Women, Marriage, and Social Security Benefits Revisited

Summary and Introduction

Introduced into the Social Security program in 1939, spouse and survivor benefits have important implications for the retirement experience of women. At the end of 2005, 12.9 million women Social Security beneficiaries aged 62 or older (59 percent) received at least part of their benefit as wives or widows of entitled workers.¹ For these women, a spouse or widow benefit provided a larger payment than that of their own earnings record (SSA 2007, Table 5.A14).

A number of dynamic factors influence the number of women (and men) eligible for, and claiming, Social Security spouse or survivor benefits at retirement. Past and present marital status determines potential benefit eligibility (typically, one must be currently married, widowed, or have had a 10-year marriage to qualify), and the benefit amount is based on a person's lifetime earnings record in relation to a current or former spouse. This article focuses on the marital history component of eligibility, while acknowledging that large-scale changes in women's workforce attachment over the past half century, such as increases in labor market participation and earnings relative to men, have led to an increase in the share of women retirees receiving at least part of their benefit based on their own earnings record.

The connection between marital history and eligibility for spouse or widow benefits is of increasing interest among policymakers and retirement analysts (Favreault and Steuerle 2007; Harrington Meyer, Wolf, and Himes 2006). Although marital trends have begun to stabilize in the United States in recent years (for example, divorce rates appear to have leveled off), an extensive literature documents dramatic changes in the marital patterns of women over the past several decades. Notable changes include increases in divorce, decreases in marriage durations, and a rise in the number of women who delay or forgo marriage.² From a retirement perspective, shifting marital patterns are important as they may mean that the share of women with the option of claiming a spouse or widow benefit at retirement may change as well. For example, recent trends show higher proportions of never-married women in the population (Tamborini 2007), which would reduce the share of women potentially eligible for spouse or widow benefits by the time they reach retirement age.

Since marital histories help establish whether spouse or widow benefits will be a retirement income option for women in old age, it is important to track marital trends, particularly as the leading edge of the baby-boom generation begins retiring. This article uses data from the 2001 Marital History Topical Module (wave 2) to the U.S. Census Bureau's Survey of Income and Program Participation (SIPP) to examine the extent to which marital patterns are changing among different age groups of women. The SIPP Marital History Module is considered one of the best surveys to gauge recent marital patterns, especially since the National Center for Health Statistics and the Current Population Survey no longer collect comprehensive data on marriage and divorce. Thanks to an agreement with the U.S. Census Bureau, the Social Security Administration has access to a Restricted-Use File of the Marital History Module, which contains the year and month of marital transition events (marriage, divorce, widow). This information is important because it allows for the estimation of length of marriages, an important component of this article's analysis.³

To explore how marital patterns have changed over the past 15 years, the article also draws on data reported in Iams and Ycas (1988), which is based on the U.S. Census Bureau's 1985 Marital History Supplement to the Current Population Survey (CPS).⁴ To facilitate comparability between the 2001 SIPP data analyzed in this article and the 1985 CPS data reported in Iams and Ycas' work, women are grouped into three age cohorts: Older Age (ages 60 to 69 in the year of analysis), Late Middle Age (ages 50 to 59 in the year of analysis), and Middle Age (ages 40 to 49 in the year of analysis).

Altogether, the article highlights important linkages between sociodemographic trends in marital patterns, the Social Security program, and its beneficiaries. By comparing the marital histories of women aged 40 to 69 in 2001 to similarly aged women in 1985, the article provides an indication of how marital trends are changing and how such changes may impact the extent to which spouse and widow benefits will be an option for future women retirees. Results indicate that the majority of women aged 40 to 69 in 2001 (over three-fourths) already had marital histories that will guarantee them the option of a spouse or widow benefit at retirement. However, a smaller proportion of these women would be potentially eligible to receive spouse or widow benefits compared with their counterparts in 1985 due to changes in current marital status or past marital duration. The magnitude of change between 1985 and 2001 varies markedly by age cohort, with the smallest shift in marital patterns having occurred among the Older Age and Late Middle Age groups and the most dramatic among the Middle Age group. Notable shifts, by and large, reflect marital pattern changes in the younger wave of the large baby-boom cohort and include a rise in the share of divorced women with shorter marriages (less than 10 years) and never-married women.

Women and Social Security Auxiliary Benefits

The retirement security of women is a clear concern among policymakers and retirement analysts. Although women's economic situation at old age has improved greatly over the past 30 years, women are still more likely than men to experience old-age poverty; partly because women earn less over their lifetimes and live longer than men. In 2004, the poverty rate for women aged 65 or older was 12 percent, compared with 7 percent for men (SSA 2006, Table 8.1).

