The Gerontologist Vol. 47, No. 2, 224–234 Copyright 2007 by The Gerontological Society of America The Economic Consequences of Widowhood for Older Minority Women Jacqueline L. Angel, PhD,1,2 Maren A. Jimenez, MA,2 and Ronald J. Angel, PhD2 ´ Purpose: We compare the economic consequences of widowhood for preretirement age and earlyretirement age Black, Hispanic, and non-Hispanic White women. Methods: We use the 1992 and 2000 waves of the Health and Retirement Study to assess the effects of widowhood on the household incomes and assets of non-Hispanic White, Black, and Hispanic women who were 51 years of age or older at baseline (N = 4,544). Results: For women of all racial and ethnic groups, marital disruption, including widowhood, results in a substantial decline in household income and assets. Net of demographic controls, the relative loss is far greater for Black and Hispanic widows than for non-Hispanic White widows. Implications: The data reveal a substantial widowhood penalty for total household income and net worth for women in each racial and ethnic group. However, the findings suggest that minority widows are at a particularly high risk of poverty in late life, given that they have lower incomes and fewer assets to begin with. Implications of the results for the financial security of women approaching retirement are discussed. Key Words: Assets, Black and Hispanic Women, Health and Retirement Study, Income, Widowhood During the closing decades of the twentieth century, the normative marital and family situations of women changed dramatically. Today a lifelong marriage to a single partner has become less common than it was in earlier times for all income, racial, and ethnic groups. These changes in marital patterns have potentially profound implications for women’s economic security in late life. For an increasing number of women, and particularly Address correspondence to Jacqueline L. Angel, University of Texas at Austin, Lyndon B. Johnson School of Public Affairs, PO Box Y, Austin, TX 78713. E-mail: jangel@mail.utexas.edu 1 Lyndon B. Johnson School of Public Affairs, The University of Texas at Austin. 2 Population Research Center, The University of Texas at Austin. 224 minority group women, marriage does not represent a guarantee of financial stability in old age (Harrington Meyer, Wolf, & Himes, 2006). Despite this new social and family reality, though, many women who are currently approaching retirement age spent their adulthood under the old normative system in which a woman’s economic security depended on that of her husband. For these older women, retirement security is closely tied to their husband’s level of asset accumulation, as well as their own marital histories (Smock, 1993; Waite & Gallagher, 2000). In the traditional marital and work systems, a husband occupied the role of primary breadwinner and a wife’s labor force activity, if she worked outside the home at all, remained largely supplemental. Today, even middle-class women often find that they cannot rely on a husband to guarantee their financial security in later life. By the year 2000, approximately 13% of women over the age of 40 were divorced (U.S. Census Bureau, 2005). Nevertheless, women are less likely than men to have their own defined-benefit or defined-contribution retirement plan, and even when they do have one, they receive significantly lower pensions (Wilmoth & Koso, 2002). For older women, widowhood remains the most common marital transition, and for minority group women, as for those who have never worked themselves, or those whose husbands spent their lives in manual labor with low incomes and episodic work histories, economic insecurity in old age has always been a fact of life and widowhood has meant only further hardship (Holden & Kuo, 1996; Holden & Smeeding, 1990; Holden & Smock, 1991; Jenkins 2003; Wise, 1996; Zick & Smith, 1991). Increasingly, though, marital disruption occurs early in a woman’s life, well before a woman faces the elevated risk of widowhood associated with aging, and has important implications for her old age economic security. In what follows we compare the household income and assets of Black, Hispanic, and non-Hispanic White widows with those of similarly aged women in other common marital statuses in the preretirement and early-retirement years. Given the complexity of contemporary marital histories, an understanding of The Gerontologist the economic consequences of widowhood and other marital statuses on women’s economic security in later life has important policy implications. We focus on minority group status and widowhood because of the fact that, given their lifelong economic disadvantage and the inability of many minority couples to accumulate substantial wealth, widowhood can represent a particularly serious threat to economic security for older minority women. Although the overall economic situation of elderly Americans has improved greatly in recent decades, leaving less than 9% below poverty, Black and Hispanic women remain particularly vulnerable to poverty in old age (Ozawa, 1995). Several significant racial and ethnic trends may account for economic hardship among women. Black women have higher employment rates, defined in terms of the proportion of the population group in the labor force, than do Hispanic women (63.2% vs 56.9%) in the 2000 Census (Fullerton & Toosi, 2001). Disparities also exist for Black and Hispanic women’s median weekly earnings ($491 and $410, respectively) compared with non-Hispanic White women ($567) in 2003 (U.S. Department of Labor, 2004). The late-life economic consequences of the double jeopardy of minority group status and being female, then, reflect the lifelong disadvantages that begin in the earliest years of the life course (Hatch, 1999). Many of these women have limited or no paid work histories, low lifetime earnings, and no personal pensions (Angel & Angel, 1996; Kijakazi, 2001; Lupton & Smith, 2003; Smith, 1995; Smith & Kington, 1997). Data from the Current Population Survey indicates that Hispanic widows were the least likely to receive private pension income between 1994 and 1998 (10.1%), followed by Blacks (16.4%) and non-Hispanic White widows (22.8%; see Kijakazi). In addition, minority women are also less likely than non-Hispanic White women to have adequate health insurance coverage throughout the life course. This lack of insurance reflects their lifelong economic vulnerability as well as their own and their spouse’s labor force disadvantage. The lack of insurance feeds back in a vicious cycle that includes compromised health and diminished life chances (Institute of Medicine, 2001). Black and Hispanic women often do not qualify for Social Security on the basis of their own employment, because they have not made the required 10 years (or 40 quarters) of contributions (U.S. Government Accounting Office, 1998). In addition, many do not claim spousal or widow (noncontributory) benefits because they are either ineligible as a result of the marriage requirement (in the case of divorcees), or a spouse was not fully covered at the time of his death. Given current marital trends, fewer minority women will be eligible for spouse and widow benefits in the future (Harrington Meyer, Wolf, & Himes, 2004). Even when they receive Social Security, minority