| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
a Department of Family and Consumer Studies, University of Utah, Salt Lake City.
b LaFollette Institute of Public Policy and Department of Consumer Sciences, University of Wisconsin, Madison.
Cathleen D. Zick, Department of Family and Consumer Studies, 225 S. 1400 E. Room 228, University of Utah, Salt Lake City, UT 84112 E-mail: zick{at}fcs.utah.edu.
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. Data from the Survey of Income and Program Participation are used to investigate the amount and composition of wealth held by four different groups: always married women, about-to-be-widowed women, recent widows, and long-standing widows. Regression analyses assess the impact of group membership on wealth holding controlling for other sociodemographic factors, and annuity calculations assess the potential for wealth to augment income.
Results. About-to-be widowed women have fewer assets than intact couples, and there is a further decline in assets at the time of the husbands' deaths and in the ensuing period. Estimates of the annuity value of widows' wealth show that its liquidation would do little to improve the economic situation of the poorest widows.
Discussion. These findings parallel what is known about income changes that surround the death of a spouse. They also point to the need for additional research on the relationship between wealth holdings and mortality as well as the roles that health care costs, life insurance, and bequests may play in altering widows' wealth.
IT is generally agreed that the incomes of widowed women have lagged behind their married counterparts over the latter half of this century. A recent study found that in the early 1990s, newly widowed women had needs-adjusted incomes that averaged only 68% of similarly aged married couples (Holden and Zick 1997
). The decline in needs-adjusted income that is typically experienced by women after their husbands' deaths is somewhat puzzling in light of the rising labor force participation rates of married women and the expansion of the insurance and pension industries during the latter part of this century. Occasionally, it has been suggested that widows' lower incomes may be partially offset by higher wealth holdings. Yet, very little is known about the wealth holdings of widows.
The death of a spouse is an event that may precipitate a large decline in wealth because the household may need to liquidate assets to cover health care and/or burial costs. In addition, if the individual who died willed assets to anyone other than his or her spouse, this too could drain away some wealth. Life insurance policies and/or lump sum distributions from pension plans may partially or fully offset this decline in assets when the listed beneficiary is the spouse. But, after the death of a spouse, the magnitude of the resultant decline in wealth and the ability to minimize any decline with insurance and/or pension distributions is not well known.
In this article, we use data from the 1990, 1991, and 1992 Survey of Income and Program Participation (SIPP) panels to learn more about the amount and composition of traditional forms of wealth held by four different groups: always married women, women who would soon be widowed, recent widows, and long-standing widows. These comparisons provide insights regarding the magnitude of changes in wealth that occur when a husband dies and the potential for using wealth to augment the income of recently widowed women who might otherwise risk falling into poverty.
Some insights about the wealth holdings of widows can be gleaned by reviewing the literature that focuses on comparing the wealth holdings of elderly and nonelderly households, as the majority of widows are aged 60 and older. Descriptive investigations of asset holdings reveal a considerable dispersion of wealth within the elderly population (Hurd 1990; Smith 1997a, Smith 1997b). For example, Smith 1997a reports that the bottom 10% of individuals in the Asset and Health Dynamics among the Oldest Old (AHEAD) Survey have less than 1% of the wealth of the median AHEAD respondent. Multivariate analyses suggest that this variation in wealth holdings may be partially attributable to differences in income (Smith 1995, Smith 1997b; Wolff 1992), education (Smith 1995, Smith 1997b), race/ethnicity (Smith 1995, Smith 1997b), health status (Smith 1995, Smith 1997b) and marital status (Hurd and Wise 1989; Smith 1995).
Investigations that explicitly focus on the relationship between widowhood and wealth have been limited to three studies to date. Hurd and Wise 1989 use data from the 1969–1979 Retirement History Survey (RHS) of heads of household who were born between 1905 and 1911 to examine pre- and postwidowhood wealth holdings. Their analyses reveal that the typical woman who became a widow between the 1975 and 1977 interviews experienced about a 30% decline in household wealth during that two-year period. Yet, they also note evidence of long-standing differences in wealth holdings, suggesting that the drop in wealth may not generally be attributable to rising medical expenses immediately prior to the death. Nor did they find any evidence that would support the hypothesis that the wealth differences are a function of bequests to children.
Descriptive work by Radner 1993 uses data from respondents aged 65 and older in the 1984 SIPP to estimate differences in wealth holdings among the oldest old (i.e., those individuals aged 80 and older) by gender and marital status. He finds that older widowed women, on average, have a moderately lower net worth, fewer financial assets, and a larger percentage of their assets tied up in housing than do similarly aged men or similarly aged "other women" (i.e., married women, women living with others).
