Earnings Sharing in the U.S. Social Security System: A Microsimulation Analysis of Future Female Retirees

Transcription

1 The Gerontologist Published by Oxford University Press on behalf of The Gerontological Society of America Vol. 50, No. 4, All rights reserved. For permissions, please doi: /geront/gnq032 Advance Access publication on May 3, 2010 Earnings Sharing in the U.S. Social Security System: A Microsimulation Analysis of Future Female Retirees Howard M. Iams, PhD,*,1 Gayle L. Reznik, PhD, 2 and Christopher R. Tamborini, PhD 2 1 Office of Research, Evaluation, and Statistics, U.S. Social Security Administration, Washington, DC. 2 Office of Retirement Policy, U.S. Social Security Administration, Washington, DC. *Address correspondence to Howard M. Iams, PhD, Office of Research, Evaluation, and Statistics, U.S. Social Security Administration, 500 E Street, S.W. Room 914, Washington, DC Howard.M.Iams@ssa.gov Received November 8, 2009; Accepted March 18, 2010 Decision Editor: William J. McAuley, PhD Purpose: As part of an ongoing effort to analyze the distributional implications of potential policy reforms to the U.S. Social Security system, we consider the widely discussed reform of earnings sharing. Such an approach has been viewed as a way to update Social Security s family benefits based on marital status and as a means to make the system more marriage neutral. Design and Methods: We use the Social Security Administration s Modeling Income in the Near Term model to simulate the distributional effects of two earnings sharing alternatives, a basic option and a modified inheritance option, on the projected retiree population in We focus our analysis on uncovering how different subgroups of women would be affected under the alternative system. Results: Widows and surviving divorced women, particularly those in lower socioeconomic groups, would face some of the steepest benefit decreases. Adding an inheritance provision would mitigate reductions across the board; however, it may not provide as much benefit adequacy to widows as might be expected. Implication: Certain groups of women would fare far better under the current Social Security system than under earnings sharing. Linking benefits with earnings sharing would be particularly problematic for lower income widows and single-earner married couples. Nevermarried persons are unaffected. Key Words: Social security, Earnings sharing, Women, Microsimulation As part of an ongoing effort to analyze the distributional implications of potential policy reforms to the U.S. Social Security system, this article considers the widely discussed option of earnings sharing. Social Security provides benefits to qualified retired workers and the spouses and survivors of these workers who may be eligible for an auxiliary benefit. Under earnings sharing, benefits would be based on the combined earnings records of spouses, which would be split equally during each year of marriage. Such an approach has been viewed as a means to make Social Security s family benefits more marriage neutral in the context of recent demographic changes (Favreault, Sammartino, & Steuerle, 2002; Favreault & Steuerle, 2007). This study s central goal was to assess how earnings sharing would affect the benefit adequacy of future retirees. Because Social Security is an especially important income source for aged women, many of whom continue to receive at least part of their benefit based on their marital status as wife or widow, a key question is how linking benefits with earnings sharing would affect different groups of female retirees. A key contribution of this study is the assessment of the impacts on women, by marital status, and with various levels of retirement income, lifetime earnings, educational attainment, and minority status. Although prior analyses suggest that widows would face particularly steep Vol. 50, No. 4,

2 reductions from earnings sharing, the literature has overlooked the impacts across different subgroups of women, such as widows with different economic resource levels. This is an important gap because policy changes to Social Security are likely to have varied consequences for the benefit adequacy of retirees. The study also sheds new light on differences in the way married women from one-earner and two-earner households would fare under the earnings sharing options. Using the Social Security Administration s (SSA) Modeling Income in the Near Term (MINT) microsimulation model, we simulate two earnings sharing scenarios: a basic design with no auxiliary spouse or survivor benefits and a modified design that allows inheritance of earnings by surviving spouses. The effects are estimated on the projected retiree population aged 62 years and older in 2030, the approximate year in which all baby boomers ( ) will reach their full Social Security retirement age. Policy Background: Earnings Sharing and Social Security The U.S. Social Security program is a social insurance system that provides monthly benefits to workers and their dependents when the worker retires, becomes disabled, or dies. Covered workers represent about 96% of the labor force and contribute to Social Security through payroll taxes. Over recent years, demographic changes in the population, concerns about benefit adequacy, coupled with long-range financial challenges to the Social Security system, have encouraged extensive debate about policy alternatives to Social Security s benefit structure. A wide range of options have been discussed, including providing credits for unpaid work caring for young children or other dependents (Herd, 2006, 2009), instituting a special minimum benefit for lifelong low earners (Favreault & Steuerle, 2007; Herd, 2005), shortening the marital duration requirements for divorced spouses (Tamborini & Whitman, in press), adding private individual accounts (Mitchell & Zeldes, 1996), and increasing the early eligibility age for benefits (Munnell, Meme, Jivan, & Cahill, 2004; Smith, 1999). Another commonly proposed reform, and the one in focus herein, is earnings sharing (Favreault & Steuerle; Forman, 2006; Iams, Reznik, & Tamborini, 2009). Part of the impetus behind earnings sharing is the simplification of Social Security s benefit structure. Under the current system, retirement benefits are based on an individual s earnings history in conjunction with their marital history and the earnings of current and previous spouses. To be eligible for a retired-worker benefit, typically a person must have worked at least 10 years (earning 40 quarters) in Social Security covered earnings. The benefit amount is based on the worker s Primary Insurance Amount (PIA), a function of the worker s average indexed monthly earnings based on the 35 highest years of covered earnings. Auxiliary spouse and survivor benefits provide monthly payments to qualified spouses and survivors of insured workers. To receive these benefits, no individual contribution is required from the auxiliary beneficiary. Spouses of retired workers may be eligible for a spouse benefit, equal up to 50% of the PIA of the living spouse s work record. Spouses of deceased workers may be eligible for a survivor benefit beginning at age 60 or age 50, if disabled, equal up to 100% of the deceased worker s benefit. Divorced individuals may also qualify for a spouse or survivor benefit based on the work record of their ex-spouse provided that the marriage lasted at least 10 years. A surviving spouse must be unmarried or have remarried at or after age 60 to qualify for a survivor benefit. Persons who qualify for both a retired-worker benefit based on their own earnings record and an auxiliary benefit (spouse or survivor) based on their spouse s earnings record receive the higher of the two. Should the retired-worker benefit equal less than the auxiliary benefit, the beneficiary is said to be dually entitled, meaning part of a spouse or survivor benefit is added to their own retired-worker benefit to reach the full auxiliary benefit. If the retired-worker benefit is higher than the auxiliary benefit, a person would receive their own retired-worker benefit. Useful summaries of the history of auxiliary benefits can be found in Berkowitz (2002) and DeWitt, Béland, and Berkowitz (2008). Incorporating earnings sharing into the Social Security program would alter the aforementioned computation of benefits. In its most basic form, earnings sharing would eliminate benefits linked to marital status. Individuals would be eligible for a retired-worker benefit based on their own lifetime earnings record, which would reflect half of the total combined couple earnings in each year of marriage and their own earnings in years not married. In the case of multiple marriages, the sharing would occur with different spouses. 496 The Gerontologist

