Hybrid seminar: Predictability and (co-)incidence of labor and health shocks

This paper employs machine learning techniques to estimate the ex-ante probabilities that individuals will be confronted with adverse labor and health events. Using rich administrative data on the entire Dutch population, we document that shock incidence is predictable especially in the labor domain and to a lesser degree in the domain of mental and physical health. Moving from ex-post incidence to ex-ante probabilities allows for an elaborate study of the risks that individuals face. We bring together risk estimates for many different shocks and uncover that risk concurrence is sizable, monotone and extending across domains. We show that various socioeconomic characteristics are over-represented in the upper tails of the estimated risk distributions. Lastly, we discuss the implications of our findings for targeted prevention policies based on individual risk estimates, shock incidence and person characteristics.