The Implications of Marital History Change on Women's Eligibility for Social Security Wife and Widow Benefits, 1990–2009

Introduction

Over the six decades following World War II, major sociodemographic changes occurred in the American family. An important research and public policy subject is to document those changes and their implications for retirement outcomes, particularly for the baby boom generation now entering retirement. In this article, we examine the change in women's marital patterns at different stages of the life course over the past two decades and its implications for women's eligibility status for Social Security spouse and survivor benefits at retirement age.

Women's financial circumstance in old age is a longstanding concern among policymakers and researchers (Ekerdt 2010; Favreault and Steuerle 2007; Government Accountability Office 2007; Holden and Fontes 2009; Lusardi and Mitchell 2008; Weaver 1997, 2010). Although the retirement security of women has improved significantly over the past 30 years, women have higher poverty rates in old age than men, in large part because they earn less over a lifetime and live longer (Blau and Kahn 2006; Weinberg 2007; Bureau of Labor Statistics 2010). Greater longevity also means that many women will spend some time during their life course as widows. Estimates from the Current Population Survey show that the poverty rate for women aged 65 or older in 2008 was almost double (11.9 percent) that of men (6.7 percent), with the unmarried group being particularly vulnerable to poverty (16.9 percent of single women compared with 5.0 percent of married women); see SSA (2010, Table 11.1).

A growing body of research has shown a connection between women's lifetime marital experiences and retirement outcomes (Couch and others 2011; Wilmoth and Koso 2002; Tamborini, Iams, and Whitman 2009; Zissimopoulos, Karney, and Rauer 2008). The main purpose of this study is to determine the implications of changing marital patterns on women's eligibility for Social Security marriage-based benefits at retirement. The Social Security program not only provides women with income as retired or disabled workers but also as spouses or widows of insured workers. Generally, a person must be currently married, widowed, or divorced from a 10-year marriage to qualify for a spouse or widow(er) benefit. This link between marital history and benefit eligibility means that the distribution of women who have the option of claiming spouse or widow benefits at retirement is subject to fundamental changes in marriage behavior.

In this article, we assess the evolution of women's potential eligibility for Social Security spouse and widow benefits over the past two decades. We draw from the 1990 and the recently released 2009 Marital History Modules to the Census Bureau's Survey of Income and Program Participation (SIPP). Our analysis compares the marital histories of women in their thirties, forties, and fifties in 2009 with similarly aged women in 1990 using descriptive and regression methods. Building on recent studies that have found declining eligibility for spouse and widow benefits because of changes in marriage patterns (Harrington Meyer, Wolf, and Himes 2006; Tamborini, Iams, and Whitman 2009), we consider how current marital trends may reshape women's potential eligibility for spouse or widow benefits and how those patterns vary across race and ethnic subgroups.

The results contribute to the literature in several ways. First, by using more recent nationally representative data than previously available, we provide up-to-date evidence indicating how women's eligibility for spouse and widow benefits is changing because of changes in marital patterns. Second, we offer insight into the early life marital experiences of the generation X (GenX) cohorts.¹ Research has just begun to consider family trends in postboomer cohorts (Tamborini and Iams 2011), and no studies to date have examined marital trends in GenX cohorts as they relate to Social Security marriage-based benefits.

A third contribution of the present study is that it advances understanding of how eligibility for Social Security marriage-based benefits varies across race and ethnic subgroups. Although the Social Security program is gender and race neutral and handles persons with identical earnings or marital histories in the same way, racial differences in marital patterns can result in differences in eligibility for marriage-based Social Security benefits. This article addresses whether the racial differentials in potential benefit eligibility resulting from differences in marital history have continued from those observed through 1995 by Harrington Meyer, Wolf, and Himes (2006) and those observed through 2004 by Tamborini, Iams, and Whitman (2009). We also estimate the eligibility patterns of Hispanic women, by nativity, which is a useful focus given the lack of research on this issue combined with their growing share of the population in the United States (US).

The next section elaborates on the background guiding this study. We then present the data and methods, followed by the empirical results. Finally, a conclusion is provided with a summary of the findings and their broader implications.

