Rural tourism evaluation has evolved from simple economic impact assessments to comprehensive sustainability frameworks incorporating multiple stakeholder perspectives (Zhu et al., 2025b). Early approaches focused primarily on visitor numbers and revenue generation, reflecting the dominant economic paradigm of tourism development (Sutomo et al., 2024). However, growing recognition of tourism's complex interactions with local communities, environments, and cultures has driven the development of more holistic evaluation frameworks (Muda et al., 2024).

Contemporary rural tourism evaluation encompasses four primary dimensions: economic impacts including employment generation, income distribution, and market development; environmental effects covering ecosystem preservation, resource utilization, and pollution management; social outcomes involving community empowerment, cultural preservation, and quality of life improvements; and institutional factors including governance structures, policy effectiveness, and stakeholder coordination (Vargas et al., 2024). This multidimensional approach reflects the complex nature of rural tourism as a socio-economic phenomenon requiring integrated assessment methodologies (Wu et al., 2024).

Recent studies have emphasized the importance of indicator selection, weighting methodologies, and temporal analysis in rural tourism evaluation (Li, 2024). Key challenges include managing indicator interdependencies, addressing scale effects across different administrative levels, and incorporating stakeholder perspectives in weight determination (Weng et al., 2023). The COVID-19 pandemic has further highlighted the need for evaluation frameworks capable of capturing disruption, adaptation, and recovery dynamics in rural tourism systems (Yan et al., 2023).

Multi-criteria decision analysis (MCDA) has become increasingly prominent in tourism research due to its capacity to handle complex decision problems involving multiple, often conflicting objectives (Saputro et al., 2023). The tourism literature has applied various MCDA approaches including Analytic Hierarchy Process (AHP), Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), and Data Envelopment Analysis (DEA) for destination competitiveness assessment, site selection, and policy evaluation (Tang et al., 2023a).

However, traditional MCDA approaches face limitations in weight determination, particularly regarding the integration of subjective expert judgment with objective data characteristics (Liu L. et al., 2023). Purely subjective approaches may suffer from inconsistency and bias, while purely objective methods may overlook important contextual factors and stakeholder preferences (Hu and Xu, 2024). This has driven interest in hybrid approaches that combine multiple weighting methodologies to leverage their respective strengths (Fang L. et al., 2023).

The entropy weighting method has gained attention for its ability to determine objective weights based on information content and variability (Scorza and Gatto, 2023). Shannon entropy, originally developed in information theory (Shannon, 1948), measures the uncertainty or randomness in a dataset, with lower entropy indicating more concentrated information and higher discriminatory power (Liu N. et al., 2023). In MCDA applications, entropy weights reflect the relative importance of criteria based on their ability to differentiate between alternatives (Wang et al., 2024).

The Delphi method, developed by the RAND Corporation in the 1950s (Linstone and Turoff, 1975), has become a widely accepted approach for building expert consensus on complex issues characterized by uncertainty and incomplete information (Zhang et al., 2025). In tourism research, Delphi studies have been applied to forecast tourism trends, identify critical success factors, and develop evaluation frameworks (Cebrian and Domenech, 2024). The method's strength lies in its ability to harness collective expert knowledge while minimizing the influence of dominant personalities and groupthink (Wei and Zheng, 2023).

Key considerations in Delphi applications include expert selection criteria, round design, consensus measurement, and result interpretation (Yang and Ning, 2025). Expert selection typically involves identifying individuals with relevant expertise, experience, and willingness to participate throughout multiple rounds (Ding et al., 2025). Consensus measurement approaches range from statistical measures such as coefficient of variation to qualitative assessments of convergence (Fernandez et al., 2025).

Recent tourism studies have emphasized the importance of structured questionnaire design, clear definition of consensus criteria, and transparent reporting of expert characteristics (Jena and Dwivedi, 2023). The integration of Delphi results with other methodological approaches represents an emerging trend in tourism research, particularly for complex evaluation problems requiring both expert judgment and empirical analysis (Liu, 2024). Recent applications in European rural tourism contexts have demonstrated the method's effectiveness in capturing diverse stakeholder perspectives, including farmer priorities in Italian Ligurian destinations (Peira et al., 2021) and heritage valorization strategies in the Piedmont Canavese area (Beltramo et al., 2021), providing valuable insights for community-based tourism development.

Despite substantial progress in rural tourism evaluation methodologies, several significant gaps remain. First, limited research has systematically integrated Delphi expert consensus with entropy-based objective weighting in tourism contexts (Wang and Liu, 2024). While both methods have been applied individually, their combination offers potential synergies that remain largely unexplored (Zhang J. et al., 2024).

