Welfare Regimes in Asia: Convergent or Divergent?

The study addresses a major gap in welfare state research, which has largely centered on advanced Western economies, by analyzing how welfare regimes have evolved across 20 Asian countries since the 2000s. The research question asks whether Asian welfare regimes are diverging (as many comparative political economy theories suggest) or converging toward commodification under globalization (as international political economy predicts), or whether they exhibit another pattern. The context is the Asia-Pacific’s integration into global markets, exposure to external shocks, and rising within-country inequality despite reductions in global poverty. The purpose is to classify welfare regime types and track their changes from the 2000s to the 2010s using data-driven methods, thereby illuminating how these countries balance commodification (market-oriented capacity building) and decommodification (social insurance and assistance). The study’s importance lies in challenging established theoretical expectations and providing a more nuanced account of welfare state trajectories in a region with diverse economic, political, and cultural conditions.

The paper situates its inquiry within comparative political economy (CPE) and international political economy (IPE) literatures. CPE traditions—structuralist, rationalist, and culturalist—have emphasized decommodification as a defining attribute of Western welfare regimes (Esping-Andersen, 1990), but this framework has limited geographic scope and historical specificity. For late industrializers and partially commodified economies, Rudra (2007) highlights how pre-existing institutional arrangements shape the balance of commodification and decommodification. Early CPE work correlates economic development with social spending and considers political regime type and partisan dynamics as key determinants of welfare expansions. Authoritarian regimes in Asia may also expand welfare for development and stability, complicating democracy-centric narratives. IPE arguments stress that globalization and capital mobility push countries toward neoliberal, commodification-centric policies, tax competition, and reduced social spending, implying institutional convergence. Yet Asian contexts feature heterogeneous development levels, export/FDI exposure, regime types, and rising internal inequality, suggesting pressures both toward commodification and toward protective decommodification responses to shocks and distributional tensions. The paper advances a dynamic equilibrium perspective that integrates CPE and IPE forces, anticipating a balanced mix rather than pure divergence or convergence.

The study uses a four-step data-driven approach for descriptive inference: (1) dimension reduction via Principal Component Analysis (PCA); (2) clustering using K-means++ as the primary algorithm, with robustness checks using Hierarchical Clustering (Ward linkage) and DBSCAN; (3) determining the optimal number of clusters via Silhouette Scores; and (4) classifying countries and comparing clusters across decades (2000–2009 vs. 2010–2019). Variables are organized into commodification and decommodification constructs. Commodification includes public expenditures and outcomes related to health and education, literacy, life expectancy, immunization, and a gender equality indicator (sources: World Bank WDI; UNDP HDI). Decommodification includes social insurance (pensions, health insurance elements, unemployment insurance) and social assistance (cash/in-kind benefits) from the ADB Social Protection Indicator; general health expenditures are excluded from decommodification to avoid redundancy, following Rudra (2007). The number of ILO conventions ratified is also included. PCA is applied separately to the commodification and decommodification variable sets to extract principal components (PC1 primarily) that capture most variance and reduce noise due to high inter-variable correlations. K-means++ clustering is run on the PCA-derived components, and the optimal K is chosen using Silhouette Scores; hierarchical and DBSCAN results are used to assess robustness and cluster density/consolidation across decades.

• PCA captured most variance in single components: PC1 explained 77.7% (commodification, 2000s), 91.4% (decommodification, 2000s), 81.1% (commodification, 2010s), and 93.0% (decommodification, 2010s), supporting dimensionality reduction before clustering.
• Optimal clustering by Silhouette Score: 2000s: K=5 (score 0.693) outperformed K=2 (0.677), K=3 (0.661), K=4 (0.651), K=6 (0.569). 2010s: K=3 (score 0.663) outperformed K=2 (0.567), K=4 (0.627), K=5 (0.597), K=6 (0.548). This indicates consolidation from five clusters in the 2000s to three clusters in the 2010s.
• Cluster characteristics: In the 2000s, one cluster nearest to protective/decommodifying features included all Northeast Asian countries alongside several Southeast and South Asian countries; another cluster closest to commodification features included Afghanistan and India. In the 2010s, three clusters emerged, with one cluster closest to commodification (including Afghanistan and Pakistan). Across both decades, clusters cut across regime types (authoritarian, hybrid, and democratic states co-located), indicating political regime was not the primary basis of grouping.
• Robustness checks: Hierarchical clustering placed 2010s clusters at lower hierarchical levels than in the 2000s, and DBSCAN produced fewer clusters in the 2010s at comparable epsilon values; at epsilon=15 in the 2010s, all observations formed a single dense cluster, a pattern not observed in the 2000s. These results corroborate increased interconnection and consolidation between countries’ welfare features over time.
• No convergence toward pure commodification or divergence to highly heterogeneous forms: Despite global market pressures, countries did not collectively shift away from decommodification; nor did they display systematic divergence. Instead, they maintained a balanced mix of commodification and decommodification.
• Interpretation: Shared external shocks, productivity competition, and rising internal inequality likely drove policy mixes that balance capability-building (commodification) with protective measures (decommodification), regardless of regime type or income level.

The findings directly address the research question by showing that Asian welfare regimes neither follow CPE’s expectation of persistent, highly heterogeneous divergence nor IPE’s expectation of convergence toward commodification. Instead, countries converge into fewer clusters that integrate both commodification and decommodification, suggesting a dynamic equilibrium shaped by joint external (globalization, capital mobility, crises) and internal (rising inequality, social demands) pressures. This balance supports policy flexibility: commodification investments bolster competitiveness, while decommodification mitigates vulnerabilities from shocks and distributional tensions. The cross-regime clustering indicates that institutional form (democracy vs. authoritarianism) is not determinative of welfare configuration in this context; rather, similar economic and social challenges yield comparable policy mixes. These results nuance theories of welfare regime development and varieties of capitalism by highlighting an adaptive equilibrium rather than linear convergence or divergence.

The study extends welfare regime analysis to 20 Asian countries and, using PCA and multiple clustering algorithms, documents a reduction in clusters from five (2000s) to three (2010s) without a shift toward purely commodification-oriented models or highly heterogeneous divergence. It contributes a dynamic equilibrium perspective, wherein external market pressures and internal inequality jointly produce balanced welfare mixes that cut across political regime types. This challenges rigid categorizations and enriches Varieties of Capitalism discussions by emphasizing adaptive, context-sensitive welfare configurations. Future research should expand longitudinal and sectoral data coverage, refine clustering strategies, and probe causal mechanisms behind country shifts and outliers to better link market integration, inequality, and welfare policy dynamics in the region.

• Data constraints: Decommodification measures rely on ADB Social Protection Indicator; commodification proxies draw from multiple international sources (World Bank, ADB, UNDP). Broader, more granular, and longer time-series data would enable stronger inference.
• Methodological choices: Results depend on PCA variable selection, clustering algorithms, and the determination of optimal K. Although robustness checks (Ward hierarchical, DBSCAN) were used, alternative specifications could yield different partitions.
• Assumptions: The presumption of a positive inequality–welfare expansion link is debated; some studies find contrary relationships.
• Explanatory gaps and outliers: The study does not fully explain country-specific shifts; Japan’s stronger decommodification tilt and Afghanistan’s distance from equilibrium are noted as outliers driven by distinctive development and conflict/integration contexts.
• Measurement: Clustering reflects relative commitment (shares of GDP) rather than absolute welfare levels, so co-clustering does not imply equal service provision.