Ageing, health and predicting future employment exits: a penalised regression approach

We examine the role of baseline health in predicting future employment exits, alongside established socioeconomic, job-related and demographic predictors. Using UKHLS, we track employed respondents over 10 years to assess subsequent employment exits. Baseline health is captured using an unusually rich set of measures: self-assessed health (SAH), self-reported diagnosed conditions, psychological distress, allostatic load (composite biomarker index), and epigenetic biological age. Applying a LASSO penalised regression approach, we find that epigenetic biological age and SAH, rather than self-reported conditions, psychological distress, or allostatic load, predict subsequent employment exits, independent of other predictors. A Shapley-Shorrocks decomposition highlights epigenetic biological age as a stronger predictor than SAH. Nevertheless, chronological age is the dominant predictor of future employment exits. Epigenetic biological age measures do allow us to disentangle the role of chronological age, mainly reflecting institutional structures such as retirement eligibility and societal norms, from other contributions that capture age-related health decline that are more directly reflected in epigenetic biological age measures.