NBER Reporter 2012 Number 1: Research Summary
Wealth After Retirement
Steven Venti*
Over the past two decades, the aging of the "Baby Boomers" has focused attention on how members of this generation accumulated assets during their working years. Now that the leading edge of this group has passed into retirement, the focus of researchers -- as well as policymakers and the financial services industry -- is shifting to the drawdown of financial resources in later life. My recent research, much of which is co-authored with James M. Poterba and David A. Wise, focuses on the factors that shape the age profile of wealth after retirement. The goal of this work is to better understand what households do with their assets after retirement and, in particular, to understand how asset draw-down decisions are affected by health, education, and the structure of public and private annuities.
Wealth at Retirement
Retired households depend primarily on three sources of financial support in retirement: benefits from the Social Security system; payments from private defined benefit (DB) pension plans; and withdrawals from household savings, including withdrawals from personal retirement accounts (PRAs) such as IRAs, Keoghs, 401(k)s and similar defined contribution plans. Benefits from Social Security and DB pensions are in the form of annuities that provide a stream of payouts until death. Assets held in PRAs or financial assets held outside of retirement accounts are typically not annuitized and are instead spent or saved at the owners’ discretion. In a recent paper, we describe the balance sheets for households headed by someone between the ages of 65 and 69 in the 2008 wave of the Health and Retirement Study (HRS).1 To facilitate comparison of the various portfolio components, we capitalize Social Security and DB pension payouts. Averaged over all households, the capitalized value of Social Security benefits accounts for about one-third of all household wealth and housing and other real estate account for another one-quarter of wealth. The capitalized value of DB pension benefits, assets held in PRAs, and financial assets held outside PRAs each account for an additional 10 to 15 percent of total wealth.
These averages hide substantial differences in both the level of total wealth and its composition. At the 90th percentile of the wealth distribution, financial assets (including PRAs) and DB pension wealth account for over half of all balance sheet wealth and Social Security is relatively unimportant. At the other extreme, in the lower part of the wealth distribution, many households have few assets outside of Social Security and housing. Half of all households headed by someone between 65 and 69 had total financial assets, including 401(k)s and IRAs, of less than $52,000 in 2008. Thus, a large fraction of households have few assets, with the possible exception of housing equity, to supplement annuity income -- primarily Social Security -- in retirement. For example, only 47 percent of these households have sufficient financial assets to purchase a private annuity that would increase their life-contingent income by more than $5,000 per year. 2 Thus it is perhaps not a surprise that we observe that in later years 19 percent of all persons die with zero financial assets and 46 percent of all persons have less than $10,000 of financial assets at death.3
The Trajectory of Assets after Retirement
What happens to non-annuitized assets after retirement? The standard life-cycle model suggests that households will gradually draw down non-annuitized assets to finance consumption in retirement. In an earlier paper, Wise and I looked at the trajectory of home equity -- the primary non-annuity asset for most households -- in retirement.4 We found that most of the decline in home equity was accounted for by changes in health status (particularly nursing home entry) or the death of a spouse. Households that did not experience these shocks reduced housing equity little, if any. Thus households appear to conserve equity in homes to tap in the event of substantial expenses rather than to use this equity for day-to-day consumption needs. Indeed, when asked in surveys, most households plan, desire, and expect to die in their homes.
More recently we looked at how personal retirement accounts were drawn down in retirement.5 Unlike housing, which provides housing services while also being a store of wealth, personal retirement accounts are designed and promoted as a means of saving to finance retirement expenditure. We find a rather modest rate of withdrawal prior to age 70 1/2 when households are required to take minimum distributions. Only 17 percent of households between the ages of 60 and 69 who own a personal retirement account make any withdrawal in a typical year. The rate of withdrawal rises sharply after age 70 1/2, suggesting that many households in their early seventies would not make withdrawals if it were not for the distribution rules. Even after age 70 1/2, the percentage of assets withdrawn in a typical year is less than the rate of return earned on PRA assets, generating an age profile of increasing personal retirement account balances for many households over the 1990s and the first half of the last decade.
In a subsequent paper, we broadened our analysis to consider the evolution of total non-annuitized assets after retirement.6 We found that the evolution of assets is strongly related to family status transitions. For both single individuals and married couples who do not experience death, separation or widowhood, average total assets increase well into old age. In contrast, married individuals that experience a family status transition exhibit much slower asset growth prior to the transition and often experience a large decline in asset values at the time of the transition.
The Role of Health
Health care costs are a major financial concern for elderly households. A substantial portion of these costs is paid for by Medicare. For some households Medicaid and private insurance are also important sources of payment. However, none of these programs cover all of the costs of poor health, particularly costs associated with long-term poor health rather than with specific health events. Such costs include expenditures associated with home relocation, home alterations, transportation, the need to hire a cleaning service, and the like. For these expenses, households must either pay out of current income or spend down assets.
A complete accounting of all of the costs associated with poor health is difficult to obtain. Surveys of out-of-pocket medical expenditures typically fail to elicit all of these costs, especially those that are tangentially related costs of poor health. Moreover, most survey questions focus on the costs of specific health events and do not capture the continuing effects of poor health on household expenditures that are likely to persist well after a specific event occurs. We attempt to infer the "full" cost of poor health by estimating the cumulative effect on assets of all of the adverse consequences of poor health over a long period of time. 7 Using data from the HRS, we compare the asset growth between 1992 and 2008 of persons with similar assets in 1992, but different levels of health. The analysis uses an index of health constructed from a large number of self-reported health questions concerning functional limitations and the presence of health conditions. The results indicate that the asset cost of poor health may be quite large, substantially larger than most survey-based estimates of out-of-pocket expenditures. For example, within each asset quintile, the healthiest individuals (those in the top third of the health distribution) accumulate at least 50 percent more assets by 2008 than do the least healthy (those in the bottom third of the distribution). The dollar differences in wealth accumulation associated with differences in health are substantial: for households near the median of the wealth distribution in 1992, the healthiest households accumulated $135,000 more assets by 2008 than did the least healthy households. For households with similar assets who were in the top asset quintile in 1992, the gap between households in the top and bottom thirds of the health distribution was over $470,000 by 2008.