group women and widows Vol. 47, No. 2, 2007 225 often receive only the minimal payment, and they lack other retirement income sources (Social Security Administration, 2002).1 For affluent women, the loss of a husband may result in a substantial drop in income, but if the couple has substantial community property then the loss of income can be compensated for through the liquidation of assets. Fewer Black and Hispanic women than non-Hispanic White women have such accumulated assets (Honig, 2000). Nevertheless, the majority of women, including Hispanic and Black women, will spend many years without a husband in old age. In 2003 only 25.4% of Black women and 39.9% of Hispanic women aged 65 years and older lived with a spouse (He, Sengupta, Velkoff, & DeBarros, 2005). In addition to widowhood, these living arrangement patterns reflect high rates of marital disruption among minority couples, as well as low rates of marriage among Blacks (Holden & Kuo, 1996; Kreider & Fields, 2002). The situation is compounded by the fact that Hispanic and Black women are also less likely than non-Hispanic Whites to remarry after divorce (Norton & Miller, 1992). As they are forced to liquidate assets, many widows suffer a long-term decline in income and net worth. The level of accumulation during the marriage, therefore, is particularly important, because on average the longer a couple is married, the more communal wealth they accumulate (Holden & Kuo, 1996; Holden & Smock, 1991; Smith & Kington, 1997; Wilmoth & Koso, 2002; Zick & Holden, 2000). If a couple that has paid for their home and acquired other property remains together until the husband dies, then the widow inherits the communal property and often some portion of her husband’s retirement if her husband had a private pension and if it included survivor’s benefits. She also usually qualifies for survivor’s benefits or her own Social Security benefit payment based on her employment history (Harrington Meyer et al., 2006). Given the lifelong labor force disadvantage faced by Blacks and Hispanics, a minority couple’s asset accumulation often remains inadequate, even if they remain married; they risk serious poverty in old age (Crystal & Shea, 2003; Holden & Smock, 1991; Johnson, Sambamoorthi, & Crystal, 2003; Wolff, 2003). In 2001, the income of Black couples was only 80% that of non-Hispanic White couples, and Black couples had accumulated only one fourth the net worth of White couples (Shapiro, 2003). Few studies have examined Hispanic ethnicity differences in wealth, but data from the Health and Retirement Study (HRS) show that Hispanic households fare worse than Black households (Angel & Angel, 1996), averaging 70% of non-Hispanic White couple household income, and only 30% of White household wealth (Smith, 1995). Although marriage remains a major source of financial protection for women in old age, their own retirement income can play an important role. All else being equal, those women who have been employed tend to be better off economically in old age than women who have not been employed (Willson & Hardy, 2002). Despite the episodic and often casual nature of the work in which they engage, their own employment is particularly important among Black women (Willson, 2003). For a disproportionate fraction of Black and Hispanic women, retirement planning is particularly problematic and may be more difficult because they are low paid and poorly educated (Lichter, LeClere, & McLaughlin, 1991). Episodic employment, which is typical of the employment histories of low-income minority women, provides little by way of retirement security (Bound, Schoenbaum, & Waidmann, 1996; Willson, 2003). The literature, then, identifies various key factors that increase the probability of economic security for older women. A lifelong marriage to a husband with a high income and a generous retirement plan is clearly an asset. Having one’s own retirement plan is increasingly important for women as marital dissolution becomes more common. The combination of a large communal estate and income from both private retirement plans and Social Security undoubtedly guarantees an older woman’s financial wellbeing. Unfortunately, for many minority group women nearing retirement, few of these sources of economic security are available. In what follows we assess the differential impact of widowhood on the economic situation of Black, Hispanic, and nonHispanic White women in the preretirement and early retirement years. In order to assess the relative impact of widowhood on economic well-being, we include divorced women as a comparison in order to compare the magnitude of any decrement. Although the literature points to potentially deleterious effects of widowhood on economic well-being, the differential impact of widowhood for different racial and ethnic groups remains poorly understood (cf. Wilmoth & Koso, 2002). From what is known, we might expect that although non-Hispanic White women may suffer larger absolute declines in income and assets as the result of widowhood, the relative loss and the practical consequences for economic security will be larger for Black and Hispanic women who enter the retirement years with lower incomes and fewer assets. Data and Methods In order to investigate the absolute and relative economic consequences of widowhood for nonHispanic White, Hispanic, and Black women, we use two waves (1992 and 2000) of the HRS, a longitudinal dataset consisting of a nationwide representative sample of individuals who were born between 1931 and 1941 and their spouses (Heeringa & Connor, 1995). The HRS consists of five waves collected in 1992, 1994, 1996, 1998, and 2000, but we use only the first and last waves. Given our focus on 226 the relative impact of widowhood for Black and Hispanic women, more detailed analyses of different patterns of change are not possible given the small sample sizes and small number of transitions other than widowhood during the interim. As a result, we focus only on net changes over the 8-year period between 1992 and 2000. However, we do include divorced women as a comparison group to contrast them to the magnitude of income and asset loss associated with widowhood. The HRS respondents were between the ages of 51 and 61 years at the time of the baseline survey in 1992. As part of the survey, researchers collected information on the respondents’ employment and marital histories, as well as their incomes and financial assets. They collected similar information for their spouses. Researchers reinterviewed the respondents in 1994, 1996, 1998 and 2000. At the time of the 2000 interview, the original respondents were between the ages of 59 and 69. Their spouses could have been of any age. More than 81% of the 12,521 respondents interviewed at baseline were also interviewed in 2000. Of the original sample interviewed, 8.7% was lost as a result of mortality by the fifth wave in 2000 (Health and Retirement Study, 2002). Comparisons of the demographic and economic characteristics of those lost to follow-up and those recontacted showed that there were no statistically significant group differences (t test) in total household income and total household net worth. Respondents who were lost to follow-up were slightly older and had fewer years of education completed than did