Summary
Poterba, Venti, and Wise use twelve waves of Health and Retirement Study (HRS) panel data (1992–2008) to decompose the education-wealth gap at retirement into five pathways: health levels at retirement, portfolio returns, Social Security (SS) income, defined benefit (DB) pension income, and a residual "additional education" effect. The paper's central finding is that education sets health levels pre-retirement (a 21.5 percentage point (ppt) gap between college graduates and high school (HS) dropouts), but post-retirement health trajectories are essentially education-independent. A large portfolio return gradient — college graduates earning 14.4% vs. 6.6% two-year average returns for the lowest education group — is the single most powerful individual pathway.
Key Claims
- Health levels vs. trajectories: Education produces a large gap in health status at retirement entry (21.5 ppt between college and <HS), but post-retirement health trajectories are education-independent conditional on entry health. This implies health selection into retirement age, not health divergence during retirement.
- Portfolio return gradient: The two-year average portfolio return ranges from 6.6% (<HS) to 14.4% (college+), a 2.7× differential. This return gap is the single most powerful identified pathway to the education-wealth gap.
- Social Security income is the most protective asset: A $10,000/year increase in Social Security income is associated with $7,000–$27,000 more in non-annuity assets (by asset quintile), substantially exceeding the protective effect of DB pensions. SS income substitutes for precautionary drawdown.
- Education-wealth gap magnitudes: Total gap ranges from ~$82,000 (Q1 assets) to ~$605,000 (Q5 assets). The identified pathways (health, returns, SS, DB) explain ~40% of the gap in the upper quintiles but only ~40% in Q1, leaving 60% unexplained at lower wealth levels.
- Subjective mortality null result: Subjective survival probability has no statistically significant effect on asset drawdown trajectories — a puzzle for the lifecycle model, which predicts that individuals expecting shorter lives should draw down faster.
- No causal identification: Education is treated as predetermined; no instrumental variable for education is employed. All estimates are descriptive associations conditional on the pathway variables.
Concepts Introduced or Extended
- Income-Mortality Gradient — education-wealth gradient at and through retirement; portfolio return mechanism; SS income as asset-protective
- SES Health Behavior Gradient — health level vs. health trajectory distinction; education sets health at retirement, not post-retirement trajectory
Entities Mentioned
Quotes
"We find that the main channel through which education affects the evolution of assets in the post-retirement period is through the level of health when retired, not through the evolution of health after retirement."
"Households with more Social Security income hold substantially more assets, suggesting that Social Security income reduces the need to draw down retirement assets."
My Take
The paper's clean decomposition of five pathways is useful, but the no-causal-identification caveat is large. Portfolio returns and Social Security income are the most actionable findings; the health pathway finding (levels not trajectories) is theoretically important for lifecycle models. The subjective mortality null is underemphasized — it's a direct rejection of a core lifecycle model prediction that deserves more attention. The sample (HRS 1992–2008, before the Global Financial Crisis's (GFC) full impact and before the explosion of 401(k) concentration) may not generalize to post-2008 cohorts with more heterogeneous defined contribution (DC) plan portfolios and lower SS replacement rates for high earners.