Poterba Venti and Wise 2010 — Family Status Transitions, Latent Health, and the Post-Retirement Evolution of Assets

Summary

Using panel data from the Health and Retirement Study (HRS; 1992–2006), the Asset and Health Dynamics Among the Oldest Old (AHEAD; 1993–2006), and the Survey of Income and Program Participation (SIPP; 1997–2005), this paper tracks how total household assets (housing, retirement accounts, and financial assets) evolve after retirement, conditioning on family status transitions between waves. For households that remain continuously married or continuously single, assets grow throughout the period; households that experience a divorce see large, statistically significant asset declines of 27–48 percent, while households entering widowhood show small, generally insignificant changes. A latent health index constructed via principal component analysis (PCA) of 28 health indicators reveals enormous health–wealth stratification: the top–bottom quintile asset ratio grows from 1.7 to 2.2 over the observation window for continuously married HRS households.

Key Claims

Continuing two-person households (2→2) accumulate assets through retirement: +6.3% per biennial interval for HRS 56–61 cohort; +4.6% for AHEAD 70–75 cohort.
Continuing one-person households (1→1) also accumulate: +4.2% (HRS 56–61); +1.4% (AHEAD 70–75); median change not statistically different from zero for the AHEAD cohort.
Divorce triggers large, statistically significant asset declines: −32% (trimmed mean, HRS 51–55); −48% (HRS 56–61).
Widowhood effects are smaller and not statistically distinguishable from zero for HRS cohorts; assets decline for the older AHEAD cohort (−11.7% trimmed mean).
Pre-transition selection: households that will become widowed or divorced already hold 55–65% of continuing 2-person household assets before the transition occurs, making it necessary to control for initial asset levels.
Latent health gradient widens with age: top/bottom quintile ratio rises from 1.7 (1992) to 2.2 (2006) for HRS 56–61 continuing married households; from 1.8 to 2.7 for HRS 51–55; from 1.5 to 1.9 for AHEAD 70–75.
1→1 households show an even larger gradient: ratio = 2.8 (1992) → 4.1 (2006) for HRS 56–61.
Cohort effects are large: HRS cohort households at age 70–75 in 2006 held substantially higher real assets than AHEAD households at the same age in 1993.
SIPP (which includes 401(k) assets excluded from HRS) shows larger within-interval increases for 2→2 and 1→1 groups, consistent with pre-retirement 401(k) accumulation.
Rising assets through retirement do not imply over-saving; households may hold assets as self-insurance against uncertain future health costs (Marshall, McGarry, and Skinner 2010).

Concepts Introduced or Extended

Post-Retirement Asset Drawdown

Entities Mentioned

Quotes

"Households on average seem not to reduce their asset holdings in old age except at the time of changes in family status."

"The ratio of assets of persons in the top health quintile to the assets of persons in the bottom quintile is 1.7 in 1992. The assets of persons in the top quintile increased much more between 1992 and 2006 than the assets of persons in the fifth quintile. By the end of 2006 the ratio of assets in the top quintile to assets in the bottom quintile was over 2.2."

My Take

The key methodological contribution is conditioning on within-interval family status transition groups rather than point-in-time marital status, which prevents confounding asset-level differences with transition effects. The pre-transition selection finding — that future widows already have lower assets years before widowhood — is particularly important for causal interpretation of widowhood wealth effects. The latent health index (28 indicators, evolving PCA quintiles) is a precursor to the more formal health–wealth decomposition in Poterba, Venti, and Wise (2013a). The paper is a working-paper draft; the published version appeared as a chapter in the NBER Explorations in the Economics of Aging volume (Wise, ed., 2011, pp. 23–69).