women who were interviewed at both waves.2 For the purposes of analysis, we exclude spouses who were younger than 51 years of age at the baseline survey in order to more clearly focus on the experiences of women near or at retirement age (n = 1,049). We include spouses who were older than 61 years of age (n=109) in 1992 in order to maximize the Black and Hispanic samples. Because the HRS oversampled Blacks, Hispanics, and Florida residents,3 it provides a unique opportunity to examine racial and ethnic differences in women’s economic well-being (Jackson, Lockery, & Juster, 1996). Women of racial and ethnic categories other than non-Hispanic White, Black, or Hispanic (n = 133) are dropped from the analysis. Finally, because of the fact that never married women and those who remarry in late life are theoretically different than married, divorced, or widowed women on multiple dimensions, they are excluded from the multivariate analyses (Rubenstein, 1987). This category represents a nonnormative marital pattern for this historical period. Measures Economic disadvantage among mature minority women is associated with a variety of factors, including prior marital disruption. By employing The Gerontologist the HRS, we can observe the income and wealth effects associated with minority group and marital statuses in late middle adulthood. Midlife is a critical point in the life course at which a couple’s or an individual’s income is close to its maximum and a time by which a large portion of one’s lifetime asset accumulation has already occurred. It is also a time of the life course at which expenses can increase or adversity can take an economic toll. The need to help children financially, a medical crisis, the onset of chronic diseases such as diabetes or hypertension, or the loss of a long-term job can undermine a couple’s economic security. By the middle and late 50s, one’s possibilities for an economically secure retirement are largely determined. It is an ideal period in the adult life course, therefore, to compare the situations of Hispanic, Black, and non-Hispanic White women. Dependent Variable A variety of income and wealth measures are available at each wave of the HRS (Smith, 1995). Total household income includes income from all sources received during the year preceding the survey, which includes the following: respondent’s earnings, household capital income, income from employer pension or annuity, income from Social Security DI or SSI, income from Social Security retirement, unemployment or worker’s compensation, income from other government transfers, and the income of all other household members. Total household net worth is measured as the value of all assets minus total debt. More specifically, it is the sum of the net value of real estate, vehicles, and businesses; IRA/Keogh accounts, stocks, mutual funds, and investment trusts; the value of checking, savings, or money market accounts; of CD, government savings bonds, and T-bills; the net value of bonds and bond funds; and the net value of all other savings, minus the value of other debt, including mortgages and other home loans. Total household income and total household net worth for Time 1 and Time 2 are estimates developed by the RAND Corporation and are based on a multistage approach to imputing values for missing income and assets for those respondents who did not provide answers to these questions. The RAND HRS Data file is a user-friendly version of the longitudinal HRS data. It is particularly useful when one is dealing with income and asset measures, because it combines the cumbersome number of measures included in the original version of the HRS into fewer variables. In addition, we set negative assets to zero, and we inflated income and assets for Time 1 to 1999 and 2000 values, respectively. The inflation factor we used is based on the Bureau of Labor Statistics Consumer Price Index and the year in which the income (1991) and asset (1992) data were collected. Because of the skewed distribution Vol. 47, No. 2, 2007 227 of income and assets, we truncate each at the 90th percentile. Independent Variables We used several well-known factors to characterize a woman’s (respondent’s) economic situation, including marital history, age, education, race, and Hispanic ethnicity. To differentiate among marital statuses between 1992 and 2000, we created four categories. Two refer to widows: (a) no change, which included those respondents who were widowed at both Time 1 and Time 2; and (b) newly widowed women who were married in Time 1 but widowed by Time 2. We include two other marital statuses in order to compare widows with women in other common marital situations: (c) women who were divorced or separated at both Time 1 and Time 2, and (d) women who were married at both waves. Initially we included a dichotomous variable for those who had experienced a marital disruption prior to 1992, as well as continuous variables for the number of years a woman had been married prior to the first wave and the total number of children she had had. None of these variables were significant, and thus we do not present them in the final models. Indicators of women’s paid employment and current occupation were also not significant, so we do not present these results either. The remaining independent variables are all measured at Time 1. Demographic controls include age in years; education, categorized as less than high school, high school, and more than high school; and race and ethnicity. Furthermore, we include controls for total household income and total household net worth at baseline to account for initial levels of income and assets in change models. Our multivariate analyses are based on ordinary least squares regression (Jaccard, Turrisi, & Wan, 1990). Results Table 1 presents weighted baseline sample characteristics for Hispanic, Black, and non-Hispanic White women. Hispanic and Black women have lower educational levels, lower income, and approximately half the assets of non-Hispanic Whites. All groups are of roughly similar age. Black women are twice as likely as non-Hispanic White or Hispanic women to be widowed in late middle age. Table 2 presents information on marital status at Time 1 and Time 2 by race and Hispanic ethnicity. The first row shows that Black women were far less likely than either non-Hispanic White or Hispanic women to have been married at both interviews. A far larger proportion of Black women than either of the other two groups were divorced at both times. Hispanic women are somewhat more likely than Table 1. Weighted Descriptive Statistics at Time 1, 1992 Sample Characteristics Non-Hispanic White Black Table 2. Stability and Change in Marital Status Between 1992 and 2000 by Race and Hispanic Ethnicity Hispanic Demographic characteristics Age (years) 55.9 55.7 55.5 Education Less than high school High school More than high school 18.2 43.8 38.0 39.0 32.1 28.9 Race or Hispanic Ethnicity 64.7 17.6 17.7 Marital status Married Divorced or separated Widowed Never married Income and assets Total household income Total household net worth Unweighted n 78.7 11.2 7.4 2.7 50,802.77 268,496.60 3039 48.8 25.6 17.6 8.0 28,356.39 Type of Marital Status Change Married T1 and T2 Divorced T1 and T2 Widowed T1 and T2 Widowed between T1 and T2 Married between T1 and T2 Divorced or separated between T1 and T2 Never married 68.3 19.5 8.4 3.8 26,622.38 Non-Hispanic White 68.3 (2103) 10.4 (301) 6.9 (196) Black Hispanic 37.6 (283) 58.7 (218) 24.8 (153) 17.8 (59) 17.4 (111) 8.4 (28) 8.6 (275) 8.7 (60) 7.2 (24) 1.4 (38) 1.2 (9) 1.8 (6) 1.8 (53) 2.6 (73) 2.5 (16) 7.6 (47) 2.1 (7) 3.8 (13) Notes: Unweighted n sizes are given in parentheses. Information is from the Health and Retirement Study Public Use Dataset, 1992–2000 (2002). T1 = Time 1; T2 = Time 2. 