3 The perceived desirability of earnings sharing can be understood against the dual historic goals of the Social Security program: balancing the promotion of benefit adequacy (i.e., providing basic protection from financial hardship in old age) with individual equity (i.e., linking benefits to contributions). In this light, earnings sharing proposals often intersect with broader discussions about whether Social Security can continue to provide adequate protection to the neediest retirees while treating groups equitably, particularly in the context of recent sociodemographic changes (Favreault & Steuerle, 2007). This concern has been underscored by recent evidence showing modest declines in the potential eligibility rates of future female retirees for spouse or widow benefits due to shorter marriages prior to divorce and lower marriage rates, particularly among Black women (Harrington Meyer, Wolf, & Himes, 2006; Tamborini, Iams, & Whitman, 2009). Insofar as declining auxiliary benefit eligibility is concentrated in groups with limited economic resources, channeling income support contingent on marriage, some claim, may be increasingly disadvantageous to some segments of the population (Harrington Meyer, 1996). Above all, earnings sharing is cast as a strategy to inject more equity into Social Security. The central issue is the treatment of one-earner married households relative to single persons and two-earner married households (Berkowitz, 2002; Steuerle & Bakija, 1994). Due to spouse and survivor benefits under the current system, total benefits paid to one-earner couples with the same total lifetime earnings as two-earner couples are generally higher (Iams et al., 2009; Table 1). This outcome, some argue, is inconsistent with the program s principle of equity. Moreover, insofar as one-earner households are characterized by one relatively high earner, the system may not be well targeted in terms of promoting benefit adequacy among the needy. Critics have argued, however, that earnings sharing would yield outcomes inconsistent with the goal of providing adequate benefits to workers and their families. Earnings sharing proposals, it is pointed out, would require policymakers to face tough trade-offs, such as benefit reductions for survivors. Another issue centers on the costly and difficult transition to, and the administration of, earnings sharing. A number of studies have evaluated the distributional implications of earning sharing over the past decades, particularly in the 1980s (Favreault & Steuerle, 2007; Fierst & Campbell, 1988; Flowers & Horowitz, 1993; Iams et al., 2009; U.S. Congressional Budget Office, 1986; U.S. Department of Health and Human Services [HHS], 1985; Zedlewski, 1984). Although varied, most studies agree that widowed men and women would experience some of the steepest benefit reductions under earnings sharing. As a result, most earnings sharing plans often contain add-on enhancements. Few existing studies, however, have focused on the potential effects of earnings sharing across different subgroups of women, as in the current study. The implications of policy alternatives to Social Security for women are of particular interest. Women are especially reliant on Social Security benefits for retirement income (Butrica & Iams, 2003; Smeeding, 1999; Williamson & Rix, 2000). Moreover, auxiliary benefits remain important for aged women despite dramatic increases in labor force participation and earnings over recent decades. In 2008, approximately 56% (12.8 million) of female beneficiaries aged 62 years and older received some type of benefit as a wife or widow (U.S. Social Security Administration, 2010; Table 5.A.14). Moreover, elderly women continue to experience relatively high poverty rates primarily because of their greater longevity and lower lifetime earnings than men (Herd, 2006). Estimates from the March Current Population Survey (CPS) put the 2008 poverty rate of women aged 65 years or older at 11.9%, almost double the 6.7% poverty rate of comparable men (U.S. Census Bureau, 2009). Data and Method The MINT Model Estimating the impact of social policy changes on individuals and their families often requires a sophisticated microsimulation data system (Martini & Trivellato, 1997; Mitton, Sutherland, & Weeks, 2000). Program complexity makes it particularly challenging to estimate changes to the U.S. Social Security system. Furthermore, because policy changes are almost always proposed for future retirees, analysis must capture socioeconomic and demographic changes that will affect the composition and expected retirement income of future retirees. The MINT microsimulation model is designed to meet these high data requirements. Developed by the SSA s Office of Research, Evaluation, and Statistics with assistance from the Brookings Institution, the RAND Corporation, and the Urban Institute, the MINT model uses Vol. 50, No. 4,