Finally, Smith 1995 uses data from the 1992 baseline interviews of men and women aged 51 to 61 in the Health and Retirement Survey (HRS) to assess the correlates of wealth in a multivariate context. In his analysis, he finds that widows continue to have significantly lower wealth holdings than their married or divorced counterparts even once one controls for current income, health status, education, employment status, region of residence, financial time horizon, subjective life expectancy, and intentions regarding bequests.
At this point, the evidence suggests that the wealth holdings of widows are typically lower than those of similarly aged married couples. Estimates of the magnitude of the difference, however, appear to vary depending on the data used and the analysis approach. All three of the analyses done to date make use of samples that contain age restrictions that may further affect the results. Two of the three also make use of panel data sets that are becoming quite dated (i.e., one is 15 years old and the other is over 20 years old). In order to gain a better understanding of how the death of a spouse affects the wealth holdings of widows, we use data from a sample of recent widows that span a wide age range and compare their wealth holdings to otherwise similar married couples.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Each SIPP panel is a nationally representative sample of households whose members are interviewed at 4-month intervals over approximately a 32-month period. At each interview, data are collected on household composition and the incomes of each household member over the 4 preceding months. In addition, questions from special topical modules, including household wealth and its composition, are asked in specific interviews.
Although income data (including income from assets) are gathered for each month, wealth values are asked only once. The wealth module is part of the Wave 4 interview in the 1990 and 1992 panels, and the Wave 7 interview in the 1991 panel. This creates a challenge in terms of identifying recently widowed individuals across the panels relative to when the wealth questions are asked. We deal with this by classifying widows into three groups: (a) women who enter the SIPP panel as widows and who report in the marital history module that their spouse died at some point in the preceding 3 years (i.e., the Wave 1 widows), (b) women who are married when they enter the SIPP panel but whose husbands die prior to the interviewing wave in which the wealth questions are asked (i.e., the recently widowed), and (c) women who are married when they enter the SIPP panel and who are still married when the wealth questions are asked, but who report that their spouse has died at some subsequent interview (i.e., the about-to-be widowed). As a point of comparison, we also include a random one-quarter sample of similarly aged married women who are married when they enter a SIPP panel and who remain married throughout the 32-month period (i.e., the always married). The only further restriction we place on the respondents in these four groups is that they be aged 40 or older at the time that they enter SIPP. With these restrictions in place, the analyses that follow contain 784 Wave 1 widows, 219 recently widowed women, 293 about-to-be widowed women, and 3,398 always married women.
A second challenge of the data is that the married women in our sample are generally younger than the widowed women. Thus, for the purposes of this study, we developed a weight that is applied to the married households in conjunction with the SIPP panel weights, making the age distribution of the always married women equal to that of the (older) widowed groups. This weight is constructed based on age counts in 5-year intervals. For example, if 10% of the sample of always married women were between the ages of 45 and 50, but only 5% of the widowed sample were in that age range, then the age weight for the always married sample members in this age range would be one half (i.e., 5%/10%). The widowed households are simply weighted by the SIPP panel weights. Thus, in the analyses that follow, any observed descriptive differences in wealth are net of both sampling differences and age differences across the groups.
A final challenge of using SIPP arises because there is a relatively high percentage of missing data on various wealth components. In the sample we use, 43% of the households have missing data for one or more of the asset categories used to construct our wealth measures. Sixty-five percent of those with missing information are missing information on one or two of the wealth questions. Always married households and the about-to-be widowed households are about twice as likely to have missing data on one or more wealth components compared to the Wave 1 widows and the recent widows. This is to be expected because the original questions are asked in terms of each household member's wealth holdings within a particular wealth category. For example, in a married couple household, the respondent reports on the husband's business equity and the wife's business equity separately. Whereas in a widowed household, the respondent reports only on her own business equity. Thus, respondents in married couple households are asked twice as many wealth questions as are respondents in widowed households; this provides them with greater opportunity for nonresponse.
Once we aggregate to get information on wealth components at the household level, we find that, across all four groups, the most common category for missing information is home equity (18%). This is followed by the value of other property (10%), other financial investments (10%), debt (9%), stock and bond holdings (7%), IRA and Keogh account holdings (6%), car value (5%), business equity (2%), and mortgage debt (<1%).
The high percentage of missing data on wealth questions is not unique to SIPP, as respondents in many surveys are either reluctant to answer or do not know the answer to questions about the dollar value of their wealth holdings. Panels like the HRS combat this problem by using bracketed follow-up questions for those respondents who are unable to give a dollar value at first (Smith 1995, Smith 1997b). Unfortunately, SIPP does not utilize bracketed follow-up questions.
