Double Scaffold of Retirement on Subsequent Health: Revisiting Spatio-Temporality and Finance Through the Feature-Wide Effect Heterogeneity

This study investigates how retirement's effect on subsequent health varies by individual attributes and their interaction. Instead of spotlighting heterogeneity for a specific, theoretically predetermined feature, I consider the feature-wide effect heterogeneity to find possible data-driven candidates that generate this heterogeneity. To accomplish this, I use causal forests and harmonized data from the Gateway to Global Aging Data (the Gateway). The results show that temporal dimensions (such as age and birth cohort) and financial dimensions (such as income and wealth) are the most important features in heterogenizing retirement's effect on health. Based on these findings, I suggest the "double scaffold" model to comprehensively explain feature-wide effect heterogeneity.