Investigating the spatial effect of operational performance in China's regional tourism system

The paper studies how to more accurately evaluate the operational performance of China’s regional tourism systems by accounting for inter-stage interactions, dynamics across time, and spatial dependence. Traditional performance assessments often examine single stages (e.g., hotels or attractions) and ignore the networked, perishable nature of tourism services and the intermediation role of travel agencies. The authors propose that understanding where inefficiencies arise—across travel agencies, hotels, attractions, and dining—and how external factors (transportation infrastructure, urbanization, environmental management) and spatial spillovers shape performance is essential for improving competitiveness and guiding policy. The study aims to build a dynamic network DEA framework with carry-over and intermediate linkages and to estimate spatial effects on stage efficiencies using Tobit-based spatial models, thereby informing resource allocation and coordinated regional development.

The literature review covers three strands. Tourism systems research emphasizes interdependent components (attractions, services, transport, information, promotion) and spatial mobility, affected by external human, sociocultural, economic, technological, political, and environmental factors. DEA in tourism has been widely used, but most studies evaluate single stages or use single-stage models, rarely modelling multi-stage interdependencies or dynamics; Huang (2018) used a serial NDEA model but did not include carry-over effects. Recent work also links performance to spatial factors (e.g., HSR effects, environmental policies, urbanization), suggesting spatial heterogeneity and spillovers. This study fills gaps by constructing a multistage, dynamic network with carry-over and intermediate variables, and by integrating spatial econometrics to quantify how external, spatially distributed factors influence stage-specific efficiencies.

Design: A two-stage framework combines (1) dynamic network DEA to compute operational efficiency scores for each regional tourism system and its four service stages, and (2) panel Tobit/spatial Tobit (Durbin) models to estimate how external factors influence those efficiencies, including spatial spillovers. Sample and period: 30 provincial-level administrative regions in mainland China (excluding Tibet, Macau, Hong Kong, and Taiwan for data and institutional comparability), 2012–2016. Data sources: China Statistical Yearbook and China Tourism Statistics 2013–2017. Monetary values deflated to 2012 RMB. Initial carry-over values are the numbers of agencies, hotels, spots, and restaurants operating at end of 2011. SBM-DNDEA framework: Adopts Tone and Tsutsui’s slacks-based, dynamic network DEA allowing multiple inputs/outputs, intermediate flows, and carry-overs across periods. Four stages: Travel agency (TA), Hotel (H), Attraction (A), Dining (DI). Separate input at each stage is full-time employees. Carry-over inputs/outputs per stage are the numbers of entities still operating at period bounds (agencies, hotels, spots, restaurants). The key intermediate linkage is the number of tourists received by travel agencies (demand) feeding into H, A, and DI within a period. Outputs are stage revenues: travel agency service revenue, hotel revenue, attraction revenue, dining sales revenue. The model yields stage-level efficiencies and a system-level efficiency as a weighted composition across stages and years. Period weights follow sum-of-years-digits (2012=0.067, 2013=0.133, 2014=0.200, 2015=0.267, 2016=0.333), emphasizing recent performance; stage weights are equal (0.25 each) under equal-importance assumption. Spatial econometric specification: Dependent variables are stage-specific efficiencies (bounded 0–1) from SBM-DNDEA. Explanatory variables (lagged 1 year) proxy traffic development and external conditions: (i) regional government support for transport (Rgs: transport budget share of fiscal budget), (ii) regional rail infrastructure intensity (Rrai: rail length per area), (iii) regional road infrastructure intensity (Rroi: road length per area), (iv) regional urbanization level (Rubl: urban population share), (v) regional environmental protection investment (Repe: total investment in environmental pollution control). Spatial weights matrix W is contiguity-based (1 for adjacent provinces, 0 otherwise). Spatial dependence is tested via Global Moran’s I. If no spatial dependence, a panel Tobit is used; if present, panel Tobit spatial models are considered: lag (SLM), error (SEM), or Durbin (SDM), with selection by Wald and LR tests. For travel agency and dining stages, if spatial coefficient (rho) is insignificant, ordinary panel Tobit with random effects (per LR tests) is applied. Robust standard errors are reported. Model fit and spatial diagnostics include Moran’s I, Wald and LR tests comparing SDM vs SLM/SEM, and additional SAR/OLS contrasts where applicable. Managerial benchmarking: DEA frontier projections for 2016 quantify required proportional changes in inputs, intermediates, and carry-overs by province/stage to reach efficiency, informing policy levers (e.g., adjusting employment or facility counts, boosting tourist flows).

