because digitalization becomes counterproductive, but because baseline capacities and tourism demand are already saturated. In this sense, the attenuated policy effect in super-large cities reflects saturation and institutional redundancy rather than a reversal of the digitalization–tourism relationship. Furthermore, distributional evidence from quantile regressions indicates that the policy effect is most pronounced at medium-to-high levels of tourism development, while remaining weak or insignificant at the lower and upper extremes of the distribution, underscoring the nonlinear nature of digital policy impacts.

Third, mechanism analysis identifies industrial structure optimization as the core transmission channel through which the NBDCPZ policy promotes tourism economic development. The policy significantly increases the ratio of tertiary to secondary industry output, and mediation tests confirm that industrial upgrading plays a partial but substantive role in transmitting policy effects. It should be emphasized that the identified mediating role of industrial structure optimization reflects a macro-level structural transmission channel, rather than a direct observation of firm- or consumer-level behavioral mechanisms. Micro-level data would be required to further unpack the behavioral foundations of this structural effect. By contrast, the mediating role of information infrastructure (measured by postal and telecommunication business volume) is insignificant, suggesting that basic connectivity expansion cannot capture the deeper data-driven innovation (Van Nuenen & Scarles, 2021), precision marketing (Li et al., 2022), and tourist analytics emphasized by the policy (Zhang & Cheng, 2024), reflecting a stage-of-development distinction. Basic connectivity indicators capture infrastructure conditions that have largely converged across cities during the sample period, whereas the NBDCPZ policy operates on higher-order data governance and analytics capabilities that remain unevenly distributed and are directly relevant for decision-making optimization rather than mere information transmission within the tourism sector.

Finally, moderation analysis reveals heterogeneous conditioning effects across enabling factors. Human capital endowment significantly amplifies the tourism-promoting effect of the NBDCPZ policy, indicating that regions with stronger higher education capacity are better positioned to translate data resources into tourism productivity gains. In contrast, although financial development exhibits a positive direct association with tourism growth, its interaction with the NBDCPZ policy is statistically insignificant, suggesting that financial depth does not systematically strengthen policy effectiveness once data-driven mechanisms are in place. This pattern implies that, at the city level, financial resources may primarily support general tourism expansion rather than the specific data integration, analytics, and governance capacities targeted by the NBDCPZ policy. When core data infrastructures and institutional arrangements are already determined by policy designation, additional financial deepening alone is insufficient to further amplify marginal policy effects.

Contributions and Policy Implications

The marginal contributions of this study are twofold. First, through systematic empirical analysis, it establishes a direct nexus between the National Big Data Comprehensive Pilot Zone (NBDCPZ) policy and tourism economic development, identifying industrial structure optimization and human capital conditions as the core mediating mechanism. This finding advances the theoretical understanding of how data elements reshape the tourism value chain through sectoral upgrading and capability enhancement mechanisms. Second, by integrating multi-period DID, spatial Durbin, and quantile regression models within a multidimensional analytical framework, the study captures the dynamic, spatial, and distributional characteristics of policy effects, offering a methodological paradigm for evaluating digital economy policies. Nevertheless, as an initial exploration of the NBDCPZ’s impact on the tourism economy, this research has limitations that merit further refinement. The depth of micro-level mechanism analysis remains limited: although the mediating effect of industrial upgrading is verified, the micro-dynamics of data-driven value chain restructuring require finer-grained investigation. Future research should enhance analytical precision through micro-level data collection. Moreover, the data dimension is constrained by reliance on traditional statistical indicators, lacking real-time measures of digital consumption behavior. Subsequent studies should incorporate multi-source dynamic data to construct a tourism digital vitality index and capture the time-varying impacts of digital technology on market recovery, thereby providing stronger empirical support for policy innovation.