Second, existing evaluation frameworks often lack comprehensive temporal analysis capabilities, particularly for capturing disruption and recovery dynamics (Hussain et al., 2023). The COVID-19 pandemic has highlighted the importance of resilience assessment, yet few studies have developed methodologies specifically designed to analyze these patterns (Gherdan et al., 2025).

Third, while numerous evaluation frameworks have been proposed, few have been rigorously tested in specific regional contexts with comprehensive empirical validation (Li et al., 2024). This limits their practical applicability and policy relevance (Zhu et al., 2025a). Finally, existing methodologies often focus on single administrative levels, with limited consideration of multi-scale analysis and cross-regional comparison capabilities (Huang et al., 2023).

This research addresses these gaps by developing an integrated Delphi-entropy framework specifically designed for rural tourism evaluation, applying it to a comprehensive empirical case study, and demonstrating its capacity to capture pandemic-related disruptions and recovery patterns (Yang M. et al., 2024).

The integrated Delphi-entropy evaluation framework combines subjective expert knowledge with objective information theory to address the limitations of single-method approaches (Sun and Zhou, 2025). The conceptual framework comprises four key components: (1) indicator system development through literature review and expert consultation; (2) dimensional weight determination via Delphi methodology; (3) indicator weight calculation using Shannon entropy theory; and (4) integration of weights for comprehensive evaluation (Wang and Chen, 2024).

The framework assumes that sustainable rural tourism development requires balanced performance across multiple dimensions, with relative importance varying according to local contexts and development stages (Yan et al., 2023). Expert knowledge provides valuable insights into contextual factors, policy priorities, and stakeholder preferences, while entropy weights reflect the objective information content and discriminatory power of individual indicators (Saputro et al., 2023).

The indicator system development process followed a systematic approach integrating literature review, expert consultation, and data availability assessment. Initial indicator identification drew from 127 peer-reviewed articles published in SSCI/SCI journals between 2018-2023, focusing on rural tourism evaluation, sustainability assessment, and regional development measurement.

Through content analysis and thematic coding, we identified 45 potential indicators across six broad categories: economic development, infrastructure and accessibility, environmental quality, social welfare, cultural preservation, and tourism performance. These indicators were subsequently refined through expert consultation and data availability constraints to produce a final system of 30 indicators organized into four primary dimensions.

The four dimensions reflect the integrated nature of rural tourism as a socio-economic phenomenon: Innovation & Culture encompasses technological advancement, educational investment, and cultural development; Economy & Infrastructure covers transportation systems, urbanization processes, and economic structure; Eco-Environment addresses environmental protection, pollution control, and ecological conservation; Tourism & Livelihood focuses on tourism performance, community welfare, and social development.

Each indicator was classified as either positive (benefit-type) or negative (cost-type) based on its theoretical relationship with sustainable rural tourism development. Positive indicators contribute directly to development objectives, while negative indicators represent constraints or negative impacts requiring minimization.

The Delphi process involved two structured rounds designed to achieve expert consensus on dimensional weights and indicator relevance. Expert panel selection followed established criteria including: (1) advanced degree (Ph.D. or Master's) in tourism, economics, environmental science, or related fields; (2) minimum professional experience differentiated by role—university professors (10+ years research experience), government policy analysts (7+ years policy formulation experience), industry consultants (8+ years practical experience), NGO specialists (6+ years community engagement); (3) at least 3 peer-reviewed publications or major policy reports related to rural tourism or sustainability assessment; and (4) geographical representation across major Chinese regions to ensure diverse regional perspectives.

The final panel comprised 11 experts with the following characteristics: 4 university professors (2 from tourism management, 1 from regional economics, 1 from environmental planning; located in Beijing, Shanghai, Xi'an, and Chengdu representing eastern, central, and western China); 3 government policy analysts (from Ministry of Culture and Tourism, Shaanxi Provincial Development and Reform Commission, and County-level Tourism Bureau); 2 tourism industry consultants (specializing in rural destination development and heritage tourism); and 2 NGO specialists (focusing on community-based tourism and sustainable livelihoods). Panel composition balanced academic expertise (36%), policy experience (27%), practical implementation knowledge (18%), and community perspectives (18%), with geographical diversity across 7 provinces.

Round 1 utilized a structured questionnaire requesting experts to: (1) rate the importance of each dimension on a 10-point Likert scale with clear anchor definitions; (2) assess indicator relevance within each dimension using a 5-point scale; (3) suggest additional indicators or modifications with written justification; and (4) provide qualitative comments on framework structure and completeness. The questionnaire was distributed electronically via professional survey platform with a 3-week response period. Response rate achieved 100% (11/11 experts).

