Ensuring a Minimum: Social Security Reform and Women

Transcription

1 The Gerontologist Vol. 45, No. 1, Copyright 2005 by The Gerontological Society of America Ensuring a Minimum: Social Security Reform and Women Pamela Herd, PhD 1 Purpose: The potential effects of implementing three different minimum benefits in Social Security, which have accompanied proposals to privatize the program and reform family benefits, are examined in relation to the adequacy of benefits for women reaching age 62 between 2020 and Design and Methods: The 1992 Health and Retirement Study is used to conduct a simplified microsimulation. Results: The minimum benefit proposal accompanying privatization proposals, which requires 40 earnings years for a poverty level benefit, fails to cover significant numbers of vulnerable women. The elimination of spousal benefits, criticized for being outdated and regressive, helps offset the costs of more generous minimum benefits, such as those that require residency or 10 earnings years for eligibility. Implications: Noncontributory benefits distributed based on marital status are not as effective at protecting poorer women, as well as a new generation of women that is less likely to be married, than are minimum benefits where eligibility is tied to U.S. residency or simply Social Security eligibility. Key Words: Gender, Retirement, Income supports Concerns about the risks associated with individual accounts and the problems with the current family benefit structure have prompted politicians and academics to suggest the reintroduction of a minimum benefit into Social Security. Older women are a cause for concern, and impetus for reform, because Social The research reported herein was supported (in part) by the Center for Retirement Research at Boston College pursuant to a grant from the U.S. Social Security Administration funded as part of the Retirement Research Consortium. The opinions and conclusions are solely those of the author and should not be construed as representing the opinions or policy of the Social Security Administration or any agency of the federal government, or the Center for Retirement Research at Boston College. The author would like to acknowledge Madonna Harrington Meyer, Doug Wolf, Tim Smeeding, Brian Goesling, and Don Moynihan for their thoughtful comments. Address correspondence to Dr. Pamela Herd, LBJ School of Public Affairs, University of Texas, Austin, PO Box 4, Austin, TX pherd@mail.utexas.edu 1 LBJ School of Public Affairs, University of Texas, Austin. Security comprises 60% of their income, and for one in five women, all of their income. Moreover, 12% of older women compared to 7% of older men are poor (Smeeding, Estes, & Glasse, 1999). Yet, while significant attention is paid to the gender gap in poverty, even more striking are the poverty gaps among women. Approximately 25% of elderly Black women live below the poverty line compared to 11% of White older women. Relative to 4% of older married women, 20% of unmarried older women are poor (Proctor & Dalaker, 2002). A new minimum benefit would affect these women most because they rely most heavily on Social Security for retirement income. But how should minimum benefits be constructed? This study examines the three most common proposals. Each option reflects a larger ideological presumption about how the welfare state should distribute benefits. Should policies target certain groups or be universal and available to all residents? One minimum benefit is tightly targeted to the employed poor, one lies in the middle of the continuum requiring 10 earnings years, whereas the last is universal and requires only residency in the United States. The ways in which these proposals would affect the benefits of women reaching age 62 between 2020 and 2030 are evaluated. Given that benefit adequacy (protecting the neediest beneficiaries) is arguably Social Security s most important goal, this analysis centers on how the reforms affect Social Security s progressiveness. But attention is also paid to how these proposals affect program costs and benefit equity, i.e., the tie between an individual s contribution to the program through her employment and the benefits she ultimately receives. This link between earnings and benefits has helped keep the program popular. The Theoretical Background Should policies be targeted or universal? Should only targeted groups be eligible? Or should all residents be eligible? Income guarantees, under which minimum benefits fall, vary in the extent to which they are targeted or universal. The minimum 12 The Gerontologist

2 benefit accompanying proposals to privatize Social Security targets the employed poor by requiring 40 earnings years to qualify for a poverty-level benefit. At the other extreme, minimum benefits are distributed universally, where all residents are eligible for them. In particular, they require no earnings history for eligibility. This is common to the Swedish welfare state, and some argue should be adopted by the United States. Countries that have more universal policies have lower poverty rates than countries with targeted policies (Korpi & Palme, 1998). Others argue that universalism is inefficient, distributing resources to many people who do not need them. This study explores how these approaches to minimum benefits would affect women s benefits. In particular, how do they affect the progressiveness of women s benefits? Universal approaches are often beneficial for women because they work and earn less as a result of workplace discrimination and their disproportionate responsibility for raising children (Korpi, 2000). But would a universal benefit simply