Machine learning in the prediction of human wellbeing

Area of Science:

• Social Sciences
• Computational Social Science
• Psychology

Background:

• Subjective wellbeing data is increasingly utilized in social sciences.
• Existing modeling approaches for wellbeing data have limited predictive power.

Purpose of the Study:

• To enhance the understanding of self-reported wellbeing using Machine Learning (ML).
• To identify key drivers of evaluative wellbeing.
• To examine functional forms of wellbeing predictors, including satiation and U-shaped effects.

Main Methods:

• Application of tree-based Machine Learning algorithms.
• Analysis of large-scale, representative survey data from Germany, UK, and US (2010-2018).
• Comparison of ML predictive performance against standard modeling approaches.

Main Results:

• ML algorithms demonstrate superior predictive performance for wellbeing scores compared to traditional methods.
• Key drivers of evaluative wellbeing identified by ML align with existing literature.
• ML analysis reveals nuances in predictor relationships, such as income satiation and age-wellbeing U-shaped curves.

Conclusions:

• Machine Learning offers enhanced predictive capabilities for subjective wellbeing research.
• ML provides valuable insights into the primary determinants and complex functional forms influencing wellbeing.
• This study establishes an upper bound for wellbeing predictability using survey data.