Round 2 presented aggregated results from Round 1, including mean scores, standard deviations, interquartile ranges, and synthesized qualitative feedback. Experts were asked to reconsider their ratings in light of group responses and provide final assessments with optional explanations for maintained divergent views. Consensus criteria required coefficient of variation below 0.25 for dimensional weights and above 70% agreement on indicator inclusion. Expert consensus improved significantly between rounds, with average coefficient of variation decreasing from 0.31 (Round 1) to 0.21 (Round 2). Kendall's coefficient of concordance confirmed strong agreement among experts (W = 0.68, χ² = 29.8, p < 0.001).

Final dimensional weights achieved strong consensus with coefficients of variation ranging from 0.18 to 0.23: Innovation & Culture (30%), Economy & Infrastructure (30%), Eco-Environment (25%), and Tourism & Livelihood (15%). These weights reflect expert assessment of relative importance for sustainable rural tourism development in the Chinese context, with Innovation and Economy dimensions prioritized equally due to their foundational role in long-term sustainability.

To support transparency and generalizability, we explicitly acknowledge in the Limitations that our expert panel size was n = 11 and discuss implications for external validity and future scaling (see Limitations within Conclusions).

Shannon entropy theory provides an objective method for determining indicator weights based on information content and variability (Chen et al., 2024). The entropy weight calculation follows a five-step process designed to ensure mathematical rigor and interpretability.

The final integrated weights combine Delphi dimensional weights with entropy indicator weights through multiplicative integration:

where WgDelphi represents the Delphi weight for dimension g containing indicator j.

The comprehensive evaluation index for each year is calculated as:

Dimensional sub-indices enable analysis of performance patterns across different aspects of rural tourism development:

Shaanxi Province was selected as the empirical case study based on several criteria: (1) significant rural tourism development over the past decade; (2) diverse geographical and cultural characteristics; (3) comprehensive statistical data availability; and (4) policy relevance for western China development strategies.

Located in northwest China, Shaanxi covers 205,800 km² with a population of 39.5 million. The province encompasses diverse geographical zones including the Qinling Mountains, Wei River Valley, and northern Loess Plateau. Cultural heritage includes 13 UNESCO World Heritage sites, reflecting over 3,000 years of Chinese civilization.

Data collection utilized official statistical sources including Shaanxi Statistical Yearbook (2019–2024), China Tourism Statistical Yearbook (2019–2024), China Statistical Yearbook on Environment (2019–2024), and specialized reports from Shaanxi Provincial Department of Culture and Tourism and Department of Ecology and Environment. The temporal scope covers 2018–2023, providing pre-pandemic, pandemic, and recovery period analysis.

The entropy weight calculation reveals significant variations in indicator importance based on information content and discriminatory power (Figure 1). The stacked bar chart illustrates the cumulative weight distribution across four dimensions, with each bar segment representing individual indicator contributions within their respective dimensions. The top-weighted indicator, C14 (culture industry investment growth rate), accounts for 25.1% of total weight, reflecting its high volatility and discriminatory capacity across the analysis period. This finding highlights the critical importance of cultural investment dynamics in differentiating rural tourism development performance.

Infrastructure investment growth (C26, 0.9%) and international tourism revenue (C43, 0.6%) rank second and third respectively, emphasizing the significance of investment volatility and international market performance in rural tourism evaluation. The weight distribution follows a power-law pattern, with the top 5 indicators accounting for 27.5% of total weight while the bottom 15 indicators contribute less than 0.1% each. The stacked visualization clearly demonstrates the concentration of weights within the Tourism & Livelihood dimension, where multiple indicators collectively contribute to its substantial 58.6% share.

Dimensional analysis reveals balanced weight distribution across Innovation & Culture (18.4%) and Economy & Infrastructure (17.8%), while Eco-Environment (5.2%) and Tourism & Livelihood (58.6%) show contrasting patterns. The stacked bar representation allows for direct comparison of both dimensional totals and the relative contribution of individual indicators within each dimension. The high concentration in Tourism & Livelihood reflects the volatile nature of tourism-related indicators, particularly during the pandemic period, with multiple indicators (C42, C43, C44, C45, C46, and C47) contributing to this dimension's dominant weight share.

The comprehensive indicator system demonstrates strong statistical properties with appropriate variability across indicators (Table 1). Coefficient of variation ranges from 0.001 for waste treatment rate to 2.847 for culture industry investment growth, reflecting diverse indicator characteristics and discriminatory capacity.

The two-round Delphi process achieved strong expert consensus on dimensional weights through systematic convergence (Figure 2). Round 1 demonstrated initial expert assessments with moderate dispersion (mean coefficient of variation CV = 0.31), reflecting diverse perspectives across the 11-member panel comprising university professors, government policy analysts, industry consultants, and NGO specialists. Round 2 exhibited significant convergence (mean CV = 0.21), with experts adjusting their ratings after reviewing aggregated feedback, statistical summaries, and synthesized qualitative comments from Round 1.