Established in the 1939 Social Security Amendments, spouse and widow(er) benefits, sometimes called dependent or auxiliary benefits, provide monthly payments to qualified spouses and survivors of insured workers (Martin and Weaver 2005).⁵ Receipt of a Social Security auxiliary benefit is contingent on two main factors: marital status/history and lifetime earnings relative to a current or previous spouse. This link between benefit eligibility and marital history has become increasingly important given recent evidence of changing marital patterns among the baby-boom and younger birth cohorts (Butrica and Iams 2000; Favreault and Steuerle 2007; Goldstein and Kenney 2001; Harrington Meyer, Wolf, and Himes 2006; Kreider and Simmons 2003).

Table 1 summarizes the current marital requirements for spouse and widow benefits. Under current law, married women are eligible for a spouse benefit at retirement age. The benefit equals 50 percent of their spouse's primary insurance amount (PIA) if claimed at the full retirement age (FRA). Divorcees may be eligible for a divorced spouse benefit based on their previous spouse's lifetime earnings provided that they had a 10-year (120 months) marriage. If a woman qualifies for benefits based on her own work record and her spouse's work record, she will receive the higher amount of the two (her own PIA or 50 percent of her husband's PIA). Qualifying spouse beneficiaries must be married to the retiring spouse for at least one continuous year prior to applying for benefits, with certain exceptions.⁶

The Social Security program also provides benefits to widow(er)s of deceased workers.⁷ The benefit for widow(er)s can equal 100 percent of a deceased spouse's PIA if the surviving spouse begins to collect benefits at the FRA. A reduced benefit (from 71 percent to 99 percent of the deceased's PIA) is available as early as age 60 (age 50 if disabled).⁸ Divorced surviving spouses may also qualify if their marriage lasted at least 10 years. Qualifying widow(er)s must have been married to the deceased spouse for at least 9 months and have not remarried before the age of 60 (50 if disabled) (SSA 2007b). Since these duration requirements are fairly short, this article assumes that they have been met.

It is important to note that spouse and widow benefit rules have been modified over time.⁹ The 1939 Social Security Amendments originally established a spouse benefit equal to one-half of the retired-worker benefit of the present spouse and a survivor's benefit equal to three-fourths of the deceased spouse's worker benefit. Benefits were extended to divorced widows with children and dependent widowers in the 1950 amendments. In 1965, divorced wives and surviving divorced spouses without children became eligible for payments, but there was a dependency and a 20-year length of marriage requirement.¹⁰ Congress raised widow(er)s benefits in 1972 from 82.5 percent of the deceased worker's benefit to 100 percent (Martin and Weaver 2005). In 1977, the length of marriage requirement was reduced from 20 years (240 months) to 10 years (120 months), and remarried women were allowed to receive a regular survivor benefit from the deceased spouse's record if the remarriage occurred after age 60 (age 50 if disabled).

To date, Social Security spouse and widow benefits remain a major source of income for elderly women (Butrica, Iams, and Sandell 1999; Favreault, Sammartino, and Steuerle 2002; Favreault and Steuerle 2007; Harrington Meyer, Wolf, and Himes 2006; Weaver 1997). As shown in Chart 1, over the past 45 years the majority of women Social Security beneficiaries aged 62 or older received a benefit based at least partly on the earnings record of their spouse or previous spouse, 61 percent (4.1 million) in 1960 and 59 percent (12.9 million) in 2005.¹¹

That said, increased labor market participation among women and a rise in their earnings relative to men (Blau, Ferber, and Winkler 2006; Bureau of Labor Statistics 2005; Clark and Weismantle 2003),¹² have led to a decline in the share of women receiving benefits as wives or widows only, and correspondingly, an increase in the share of dually entitled women—that is, women who qualify for benefits as retired workers and as spouses and surviving spouses. This development is highlighted in Chart 1, which shows that between 1960 and 1985, the share of female beneficiaries aged 62 or older who were dually entitled grew from 5 percent to 20 percent. By 2005, 28 percent of women beneficiaries 62 or older were dually entitled; 12 percent therein were dually entitled spouses (2.7 million) and 16 percent were dually entitled widows (3.5 million).¹³ The average monthly benefit in 2005 for women aged 65 or older who were dually entitled spouse beneficiaries was $601 and for dually entitled survivor beneficiaries it was $1,131; for spouse-only beneficiaries the monthly benefit averaged $507, and for widow-only beneficiaries the average was $969 (SSA 2007, Table 5.A15).

