| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Richard W. Johnson, Senior Research Associate, The Urban Institute, 2100 M Street, NW, Washington, DC 20037. E-mail: rjohnson{at}ui.urban.org
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Pensions • Retirement annuities • Survivor protection • Widows • Economic well-being
Survivor protection can be especially important for older women. Despite recent improvements in the economic status of the elderly population, many older widows have low levels of income. In 2000, 17% of widowed women aged 65 and older lived in poverty, compared with only 4% of married elderly women (Social Security Administration, 2002). These high poverty rates result in part from selective mortality: Poor men tend to die at younger ages than wealthy men, leaving behind poor widows. However, many women in general also lose substantial income when their husbands die. In the early 1990s, for example, average incomes for older widows fell 47% following the death of their husbands, in part because many women lost pension income when their husbands died (Holden & Zick, 2000). In addition, more than 40% of the widows of pensioners reported no pension income after their husbands' deaths. Improving access to survivor benefits in employer-sponsored pension plans could raise incomes for older widows and lift some of them out of poverty (although widows with the lowest incomes would not be helped much, because few of them were married to men with pension coverage).
Federal Law on Spousal Pension Rights
Federal pension law encourages joint and survivor annuity payouts. The Employee Retirement Income Security Act (ERISA) of 1974 requires employers that sponsor defined-benefit pension plans to offer joint and survivor annuities as the default option for retiring workers. Before Congress enacted these rules, employer-sponsored plans did not have to offer survivor protection at all. The Retirement Equity Act (REA) of 1984 further requires beneficiaries to obtain the written consent of their spouses before they can decline survivor benefits.
These laws appear to have improved access to retirement benefits for surviving spouses. For example, among men with pension income in 1981, 64% of those who began collecting benefits after the passage of ERISA in 1974 chose a joint and survivor annuity, compared with only 48% of those who began collecting benefits beforehand (Holden & Nicholson, 1998). The passage of the REA appears to have raised the take-up of survivor annuities even further. Survivor benefit coverage for the wives of retirees with employer-sponsored pension coverage increased from about 70% in the 4 years before 1986, when REA went into effect, to about 77% in the 4 years following enactment (Aura, 2001). Legislation introduced in Congress in 2002 and 2003, but not yet passed into law, would extend spousal protections to employer-sponsored defined-contribution plans, which offer workers tax-deferred retirement savings accounts. These plans have grown rapidly during the past 20 years and now cover more workers than defined-benefit plans (Copeland, 2002). Other recently introduced congressional legislation would permit the electronic transmission of spousal consent to waive survivor protection, which some advocates fear would weaken spousal pension rights (National Women's Law Center, 2003).
The Pension Payout Decision
Foregoing survivor protection may be the best option for some couples, even those fully committed to each other's welfare. For example, some couples have access to other resources that could protect the surviving spouse in the case of widowhood, making joint and survivor annuities redundant. Pension coverage in one's own name probably offers the most security. Women who receive their own pension income are less likely to need the husband's survivor benefits to live comfortably in retirement if he should die first. Older women are still less likely than men to receive pension income, but the gap is closing as women's employment continues to rise (Johnson, Sambamoorthi, & Crystal, 1999). Couples also can purchase life insurance to replace the income of the pensioner if he dies first, or they can save more to finance consumption needs in the event of widowhood. Several studies have in fact found that men are less likely to choose joint and survivor annuities if their wives have their own pensions or other assets that can provide financial protection in case the husband dies first (Aura, 2001; General Accounting Office [GAO], 1988; Holden & Nicholson, 1998).
Opting for survivor protection does not necessarily make sense for couples with limited financial resources, because joint and survivor annuities pay lower monthly benefits than single life annuities. Earlier work shows that couples with small pensions and low income tend to elect single life annuities, maximizing their pension income when both spouses are alive and assuming the risk that the spouse will outlive the pensioner (GAO, 1988, 1992; Holden, Burkhauser, & Myers, 1986). White men and well-educated men are more likely than men of color and those with limited education to choose joint and survivor annuities (Aura, 2001; Holden & Nicholson, 1998), perhaps because White and well-educated men can generally better afford survivor protection.