69,166.56 127,965.08 681 356 Note: Information is from the Health and Retirement Study Public Use Dataset, 1992–2000 (2002). non-Hispanic White women to have been divorced at both times. Black women were also more likely than the other two groups to have been widowed at both waves. The proportion of women who were married at Time 1 but widowed by Time 2 is similar for all groups, roughly 8%. Only small fractions of women of any group divorced or married between the two waves. Blacks were also twice as likely as the other racial or ethnic groups to never have married. Tables 3 and 4 present measures of income and assets by marital status at Time 1 and the amount of change in income and assets between Time 1 and Time 2 in 1999 and 2000 dollars, the years the information was collected. In these tables, neither assets nor income is truncated. Table 3 presents information on total household income in 1992 (top panel) and the change in household income between 1992 and 2000 (bottom panel). The first row of the top panel reveals that among women who were married at both waves, non-Hispanic White women had far higher incomes than either Black or Hispanic women in 1992. Among those who were widowed at both waves, non-Hispanic White women had the highest incomes, although their income was far lower than that of non-Hispanic White women who remained married. Among those who were divorced at both waves, Hispanic women had the lowest initial incomes. The first row of the second panel shows that among those who were married at both waves, nonHispanic women experienced a substantially smaller absolute loss ($1,150 or 1.6%) in income, a loss that was far smaller than that of Blacks ($6,737 or 13.6%) or Hispanic women ($6,158 or 15.3%). Among those women who became widows between 1992 and 2000, 228 all groups suffered substantial losses in income, but these were larger for non-Hispanic White women ($22,534 or 39.7%) and Black women ($22,121 or 54.6%) than for Hispanic women ($10,933 or 27.1%). Among those women who were widowed at both waves, non-Hispanic women ($5,164) suffered the largest loss in income, and among the small number of women who were divorced at both waves, Blacks suffered the largest loss ($4,830 or 19%). Table 4 presents similar information for total household net worth. The first row of the top panel reveals that among those women who were married at both waves, non-Hispanic White women lived in households with far greater assets in 1992 than did either Black or Hispanic women. Among those who were widowed at both Time 1 and Time 2, nonHispanic White women also had the highest net worth at the first wave compared with Black women. In contrast, continuously married Hispanic women suffered a slight drop in household assets at the follow-up ($25,604 or 12.7%). Among those women who became widowed, as among the comparison group of divorced women, non-Hispanic White women had the highest initial net worth, followed by Hispanic women. The second panel of Table 4 shows changes between 1992 and 2000 in net worth for the different marital statuses. The first row reveals that among those women who were married at both Time 1 and Time 2, non-Hispanic Whites experienced an increase in net worth of $174,261 or 44%). Similarly, continuously married Black women also reported an increase in assets ($38,318 or 35.3%). In contrast, Hispanic women experienced a net loss of $25,605 or 12.7%. Non-Hispanic White divorced or separated women experienced modest increases in net worth ($43,209 or 37.1%), whereas Hispanic women experienced a loss of $3,227 or 7.7%. These patterns are similar to the comparison group of divorced The Gerontologist Table 3. 1992 Household Income and Change in Household Income Between 1992 and 2000 (Year 1999 Dollars) Type of Marital Status Change Non-Hispanic White Black Hispanic 72,113 31,966 56,793 27,660 49,670 16,252 40,539 25,406 40,157 12,072 40,614 13,514 1992 income Married T1 and T2 Widowed T1 and T2 Widowed between T1 and T2 Divorced or separated T1 and T2 Change in income between 1992 and 2000 Married T1 and T2 Widowed T1 and T2 Widowed between T1 and T2 Divorced or separated at T1 and T2 À1,150 À5,164 À22,534 1,124 (À1.6%) (À16.1%) (À39.7%) (3.0%) À6,737 À1,452 À22,121 À4,830 (À13.6%) (À8.9%) (À54.6%) (À19.0%) À6,158 4,813 À10,933 554 (À15.3%) (39.9%) (À27.1%) (4.1%) Notes: T1 = Time 1; T2 = Time 2. Information is from the Health and Retirement Study Public Use Dataset, 1992–2000 (2002). women. Among women who became widowed between the two interviews, all three groups experience losses in net worth, and the amount is largest for Hispanic women, $113,401, compared with $34,545 for non-Hispanic White widows and $24,826 for Black widows. In Table 5 we present ordinary least square regressions of total household income for 1999 (Models 1 and 2) and change in total household income between 1992 and 1999 (Models 3 and 4) on the various marital status comparisons and demographic controls. Models 2 and 4 include multiplicative interaction terms based on race and Hispanic ethnicity and marital status between the two waves. Non-Hispanic White women who were married at both Time 1 and Time 2 serve as the reference category. Model 1 explains 34% of the variance and demonstrates that women who were divorced or widowed at both Time 1 and Time 2 and those who became widowed between Time 1 and Time 2 had over $20,000 less in household income than did women who were married at both Time 1 and Time 2, the reference category. In addition, this model has significant independent negative effects associated with both race and Hispanic ethnicity. In Model 2 the interaction terms based on race, Hispanic ethnicity, and marital status produce only one significant interaction, that between Black and divorce, at both Time 1 and Time 2. Models 3 and 4, the last two columns of Table 5, predict the net total household income at Wave 5 net of Time 1 income. Model 4 includes the interaction terms between race or ethnicity and marital status. Model 3 reveals that, compared with women who were married at both Time 1 and Time 2, women in all other marital statuses suffered losses in total household income between Time 1 and Time 2. Women who became widows between the two waves suffered the largest decline, $18,062. In Model 4, none of the interaction terms are significant. The models in Table 6 are similar to those in Table 5 except they predict the influence of changes in marital status between Time 1 and Time 2 on household net worth at Time 2 (in Year 2000 dollars). Model 1 explains 26% of the variation in household net worth in 2000 and reveals that compared with women who were married at both Time 1 and Time 2, women in all other marital categories experienced large losses in net worth. Women who were divorced at both Time 1 and Time 2 reported the largest loss ($159,743). This model also reveals large independent losses in net worth for both Hispanic and black women. In Model 2 four interaction terms are significant, Table 4. 