Women's Marital Patterns, 1985 and 2001

The estimates reported in this article are based on a Restricted-Use File of the 2001 Marital History Module to the U.S. Census Bureau's Survey of Income and Program Participation (SIPP).¹⁴

Data

The SIPP is a household survey of the civilian noninstitutionalized resident U.S. population made up of a series of panels, each of which represents a new sample. The 2001 panel began in 2001 and extended through 2003 with interviews conducted every 4 months. The survey can be used as a longitudinal sample or cross-sectional study, as is done here.

SIPP's topical modules represent a variety of topics and are implemented on a rotating basis in survey panels. In this article, we rely on data from the Marital History Module, which was administered in wave two (the second interview) of the 2001 panel. The supplement is designed to collect comprehensive information on the marital histories of persons, families, and households in the United States. Note that, because marital event dates are suppressed in the Public-Use File of the 2001 Marital History Module, the following analysis relies on a Restricted-Use File, which contains the dates (in year and months) at which a respondent's marriage (up to three) began and ended. Such information allows us to estimate length of marriage(s), an important factor determining potential eligibility for Social Security auxiliary benefits. Similar to other topical modules, the marital history module is collected in combination with the core questionnaire, which covers general demographic, socioeconomic, and program participation information across all waves of each panel. All users of restricted-use data must be granted permission by the U.S. Census Bureau and the Social Security Administration, and the data must be drawn for research purposes only.¹⁵

To ascertain how women's marital histories are changing, the article also draws on data from the June 1985 Marital History Supplement to the U.S. Census Bureau's Current Population Survey (CPS), as reported by Iams and Ycas (1988). An important asset of this data is that the marital history supplement questions were essentially the same questions used in the 2001 SIPP panel. In both cases, women were asked a sequence of similar questions relating to the dates that marriage events started and ended, in months and years (up to three marriages).¹⁶ Given that the Current Population Survey (after 1995) and the National Center for Health Statistics no longer collect comprehensive data on marriage and divorce, the SIPP Marital History Module represents one of the best available microlevel data sources to gauge recent marital trends.

All declarations of numeric change between 1985 and 2001 have undergone statistical tests of difference, and only those differences that are statistically significant at a 95-percent level are cited.¹⁷ To lessen the magnitude of sampling error, statistics from a weighted population base of less than 200,000 are not reported. Nevertheless, particular caution should be used when comparing differences between estimated percentages across data sources. Appendix A discusses the methodological issues of this study in greater detail. Appendix B provides the standard errors of the estimates reported in this article. All estimates are weighted using the final person weights supplied in the data set.

Tables 2–6 report the results, which are organized by age cohort, beginning with the Older Age group.

Older Age Cohort: Age 60–69

The Older Age cohort in 2001 is comprised of women born between 1932 and 1941. Women's marital histories at this age are particularly salient as these women have already reached, or are on the verge of, retirement age. As a starting point, the share of Older Age women potentially eligible for spouse or widow benefits on the basis of current marital status was determined (Table 2). In 2001, 63.7 percent of Older Age women were currently married (including separated) and 19.7 percent were widows, compared with 63.0 percent and 25.4 percent, respectively, in 1985 (Table 2).¹⁸

In terms of women who would not be potentially eligible for spouse or widow benefits, Table 2 shows that roughly 4 percent of Older Age women were never-married in 2001, which is consistent with what Iams and Ycas found in the 1985 CPS. While there is no observable change in the share of never-married Older Age women, the data reveal important changes in divorce patterns such that a larger share of Older Age women in 2001 were currently divorced (12.6 percent) or had ever-divorced (28.3 percent) than was recorded in the 1985 CPS estimates (7.4 percent and 17.6 percent, respectively) (Table 3). However, potential eligibility for spouse or widow benefits among divorced women cannot be determined by current marital status alone.

To qualify for the option of receiving a divorced spouse or surviving divorced spouse benefit, currently divorced women must have had at least one marriage lasting 10 years. Among ever-married Older Age women in 2001, roughly 2 percent were both currently divorced and without a 10-year marriage (Table 4). Stated differently, 17 percent of currently divorced Older Age women in 2001 had less than 10 years (120 months) in any marriage. These figures show no significant deviation from Older Age women in 1985.