Retirees in poor health and with shorter life expectancies than their spouses are especially likely to select joint and survivor annuities over single life annuities (Holden et al., 1986; Holden & Nicholson, 1998; Turner, 1988). The chances that the spouse will eventually receive survivor benefits generally increase when the pensioner has health problems.
Although the existing literature provides valuable insights into annuitization choices, most previous studies use old data that predate the enactment of the REA and in some cases even ERISA (e.g., GAO, 1988; Turner, 1988). It is important to reexamine annuitization decisions by using more recent data that better describe the current legal, social, and economic environment. In addition, because most of the existing research was motivated by concerns about the retirement security of widows, most studies focus exclusively on the annuitization decisions of married men (e.g., Holden et al., 1986; Holden & Nicholson, 1998). However, as more married women retire with substantial work experience and pension coverage in their own names, their annuitization choices are becoming increasingly important.
Finally, few studies examine the annuitization decision at the time retiring workers are choosing between single life and joint and survivor annuities. Instead, most studies examine a sample of retirees with pension income, and they relate the presence of survivor protection to current characteristics, sometimes many years after the pensioner made the annuitization decision (e.g., GAO, 1988; Holden & Nicholson, 1998). A limitation to this approach is that it understates the share of pensioners who choose joint and survivor annuities if those who elect survivor protection exhibit higher mortality rates than those who elect single life annuities, because the analysis includes only those who survive from the time of initial benefit receipt to the time of the survey. In addition, the characteristics of the couple may change over time, so that certain factors (such as strong physical or financial health) that motivated the choice of a particular payout option may no longer exist when the couple is surveyed years later. In particular, health problems that develop unexpectedly in retirement can obscure the estimated impact of health on annuitization decisions.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The covariates in the model consist of variables designed to capture both the financial and nonfinancial costs and benefits of survivor protection. Financial variables in the model include measures of the value of the pension, the financial resources of the couple, and the availability of alternative sources of survivor protection for the spouse. We expect that couples are less likely to forego survivor protection as the size of the monthly pension increases, reducing the financial pain of trading current pension income for survivor protection. Our hypotheses about the effects of financial resources and the availability of alternative sources of survivor protection for the spouse (such as coverage in her own name) are less definitive. On one hand, by providing financial protection for widowed spouses, wealth and pension coverage in the spouse's own name diminish the importance of joint and survivor annuities to the spouse's economic well-being in widowhood and could reduce the likelihood that pensioners accept joint and survivor annuities. On the other hand, increases in the family's financial resources allow the couple to take a joint and survivor annuity and reduce their current income without necessarily limiting their consumption, which could increase the odds that they accept survivor protection.
The model also includes a number of nonfinancial factors that are likely to play a role in the annuity payout decision, including the health of the pensioner relative to the spouse, attitudes about risk and the importance of providing survivor protection, and the length and quality of the marital relationship. We expect that pensioners tend to reject joint and survivor annuities when their health is better than their spouses' health, decreasing the likelihood that the spouse will become widowed and need survivor protection. Pensioners may be especially likely to elect joint and survivor annuities when they believe in the moral imperative of providing financial protection for their widows and when they are unwilling to take risks.
The quality of the marriage might also affect the payout decision. Pensioners who are less committed to their spouses may be less willing to accept lower monthly payments when they are alive in order to secure survivor protection for their spouses, and they may instead opt for single life annuities. Thus, if men and women whose marriages have not lasted for many years or who have been previously divorced feel less responsible for each other than those who have been married for many years and those who have been married only once, then the length of the marriage and past marital history may influence the payout decision. However, if federal laws requiring workers to obtain the consent of their spouses before they can turn down survivor protection are effective, then the quality of the marriage may not have any impact on the decision to accept joint and survivor annuities.
The model also controls for union membership (because unions may educate their members about the benefits of survivor protection), race (which may reflect unobservable differences in income and health), and educational attainment (which may reflect differences in financial savvy or attitudes about risk). Because some of the variables we examine may be collinear, we report results from three different specifications of the model, which include different combinations of covariates.