Despite the drawbacks of the wealth measures contained in SIPP, we elect to use it in the currently analysis for three reasons. First, it is the only panel that can provide information on wealth (and its components) for recent widows of all ages. Second, SIPP does employ a cross-sectional imputation procedure to deal with missing data on total wealth and its components; thus, households who initially had missing wealth data are still included in the current analyses. In the multivariate analyses, we include a dummy variable to control for the impact of one or more imputations of specific asset components. Finally, scholars who have assessed the general quality of SIPP wealth data have concluded that, despite its limitations, SIPP provides high quality measures of wealth and its components for the overall population and various subgroups (Curtin, Juster, and Morgan 1989, McNeil and Lamas 1989).
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Table 1 contains information on the percentage of households holding various types of assets by widowhood group. The married couple households in this sample are more likely to own a home, have money market accounts, and have certificates of deposit than were similarly aged households (of all types) in the early 1980s (see Levy and Michel 1991, Table 5.7). Yet, those who entered the panel widowed or who became widowed at some point during the panel, have significantly lower probabilities of holding any of these six traditional kinds of wealth compared to their married counterparts. This suggests that perhaps the husbands in these widowed households bequeathed certain assets to individuals outside of their immediate families when they died. Or, perhaps these households did not experience the growth in various asset ownership rates over the past decade that continuously married households have enjoyed. In either case, these differences in ownership are consistent with the long-term income differences between eventual and Wave 1 widows reported in Holden and Zick 1997.
|
|
Although we age-weight the always married households to achieve age distributions similar to those of widowed households, other factors may cause differences in wealth across these groups that, in the bivariate examinations, appear to be due to their widow status. In Table 3 , we use a multivariate approach to control for other characteristics that could affect wealth holdings. Regressions are estimated for all three of the wealth variables detailed in Table 2 ; total wealth, net worth, and financial wealth. Initially, both main effect models and models that included interactions between widowhood group and age, education, and race, were estimated. F tests for the change in goodness of fit between the two models revealed that the more parsimonious main effects models are to be preferred; thus, they are the ones presented in Table 3 .
|
The estimates associated with group membership suggest that the widowhood event precipitates a substantial decline in wealth (i.e., the large negative coefficient associated with being a new widow) with little prospect for economic recovery (i.e., the estimated coefficient associated with being a Wave 1 widow is negative, statistically significant, and larger than the coefficient associated with the recently widowed dummy variable). These results are consistent with the hypothesis that some wealth may be bequeathed to individuals outside of the immediate household at the time of the husband's death and that this loss of wealth is not fully compensated for by the receipt of any life insurance benefits.
What is perhaps more unexpected is the finding that total wealth holdings appear to be lower in widowed households even prior to the death (i.e., the statistically significant coefficient associated with the about-to-be widowed dummy). Yet, this finding is consistent with the work of Hurd and Wise 1989, who also observed relatively long-standing differences in wealth holdings between always married and widowed groups in the RHS data.
The negative coefficients associated with the about-to-be-widowed group may be explained by one or more underlying factors. It may be that husbands who are about to die are in poorer health than their counterparts in the always married households. Smith 1995, Smith 1997b finds that health status is one of the strongest predictors of wealth holdings. Husbands in poorer health may have had lower life-time earnings, which would lead to lower asset accumulation levels. Husbands in poorer health may also have larger health care expenses that their households meet by liquidating assets. In addition, it may be that the husbands in the to-be-widowed households are older than the husbands in the married couple households (even though these regressions control for the woman's age). Older husbands are more likely to be at a stage in the life cycle where they are drawing down on assets because they are retired or have reduced their market work. Finally, it may be that the lower wealth holdings of the widowed households actually contribute to the earlier deaths of these husbands, as prior research on the correlates of mortality have found an association between lower levels of needs-adjusted income and higher risks of mortality (Duleep 1986; Hadley 1982; Zick and Smith 1991a). Singularly or in combination, all of these factors may explain why the typical wealth holdings of the to-be-widowed households are lower than the typical wealth holdings of the always married households.
Control variables included in the regressions merit brief discussion. Across all equations, we find that, holding other factors constant, households in which women are White, more highly educated, and younger all have higher wealth holdings relative to households with non-White women, women with less education, and older women. The findings with respect to race and education are consistent with past multivariate analyses conducted by Smith 1995, Smith 1997b that have examined the correlates of wealth among elderly adults in general. He argues that racial differences in wealth holdings are likely to be a function of many contributing factors, including the fact that members of racial/ethnic minorities typically have lower lifetime earnings, lower inheritances, poorer health, and/or may have different bequest preferences than their otherwise comparable White counterparts.
