This article presents a new approach to subjective well-being research using machine learning methods by Associate Professor Naoto Yokoyama of the National Graduate Institute for Policy Studies.
As efforts to measure people's well-being that cannot be captured by GDP indicators and utilize it in policy expand globally, this research conducts empirical analysis using Random Forest (RF), one of the machine learning methods, regarding fields and factors related to subjective well-being in Japan. RF has been proven capable of flexibly capturing well-being related factors from multifaceted perspectives such as "breadth," "impact," and "non-linearity" compared to conventional statistical methods, demonstrating its potential as an effective tool for understanding Japan's well-being characteristics.
As a particularly noteworthy analysis result, detailed examination of factors contributing to increased life satisfaction in the mid-60s revealed that changes in work-life balance have the greatest impact. This finding suggests that adjustment of employment patterns and working hours has important policy implications for improving quality of life in later years. The research sets Japanese and English keywords including well-being, life satisfaction, happiness, Beyond GDP, machine learning, random forest, life satisfaction, happiness, machine learning, and random forests, with a structure conscious of connections to international research trends.
This research was published as "GRIPS Discussion Papers Report No. 25-7" by the GRIPS Policy Research Center in July 2025 and assigned a DOI (10.24545/0002000200), ensuring visibility in international academic information distribution.
The article concludes that machine learning methods demonstrate the possibility of establishing analysis tools that complement conventional methods in well-being research and provide deeper and more practical policy implications, presenting a new methodological foundation for happiness research in Japan.