prove to be a boon for wealthy women who do not need to work? Contrastingly, would targeted approaches be enough of a boon for poor women? Do poor women work enough to qualify for a benefit that requires 40 earnings years? Many argue that poor mothers face significant difficulties balancing work and family (Edin & Lein, 1997). Their lowpaying jobs fail to provide them the money, time, and flexibility to meet conflicting demands, thus often forcing them out of work altogether. In order to lend insights into these questions and issues that is, the ramifications of targeting versus universalism each proposal represents a point on the targeted universal continuum. But before detailing the proposals, how did minimum benefits end up on the Social Security reform agenda in the first place? The Policy Background While Social Security does not currently have a minimum benefit, this was not always the case. The 1939 amendments, which also fashioned spousal and survivor benefits, added a minimum benefit. Though never generous, it was eliminated in the 1981 Social Security amendments for budgetary reasons and claims that it was benefiting housewives and individuals with limited work histories (Government Accounting Office [GAO], 1979). Amendments to the program in 1972 instituted two measures to protect the low-income elderly population, but each has serious shortcomings. The first was a special minimum benefit for which few qualify because it requires such a long, low earnings history (Congressional Research Service, 2001). The second created Supplemental Security Income (SSI) to provide a safety net for poor elderly and disabled people. However, federal benefits are below the poverty line (though some states supplement them), payments are reduced significantly if the recipient lives with someone, and only a few thousand dollars are allowed in savings. But the most damning critique is that only about half of those eligible for SSI receive benefits (McGarry, 2000). Reintroducing a minimum benefit was suggested in the late 1990s when privatization proposals were gaining credibility. Critics of partial privatization argued that individual accounts would reduce benefit adequacy. As payroll taxes are shifted to individual accounts, there is no opportunity for redistribution of the resources within individual accounts from high-earner to low-earner beneficiaries, many of whom are women (Williamson & Rix, 1999). Further, a bear market would hurt those who rely primarily on Social Security as their retirement income (Porter, Larin, & Primus, 1999). Proponents of privatization responded by including a minimum benefit within their proposals (see Kolbe-Stenholm Plan, H.R [2001]; President s Commission to Strengthen Social Security, 2001). Calls to reintroduce a minimum benefit have not only accompanied partial privatization proposals. Since the 1970s they have been mentioned in debates about how to update spousal and survivor benefits (GAO, 1996; Harrington Meyer, 1996). Currently, individuals receive either contributory benefits as workers, based on their highest 35 earnings years, or non-contributory benefits as the wives and widows of workers, which require no earnings history. A woman s spousal benefit is half her husband s worker benefit, and, should he die, her survivor benefit is 100% of his benefit. Two-thirds of women receive family benefits, and women comprise 98% of family benefit recipients. Called a breadwinner benefit, the women who fare best are married and are never employed. Of course, very few women fall into this category today (GAO, 1996). Noncontributory benefits aim to improve benefit adequacy by subsidizing married women s exits from work. Contrastingly, while the worker benefit is progressive, replacing a higher percentage of income for low earners, it places a heavier emphasis on benefit equity by tying eligibility and benefit levels to earnings (Harrington Meyer, 1996). While the current critique of breadwinner family benefits is that they have not kept up with changes in the American family, including more women working and fewer women marrying, one overlooked problem is that spousal and survivor benefits cannot protect many poor American women. This is particularly true for many unmarried mothers, who comprise a growing proportion of the population; the percentage of single mothers rose from 12% in 1970 to 26% in 2000 (Bachu & O Connell, 2000). Those who most need the supplementary income, poor and Black women, are less likely to marry and be eligible for noncontributory benefits (Goldstein & Kenney, 2001). Spousal benefits also offset the progressive worker benefit. Because they require no earnings Vol. 45, No. 1,

3 history, a woman who worked in low-wage jobs for most of her life may receive a smaller benefit than a woman who was never employed but whose husband had substantial earnings. Many Black and poor women, who are more likely to be employed, are effectively subsidizing the noncontributory benefits of upper income women who do no paid work (Gustman & Steinmeier, 2000; Harrington Meyer, 1996). Reform Options There are three general approaches to reforming women s benefits: altering spousal and survivor benefits to better protect unmarried women; introducing care credits to reward women for raising children; and introducing a minimum benefit. Each of these general options, in turn, could be implemented in numerous ways. Given the range of potential reforms, only minimum benefits are examined in this article. The minimum benefits are based on the three most common reform proposals. Each proposal falls along the targeted universal continuum and addresses some of the