Like race, the role of education in explaining wealth differences in later life is also likely to be multifaceted. Women with high levels of education have a large stock of human capital. This means that it is more likely that they have had higher lifetime earnings, married men with higher lifetime earnings, and/or enjoyed better health (which can translate into higher earnings) than their otherwise comparable counterparts who have less education. In addition, more highly educated women may have enjoyed a larger inheritance and/or may have a stronger bequest preference than women with less education.
Although we weight the samples so that the married and widowed women have similar age distributions, we include a dummy variable that measures whether or not the woman was aged 60 or younger to control for the impact that life cycle stage may has on wealth holdings. Those households in which the woman is aged 60 or older are more likely to have one or both adults retired from the labor force. As a consequence, these households may be purposefully drawing down on wealth as they enter the later stages of the life cycle. This explanation is consistent with the statistically significant negative coefficient we observe for the age dummy.
Table 4 , Table 5 , and Table 6 show estimates of the effect of economic well-being if the annuity value of wealth holdings is included in income. For each couple and widow, we estimate what income would be obtained if net worth or financial assets alone were annuitized—that is, paid out over the expected lifetime of the woman. The life expectancy estimates used in these annuity payment calculations are based on age-specific, cross-sectional estimates of life expectancy in 1990 (Bell, Wade and Goss 1992); thus, they probably underestimate the actual life expectancies of these women. In addition, these annuity calculations assume no one-time or recurring costs associated with the purchase of the annuity. Both choices simplify our calculations, but they also likely bias the estimates upward.
|
|
|
In Table 5 , we estimate how economic well-being would change if the annuity value of assets were added to income. Among Wave 1 widows, for example, the poverty rate would fall from the 17.8% to 16.9% if financial assets were annuitized. But poverty and near-poverty rates would also fall for the always married couples. Indeed, annuitized financial wealth makes only a small difference in both the absolute and relative poverty rates across these groups. Compared to the always married couples, poverty rates remain 4 times higher for Wave 1 widows and recently widowed women and almost twice as high for the about-to-be widowed group.
To see how the addition of annuitized financial wealth might change the economic situation at the extreme ends of the distributions, Table 6 presents the mean income-to-needs ratio and the income-to-needs adjusted for annuitized financial wealth for those in the top 10% and those in the bottom 10% of the wealth holdings. The figures presented in this table reveal that the addition of annuitized financial wealth has almost no effect on the economic position of those at the bottom of the wealth distribution while it raises the income-to-needs ratio of those at the top of the wealth distribution by more than 50% on average.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Although this research extends previous work on widowhood and wealth, the findings of this study should be interpreted with some caution because of concerns regarding the accuracy of the SIPP wealth data. Specifically, many of the households included in the current analyses had missing information on one or two of the questions used to construct the total wealth, net wealth, and financial wealth measures. Cross-sectional imputations prevented the deletion of these observations, but it is likely that the wealth reports for these households have some measurement error that reduces the validity of our findings. In addition, there may be a greater measurement error in those instances where the wealth questions were answered but the wealth module was administered near the time of the husband's death. A woman whose husband is gravely ill or whose husband has just died may be preoccupied with other more pressing issues and thus may inadvertently provide less accurate responses to these relatively taxing survey questions.
A second reason for caution stems from the fact that the SIPP wealth data are cross-sectional. We exploit the panel information to categorize households by prior or impending demographic changes, and this allows us to draw some tentative inferences about the impact of widowhood on wealth dynamics. However, a more definitive description of widowhood and wealth dynamics would require the analysis of panel data that contains wealth measures on married couples and widowed households at multiple points in time.
With the above caveats in mind, we draw three conclusions from this descriptive examination of widows' wealth holdings. First, our analyses reaffirm the findings of others (Smith 1995, Smith 1997b) who have noted greater wealth disparities among elderly persons relative to nonelderly persons. We find considerable variance in the wealth holdings of widows, suggesting that this primarily elderly subgroup may be contributing disproportionately to the variance in the wealth holdings of the more general elderly population.
Second, from our cross-sectional analyses, we infer that the decrement in wealth attributable to the death of a spouse may happen over a period of time rather than suddenly, at the time of the death. This is consistent with findings based on data gathered in the late 1970s and early 1980s (Hurd and Wise 1989; Radner 1993), suggesting that little may have changed in the relationship between wealth and widowhood over the past couple of decades. As is the case for income (Holden and Zick 1997; Zick and Smith 1991b), couples who are about to experience widowhood appear to have somewhat lower wealth holdings than do couples who remain intact. There appears to be a further decline in wealth when husbands die, and, afterward, wealth appears to continue to fall. Whether the impending death precipitates the decline in wealth or the decline in wealth contributes to the timing of the death, or whether both are simultaneously determined by other factors, we cannot say. A definitive answer to the question of causality must await the analysis of panel data that contain measures of wealth and household composition over time.