4 Table 1. Earnings Sharing Results for Individuals in 2030 Measure ES1 ES2 Average current law benefit $14,877 $14,877 Average benefit under policy $13,609 $14,980 Average individual percent 8 1 difference (%) a Individuals with no benefit change relative to current law Percent of population (%) Average current-law benefit $15,315 $15,608 Individuals with benefit increases relative to current law Percent of population (%) Average current-law benefit $14,401 $14,584 Average benefit under policy $15,512 $16,140 Average individual percent difference (%) a Individuals with benefit decreases relative to current law Percent of population (%) Average current-law benefit $15,011 $15,014 Average benefit under policy $12,439 $13,315 Average individual percent difference (%) a Total population (in thousands) 53,849 53,849 Note: ES1 = basic earnings sharing (no auxiliary benefits); ES2 = earnings sharing with inheritance of earnings for survivors. Includes married, divorced, or widowed individuals aged 62 years or older who receive benefits under current law. Married individuals are only included if the spouse receives benefits under current law. No change in benefits is defined as benefits within 1% of current law. Benefits that change by one percent or more above or below current law are defined as either increases or decreases. For divorced and widowed individuals, benefits are annual individual benefits. For married individuals, benefits are the per capita benefit (half the combined annual couple benefit). Benefit amounts are expressed in 2005 dollars. The results are based on authors calculations using Modeling Income in the Near Term. a Table 1 and the subsequent tables report the average individual percent difference to measure the change in benefits from current law. The average individual percent difference is produced by calculating the percent differences for each individual and then taking a weighted average of the individual values. This differs from the percent average difference, which involves summing the weighted benefit amounts of all the individuals in each subgroup and finding the percent difference of the summed benefits of that subgroup under the proposal. In calculating the percent average difference, those with relatively higher benefit amounts will have a greater effect on the average than those with relatively lower benefit amounts. For a distributional analysis of Social Security benefits, where the effects on low-earners and the progressivity of the program are of concern, the average individual percent difference better illustrates the distributional results for groups with low benefit amounts that might be lost using the percent average difference measure. See Iams, Reznik, and Tamborini (2009) for an analysis using the percent average difference. The results are qualitatively similar. observed and estimated population characteristics to project the demographic characteristics and expected economic status of future retirees. MINT begins with nationally representative microdata of respondents born between 1926 and 1972 from the Census Bureau s 1990 to 1993 and 1996 Survey of Income and Program Participation (SIPP), which are matched to SSA longitudinal earnings and benefit administrative data through For the baby-boom cohort in 2030, MINT uses survey data through the first third to first half of respondents lives and statistically projects their characteristics forward, adjusting for expected demographic and socioeconomic changes. To calculate benefits, MINT draws from a comprehensive Social Security benefit calculator built into the model. Accordingly, MINT data include the characteristics of current, former, and future spouses, most importantly their earnings and marital histories. Because MINT matches up spouses, researchers are able to calculate shared lifetime earnings as well as the total value of the couple s benefits under current or alternative benefit designs. All the measures and projections in MINT have undergone thorough benchmarking. More detailed information, including a series of validation and sensitivity tests for MINT projections, is presented in Toder and colleagues (2002) and Smith, Cashin, and Favreault (2005). Details on marriage projections are contained in Panis and Lillard (1999). Simulating Two Earnings Sharing Designs Using MINT, we simulate two earnings sharing designs. The first policy, ES1, is basic earnings sharing, which does not include any auxiliary spouse or survivor benefits. The Social Security earnings of married couples are equally divided during each year of marriage to calculate a retired-worker benefit for each spouse. In years when individuals are not married, earnings are based on the individual s own earnings record. If either spouse is still working, even after age 62 and after collecting Social Security benefits, the sharing of earnings still occurs. The second policy, ES2, includes an inheritance provision in which survivors can inherit the Social Security earnings of all deceased spouses, including deceased ex-spouses, in addition to their own earnings for each year of marriage. Thus, instead of counting half of the combined earnings for each year of marriage to a deceased spouse, survivors can count all combined earnings during each year of marriage to a deceased spouse. This option 498 The Gerontologist

5 attempts to address sharp benefit reductions among some survivors under basic earnings sharing and is similar to the provision modeled in HHS (1985). Note that another possible modification may be to retain survivor benefits but calculate those benefits under earnings sharing. However, as Iams and colleagues (2009) shows, keeping survivor benefits under a completely shared earnings framework would not result in substantially different outcomes than a basic design. Our analysis focuses on the impact of the policy scenarios on the Social Security benefits of women projected to be aged 62 years and older in We start at age 62 because that is the age at which individuals are entitled to receive retired-worker benefits. The selection of the year 2030 serves several purposes. It could be expected that large policy changes would be applied only to future retirees to permit adequate time to financial planning. The year 2030 reflects the approximate year when the baby-boom cohort will reach full retirement age for Social Security. Because the baby-boom cohort has been at the forefront of demographic changes in the American population, they would be affected by policy reforms differently than current retirees. The baby-boom cohort makes up about 70% of the analyzed retiree population. Individuals from pre-baby boom cohorts are also included in the analysis to allow for additional widows and widowers, a group greatly affected by earnings sharing. Note that our analysis focuses on the impact of earnings sharing at one point in time (in 2030) rather than on lifetime benefits. Although this is a more straightforward approach, a person s benefit can increase or decrease over time depending on changes in marital status and earnings. To gauge how different subgroups fare under earnings sharing, we analyze outcomes across an array of socioeconomic and demographic characteristics by marital status. Among married women, an important distinction is one-earner and twoearner couples. Observed characteristics within married, widowed, and divorced women subsets include benefit type, current retirement income, shared (combined) lifetime earnings, education, age, and race/ethnicity. Benefit type essentially reflects a proxy for relative lifetime earnings between spouses; they include retired worker (receives only a benefit based on one s own earnings record), spouse and worker (receives a retired-worker benefit and a partial aged spouse benefit), spouse only (receives an aged spouse benefit), survivor and worker (receives a retired-worker benefit and a partial aged survivor benefit), and survivor only (receives an aged survivor benefit). In essence, spouse-only or survivor-only beneficiaries did not have at least 10 years of covered employment during their working lives. Projected retirement income includes annual household earnings, asset income, Social Security, Supplemental Security Income (SSI), and income from nonspousal coresidents. Lifetime earnings are the present value of real (inflation-adjusted) shared lifetime total earnings, which is calculated using a discount rate equal to the effective rate for the Social Security trust funds. Education is based on the number of years of education. The results for the never-married population are not reported because they would not be affected by earnings sharing, by definition. Individuals and married couples projected to be eligible for disability benefits are also excluded because their earnings histories are incomplete. All beneficiaries are assumed to share earnings over their entire working and married lifetimes. Estimates are weighted to reflect the national population, and all income is set to 2005 dollars. The unweighted sample size is 45,514 individuals. It is important to point out that MINT, like any microsimulation model, faces certain limitations. Because MINT is based on the SIPP panels, the projections contain a margin of error associated with differences between the sample population and the actual population to which the projection is applied. There is also projection error related to MINT s assumptions of the future. Because MINT results are derived from a complex system of statistical equations projecting five independent panels of SIPP, sampling variation cannot be measured through normal estimation processes. Small differences in the results should be viewed with caution, and the actual effects could be somewhat different than presented here. Additionally, MINT does not assume any type of behavioral response to earnings sharing. Finally, it is outside the scope of the article to analyze the financial implications of earnings sharing on the Social Security Trust Fund, the economy, or the transition costs. Any modification to Social Security needs to consider such issues. Results Overall Change in Benefits Table 1 examines the options impact on the overall ever-married population aged 62 years and older in 2030 (currently married individuals are included if both spouses are receiving benefits Vol. 50, No. 4,