Poor health can trigger asset drawdown in many ways. The most direct pathway is through greater health-related expenditures. However, the effect of poor health and associated expenditures may differ, even among persons with the same wealth. For example, we show that Social Security benefits and DB pension benefits can be "protective" of assets in retirement.8 Among households entering retirement with the same health and wealth, those households receiving higher annuity benefits have substantially lower rates of asset drawdown. The magnitude of these effects varies considerably by level of wealth, but on average an additional $10,000 of Social Security benefits is associated with a $10-15,000 increase in assets over the two-year intervals between waves of the HRS.
There are also important differences in the post-retirement asset drawdown of households with different levels of education. Education clearly affects the level of assets a household accumulates before retirement through its effect on pre-retirement earnings and health. In another paper, we focus on two potential routes through which education may affect asset drawdown after retirement.9 One route is through the effect of education on the trajectory of health after retirement. The other is through the effect of education on asset allocation and investment returns. It is possible that persons with higher levels of education will earn greater returns on their investments, either because they will choose to allocate their assets differently, and to hold more risky but higher expected return assets, or they will hold assets within each asset class that generate higher returns, than their counterparts with lower levels of educational attainment. Analysis of data from several HRS cohorts indicates that effect of education on asset growth through both the health and asset return channels is substantial. The effect of education on the two-year change in assets varies by the initial level of assets and by marital status. For married persons, the two-year increase in assets is $15,000 to $36,000 greater for persons having a college degree than for persons without a high school diploma when both channels are combined.
Summary
There is substantial diversity in the financial circumstances of households entering retirement. This means that households in different parts of the wealth distribution face very different decisions concerning asset drawdown after retirement. We find that for most households in the bottom half of the wealth distribution, there is no drawdown decision to be made. Other than housing, these households hold few assets that could be drawn down or annuitized to supplement other sources of income. Moreover, these households appear to treat their home equity as a source of reserve wealth to be tapped in extreme circumstances rather than as a source of income for day-to-day consumption. The meager level of non-housing assets among many households is also, in part, an explanation for low rates of participation in private annuity markets. Many annuity providers require a minimum investment. Forty-three percent of the households aged 65 to 69 would not be able to make a $25,000 minimum investment even if they liquidated all of their financial assets, including personal retirement accounts.
For households that enter retirement with substantial assets, the late life financial planning problem is multifaceted. At least early in retirement, households appear reluctant to spend down assets as predicted by simple life-cycle models. Surprisingly, households are even reluctant to spend down assets held in IRAs, Keoghs, and 401(k) plans -- perhaps the assets one might expect households to rely on first to meet consumption needs. Only 17 percent of persons between the ages of 60 and 69 make a withdrawal from a PRA, and even after required minimum distributions begin at age 70 and 1/2, balances in these accounts often continue to grow with age. For households with substantial assets that are observed to drawdown assets, most of the drawdown is associated with changes in family status, such as divorce or widowhood, and poor health. The long-term effect of health on the level of assets -- what we call the "asset cost" of poor health -- can be substantial. Among households with the same initial wealth in 1992, those in the top third of the distribution of health accumulate at least 50 percent more assets by 2008 than do those in the bottom third. But even among households with the same initial wealth, the asset cost of poor health varies substantially. The size of the asset decline clearly depends on the severity and nature of the decline in health faced by each household, as well as on the presence of annuity income from Social Security or DB pensions that can substitute for asset withdrawals.
* Venti is a Research Associate in the NBER's Program on the Economics of Aging and the DeWalt Ankeny Professor of Economic Policy at Dartmouth College.
1. J. Poterba, S. Venti, and D. Wise, "The Composition and Drawdown of Wealth in Retirement", NBER Working Paper No. 17536, October 2011. A shorter version appears under the same title in Journal of Economic Perspectives, 25(4), fall 2011.
2. Ibid.
3. J. Poterba, S. Venti, and D. Wise, "Were They Prepared for Retirement? Financial Status at Advanced Ages in the HRS and AHEAD Cohorts", NBER Working Paper No. 17824, January 2012, presented at an NBER Conference on Aging, Carefree AZ, May 2011.
4. S. Venti and D. Wise, "Aging and Housing Equity: Another Look", NBER Working Paper No. 8608, November 2001, and Perspectives in the Economics of Aging, D. Wise, ed. University of Chicago Press, 2004.
5. J. Poterba, S. Venti, and D. Wise, "The Drawdown of Personal Retirement Assets", NBER Working Paper No. 16675, January 2011.
6. J. Poterba, S. Venti, and D. Wise, "Family Status Transitions. Latent Health, and the Post-Retirement Evolution of Assets", NBER Working Paper No. 15789, February 2010, and Explorations in the Economics of Aging, D. Wise, ed., University of Chicago Press, 2011.
7. J. Poterba, S. Venti, and D. Wise, "The Asset Cost of Poor Health", NBER Working Paper No. 16389, September 2010.
8. J. Poterba, S. Venti, and D. Wise, "The Role of Social Security Benefits in the Asset Cost of Poor Health," draft submitted to SSA, September 2011.
9. J. Poterba, S. Venti, and D. Wise, "Health, Education, and the Post-Retirement Evolution of Household Assets," prepared for the RAND Conference on Health, Aging, and Human Capital, November 2011.