When Health Shocks Hit Home

Abstract

Older European adults experiencing acute adverse health shocks — heart attack, stroke, hip fracture — diverge sharply in their post-shock financial trajectories within twelve months. Some recover to baseline vulnerability levels; others remain in elevated-hardship states for years. Country fixed-effects and household-income controls reduce but do not eliminate the cross-respondent dispersion, and the residual heterogeneity correlates with measurable differences in **social engagement** — participation in volunteer organisations, community activities, and informal community networks.

The health-economics literature has long suspected that social capital operates as a buffer against the financial consequences of health shocks; what has not been clearly identified is the magnitude of this buffer across the heterogeneous European welfare regimes, where the same health shock can translate into very different financial outcomes depending on national income-support generosity. The strategic hook is that shocks should not have equal financial consequences across countries — and the welfare-regime moderation is the testable cross-national claim.

The empirical setting is SHARE restricted to adults aged 50–80 with the acute-health-shock module fielded. Acute health shocks include first-incident heart attack, stroke, and hip fracture, with shock timing identified by SHARE's biennial wave structure. The existing analysis applies the classical identification stack: difference-in-differences around acute-health-shock onset, panel fixed-effects absorbing time-invariant unobservables, Cox proportional-hazards modelling for the post-shock vulnerability spell, and Oster endogeneity bounds for omitted-variable robustness. The existing evidence already documents sustained increases in financial vulnerability after adverse health shocks.

We propose extending the existing analysis with three ML layers: a Causal Forest by welfare-regime that would recover the conditional shock-effect across the Esping-Andersen typology (replacing the country FE constant-effect with a regime-specific CATE distribution); a Random Survival Forest that would estimate the financial-vulnerability spell duration without the proportional-hazards constraint; and a planned network-analysis layer on the SHARE social-engagement battery that would characterise the underlying participation structure and test whether engagement-network centrality (rather than engagement frequency) is the buffer mechanism. The proposed extensions, once implemented, would supply a cross-national policy-learning rule for countries in the Mediterranean / Southern regime where formal income support is thinner.

Data & Methods