1992 Household Assets and Change in Household Assets Between 1992 and 2000 (Year 2000 Dollars) Type of Marital Status Change 1992 Net Assets Married T1 and T2 Widowed T1 and T2 Widowed between T1 and T2 Divorced or separated T1 and T2 Change in net assets between 1992 and 2000 Married T1 and T2 Widowed T1 and T2 Widowed between T1 and T2 Divorced or separated T1 and T2 Non-Hispanic White Black Hispanic 396,242 174,073 282,223 116,354 108,450 46,457 40,539 61,997 201,779 79,618 238,403 41,789 174,261 10,852 À34,545 43,209 (44%) (6.2%) (À12.2%) (37.1%) 38,318 À838 À24,826 490 (35.3%) (À2.0%) (À61.2%) (7.9%) À25,605 À18,027 À113,401 À3,227 (À12.7%) (À22.6%) (À47.6%) (À7.7%) Notes: T1 = Time 1; T2 = Time 2. Information is from the Health and Retirement Study Public Use Dataset, 1992–2000 (2002). Vol. 47, No. 2, 2007 229 Table 5. Unweighted Linear Regressions Predicting Household Income at Time 2 (Year 1999 Dollars) Total Household Income Comparison or Control Model 1 Model 2 Model 3 Model 4 Intercept Net income T1 Marital change Widowed T1 and T2 Widowed between T1 and T2 Divorced or separated T1 and T2 103,847*** — 104,093*** — 69,508*** 0.39*** 69,774*** 0.39*** À22,149*** À21,221*** À24,016*** À23,898*** À22,144*** À26,729*** À11,129*** À17,417*** À12,130*** À12,701*** À18,062*** À13,410*** À990*** À985*** À766*** À765*** Education Less than high school More than high school À10,919*** 15,433*** À10,830*** 15,383*** À6,328*** 8,563*** À6,309*** 8,535*** Race or ethnicity Hispanic Black À11,212*** À9,043*** À13,499*** À11,818*** À6,145*** À5,950*** À7,709*** À7,479*** — — — — — — 3,950 1,761 3,862 6,485 5,806 3,621 0.46 3,814 0.46 3,814 Demographics Age Interaction terms Black 3 Widowed Black 3 Widowed T1 and T2 Black 3 Divorced or separated Hispanic 3 Widowed Hispanic 3 Widowed T1 and T2 Hispanic 3 Divorced or Separated Adjusted R2 N — — — — — 5,058 3,663 8,241** 7,928 6,262 7,350 0.34 3,814 0.34 3,814 Notes: Unweighted linear regressions predicting household income at Time 2 (Year 1999 dollars) are truncated at the 90th percentile (income = $109,140). For marital change, reference = married; for education, reference = high school; for race or ethnicity, reference = Non-Hispanic White. T1 = Time 1; T2 = Time 2. Information is from the Health and Retirement Study Public Use Dataset, 1992–2000 (2002). *p .05; **p .01; ***p .0001. those for (1) black women who were divorced at both Time 1 and Time 2, who lost; (2) black women who were widowed at both Time 1 and Time 2; (3) black women who were widowed between Time 1 and Time 2; and (4) Hispanic women who were divorced at both Time 1 and Time 2. All of the interactions are positive and reduce the net losses associated with marital status, race and ethnicity. In this model the main effects for race and ethnicity, as well as for marital status, decrease. Between Time 1 and Time 2, the loss associated with remaining divorced for black women is $202,372 plus an additional loss of $185,599 associated with being black, minus $131,402 for the interaction, resulting in a total penalty of $256,569 in net worth. At Time 2, for divorced Hispanic women who did not remarry the penalty is $244,038; for black women who became widowed the penalty is $228,679; and for black widows who did not remarry the penalty is $244,525. Model 3 controls for assets at Time 1. It reveals significant losses in net worth associated with all of the marital statuses other than married at both Time 1 and Time 2. As in the previous models black and Hispanic women experience large losses in net worth in all marital statuses. Model 4, introduces the 230 interactions between race, Hispanic ethnicity, and marital status changes between Time 1 and Time 2. It reveals three significant effects, (a) black women who were divorced at both Time 1 and Time 2; (b) black women who were widowed at both Time 1 and Time 2; and (c) Hispanic women who were divorced at both Time 1 and Time 2. Again these coefficients are positive and reduce the net impacts of marital status, race, and ethnicity but do not eliminate the penalty suffered by black and Hispanic widowed and divorced women. For non-Hispanic black women, the effect of being continuously divorced (as opposed to being continuously married) is less than that for nonHispanic white women (À$14,005 vs À$62,023). Hispanic women who remain divorced lose less than other groups ($6,100). Furthermore, among non-Hispanic black women the effect of being continuously widowed on assets are less than that those for non-Hispanic white woman ($13,556 vs $64,207). Among Hispanic women, the effect of being continuously widowed on assets is not significant. Additionally, non-Hispanic black women who were widowed at both times experienced a loss of $86,323. Combining the coefficients for marital status, race or ethnicity, and the significant inThe Gerontologist Table 6. Unweighted Linear Regressions Predicting Household Net Worth at Time 2 (Year 2000 Dollars) Total Net Household Assets Comparison or Control Intercept Net Assets T1 Marital change Widowed T1 and T2 Widowed between T1 and T2 Divorced or separated T1 and T2 Demographics Age Education Less than high school More than high school Race or ethnicity Hispanic Black Interaction terms Black 3 Widowed Black 3 Widowed T1 and T2 Black 3 Divorced or separated Hispanic 3 Widowed Hispanic 3 Widowed T1 and T2 Hispanic 3 Divorced or separated Adjusted R2 N Model 1 Model 2 255,727.00*** — 260,539.00*** — 252,129.00*** 0.92*** 254,478.00*** 0.91*** À122,372*** À101,299*** À159,743*** À148,959*** À117,946*** À202,372*** À47,295*** À58,956*** À45,181*** À63,554*** À67,776*** À61,985*** À2,849** À2,804** 899 995 Model 3 Model 4 À90,632*** 112,951*** À89,468*** 112,141*** À29,238*** 34,561*** À29,159*** 34,406*** À126,109*** À137,832*** À158,250*** À185,599*** À62,245*** À48,721*** À79,557*** À69,282*** — — — — — — 90,033** 74,866** 131,402*** 83,744 92,179 116,584** — — — — — — 50,034* 32,015 47,507* 43,239 60,392 55,428* 0.26 3,814 0.27 3,814 0.61 3,814 0.61 3,814 Notes: Unweighted linear regressions predicting household net worth at Time 2 (Year 2000 dollars) are truncated at the 90th percentile (assets = $591,914). For marital change, reference = married; for education, reference = high school; for race or ethnicity, reference = Non-Hispanic White. T1 = Time 1; T2 = Time 2. Information is from the Health and Retirement Study Public Use Dataset, 1992–2000 (2002). *p .05; **p .01; ***p .0001. teraction terms, we find that the penalty for black women who were divorced at both Time 1 and Time 2 is $83,760 and for Hispanics who were divorced at both interviews the penalty is $86,114. For black women who remain widowed between Time 1 and Time 2 the penalty is $82,802. Discussion Previous research and our data indicate clearly that a woman’s financial situation in the preretirement years and later is determined by her marital history, as well as her current marital status (e.g., Holden & Smock, 1991; Weir et al., 2002). Although the results of the analysis reveal clear economic penalties associated with all marital statuses other than continuously