6 Table 2. Earnings Sharing Results for Individuals in 2030 by Sex and Marital Status ES1 ES2 a Measure Married Widowed Divorced Total Widowed Divorced Men Average benefit under policy $14,320 $13,288 $14,951 $14,336 $17,097 $15,894 Average individual percent difference (%) b Individuals with no benefit change relative to current law Percent of population (%) Average current-law benefit $14,733 $16,455 $16,374 $15,036 $17,951 $17,585 Individuals with benefit increases relative to current law Percent of population (%) Average benefit under policy $16,540 $15,129 $12,788 $16,233 $17,649 $15,413 Average individual percent difference (%) b Individuals with benefit decreases relative to current law Percent of population (%) Average benefit under policy $12,590 $13,178 $15,088 $13,283 $13,978 $15,716 Average individual percent difference (%) b Total population (in thousands) 14,298 1,968 3,702 19,967 1,968 3,702 Women Average benefit under policy $14,303 $11,702 $13,412 $13,181 $15,588 $14,866 Average individual percent difference (%) b Individuals with no benefit change relative to current law Percent of population (%) Average current-law benefit $14,701 $14,080 $17,302 $15,498 $15,931 $16,995 Individuals with benefit increases relative to current law Percent of population (%) Average benefit under policy $16,136 $14,024 $13,013 $15,062 $16,626 $15,037 Average individual percent difference (%) b Individuals with benefit decreases relative to current law Percent of population (%) Average benefit under policy $12,541 $11,525 $12,326 $11,964 $14,161 $13,552 Average individual percent difference (%) b Total population (in thousands) 14,530 12,136 7,217 33,882 12,136 7,217 Notes: ES1 = basic earnings sharing (no auxiliary benefits); ES2 = earnings sharing with inheritance of earnings for survivors. Includes only married, divorced, or widowed individuals aged 62 years or older who receive benefits under current law. Married individuals are only included if the spouse receives benefits under current law. No change in benefits is defined as benefits within 1% of current law. Benefits that change by 1% or more above or below current law are defined as either increases or decreases. For divorced and widowed individuals, benefits are annual individual benefits. For married individuals, benefits are the per capita benefit (half the combined annual couple benefit). Benefit amounts are expressed in 2005 dollars. Results are based on authors calculations using the MINT model. a The results for ES2 are not reported for married individuals because they generally do not qualify for a survivor benefit. Married individuals can receive a survivor benefit from a deceased spouse if the remarriage occurred past age 60. b The average individual percent difference is produced by calculating the percent differences for each individual and then taking a weighted average of the individual values. under current law). Estimates indicate that ES1 would lead to an average overall benefit decrease of 8% of current law, whereas ES2 would lead to an average benefit increase of 1%. As might be expected, a higher share of retirees would receive reduced benefits under ES1 (61%) than under ES2 (39%). Under ES2, the percentage receiving benefit increases would rise to 49% due to permitting aged beneficiaries to inherit the earnings of their deceased spouse. Table 2 separates out the effects for men and women by marital group. Note that because MINT can track lifetime earnings of both husband and wife, estimates for married individuals reflect 500 The Gerontologist