With respect to length of marriage, estimates also show that nearly 80 percent of ever-married Older Age women had reached their 20-year anniversary in 2001 (Table 5). Older Age women in 2001 had spent a similar average number of years (36.1 years) in marriage (as a total, not just a single marriage) as their counterparts in 1985 (35.4 years) (Table 6).

Overall, estimates indicate relatively small changes between the marital trajectories of Older Age women in the 2001 SIPP and their counterparts in the 1985 CPS. Those Older Age women who were not potentially eligible for auxiliary benefits in 2001 consisted of the 4.1 percent never-married and the 2.2 percent ever-married women who were divorced with less than 10 years in any marriage.¹⁹ This modest continuity in marital patterns between 1985 and 2001 was not as evident among the Late Middle Age and Middle Age groups.

Late Middle Age Cohort: Age 50–59

The Late Middle Age cohort in 2001 consists of women born between 1942 and 1951. The oldest of these women reached age 62 in 2004 and the youngest will be 62 in 2013. In terms of current marital status, Table 2 shows that 68.7 percent of Late Middle Age women in 2001 were potentially eligible for spouse or widow benefits as a result of being married or separated and 7.1 percent due to their current status as widows, which are lower than the equivalent percentages in the 1985 CPS data.²⁰

Estimates also show that the share of this group who were never-married increased by more than 2 percentage points from 1985 to 2001(Table 2). While never-married women in this age group may change their current status by the time they reach retirement age, a first marriage between the ages of 50 to 59 is relatively unlikely. From 1985 to 2001, the median age at first marriage for women increased from 23.3 years to 25.1 years (U.S. Census Bureau 2005), which is 25 years younger than even the youngest members of the Late Middle Age cohort.

With respect to divorce, 38.8 percent of Late Middle Age women in the 2001 SIPP were ever-divorced (Table 3). This figure represents a substantial increase over the 22.4 percent of ever-divorced women in the 1985 CPS. Currently divorced women also represented a larger share of the Late Middle Age group in 2001 than in 1985 (17.9 percent versus 10.3 percent, respectively). That said, remarriage remained common among Late Middle Age women in 2001, with over half of the ever-divorced having at some point remarried.

The share of currently divorced Late Middle Age women without a 10-year marriage also increased between 1985 and 2001. In 2001, 31.0 percent of those who were currently divorced did not have a marriage that met the 10-year duration requirement, compared with 22.2 percent in 1985 (Table 4). Stated differently, 6.0 percent of the ever-married Late Middle Age female population were currently divorced and without a 10-year marriage in 2001, up from the 2.4 percent reported in the 1985 CPS.

Many Late Middle Age women in 2001 had already been married for a long period of time; 63.6 percent of the ever-married had been married at least 20 years (Table 5). The average number of years married among Late Middle Age women in 2001, 26.8 years, is fairly similar to Iams and Ycas' findings of 28.7 years for 1985 (Table 6).

Middle Age Cohort: Age 40–49

The Middle Age cohort represents the youngest age group considered in this study. In 2001, women in this group were born between 1952 and 1961 and are thus reflective of the younger wave of the large baby-boom cohort born between 1946 and 1964. The oldest of this group in 2001 will reach age 62 in 2014 and the youngest in 2023. Since there is less certainty about the marital paths that the Middle Age group will take as they age into retirement, findings in this section should be regarded with caution when estimating women's potential eligibility for spouse or widow benefits between 2014 and 2023.

As shown in Table 2, 70.3 percent of Middle Age women in 2001 were currently married (including separated), compared with 77.1 percent in the corresponding 1985 CPS estimates (Table 3).²¹ A particularly striking change between Middle Age women in 1985 and 2001 is the increasing share in the never-married population. Roughly 10.5 percent were never-married in 2001, while about half this amount was documented in the 1985 CPS data. An important consideration that will become evident in future years is whether marriage is simply being delayed until later in life among younger cohorts or avoided altogether (see Goldstein and Kenney 2001).

The experience of Middle Age women with divorce has also changed. In 2001, 16.8 percent were currently divorced and 35.5 percent had at least one divorce, up from 13.9 percent and 29.1 percent, respectively, from their 1985 predecessors (Table 3). Furthermore, 51.4 percent of currently divorced women aged 40 to 49 were without a 10-year marriage in 2001, up from 32.7 percent in 1985 (Table 4). Looking at the ever-married group, 9.7 percent of ever-married Middle Age women were divorced and without a qualifying marriage in 2001, compared with 4.8 percent in 1985 (Table 4). However, some of these currently ineligible women may remarry before retirement.