married, our primary focus was on widowhood, and the data unambiguously show that widows, including those who were widowed before our initial interview and those who became widowed during the study period, experienced significant and substantial losses in income and net worth. For minority women, this decline began from a lower point because minority couples in general are less able than non-Hispanic White couples to Vol. 47, No. 2, 2007 231 accumulate assets during their working years. The picture that emerges, then, is complicated, because the absolute loss in income resulting from widowhood can be larger for women whose husbands had substantial incomes than for women whose husbands made less. If, as is often the case, higher income results in substantial assets, the widow can draw upon those to ensure her economic security in middle through early old ages. Fewer Black and Hispanic women than non-Hispanic White women find themselves in such a favorable situation. Consistent with other HRS research (cf. Honig, 2000), our research finds that this differential economic vulnerability in late life arises from a set of factors that operate over the life course to increase the risk of poverty for minority group women in later life. These include lower levels of human capital among minority individuals, less stable employment in low-wage jobs, low earnings for both husbands and wives, impaired asset accumulation, and more frequent marital disruption. Women who have not had careers or worked in jobs in which they were able to vest a pension are almost totally dependent on their husband’s income and on limited community assets. Even if their absolute drop in income is less than that of more affluent non-Hispanic White women, with few assets to liquidate or borrow against, they can end up far worse off financially. The fact that a growing proportion of women, and especially minority group women, find themselves alone and with few economic resources has important implications for public policy, as well as for norms concerning paid work and family for women. Marriage has simply ceased to be the guarantee of retirement-age security for women that it might have been in earlier marriage systems. The clear negative financial impact of marital disruption of all sorts, including widowhood, means that women must rely more on their own income and asset accumulation, and public policy must focus more on individuals and abandon the older malebreadwinner model in which the couple is the focus of old-age support. As Harrington Meyer (2004, p. 7) and her colleagues note, with each successive cohort we can expect ‘‘the proportion of women who will not be eligible for widow benefits [to increase] modestly for Whites and Hispanics, but dramatically for Blacks. The growing race gap in marital rates means that older Black women will be particularly unlikely to qualify for these benefits.’’ The result may be greater income inequalities related to race and ethnicity in the future. Johnson (1999, p. 6), for example, suggests the following: Comparisons of the size and composition of assets held by married couples approaching retirement in 1992 with those held by couples in 1969 indicate that wives who become widowed today would be substantially better off than widows in the past. The improvement in the economic security of wives appears to come primarily from increases in private pensions, conventional assets, and women’s earnings. As our data and others show, these privileged widows are more likely to be non-Hispanic White than minority group members. As Johnson notes, financial security in old age depends on multiple income streams. It requires an adequate income and enough assets with which to purchase long-term care and other age-based necessities. Given the labor force and earning disadvantages faced by Black and Hispanic men and women, these groups will remain at an elevated risk of inadequate financial security in old age. In the future, racial and Hispanic differences in income and wealth may be exacerbated by differential marital disruption. Future public policy requires new research to assess the potential magnitude of the effect of different types of marital transitions on the financial well-being of older minority women. Many of these women will reach old age with little income, no savings or retirement plans, and very few assets. In such a world, the difference in the financial situation of the advantaged and the disadvantaged could increase greatly. In addition, future studies should also examine never married women, because 232 they will comprise an ever greater proportion of women’s marital status and they are an important component of understanding Black women’s experiences in particular. Policy Implications These data demonstrate clear financial vulnerabilities among older minority group women. Although marriage provides some economic stability, regardless of marital status, Black and Hispanic women have lower earnings, fewer assets, and less adequate pension coverage than do non-Hispanic White women of similar age. In addition, the results indicate that the situation may be getting worse for minority women for whom marital disruption is increasing and for whom marriage has never been a route to economic security. Given the limited work opportunities available to many of these women, their own employment is unlikely to compensate for the economic weakness of marriage. Few of the service sector jobs for which they qualify are likely to provide even minimal benefit packages that include a retirement plan. Social Security was based on a male-breadwinner model that assumes that a woman’s retirement security would be ensured by marriage (Herd, 2005). As only one pillar of the theoretical retirement income scheme, Social Security was intended to supplement private pensions and individual savings and not serve as the sole source of retirement income that it has become for many older minority Americans. The fact that Black and Hispanic men have faced serious barriers to the accumulation of assets and have often spent their working years in jobs with low wages and no retirement plans made this model less appropriate for minority women, even prior to the decline in marriage that we have witnessed in recent decades (Herd, 2006). The irrelevance of the male-breadwinner model for many poor women, combined with their own limited economic opportunities, means that they are highly vulnerable to poverty in old age (Sevak, Weir, & Willis, 2003–2004). Despite its shortcomings in terms of providing an adequate income, then, Social Security represents the major income source for a large fraction of poor women. The centrality of Social Security is made clearer by the fact that the average woman gains relatively more than the average male from Social Security, because the benefit structure favors dependents and low-wage earners, most of whom are women or children. The evidence we have presented leads to the conclusion that the male-breadwinner model of retirement security should be reconsidered, especially for minority women. Regardless of their race or ethnicity or even their social class, future cohorts of working-age women will be forced to take responsibility for their own welfare and retirement The Gerontologist incomes. Some will clearly be in a