7 changes in their couple benefits rather than individual benefits. This is an important distinction given that earnings sharing typically results in the transfer of earnings from husband to wife as a result of women s historically lower lifetime earnings. Thus, a married person may individually have lower benefits under earnings sharing, but the couple may have higher benefits. To make estimates for married individuals more comparable with other groups, the couple benefit is halved and referred to as the per capita benefit. For ES1, the overall percentage of women and men experiencing benefit reductions would be relatively equal; however, among those receiving lower benefits, the average percentage reduction would be higher for women than for men (21% and 11%, respectively). As expected, basic earnings sharing would be particularly problematic for widows and widowers. Under ES1, benefits are projected to decrease for 93% of widows and for 96% of widowers by an average percentage difference of 27% and 20%, respectively. Widows would receive such sharp decreases because under current law, the death of a husband initiates a survivor benefit of up to 100% of the deceased husband s benefit based on unshared earnings. Assuming that the husband was the higher earner, earnings sharing would reduce the earnings credited to the husband and increase the earnings credited to the wife. However, each spouse s shared earnings would be lower than the total unshared earnings credited to the husband under current law. The addition of an inheritance provision (ES2) would substantially reduce the percentage of widows and widowers with benefit decreases. Nevertheless, even with this enhancement, 17% of widowers and 40% of widows would receive reduced benefits. Many widows would continue to experience benefit decreases because inherited earnings would only come from years of marriage, and the deceased husband could have higher earnings than the wife in years before marriage. These unmarried earnings contribute to the currentlaw survivor benefit but would be omitted from the inherited survivor benefit. Earnings sharing would also affect divorced men and women. Divorced men would fare, on average, worse than divorced women. Divorced women would fare much better under ES2 than under ES1 due to the fact that many divorced women receive survivor benefits based on a deceased ex-husband s record. Such women, as later analysis will show, would therefore fare much better under an approach that includes some type of add-on inheritance provision. Benefit Change Within Different Subgroups of Women Table 3 shows estimates of ES1 on married women across various demographic and socioeconomic characteristics. The small numbers of women who are not affected by earnings sharing are not included in Tables 3 5; consequently, the percentages do not necessarily sum to 100%. Married women who did not have 10 years of covered employment (spouse only) would be the hardest hit benefit type under earnings sharing; 73% would receive reduced benefits by an average of 27%. This would occur because the total couple benefit under ES1 would be based on total combined earnings during marriage, plus any earnings outside of marriage. In contrast, under current law, the total benefit for these women is based on half of the husband s unshared retired-worker benefit along with the retired-worker benefit of the husband. Analysis of socioeconomic characteristics highlights whether earnings sharing yields progressive outcomes for married women. The pattern is generally similar for the three measures; married women in the lowest quintiles of income and lifetime earnings and with less than 12 years of education would have the highest prevalence and percentage reduction in benefits. In other words, earnings sharing would be most problematic for lower income married couples, suggesting that current-law benefits are more progressive for married couples. In general, the impact of ES1 would be only modestly different for married women by age. However, a markedly higher percentage of the oldest married women (aged 90 years and older) would receive benefit decreases. This is related to the lower labor force participation among women in this age group. In terms of race/ethnicity, White married women would be somewhat more likely to have benefit increases and less likely to have benefit reductions compared with their Black and Hispanic counterparts. Figures 1 and 2 depict the effects of earnings sharing on married women by household earner status. Our one-earner and two-earner designation is defined by the current-law benefit types of the husband and wife. In one-earner couples, one spouse receives a retired-worker benefit and the other a Vol. 50, No. 4,

8 Table 3. Earnings Sharing Results for Married Women in 2030 by Demographic Characteristics ES1 Individuals with benefit increases Individuals with benefit decreases Demographic characteristics Total population (in thousands) Percent of population Average individual percent difference a Percent of population Average individual percent difference a All married women 14, Current-law benefit type b Retired worker 9, Spouse and worker (dually entitled) 3, Spouse only 1, Current per capita income quintiles (current law) Highest quintile 2, Second highest quintile 2, Middle quintile 3, Second lowest quintile 2, Lowest quintile 2, Present value shared lifetime earnings quintiles (current law) c Highest quintile 3, Second highest quintile 3, Middle quintile 2, Second lowest quintile 2, Lowest quintile 2, Education Graduate 1, Bachelor 2, Associate 2, High school 6, Less than 12 years 1, Age (years) , , , Race/ethnicity White 11, Black Hispanic 1, Other Notes: ES1 = basic earnings sharing (no auxiliary benefits); ES2 = earnings sharing with inheritance of earnings for survivors. Includes married women where both the husband and the wife receive benefits under current law. Benefits are the per capita benefit (half the combined annual couple benefit). Quintiles are defined for married (if the husband also receives benefits under current law), divorced, and widowed women (aged 62 years and older) in 2030 who have benefits under current law. Benefits that change by 1% or more above or below current law are defined as either increases or decreases. Benefit amounts are expressed in 2005 dollars. Results are based on authors calculations using the MINT model. a The average individual percent difference is produced by calculating the percent differences for each individual and then taking a weighted average of the individual values. b The table does not report separately the small group of married women with survivor and worker (dually entitled) or survivoronly current-law benefit types; however, these groups are not excluded from the other rows of the table. A married woman must remarry after age 60 to qualify for these benefit types. c The present value is calculated using a discount rate equal to the effective rate for the Social Security trust funds. spouse benefit. Two-earner couples are separated into couples in which both the husband and the wife receive retired-worker benefits and couples in which one spouse receives a retired-worker benefit and the other a dually entitled spouse benefit. For computational reasons, Figures 1 and 2 are restricted to married women in their first marriage with both spouses receiving current-law benefits. To recall, a major rationale for earnings sharing is to improve equity between one-earner and 502 The Gerontologist

9 Figure 1. Earnings sharing results under ES1 for married women aged 62 years and older in 2030, by earner status of couple. Notes. ES1 = basic earnings sharing (no auxiliary benefits). Includes married women from first marriages only, where both the husband and the wife receive benefits under current law. Benefits are the per capita benefit (half the combined annual couple benefit). Benefits that change by 1% or more above or below current law are defined as either increases or decreases. Benefit amounts are expressed in 2005 dollars. Results are based on authors calculations using the MINT model. two-earner married couples. Our results confirm that earnings sharing would improve equity insofar as one-earner couples would be more likely to receive benefit decreases and relatively greater reductions due to the elimination of the spouse benefit. Additionally, couple benefits would increase for about 40% of married women in the two-earner status compared with 23% in the one-earner status. Table 4 addresses how different types of widows would fare under the options. Across benefit types, estimates show that almost all dually entitled and survivor-only widows, the vast majority of widows, would receive reduced benefits under ES1. Survivoronly widows (i.e., women without 10 years in covered employment) would experience some of the most dramatic reductions. Under current law, these widows receive up to 100% of their husband s benefit based completely on his unshared earnings, but under ES1, they would receive a benefit based on half of their husband s unshared earnings during years of marriage. Because retired-worker and dually entitled widows have their own earnings, the effects of basic earnings sharing would not be as dramatic. Moving to socioeconomic status, we observe that earnings sharing would be particularly problematic Figure 2. Average individual percent difference in total couple benefit under ES1 among married women aged 62 years and older in 2030, by earner status of couple. Notes. ES1 = basic earnings sharing (no auxiliary benefits). Includes married women from first marriages only, where both the husband and the wife receive benefits under current law. Benefits are the per capita benefit (half the combined annual couple benefit). Benefits that change by 1% or more above or below current law are defined as either increases or decreases. Benefit amounts are expressed in 2005 dollars. Results are based on authors calculations using the MINT model. Vol. 50, No. 4,