problems discussed above. Privatization Minimum The first proposal targets the employed poor (President s Commission to Strengthen Social Security, 2001). The proposal, which has gained attention through its association with privatization proposals, attempts to enhance benefit adequacy with a minimum benefit that maintains a strong emphasis on benefit equity. Despite this emphasis on equity, the reform does not alter noncontributory marital status benefits, which have no link between an individual s employment history and benefit. It requires 20 earnings years to guarantee a benefit equal to 60% of the poverty line. For each additional year the benefit rises 2 percentage points until 40 earnings years, where the benefit equals the poverty line (Kolbe-Stenholm plan, H.R. 1793). The generosity of the minimum will ultimately determine how women fare. For example, wage indexing the poverty level would produce a higher number of benefit gainers than will this proposal (see Favreault et al., 2002). The next two minimum benefits move away from breadwinner family benefits by eliminating noncontributory spousal benefits and shifting these revenues to subsidize the costs of minimum benefits that are more universal than the Privatization minimum. Worker Minimum This proposal is neither targeted nor exactly universal. It reintroduces the minimum benefit eliminated in 1981 but makes it more generous. Eligibility requires 10 years of earnings, the qualification for the Social Security worker benefit. The benefit equals the federal poverty level, which in 2001 was $751 a month for one-person households with a head age 65þ. Spousal benefits are eliminated. Resident Minimum This proposal is purely universal. Any older U.S. resident is eligible (Advisory Council on Social Security, 1979). Unlike the Worker minimum, eligibility for Social Security is not required. The benefit is set at the federal SSI level, which was $531 a month in Spousal benefits are eliminated. In regards to other work on the topic, Faverault, Sammartino, and Steuerle (2002) analyze a number of different approaches to instituting minimum benefits. One key difference is that they evaluate progressivity in terms of individual earnings, as opposed to household assets. This study evaluates progressiveness in terms of household assets because many wealthy women are not employed. Just because women have low earnings does not mean they are poor. Earnings are a better measure of progressivity for men because most men who have low earnings are poor. Methods and Data The study s goal is to measure differences in the relative impact three minimum benefit proposals have on women reaching age 62 in 2020 to The policy simulation is carried out on the 1992 Health and Retirement Study (HRS) with restricted access administrative Social Security earnings data. The HRS is a nationally representative longitudinal study of more than 9,700 Americans and their spouses born between 1931 and Thus, the total sample was 12,600 individuals of whom slightly more than 9,700 were age eligible. To ensure the sample is representative, only age-eligible individuals, or rather those individuals born between 1931 and 1941, are included. The HRS is the most comprehensive survey of aging Americans, asking questions on family, income, current and past employment, and health. The sample for this study includes age-eligible women in Wave 1 of the HRS, but excludes women without Social Security records because these are essential to create Social Security benefits. Weights were used to adjust for survey design as well as to account for these women. The final sample was 3,427 women. To analyze changes to family benefits, data must include earnings information on the spouses and former spouses of individuals to calculate spousal and survivor benefits. In the HRS, Social Security earnings records, which are imperative to benefit calculations, can be merged to both individuals within married couples. The earnings records of former spouses, however, are not available. Matching multiple imputations generated earnings histories for former spouses of divorced and widowed women (Rubin, 1987). A benefit calculator was 14 The Gerontologist

4 created to calculate reformed benefits. Benefit calculations were equivalent to those produced by the Social Security Administration s primary insurance amount (ANYPIA) calculator. Because this study is only interested in relative changes, there are some precise pieces that were not included, such as a benefit supplement for the survivors of railroad workers. Benefits were calculated based on retirement at age 62 without adjusting for benefit reductions normally associated with early retirement because the point is to consider women s options rather than their choices. Age 62 was chosen rather than 67 because there is significant volatility in peoples earnings after age 62 due to retirement. Again, the study goal is to evaluate relative outcomes so their benefits must be comparable. Nonparametric Simulation To simulate demographic changes related to women s work and family and to avoid the complexities associated with a full-blown microsimulation model, a nonparametric cohort splicing approach was adopted to account for demographic changes between cohorts. This method is not meant to be exact, but to be illustrative of how changing demographic trends influence the reforms effects on a younger cohort of women. The underlying strategy for this method is similar to more complicated microsimulation models. Observed histories of the present cohort of women beneficiaries represent the unobserved histories of younger cohorts of women (Iams & Sandell, 