Dynamic evaluation of global sustainability: longitudinal ranking and spatial profiling across countries Authors: Yin-Yin Huang, Xiangyan Wang, Lei Jiang, Tsai-Sung Lin, Ruey-Chyn Tsaur Published: 2026 ABSTRACT AbstractThis study addresses the limitations of static evaluation models in global sustainability assessment by tracing longitudinal sustainability trajectories across 103 countries. We develop an integrated analytical framework that combines Dynamic TOPSIS with time-decayed entropy weighting to capture the temporal evolution of sustainability performance from 2014 to 2024. To enhance robustness, linear ranking results are cross-validated through non-linear spatial classification using K-means clustering with k-means + + initialization. Seven core indicators spanning economic, social, and environmental dimensions are employed to characterize the structural profiles of sustainability tiers. The results reveal a pronounced staircase effect in global sustainability, with countries converging into three stable tiers. The three-cluster solution is supported by a reduction in the Sum of Squared Errors from 26.34 to 8.06, with diminishing marginal improvements thereafter. Robustness tests show stable rankings, with no rank reversal among the top 20 countries and consistent positions within the top five. A notable gap is found in the Lagging Tier: despite relatively high environmental efficiency driven by low industrialization, sustainability progress is held back by weak human capital and institutional quality. The Leading Tier, by contrast, maintains balanced performance through sound governance and environmental management. By giving greater weight to recent observations through time-decayed weighting and cross-checking results through both linear and spatial methods, this study offers a dynamic alternative to static sustainability rankings. The findings point to education and governance gaps as key barriers in lower-tier nations, with implications for international interventions that go beyond purely economic assistance.Keywords: Dynamic TOPSIS, K-means clustering, Time-decayed weights, Sustainability tiers, Structural bottlenecksSubject terms: Environmental sciences, Environmental social sciences, Geography, Geography, Mathematics and computing Introduction The implementation of the 2030 Agenda for Sustainable Development requires high-precision evaluation frameworks to address regional disparities1–3. While GDP per capita has traditionally served as a primary developmental metric, its inability to reflect ecological and social dimensions has led to the exploration of multidimensional indexes4–6. Recent global instability underscores the need to integrate climate risk and governance quality into national performance assessments7–9. Furthermore, identifying institutional drivers such as Rule of Law and Human Capital are increasingly seen as prerequisites for nations to escape developmental stagnation10–12. Despite these trends, existing evaluation frameworks remain limited by three critical issues: (1) a continued reliance on static GDP-centric models that mask ecological trade-offs4,6; (2) a dilution effect where the aggregation of longitudinal data without temporal decay obscures recent policy progress13,14; and (3) a lack of structural validation to identify institutional bottlenecks hindering the transition between development tiers11,15. While TOPSIS is a standard tool for national ranking, current models often overlook the evolving utility of temporal information and fail to provide spatial validation for identified rankings16–19. In response to these gaps, this study develops an integrated evaluation framework leveraging six methodological scenarios (M1-M6) to cross-verify national sustainability performance. As detailed in the subsequent comparative analysis (Sect. "Comparative analysis of experimental scenarios and global rankings"), we establish static baselines (M1, M2) to contrast with our primary Dynamic Main Model (M3), while incorporating consistency (M4) and stability (M5) validations (utilizing Grey Relational Analysis and Monte Carlo simulations) alongside tier-based clustering (M6)16,20–22. This research seeks to solve: (1) how global trajectories evolve when prioritizing contemporary information utility; (2) the stability of rankings under stochastic perturbations; and (3) the structural indicators constituting the primary bottlenecks for lagging nations. This work provides three original contributions spanning both theoretical and applied dimensions of sustainability evaluation. First, we develop a framework integrating six methodological scenarios (M1-M6) as a theoretically grounded cross-verification protocol that extends beyond conventional single-scenario assessments23. Second, this research introduces a dynamic weighting mechanism within the M3 model that captures the evolving utility of information, providing a more responsive trajectory analysis than traditional models14. Finally, our findings leverage the integration of dynamic ranking and spatial clustering to pinpoint the exact institutional thresholds required for tier transition, offering a level of policy granularity missing from aggregate ranking systems7,24. Literature review Multidimensional indicators for sustainable development Following the thematic framework established by the United Nations3, this study selects seven core proxy indicators to capture the multifaceted nature of global sustainability. Given the multifaceted nature of sustainable development, these metrics serve as representative proxies for three main pillars: Economic (GDP per capita and Adjusted Net Savings), Social (Education Index, Life Expectancy, and Rule of Law), and Environmental (CO2 Emissions and the Environmental Performance Index). This selection aligns with the multidimensional evaluation approaches validated in recent literature4–6,25,26, so that the selected proxies cover both human development and ecological efficiency1,27. Economic and social dimensions: beyond income metrics The transition from traditional GDP per capita centric evaluations to multidimensional sustainability frameworks represents a fundamental shift in development economics4,6. Scholars argue that while GDP per capita reflects average economic output (SDG 8 & 9), it fails to account for the depreciation of natural capital and the long-term social costs of rapid industrialization5,11,28. In this context, Adjusted Net Savings (ANS) has emerged as a useful metric for evaluating genuine wealth, as it integrates environmental depletion and human capital investments into a nation’s savings profile1,2. These metrics together reflect the interplay between human development and institutional strength10–12. Furthermore, the Rule of Law (SDG 16) and institutional quality are identified as essential prerequisites for stable progress, as environmental regulations and democratic governance significantly moderate the impact of economic activities on sustainability7,10–12. Environmental dimension: carbon intensity and ecological efficiency The inclusion of CO2 Emissions and the Environmental Performance Index (EPI) is essential for tracking long-term sustainability trajectories11,26,27. As countries transition toward higher income levels, the challenge of decoupling economic growth from carbon intensity becomes paramount6,29–32. Furthermore, environmental performance is significantly moderated by institutional regulations and sectoral sustainability targets10,33,34. Consequently, robust policy frameworks are required to maintain ecosystem health and ensure that national trajectories remain statistically robust15,16,24. Methodological evolution in sustainability assessment Multi-criteria decision making (MCDM) in SDGs The complexity of evaluating Sustainable Development Goals (SDGs), characterized by conflicting economic and environmental objectives, necessitates the use of Multi-Criteria Decision-Making (MCDM) methodologies to synthesize heterogeneous data17,23,35–37. Among various MCDM techniques, the TOPSIS method(originally developed by38 is widely used to identify the shortest distance to ideal development targets13,18,19,38,39. To reduce subjectivity, researchers have used the Entropy Weight Method (EWM) to determine weights based on data dispersion14,28,39. These frameworks provide the foundation for assessing institutional commitment and ecological efficiency through a unified, evidence-based lens23,40,41. The shift toward dynamic and longitudinal models A significant evolution in recent literature is the movement beyond static averages toward dynamic longitudinal frameworks. Wang et al.14, have identified a dilution effect in traditional studies, arguing