10 Table 4. Earnings Sharing Results for Widowed Women in 2030 by Demographic Characteristics ES1 ES2 Demographic characteristics Total population (in thousands) Individuals with benefit increases Individuals with benefit decreases Individuals with benefit increases Individuals with benefit decreases Percent of population Average individual Percent of percent difference a population Average individual Percent of percent difference a population Average individual Percent of percent difference a population Average individual percent difference a All widowed women 12, Current-law benefit type b Retired worker 2, Survivor and worker (dually entitled) 8, Survivor only 1, Current individual income quintiles (current law) Highest quintile 2, Second highest quintile 2, Middle quintile 2, Second lowest quintile 2, Lowest quintile 2, Present value shared lifetime earnings quintiles (current law) c Highest quintile 1, Second highest quintile 2, Middle quintile 2, Second lowest quintile 2, Lowest quintile 2, Education Graduate 1, Bachelor 1, Associate 1, High school 5, Less than 12 years 1, Age (years) , , , , Race/ethnicity White 9, Black Hispanic 1, Other Notes: Includes widowed women who receive benefits under current law. Benefits are annual individual benefits. Quintiles are defined for married (if the husband also receives benefits under current law), divorced, and widowed women (aged 62 years and older) in 2030 who have benefits under current law. Benefits that change by 1% or more above or below current law are defined as either increases or decreases. Benefit amounts are expressed in 2005 dollars. Results are based on authors calculations using the MINT model. ES1 = basic earnings sharing (no auxiliary benefits); ES2 = earnings sharing with inheritance of earnings for survivors; = sample size is too small to report. a The average individual percent difference is produced by calculating the percent differences for each individual and then taking a weighted average of the individual values. b The table does not report separately the small group of widowed women with spouse and worker (dually entitled) current-law benefit type; however, this group is not excluded from the other rows of the table. c The present value is calculated using a discount rate equal to the effective rate for the Social Security trust funds. 504 The Gerontologist

11 Table 5. Earnings Sharing Results for Divorced Women in 2030 by Demographic Characteristics ES1 ES2 Demographic characteristics Total population (in thousands) Individuals with benefit increases Individuals with benefit decreases Individuals with benefit increases Individuals with benefit decreases Percent of population Average individual Percent of percent difference a population Average individual Percent of percent difference a population Average individual Percent of percent difference a population Average individual percent difference a All divorced women 7, Current-law benefit type Retired worker 4, Spouse and worker (dually entitled) Spouse only Survivor and worker (dually entitled) 2, Survivor only Current individual income quintiles (current law) Highest quintile 1, Second highest quintile 1, Middle quintile 1, Second lowest quintile 1, Lowest quintile 1, Present value shared lifetime earnings quintiles (current law) b Highest quintile 1, Second highest quintile 1, Middle quintile 1, Second lowest quintile 1, Lowest quintile 1, Education Graduate Bachelor 1, Associate 1, High school 3, Less than 12 years Age (years) , , , Race/ethnicity White 5, Black Hispanic Other Notes: Includes divorced women who receive benefits under current law. Benefits are annual individual benefits. Quintiles are defined for married (if the husband also receives benefits under current law), divorced, and widowed women (aged 62 years and older) in 2030 who have benefits under current law. Benefits that change by 1% or more above or below current law are defined as either increases or decreases. Benefit amounts are expressed in 2005 dollars. Results are based on authors calculations using the MINT model. ES1 = basic earnings sharing (no auxiliary benefits); ES2 = earnings sharing with inheritance of earnings for survivors; = sample size is too small to report. a The average individual percent difference is produced by calculating the percent differences for each individual and then taking a weighted average of the individual values. b The present value is calculated using a discount rate equal to the effective rate for the Social Security trust funds. Vol. 50, No. 4,

