Regional and spatial dependence of poverty factors in Thailand, and its use into Bayesian hierarchical regression analysis

Poverty is a serious issue that harms humanity progression. The simplest solution is to use one-shirt-size policy to alleviate it. Nevertheless, each region has its unique issues, which require a unique solution to solve them. In the aspect of spatial analysis, neighbor regions can provide useful information to analyze issues of a given region. In this work, we proposed inferred boundaries of regions of Thailand that can explain better the poverty dynamics, instead of the usual government administrative regions. The proposed regions maximize a trade-off between poverty-related features and geographical coherence. We use a spatial analysis together with Moran’s cluster algorithms and Bayesian hierarchical regression models, with the potential of assist the implementation of the right policy to alleviate the poverty phenomenon. We found that all variables considered show a positive spatial autocorrelation. The results of analysis illustrate that 1) Northern, Northeastern Thailand, and in less extend Northcentral Thailand are the regions that require more attention in the aspect of poverty issues, 2) Northcentral, Northeastern, Northern and Southern Thailand present dramatically low levels of education, income and amount of savings contrasted with large cities such as Bangkok-Pattaya and Central Thailand, and 3) Bangkok-Pattaya is the only region whose average years of education is above 12 years, which corresponds (approx.) with a complete senior high school.

💡 Research Summary

This paper tackles the persistent problem of poverty in Thailand by moving beyond the conventional use of administrative regions for policy design. The authors propose a data‑driven regionalization that simultaneously maximizes similarity in poverty‑related indicators and geographic coherence, thereby generating “inferred boundaries” that better reflect the underlying socioeconomic dynamics.

The study uses a massive household‑level dataset collected in 2022, comprising 12,983,145 observations across all Thai provinces. Key variables include monthly average household income, average years of education, and household savings. First, the authors assess spatial dependence for each variable using Global Moran’s I. All indicators display significant positive spatial autocorrelation, indicating that provinces with high (or low) values tend to cluster geographically. Spatial weights are constructed via a 5‑nearest‑neighbor (k‑nn) scheme, and the significance of Moran’s I is evaluated through 1,000 permutation tests, providing robust evidence of spatial structure.

To identify meaningful spatial clusters, the authors initially apply the Fisher‑Jenks algorithm, which partitions provinces based purely on attribute values. However, this method often yields geographically fragmented groups. To enforce spatial contiguity, they introduce a second step: an agglomerative hierarchical clustering with Ward linkage that incorporates the k‑nn weight matrix as a spatial constraint. The resulting clusters are evaluated on two fronts: (1) geographic coherence, measured by the isoperimetric quotient (higher values indicate shapes closer to a circle, i.e., less “wiggly” boundaries); and (2) attribute coherence, assessed using silhouette scores and the Calinski‑Harabasz index. The final solution consists of four to five spatially continuous regions that differ markedly from the official administrative divisions. Notably, the northern, northeastern, and to a lesser extent north‑central provinces form a low‑income, low‑education cluster, whereas Bangkok‑Pattaya and the central/eastern provinces constitute a high‑income, high‑education cluster.

With these inferred regions in hand, the authors fit a Bayesian hierarchical regression model to explore the relationship between income and education at both the provincial (level‑1) and regional (level‑2) levels. The model specifies household income as the dependent variable and average years of education as the primary predictor, while allowing region‑specific intercepts and slopes to capture local heterogeneity. Non‑informative normal priors are placed on fixed effects, and inverse‑gamma priors on variance components. Posterior inference is obtained via Markov Chain Monte Carlo sampling. Results reveal a robust positive association: each additional year of education corresponds to an approximate 5–7 % increase in average household income. The regional random effects confirm that the northern and northeastern clusters suffer from both lower education (average ≈10.5 years) and lower income, marking them as priority areas for poverty‑reduction interventions. Conversely, Bangkok‑Pattaya exhibits average education above 12 years and the highest income levels, underscoring a stark regional disparity.

The paper’s contributions are threefold: (1) it demonstrates a rigorous workflow for detecting and quantifying spatial autocorrelation in poverty‑related data; (2) it introduces a spatially constrained clustering approach that yields geographically coherent, policy‑relevant regions; and (3) it integrates these regions into a Bayesian hierarchical framework, allowing simultaneous estimation of national‑level relationships and regional deviations.

Nevertheless, several limitations merit attention. The choice of k = 5 for the nearest‑neighbor weight matrix is somewhat arbitrary; a sensitivity analysis across different k values would strengthen the robustness claim. The variable set is limited to income, education, and savings, omitting health, housing, and access‑to‑services dimensions that are central to multidimensional poverty indices (e.g., MPI). The analysis is cross‑sectional, precluding assessment of temporal dynamics or policy impact over time. Model validation is confined to posterior diagnostics; external validation (e.g., out‑of‑sample prediction) is absent. Finally, while the inferred regions are analytically sound, the paper does not discuss how these new boundaries align with existing administrative, fiscal, or political structures, which could affect the feasibility of implementing region‑specific policies.

In summary, the study offers a compelling methodological blueprint for spatially informed poverty analysis and provides actionable insights for Thai policymakers: prioritize the northern and northeastern clusters for education‑focused interventions, and leverage the high‑education, high‑income characteristics of Bangkok‑Pattaya as a potential growth engine. With further refinement—particularly broader variable inclusion, temporal extension, and alignment with governance frameworks—the approach could be adapted to other developing nations seeking to design more nuanced, region‑tailored poverty alleviation strategies.