that equal weighting of longitudinal data obscures recent policy gains13,14. To address this issue, current research increasingly adopts dynamic weight allocation mechanisms that adjust indicator importance across multiple time points to capture fluctuating development trajectories14. These scholars emphasize the concept of Information Utility, where contemporary observations are assigned higher coefficients through time-decayed algorithms so that the evaluation reflects recent national trajectories14. This theoretical shift establishes Information Utility as a primary criterion for longitudinal assessments, emphasizing that the temporal relevance of data is as critical as its statistical weight in capturing evolving sustainability trends21,42. Robustness, stability, and spatial validation: ensuring scientific rigor Uncertainty and stochastic stability testing The reliability of national rankings in complex MCDM models is frequently challenged by the sensitivity of weight assignments and inherent data noise16,21,40,41. To address these challenges, Monte Carlo simulation has emerged as a key technique for assessing the stochastic stability of multi-criteria assessments16,43. Baudry et al.43, combined Monte Carlo simulation with multi-actor multi-criteria analysis to account for parametric uncertainty, showing that probabilistic outputs improve decision confidence. Building on this, Cui et al.44, have extended this framework to hybrid MCDM models, where Monte Carlo iterations are coupled with weighted data structures to improve both ranking stability and decision reliability. By introducing random perturbations to indicator weights (typically 1,000 to 10,000 iterations) this method allows researchers to observe the probabilistic distribution of rankings rather than a single deterministic result16,21,42–44. Furthermore, such stochastic modeling ensures that performance benchmarks are statistically robust across diverse weighting configurations, significantly enhancing the credibility of policy recommendations40,41. Consistency analysis in hierarchical rankings In addition to stochastic stability, existing literature emphasizes that ensuring ordinal consistency between different methodological configurations is a prerequisite for scientific cross-verification13,23. Studies have shown that traditional correlation coefficients often fail to capture the non-linear nuances of ranking shifts in longitudinal data45. To address this methodological gap, Grey Relational Analysis (GRA) offers an alternative by evaluating the geometric similarity and trajectory overlaps between various ranking outputs13,41. This helps determine whether ranking changes reflect genuine performance shifts or are artifacts of weighting sensitivity. Consequently, the application of GRA has become a standard protocol for validating the transition between static baselines and more complex dynamic evaluation models19,23. Unsupervised machine learning for performance tiering Complementing ordinal assessments, scholars have increasingly integrated unsupervised machine learning to add a spatial and structural dimension to sustainability analysis20,22,46. Recent studies suggest that the K-means + + clustering algorithm offers greater stability in identifying performance tiers compared to traditional grouping methods20,22. By categorizing nations based on multidimensional structural similarities, this approach allows examination of how institutional variables differentiate high-performing from lower-performing clusters11,12,20. Identification of research gaps A critical synthesis of the current literature identifies three fundamental gaps that this research addresses: Temporal Utility Deficit: While longitudinal data is abundant, existing frameworks often fail to integrate dynamic weighting mechanisms that prioritize contemporary information utility13,14. This limitation results in evaluations that undervalue recent policy advancements and developmental trajectories, potentially leading to outdated policy insights21,42. Temporal Utility Deficit: While longitudinal data is abundant, existing frameworks often fail to integrate dynamic weighting mechanisms that prioritize contemporary information utility13,14. This limitation results in evaluations that undervalue recent policy advancements and developmental trajectories, potentially leading to outdated policy insights21,42. Lack of Multi-Scenario Validation: Few studies establish a systematic protocol to compare static baselines against dynamic models through both stochastic and relational consistency tests16,41. Most studies rely on single-method outputs, limiting the ability to detect methodological artifacts23,40. Lack of Multi-Scenario Validation: Few studies establish a systematic protocol to compare static baselines against dynamic models through both stochastic and relational consistency tests16,41. Most studies rely on single-method outputs, limiting the ability to detect methodological artifacts23,40. Fragmented Linkage between Ranking and Tiering: Current studies often treat hierarchical ranking and spatial clustering as separate tasks, and often miss the chance to identify the specific institutional bottlenecks (e.g., Rule of Law and Education) that define the boundaries between global development tiers11,15. This disconnection hinders a holistic understanding of structural barriers and the pathways for tier transitions12,20,46. Fragmented Linkage between Ranking and Tiering: Current studies often treat hierarchical ranking and spatial clustering as separate tasks, and often miss the chance to identify the specific institutional bottlenecks (e.g., Rule of Law and Education) that define the boundaries between global development tiers11,15. This disconnection hinders a holistic understanding of structural barriers and the pathways for tier transitions12,20,46. Methodology This study proposes a comprehensive dynamic multi-criteria decision-making (MCDM) framework to evaluate the sustainable development performance of 103 countries. Unlike conventional static assessments that rely on single-period snapshots, our approach captures both the magnitude and the temporal trajectory of sustainability across an 11-year observation window from 2014 to 2024. The systematic research framework, illustrated in Fig. 1, begins with data preprocessing of seven core SDG indicators. Logarithmic transformation is applied to reduce skewness and variance in economic and environmental metrics4–6. The core analysis contrasts static baselines (M1 and M2) with the dynamic main model (M3), which integrates a Dynamic Entropy Weight Method (DEWM) with time-decayed information utility to give greater weight to recent development trends14,28. Systematic Methodological Framework for Dynamic Sustainability Evaluation (M1–M6). Robustness is assessed through Grey Relational Analysis (M4) for ordinal consistency13,41 and Monte Carlo simulations (M5) for stochastic stability testing under weight perturbations16,21. The framework also uses K-means + + clustering (M6) to categorize global development tiers and identify institutional bottlenecks that hinder sustainable transitions20,22,46. Data collection and preprocessing The research dataset centers on 103 representative countries, categorized according to the World Bank’s income classification. In order to uphold scientific rigor and the validity of the longitudinal analysis, the initial data pool from 195 nations reported by the World Bank was subjected to a strict screening process. Nations with missing values in any of the seven indicators across the entire 2014–2024 observation window were excluded, resulting in a balanced panel of 103 countries (1,133 observations). This eliminates imputation bias and supports the reliability of subsequent entropy weight calculations. Despite the exclusion of certain regions due to data incompleteness, the final sample spans all major continents and development stages, covering a broad cross-section of global sustainability conditions. Seven core indicators aligned with the United Nations Sustainable Development Goals (SDGs) were selected to provide a holistic view of national performance: Economic Dimension: GDP per capita and Adjusted Net Savings (ANS) (SDG 8 & 9). Economic Dimension: GDP per capita and Adjusted Net Savings (ANS) (SDG 8 & 9). Social Dimension: Education Index, Life Expectancy, and Rule of Law (SDG 3, 4 & 16). Social Dimension: Education Index, Life Expectancy, and Rule of Law (SDG 3, 4 & 16). Environmental Dimension: CO₂ emissions (per capita) and the Environmental Performance Index (EPI) (SDG 13 & 15). Environmental Dimension: CO₂ emissions (per capita) and the Environmental Performance Index (EPI) (SDG 13 & 15). The theoretical justification for each indicator is detailed in the section "Multidimensional indicators for sustainable development". The selection was further guided by the five criteria proposed by Keeney and Raiffa47: completeness, operationality, decomposability, non-redundancy, and minimum size. Each indicator addresses a distinct dimension of sustainable development without conceptual overlap, remains consistently measurable across the 103-country panel over the full 2014–2024 window, and has established precedent in validated sustainability assessment frameworks4,6,25. The number of indicators also falls within the 7 ± 2 range recommended by Saaty48, who drew on Miller’s (1956) cognitive manageability principle to suggest that decision matrices with fewer than nine criteria maintain greater judgemental consistency in MCDM applications. Data were collected from the World Bank, Yale University, and the UNDP. For cost-type indicators (e.g., CO₂ emissions per capita), the raw data were first log-transformed (with a small constant added if necessary to handle non-positive values), then inverted to align with benefit-type criteria in MCDM. This inversion ensures that higher normalized scores represent lower per capita emissions, as suggested by Dědeček and Dudzich4, DiMaria5, and Yan et al.6. All variables were subsequently standardized using min-max normalization to the range [0.0001, 1.0001] (a small positive constant ε = 0.0001 was applied to prevent exact zeros, ensuring numerical stability in entropy weighting and TOPSIS computations): where denotes the original value of the -th indicator for the -th country, and represent the extreme values across the sample, and is the resulting normalized value. This preprocessing step ensures the objectivity of the subsequent DEWM and TOPSIS-GRA calculations. Dynamic weighting mechanism To ensure the objectivity of the evaluation, this study employs the Dynamic Entropy Weight Method (DEWM) to determine the relative importance of the seven SDG-aligned indicators. Traditional weighting methods often overlook the temporal fluctuations in panel data; however, DEWM captures the information utility by considering the variability of indicators across both the cross-sectional (103 countries) and time-series (11 years) dimensions. Step 1: Calculation of Indicator Proportions. For each normalized indicator , the proportion is calculated to represent its relative contribution within the entire observation window: Step 2: Information Entropy Measurement. The information entropy for the -th indicator is computed as follows: where n = 103 and T = 11. A higher entropy value indicates lower information utility and, consequently, a smaller weight. Step 3: Weight Derivation. The redundancy of information is defined as . The final dynamic indicator weights () are then derived: By utilizing this dynamic weighting scheme, the model automatically adjusts the influence of each indicator based on its actual data distribution from 2014 to 2024, ensuring that the final ranking is driven by the most informative sustainable development trajectories. Step 4: Integration of Temporal Significance Factors. To prioritize contemporary data in sustainability transitions, a linear time-decayed significance factor is introduced to emphasize recent information utility. For each year t in the 11-year window, the temporal weight is defined as: This linear progression (from 1/66 to 11/66) provides a transparent and objective escalation of temporal importance, avoiding the sensitivity issues often associated with exponential decay functions that depend on arbitrary decay constants. The final hybrid weight for indicator j in year t is obtained by integrating the static entropy-derived weight with the temporal factor , followed by normalization: This mechanism ensures that the model prioritizes recent sustainable development trajectories while preserving the objective information utility of the indicators across the panel. Stability tests in M5 further confirm that this linear weighting effectively captures contemporary progress without compromising long-term structural integrity. Dynamic TOPSIS method for longitudinal ranking To evaluate the global sustainable development performance across a decade, this study employs the Dynamic TOPSIS method, a sophisticated extension of the multi-criteria decision-making framework. Unlike conventional static assessments that provide only cross-sectional snapshots, our approach constructs a three-dimensional decision matrixintegrating dimensions of Country (i = 1, …, n), indicators (j = 1, …, m), and Time (t = 1, …, T). This spatio-temporal structure supports a longitudinal assessment of 103 countries over an 11-year period (2014–2024) using 7 SDG-aligned indicators. By capturing fluctuations within the panel data, the model ensures that the final relative closeness () scores reflect both a nation’s performance at specific points in time and its trajectory across the observation window. The formalized ranking process is detailed through the following refined steps: Step 1: Construction of the Dynamic Weighted Normalized Matrix. The standardized values () are multiplied by the hybrid weights () defined in the section "Dynamic weighting mechanism" to generate the weighted normalized decision matrix V, where each element represents the performance of country i for indicator j in year t: Step 2: Identification of the Dynamic Positive and Negative Ideal Solutions. In the dynamic context, the Positive Ideal Solution () and Negative Ideal Solution () are determined by identifying the optimal and least favorable values for each indicator across the entire spatio-temporal scope: , for j = 1m (8). Step 3: Calculation of Longitudinal Euclidean Distances. To capture the cumulative performance over the 11-year window, the Euclidean distance for each country i from the dynamic ideal solutions is calculated by integrating the differences across all indicators and time points: Step 4: Derivation of the Relative Closeness Coefficient (). The final sustainability score () is computed to rank the 103 countries. A higher indicates that a country’s overall trajectory is closer to the positive ideal path: where Unlike traditional separate rankings, this unified allows for a direct comparison of all nations regardless of their income levels. K-means clustering for tiered classification To identify the latent structural patterns of global sustainability and validate the consistency of the Dynamic TOPSIS results, this study employs the K-means clustering algorithm. Instead of relying solely on traditional income-based groupings, this unsupervised learning approach partitions the 103 countries into distinct performance tiers based on their 11-year dynamic score vectors. The overall objective function of the K-means clustering is to minimize the total within-cluster sum of squares (WCSS): where represents the dynamic performance vector of country i over T = 11 years, and is the centroid of cluster k. The squared Euclidean distance is calculated as: Operational Steps for Tiered Analysis: Feature Selection: The input features are the longitudinal scores derived from the Dynamic TOPSIS model, capturing the comprehensive sustainability trajectory of each nation. Feature Selection: The input features are the longitudinal scores derived from the Dynamic TOPSIS model, capturing the comprehensive sustainability trajectory of each nation. Optimal Cluster Determination: While the World Bank utilizes four income categories, the Elbow Method was applied to our dynamic dataset to ensure statistical rigor. By plotting the total Within-Cluster Sum of Squares (WCSS) against K values, we identified the elbow point where the rate of variance reduction begins to plateau. The results indicated that K = 3 provides the most appropriate partition, effectively categorizing the 103 nations into Leading, Transition, and Lagging tiers. Optimal Cluster Determination: While the World Bank utilizes four income categories, the Elbow Method was applied to our dynamic dataset to ensure statistical rigor. By plotting the total Within-Cluster Sum of Squares (WCSS) against K values, we identified the elbow point where the rate of variance reduction begins to plateau. The results indicated that K = 3 provides the most appropriate partition, effectively categorizing the 103 nations into Leading, Transition, and Lagging tiers. Centroid Initialization via k-means++: To reduce the risk of converging to local optima, a common limitation of random initialization, this study utilizes the k-means + + seeding technique. This algorithm selects initial centroids that are spread far apart in the feature