12 for low-income widows. Among those who receive lower benefits under ES2, the average percentage reduction would be far greater for low-income widows. This reduction would likely be detrimental to the retirement income security of these retirees because Social Security generally makes up a higher share of the total retirement income of low-income groups. The results for education are similar; the average percentage reduction among those with lower benefits would be greater for less educated widows. Our microsimulation results also indicate that younger widows would be slightly better off, on average, than older widows under both options. This is largely a function of younger cohorts having higher labor force participation rates during their working lives. Additionally, under ES2, Black and Hispanic widows would have slightly less propensity to experience decreases than White widows, but the average decrease among those receiving reductions is markedly greater. Table 5 looks at different subgroups of divorced women. Under current law, many divorced women are eligible for a divorced survivor benefit based on the earnings record of a deceased ex-husband. ES2 is therefore particularly important to consider. The results confirm the importance of current-law benefit type. In general, divorced women with their own worker benefits or a living ex-spouse would fare better than their counterparts without worker benefits and with a deceased ex-spouse. A relatively large percentage of divorced women fare well because under earnings sharing, the earnings credits received by divorced women during years of marriage would be portable, regardless of the length of marriage. In essence, earnings sharing would allow divorced women who have marriages of less than 10 years to count the earnings of their ex-husbands during years of marriage. However, most divorced women with deceased ex-husbands would experience benefit declines under ES1, and the reductions among those receiving lower benefits would be, on average, 28% for dually entitled survivors and 46% for survivors only. This pattern results from the fact that many divorced women switch from a retired-worker benefit to a dually entitled survivor benefit upon the death of their ex-husband (Butrica & Iams, 2000). This relationship is observed in the large mitigating effect of ES2 on the reductions experienced by divorced surviving spouses. The effects of both earnings sharing options on divorced women with different characteristics are mixed. On the one hand, less educated divorced women would be slightly more likely to receive reduced benefits, but those in the lowest income and lifetime earnings quintiles would be slightly more likely to receive increased benefits than their higher income counterparts. On the other hand, these women would average sharply greater percentage reductions than the higher education and income quintiles when they receive reduced benefits. With respect to age, older divorced women would be more likely to receive benefit decreases. This is not surprising given that older divorced women are more likely to have a deceased ex-husband and therefore have a higher probability of receiving a divorced survivor benefit under current law. Relative to other race/ethnicity groups, Hispanic divorced women would be slightly more likely to receive benefit decreases under ES1, and under ES2, they would experience modestly larger benefit reductions among those receiving lower benefits. Discussion Concerns over benefit adequacy and equity, demographic changes in the U.S. population, and long-range finances, have prompted a number of policy proposals to reform the Social Security program. This article extends existing research on one such proposal, earnings sharing, by using the MINT microsimulation model to assess the policy s impact on future female retirees by marital status and with different economic resource levels, educational attainment, and minority status. Overall, our results underscore the notion that structural reforms to Social Security have important, but often varying, impacts on the benefit adequacy of different groups of retirees. Several themes encapsulate our findings. First, consistent with prior research, married women from one-earner households and surviving spouses would experience some of the steepest benefit reductions if benefits were linked to earnings sharing. Second, adding an inheritance provision helps mitigate the incidence and magnitude of these reductions; however, it does not protect as many surviving spouses as might be expected. Third, Social Security s current system of family benefits linked to marital status would provide substantially greater retirement income for many future female retirees than benefits linked to earnings sharing. Earnings sharing, our results suggest, would exacerbate economic hardship among some of the neediest widows and divorced women. Even 506 The Gerontologist

13 with inheritance of earnings, benefits under earnings sharing would often be an incomplete substitute for current-law survivor benefits received by many low-income widows. Moreover, due in part to having lower lifetime earnings, Black and Hispanic widows experiencing benefit decreases under earnings sharing with inheritance would experience particularly large reductions. These results suggest several policy insights. Earnings sharing options would generally improve individual equity by providing a more direct link between earnings histories and benefits than the current system. In our analysis, one way this can be observed is through the impact of the policy on married women by household earner status. In general, married women in two-earner households would fare better under earnings sharing than those in one-earner households due to changes in spouse and survivor benefits. However, at the same time, earnings sharing would reduce benefit adequacy, that is, whether benefits provide basic protection from financial hardship in old age. Our results show that some of the most economically vulnerable subgroups of women, such as low-income widows, would receive sharp benefit reductions under earnings sharing. One way to protect economically vulnerable groups would be to expand the current SSI program for the aged population, a program whose resource and income limits have not been changed in two decades. Other proposals may include a minimum benefit for very low income retirees and increased survivor benefits for the oldest beneficiaries. Finally, the study provides one example of how microsimulation methodology can be a useful approach to help assess the distributional implications of hypothetical changes to social policy. The use of microdata in the MINT model allows estimates of subgroups often overlooked in analyses using stylized individuals and thus permits a better representation of the complex economic and social lives of the population. However, the limitations associated with microsimulation methodology are important to keep in mind. In particular, our results are sensitive to the model s assumptions of the future. Because MINT s assumptions are fixed, extraordinarily dramatic structural changes over time, such as a long economic downturn characterized by high unemployment and wage stagnation, could affect projections 20 years forward. The 2008 economic downturn, however, should not substantively alter the estimated relative distributional effects. Analysis by Gustman, Steinmeier, and Tabatabai (2009) suggests that the impact of the 2008 economic downturn on the retirement income security of many early baby boomers is not likely to be substantial, in large part, because the majority of their retirement wealth is Social Security wealth (a function of the 35 highest years of covered employment) and in defined benefit pensions rather than stocks. Moreover, as previously noted, MINT does not assume a behavioral response to policy changes. However, we suspect that the modest changes overall in benefits from earnings sharing would prompt only minor shifts in behavior. Policymakers would also need to consider other factors not addressed in this article, such as the long-range cost of policy alternatives. At a time of considerable focus on ways to address Social Security s long-term financing challenges, further analysis of other potential policy reforms would be valuable. Given the potential escalation of retirement risks in the future due to the erosion of the traditional defined benefit plan, relatively low retirement savings rates, and possibly a sustained economic downturn, developing a better understanding of the distributional implications of alternative Social Security benefit structures is increasingly relevant. Acknowledgments The findings and conclusions presented in this article are those of the authors and do not represent the views or opinions of the U.S. Social Security Administration or any agency of the federal government. The MINT model data are restricted for use; all users must receive approval of the Social Security Administration and the U.S. Census Bureau. We thank the Editor and the anonymous reviewers of the Gerontologist for helpful comments. Thanks also to Benjamin Bridges, Dave Shoffner, Lee Cohen, Susan Grad, Jason Fichtner, and David Weaver for comments. An early version of this article was presented at the 2009 General Conference of the International Microsimulation Association, Ottawa, Canada. References Berkowitz, E. D. (2002). Family benefits in Social Security: A historical commentary. In M. M. Favreault, F. J. Sammartino, & C. E. Steuerle (Eds.), Social Security and the family: Addressing unmet needs in an underfunded system (pp ). Washington, DC: Urban Institute Press. Butrica, B. A., & Iams, H. M. (2000). Divorced women at retirement: Projections of economic well-being in the near future. Social Security Bulletin, 63(3), Butrica, B. A., & Iams, H. M. (2003). The impact of minority group status on the projected retirement income of divorced women in the baby boom cohort. Journal of Women & Aging, 15(2/3), Dewitt, L. W., Béland, D., & Berkowitz, E. D. (2008). Social Security: A documentary history. Washington, DC: Congressional Quarterly Press. Favreault, M. M., Sammartino, F. J., & Steuerle, C. E. (2002). Social Security benefits for spouses and survivors: Options for change. In M. M. Favreault, F. J. Sammartino, & C. E. Steuerle (Eds.), Social Security and the family: Addressing unmet needs in an underfunded system (pp ). Washington, DC: Urban Institute Press. Favreault, M. M., & Steuerle, C. E. (2007). Social Security spouse and survivor benefits for the modern family. (Retirement Project Discussion Paper No ). Washington, DC: Urban Institute Press. Fierst, E. U., & Campbell, N. D. (Eds.). (1988). Earnings sharing in Social Security: A model for reform: Report of the Technical Committee on Earnings Sharing. Washington, DC: Center for Women Policy Studies. Vol. 50, No. 4,