better position to do so than others. Increasingly, concerns for a career and income generation are as relevant for women as for men. Because of the labor force disadvantages they face, though, the situation of low-wage service sector workers and the unemployed and underemployed requires special attention. For these women, employment will never serve as a route out of poverty or ensure an adequate retirement income. For this reason, any changes to be made to the present Social Security system will have to take these vulnerabilities into account. The long-term fiscal health of the Social Security system will require some retrenchment, including higher payroll taxes and reduced benefits. Proposals aimed at cutting benefits as opposed to generating revenue would have particularly serious implications for poor and minority women. For better or worse, the situation of these women argues for greater means testing of benefit outlays. Such a move, of course, runs the risk of transforming Social Security into more of a welfare program with the potential stigma that welfare entails. In addition to the inevitable reforms to Social Security that we will see in the future, efforts to improve access to private pensions and increased personal savings and investment by low-wage women employees are clearly warranted, even if the possibilities remain limited (Herd, 2005). The partial privatization of Social Security has been offered as a means of increasing individual returns and responsibilities. Unfortunately, such reforms would in all likelihood not improve the old age security of low-wage workers or those with noncontinuous work histories. The privatization of Social Security without other reforms to ensure an adequate income to those without adequate savings would only increase the risk for minority paid workers. Low-income households, especially female-headed families, have limited experience with private investments (Choudhury, 2001–2002). Privatization of Social Security retirement presents additional serious disadvantages for women, because women tend to be more conservative investors than men and such conservative choices could easily result in inadequate returns (Trout, 1997). What remains clear is that the old-age economic security of women is greatly affected by race and Hispanic ethnicity. For many African American and Hispanic women, restricted employment opportunities and low educational levels make retirement planning irrelevant, and for many of these women marriage is no assurance of security. The restricted earnings capacity of many Black and Hispanic husbands means that the couple does not have the opportunity to accumulate assets. In such cases, when a husband dies he leaves his wife little wealth and no long-term financial security. However, even for these women the problem is compounded by the increasing risks associated with different and unVol. 47, No. 2, 2007 233 equal pathways to old age economic security, including marital disruption earlier in life. Policies intended to ensure the economic welfare of women approaching retirement must address persistent inequities in educational and labor market opportunities associated with race and ethnicity. Toward that end, research must examine lifelong processes and challenges associated with economic insecurity and their cumulative impact by retirement age (Hatch, 1999). General reforms to public and private retirement systems may well increase the choices available to affluent paid workers, but they are not likely to increase the economic security of low-wage or minority paid workers. Targeted programs that focus on the most vulnerable individuals could more directly address the unique needs of older Black and Hispanic women, but again, like targeted programs in general, they run the risk of stigmatizing the recipients and of generating political opposition. Whatever the future holds, public policy focused on old age economic security cannot ignore the complex social and economic vulnerabilities associated with minority group status and gender. References Angel, R. J., & Angel, J. L. (1996). The extent of private and public health insurance coverage among adult Hispanics. The Gerontologist, 36, 332– 340. Bound, J., Schoenbaum, M., & Waidmann, T. (1996). Race differences in labor force attachment and disability status. The Gerontologist, 36, 311–321. Choudhury, S. (2001–2002). Racial and ethnic differences in wealth and asset choices. Social Security Bulletin, 64, 1–15. Crystal, S., & Shea, D. (Eds.) (2003). Introduction: Cumulative Advantage, Public policy, and Inequality in Later Life. In S. Crystal and D. Shea (Eds.), Annual Review of Gerontology and Geriatrics (pp. 1–13). New York: Springer. Fullerton, H. N., & Toosi, M. (2001). Labor force projections to 2010: Steady growth and changing composition. Monthly Labor Review, 124, 21–38. Harrington Meyer, M., Wolf, D. A., & Himes, C. L. (2004). Linking benefits to marital status: Race and diminishing access to Social Security spouse and widow benefits in the U.S. (Center for Retirement Research Working Paper Series, WP2004-05). Chestnut Hill, MA: Boston College. Harrington Meyer, M., Wolf, D. A., & Himes, C. L. (2006). Declining eligibility for Social Security spouse and widow benefits in the U.S? Research on Aging, 28, 240–260. Hatch, L. (1999). Beyond gender differences: Adaptation to aging in life course perspective. Amityville, NY: Baywood. He, W., Sengupta, M., Velkoff, V., & Debarros, K. (2005). 65þ in the United States: 2005 (Current Population Reports, Series P23–209). Washington, DC: U.S. Government Printing Office. Health and Retirement Study. (2002). Background information. Available online at http://hrsonline.isr.umich.edu/intro/sho_intro.php?hfyle=uinfo Health and Retirement Study public use dataset, 1992–2000. (2002). Ann Arbor, MI: University of Michigan (with funding from the National Institute on Aging, Grant NIA U01AG009740). Heeringa, S. G., & Conner, J. (1995). Technical Description of Health and Retirement Study Sample Design (HRS/AHEAD Documentation Report DR-002). Ann Arbor, MI: University of Michigan, Institute for Social Research. Herd, P. (2005). Ensuring a minimum: Social Security reform and women. The Gerontologist, 45, 12–25. Herd, P. (2006). Crediting care or marriage? Reforming Social Security family benefits. Journal of Gerontology: Social Sciences, 61B, S24–S34. Holden, K., & Kuo, D. (1996). Complex marital histories and economic wellbeing: The continuing legacy of divorce and widowhood as the HRS cohort approaches retirement. The Gerontologist, 36, 383–390. Holden, K. C., & Smeeding, T. M. (1990). The poor, the rich, and the insecure caught in between. The Milbank Quarterly, 68, 191–219. Holden, K. C., & Smock, P. J. (1991). The economic costs of marital dissolution: Why do women bear a disproportionate cost? Annual Review of Sociology, 17, 51–78. Honig, M. (2000). Minorities face retirement: Worklife disparities repeated? In B. Hammond, O. S. Mitchell, & A. Rappaport (Eds.), Forecasting retirement needs and retirement wealth (pp. 235–252). Philadelphia: University of Pennsylvania Press. Institute of Medicine. (2001). Coverage matters: Insurance and health care. Washington, DC: National Academy Press. Jaccard, J., Turrisi, R., & Wan, C. K. (1990). Interaction effects in multiple