1997). Essentially, this is a cell reweighting exercise. The cohort of women in the HRS (using the Social Security adjusted weight) are categorized into descriptive cells including race, marital status, labor force participation, education, and parenting status based on their characteristics from ages 30 to 40. Women aged 30 to 40 in the 2000 Current Population Survey and Fertility Supplement (using the CPS person level weight) are categorized into the same descriptive cells. The women in the CPS provide the data to reweight the cells in the HRS. See Appendix Table 1 for the new weights and the specific categories. Appendix Table 2 shows that these results were comparable to more complicated microsimulation models such as Modeling Income in the Near Term (MINT), developed by the Social Security Administration to project retirement income for those born between 1931 and 1960, and demographic forecasts using the CPS (Goldstein & Kenney, 2001; Toder et al., 1999). See Herd (in press) for a more detailed comparison between these two approaches. The Model To estimate who benefits and loses under each reform, simple cross tabulations that control for only one or two characteristics were used, as well as regression analyses that can control for multiple characteristics. The dependent variable is the proposed benefit subtracted from the benefit under current rules. The categorical variable is based on three categories where Y = 1 for benefit decreases, Y = 2 for no change in benefit, and Y = 3 for benefit increases. The dependent variable is a dummy variable for the Privatization minimum, where Y = 1 for a benefit increase, because spousal benefits are not eliminated. The independent variables are a series of demographic, socioeconomic, employment, and family variables, which show how different groups fare under the proposed reforms. Demographic variables include race and age. Socioeconomic variables are imperative to sort out the progressivity of each reform. They include education (no high school, high school diploma, college diploma, or above) and household assets. Household assets include all housing and nonhousing equity. Assets were divided into quartiles to provide a relative measure of peoples resources. The study was most concerned with how these proposals impacted those in the bottom two quartiles to get a sense of the proposals progressiveness. Employment variables include the number of zero earnings years in prime earning years and whether zero earnings years occurred early or late in the life course. Family variables include marital status and whether or not individuals have children. For the regression analyses, an altered two-part model (Duan, Manning, Morris, & Newhouse, 1983) was used. The model s first part is an ordered probit that estimates the probability of receiving an increase, no change, or decrease in benefit. Coefficient estimates from the probit model are used to estimate predicted probabilities: Pr(Y = 3) = U(Sj k 2 ) Pr(Y = 2) = U(k 2 Sj) U(k 1 Sj) Pr(Y = 1) = 1 Pr(Y = 3) Pr(Y = 2) where U is standard normal cumulative function, Sj is the sum of the model coefficients multiplied by the chosen variable values (i.e., to predict the probability of benefit increase for a Black women, multiply the race coefficient by 1, where Black = 1 and White = 0, and so forth, then summing them), and k 1 and k 2 are the respective cut points. This equation is altered for the Privatization minimum because there are only two outcomes (no change or benefit increase) and one cut point. The model s second part is an ordinary least squares (OLS) regression that estimates the magnitude of benefit change. The dependent variable is the natural log of the difference between an individual s benefit under the current rules and the proposed benefit reform. Separate models are run for those receiving increases and decreases. Vol. 45, No. 1,

5 Table 1. Description of Policy Proposals Policy Option Who s Eligible Monthly Benefit Size Cost Offset Targeted/Universal Privatization Worker minimum Resident minimum Those who work years Those who work a minimum of 10 years Any citizen or resident of the United States The benefits is 60% None of poverty level for those working 20 years up to 100% of poverty level for those working 40 years. Poverty level is set at that for a single elderly person ($751). $545 Eliminate spousal benefits Poverty level for single elderly person ($751) Eliminate spousal benefits Targeted Semi-universal Universal Results The following results detail how the targeted Privatization minimum (requires 40 years of earnings for a poverty-level benefit), the partially universal Worker minimum (eliminates spousal benefits and requires 10 years of earnings for a poverty-level benefit) and the universal Resident minimum (eliminates spousal benefits and requires no earnings history for a benefit set at 71% of the poverty level) affect benefit adequacy and equity within the program as a whole, as well as for individual beneficiaries. Table 1 provides an overview of each policy proposal. The results show that the Privatization minimum is less successful than the Resident and Worker minimums at protecting the poorest women beneficiaries. Figure 1 shows how each proposal affects Social Security s overall progressivity for women. Each bar represents the total amount of money spent on noncontributory benefits, benefits that are unrelated to individuals earnings and contributions to the system. The first bar reflects money spent on noncontributory spousal benefits under the current structure. Just 14% of noncontributory spousal benefits (benefits not tied to earnings) are distributed to women in the bottom asset quartile, while 40% are distributed to women in the top asset quartile. The next bar represents the Privatization minimum and Figure 1. Distribution of noncontributory benefits under current rules and reforms for women beneficiaries reaching age 62 between 2020 and 2030, by household asset quartile. The overall sum of noncontributory benefits varies between these proposals. Compared to current spousal benefits, Privatization is 1% higher, Worker is 1% lower, and Resident is 0.005% higher. 16 The Gerontologist