Figure 2 presents detailed comparison of rating distributions across both rounds. Innovation & Culture and Economy & Infrastructure dimensions received consistently high ratings (mean scores 8.7–9.2 out of 10), reflecting expert consensus on their foundational importance for long-term sustainable development. Eco-Environment achieved moderate-high ratings (mean 7.8–8.1), while Tourism & Livelihood received relatively lower dimensional priority (mean 5.2–5.8), consistent with its inherent short-term volatility and dependence on stable foundations in other dimensions. The visual convergence—evidenced by tighter box ranges, reduced whisker spans, and decreased scatter in Round 2—confirms the Delphi method's effectiveness in building structured expert agreement.

Expert rating patterns revealed meaningful cross-group differences while maintaining overall convergence (Figure 3). University professors (n = 4) assigned highest weights to Innovation & Culture (mean 9.1), emphasizing knowledge production, human capital development, and institutional capacity building as prerequisites for sustainable tourism. Government policy analysts (n = 3) prioritized balanced weights across all dimensions (means 8.2–8.9), reflecting holistic policy design requirements and cross-sectoral coordination imperatives. Industry consultants (n = 2) emphasized Economy & Infrastructure (mean 9.3), highlighting practical implementation prerequisites including transportation access, market connectivity, and service delivery systems. NGO specialists (n = 2) assigned relatively higher weights to Eco-Environment (mean 8.7) and Tourism & Livelihood (mean 6.8), reflecting community sustainability priorities, environmental justice concerns, and direct livelihood impacts.

Consensus quality metrics improved substantially between rounds (Figure 4), providing quantitative evidence of convergence effectiveness. Coefficient of variation decreased by 32% on average (from mean 0.31 to 0.21 across dimensions), demonstrating reduced relative dispersion in expert opinions. Standard deviation declined by 28% (from mean 1.82 to 1.31 on the 10-point scale), indicating absolute reduction in rating spread. Interquartile range narrowed by 35% (from mean 2.15 to 1.40), showing tightened middle-50% consensus zone. Overall score range contracted by 41% (from mean 4.73 to 2.79), reflecting convergence of extreme positions toward group median.

Kendall's coefficient of concordance (W = 0.68, χ² = 29.8, p < 0.001, df = 3) confirms strong overall agreement among the 11 experts, validating the Delphi process effectiveness. This concordance value, substantially above the conventional threshold of 0.5, indicates that expert rankings of dimensional importance achieved high consistency despite diverse professional backgrounds. The statistically significant chi-square test rejects the null hypothesis of random ranking, providing formal evidence that observed agreement patterns reflect genuine expert consensus rather than chance alignment.

Final dimensional weights integrated expert consensus with theoretical frameworks and regional development priorities: Innovation & Culture (30%) and Economy & Infrastructure (30%) received equal highest priority, reflecting their foundational role in enabling sustainable tourism development; Eco-Environment (25%) received substantial weight acknowledging environmental sustainability imperatives and ecological carrying capacity constraints; Tourism & Livelihood (15%) received lower dimensional weight while paradoxically containing the highest-weighted individual indicator (C14: culture industry investment growth at 25.1% of total system weight). This apparent discrepancy demonstrates the framework's sophisticated capacity to capture both dimensional importance (reflecting stable long-term priority) and indicator-level discriminatory power (reflecting temporal volatility and information content). The dimensional weights guide the entropy-based indicator weighting process, ensuring that final integrated weights balance expert judgment on strategic priorities with objective data characteristics.

The comprehensive development index exhibits a pronounced U-shaped trajectory over the 2018–2023 period, with distinct phases of decline, stabilization, and gradual recovery (Figure 5). The index declined from 0.271 in 2018 to 0.018 in 2020, representing a 93% decrease during the initial pandemic impact. This dramatic decline primarily reflects the collapse of international tourism indicators and reduced investment in culture industry projects.

The 2020–2021 period shows signs of recovery, with the index increasing to 0.117 in 2021, though remaining well below pre-pandemic levels. However, a secondary decline occurs in 2022 (0.015), followed by modest improvement in 2023 (0.044). This pattern reflects the complex and prolonged nature of tourism recovery, influenced by ongoing travel restrictions, changing consumer behavior, and economic uncertainties.

