Examining 81 Predictors of Self-Esteem Using Machine Learning

The purpose of this study was to identify and rank the most important predictors of self-esteem. Data were drawn from the Midlife in the United States (MIDUS) study, a nationally representative survey of American adults. A total of 81 potential predictors, including psychological, sociodemographic, and health-related variables, were included. The Random Forest machine learning algorithm was used for data analysis. Environmental mastery emerged as the strongest predictor, followed by negative affect, sense of personal growth and positive affect. Agency-related and affective variables ranked among the top predictors, surpassing socio-demographic, health-related, relational and status-related factors. These findings are inconsistent with some theoretical frameworks that emphasise social validation and status as primary drivers of self-esteem, suggesting that self-esteem is more strongly linked to personal agency, a subjective sense of growth and affective experiences. The results contribute to ongoing theoretical development and offer direction for future theorising and empirical research on the nature and predictors of self-esteem.