space, improving the stability and replicability of the cluster assignments. For each country, the Euclidean distance to these centroids is calculated, and the country is assigned to the cluster with the nearest mean. Centroid Initialization via k-means++: To reduce the risk of converging to local optima, a common limitation of random initialization, this study utilizes the k-means + + seeding technique. This algorithm selects initial centroids that are spread far apart in the feature space, improving the stability and replicability of the cluster assignments. For each country, the Euclidean distance to these centroids is calculated, and the country is assigned to the cluster with the nearest mean. Update and Convergence: The centroids are recalculated as the mean of all countries assigned to each cluster. This process repeats until the assignments stabilize or a maximum of 100 iterations is reached. Update and Convergence: The centroids are recalculated as the mean of all countries assigned to each cluster. This process repeats until the assignments stabilize or a maximum of 100 iterations is reached. Structural Validation: The resulting clusters are cross-referenced with the rankings. A high degree of overlap between top-ranked countries and the “Leading Tier” confirms the mathematical stability of the evaluation framework. Structural Validation: The resulting clusters are cross-referenced with the rankings. A high degree of overlap between top-ranked countries and the “Leading Tier” confirms the mathematical stability of the evaluation framework. Robustness and consistency tests To ensure the scientific validity and reliability of the dynamic evaluation framework, this study conducts a dual-layered robustness test comprising Grey Relational Analysis (GRA) and Monte Carlo simulation. Consistency check via GRA While Dynamic TOPSIS measures the distance from ideal solutions, GRA is employed to assess the similarity between the development trajectories of the 103 countries and the positive ideal path. The Grey Relational Grade () is calculated to provide an alternative ranking: where is the relational coefficient. The consistency between the TOPSIS ranking () and the GRA ranking () is quantified using Spearman’s rank correlation coefficient (): A correlation coefficient of 0.9989 confirms that the ranking results are highly consistent and independent of the specific mathematical distance metric used. Stability check via Monte Carlo simulation To test the sensitivity of the identified tiers (Leading, Transition, and Lagging) to potential variations in indicator weights, a Monte Carlo simulation with 1,000 iterations was executed. In each simulation, a stochastic disturbance () is introduced to the dynamic weights (): where follows a random uniform distribution. By analyzing the distribution of 1,000 simulated rankings, the Stability Index is derived. The results demonstrate that the cluster assignments remain robust against weight fluctuations, ensuring the objectivity of the tiered classification. Results and discussion Data overview and descriptive statistics To ensure mathematical consistency and mitigate skewness, a Conditional Logarithmic Transformation was applied to the raw data prior to normalization. The transformation is defined as y = ln(x ++ 1), where x represents the raw observation. For indicators with potential negative values (ANS and ROL), the shifting parameter is set to the absolute minimum value of the entire panel (). For all other indicators, is set to zero. This unified approach stabilizes the variance across the longitudinal dataset (M3) while providing a robust foundation for static comparisons (M1, M2). Data were compiled from the World Bank (WDI) and the Sustainable Development Solutions Network (SDSN). Data were compiled from the World Bank (WDI) and the Sustainable Development Solutions Network (SDSN). Indicators: GDP: GDP per capita; ANS: Adjusted Net Savings; EDU: Education Index; LIF: Life Expectancy; RUL: Rule of Law Index; CO2: CO2 Emissions Intensity; EPI: Environmental Performance Index. Indicators: GDP: GDP per capita; ANS: Adjusted Net Savings; EDU: Education Index; LIF: Life Expectancy; RUL: Rule of Law Index; CO2: CO2 Emissions Intensity; EPI: Environmental Performance Index. Data were processed via logarithmic scaling and Min-Max normalization to [0.0001, 1.0001] to ensure computational stability. Data were processed via logarithmic scaling and Min-Max normalization to [0.0001, 1.0001] to ensure computational stability. The statistical profile in Table 1 reveals several critical characteristics: Table 1Descriptive Statistics of Normalized Sustainability Indicators (2014–2024).IndicatorGDPANSEDULIFRULCO2EPIMean0.55300.81180.65220.67960.60620.58360.6731Std. Dev.0.22910.07750.18620.22050.22960.20370.1963Min0.00010.00010.00010.00010.00010.00010.0001Max1.00011.00011.00011.00011.00011.00011.0001Skewness−0.1488−3.7797−1.0742−0.8269−0.1711−0.0847−0.5416Kurtosis−0.793632.80750.57260.2226−0.8417−0.2141−0.3925Note:. Descriptive Statistics of Normalized Sustainability Indicators (2014–2024). Note:. Normalization and Normality: After logarithmic scaling, the GDP index exhibits a near-normal distribution with a skewness of −0.1488, a improvement from the raw data’s extreme positive skewness. This allows the DEWM-TOPSIS model to better differentiate between middle-income and high-income nations without undue influence from extreme outliers. Normalization and Normality: After logarithmic scaling, the GDP index exhibits a near-normal distribution with a skewness of −0.1488, a improvement from the raw data’s extreme positive skewness. This allows the DEWM-TOPSIS model to better differentiate between middle-income and high-income nations without undue influence from extreme outliers. Savings Sensitivity: Adjusted Net Savings (ANS) exhibits a high kurtosis (32.8075), indicating that while most nations maintain consistent savings patterns, specific economies experience sharp fluctuations. The inclusion of a shifting constant during transformation captured these variances without affecting mathematical stability. Savings Sensitivity: Adjusted Net Savings (ANS) exhibits a high kurtosis (32.8075), indicating that while most nations maintain consistent savings patterns, specific economies experience sharp fluctuations. The inclusion of a shifting constant during transformation captured these variances without affecting mathematical stability. Institutional and Environmental Variance: The Rule of Law (RUL) and CO2 Intensity indicators maintain substantial standard deviations (0.2296 and 0.2037, respectively). This high variance contributes significant informational entropy to the DEWM model, supporting their roles as key differentiators in sustainability rankings. Institutional and Environmental Variance: The Rule of Law (RUL) and CO2 Intensity indicators maintain substantial standard deviations (0.2296 and 0.2037, respectively). This high variance contributes significant informational entropy to the DEWM model, supporting their roles as key differentiators in sustainability rankings. Comparative analysis of experimental scenarios and global rankings Methodological framework of analysis scenarios To ensure the robustness and objectivity of the sustainability evaluation, this study designed six experimental scenarios (M1-M6) to cross-validate the results from different methodological perspectives. As outlined in Table 2, the Dynamic (M3) scenario serves as the primary longitudinal model, utilizing the Dynamic Entropy Weight Method (DEWM) to capture the evolving trajectories of 103 nations over 11 years. Scenarios M1, M2, and M4-M6 provide critical benchmarks for static comparison, weight sensitivity, and structural stability. Table 2Summary of Experimental Scenarios and Methodological Frameworks.ScenarioWeighting MethodTemporal FrameworkDecision ModelPurpose of AnalysisM1: Static BaselineEntropy11-year AverageStatic TOPSISEstablish a cross-sectional baseline for comparisonM2: Equal BaselineEqual11-year AverageStatic TOPSISTest weight sensitivity in traditional static modelsM3: Dynamic (Main)DEWMLongitudinal (11y)Dynamic TOPSISPrimary model for capturing longitudinal trajectoriesM4: ConsistencyDEWMLongitudinal (11y)GRAValidate ranking consistency via curve similarityM5: StabilityStochasticLongitudinal (11y)Monte CarloTest robustness against weight fluctuationsM6: Clustering–Longitudinal (11y)K-MeansCategorize countries into distinct performance tiers Summary of Experimental Scenarios and Methodological Frameworks. Analysis of global sustainability rankings The implementation of the M3 model yielded the Dynamic Closeness Scores (), representing the relative sustainability