Tourism in pandemic: the role of digital travel vouchers in China

The study investigates whether and how government-issued digital travel vouchers can spur domestic tourism recovery in China during COVID-19. Tourism—which supports growth, employment, and poverty reduction—was severely disrupted by lockdowns and risk perceptions. Despite policy efforts globally, uncertainties and psychological barriers persist, particularly in China with stricter controls. Grounding the analysis in the Theory of Planned Behavior (TPB), the authors hypothesize that vouchers can influence intentions and behavior by shifting perceptions and perceived control. They also consider localization dynamics and regional heterogeneity in China, motivating an examination of spatial spillovers. The research questions are: Do digital travel vouchers increase tourist inflows? Which voucher characteristics enhance effectiveness? Do vouchers generate spatial spillovers across neighboring cities? The paper contributes by assembling novel city-month data on traveler flows and voucher issuance, examining digital issuance features, and modeling spatial effects.

The authors develop three hypotheses based on economic theory, consumer behavior, and digitization. Tourism stimulus policies are justified when markets face shocks; demand fell not only due to income loss but also perceived health risks. TPB suggests behavioral, normative, and control beliefs shape intentions and actions; government vouchers can rebuild confidence and stimulate demand (H1). Prior work on vouchers shows mixed results influenced by design; digitization in China (mobile payments, platforms, online booking) enhances usability and information advantages of digital vouchers, implying heterogeneous effectiveness by platform credibility, issuance frequency, and digital inclusiveness (H2). Given city-level issuance in a geographically large, heterogeneous country, and the tendency for shorter-distance travel post-pandemic, digital vouchers may transcend administrative borders and produce spatial spillovers in nearby cities (H3). The literature also points to regional imbalances, infrastructure, and digital finance as moderators of tourism outcomes.

Data and variables: - Dependent variable: Tourist inflows proxied by Baidu Migration Index population inflows (monthly average), collected via Python web-crawling of Baidu Migration for 306 cities from Sep 2020 to Dec 2021 (16 months). Random manual checks on 50 cities validate the crawler data. - Core explanatory variable: City-month value of digital travel vouchers, manually collected via Baidu search of official and third-party announcements (e.g., validity, value, participating businesses, platforms). In total, 171 issuance events across 69 cities; issuance is seasonal with peaks in October. - Voucher characteristics: Issuing platform (1 = large platforms such as Alipay/Ctrip; 0 = WeChat Mini Apps); issuance frequency (order of each issuance, logged); publicity (logged count of Baidu news articles on the campaign); availability (logged number of participating businesses). - Controls: City population, per capita GDP, number of high-speed rail lines, municipal utility site area, fixed-asset investment in urban appearance, confirmed COVID-19 cases; internet penetration (broadband access users per 100 people) as a proxy for digital inclusiveness. Econometric models: - Baseline effect: Poisson Pseudo-maximum Likelihood (PPML) with province and month fixed effects to handle heteroscedasticity, measurement error, and zeros. Tour_Arrival_it = exp(α lnVALUE_it + Γ lnControls_it + η_i + ν_t + ε_it). - Treatment effect: Staggered Difference-in-Differences (DID) using a voucher issuance indicator to compare issuing vs. non-issuing cities pre/post first issuance, with province and month fixed effects. - Extended models: Interaction terms to test moderating effects of platform, issuance frequency, and internet penetration. Variables are demeaned (centralized) to ensure interpretability and comparability of interaction coefficients. - Spatial spillovers: Spatial Durbin Model (log-linear) with spatially lagged dependent and explanatory variables, using (i) geographical adjacency matrix W and (ii) a nested matrix W^de that multiplies inverse geographic distance and inverse GDP-per-capita difference to incorporate economic similarity. Precondition assessed via Moran’s I tests for spatial autocorrelation in monthly tourist inflows. - Robustness: Alternative voucher intensity proxies (publicity, availability); propensity score matching with staggered DID; Callaway and Sant’Anna (2021) CSDID for multiple treatment timing; instrumental variables (2SLS and GMM) using as instrument the one-period lagged sum of voucher values issued by other cities in the same province; parallel trends and placebo tests. Sample sizes: Full city-month panel (up to 4,896 observations); subsamples exclude extreme months or major holidays; spatial models use city-months with complete covariates.