14 Flowers, M. R., & Horowitz, J. B. (1993). Distributional implications of the Social Security spouse benefit. Southern Economic Journal, 60, Forman, J. B. (2006). Making America work. Washington, DC: Urban Institute Press. Gustman, A. L., Steinmeier, T. L., & Tabatabai, N. (2009). What the stock market decline means for the financial security and retirement choices of the near-retirement population. (NBER Working Paper No ). Cambridge, MA: National Bureau of Economic Research. Harrington Meyer, M. (1996). Making claims as workers or wives: the distribution of Social Security benefits. American Sociological Review, 61, Harrington Meyer, M., Wolf, D. A., & Himes, C. L. (2006). Declining eligibility for Social Security spouse and widow benefits in the United States? Research on Aging, 28, Herd, P. (2005). Ensuring a minimum: Social Security reform and women. The Gerontologist, 45, Herd, P. (2006). Crediting care or marriage? Reforming Social Security family benefits. Journal of Gerontology: Social Sciences, 61B, S24 S34. Herd, P. (2009). Crediting care in Social Security: A proposal for an income tested care credit (NASI Working Paper). Washington, DC: National Academy of Social Insurance. Iams, H. M., Reznik, G. L., & Tamborini, C. R. (2009). The impact of three earnings sharing scenarios using the MINT model. Social Security Bulletin, 69(1), Martini, A., & Trivellato, U. (1997). The role of survey data in microsimulation models for social policy analysis. Labour, 11(1), Mitchell, O., & Zeldes, S. (1996). Social Security privatization: A structure for analysis. American Economic Review, 86, Mitton, L., Sutherland, H., & Weeks, M. (Eds.). 2000). Microsimulation modelling for policy analysis. Cambridge, England: Cambridge University Press. Munnell, A. H., Meme, K. B., Jivan, N., & Cahill, K. E. (2004). Should we raise Social Security s earliest eligibility age? (Issue in Brief 18). Chestnut Hill, MA: Center for Retirement Research at Boston College. Panis, C., & Lillard, L. (1999). Near-term model development part II. (Final Report, SSA Contract No ). Santa Monica, CA: RAND. Smeeding, T. M. (1999). Social Security reform: Improving benefit adequacy and economic security for women. (Working Paper No. 16). Chestnut Hill, MA: Center for Retirement Research at Boston College. Smith, K. E., Cashin, D. B., & Favreault, M. M. (2005). Modeling income in the near term 4. (Final Report, SSA Contract No ). Washington, DC: Urban Institute. Smith, R. (1999). Raising the earliest eligibility age for Social Security benefits. (CBO Papers). Washington, DC: Congressional Budget Office. Steuerle, C. E., & Bakija, J. M. (1994). Retooling Social Security for the 21st century: Right and wrong approaches to reform. Washington, DC: Urban Institute Press. Tamborini, C. R., Iams, H. M., & Whitman, K. (2009). Marital history, race, and Social Security spouse and widow benefit eligibility in the United States. Research on Aging, 31, Tamborini, C. R., & Whitman, K. (in press). Lowering Social Security s duration-of-marriage requirement: Distributional effects for future female retirees. Journal of Women & Aging. Toder, E., Thompson, L., Favreault, M., Johnson, R., Pevese, K., Ratcliffe, C., et al. (2002). Modeling income in the near term: Revised projections of retirement income through 2020 for the birth cohorts. (Final Report, SSA Contract No ). Washington, DC: Urban Institute Press. U.S. Census Bureau. (2009). Current population survey, 2009 annual social and economic supplement. Retrieved October 22, 2009, from U.S. Congressional Budget Office. (1986). Earnings sharing options for the Social Security system. Washington, DC: Author. U.S. Department of Health and Human Services, Social Security Administration. (1985). Report on earnings sharing implementation study (House Committee on Ways and Means Print No. 99-4). Washington, DC: Government Printing Office. U.S. Social Security Administration. (2010). Annual statistical supplement to the Social Security bulletin, Washington, DC: Author. Williamson, J. B., & Rix, S. E. (2000). Social Security reform: Implications for women. Journal of Aging & Social Policy, 11(4), Zedlewski, S. R. (1984). The distributional consequences of an earnings sharing proposal (Project Report No. 3344). Washington, DC: Urban Institute Press. 508 The Gerontologist