Data quality note: The 2023 comprehensive index incorporates estimated values for C14 (culture industry investment growth, the highest-weighted indicator at 25.1%), derived from Q1–Q3 reported data using historical seasonal adjustment factors. While this estimation introduces potential measurement uncertainty, sensitivity analysis demonstrates that alternative estimation methods (last observation carried forward, linear trend extrapolation) yield highly correlated results (Pearson's r>0.94), maintaining temporal pattern validity. The U-shaped trajectory remains statistically significant across all estimation scenarios, confirming that principal conclusions are robust to this data constraint.

The 2023 comprehensive index relies on estimated values for C14 (culture industry investment growth, 25.1% weight), derived from Q1-Q3 data with historical seasonal adjustment. To assess estimation impact on principal findings:

This analysis confirms that while 2023 estimates introduce measurement uncertainty, they do not materially affect the study's substantive conclusions regarding incomplete recovery and policy priorities.

Dimensional performance analysis reveals significant heterogeneity across the four evaluation dimensions. The analysis examines performance patterns for key years (2018, 2020, 2022, and 2023) across normalized scales, enabling direct comparison of dimensional contributions to overall development.

Eco-Environment consistently demonstrates the strongest performance, with normalized scores above 0.7 throughout the analysis period. This reflects Shaanxi's sustained commitment to environmental protection and pollution control, including improvements in air quality, waste treatment, and green coverage. The dimension shows steady improvement from 0.72 (2018) to 0.89 (2023), indicating successful environmental governance.

Innovation & Culture shows moderate stability with scores ranging from 0.45 to 0.65, though with notable volatility in 2019–2020 reflecting disruptions to educational and cultural investments during the pandemic. Recovery becomes apparent from 2021 onwards, suggesting resilience in innovation systems and cultural development programs.

Economy & Infrastructure demonstrates intermediate performance with gradual improvement over time, from 0.38 (2018) to 0.52 (2023). This reflects ongoing urbanization, infrastructure development, and economic structural transformation in Shaanxi Province. The relatively steady trajectory suggests that infrastructure development maintained momentum despite pandemic disruptions.

Tourism & Livelihood exhibits the highest volatility and most severe pandemic impacts, declining from 0.61 (2018) to 0.09 (2020) before showing modest recovery to 0.31 (2023). This pattern directly reflects the tourism sector's vulnerability to mobility restrictions and changing travel patterns during the pandemic period.

The correlation network analysis reveals complex interdependencies among the top weighted indicators (Figure 6). To improve readability, we apply a conservative threshold (|r|≥0.50) and enlarge labels for highly weighted nodes; the full network with lower thresholds is available in the Supplementary Material. Strong positive correlations emerge between innovation-related indicators (C11, C13, C15) with coefficients above 0.6, suggesting synergistic relationships in the provincial innovation ecosystem.

Tourism indicators (C42, C43, and C47) demonstrate high positive correlations (r > 0.8), confirming their interdependent nature and shared vulnerability to external shocks. This clustering supports the appropriateness of their grouping within the Tourism & Livelihood dimension and explains their collective response to pandemic impacts.

Negative correlations appear between infrastructure indicators (C21 and C26) and some tourism indicators, potentially reflecting trade-offs between development investment and short-term tourism performance. Environmental indicators show weak correlations with other dimensions, suggesting their relative independence and different driving factors.

The correlation structure informs understanding of indicator redundancy and complementarity. High correlations within dimensions support the dimensional structure while revealing potential opportunities for indicator consolidation in future applications. Cross-dimensional correlations suggest important interaction effects requiring consideration in policy design.

Principal component analysis reveals the underlying structure of indicator relationships and temporal development patterns (Figure 7). The three-dimensional trajectory visualization displays year positions in PC1-PC2-PC3 space, capturing 97.3% of total variance (PC1: 79.1%, PC2: 18.2%). This representation provides intuitive understanding of how Shaanxi's rural tourism development evolved through pre-pandemic prosperity, pandemic crisis, and post-pandemic recovery phases.

Scenario analysis examines potential impacts of targeted policy interventions across different dimensions over the 2018–2023 period (Figure 8). The stacked area visualization displays cumulative effects of three policy scenarios—20% innovation enhancement, 15% environmental improvement, and 30% tourism recovery—layered above the baseline development trajectory. This representation enables simultaneous assessment of individual scenario contributions and combined potential impacts across the entire analysis period.

The temporal visualization reveals several critical patterns. First, the baseline trajectory (gray area) demonstrates the U-shaped pandemic impact clearly, with the comprehensive index declining from 0.27 (2018) to near zero (2020) before gradual recovery. Second, the relative contribution of each scenario varies substantially across years: during pre-pandemic prosperity (2018–2019), tourism recovery scenarios (orange) provide the largest incremental gains; during crisis years (2020–2022), innovation enhancement (blue) and environmental improvement (green) maintain more stable contributions despite reduced absolute magnitudes; in the recovery phase (2023), tourism scenarios again dominate potential improvements.