Identifying Annuitants Who do not Adequately Provide for Their Spouses
We also estimate the share of employer-sponsored pension recipients who appear to provide inadequate financial protection to their spouses. We identify pensioners who have potentially compelling reasons to reject joint and survivor annuities: having a spouse with adequate access to other types of financial protection (and thus unlikely to need pension survivor protection), having only limited pension benefits or household income (and thus unable to afford survivor protection), and being in better health than the spouse (and thus likely to outlive their spouses). We then compute the cumulative share of pensioners who reject joint and survivor annuities and who do not fall within any of these categories that could potentially justify their decisions.
The analysis considers spouses to have adequate access to alternative types of survivor protection if they receive pension income in their own names, equal to at least 50% of the value of the pension income received by the pensioner; if they have pension coverage in their own names from a job that they have held for at least 20 years (because most pensions from jobs held for less time do not generate substantial benefits); if the couple is in the top quintile of the wealth distribution; or if the pensioner has adequate life insurance coverage, which we define as policies with face values that exceed five times the annual pension income. We define a small pension as one that provides less than $250 per month in benefits, and we classify a couple as having limited income if their incomes fall below 125% of the federal poverty level.
Data and Sample
The data for the project come from the 1992–2000 waves of the Health and Retirement Study (HRS). Designed and fielded by the University of Michigan for the National Institute on Aging, the HRS interviews a large, nationally representative sample of Americans at midlife and older ages and follows them over time. We use data on the original HRS cohort, which consists of 9,741 respondents born between 1931 and 1941. The HRS also surveys the spouses of all respondents, even if they are not members of the specified birth cohort. Respondents were first interviewed in 1992, when they were between the ages of 51 and 61, and they were reinterviewed every 2 years. The survey oversamples African Americans, Hispanics, and Florida residents, and the analysis uses sample weights so that our estimates represent the underlying national population.
At each wave, the HRS asks respondents whether they are receiving income from retirement pensions. For the two largest pensions, respondents report how much they received in the previous month, whether payments will continue until they die, whether the annuity payment adjusts automatically for changes in the cost of living, and whether their spouses would continue to receive payments after they die. We include in our sample all married men and women who were employed and not receiving employer-sponsored pension income in 1992 and then began receiving life annuities by the year 2000. The vast majority of pensioners in our sample participated in defined-benefit plans, although a few were in defined-contribution plans that annuitized their retirement benefits. After we eliminate 45 cases with missing data on pension income, our sample consists of 763 men and 386 women.
Key Measures
We set the dependent binary variable, identifying adults who forego survivor protection for their spouses, equal to 1 if the spouse would not continue to receive any of the respondent's pension income if the pensioner died first. We base the measure on information collected at the time the respondent first reported receiving employer-sponsored pension income. Unless otherwise noted, we base all other variables on data collected at the interview immediately preceding the initial report of pension receipt, which roughly corresponds to the time the pensioner made the payout decision. We express all of the financial measures in our model in constant 2000 dollars, adjusted by changes in the Consumer Price Index.
We measure the size of the pension benefit as the sum of all monthly pension income at the time that respondents first report pension receipt. We express benefits received from joint and survivor annuities as the monthly benefit that would have been paid by an actuarially equivalent single life annuity, which pays higher monthly benefits for a given plan. The adjustment factor depends on the ages of both the pensioner and spouse and on whether the annuity includes automatic cost-of-living adjustments (COLA). For example, our adjustment algorithm indicates that the monthly benefit from a plan without COLA provisions would be 8.7% larger if taken as a single life annuity than as a joint and survivor annuity for married adults who begin collecting benefits at age 65 and are 3 years older than their spouses. Because the size of the pension benefit is likely to raise the likelihood of accepting spousal survivor protection, but at a diminishing rate, we include the natural logarithm of the size of the monthly pension in the models.
We measure financial resources by household wealth and monthly nonpension income. Household wealth equals the net value of all assets held by the respondent and spouse at the time of the survey, including housing, expressed as a natural logarithm. It excludes, however, assets in employer-sponsored pension plans and the value of future Social Security benefits. Monthly nonpension income includes all sources of income received by the respondent and spouse, except the respondent's pension income. Because we want a measure of resources available in retirement, we measure income at the interview in which the respondent first reports pension income (not the previous interview).
The analysis measures alternative sources of spousal survivor protection by whether the spouse has pension coverage from the spouse's own past or present employers and by the value of life insurance held by the pensioner. However, we do not include a measure of life insurance in the final models because it was not significant in preliminary estimates.