- Baseline effects: Digital travel vouchers significantly increase tourist inflows. In PPML, lnVALUE has a positive and significant coefficient (e.g., 0.090, p<0.01). The staggered DID indicates issuance increases inflows by about 0.078 (significant at 5%). Effects remain across subsamples excluding extreme months/holidays and focusing on issuing cities. COVID-19 cases have a significant negative effect. Larger population and better infrastructure correlate with higher inflows; high-speed rail effects are not robust. - Extended models (H2): • Platform: Issuing via larger, reputable platforms significantly raises inflows and amplifies the value effect (positive, significant interaction). • Frequency: More frequent issuance directly increases inflows (positive, significant), but frequency reduces the marginal effectiveness of larger voucher values (negative interaction), implying many small issuances can be more attractive than fewer large ones. • Internet penetration: The interaction between lnVALUE and internet penetration is positive and significant (5%), confirming greater effectiveness in digitally inclusive cities. - Heterogeneity: • City reputation has a marginal direct effect; its interaction with voucher value is positive and significant at 10%, indicating campaigns in well-known destinations gain more from voucher value. • Digital finance: Breadth of digital financial inclusion is directly positive (and its interaction with value is positive at 1%). Depth is not directly significant, but its interaction with value is positive and significant (1%), highlighting that both broader coverage and deeper usage strengthen voucher effectiveness. - Spatial spillovers (H3): Moran’s I for monthly inflows is positive and significant (≈0.36–0.63, p<0.01), confirming spatial clustering. Spatial Durbin results show significant and positive indirect (spillover) effects: voucher issuance in one city increases inflows in neighboring cities. Using the nested matrix, the total effect of lnVALUE on inflows is about 0.122 (p<0.01). Neighboring city population exerts a negative spillover (siphon effect), consistent with larger cities drawing tourists away from nearby locales. - Robustness: Alternative voucher proxies (publicity, availability) have positive and significant effects. Matched PSM-staggered DID and CSDID confirm positive impacts. IV-2SLS and IV-GMM yield positive, significant effects of voucher value; first-stage diagnostics (Kleibergen-Paap LM and Cragg-Donald F) indicate strong, valid instruments. Parallel trend and placebo tests support identification.

Findings support TPB mechanisms: digital travel vouchers likely enhance perceived behavioral control and normative beliefs, translating into higher travel intentions and realized trips. Design features are crucial—issuing via large, credible platforms improves user trust and awareness, frequency sustains salience, and digital inclusiveness (internet penetration and digital finance) reduces frictions in discovery, purchase, and redemption. The negative spillover from neighboring city size indicates competitive siphoning by large urban centers, while positive cross-border spillovers from voucher issuance argue for coordinated regional strategies. The results suggest that recovering tourism demand requires integrating economic incentives with psychological and technological enablers. The heterogeneity analyses imply that marketing in reputed destinations and in digitally advanced environments yields larger returns to voucher value, guiding targeted policy deployment.

The study documents that digital travel vouchers significantly increase tourist inflows in China during COVID-19, with effects robust to multiple specifications and identification strategies. Effectiveness is enhanced when vouchers are issued via large platforms, when issuance is frequent (with smaller values spread over time), and in cities with higher internet penetration and broader digital finance coverage. Spatial analyses reveal positive spillovers to neighboring cities, underscoring the need for inter-city coordination. Policy implications include: partnering with major platforms; structuring campaigns with repeated, smaller issuances; improving digital infrastructure and inclusiveness; facilitating online purchasing and convenient redemption; and designing regionally coordinated voucher programs to capitalize on spillovers and mitigate siphoning. The paper contributes by combining web-crawled mobility data with detailed voucher information and spatial econometrics to provide evidence aligned with TPB-based behavioral mechanisms. Future research should collect direct voucher usage data, expand city typologies, and conduct cross-country comparisons to generalize and tailor voucher designs to different socio-cultural contexts.

- Tourist inflows are proxied by Baidu Migration population inflows, which cannot perfectly separate tourists from business travelers or commuters. - Voucher data reflect issuance amounts; multi-app fragmentation and privacy protections prevent observing actual redemption and usage patterns. - Other concurrent voucher types or policies may influence outcomes, but comprehensive data are hard to assemble. - City heterogeneity characterization is limited; additional features could refine heterogeneity analyses. - Results are within China’s policy and digital context; external validity to other countries may require comparative studies.