performance of 103 countries. Table 3 presents the top, median, and bottom performers, revealing a diverse global landscape of sustainable development. Table 3Global Sustainability Rankings Based on the Dynamic M3 Model.RankCountryScoreRegion (World Bank)Income GroupSustainability Tier (M6)Top 101Switzerland0.8160Europe & Central AsiaHigh incomeLeading2Sweden0.8116Europe & Central AsiaHigh incomeLeading3Denmark0.7939Europe & Central AsiaHigh incomeLeading4United Kingdom0.7790Europe & Central AsiaHigh incomeLeading5Norway0.7762Europe & Central AsiaHigh incomeLeading6Finland0.7760Europe & Central AsiaHigh incomeLeading7France0.7736Europe & Central AsiaHigh incomeLeading8Malta0.7706Europe & Central AsiaHigh incomeLeading9Ireland0.7694Europe & Central AsiaHigh incomeLeading10Austria0.7667Europe & Central AsiaHigh incomeLeadingMedian 1047Qatar0.6042Middle East & North AfricaHigh incomeTransition48Colombia0.6017Latin America & CaribbeanUpper middle incomeTransition49Georgia0.6008Europe & Central AsiaUpper middle incomeTransition50Malaysia0.6008East Asia & PacificUpper middle incomeTransition51Bahrain0.6007Middle East & North AfricaHigh incomeTransition52Saudi Arabia0.6004Middle East & North AfricaHigh incomeTransition53Dominican Republic0.5997Latin America & CaribbeanUpper middle incomeTransition54Turkiye0.5993Europe & Central AsiaUpper middle incomeTransition55Thailand0.5962East Asia & PacificUpper middle incomeTransition56Jordan0.5961Middle East & North AfricaLower middle incomeTransitionBottom 1094Cambodia0.4397East Asia & PacificLower middle incomeLagging95Tajikistan0.4368Europe & Central AsiaLower middle incomeLagging96Tanzania0.4334Sub-Saharan AfricaLower middle incomeLagging97Cote d’Ivoire0.4318Sub-Saharan AfricaLower middle incomeLagging98Pakistan0.4158South AsiaLower middle incomeLagging99Lao PDR0.4119East Asia & PacificLower middle incomeLagging100Cameroon0.4054Sub-Saharan AfricaLower middle incomeLagging101Benin0.4035Sub-Saharan AfricaLower middle incomeLagging102Burkina Faso0.4035Sub-Saharan AfricaLow incomeLagging103Madagascar0.3707Sub-Saharan AfricaLow incomeLagging Global Sustainability Rankings Based on the Dynamic M3 Model. Key findings from the refined evaluation The revised ranking results reflect three major insights that distinguish this study from traditional sustainability assessments: Dominance of Institutional and Environmental Stability: Under the log-transformed DEWM framework, European nations consistently occupy the top 10 positions (e.g., Switzerland, Sweden). This dominance is attributed not only to economic output but primarily to their superior performance in the Rule of Law (RUL) and CO2 Intensity metrics. Notably, after mitigating extreme skewness through logarithmic scaling, the DEWM model identifies institutional quality as a more significant differentiator of sustainability than raw economic volume. Dominance of Institutional and Environmental Stability: Under the log-transformed DEWM framework, European nations consistently occupy the top 10 positions (e.g., Switzerland, Sweden). This dominance is attributed not only to economic output but primarily to their superior performance in the Rule of Law (RUL) and CO2 Intensity metrics. Notably, after mitigating extreme skewness through logarithmic scaling, the DEWM model identifies institutional quality as a more significant differentiator of sustainability than raw economic volume. Decoupling of Wealth and Sustainability: A notable finding is the relative underperformance of certain high-income nations. For instance, Qatar (Rank 47) and Saudi Arabia (Rank 52), despite their exceptionally high GDP per capita, are positioned in the Transition Tier. This suggests that high economic affluence does not automatically translate into better sustainability outcomes when heavy environmental burdens (CO2) and institutional maturity are integrated into the evaluation. Decoupling of Wealth and Sustainability: A notable finding is the relative underperformance of certain high-income nations. For instance, Qatar (Rank 47) and Saudi Arabia (Rank 52), despite their exceptionally high GDP per capita, are positioned in the Transition Tier. This suggests that high economic affluence does not automatically translate into better sustainability outcomes when heavy environmental burdens (CO2) and institutional maturity are integrated into the evaluation. Regional Disparities and Development Traps: A geographic concentration is seen among the Lagging Tier performers, with Sub-Saharan African nations (e.g., Madagascar, Burkina Faso) facing multi-dimensional challenges. The results suggest a sustainability trap where low scores in education (EDU) and life expectancy (LIF) consistently limit the overall performance of low-income economies. Regional Disparities and Development Traps: A geographic concentration is seen among the Lagging Tier performers, with Sub-Saharan African nations (e.g., Madagascar, Burkina Faso) facing multi-dimensional challenges. The results suggest a sustainability trap where low scores in education (EDU) and life expectancy (LIF) consistently limit the overall performance of low-income economies. To assess the external validity of the M3 rankings, Spearman rank correlations were computed between the M3 Dynamic TOPSIS results and two established global benchmarks: the 2025 SDG Index8 and the Human Development Index (UNDP, 2024). As summarized in Table 4, both correlations are statistically significant and positive, confirming strong ordinal consistency between the M3 framework and external assessments. The leading nations in the M3 rankings, including Switzerland, Sweden, and Denmark, consistently appear among the top performers in both benchmarks. Notably, the M3 model captures nuances not reflected in income-centric indices; for instance, Qatar ranks 47th in M3 but considerably higher in the HDI, reflecting the penalty imposed by its environmental burden under the multi-dimensional dynamic weighting scheme. These results confirm that the M3 framework aligns with established sustainability assessments while providing greater sensitivity to temporal dynamics and institutional dimensions. Table 4External Benchmark Validation of M3 Rankings.Benchmark n Spearman ρp-valueInterpretationSDG Index 202581010.835< 0.001Strong positive correlationHuman Development Index 2023 (UNDP, 2024)1020.933< 0.001Very strong positive correlationSpearman rank correlation coefficients between M3 Dynamic TOPSIS rankings and external benchmark rankings. p-values based on two-tailed tests. External Benchmark Validation of M3 Rankings. Spearman rank correlation coefficients between M3 Dynamic TOPSIS rankings and external benchmark rankings. p-values based on two-tailed tests. Robustness and comparative analysis The M3 model represents an Objective Dynamic Framework based on the Dynamic Entropy Weight Method (DEWM). It is characterized as ‘objective’ because it eliminates human bias through entropy-based weight allocation, and ‘dynamic’ because it incorporates temporal variances across the 11-year study period. Temporal dynamics: static (M1) vs. dynamic (M3) models To validate the necessity of the dynamic framework, a comparative analysis between the Static Baseline (M1) and the Dynamic TOPSIS model (M3) was conducted. While the rankings of nations in the Leading Tier remain relatively stable, the most profound rank displacements are observed within the Transition and Lagging Tiers. Following a rigorous selection process, the top 10 nations with the largest Rank Shifts (ΔR) were identified to demonstrate how temporal dynamics recalibrate sustainability evaluations. It should be noted that M1 and M3 operate on different sample scales: M1 is derived from the full global dataset of 195 countries, while M3 is calculated strictly within the 103-country balanced panel, such that Static Rank values exceeding 103 are expected and do not constitute an error. As shown in Table 5, the dynamic model (M3) identifies growth trajectories that are masked by the 11-year averages used in M1. This suggests that the Time-Entropy weighting mechanism successfully rewards incremental progress and positive trends achieved in the latter half of the 2014–2024 period. For instance, nations such as Madagascar and Pakistan exhibit substantial upward shifts, indicating that their recent sustainability efforts carry more weight in the dynamic evaluation than their historical averages. Conversely, the downward shift for Belarus (+ 26) reflects a relative stagnation when temporal decay is applied. This shows that the M3 model provides a more sensitive reflection of recent developmental trends. Table 5Analysis of Significant Rank Shifts across Sustainability Tiers (Top 10 Shifts).Sustainability TierCountryDynamic Rank (M3)Static Rank (M1)Rank Shift (ΔR)InterpretationLaggingMadagascar103138−35Significant relative improvement in dynamic trajectory.LaggingPakistan98129−31Upward momentum captured by time-weighting.LaggingBenin101130−29Performance trend exceeds long-term static average.LaggingBurkina Faso102128−26Emerging stability in recent years compared to baseline.LaggingTanzania96122−26Gradual progress in sustainability indicators over time.TransitionBangladesh87113−26Strong recent growth not fully reflected in static mean.TransitionBelarus805426Downward trend or stagnation relative to global peers.TransitionSri Lanka6288−26Resilience in recent periods improving dynamic standing.LaggingNepal93118−25Consistent incremental gains captured by the M3 model.LaggingRwanda90115−25Effective policy execution leading to ranking recovery.