Bayesian local clustering of age-period mortality surfaces across multiple countries

2025-04-07

AI Summary

Existing age-period mortality models struggle to capture localized, dynamic clustering patterns across countries in the age-period plane, limiting interpretability in cross-national mortality forecasting. This paper proposes a Bayesian joint modeling framework that introduces, for the first time, a time-varying random partition prior—enabling locally adaptive country clustering over age and period. Integrated with B-spline–based age effects, country-specific dynamic coefficients, and Gibbs sampling for posterior inference, the model jointly fits mortality surfaces for 14 countries. It successfully identifies well-established demographic phenomena (e.g., declining young-adult mortality in Nordic countries) and emerging trends (e.g., rising midlife female mortality in specific regions), markedly improving both predictive accuracy and policy interpretability. The framework provides a novel tool for regional demographic trend analysis and targeted public health interventions.

Technology Category

• Reasoning under Uncertainty: Relational Probabilistic Models
• Machine Learning: Bayesian Learning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Models for Web evolution

Abstract

Although traditional literature on mortality modeling has focused on single countries in isolation, recent contributions have progressively moved toward joint models for multiple countries. Besides favoring borrowing of information to improve age-period forecasts, this perspective has also potentials to infer local similarities among countries' mortality patterns in specific age classes and periods that could unveil unexplored demographic trends, while guiding the design of targeted policies. Advancements along this latter relevant direction are currently undermined by the lack of a multi-country model capable of incorporating the core structures of age-period mortality surfaces together with clustering patterns among countries that are not global, but rather vary locally across different combinations of ages and periods. We cover this gap by developing a novel Bayesian model for log-mortality rates that characterizes the age structure of mortality through a B-spline expansion whose country-specific dynamic coefficients encode both changes of this age structure across periods and also local clustering patterns among countries under a time-dependent random partition prior for these country-specific dynamic coefficients. While flexible, this formulation admits tractable posterior inference leveraging a suitably-designed Gibbs-sampler. The application to mortality data from 14 countries unveils local similarities highlighting both previously-recognized demographic phenomena and also yet-unexplored trends.

Problem

• Model local clustering of mortality patterns across multiple countries
• Incorporate age-period structures with varying country similarities
• Improve mortality forecasts by identifying demographic trends and policies

Innovation

• Bayesian model for multi-country mortality clustering
• B-spline expansion for age structure modeling
• Time-dependent random partition prior