6 Table 2. Percentage of Those Receiving Increases and Decreases in Benefit and Percentage Differences in Benefits Privatization Worker Resident Increase Increase Decrease Increase Decrease Variable Total % Increase Total % Increase Total % Decrease Total % Increase Total % Decrease Total (%) Race Black White Asset quartiles Bottom Second Third Top Marital status Married Divorced Widowed Never married Single women and mothers Never married mothers Never married nonmothers Divorced mothers Divorced nonmothers Bottom lifetime income quartile Never married mothers Never married nonmothers Divorced mothers Divorced nonmothers includes the money spent on noncontributory spousal benefits as well as the new minimum benefit. The Privatization minimum alters the system so that one fifth of noncontributory benefits are distributed to women in the bottom quartile. The last two bars represent the Worker and the Resident minimum. Because these proposals eliminated spousal benefits the bars only represent the dollars spent on the minimum benefits. The Worker and Resident minimum adjust the system so that more than 30% of noncontributory benefits are distributed to women in the bottom asset quartile. Figure 1 is important because it shows how each of the proposals changes the overall distribution of benefits among women. It shows the result of shifting resources devoted to noncontributory spousal benefits to the Worker and Resident minimum proposals. The Resident and Worker minimums, as compared to the current system and the Privatization minimum, are most effective at targeting resources towards women in the lowest asset households. Table 2 compares how the reforms would change women s benefits as compared to their benefits under the current rules. It shows the percentages of who receives increases and decreases and the percentage size of those increases and decreases. The Privatization minimum raises benefits by 15% for 4% of women. Almost 30% of women receive an average 25% increase under the Worker minimum. The Resident minimum causes 14% of women to receive an average 56% increase. In sum, the fewest women receive an increase from the targeted Privatization minimum, the largest number receive an increase from the partially universal Worker minimum, while the fully universal Resident minimum leads to the largest benefit increases. Concerning adequacy, how do women in the bottom versus top asset quartile fare under these proposals? The Privatization minimum is most effective at targeting benefit increases only to poor women, while benefiting few wealthy women. However, many more poor women gain benefit improvements under the Worker and Resident minimum benefits compared to the Privatization minimum. Benefit increases go to 11% versus 1.6% of those in the bottom and top asset quartiles respectively with the Privatization minimum. The average size of the increases was 19% and 3% respectively. With the Worker minimum, more than 40% of those in the bottom asset quartile receive an average 36% benefit increase Vol. 45, No. 1,

7 Table 3. Distribution of Those Ineligible for Worker Minimum Ineligible Total 11.3 Composition of ineligible Marital status Married 89 Divorced 8.2 Never married 2.7 Asset quartiles Bottom 25.6 Second 19.8 Third 23.1 Top 33.5 compared to an average 17% increase for 27% of those in the top asset quartile. The Resident minimum provides benefit increases to 32% of those in the bottom asset quartile and 8% of those in the top asset quartile. Another way to explore how these policies affect needier women is to focus on women who had been unmarried mothers. Single mothers are a growing percentage of the population and have a high risk of poverty, both when young and in old age. Once again, the Privatization minimum benefits significantly fewer women than the Worker and Resident minimums. Just 12% of never-married mothers in the lowest lifetime earnings quartile can expect a benefit increase under the Privatization minimum compared to 63% and 85% with the Worker and Resident minimums respectively. The respective size of those benefit increases were 11%, 74%, and 167% respectively. Single mothers generally fared better the more universal the policy. While 40% of nevermarried nonmothers in the bottom lifetime income quartile had benefit increases under the Privatization minimum, just 12% of never-married mothers did. Further, the size of the nonmothers increase was twice as large. Under the Worker and Resident minimums single mothers fared relatively equally or even better than nonmothers. Two of the proposals lead to benefit decreases. The Worker and Resident minimums eliminate noncontributory spousal benefits and shift those resources, in theory, toward poorer women. Do these two proposals keep benefit cuts confined to well-off women? Under the Worker and Resident minimums, 11% and 21% of women, respectively, receive benefit cuts. The Worker minimum generates the largest cuts because it requires at least 10 years of earnings to receive the $751 benefit, otherwise no benefit is received. Table 3 displays the composition of those ineligible. About 11% of women who