ΔR = M3 Rank - M1 Rank. A negative value indicates an improved standing in the dynamic model compared to the global static baseline. Static Rank (M1) represents the nation’s position within the broader global dataset of 195 countries prior to sample refinement to the balanced panel of 103 countries used in the dynamic analysis. Values exceeding 103 are therefore expected. Analysis of Significant Rank Shifts across Sustainability Tiers (Top 10 Shifts). ΔR = M3 Rank - M1 Rank. A negative value indicates an improved standing in the dynamic model compared to the global static baseline. Static Rank (M1) represents the nation’s position within the broader global dataset of 195 countries prior to sample refinement to the balanced panel of 103 countries used in the dynamic analysis. Values exceeding 103 are therefore expected. Weighting logic: equal (M2) vs. objective (M3) weighting To evaluate the sensitivity of the results to the weighting mechanism, this study compared the Equal Weighting baseline (M2) with the Objective Dynamic Weighting (M3). In M2, all seven sustainability indicators are assigned a uniform weight of approximately 1/7 0.1428. In contrast, M3 utilizes the DEWM framework to derive objective weights based on informational entropy, allowing the data’s inherent dispersion to dictate the importance of each metric. As summarized in Table 6 the transition to objective weights triggers substantial rank displacements, particularly for nations whose rankings are heavily anchored in specific high-impact metrics. Following the same selection criteria as the section "Temporal dynamics: static (M1) vs. dynamic (M3) models", the top 10 nations with the largest rank shifts between M2 and M3 are analyzed. Table 6Analysis of Significant Rank Shifts across Weighting Schemes (Top 10 Impacted).Sustainability TierCountryDynamic Rank (M3)Equal Rank (M2)Rank Shift (ΔR)InterpretationTransitionLebanon76111−35Higher performance in entropy-weighted social dimensions.LaggingMadagascar103135−32Baseline stability remains low regardless of weight allocation.TransitionQatar4779−32De-wealth effect: Ranking improves when GDP weight is balanced.LaggingBenin101130−29Structural weaknesses across all indicators lead to persistent low rank.LaggingBurkina Faso102129−27Minimum variability across scenarios suggests a stable lagging status.LaggingCameroon100127−27Balanced weights (M2) further highlight institutional deficiencies.LaggingLaos99126−27Consistency in ranking across scenarios indicates model robustness.LaggingPakistan98123−25Significant gap between specific strengths and overall average.TransitionNamibia7093−23Middle-tier stability confirmed through weight triangulation.LaggingRwanda90112−22Sensitivity to weights reveals uneven development across indicators.ΔR = M3 Rank - M2 Rank. A negative value suggests that the country’s sustainability profile is better captured by the information-dense DEWM weighting than by a uniform average. Analysis of Significant Rank Shifts across Weighting Schemes (Top 10 Impacted). ΔR = M3 Rank - M2 Rank. A negative value suggests that the country’s sustainability profile is better captured by the information-dense DEWM weighting than by a uniform average. The rank displacements in Table 6 confirm that an equal-weight approach (M2) tends to homogenize performance, failing to account for the unique informational value of different indicators. For instance, nations like Qatar (−32) show notable shifts, suggesting that their sustainability standing is sensitive to the logarithmic compression of economic indicators and the subsequent entropy-based weighting. In M3, the objective weighting mechanism identifies indicators with higher informational entropy (such as Rule of Law and CO2 Intensity in the refined model) and assigns them priority, which recalibrates the positions of these metric-sensitive nations. M2, by contrast, dilutes these differences through uniform weighting, producing a less precise hierarchy. This methodological comparison suggests that the DEWM-based M3 model provides a more detailed and justified evaluation of global sustainability than traditional equal-weighting methods. Stability and consistency analysis: GRA verification (M4) The mathematical robustness of the Dynamic DEWM-TOPSIS (M3) framework was cross-verified using Grey Relational Analysis (M4) as a benchmark. While TOPSIS evaluates the Euclidean distance to ideal solutions, GRA captures the similarity of developmental trajectories, providing a complementary lens on ranking stability. Results show high consistency between the two models, with a Spearman Rank Correlation Coefficient () of 0.9797, with no rank reversal observed among the top 20 nations. This indicates that the leading tier’s classification is consistent across both distance-based (TOPSIS) and similarity-based (GRA) algorithms. As detailed in Table 7, rank displacements remain within a narrowrange even among nations with the most visible fluctuations, supporting the stability of the sustainability hierarchy across methods. Table 7Validation of Ranking Consistency: Dynamic TOPSIS (M3) vs. GRA (M4).Sustainability TierCountryDynamic Rank (M3)GRA Rank (M4)Rank Shift (ΔR)InterpretationLaggingRwanda907119High synchronization in developmental trajectories across indicators.TransitionQatar472918Strong trend consistency despite medium absolute performance.LeadingLuxembourg23518Exceptional synergy among environmental and economic trends.TransitionBrunei Darussalam574314Coordinated progress in specific sustainability dimensions.LeadingAustralia21813High structural stability in longitudinal performance curves.LeadingUnited States322012Dynamic trends exhibit strong resilience across all metrics.TransitionBahrain514011Positive correlation between institutional and social improvements.TransitionSaudi Arabia524210Emerging trend alignment in recent energy efficiency indicators.LeadingUruguay2434−10Slight decoupling between trend direction and absolute score.LeadingMalta818−10Absolute lead remains high despite minor trend fluctuations. Validation of Ranking Consistency: Dynamic TOPSIS (M3) vs. GRA (M4). The broad convergence between M3 and M4 outcomes provides further validation. This high Spearman correlation suggests that the global ranking is a reflection of underlying data structures rather than an artifact of algorithmic bias. As illustrated in Table 7, these displacements are statistically negligible within the context of the 103-country sample. Whether assessed by distance to the ideal (TOPSIS) or relational similarity (GRA), the core evaluative findings remain unchanged. Leading nations such as Luxembourg and Malta consistently maintain their top positions, while the classification of Transition economies like Qatar and Saudi Arabia remains consistent. This stability suggests that the framework provides a reasonable foundation for the policy recommendations discussed in the subsequent chapter. Uncertainty and probabilistic stability: Monte Carlo simulation (M5) To address potential data measurement errors and the inherent uncertainty in indicator performance, a Monte Carlo Simulation (M5) was executed with 1,000 iterations. By introducing 10% random perturbations to the weights of the seven sustainability indicators, the study rigorously evaluated the ranking persistence of the top-performing nations. This probabilistic validation ensures that the M3 results are not sensitive to minor weighting fluctuations and represent a robust sustainability hierarchy. As summarized in Table 8, the simulation confirms an absolute degree of probabilistic stability for the leading nations across all 1,000 scenarios. Table 8Rank Distribution and Stability of Top 5 Nations (n = 1,000).RankSwitzerlandNorwayIrelandLuxembourgAustraliaNO.110000000NO.201000000NO.300001000NO.400100000NO.500010000 Rank Distribution and