The tourism enhancement scenario demonstrates the largest potential impact, with comprehensive indices increasing by an average of 45% across the analysis period. This reflects both the high weight of tourism indicators and their current below-potential performance levels. However, the scenario also exhibits highest volatility, confirming tourism's vulnerability to external shocks.

Innovation enhancement shows moderate but consistent improvements (average 15% increase) with lower volatility than tourism scenarios. This suggests that innovation-focused policies offer stable development benefits while building long-term competitive advantages. The steady improvement pattern supports innovation as a foundation for sustainable rural tourism development.

Environmental enhancement produces the smallest aggregate impact (8% average increase) but demonstrates exceptional stability and consistent improvement trends. This reflects the already strong environmental performance in Shaanxi Province and suggests that environmental policies offer reliable but incremental benefits.

The dimensional improvement potential analysis reveals Tourism & Livelihood as having the greatest development opportunity, with potential scores 25% above current levels. Innovation & Culture and Economy & Infrastructure show moderate improvement potential (15%–18%), while Eco-Environment demonstrates limited but stable enhancement opportunities (12%).

Annual growth rate analysis confirms the volatility patterns, with tourism-related growth rates ranging from –75% to +156% while environmental improvements maintain steady 2-5% annual growth. Infrastructure development shows intermediate volatility (±15% annually), suggesting balanced risk-return characteristics for infrastructure-focused policies.

The development stability analysis using rolling standard deviation reveals increasing stability over time, from 0.089 (2018–2020) to 0.021 (2021–2023). This pattern suggests improving resilience and reduced sensitivity to external shocks, potentially reflecting adaptive capacity development and policy learning effects.

The annual performance matrix provides comprehensive assessment of development patterns across dimensions and overall performance (Table 2). Rankings reveal 2018 as the peak performance year (rank 1), followed by 2019 (rank 2), with 2021 (rank 3) showing initial recovery, while 2020 showing the poorest performance (rank 6) reflecting pandemic impacts.

Innovation & Culture performance demonstrates gradual improvement from 2020 onwards, with 2023 achieving the second-highest score (0.347) after 2018 (0.384). This pattern suggests successful adaptation and recovery in educational investments, technological development, and cultural programs despite initial pandemic disruptions.

Economy & Infrastructure shows steady improvement throughout the period, with monotonic increases from 0.301 (2018) to 0.478 (2023). This consistent trajectory reflects sustained infrastructure investment and urbanization progress, suggesting that this dimension maintained development momentum despite external challenges.

Eco-Environment achieves consistently high performance with gradual improvement from 0.721 (2018) to 0.889 (2023). The strong and stable performance confirms Shaanxi's environmental protection achievements and suggests that environmental governance remained a policy priority throughout the analysis period.

Tourism & Livelihood exhibits extreme volatility, ranging from 0.605 (2018) to 0.089 (2020) before partial recovery to 0.314 (2023). The dramatic fluctuations reflect the sector's vulnerability to mobility restrictions, border controls, and changing travel behaviors during the pandemic period.

To assess the robustness of findings and address potential methodological concerns, we conducted comprehensive sensitivity analysis examining weight variation impacts, temporal stability, and methodological alternatives.

The integrated Delphi-entropy framework addresses several critical limitations in existing rural tourism evaluation methodologies. By combining subjective expert consensus with objective information-theoretic measures, the approach overcomes the bias limitations of purely subjective methods while incorporating contextual knowledge that purely objective approaches may overlook (Chen et al., 2024). Building upon established evaluation frameworks for rural ecotourism resources (Liu et al., 2020), the two-stage weight determination process ensures both theoretical validity and practical relevance in indicator importance assignment. This integrated approach aligns with recent methodological advances in sustainable rural development assessment, particularly frameworks oriented toward Sustainable Development Goals (Zhang W. et al., 2024; Chen et al., 2023), which emphasize the importance of comprehensive indicator systems capturing economic, environmental, and social dimensions of sustainability.

The entropy weighting component provides several advantages over traditional equal weighting or subjective assignment approaches. Indicators with higher variability and discriminatory power receive greater weight, ensuring that evaluation outcomes reflect actual performance differences rather than arbitrary assumptions (Yang S. et al., 2024). This is particularly valuable for rural tourism evaluation, where indicators may exhibit vastly different volatility patterns and information content. Contemporary studies on rural human settlement sustainability (Chen et al., 2023) similarly demonstrate the effectiveness of multi-criteria evaluation approaches in capturing complex development dynamics across different regional contexts.

