Using Health and Retirement Study (HRS) panel data (1992–2012, ages 50–62), Poterba, Venti, and Wise decompose education's strong negative correlation with disability insurance (DI) participation into a direct effect and indirect effects operating through four "pathways": health, wealth, occupation, and employment status. Health is the dominant pathway for women (accounting for 74.2% of the education-DI gap), and wealth is second (23.8%); the four pathways together explain virtually all (99%) of the education-DI gap for women and 57% for men. A time-series extension using National Health Interview Survey (NHIS), Current Population Survey (CPS), and Survey of Consumer Finances (SCF) data back to 1972 shows that the education-health gap has widened substantially, but rising average educational attainment in the population has offset the rising DI rates among the less-educated. The relationships are explicitly presented as correlational, not causal.
"The long reach of education extends to DI participation through multiple pathways — health, wealth, occupation, and employment — with health dominating for women."
"The education gradient in health has widened substantially since the early 1970s. But at the same time, the educational composition of the workforce has shifted substantially toward more education."
The paper's main methodological contribution is the pathway decomposition: systematically absorbing each mediator in sequence to apportion the raw education-DI correlation. The approach is explicitly descriptive — the authors are careful not to claim causal identification of each pathway, which is appropriate given the selection problems at each stage. The key substantive finding is the dominance of the health pathway, especially for women, which suggests that education-based DI disparities largely reflect pre-retirement health accumulation rather than financial or occupational channels alone. The men's residual (43% unexplained after all four pathways) is striking and merits explanation — it likely reflects unmeasured health heterogeneity, selection into education itself, or pathways not captured by HRS measures. The time-series analysis adds a useful population-level perspective: rising attainment is a secular force that moderates the DI growth pressure from widening education gradients.