will reach age 62 from 2020 through 2030 will be ineligible for the Worker minimum. Almost 90% of them are married, and 57% are in the top two asset quartiles. As mentioned above, only married and Table 4. Replacement Rates for Married Couples. Income Quartile Current Privatization Worker Resident Bottom Second Third Top divorced women receive spousal benefits and have cuts. Though not listed in the table, further calculations show, overall, 5% of women in the bottom two asset quartiles lose their Social Security benefits altogether. They can draw on SSI, but as previously discussed, SSI is an imperfect program. The cuts under the Resident minimum are not as severe as cuts under the Worker minimum because everyone receives a minimum $545 benefit. The average benefit cut is 16%. The highest spousal benefit is $830; thus one could not lose more than $285. Approximately 13% of those in the bottom asset quartile received an average 16% benefit cut, which is problematic because they likely have little income outside of Social Security. Nonetheless, 32% of women in the bottom asset quartile receive an average 72% increase in their benefit under the Resident minimum. One effective test of how benefit changes impact progressiveness is the replacement rate the percentage of an individual s average lifetime monthly income replaced by the benefit. Spousal benefits distort replacement rates in ways that provide preference to high-wage earners in single-income households relative to two-wage earners in a lowincome household because the spousal benefit increases the replacement rate by 50% in a single-earner household. Table 4 shows changes in replacement rates for married couples. The replacement rate for couples in the bottom income quartile rises from 64% under the current formula to 68% under the Privatization minimum. In contrast, it rises to 75% and 109% under the Worker and Resident minimums, respectively. All of the previous analyses are descriptive and do not control for more than one background factor at a time. Figures 2 and 3 display predicted probabilities, which show how assets, employment history, and marital status, controlling for other factors, impact the probability of receiving benefit increases and decreases. Appendix Table 3 displays the ordered probit coefficients and their significance levels. The predicted probabilities are useful because the coefficients have little intuitive meaning. Generally, however, Appendix Table 3 produces similar results to more descriptive analyses. Women who are never married, Black, and live in low-asset households are statistically significantly more likely to receive benefit increases than are married, White, and wealthier women under all three proposals, though the effect of 18 The Gerontologist

8 Figure 2. Probability of benefit change for a low-asset, single, working mother (high school degree, never married, born in 1931, Black, 5 9 zero earnings years during 20s, 30s, and 40s, low-asset household). Ordered prohibit equations available on request. Author s calculations based on the Health and Retirement Study (HRS). marital status was not significant under the Privatization minimum. Moreover, women s work histories had statistically significant impacts on their probability of benefit increase under all three proposals. Thus, in thinking about adequacy and equity issues, a useful comparison is between a single mother in a low-asset household with high employment and a married mother with limited employment in a high-asset household. Older women who had been single mothers are of particular concern given their high rates of poverty and because their numbers are growing. It is important to note that the precise estimates displayed here are less important than the general differences between the proposals that they convey. Figure 2 shows that the low-asset, never-married working mother has a 76% probability of receiving a benefit increase under the Worker minimum and a 57% probability under the Resident minimum, compared to a 12% probability under the Privatization minimum. Thus, she fared much better under the Resident and Worker minimums compared to the Privatization minimum. She has no probability of receiving a benefit decrease under any of these proposals. Figure 3 examines how these proposals affect a married mother in a high-asset household who is rarely employed. She has less than a 1% probability of benefit increase under the Privatization minimum and a 6% and 12% probability of increase with the Worker and Resident minimums, respectively. She has a significant probability of benefit decrease under the Worker (61%) and Resident (39%) minimums. This demonstrates how the Worker and Resident reforms shift noncontributory spousal benefits from Figure 3. Probability of benefit change for a high-asset, married, nonworking mother (high school degree, married, born in 1931, White, zero earnings years during 20s, 30s, and 40s, high-asset household). Ordered prohibit equations available on request. Author s calculations based on the Health and Retirement Study (HRS). Vol. 45, No. 1,