Third, while income-to-needs ratios rise and poverty rates fall somewhat when the annuity value of wealth holdings are added to income, our estimates suggest that the gain is likely to be small and is unlikely to alter the relative differences in economic well-being across these groups. It would appear that wealthier households have taken steps to minimize the decline in income (adjusted for needs) experienced by the wife when the husband dies. Thus, the annuitization of the wealth holdings in these households further strengthens their economic position relative to similarly aged widows who have lower levels of need-adjusted income. As such, we cannot assume that the typical low-income widow could improve her economic position by annuitizing her wealth. In this respect, the findings of this study reinforce the contention that public transfer programs (i.e., primarily Social Security survivor benefits) play a critically important role in insuring the economic well-being of low-income widows.
Finally, the descriptive findings presented in this article provide a rich agenda for future research. Why is it that widowed women appear to have lower levels of wealth than their otherwise similar married counterparts? For instance, is it possible that out-of-pocket health care expenditures are part of the answer? Descriptive statistics from the Consumer Expenditure Survey show that out-of-pocket health expenditures rise with the age of the household head and, for older household heads, have increased as a percentage of all household expenditures between 1980 and 1992 (Acs and Sabelhaus 1995). It may be that some wealth is liquidated to pay these rising health care costs in older households where the husband is at greater risk of dying. Likewise, some wealth may be used to cover large funeral expenses, as recent estimates place the average funeral costs at $5,500 (Vargas 1999).
Bequests made by the husband to individuals outside of the immediate household and/or to organizations may also be responsible for some of the decline in wealth. Empirical research on bequest behavior is scant at best. Data from federal estate tax records do show that marital bequests are 79.2% of all deductions for nontaxable returns (i.e., estates of less than $600,000) and 45.7% of all deductions for taxable returns (i.e., estates of $600,000 or more; Eller 1997). Unfortunately, these figures are not presented separately by marital status of the decedent. This makes it impossible to assess how much wealth is bequeathed to someone other than a spouse, but these aggregate figures nonetheless suggest that, in married couple households, some wealth may be lost through transfers to individuals other than the surviving spouse.
If some wealth is used to cover health expenditures and funeral expenses or if some wealth is bequeathed to someone other than the spouse, it is still possible that the loss could be offset by life insurance and/or lump sum pension distributions if the wife is listed as the beneficiary. Again, little research has been done on this topic. For example, we know that in 1990, 70.0% of all decedents who had an ordinary life insurance policy were men, but only 48.7% of the time was the wife listed as the beneficiary (American Council of Life Insurance 1997). Figures such as these hint at the possibility that the wife may not always be the beneficiary on a husband's insurance policy but additional research is clearly needed before such conclusions could be confidently drawn.
Investigations focusing on the issues of health expenditures, bequests, life insurance, and pension distributions are needed if we are to gain a clearer picture of the relationship between widowhood and wealth dynamics. The findings from such studies would be useful for financial educators, gerontologists, and policymakers who work on issues related to ensuring the economic well-being of the elderly. In addition, they would also likely help researchers better understand why the association between widowhood and wealth appears to have changed little over the past couple of decades.
|
Acknowledgments |
|---|
Received for publication December 29, 1998. Accepted for publication September 22, 1999.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  J. D. Fisher, D. S. Johnson, J. T. Marchand, T. M. Smeeding, and B. B. Torrey Identifying the Poorest Older Americans J Gerontol B Psychol Sci Soc Sci, November 1, 2009; 64B(6): 758 - 766. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Pai and A. E. Barrett Long-Term Payoffs of Work? Women's Past Involvement in Paid Work and Mental Health in Widowhood Research on Aging, September 1, 2007; 29(5): 436 - 456. [Abstract] [PDF]
|
|
|
|
 |
|
 J. L. Angel, M. A. Jimenez, and R. J. Angel The Economic Consequences of Widowhood for Older Minority Women Gerontologist, April 1, 2007; 47(2): 224 - 234. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. V. Burkhauser, P. Giles, D. R. Lillard, and J. Schwarze Until Death Do Us Part: An Analysis of the Economic Well-Being of Widows in Four Countries J. Gerontol. B. Psychol. Sci. Soc. Sci., September 1, 2005; 60(5): S238 - S246. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