The dimensional structure balances comprehensiveness with parsimony, addressing the common challenge of indicator proliferation in sustainability assessment frameworks (Wang, 2024). The four-dimension structure captures the essential aspects of rural tourism development while maintaining analytical tractability and policy relevance. The hierarchical weight integration enables both aggregate assessment and dimension-specific analysis, supporting multi-level policy design.

However, the methodology also presents certain limitations requiring consideration. The entropy weighting approach favors indicators with high volatility, which may overemphasize short-term fluctuations relative to long-term trends (Jiang et al., 2025). This is particularly relevant for tourism indicators, which demonstrated extreme volatility during the pandemic period. Future applications might consider modified entropy approaches that incorporate temporal smoothing or trend adjustment mechanisms.

The Delphi expert selection and consensus-building process, while systematic, reflects the perspectives of a specific expert panel at a particular time (Gonzalez et al., 2023). Different expert compositions or temporal contexts might yield different dimensional weights, affecting evaluation outcomes. Sensitivity analysis of weight variations would strengthen the methodology's robustness and generalizability.

The U-shaped development trajectory observed in Shaanxi Province reflects broader patterns documented in tourism literature regarding pandemic impacts and recovery dynamics (Wang et al., 2025). However, the magnitude of decline (93% from peak to trough) exceeds many international benchmarks, potentially reflecting China's particularly stringent mobility restrictions and border controls during 2020–2021.

The dominance of culture industry investment growth in the weight structure (25.1%) highlights the critical importance of cultural investment volatility in differentiating regional performance (Meng and Yang, 2024). This finding suggests that rural tourism development is particularly sensitive to cultural sector investment patterns, emphasizing the need for stable, long-term cultural development funding mechanisms rather than project-based or cyclical investment approaches.

The strong and consistent environmental performance throughout the analysis period demonstrates Shaanxi's successful implementation of ecological civilization policies and environmental protection measures (Yang and Wang, 2025). This performance provides a foundation for sustainable tourism development and suggests that environmental quality can be maintained even during economic stress periods with appropriate policy frameworks.

The extreme volatility in Tourism & Livelihood indicators reflects both the sector's vulnerability and its potential for rapid recovery under favorable conditions (Martinez and Garrido, 2024). The partial recovery observed in 2023 suggests that tourism systems possess inherent resilience mechanisms, though full recovery may require sustained effort and favorable external conditions.

The scenario analysis and empirical findings support a prioritized, multi-pathway approach to rural tourism recovery and sustainable development. We propose specific, actionable recommendations organized by priority tier and implementation timeline.

Several methodological extensions could enhance the framework's analytical capacity and practical applicability (Sutomo et al., 2024). Dynamic (time-varying) weighting and fuzzy extensions (e.g., fuzzy Delphi, fuzzy entropy) can better accommodate uncertainty and phase-specific shifts in indicator salience. Machine learning techniques might identify optimal weight combinations for different contexts or objectives, and Bayesian aggregation could fuse multiple weighting priors while quantifying uncertainty.

Spatial analysis integration could extend the framework to multi-regional or hierarchical assessment applications (Muda et al., 2024). Geographic information systems (GIS) integration would enable spatial visualization of performance patterns and identification of geographical clusters or hotspots requiring targeted intervention.

Stakeholder perspective integration beyond expert opinion could strengthen the framework's legitimacy and social acceptance (Vargas et al., 2024). Community consultation, tourist survey integration, and business stakeholder input could provide additional perspectives on indicator relevance and weighting preferences.

Uncertainty and sensitivity analysis development would strengthen the framework's reliability and robustness (Wu et al., 2024). Monte Carlo simulation, fuzzy logic integration, or scenario-based sensitivity testing could assess result stability under different assumptions and parameter variations.

Longitudinal extension to longer time series would enable trend analysis, cyclical pattern identification, and forecasting capability development (Li, 2024). Integration with economic forecasting models or tourism demand projection systems could support strategic planning and policy scenario testing applications.

The hierarchical Delphi-entropy integration successfully synthesizes subjective expert consensus with objective information-theoretic measures, demonstrating superior performance across discriminatory power (CV: 0.847), expert validity (82% agreement), and temporal stability metrics compared to alternative approaches. Comprehensive sensitivity analysis—including weight perturbations (±30%, correlations >0.92), temporal window variations (C14 weight stable at 18.3%–25.1%), methodological comparisons (overall score 8.5/10 vs. 4.2–7.5/10 for alternatives), and bootstrap confidence intervals—confirms robustness under diverse assumptions. This methodological contribution provides a replicable template extending beyond rural tourism to broader regional development assessment contexts.