The HRS survey asks all respondents to rate their overall health as excellent, very good, good, fair, or poor. We use these rankings to identify respondents who report better health than their spouses. Although these ratings are clearly subjective and the criteria that individuals use to rate their health probably varies substantially, previous research finds that self-assessed health status is a strong predictor of future mortality (Idler & Benyamini, 1997). We experimented with alternative measures of relative health status, based on questions in the HRS about functional limitations, the presence of chronic health conditions, and expectations about surviving to ages 75 and 85, but none of these alternatives were related to the annuity payout decision.
We also construct several measures designed to reflect the quality of the marital relationship. The HRS asks respondents whether they generally enjoy the time they spend with their spouses. We classify pensioners as not enjoying time with their spouses if they said the time together was "not too enjoyable" or only "somewhat enjoyable," instead of extremely or very enjoyable. We compute the length of the marriage in years, expressed as a natural logarithm to capture nonlinear effects of marriage duration on pension payout decisions. Couples whose marriages have endured for many years may have stronger relationships and may feel more responsible for each other than those who wed more recently. We experimented with indicators of a history of divorce, but we dropped them from the final model because they were not significant in preliminary estimates. The model also includes an indicator identifying pensioners who report attending religious services more than one time per week. People who worship frequently may believe that they have a moral obligation to provide survivor protection to their spouses.
The HRS includes a question designed to measure respondents' willingness to assume risk. In 1992, the survey asked respondents to imagine a scenario in which they had to choose between a job that guaranteed them their current family income every year for life and a more risky job that, with equal probabilities, would either double their income or cut it by a third. The survey then asked those who declined the risky job if they would accept it if their income would fall by only 20% if the less desirable outcome were realized. We classify those who accepted the initial job offer as not risk averse, those who accepted the second job offer but not the first as moderately risk averse, and those who rejected both offers as very risk averse. Previous research has found that HRS respondents who are unwilling to accept these hypothetical risky jobs are unlikely to engage in actual risky behaviors, such as smoking, drinking, foregoing insurance, and compiling a financial portfolio that is more heavily weighted toward stocks than Treasury bills (Barsky, Juster, Kimball, & Shapiro, 1997).
Indicators for spousal pension coverage, health status, religious service attendance, risk aversion, and enjoyment of time spent with the spouse are missing for a few members of our sample because the HRS surveyors were unable to interview the spouses of some respondents and other respondents refused or were unable to answer some of the survey questions. Instead of dropping these cases from our analysis, which would reduce our sample size and could bias our estimates if the likelihood of responding is not random, we set the indicator for the measure that is missing equal to 0 and created another indicator to identify respondents with missing data. The models include the missing data indicators, none of which are statistically significant.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
Multivariate Estimates for Married Men
Table 2 reports results from the probit model for married men, for three different specifications. The first specification includes only financial and health variables; the second specification adds variables designed to measure marriage quality, attendance at religious services, and risk aversion; and the final specification adds controls for union membership, race, and educational attainment. The table reports marginal effects, which show the percentage point increase in the probability of rejecting joint and survivor annuities associated with the given variable, holding all other variables in the model constant at their sample means. Standard errors are in parentheses, and asterisks identify effects that differ significantly from zero.
|
Nonfinancial factors are also important determinants of the pension payout decision. Consistent with our hypothesis that pensioners who expect to outlive their spouses tend to reject joint and survive annuities, we find that men who report better health than their wives are about 6 percentage points more likely to forgo spousal survivor protection than men whose health is worse or no better than their wives' health, although the effect becomes insignificant when we control for marriage duration. The length of the marriage significantly decreases the likelihood that men will reject survivor protection for their spouses, perhaps because men in long-term marriages tend to feel more responsible for their wives' financial well-being than those who have been married for only a few years. Men who do not enjoy spending time with their spouses and who are not risk averse are more likely to reject spousal survivor protection than other men, but the effects are not significant. Men who attend religious services more than once per week are less likely to reject joint and survivor annuities than men who attend less frequently or not at all, perhaps because they feel a strong moral obligation to protect their spouses. African American men are significantly more likely than non-Hispanic White men to reject survivor protection, even after financial resources, health status, and spouse's pension coverage have been controlled for. Educational attainment and union membership do not significantly affect pension payout decisions.