Background

Social Security benefits are an important source of retirement income for women. According to the Current Population Survey, Social Security made up at least half of the family income for 58 percent of women aged 65 or older in 2008 (SSA 2010, Table 9.B1). Benefits are particularly important to the financial security of the single, elderly population. Among women aged 65 or older in beneficiary families in 2008, Social Security accounted for at least 90 percent of the family income of 35 percent of unmarried women compared with 21 percent of married women (ibid., Table 9.B3).

Originally, Social Security benefits were based on a worker's own lifetime earnings covered by the program. In the 1939 Social Security Amendments, Congress established wife and widow benefits to provide monthly payments to qualified spouses and survivors of insured workers (Martin and Weaver 2005). As Berkowitz (2002) noted, the design of Social Security marriage-based benefits was essentially based on a family structure common to the Depression era, when the Social Security program began; see also Cherlin (2009, 69). Typical to that era would be a one-earner family with an employed husband supporting a wife and children in a marriage that continued until the death of one of the spouses. Over time, those benefits were extended to divorced spouses and divorced surviving spouses with 10 years in the marriage and to children.²

Potential eligibility for Social Security spouse or widow(er) benefits reflects a person's past and present marital status. The amount reflects the work record of that person in relation to the record of their current or former spouse(s). In other words, the amount of Social Security spouse or widow(er) benefits is a function of the lifetime earnings of workers and their spouses (or former spouses), while eligibility is based on marital history. Table 1 provides a brief overview of current marital requirements.

Under current law, an aged individual who is currently married to an insured worker is eligible for a spouse benefit equal to a maximum of half of his or her spouse's primary insurance amount (PIA), beginning at the early retirement age of 62.³ Eligible spouses generally must be married to the retiring worker for at least 1 continuous year before they can receive benefits based on their entitled spouse's record. An aged divorced person may be eligible for a divorced spouse benefit equal to up to half of his or her ex-spouse's PIA if the marriage was 10 years (120 months) in duration and the ex-spouse is alive. If the aged person qualifies for his or her own retired-worker benefit and a spouse or divorced spouse benefit, that person will receive the higher amount of the two benefits. If the spouse benefit exceeds one's own retired-worker benefit, that person becomes dually entitled and will receive his or her full retired-worker benefit with the spouse benefit supplementing the difference. If the retired-worker benefit is higher than the spouse benefit, then that person will receive only his or her own worker benefit.

Social Security also provides widow(er) benefits. If a surviving spouse begins to receive benefits at the full retirement age, the widow(er) benefit can equal up to 100 percent of a deceased spouse's PIA. A reduced benefit (from 71 percent to 99 percent of the deceased worker's PIA) is available as early as age 60 (age 50 if the survivor is disabled).⁴ Divorced persons with a 10-year marriage may qualify for a surviving divorced spouse benefit if their ex-spouse dies. If the widow(er) benefit exceeds a person's own retired-worker benefit, that person is dually entitled and will receive his or her full retired-worker benefit with the widow(er) benefit supplementing the difference.⁵ If the retired-worker benefit is larger than the widow(er) benefit, then that person would be eligible for only a retired-worker benefit.

Over the past five decades, women increasingly receive their own earned benefits, but spouse and survivor benefits still constitute a major part of the Social Security program. To illustrate, Chart 1 compares the distribution of benefits among aged female Social Security beneficiaries from 1960 through 2009. Several noteworthy trends emerge. First, the majority of aged women receive at least part of their Social Security benefit as a wife or widow. In 1960, around 61 percent (4.1 million) of female beneficiaries aged 62 or older received some type of benefit as a wife or widow, and in 2009 this figure was 55 percent (12.9 million). Second, the share of retired-worker-only beneficiaries remained relatively constant over the period (at about 40 percent), until recently increasing in 2009 (to about 45 percent).

Third, there has been a shift away from women's entitlement based solely on their husband's record to dual entitlement—retired-worker benefits that are supplemented by a higher auxiliary spouse or survivor benefit. Thus, the share of dually entitled women rose from 5 percent in 1960 to 20 percent in 1985 and to 27 percent in 2009. Rising female labor force participation has largely driven this trend. Because women usually have lower lifetime earnings than their spouses, partly because of family responsibilities (for example, child rearing or caretaking of other family members), many women who qualify for their own retired-worker benefit also are dually entitled to a spouse, or more likely, a widow benefit (Butrica, Iams, and Sandell 1999; Kingson and O'Grady-LeShane 1993).