Based on the above findings, this study proposes a three-dimensional framework for optimizing the effectiveness of the National Big Data Comprehensive Pilot Zone (NBDCPZ) policy in promoting tourism economic development. First, to address the lagged and cumulative nature of policy effects, a multidimensional monitoring system should be established to track core indicators—such as tourism economic elasticity and industrial transformation rates—in real time, thereby identifying inflection points in policy effectiveness. For pilot zones with implementation cycles exceeding 2 years, third-party evaluations should be conducted, and a policy resource elasticity allocation model should be developed to ensure the precise alignment of policy instruments with regional development stages. Second, regions with stronger economic foundations and digital absorption capacity should emphasize technological integration, innovation, and market-oriented data allocation, including the construction of cross-regional cultural and tourism collaboration platforms. Regions with relatively weaker short-term policy responses should prioritize the deployment of digital infrastructure hubs and talent cultivation systems, supported by interregional transfer mechanisms to strengthen foundational capacities. Regarding city size, large cities should foster smart tourism industrial clusters, whereas mega and extra-large cities should mitigate scale-related coordination inefficiencies through computational resource redistribution and spatial reconfiguration to enhance technology diffusion and resource sharing. Finally, to deepen the core transmission pathway of industrial structure optimization, the establishment of market-oriented platforms for tourism data elements should be accelerated, promoting the full-chain digital transformation of cultural and tourism enterprises and expanding the modern service sector. In addition, the complementary role of human capital should be strengthened by improving the alignment between higher education training and the practical needs of data-driven tourism development, thereby enhancing cities’ capacity to translate digital policies into sustained tourism growth.

In summary, this study empirically demonstrates the promotional role of national big data comprehensive pilot zones in tourism economies and clarifies the heterogeneous conditions under which such effects are most likely to materialize. It provides theoretical underpinnings and policy targets for empowering high-quality tourism development through national digital policies. Future research should deepen micro-level mechanism analysis and multi-source data integration to foster the vigorous development of new tourism productive forces within China’s modernization process.

Footnotes

ORCID iDs

Xu Lele

Author Contributions

Conceptualization: XL, YQ, and ZH-X; Methodology: XL and YQ; Software: XA; Formal analysis: XL, YQ, and ZH-X; Resources: YQ and ZH-X; Writing – original draft preparation: XL and YQ; Writing – review and editing: YQ and ZH-X; Visualization: XA. All authors have read and agreed to the published version of the manuscript.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by the Fujian Provincial Social Science Fund, grant number FJ2024C088.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The datasets used and/or analyzed in the current study are available from the corresponding author upon reasonable request.*

References

Afolabi

J. A.

(2023). Advancing digital economy in Africa: The role of critical enablers. Technology in Society, 75, Article 102367. https://doi.org/10.1016/j.techsoc.2023.102367

Baltagi

B. H.

Liu

(2016). Random effects, fixed effects and Hausman’s test for the generalized mixed regressive spatial autoregressive panel data model. Econometric Reviews, 35(4), 638–658. https://doi.org/10.1080/07474938.2014.998148.

Brauer

M. F.

(2013). The effects of short-term and long-term oriented managerial behavior on medium-term financial performance: Longitudinal evidence from Europe. Journal of Business Economics and Management, 14(2), 386–402. https://doi.org/10.3846/16111699.2012.703965

Caliendo

Kopeinig

(2008). Some practical guidance for the implementation of propensity score matching. Journal of Economic Surveys, 22(1), 31–72. https://doi.org/10.1111/j.1467-6419.2007.00527.x

Chen

Cai

(2025). Does the digital economy promote tourism eco-efficiency? – An empirical study based on Chinese cities. Polish Journal of Environmental Studies, 34(3), 2063–2079. https://doi.org/10.15244/pjoes/187608

Chen

Teng

Chen

(2022). How does FinTech affect the development of the digital economy? Evidence from China. The North American Journal of Economics and Finance, 61, Article 101697. https://doi.org/10.1016/j.najef.2022.101697

Chen

Y. Q.

Zhang

X. Y.

(2023). The impact mechanism of digital economy development on regional industrial structure optimization: Empirical analysis based on provincial panel data from 2011–2019. Collected Essays on Finance and Economics, 39(4), Article 14.

Cho

J. S.

Greenwood-Nimmo

Shin

(2023). Recent developments of the autoregressive distributed lag modelling framework. Journal of Economic Surveys, 37(1), 7–32. https://doi.org/10.1111/joes.12450

Choi

C. H.

Zhou

J. O.

(2023). Can human capital drive sustainable international trade? Evidence from BRICS countries. Sustainability, 16(1), Article 135.

10.

Gannon

Taheri

(2020). Marketing for tourism, hospitality and events: A global and digital approach. Tourism Analysis, 25(1), 191–193.

11.

Gao

Y. T.

(2024). The mechanism and realization path of big data pilot zone construction in promoting green economic development. Journal of South-Central University for Nationalities (Natural Science Edition), 43(2), 280–288. https://doi.org/10.20056/j.cnki.ZNMDZK.20240219

12.

