China's tourism industry has significantly contributed to its national economy. Effective performance evaluation of regional tourism systems is crucial for regulators to identify inefficiencies and improve competitiveness through policy planning. While past research emphasizes network collaboration, it often focuses on individual stages (e.g., hotels) rather than the interactions between them. This study aims to develop a more comprehensive framework for evaluating the performance of China's regional tourism system. This framework will incorporate network interactions among different tourism stages (travel agencies, hotels, attractions, and dining), dynamic features (carry-over effects), and spatial dependency to enhance the accuracy of performance evaluation. Understanding the operational performance of regional tourism systems is essential for both regional and central governments to control the quality of sustainable operations, clarify the sources of inefficiency, and offer direction for policy planning aimed at improving system shortcomings. The study will use the SBM-DNDEA model to measure operational performance and the Tobit spatial Durbin model to analyze the spatial effects and influencing factors.

Existing literature on tourism performance evaluation frequently employs Data Envelopment Analysis (DEA), a multi-dimensional framework for measuring relative performance of decision-making units (DMUs). While DEA applications in tourism are extensive, most studies focus on individual stages (hotels, travel agencies, attractions) neglecting the complex interrelationships within the tourism system. Huang (2018) pioneered a model for China's regional tourism system, but it lacked the exploration of carry-over items and the intermediation function of travel agencies. This study builds on this work by incorporating the dining stage (given its increasing importance), emphasizing the role of travel agencies as intermediaries, and considering carry-over items (number of firms operating at year-end) to reflect dynamic features and accumulated capital expenditure. The study also incorporates spatial econometric analysis which is needed because the external operating environment significantly influences tourism performance, including factors such as environmental awareness, technological innovations, and sustainability concerns. A two-stage methodological framework (DEA and Tobit regression) commonly used in performance evaluation will be extended to include spatial econometric models to account for spatial dependencies in the tourism system's performance.

This study employs a two-stage methodological framework. The first stage uses the slack-based measure approach of dynamic network DEA (SBM-DNDEA) to assess the relative operational performance of China's regional tourism system. The system is modeled as four interconnected stages: travel agency, hotel, attraction, and dining. The SBM-DNDEA model incorporates multiple inputs, outputs, intermediate variables (number of tourists served by travel agencies), and carry-over variables (number of firms operating at year-end) to account for the dynamic interactions between stages over time (2012-2016). Weights for each time period and stage are exogenously determined using the sum-of-the-year's digits method and equal weighting for stages respectively. The second stage uses a spatial econometric model, specifically the Tobit spatial Durbin model, to examine the factors influencing operational efficiency. The dependent variables are the efficiency scores from the SBM-DNDEA model. Independent variables include regional government support for transportation, regional rail and road infrastructure intensity, regional urbanization level, and regional environmental protection investment. The Tobit model is chosen due to the bounded nature (0-1) of the efficiency scores. Spatial dependence is assessed using Moran's I test. If spatial dependence is detected, the Tobit spatial Durbin model is employed; otherwise, an ordinary panel Tobit model is used. This model considers both direct effects of independent variables on efficiency and spillover effects from neighboring regions.

The study analyzed panel data from 30 provincial administrative regions in China from 2012 to 2016. The SBM-DNDEA model revealed that the overall operational performance of China's regional tourism system was relatively low (average score around 0.59). The attraction stage consistently outperformed other stages (travel agency, hotel, dining). Performance scores showed a gradual upward trend after 2013. Significant regional discrepancies were observed, with the eastern region exhibiting the highest performance (0.653), followed by the west (0.534), northeast (0.540), and central (0.483) regions. Guangdong consistently ranked as the most efficient region, while several regions like Inner Mongolia, Hebei, and Gansu demonstrated significantly lower performance. Analysis of individual stage performance highlighted the underperformance of the travel agency stage, suggesting it as a major source of inefficiency for the overall system. The Tobit spatial Durbin model results indicated that road infrastructure intensity had a significant positive effect on hotel and attraction performance. Urbanization level positively impacted all stages, both locally and through spillover effects to neighboring regions. However, environmental management expenditure displayed a negative spillover effect on surrounding regions' attraction performance. The analysis of the efficient frontier projection suggested potential improvements through adjustments in input variables such as workforce size and number of firms to achieve greater efficiency.

The findings highlight the need for a more holistic approach to tourism development in China. The low overall operational performance and significant regional disparities underscore the challenges in achieving efficient resource allocation and coordination among different tourism stages. The underperformance of the travel agency stage indicates a need for adaptation to changing market dynamics, particularly the rise of online travel agencies and digital technologies. The positive effects of urbanization and road infrastructure on tourism performance emphasize the importance of infrastructure development in enhancing regional connectivity and accessibility. The negative spillover effects of environmental management on adjacent attractions highlight the importance of strategic spatial planning in tourism development to avoid negative externalities. The findings of this study also suggest that promoting balanced regional development and improving the overall efficiency of the tourism system requires a multi-faceted approach that considers the interconnectedness of various tourism stages and their spatial interactions.

This study contributes to the existing literature by providing a comprehensive framework for evaluating the performance of China's regional tourism system. The findings highlight the low overall operational performance, significant regional disparities, and the crucial role of travel agencies in the system's efficiency. The results provide valuable insights for policymakers and tourism managers to develop targeted strategies for improving resource allocation, promoting balanced regional development, and adapting to the changing technological landscape. Future research could explore the effects of more granular environmental protection investment categorized by tourism stage or element. Further investigation into specific factors contributing to the underperformance of travel agencies and the exploration of innovative collaboration models among tourism stages is warranted.

The study's limitations include the use of readily available data which might restrict the inclusion of more tourism stages in the analysis. The focus on provincial-level data might mask variations at a more granular level. The model's assumptions, such as equal weighting for stages, could also influence the results. Future research could explore more complex weighting schemes based on empirical data or expert knowledge. The time frame (2012-2016) may not fully reflect the long-term impact of policy changes or technological advancements. The exclusion of some regions due to data unavailability could affect the generalizability of the results.