regression. Newbury Park, CA: Sage. Jackson, J. S., Lockery, S. A., & Juster, F. T. (1996). Introduction: Health and retirement among ethnic and racial minority groups. The Gerontologist, 36, 282–284. Jenkins, C. L. (2003). Widows and divorcees in later life: On their own again. Binghamton, NY: Haworth Press. Johnson, R. W. (1999). The gender gap in pension wealth: Is women’s progress in the labor market equalizing retirement benefits? (The Retirement Project, Brief Series No. 1). Washington, DC: Urban Institute. Johnson, R. W., Sambamoorthi, U., & Crystal, S. (2003). Gender differences in pension wealth and their impact on late-life inequality. In S. Crystal and D. Shea (Eds.), Annual Review of Gerontology and Geriatrics (pp. 116–137). New York: Springer. Kijakazi, K. (2001, June). Women’s retirement income: The case for improving Supplemental Security Income. Paper presented at the Institute for Women’s Policy Research Annual Conference. Available at http://www.cbpp.org/6-8-01socsec.htm Kreider, R. M., & Fields, J. M. (2002). Number, timing, and duration of marriages and divorces: 1996. Washington, DC: U.S. Census Bureau. Available at http://www.census.gov/prod/2002pubs/p70-80.pdf. Lichter, D. R., LeClere, F. B., & McLaughlin, D. K. (1991). Local marriage markets and the marital behavior of Black and White women. American Journal of Sociology, 96, 843–867. Lupton, J. P., & Smith, J. P. (2003). Marriage, assets, and savings. In S. Grossbard-Shecht (Ed.), Marriage and the economy: Theory and evidence from advanced industrial societies (pp. 129–152). Cambridge, England: Cambridge University Press. Norton, A. J., & Miller, L. F. (1992). Marriage, divorce, and remarriage in the 1990s (Current Population Reports, Series P23-180). Washington, DC: U.S. Census Bureau. Ozawa, M. N. (1995). The economic status of vulnerable older women. Social Work, 40, 323–331. Sevak, P., Weir, D. R., & Willis, R. J. (2003–2004). The economic consequences of a husband’s death: Evidence from the HRS and AHEAD. Social Security Bulletin, 65. Available at http://www.ssa.gov/policy/docs/ssb/v65n3/ v65n3p31.html RAND HRS Data, Version F. (2004, April). Santa Monica, CA: RAND Center for the Study of Aging (with funding from the National Institute on Aging and the Social Security Administration). Rubinstein, R. (1987). Never married elderly as a social type: Reevaluating some images. The Gerontologist, 27, 108–113. Shapiro, T. M. (2003). The hidden cost of being non-Hispanic Black: How wealth perpetuates inequality. New York: Oxford University Press. Smith, J. P. (1995). Racial and ethnic differences in wealth in the Health and Retirement Study. The Journal of Human Resources, 30, S158–S183. Smith, J. P., & Kington, R. (1997). Demographic and economic correlates of health in old age. Demography, 34, 159–170. Smock, P. (1993). The economic costs of marital disruption for young women over the past two decades. Demography, 30, 353–371. Social Security Administration. (2002). Annual Statistical Supplement (Tables 5.A6, 5.A1.2, 5.A1.3, 5.A1.6, and 5.G1). Washington, DC: U.S. Government Printing Office. Trout, P. S. (1997). Adequacy and equity of Social Security (Report of the 1994–1996 Advisory Council on Social Security, Vol. II, Findings and Recommendations). Washington, DC: U.S. Government Printing Office. U.S. Census Bureau. (2005). Marital history for people 15 years and over, by age, sex, race, and Hispanic origin: 2001 (Table 1). Available at http:// www.census.gov/population/socdemo/marr-div/p23-180/p23-180.pdf U.S. Census Bureau. (2004). Age and sex of all people, family members and unrelated individuals iterated by income-to-poverty ratio and race (Annual Demographic Survey: March Supplement; POV01). Available at http://pubdb3.census.gov/macro/032005/pov/new01_125_16.htm U.S. Department of Labor. (2004). Highlights of women’s earnings in 2003: Report 978. Washington, DC: Bureau of Labor Statistics. U.S. Government Accounting Office. (1998). Women and Social Security reform (Report GAO/HEHS-98-42). Washington, DC: U.S. General Accounting Office. Waite, L. J., & Gallagher, M. (2000). The case for marriage: Why married people are happier, healthier, and better off financially. New York: Doubleday. Weir, D., Willis, R., & Sevak, P. (2002, May). The economic consequences 234 of a husband’s death: Evidence from the HRS and AHEAD. Paper presented at the annual conference for the Retirement Research Consortium, Washington, DC. Wilmoth, J. M., & Koso, G. (2002). Does marital history matter? The effect of marital status on wealth outcomes among pre-retirement age adults. Journal of Marriage and Family, 64, 254–268. Willson, A. E. (2003). Race and women’s income trajectories: Employment, marriage, and income security over the life course. Social Problems, 50, 87–110. Willson, A. E., & Hardy, M. A. (2002). Racial disparities in income security for a cohort of aging American women. Social Forces, 80, 1283–1306. Wise, D. A. (Ed.). (1996). Advances in the economics of aging. Chicago: The University of Chicago Press. Wolff, E. N. (2003). The impact of gifts and bequests on the distribution of wealth. In A. H. Munnell & A. Sunden (Eds.), Death and dollars: The role of gifts and bequests in America (pp. 345–388). Washington, DC: Brookings Institution Press. Zick, C. D., & Holden, K. (2000). An assessment of the wealth holdings of recent widows. Journal of Gerontology: Social Sciences, 55B, S90–S97. Zick, C. D., & Smith, K. R. (1991). Patterns of economic change surrounding the death of a spouse. Journal of Gerontology: Social Sciences, 46, S310–S320. Received May 8, 2006 Accepted November 8, 2006 Decision Editor: Linda S. Noelker, PhD Endnotes 1. A substantial proportion of elderly widows (29%) rely on Social Security alone to survive (Weir, Willis, & Sevak, 2002). In fact, the Social Security program reduces the percentage of widows in poverty by more than two thirds, from 62% to 20% (Kijakazi, 2001). Without Social Security, the total poverty rate among aged beneficiaries would be 67% for Blacks and 47% for Whites (Social Security Administration, 2002). Social Security accounts for almost two thirds of total retirement income among Black and Hispanic elderly couples (Kijakazi). As a result of these cumulative disadvantages, approximately 39.8% of Black and 32.6% of Hispanic women over the age of 65 receive incomes of less than 125% of poverty level in 2004 (U.S. Census Bureau, 2004). 2. Among this group, ethnic differences were apparent in cases lost to follow-up or mortality, with non-Hispanic White women reporting somewhat lower household income, but higher household assets, than those non-Hispanic White women who remained in the sample. The opposite is true of Hispanic women—those who did not participate in the survey in 2000 had slightly higher income but significantly lower household assets than those Hispanic women who were not lost to attrition or mortality. However, non-Hispanic Black women not interviewed in 2000 had both lower income and total household assets. 3. The HRS oversampled Florida residents, because of the Congressional Appropriation language for the study, which specified that special attention be given in the HRS to areas with ‘‘high densities and numbers of older populations’’ (Health and Retirement Study, 2002). The Gerontologist