Guo

Geng

Yao

Zhao

L. X.

(2024). Can digital infrastructure enhance economic efficiency? Evidence from China. Quality & Quantity, 58(2), 1729–1752. https://doi.org/10.1007/s11135-023-01710-y

13.

Guo

X. Y.

Ming

Q. Z.

(2025). The interactive relationship between digital inclusive finance and tourism industry efficiency: Evidence from Guizhou Province. Human Geography, 40(1), 80–88. https://doi.org/10.13959/j.issn.1003-2398.2025.01.009

14.

Huang

Z. X.

Shi

S. H.

(2024). Coordinated effects and barrier factors between China’s tourism economy and ecosystem services. Tourism Tribune, 39(4), 93–108. https://doi.org/10.19765/j.cnki.1002-5006.2024.04.012

15.

Jedwab

Kerby

Moradi

(2017). History, path dependence and development: Evidence from colonial railways, settlers and cities in Kenya. The Economic Journal, 127(603), 1467–1494.

16.

Lam

Souza

P. C. L.

(2020). Estimation and selection of spatial weight matrix in a spatial lag model. Journal of Business & Economic Statistics, 38(3), 693–710. https://doi.org/10.1080/07350015.2019.1569526

17.

Liang

Rui

(2022). Research on the integration of cultural tourism industry driven by digital economy in the context of COVID-19-based on the data of 31 Chinese provinces. Frontiers in Public Health, 10, Article 780476. https://doi.org/10.3389/fpubh.2022.780476

18.

B. W.

Wei

(2014). Analysis and guidance on the development of the new tourism economy theory. Academic Forum, 37(6), 76–79. https://doi.org/10.16524/j.45-1002.2014.06.033

19.

Liang

Yan

Zhang

Q. P.

(2024). Exploring the provision, efficiency and improvements of urban green spaces on accessibility at the metropolitan scale. Journal of Environmental Management, 352, Article 120118. https://doi.org/10.1016/j.jenvman.2024.120118

20.

Liu

Y. J.

Han

Y. J.

(2020). Factor structure transformation, institutional environment, and high-quality development of tourism economy. Tourism Tribune, 35(3), 28–38. https://doi.org/10.19765/j.cnki.1002-5006.2020.03.008

21.

Lü

Y. W.

Wang

W. Q.

Zhang

J. N.

(2023). The impact and transmission mechanism of digital economy on regional green innovation efficiency. Statistics and Decision, 39(20), 120–124. https://doi.org/10.13546/j.cnki.tjyjc.2023.20.022

22.

Nguyen

T. T. T.

Choi

C. H.

(2025). Assessing the impact of ICT on export performance: A comparative analysis of ASEAN-5 and partner countries. Review of Development Economics, 29(3), 1834–1848.

23.

Park

N. R.

Yun

H. S.

Choi

C. H.

(2024). Green trade and cultural innovation: Examining the impact on GTFP and greenhouse gas emissions in OECD countries. Sustainability, 16(19), Article 8339.

24.

Park

Choi

C. H.

(2025). The role of ICT in shaping international trade dynamics: A G20 perspective. Technological and Economic Development of Economy, 31(6), 1781–1800. https://doi.org/10.3846/tede.2025.23168

25.

Phaosathianphan

Leelasantitham

(2021). An intelligent travel technology assessment model for destination impacts of tourist adoption. Tourism Management Perspectives, 40, Article 100882. https://doi.org/10.1016/j.tmp.2021.100882

26.

Shi

D. Q.

Ding

Wei

Liu

J. J.

(2018). Can smart city construction reduce environmental pollution? China Industrial Economics, 35(6), 117–135. https://doi.org/10.19581/j.cnki.ciejournal.2018.06.008,

27.

Song

Wang

Bergmann

(2020). China’s prefectural digital divide: Spatial analysis and multivariate determinants of ICT diffusion. International Journal of Information Management, 52, Article 102072. https://doi.org/10.1016/j.ijinfomgt.2020.102072

28.

Van Nuenen

Scarles

(2021). Advancements in technology and digital media in tourism. Tourist Studies, 21(1), 119–132. https://doi.org/10.1177/1468797621990410

29.

Wang

Cui

Zhao

(2021). Effect of green technology innovation on green total factor productivity in China: Evidence from spatial Durbin model analysis. Journal of Cleaner Production, 288, Article 125624. https://doi.org/10.1016/j.jclepro.2020.125624

30.

Wang

J. H.

Xia

J. C.

(2017). Spatial correlation structure and its effects of regional tourism economy in China: Based on social network analysis. Tourism Tribune, 32(7), 15–26.