Marital Trends, Race and Ethnicity, and Social Security

The present study focuses on the link between marital trends and women's eligibility for marriage-based Social Security benefits, while recognizing that women's lifetime workforce attachment will also influence the benefits they receive. A focus on marital trends is valuable given that fundamental changes in marriage behavior in the United States over the past decades will have important implications for women's retirement experience, including their Social Security benefits (Couch and others 2011).

Briefly, recent analysis of marital history data has shown declining marriage rates after the mid-1970s to "the lowest level in recorded history" by 2004 (Stevenson and Wolfers 2007, 29; Goldstein and Kenney 2001; Tamborini 2007). Less than 90 percent of women were likely to ever marry among the late-1960s birth cohort, unlike women born in the late 1940s (Kreider and Ellis 2011, 6). The divorce rate increased dramatically in the 1970s, peaked in 1981, and gradually decreased over the next 25 years (Stevenson and Wolfers 2007). Together, these trends have resulted in a sharp reduction in the percentage of currently married women between 1980 and 2009 (Kreider and Ellis 2011).

At the same time, empirical research has shown increasing heterogeneity in marital outcomes by race and ethnicity. Black women have had a "lower and slower" entry into marriage than white women, a reversal of previous patterns (Goldstein and Kenney 2001; Kreider and Ellis 2011, Table 1; Stevenson and Wolfers 2007). Research has also found black/white differentials in marital disruption (Sweeney and Phillips 2004). The trend for Hispanic women has been mixed; generally, the marital experiences of Hispanic women has been characterized as somewhere between those of black and white women (McNamee and Raley 2011). While Hispanic women's marriage rates were similar to those of white women, their divorce rates were closer to those of black women, depending on the cohort. Importantly, research has indicated substantial differences by nativity, with marriage rates being higher among foreign-born than US-born Hispanics (McNamee and Raley 2011; Oropesa and Landale 2004).⁶

Recent empirical work by Tamborini, Iams, and Whitman (2009) and Harrington Meyer, Wolf, and Himes (2006) showed that changes in marriage patterns among recent birth cohorts would lead to a decline in women's eligibility for spouse and widow benefits when those cohorts reach retirement age. A key finding in both analyses was that lower marriage rates among black women would result in sharper declines in their potential eligibility for spouse or widow benefits compared with white women. Harrington Meyer, Wolf, and Himes's (2006) projections also showed fairly similar trends in potential eligibility between white and Hispanic women using Current Population Survey marital history data through 1995.

Put together, prior research leads us to expect declines in potential eligibility for spouse or widow benefits among recent cohorts of women as a result of changing marital patterns. We also expect that the downward trend is sharper for black women than white women because of the widening racial gap in marriage rates among recent cohorts. Hispanic women are expected to show mixed patterns, with substantial differences by nativity.

Methods

The data for this analysis are from the 1990 and the recently released 2009 Marital History Topical Modules (Wave 2 of the 1990 and 2008 panels) to the Census Bureau's SIPP—a nationally representative household survey designed to measure program participation and the economic situation of persons, families, and households in the United States.

SIPP's Marital History Topical Module, implemented once within each panel, collects retrospective marital histories for every person in the household aged 15 or older. The data include information on marital transitions in months and years, which permit precise estimates of whether a respondent has a 10-year marriage. Because comprehensive marital history data are no longer collected by the National Center for Health Statistics and the Current Population Survey, the Marital History Module represents the best current data sources of marital history for the US population. Moreover, because the Census Bureau is redesigning the SIPP for 2014, the 2009 SIPP marital history data will be the most recent source for comprehensive marital histories of individuals for some time. The Social Security Administration (SSA) has funded a supplement to the redesigned SIPP to collect SIPP respondent's marital histories in the fall of 2014 and release data in 2015.

The current study relies on a restricted-use file of the 2009 SIPP Marital History Module. In contrast to 1990, marital transition dates in months are not provided in the 2009 public-use file. Because the inclusion of marital data in months allows for more precise estimates of length of marriage, particularly with respect to Social Security's 10-year marriage rule, we utilize a restricted-use file, which contains this information.⁷ To protect respondent confidentiality, all users of restricted files must have authorization from the Census Bureau, and the data must be used only for research purposes. All statistics have also been reviewed by SSA's disclosure review board.