9 Table 5. Average Benefit Increases and Decreases: OLS Estimates From the Second Part of the Two-Part Model Privatization Worker Resident Variable Increase Increase Decrease Increase Decrease Original benefit Assets (0 ¼ Top Quartile) Bottom quartile *** ** (0.392) (0.162) (0.120) (0.172) (0.165) Second quartile 0.977* 0.360*** 0.199* ** (0.392) (0.089) (0.101) (0.162) (0.090) Third quartile 0.813* (0.403) (0.081) (0.101) (0.497) (0.079) Marital status Never married ** (0.528) (0.197) (0.228) Widowed (0.826) (0.145) (0.261) Divorced * (0.360) (0.088) (0.157) (0.195) (0.113) No children 1.11*.377* (0.469) (0.158) (0.226) (0.206) (0.232) Number of zero earnings years in 20s, 30s, 40s 6 10 (0 ¼ 0 5) (0.336) (0.259) (0.393) (0.511) (0.204) * (0.368) (0.274) (0.763) (0.520) (0.199) * 1.523*** * (0.537) (0.240) (0.413) (0.524) (0.202) ** 2.161*** 1.079*.610** (0.563) (0.256) (0.406) (0.476) (0.205) * 2.374*** 1.206* 0.516* (0.553) (0.308) (0.401) (0.531) (0.209) R Notes: A dash indicates no variation. OLS = ordinary least square. *p,.05; **p,.01; ***p,.001; p,.10. women like the high-asset, married mother who is rarely employed towards the poor, working, nevermarried mother. Table 5 shows the OLS analyses of how different factors, with all else equal, affected the sizes of the benefit increases and decreases under the proposals. Again, the precise estimates are less important than the general sense they convey about the differences between the proposals. Compared to women in highasset households, women in the lowest asset quartile had 57% and 28% significantly larger benefit increases under the Worker and Resident minimums, respectively. Under the Privatization minimum, women in the second and third asset quartiles had 98% and 81% significantly smaller increases, respectively, than women in the top asset quartile. Under the Worker and Resident minimums, women in lower asset households also had smaller benefit decreases when compared to women in highest asset households, though not all were significant. The negative coefficients indicate a smaller decrease in benefits. For example, under the Resident minimum women in the bottom asset quartile had a 44% smaller benefit decrease than women in the top asset quartile. This result reinforces the progressiveness of the more universal minimums compared to the targeted Privatization minimum. Never-married women had significantly larger increases than married women under the Worker and Resident minimums, though the effect was smaller for the Worker minimum. Nonmothers had a 111% larger benefit increase than did mothers with the Privatization minimum, emphasizing the difficulty mothers face accruing a 40-year earnings history. Nonmothers had a 38% larger benefit increase under the Worker minimum compared to mothers, which only required 10 earnings years for eligibility. The work history variables well reflected the minimums employment incentives. For the Worker minimum, those with high numbers of zero earnings years (21 25), had a 69% larger increase, but also a 216% larger decrease than those with 0 5 zero earnings years. With the Resident minimum, they had a 108% and 61% larger increase and decrease, respectively. Effects were insignificant for the Privatization minimum. 20 The Gerontologist

10 Table 6. Cost Estimates Change in Average Benefit Among Men and Women Policy Option Women Men Privatization (%) Worker (%) 1 2 Resident (%) Recent proposals to reform Social Security are driven by the program s perceived budgetary problems. It is therefore important to account for budgetary implications. This study compares changes in average benefit, in the population as a whole, between the current rules and each of the three reforms. This is done for women and men because both would be covered under the new rules. As Table 6 shows, the Privatization minimum benefit increases the average benefit for women by.004% and.0001% for men. The other proposals cost more, but these costs are significantly curtailed by the elimination of spousal benefits. The Worker minimum would reduce women s average benefit by 1%. Men s benefits would increase by 2%. The Resident Minimum benefit leads to a.005% average increase in women s benefits and a 2% increase in men s benefits. Average changes in benefits are meant to be illustrative of these proposals relative costs. Behavioral changes that might result from these reforms are not taken into account, which is common for policy simulations. Nonetheless, minimum benefits that require no or limited employment could reduce employment, particularly just before retirement. But because current spousal benefits require no employment, their elimination combined with the introduction of minimum benefits would likely lead to no overall change in employment incentives for married and formerly married women. However, these minimums could reduce incentives for labor force participation for men and those who never marry as compared to the current system. Study Limitations The results from this study are based on a policy simulation. The study projected how women who reach age 62 between 2020 and 2030 would fare under three different minimum benefit proposals. The limitations of this study are comparable to other simulations projecting retirement income (i.e. Toder et al., 1999). The first aspect to the simulation entailed filling in missing earnings histories necessary to compute benefit calculations. For most individuals (about 75% of the sample) Social Security administrative earnings data were available. Most studies have found that the remaining sample is basically representative with the exception that people with very limited earnings histories are slightly underrepresented (Haider & Solon, 2000). Thus, there is the possibility that the percentage of women who have benefit cuts is underestimated (as under the Resident and Worker minimum women must rely on their own earnings histories for their benefit). These minimal disadvantages, however, are outstripped by having Social Security administrative earnings data to calculate benefits. Finally, it was necessary to impute missing earnings information for former spouses of divorced women and some widowed women. Thus, the results for divorced women