Concluding Remarks

Trends in women's retirement income have been a longstanding point of interest in the retirement literature, and Social Security spouse and widow benefits constitute an important source of income for many aged women. Since marital history affects women's potential eligibility for spouse and widow benefits, it is important to understand changes in marital trends in the context of Social Security rules, particularly as the leading edge of the baby-boom cohort begins retiring.

While caution should be used in interpreting differences across data sources, this article highlights both important changes and continuities in women's potential eligibility for spouse and widow benefits since 1985. On the one hand, despite important structural changes in marital patterns such as the "retreat from marriage" (Goldstein and Kenney 2001; Waite 1995), over three-fourths of women aged 40 to 69 in the 2001 SIPP already had marital histories that assured them the option for spouse or widow benefits at retirement. On the other hand, trends reveal a downward shift in the share of women potentially eligible for spouse or widow benefits in future years due to changing marital patterns among baby boomers. Changes are principally evident for the Middle Age cohort (aged 40–49), which had a significantly higher share of never-married and divorced women with shorter marriages in 2001 than in 1985.

As a whole, this article contributes to our understanding of sociodemographic trends affecting the Social Security program and distributional outcomes of its beneficiary population. Most aged women qualify for Social Security benefits as wives or widows today. However, a growing tendency toward never marrying, along with shorter marriages before divorce, among the younger wave of baby boomers may foreshadow a decline in the proportion of women that will be eligible for spouse or widow benefits in future years. These trends could have an important effect on the sources, and in some instances adequacy, of retirement income for some women, as well as the sociodemographic profile of the overall beneficiary population for Social Security.

Appendix A: Methodological Issues

This study's analytic approach raises several methodological issues for elaboration. To begin with, estimates based on the SIPP and CPS, as with any survey data, are subject to errors related to measurement and sampling, among other factors. Because estimates of women's marital histories may differ from the actual population, a standard error for the estimated percentages was calculated for all reported estimates using the appropriate generalized variance tables for each survey.

Special care should be taken when making numeric contrasts based on two independent data sources. In this article, estimates of women's marital histories from the 2001 SIPP are evaluated against estimates from the 1985 CPS on which the Iams and Ycas' (1988) study is based. Although both of these surveys were conducted by the Census Bureau and represent large, nationally-representative samples of U.S. households, they are designed to meet different objectives. Each survey, for example, has different sampling procedures, sample compositions, and survey reference periods, all of which affect the parameter estimates. The CPS data are based on a single cross-sectional study, whereas SIPP data are based on a longitudinal panel sample. A panel design can introduce nonsampling errors not found in cross-sectional samples, such as a "seam" effect (patterned heavy concentration of responses to adjacent months between waves) and a "historical maturity" effect.²² Another concern present in both surveys is the potential bias in respondents' retrospective report of their marital histories (see Lillard and Waite 1989; Weaver 2000).

A number of steps were taken to lessen the negative impact of the aforementioned concerns. All declarations of numeric change between 1985 and 2001 have undergone statistical tests of difference. Only those numeric differences that are statistically significant at a 95-percent level are cited. Moreover, standard error calculations were based on generalized variance tables rather than the formula for random samples. The test of difference between two estimates accounts for, in part, the impact of the individual survey design. To further reduce the magnitude of error, statistics are derived from weighted data. Any statistic from a weighted population base of less than 200,000 is not reported.²³

It is worthwhile to mention that although the estimates used in this article are derived from separate surveys, they are based on largely consistent supplemental marital history modules. Both the SIPP and CPS surveys used a similar sequence of questions about marital history, focusing on events such as duration of marriages (month and year of each marriage event) and number of times married. This consistency reduces potential bias arising in nonsampling errors between the SIPP and CPS.²⁴ While estimates of marital status from the CPS have been shown to be rather consistent with those from the SIPP,²⁵ any numeric contrast between women's marital histories in 2001 and their counterparts in 1985 should be interpreted as representing a broad directional shift in marital patterns, rather than an exact percentage change over the period of analysis.

Appendix B: Standard Errors

The following tables present the standard errors of estimated percentages reported in this article. They were approximated using the formula

$$S_{(x,p)}=\sqrt{\frac{b}{x}(p)(100-p)}$$

where x is the total number of persons, families, or households (the base of the percentage), p is the percentage, and b is the parameter associated with the characteristic in the numerator of the percentage. For parameters associated with estimated percentages based on the 2001 SIPP Marital History Topical Module (wave 2) see U.S. Census Bureau (2005b, Table 9); for reported 1985 CPS estimates see U.S. Census Bureau (1986, Table C-5).