The statistically validated U-shaped trajectory (quadratic regression: β₂ = 0.013, p = 0.035, R² = 0.785) reveals comprehensive indices declining 93.4% from 0.271 (2018) to 0.018 (2020) before gradual recovery to 0.044 (2023). This decline substantially exceeds international averages (55%–85%), with Shaanxi's recovery rate (10.3% by 2023) among the slowest globally versus North American (89%) and Mediterranean (75%) comparators. Three China-specific factors drive this exceptional severity: stringent zero-COVID policies, 2.5-year border closure, and dramatic cultural investment volatility (25.1% weight, highest among all indicators). The 25.1% weight on culture industry investment growth emphasizes critical importance of stable cultural funding in heritage-based tourism models.

Dimensional analysis demonstrates multifaceted development patterns: Eco-Environment showing consistent strength (scores 0.72–0.89), Innovation & Culture and Economy & Infrastructure displaying moderate performance with gradual improvement (0.38–0.52 for Economy), and Tourism & Livelihood exhibiting extreme volatility (0.61 to 0.09 to 0.31). These patterns, validated through structural break tests (Chow F = 12.84, p = 0.008) and paired comparisons (2019–2020 decline: t = 4.82, p < 0.001), provide evidence-based guidance for balanced policy development.

The three-tier prioritized policy framework, grounded in scenario analysis, demonstrates clear implementation pathways: Tier 1 (tourism recovery and cultural stabilization, ¥4.5 billion, 45% potential impact), Tier 2 (innovation and infrastructure, ¥2.3 billion, 15% stable improvement), Tier 3 (environmental excellence and regional cooperation, ¥750 million, 8% foundation benefits). Total ¥7.55 billion budget allocation over 2024–2028 provides specific, actionable measures with quantified outcomes, timelines, and governance structures. Policy scenario analysis confirms tourism enhancement offering highest impact (45% improvement) with volatility requiring management, innovation strategies providing middle-ground stability (15% increase), environmental policies ensuring foundation benefits (8% increase), and infrastructure demonstrating consistent moderate progress.

Methodologically, the framework contributes innovations through hierarchical weight integration combining dimensional and indicator-level analysis, entropy weighting effectively handling indicator heterogeneity while maintaining objectivity, and multi-temporal analysis capability proving valuable for capturing disruption and recovery patterns. The comprehensive validation approach—integrating sensitivity analysis, statistical testing, international comparison, and scenario modeling—establishes new standards for rigor in tourism evaluation research.

Certain limitations warrant acknowledgment: entropy weighting favors high-volatility indicators, potentially overemphasizing short-term fluctuations; data constraints (6 years, 30 indicators, 97.3% completeness) limit temporal scope and analytical depth.

The expert panel (n = 11), while systematically selected across diverse expertise domains and geographical regions, represents a relatively small sample. Larger panels (n = 20–30) would enhance statistical robustness of consensus measures and capture broader perspective diversity. However, our panel achieved strong consensus (Kendall's W = 0.68, p < 0.001) and high response rates (100% both rounds), suggesting adequate reliability for the study's purposes. Future applications in larger administrative units might benefit from expanded panel sizes.

Specifically, the 6-year time series constrains statistical power for complex modeling. Quadratic regression analysis, while achieving significance (p = 0.035), operates with limited degrees of freedom (df = 3), requiring cautious interpretation. The significant result is strengthened by convergent evidence from multiple analytical approaches (Chow structural break test, paired t-tests, bootstrap confidence intervals), which collectively support the U-shaped trajectory interpretation. Nonetheless, longer time series would enable more robust statistical inference and better capture cyclical patterns. Future longitudinal extensions with 10+ years of data would substantially enhance analytical capabilities for trend decomposition, forecasting validation, and model comparison.

Additional research opportunities include dynamic weighting approaches adjusting for development phases, spatial analysis integration for multi-regional assessment, stakeholder perspective integration beyond expert opinion, and longitudinal extension enabling trend forecasting. Empirical applications to other regions would enhance generalizability and enable systematic comparative analysis.

The framework provides immediately applicable tools for evidence-based policy formulation and resource allocation in rural tourism development. The empirical results offer specific guidance for Shaanxi Province while demonstrating broader principles transferable to similar contexts. In post-pandemic recovery and sustainable development contexts, this research contributes timely insights into resilience mechanisms, policy effectiveness, and recovery strategies.

The research ultimately demonstrates that sustainable rural tourism development requires sophisticated analytical frameworks capable of capturing complexity, uncertainty, and dynamic change. The integrated Delphi-entropy approach provides such a framework, offering both methodological rigor and practical relevance. The framework's emphasis on multidimensional assessment, comprehensive validation, and integrated policy approaches aligns with contemporary sustainability paradigms and supports evidence-based rural tourism development in an increasingly complex and uncertain world.