Analytic Approach

Our study sample consists of women aged 30–59 in the 1990 and 2009 Marital History Modules, corresponding to an unweighted sample size of 11,208 in 1990 and 20,936 in 2009. Descriptive statistics of the study sample appear in the Appendix, Table A-1.

A key aspect of our analytic approach is the stratification of the sample into three age groups: young adult (30–39), middle age (40–49), and late middle age (50–59). Comparing those age-specific groups across the two survey years permits us to identify differences across period and birth cohorts. Table 2 illustrates the birth cohorts represented for each age group in our analysis. As shown in the table, women in their forties and fifties in 2009 approximated the late baby boom/early GenX and middle baby boom birth cohorts, while women in their thirties represented the GenX birth cohort. By contrast, women in their fifties constituted the pre–baby boom cohorts in 1990, while women in their forties represented a mix of pre– and early baby boomers, and women in their thirties reflected the middle of the baby boom cohort. We recognize that younger women, particularly in their thirties, are far from finishing their marital histories. However, taking stock of women's marital experiences at the early to midlife stages of their life course across nearly two decades provides important insights into the broad direction of marital changes that may affect women's Social Security eligibility for marriage-based benefits in the future.

Our analysis uses descriptive and multivariate regression methods. Given prior research showing racial differences in eligibility for Social Security marriage-based benefits resulting from differential marital patterns, our analysis provides separate estimates by race and ethnicity. Among Hispanic women, we also consider outcomes between those who were born in the United States and those who were foreign born. Unless otherwise noted, descriptive differences discussed in the text are significant at the .05 level.

The multivariate regression analysis employs a probit model to examine the probability of not having eligibility for Social Security spouse or widow benefits based on marital history, holding important variables constant. We estimate separate probit regressions for the three age-specific groups in 1990 and 2009. This allows us to better identify differences in potential eligibility between 1990 and 2009 among women at the same stage of the life cycle. The general model can be expressed as follows:

where Y is the estimated probability of having a marital history that would imply ineligibility for Social Security spouse or widow benefits net of other characteristics; α is the intercept, βs are the probit regression coefficients, and _E is the error term. Vector RE reflects the dummy variables measuring race and ethnicity/nativity, and vector C refers to the control variables.

The dependent variable is a dichotomous measure denoting "ineligible" for Social Security spouse or widow benefits based on marital history (1 = yes, 0 = no). In this case, ineligible refers to those women who were never married or currently divorced without any 10-year marriage at the time of the survey.

The key independent variables are the set of five dummy variables indicating race and ethnicity/nativity for Hispanics: (1) white non-Hispanic, (2) black non-Hispanic, (3) US-born Hispanic, (4) foreign-born Hispanic, and (5) other. The omitted reference group is white non-Hispanic. When we use the identifier black or white, we refer to non-Hispanics. Comparing the marginal effects of the race and ethnicity dummy variables in the 1990 and 2009 estimates helps us to identify whether the association between potential eligibility and race and ethnicity has changed.

Control variables are educational attainment and age at the time of the survey. Education is measured by four dummy variables: less than high school graduate, high school graduate, some college, and bachelor's degree. High school graduate is the reference group. We measure the respondent's age in years. Income is not controlled for given that income measures in the SIPP reflect a woman's income status at the time of the survey rather than over her lifetime. Current marital status is quite likely to affect family income; for example, married couples tend to have higher family incomes than single persons. This relationship may result in biased estimators if income was included in the equation.

Throughout the analysis, we use Stata's (SVY) command to obtain standard error estimates that account for SIPP's complex survey design (StataCorp 2009).⁸ Our analysis also applies the appropriate survey person weights. For the probit regression, we report the marginal effects, which can be interpreted as the probability of ineligibility for spouse or widow benefits because of marital history, holding the other variables in the model constant. For ease of presentation, we do not report the estimates for the control variables, which are available upon request.

Findings

We begin with the descriptive analysis. Table 3 reports the marital history measures for women in 1990 and 2009 by the three age groups: young adult (30–39), middle age (40–49), and late middle age (50–59). Separate estimates for white, black, and Hispanic women are provided.