Stability of Top 5 Nations (n = 1,000). The simulation results, further visualized in the Box-and-Whisker plot (Fig. 2), show the stability of the identified rankings. The scoring distribution shows that even under stochastic variations, each leading nation maintains a narrow closeness score interval. Notably, the distinct gaps between the boxes(representing the interquartile ranges) indicate a clear margin between Switzerland, Sweden, and Norway. Stochastic Sensitivity Analysis of Sustainability Performance (n = 1,000). A key observation is the role of Norway, which consistently secures the NO.3 position in the Monte Carlo simulation. While the primary M3 model (incorporating dynamic temporal weights) ranks Norway 5th to reflect recent developmental trajectories, the 1,000-iteration simulation based on long-term averages reinforces its underlying structural strength. This discrepancy highlights the model’s high sensitivity to reflect recent developmental trajectories, reinforcing its underlying structural strength. Beyond the top 5 nations, the simulation shows consistent global ranking. No rank reversal was observed within the top 20 positions across all 1,000 iterations, with leading nations maintaining high stability probabilities. This consistency under stochastic weight perturbations (10%) suggests that the M3 results are resistant to measurement uncertainties and reflect a stable sustainability hierarchy. In conclusion, the fact that all top-tier nations maintained their respective positions in 100.0% of the iterations proves that the Dynamic DEWM-TOPSIS (M3) framework is statistically resilient to weighting bias. This “stress test” effectively mitigates concerns regarding data uncertainty, ensuring that the sustainability tiers are derived from consistent performance rather than random artifacts, thus providing a credible foundation for the subsequent policy implications. Structural consistency and tier validation: K-means clustering (M6) For the purpose of further validate the objectivity of the sustainability tiers, a K-means Clustering analysis (M6) was performed. While the Dynamic TOPSIS model (M3) provides a linear hierarchy, the K-means algorithm groups the 103 nations based on their multi-dimensional Euclidean proximity across the seven indicators. The convergence between these two independent methods ensures that the classification is statistically robust and not a byproduct of arbitrary thresholds. Cluster centroid characteristics Determining the optimal number of clusters, the Elbow Method was employed by calculating the Within-Cluster Sum of Squares (WCSS) for K ranging from 1 to 6. As illustrated in Fig. 3, the WCSS exhibits a sharp decline from 26.34 at K = 1 to 12.49 at K = 2, followed by a clear inflection point at K = 3 with a WCSS of 8.06. Beyond K = 3, the rate of decrease slows significantly, indicating diminishing returns in explaining data variance. This “elbow” at K = 3 justifies the three-tier classification as the most stable structural representation of global sustainability disparities in this dataset. Elbow Plot of WCSS Against Number of Clusters (k). As shown in Table 9, the K-means algorithm identified three distinct clusters with significant variance in their central coordinates. The spatial distribution across the seven indicators reveals the underlying structural logic of global sustainability. Table 9Multi-dimensional Centroids of Sustainability Clusters (M6).Cluster IDGDPANSEDULIFRULCO2EPICluster 1 (Leading)0.8110.8120.7930.910.8750.4490.852Cluster 2 (Transition)0.4850.8190.6860.6690.4720.6120.648Cluster 3 (Lagging)0.2180.8350.3280.3470.3840.9020.447GDP: GDP per capita; ANS: Adjusted Net Savings; EDU: Education Index; LIF: Life Expectancy; RUL: Rule of Law Index; CO2: CO2 Emissions per capita (inverted); higher scores indicate lower per capita emissions, treated as a favorable outcome following inversion preprocessing as described in the section "Data collection and preprocessing".; EPI: Environmental Performance Index. Centroids derived from the M6 scenario utilizing K-means + + clustering (K = 3). Multi-dimensional Centroids of Sustainability Clusters (M6). GDP: GDP per capita; ANS: Adjusted Net Savings; EDU: Education Index; LIF: Life Expectancy; RUL: Rule of Law Index; CO2: CO2 Emissions per capita (inverted); higher scores indicate lower per capita emissions, treated as a favorable outcome following inversion preprocessing as described in the section "Data collection and preprocessing".; EPI: Environmental Performance Index. Centroids derived from the M6 scenario utilizing K-means + + clustering (K = 3). The multi-dimensional performance gaps are visualized in the radar profile (Fig. 4). Cluster 1 (Leading) shows strong performance driven by high governance scores (0.875) and social well-being (0.910). However, its lower score in Carbon Productivity (0.449) reflects the environmental costs of highly developed economies, highlighting the difficulty of decoupling high-income lifestyles from carbon intensity. Comparative Radar Profile of Sustainability Tiers based on K-means Centroids (M7). In contrast, Cluster 3 (Lagging) exhibits a unique structural characteristic. While it achieves the highest scores in Environmental Efficiency (CO2: 0.902), primarily due to its low level of industrialization and minimal carbon footprint, it remains trapped in the ‘Lagging Tier’ due to severe deficiencies in human capital (Education: 0.328) and institutional strength (Rule of Law: 0.384). Cluster 2 (Transition) serves as the global stabilizer, occupying the moderate middle-ground across all pillars, yet still requiring substantial improvements in social indicators to bridge the gap toward the Leading Tier. Sustainability tier profile and global distribution The spatial distribution of the 103 nations across the three clusters demonstrates an exceptionally high degree of congruence with the M3 dynamic rankings. Specifically, a cross-validation between the two frameworks reveals that among the top-ranked 32 nations in the M3 model, 28 (87.5% congruence) are accurately categorized into Cluster 1 (Leading Tier). Similarly, for the bottom 32 nations, 30 (93.8% congruence) fall into Cluster 3 (Lagging Tier). This substantial overlap confirms that the tiered classification is not an artifact of arbitrary thresholds but reflects a consistent socio-economic reality across both linear and spatial methodologies. The cluster profiles and representative nations are summarized in Table 10. Table 10Global Sustainability Tier Profiles and Distribution (M6) n = 103.Cluster/TierSample Size (n)Profile CharacteristicsRepresentative NationsCluster 1: Leading32 (31.1%)Exceptional “Economy-Environment” coordination; Superior Rule of Law (0.875) and EPI (0.852).Switzerland (1), Norway (5), Spain (12), Japan (17), Canada (25)Cluster 2: Transitioning39 (37.8%)Largest global cohort; Stable institutional base but faces “human capital” and “CO2” bottlenecks.Albania (41), Brazil (44), Thailand (55), China (61), Mexico (65)Cluster 3: Lagging32 (31.1%)Severe resource constraints; Highest environmental burden (CO2 0.902) and acute institutional gaps.Cambodia (94), Pakistan (98), Benin (101), Madagascar (103) Global Sustainability Tier Profiles and Distribution (M6) n = 103. The K-means results offer profound insights into the structural barriers of global sustainability. The Leading Tier (Cluster 1) is characterized not just by wealth, but by a high Rule of Law (0.875) and EPI (0.852), suggesting that institutional stability is the primary engine of high-level sustainability. In contrast, the Lagging Tier (Cluster 3) exhibits a critical lack of human capital (Education 0.328), which may prevent these nations from effectively translating economic growth into sustainable outcomes. The fact that the Transitioning Tier (Cluster 2) constitutes nearly 38% of the sample reflects the middle ground where most global progress is occurring. The high consistency between the linear ranking (M3) and spatial clustering (M6) (with Cluster 2 serving as the Transitioning group) suggests that the proposed framework captures meaningful socio-economic groupings rather than arbitrary score-based divisions. Conclusions and policy implications Main findings This research successfully characterizes the sustainability evolution of 103 nations through a dual-validation framework involving Dynamic TOPSIS and K-means clustering. The core findings are summarized as follows: Tier Stability under Dynamic Weighting: The integration of time-decayed entropy weights into the dynamic evaluation model provides a more precise reflection of national transitions along recent development trajectories compared to traditional static