- National performance levels: The average system efficiency across 30 regions was 0.590 (rising from 0.583 in 2012 to 0.622 in 2016), indicating roughly 40% improvement room to reach the frontier. Attraction consistently outperformed other stages; travel agencies were the weakest and below the system average throughout. - Regional disparities: The east region had the highest overall performance (0.653), followed by west (0.543), northeast (0.540), and central (0.483), reflecting long-standing resource and development imbalances. Stage differences by zone were most pronounced for dining, travel agency, and hotel; attractions showed smaller deviations. - Benchmarks and under/over-capacity signals: Guangdong was fully efficient over 2012–2016 and served as a benchmark; Shanghai and Ningxia frequently led stage rankings. DEA projections (2016) suggest: nationally, increasing the number of tourists received by travel agencies by at least 1.73% could improve TA efficiency; in hotels, reducing employees by ~17.56% and hotel counts by ~23.63% on average would help reach efficiency, indicating overemployment and overcompetition, with similar patterns in attractions and dining. - Spatial dependence: Global Moran’s I indicated significant positive spatial clustering for stage efficiencies, motivating spatial Tobit models for hotels and attractions (and model testing for dining and travel agency). - Determinants and spillovers (Tobit/Tobit-SDM): - Travel agency (random-effects Tobit): Urbanization (Rubl) had a significant positive local effect; transport policy proxies (Rgs, Rrai, Rroi) and environmental investment were not significant. - Hotel (Tobit-SDM): Road infrastructure intensity (Rroi) significantly improved local hotel performance. Urbanization exhibited both significant local and positive neighboring spillover effects (W*Rubl), suggesting urban growth enhances accommodation demand across adjacent regions. - Attraction (Tobit-SDM): Road infrastructure intensity (Rroi) significantly boosted local attraction performance. Neighboring regions’ transport support and infrastructure (W*Rgs, W*Rrai, W*Rroi) exerted significant negative spillovers on local attractions, consistent with siphon effects drawing tourists to better-connected neighbors. Neighboring environmental protection investment (W*Ln(Repe)) also showed a significant negative spillover on local attraction performance. - Dining: Results indicated transport infrastructure—particularly roads (and rail, per text)—positively affected local dining performance; urbanization was strongly positive. Spatial dependence for dining was not robustly significant, so a non-spatial panel Tobit with random effects was preferred. - Policy context: The dip in 2013 performance likely relates to the China Tourism Law’s implementation, triggering declines in group tours and a transition toward independent travel, challenging traditional travel agency models.

By modelling the tourism system as a dynamic, multistage network with travel agencies as intermediators and explicitly incorporating spatial dependence, the study identifies where inefficiencies originate and how external conditions shape performance locally and across borders. The consistently weak travel agency performance pinpoints the intermediation stage as a primary constraint on system efficiency, aligning with the sector’s digital disruption and changing tourist behavior. Strong attraction performance coupled with siphon effects from neighboring transport investments highlights the need for coordinated, regionally differentiated transport and destination strategies to avoid zero-sum competition. Positive local and spillover effects of urbanization on hotels (and positive local effects for other stages) imply that urban development can stimulate tourism demand beyond municipal boundaries, supporting coordinated regional planning. DEA frontier projections reveal structural overcapacity in supply (overemployment and excessive facility counts) in several stages, suggesting the benefits of market restructuring, exit/merger mechanisms, and targeted demand stimulation. Overall, the integrated approach addresses the research question by delivering a more accurate, policy-relevant performance assessment that accounts for inter-stage linkages, temporal persistence, and spatial spillovers.

The paper develops and applies a dynamic network DEA framework with carry-over and intermediate linkages, combined with Tobit-based spatial econometrics, to evaluate and explain operational performance in China’s regional tourism systems (2012–2016). Key contributions are: (1) introducing a four-stage system (travel agency, hotel, attraction, dining) with travel agencies’ intermediation explicitly linking demand to other stages; (2) integrating dynamic carry-overs (continuing firms) and an intermediate flow (tourists served) into SBM-DNDEA; and (3) quantifying spatial effects of transport development, urbanization, and environmental investment on stage efficiencies. Empirically, system performance was modest nationally, with attractions leading and travel agencies lagging; the east region outperformed others; roads improved hotel and attraction performance; urbanization had strong positive local and spillover effects (notably for hotels); and neighboring transport/environmental investments could siphon demand from local attractions. Policy implications include upgrading travel agencies’ digital collaboration and product personalization, managing overcapacity via restructuring, and designing transport/environmental policies mindful of cross-border spillovers to foster coordinated development. Future research can extend the framework to include additional stages (e.g., education), finer-grained environmental and technological variables by stage, longer and more recent panels, and alternative spatial weight structures (e.g., economic distance, transport networks).

- Stage coverage is constrained by data availability: the tourism education stage was excluded and environmental protection investment was not disaggregated by stage/element, limiting interpretability of environmental effects. - Efficiency is evaluated at province level; micro-level heterogeneity (firm-level technologies, market structures) is not observed. - Spatial weights rely on geographic contiguity; alternative matrices (e.g., based on transport connectivity or economic ties) may capture different spillover channels. - The study period (2012–2016) predates more recent shocks and digital transformations; results may not generalize to later periods without re-estimation.