31.

Wang

Gan

(2023). Spatial spillover effects of industrial structure optimization and upgrading on tourism efficiency in China. Areal Research and Development, 42(4), 7–13.

32.

Wei

H. X.

Huang

H. Y.

(2022). Tourism economic resilience and high-quality development under the pandemic. Tourism Tribune, 37(9), 5–7. https://doi.org/10.19765/j.cnki.1002-5006.2022.09.003

33.

Wei

(2022). Digital tourism: A new model of China’s tourism economic development. Tourism Tribune, 37(4), 10–11. https://doi.org/10.19765/j.cnki.1002-5006.2022.04.006

34.

Wei

D. C.

(2025). Tourism economy, human capital compensation, and regional growth. Tourism Tribune, 40(1), 150–171. https://doi.org/10.19765/j.cnki.1002-5006.2025.01.014

35.

Siddik

A. B.

Sobhani

F. A.

Rahman

M. M.

(2024). Driving economic success: Fintech, tourism, FDI, and digitalization in the top 10 tourist destinations. Humanities and Social Sciences Communications, 11(1), Article 1549. https://doi.org/10.1057/s41599-024-03978-3

36.

Yin

Z. H.

Choi

C. H.

(2023). Does digitalization contribute to lesser income inequality? Evidence from G20 countries. Information Technology for Development, 29(1), 61–82. https://doi.org/10.1080/02681102.2022.2123443

37.

Leichtle

Zhang

Liu

Wang

Xue

Taubenböck

(2023). Does urban growth mean the loss of greenness? A multi-temporal analysis for Chinese cities. Science of the Total Environment, 898, Article 166373. https://doi.org/10.1016/j.scitotenv.2023.166373

38.

Yue

Zhou

Liu

(2024). How does green finance influence industrial green total factor productivity? Empirical research from China. Energy Reports, 11, 914–924. https://doi.org/10.1016/j.egyr.2023.12.056

39.

Yun

Park

Choi

C. H.

(2025). Behavioral insights into audience satisfaction: Analyzing emotional and cognitive factors in K-culture dance engagement. PLoS One, 20(12), Article e0337527. https://doi.org/10.1371/journal.pone.0337527

40.

Zhang

Cheng

(2024). Spatio-temporal effects and influence mechanism of digital technology on tourism efficiency in Chinese provinces. Scientific Reports, 14(1), Article 22975. https://doi.org/10.1038/s41598-024-74367-8

41.

Zhang

Yin

Z. H.

Choi

C. H.

(2025). Industrial transformation and environmental sustainability: The case of China’s “Made in China 2025” policy. Sage Open, 15(4), Article 21582440251382556.

42.

Zhang

Choi

C. H.

(2025). Will digital trade be friend or foe of the green economy? Unveiling the complexities of green growth. Journal of Applied Economics, 28(1), Article 2464591. https://doi.org/10.1080/15140326.2025.2464591

43.

Zhang

Y. H.

Guo

X. H.

(2023). Big data development and enterprise total factor productivity: Empirical evidence from national big data comprehensive pilot zones. Industrial Economics Research, (2), 69–82. https://doi.org/10.13269/j.cnki.ier.2023.02.009

44.

Zhao

Zhu

Y. N.

Zhao

(2024). National big data comprehensive pilot zones and new productive forces: Empirical evidence from 230 Chinese cities. Journal of Chongqing University (Social Sciences Edition), 30(4), 62–78.

45.

Zhi

Y. P.

Huang

L. Q.

Chen

(2021). The construction of free trade pilot zones and regional industrial structure upgrading: An empirical analysis based on panel data of 286 Chinese cities. South China Journal of Economics, 40(4), 37–54. https://doi.org/10.19592/j.cnki.scje.381305

46.

Zhi-Hong

Jing-Nan

(2006). Entropy method for determination of weight of evaluating indicators in fuzzy synthetic evaluation for water quality assessment. Journal of Environmental Sciences, 18(5), 1020–1023. https://doi.org/10.1016/S1001-0742(06)60032-6

47.

Zhou

Wen

Yang

(2021). Agglomerating or dispersing? Spatial effects of high-speed trains on regional tourism economies. Tourism Management, 87, Article 104392. https://doi.org/10.1016/j.tourman.2021.104392

48.

Zhou

Choi

C. H.

(2025). Tourism development, foreign direct investment, and environmental sustainability: How green technology innovations make a difference. SAGE Open, 15(2), Article 21582440251331003.