rankings. Results indicate a significant staircase effect in global sustainability structures, with tier distributions showing resilience across GRA and Monte Carlo stability tests. Tier Stability under Dynamic Weighting: The integration of time-decayed entropy weights into the dynamic evaluation model provides a more precise reflection of national transitions along recent development trajectories compared to traditional static rankings. Results indicate a significant staircase effect in global sustainability structures, with tier distributions showing resilience across GRA and Monte Carlo stability tests. Congruence between Ranking and Spatial Distribution: A strong alignment exists between the linear hierarchy of the M3 model and the spatial grouping results of the M6 model. This convergence suggests that the Leading, Transitioning, and Lagging tiers reflect real socio-economic differences rather than algorithmic artifacts. Congruence between Ranking and Spatial Distribution: A strong alignment exists between the linear hierarchy of the M3 model and the spatial grouping results of the M6 model. This convergence suggests that the Leading, Transitioning, and Lagging tiers reflect real socio-economic differences rather than algorithmic artifacts. Quantitative Identification of Structural Bottlenecks: Data analysis identified the main developmental barriers for each tier. While the Leading Tier (Cluster 1) is driven by institutional quality (Rule of Law: 0.875) and environmental performance (EPI: 0.852), the Lagging Tier (Cluster 3) exhibits a unique structural asymmetry. While it achieves the highest scores in environmental efficiency (CO2: 0.902), primarily a byproduct of low industrialization rather than active mitigation, its transition is limited by deficiencies in human capital (Education: 0.328) and institutional strength (Rule of Law: 0.384). Quantitative Identification of Structural Bottlenecks: Data analysis identified the main developmental barriers for each tier. While the Leading Tier (Cluster 1) is driven by institutional quality (Rule of Law: 0.875) and environmental performance (EPI: 0.852), the Lagging Tier (Cluster 3) exhibits a unique structural asymmetry. While it achieves the highest scores in environmental efficiency (CO2: 0.902), primarily a byproduct of low industrialization rather than active mitigation, its transition is limited by deficiencies in human capital (Education: 0.328) and institutional strength (Rule of Law: 0.384). Theoretical and policy implications The empirical findings of this study carry several theoretical implications for the sustainability evaluation literature. The identification of a persistent staircase effect across 103 nations over an 11-year period provides empirical support for development trap theory, suggesting that structural inequalities between tiers are self-reinforcing rather than transitory11. The Lagging Tier’s paradox of high environmental scores driven by low industrialization rather than active policy intervention offers new empirical evidence for the environmental Kuznets curve debate, demonstrating that favorable environmental indicators in low-income economies should not be conflated with genuine ecological progress6,29. Furthermore, the consistent role of Rule of Law as the primary differentiator between the Leading and Lagging Tiers reinforces the theoretical centrality of SDG 16 as a cross-cutting enabler of sustainable development, extending prior arguments on institutional prerequisites for international sustainability transitions10,12. Based on the refined centroid characteristics, this study proposes the following targeted governance strategies to bypass developmental bottlenecks and optimize global sustainability pathways: Leading Tier (Cluster 1): Deepening Green Finance and Institutional Export Nations in this tier should leverage their institutional capital to further climate resilience. Beyond maintaining superior environmental efficiency (0.852), policy focus should shift toward pioneering next-generation carbon-neutral technologies and exporting successful governance models through multilateral frameworks to assist lower-tier transitions, thereby fostering global institutional convergence11,12. Leading Tier (Cluster 1): Deepening Green Finance and Institutional Export Nations in this tier should leverage their institutional capital to further climate resilience. Beyond maintaining superior environmental efficiency (0.852), policy focus should shift toward pioneering next-generation carbon-neutral technologies and exporting successful governance models through multilateral frameworks to assist lower-tier transitions, thereby fostering global institutional convergence11,12. Transitioning Tier (Cluster 2): Structural Optimization and Bridging Social Gaps As the largest global cohort (37.8%), these nations are in a critical trade-off phase between economic growth and environmental degradation. Priorities should focus on upgrading human capital to decouple growth from carbon intensity6,15. Transitioning Tier (Cluster 2): Structural Optimization and Bridging Social Gaps As the largest global cohort (37.8%), these nations are in a critical trade-off phase between economic growth and environmental degradation. Priorities should focus on upgrading human capital to decouple growth from carbon intensity6,15. Lagging Tier (Cluster 3): Reconstructing Educational Foundations and Legal Frameworks Economic aid alone is insufficient to address systemic stagnation in this tier. Policies must prioritize basic education (0.328) and the Rule of Law (0.384) to create a ‘receptive’ environment for international sustainability funds (Tutar, 2026;16). Lagging Tier (Cluster 3): Reconstructing Educational Foundations and Legal Frameworks Economic aid alone is insufficient to address systemic stagnation in this tier. Policies must prioritize basic education (0.328) and the Rule of Law (0.384) to create a ‘receptive’ environment for international sustainability funds (Tutar, 2026;16). Research limitations and future directions Despite the methodological advancements, several avenues for further refinement remain: Dynamic Expansion of Data Dimensions: While the current model relies on seven core indicators, future research could incorporate parameters with higher temporal sensitivity, such as digital transformation indices or climate adaptation resilience scores. Dynamic Expansion of Data Dimensions: While the current model relies on seven core indicators, future research could incorporate parameters with higher temporal sensitivity, such as digital transformation indices or climate adaptation resilience scores. Micro-level Case Analysis: Subsequent studies could perform deep-dives into “outlier” nations within the Transitioning Tier that exhibit exceptional upward mobility, thereby uncovering specific policy contributions to dynamic trajectory improvements. Micro-level Case Analysis: Subsequent studies could perform deep-dives into “outlier” nations within the Transitioning Tier that exhibit exceptional upward mobility, thereby uncovering specific policy contributions to dynamic trajectory improvements. Data Constraints and Sample Selection: With the intent of maintaining a balanced panel for dynamic weighting (M3), nations with inconsistent or missing longitudinal data were excluded. While this ensures statistical rigor and eliminates imputation bias, future studies could explore hybrid models to incorporate nations with sparse data in critical regions like Sub-Saharan Africa. Data Constraints and Sample Selection: With the intent of maintaining a balanced panel for dynamic weighting (M3), nations with inconsistent or missing longitudinal data were excluded. While this ensures statistical rigor and eliminates imputation bias, future studies could explore hybrid models to incorporate nations with sparse data in critical regions like Sub-Saharan Africa. Author contributions Y.- Y. H., X. W., L. J., T.- S. L. conceptualized the analysis and monitored data preparation. T.- S. L., R.-C. T. performed MCDM analysis of the data, drafted the manuscript, and prepared the figures. T.- S. L., and X. W. analyzed the software. Y.- Y. H., T.- S. L., and R.-C. T. revised the manuscript and approved final submission. Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request. Declarations Competing interests The authors declare no competing interests. AI use statement AI-assisted tools were used solely for language editing (grammar and phrasing) in the preparation of this manuscript. All intellectual content, research design, data analysis, and conclusions are the original work of the authors. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.