Determinants of Subjective Well-Being among Chinese Older Adults: A Machine Learning Approach with SHAP Explainability

DOI: https://doi.org/10.62517/jse.202511305

Author(s)

Jia He*

Affiliation(s)

School of Mathematics and Statistics, The Center Applied Mathematics of Guangxi, Guangxi Normal University, Guangxi, Guilin, China *Corresponding Author

Abstract

China's rapidly aging population necessitates a deeper understanding of the determinants of subjective well-being (SWB) among older adults. This study employs a machine learning (ML) approach with SHAP (SHapley Additive exPlanations) explainability to identify key drivers of SWB using nationally representative data from the China Health and Retirement Longitudinal Study (CHARLS 2020, *n* = community-dwelling adults aged 60+). the CatBoost algorithm achieved superior predictive performance (AUC = 0.80) compared to traditional methods (e. g., logistic regression AUC = 0.67), highlighting ML's capacity to model complex, nonlinear relationships. SHAP analysis revealed self-rated health, loneliness, and frequency of contact with children as the top three predictors of SWB. Crucially, self-rated health exhibited a nonlinear threshold effect: ratings of "Very Good" or "Good" (scores ≤2) significantly enhanced SWB, while "Average" or worse (scores ≥3) diminished it. Depression scores showed a complex U-shaped association with SWB (scores 12–23 associated with positive effects). Socioeconomic factors (e. g., utility expenditure, insurance type) were influential but secondary to health and social connections. These findings underscore the multidimensional nature of SWB and suggest interventions targeting health perception, loneliness reduction, and family support are critical for enhancing well-being in China's aging population.

Keywords

Subjective Well-Being; Aging; Machine Learning; SHAP Explainability; Health Perception; Loneliness

References

[1] Zhao, Yaohui, Yisong Hu, James P Smith, John Strauss, Gonghuan Yang. (2014). Cohort Profile: The China Health and Retirement Longitudinal Study (CHARLS), International Journal of Epidemiology, 43(1):61–68. [2] Bowling, A., & Iliffe, S. (2019). Aging and society: Historical and contemporary perspectives. Oxford University Press. [3] Cai, F., et al. (2021). Population aging and economic growth in China. China Economic Journal, 14(1), 1–16. [4] Chen, X., et al. (2022). Machine learning applications in depression: A systematic review and meta-analysis. Journal of Affective Disorders, 300, 50–62. [5] Diener, E. (1984). Subjective well-being. Psychological Bulletin, 95(3), 542–575. [6] Hybels, C. F., et al. (2014). the health–well-being paradox in older adults: A longitudinal study. the Gerontologist, 54(3), 435–443. [7] Kohler, T., et al. (2017). Is depression always bad for well-being? A nonlinear analysis across 21 countries. Social Science & Medicine, 187, 138–146. [8] Lundberg, S. M., & Lee, S. I. (2017). A unified approach to interpreting model predictions. In Advances in neural information processing systems (pp. 4765–4774). [9] National Bureau of Statistics of China. (2024). China Statistical Yearbook 2024. China Statistics Press. [10] Ng, J., et al. (2019). Socioeconomic status and subjective well-being in later life: A cross-national analysis. the Journals of Gerontology: Series B, 74(3), 540–550. [11] Prokhorenkova, L., et al. (2018). CatBoost: unbiased boosting with categorical features. In Proceedings of the 32nd International Conference on Neural Information Processing Systems (pp. 6635–6645). [12] Williams, J. (2018). Handling missing data in clinical research. Journal of Thoracic Disease, 10(11), E98–E107. [13] World Health Organization. (2023). World report on aging and health. WHO Press. [14] Yuan, Y., et al. (2020). Machine learning-based prediction of depression: A large-scale cross-sectional study. Journal of Affective Disorders, 262, 283–290. [15] Zhang, J., & Hayflick, S. (2020). Cultural perspectives on aging and well-being. International Review of Gerontology, 34, 1–12. [16] Sun YH, Liu QJ, Lee NY, et al. A novel machine learning approach to shorten depression risk assessment for convenient uses [J]. Journal of Affective Disorders, 2022, 312:275-291. [17] Aleem S, Huda NU, Amin R, et al. Machine learning algorithms for depression: diagnosis, insights, and research directions [J]. Electronics, 2022, 11(7):1111. [18] LundbergS, LeeSI. Aunified approach to interpretingmodel predictions [M]. Neural Information Processing Systems, 2017, 4:4768-4777. [19] Liu Y. Prediction of depression in the elderly based on machine learning [D]. Jinan: Shandong University, 2023. [20] Monroe SM, Harkness KL. Major depression and its recurrences: Life course matters [J]. Annual Review of Clinical Psychology, 2022, 18:329-357. [21] Zhang CY, Chen XF, Wang S, et al. Using CatBoost algorithm toidentify middle-aged and elderly depression, National health and nutrition examination survey 2011-2018 [J]. Psychiatry Research, 2021, 306:114261. [22] Lundberg, S. M., & Lee, S. I. (2017). A Unified Approach to Interpreting Model Predictions. Advances in Neural Information Processing Systems, 30, 4765-4774. [23] Prokhorenkova, L., et al. (2018). CatBoost: Unbiased Boosting with Categorical Features. Advances in Neural Information Processing Systems, 31, 6638-6648. [24] National Bureau of Statistics of China. (2023). China Statistical Yearbook 2023. China Statistics Press.