Looking first at the late middle-aged group (50–59), the data show a trend away from marriage. Compared with 1990, late middle-aged women in 2009 were more likely to be never married, ever divorced, currently divorced, and currently divorced without a 10-year marriage. The growth in the never-married subgroup was substantial over the period, more than doubling from 4.5 percent to 9.1 percent. The increase was particularly sharp for black women (from 7.8 percent to 21.5 percent, compared with from 3.7 percent to 7.2 percent for white women and 7.4 percent to 11.0 percent for Hispanic women). The data show significant increases in the proportion of women who were currently divorced without a 10-year marriage, from 3.1 percent to 7.9 percent (among the ever married). By race and ethnicity, black women aged 50–59 had particularly sharp increases in the share who were divorced without a 10-year marriage (from 6.1 percent to 15.8 percent of ever-married black women, compared with from 2.7 percent to 7.1 percent for their white counterparts and from 3.8 percent to 7.0 percent for their Hispanic counterparts).

Among middle-aged women (40–49), substantial marital changes also occurred. Compared with 1990, middle-aged women in 2009 were more likely to be never married or currently divorced without a 10-year marriage. The percentage ever divorced remained roughly constant between 1990 and 2009, very likely reflecting the leveling of divorce rates since the 1970s.

The data reveal sharp differences among middle-aged women by race and ethnicity. Between 1990 and 2009, the percentage of ever-married black women decreased 16.3 percentage points (to 69.0 percent) compared with white women who experienced a decline of 4.7 percentage points (to 90.2 percent). The comparable figure for Hispanic women reflected a 2.4 percentage point decline (to 85.9 percent). In terms of divorce, a higher proportion of ever-married, black middle-aged women were currently divorced without a 10-year marriage than white women, but this gap narrowed between 1990 and 2009. For Hispanics, the percentage of currently divorced women without a 10-year marriage slightly declined. Thus, racial differences in ever-married women aged 40–49 seem to have expanded between 1990 and 2009, but those differences contracted in terms of the share currently divorced without a 10-year marriage.

The most substantial changes appear in the young-adult group (aged 30–39). A "retreat from marriage" was evident among all groups, but it was greatest for black women. Specifically, the percentage of never-married black women rose from 32.2 percent in 1990 to 47.3 percent in 2009, while the percentage currently married dropped from 52.4 percent to 41 percent. The proportion of ever-married black women currently divorced without a 10-year marriage also increased, from 15 percent in 1990 to 17.7 percent in 2009. Of course, women in their thirties are far from completing their lifetime marital histories, but the trend appears in the direction of less marriage and with increasing black/white disparities among more recent birth cohorts.

Chart 2 presents a summary measure showing the share of women in each age cohort and race/ethnic subgroup who lacked marital histories required for Social Security spouse or widow benefits. As previously noted, ineligible reflects the sum of never-married women and currently divorced women without a 10-year marriage. Overall, the results show that well over half of all women in each age group had qualifying marital histories in 1990 and 2009, but at the same time, a larger proportion in each age-specific group would be ineligible for Social Security marriage-based benefits in 2009 relative to 1990.

Chart 2 also reveals important differences by race and ethnicity. Black women comprised the highest share ineligible for marriage-based benefits in all age groups. Moreover, black/white differences appear to have widened between 1990 and 2009. The percentage point difference between the share of black and white women who lacked potential eligibility in the middle-aged group was greater in 2009 (that is, the gap was about 21 percentage points in 2009, corresponding to 40.9 percent (black) and 19.6 percent (white); those figures are compared with about 13 percentage points in 1990, corresponding to 24.3 percent (black) and 10.9 percent (white)). Among the young-adult group, black/white differences in ineligibility grew because of the faster rate of increase in never-married black women.