Multivariate Estimates for Married Women
Table 3 reports probit estimates for married women, for the same three specifications described in Table 2. As with men, women receiving employer-sponsored retirement annuities are more likely to reject spousal survivor protection when the spouse has pension coverage in his own name than when the only pension benefits derive from his wife's employment. Although the size of the impact is as large for women as men, the effect is not significant for women in all specifications, probably because few husbands in our sample lack pension coverage. The likelihood of rejecting joint and survivor pension annuities falls as the value of the pension increases, but household wealth and nonpension household income are not significant predictors of the pension payout decision for married women.
|
Annuitants Who do not Adequately Provide for Their Spouses
Table 4 shows the share of pension annuitants who appear to provide inadequate survivor protection for their spouses. The cell entries indicate the share who reject joint and survivor annuities and whose decisions cannot be explained by the circumstances described in the row or any rows above it in the table. As noted earlier, 28% of men receiving employer-sponsored retirement annuities forego spousal survivor protection. However, many of these men are married to women who appear to have access to adequate survivor protection through their own pension coverage, their husbands' life insurance coverage, or other assets. Accounting for these cases reduces the share of male annuitants who leave their wives with inadequate survivor protection to only 11%.
|
Among the two thirds of female annuitants who turn down joint and survivor annuities, four fifths are either married to men with pension coverage in their own names or have insurance coverage or other assets that would protect their husbands if they become widowed. As a result, only 12% of married female annuitants leave their husbands with inadequate survivor protection. Some of these women receive small pensions or have limited incomes, so that only 7% of female annuitants could afford to provide survivor protection for their husbands but choose not to protect them. Of these, many report better health than their husbands. Overall, only 3% of women reject spousal survivor protection without evidence of potentially strong reasons.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
These results suggest that additional measures to encourage workers to accept joint and survivor annuities from their employer-sponsored defined-benefit plans are not necessary. Among retirees taking an annuity, it is unlikely that additional efforts to encourage joint and survivor annuities in employer-sponsored defined-benefit plans (or to mandate them outright) would substantially improve economic outcomes for widows in later life. However, persistently high poverty rates among elderly widows may justify additional policy initiatives to improve their retirement security, such as increasing Social Security's survivor benefits or minimum benefits.
Current law may not, however, adequately protect the spouses of workers in defined-contribution plans, soon to be the dominant type of employer-sponsored retirement plan for retiring workers. Federal law does not require most employers with defined-contribution plans to offer annuity options, and the provisions of the REA do not apply to most defined-contribution plan participants who elect to take their retirement benefits as lump-sum payments instead of as annuities. Although research on payouts from defined-contribution plans is limited, the available evidence suggests that most retirees in defined-contribution plans take their benefits as lump-sum payments instead of annuities (Brown 1999; Hurd, Lillard, & Panis, 1998). Even if those who receive lump-sum payments from their employers use the proceeds to purchase annuities from insurance companies, their spouses' survivor benefits are not protected by federal pension law. (Defined-contribution plan participants who receive annuities from their employers are subject to the REA, however, and must obtain spousal consent before they can take single life annuities instead of joint and survivor annuities.) Lawmakers have recently proposed extending the spousal protections available in defined-benefit plans to defined-contribution plans, requiring plan participants to obtain spousal consent before they can take their retirement benefits as lump-sum payments instead of annuities. Given the apparent low rates of annuitization among defined-contribution plan participants, these proposals may have merit.
|
Footnotes |
|---|
1 The Urban Institute, Washington, DC.
Decision Editor: Linda S. Noelker, PhD
Received for publication August 18, 2003. Accepted for publication January 13, 2004.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  K. Henkens, H. van Solinge, and W. T. Gallo Effects of retirement voluntariness on changes in smoking, drinking and physical activity among Dutch older workers Eur J Public Health, December 1, 2008; 18(6): 644 - 649. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. van Solinge and K. Henkens Involuntary Retirement: The Role of Restrictive Circumstances, Timing, and Social Embeddedness J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2007; 62(5): S295 - S303. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. van Solinge Health Change in Retirement: A Longitudinal Study among Older Workers in the Netherlands Research on Aging, May 1, 2007; 29(3): 225 - 256. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