Hispanic women's incidence of ineligibility because of marital history is more similar to white than black women (see Chart 2). However, thus far, we have discussed Hispanic women without distinguishing between those who were US born and foreign born, two subgroups thought to have differing marital behavior. To assess possible differences by nativity, Chart 3 shows the proportion of US- and foreign-born Hispanic women without a qualifying marital history for spouse or widow benefits. The results confirm clear differences in marriage-based eligibility by nativity, with US-born Hispanic women having higher proportions of ineligible marital histories than those born outside the United States. For example, the percentage of ineligible foreign-born Hispanic women in late middle age (50–59) in 2009 was 15 percent compared with 21 percent of those who were US born. Among the middle-aged group (40–49), 28 percent of US-born Hispanic women did not have a qualifying marital history compared with 13 percent of those who were foreign born. Among the young-adult group (aged 30–39), 31 percent of US-born Hispanic women in 2009 had ineligible marital histories compared with 22 percent of those who were foreign born. These observed differences are driven by differences in marital patterns, which are shown in detail in the Appendix, Table A-2. Briefly, we find higher rates of ever marrying among foreign-born Hispanic women compared with their US-born counterparts, particularly in the young-adult group. In terms of marital dissolution, the likelihood of being divorced is substantially lower among foreign-born Hispanic women in the young-adult and middle-aged groups.

In sum, the descriptive data indicate substantial changes in the marital patterns of recent cohorts of women and in the distribution of women potentially eligible for Social Security spouse or widow benefits. There is significant differentiation by race, ethnicity, and nativity. The next section examines those trends in a multivariate context.

Probit Regression Results

Using probit regression, we estimated the probability of being ineligible for Social Security marriage-based benefits in the three age groups in 1990 and 2009. Our main focus was to determine whether women's likelihood of being ineligible for spouse or widow benefits based on marital history changed over time and how that pattern varied by race, ethnicity, and nativity among Hispanics, controlling for the variables in the model. Table 4 shows the results of separate regressions for 1990 and 2009 for the three age groups: young adult (30–39), middle age (40–49), and late middle age (50–59). We report the marginal effects, their significance level (two-tailed tests), and the standard errors, which account for SIPP's complex survey design.

Overall, the models indicate that women in 2009, across all three age groups, had a substantially higher predicted probability of not meeting the marital requirements for spouse or widow benefits than the comparable model in 1990, holding race and the other variables constant. These results capture cohort differences in marriage patterns at the same stage of life, namely relative rises in never-married and currently divorced women without 10-year marriages.

In terms of race, being black was strongly and positively associated with higher probabilities of being ineligible for marriage-based benefits (p < .001) across all age groups in 1990 and 2009. The marginal effect of being black was higher in 2009 compared with 1990, across all age groups. In other words, black women in 2009, relative to comparable white women, had higher probabilities of being ineligible for marriage-based benefits than their age-specific counterparts in 1990, all else being equal. At ages 30–39, for example, the marginal effect of being black on the probability of not having a qualifying marital history, relative to being white, was 29 percent in 2009 compared with 22 percent in 1990. That pattern reflects the faster increase in never-married black women in the young-adult cohort from 1990 to 2009.

Among Hispanic women, we find sharp differences between those born in the United States and those who were not. US-born Hispanic women in the young-adult (survey year 1990), middle-aged (survey years 1990 and 2009), and late middle-aged (survey year 2009) models were significantly more likely to have a nonqualifying marital history than their white counterparts. By contrast, the results for foreign-born Hispanic women were more mixed, with some models (for example, models 1 and 2 in 2009) showing significantly lower probabilities of those women having a nonqualifying marital history. Overall, then, Hispanic women, namely those who were born in the United States, have ineligible probabilities resulting from marital history that appear to lie between those of white and black women. Thus, US-born Hispanic women were more likely to be ineligible because of marital history than white women in the middle- and late middle-aged groups in 2009, but the magnitude was less than black/white differentials at the same stage of life. Interestingly, the probability of ineligibility because of marital history appeared to decrease among foreign-born Hispanic women in the middle- and late middle-aged groups in 2009, shown by the changing sign in the marginal effect between 1990 and 2009. This outcome likely picks up, in part, different patterns of immigration waves over time (see note 6).

Conclusions

The majority of aged women currently receive a Social Security benefit at least partly based on the earnings record of a present or past spouse. The distribution of women's potential eligibility for Social Security spouse or widow benefits is, however, dynamic and subject to changes in marital trends. In this study, we used data from the Census Bureau's SIPP Marital History Modules in 1990 and 2009 to shed light on the implications of trends in women's lifetime marital experiences for Social Security spouse and widow benefit eligibility. Overall, we find that most women are approaching retirement age with marital histories that make them potentially eligible for spouse or widow benefits in the future. However, fundamental changes in marital patterns, starting with the leading edge of the baby boom cohort, will prompt a decline in the share of women potentially eligible to receive these benefits.

Our findings reveal considerable changes in the marital patterns of women aged 30–59 between 1990 and 2009. Those changes have been marked by increasing proportions of women who were never married, ever divorced, and currently divorced with shorter marriages (less than 10 years). Such patterns very likely reflect a mix of period and birth cohort effects as the sample moves from the large baby boom generation (born in the 1946–1964 period) through GenX (born in the 1968–1979 period). The results also indicate important differences by race, ethnicity, and nativity among Hispanic women. We find a sharper retreat from marriage for black women among more recent birth cohorts (Goldstein and Kenney 2001; Harrington Meyer, Wolf, and Himes 2006; Tamborini, Iams, and Whitman 2009). Consistent with the literature, the marital histories of Hispanic women lie somewhere between their white and black counterparts. We also find distinct marital patterns among US- and foreign-born Hispanic women.

The observed changes in women's marital histories are associated with an increase in the share without a qualifying marriage for Social Security spouse and widow benefits. Thus, the increase in women's ineligibility for Social Security spouse or widow benefits found in past research (Harrington Meyer, Wolf, and Himes 2006; Tamborini, Iams, and Whitman 2009) continues among recent cohorts of women, including those in GenX. In short, GenX women, although just in their thirties, appear to be continuing the retreat from marriage. Although the direction of marital patterns among those women could change as they age, the postboomer cohort appears to be following the trend of increasing proportions of women with nonqualifying marital histories for spouse or widow benefits, at least in an earlier stage of the adult life course.

The results also document sharp differences in benefit ineligibility by race and ethnicity. Across all of the age groups, black women had substantially higher proportions without a qualifying marital history for Social Security spouse or widow benefits than comparable white women. Moreover, we find that the widening black/white differentials in potential ineligibility shown in previous studies (Tamborini, Iams, and Whitman 2009) continue in the young-adult (aged 30–39) cohorts. A key driver of this trend is the increase in never-married black women.

Among Hispanic women, a central finding is that eligibility for Social Security spouse and widow benefits based on marital history varies by nativity. The probability of not having a qualifying marriage was greater among US-born Hispanic women than their foreign-born counterparts, particularly among the younger cohorts. However, differences between US-born Hispanic women and comparable white women were not as wide as black/white differences. Hispanic women may assimilate over time to the increasing retreat from marriage, as suggested by Oropesa and Landale (2004), which may result in widening differences in the marital experiences between white and Hispanic women in future years. Complicating interpretation, Hispanic immigrants come from many countries with different cultures, and country of origin shifted across the time period examined in this study. Our sample size is not large enough to identify country of origin reliably for immigrants and parents of US-born Hispanics. We also did not consider differences among Hispanics in terms of their immigration history (that is, first generation US born versus second or greater generations).

Put together, the changing distribution of women potentially eligible for spouse or widow benefits can have important consequences for their retirement resources and for the Social Security program as a whole. Recent microsimulation projections based on the Modeling Income in the Near Term (MINT) data system have shown that women in the baby boom and GenX cohorts will be more likely to receive retired-worker benefits when their spouse is alive, but most will continue to take up widow benefits (dually entitled) if they outlive their husband because of lower lifetime earnings (Butrica and Smith 2012a). However, as this study suggests, a smaller share of future female retirees will have the option of augmenting their retirement benefit based on the work record of a deceased spouse because of shifts in marital patterns. That trend will be particularly pronounced among future black female retirees.⁹ There is some reason for concern that those women may be at higher risk of economic vulnerability in old age. Microsimulation projections of the retirement-age population in 2030, for example, have suggested disproportionately high rates of poverty and near-poor status among divorced female retirees with less than 10 years of marriage (Tamborini and Whitman 2010) as well as the never married (Tamborini 2007).

Future empirical work could clarify the consequences of changing eligibility for Social Security spouse or widow benefits for women's retirement income security. For example, the extent to which declining eligibility is concentrated in minority or less-educated subgroups may influence the effect of those changes on economic outcomes. Additionally, a further retreat from marriage among GenX cohorts, if sustained over the life course, may have important consequences for women's retirement resources in future years. The relationship between marriage and retirement outcomes among Hispanic women may be a salient topic given the lack of research in this area.

Appendix