in particular should be interpreted more warily (see Herd, in press, for more details). Second, like all simulations, the past (that is the histories of women in the current HRS cohort) is used to project the future. This method is not meant to be exact, but illustrative of how changing demographic trends influence the reforms effects on a younger cohort of women. The major limitation is that 100% certainty about the benefits of younger women cannot be achieved until these women actually reach the eligibility age for Social Security. The simulation provides an educated estimate. The likeliest error to result from this approach is that the percentage of women who had no benefit change is underestimated because the wages of women have improved between the HRS cohort and younger cohorts of women. Younger cohorts of women have higher earnings and may be less likely to rely on any form of noncontributory benefits. Therefore, these estimates are strongest when thinking about the differences between women, i.e., how married women fare relative to unmarried women. Finally, a more general and conceptual issue with this study is that only minimum benefits are analyzed as a way to reform Social Security s family benefits. Other reform suggestions have included altering spousal and widow benefits to better protect divorced women or rewarding parents with care credits. (This paper is a component of a larger project that examines these other two approaches as well [Herd, in press].) In general, more universal minimum benefits (i.e. the Worker and Resident minimums) are more progressive, in terms of benefiting the poorest women, than are care credits or improvements to spousal and widow benefits. Discussion Concerns with individual accounts and problems with family benefits, as well as high poverty rates among subgroups of women, have prompted politicians and academics to suggest implementing a minimum benefit into Social Security. This study sheds light on how the three most common proposals would affect the progressiveness of older women s benefits. The tightly targeted Privatization minimum, which requires 40 years of earnings for a povertylevel benefit, is significantly less successful at Vol. 45, No. 1,

11 improving Social Security s adequacy goals than the more universal Worker and Resident minimums, which require 10 earnings years and U.S. residency, respectively, for eligibility. The Worker and Resident minimums shifted the money spent on noncontributory spousal benefits, which most heavily benefit upper-income women, toward more generous minimums that benefit poor women. About 14% of the money spent on spousal benefits ends up with women in the lowest asset quartiles. The Privatization minimum increases the combination of noncontributory spousal and minimum benefit dollars directed at the poorest beneficiaries to 20%. The Worker and the Resident minimums, by diverting all of the money spent on spousal benefits toward minimum benefits, leave one third of the money spent on noncontributory benefits with women in the lowest asset quartile. Poor women were also less likely to benefit, and received smaller increases, from the Privatization minimum, than the Resident and Worker minimums. Ultimately, most women, even poor women who are more likely to work, cannot muster 40 years of earnings largely due to their disproportionate responsibility for raising children. Universal policies are beneficial to women for this very reason. Using funding spent on spousal benefits to implement a minimum benefit has not been seriously considered, but in a budget-driven policy debate this could be a viable alternative. Spousal benefits are most beneficial for married women with high-earning spouses. But, the elimination of spousal benefits does cause some women to receive lower benefits. Most women with limited means face no benefit reduction, but some women with limited assets, particularly divorced women, receive smaller benefits than they would have under the current benefit rules. Those receiving benefit cuts are safest under the Resident minimum because their benefit cannot go below $545. Contrastingly, under the Worker minimum, those without 10 years of earnings could receive no benefit whatsoever. Ideally, all beneficiaries should have an adequate benefit. But resources are limited. Noncontributory benefits should be distributed more equally among women. Should a divorced mother who never worked have an $830 benefit while a nevermarried mother who did low-wage intermittent work have a $400 benefit? By eliminating spousal benefits and instituting a more universal minimum benefit we level the playing field for these two women. Some may argue that spousal benefits subsidize women who stay at home to raise children. Women do work for these benefits. But spousal benefits are not a reward for raising children, one need not have children to receive them, and many single parents will never access them. It is an indirect approach to rewarding parents that does not reward all parents. Moreover, women are rewarded differently depending on how much their husbands earn. One could argue minimum benefits are a more equitable way to distribute benefits that are unassociated with earnings and indirectly benefit parents who took significant chunks of time out of employment than are spousal benefits. Another concern is how the Worker and Resident minimums would affect work incentives. The Worker minimum would improve work incentives for married women, relative to spousal benefits, because they require some earnings history for receipt. The Resident minimum, like spousal benefits, would require no earnings history. But eliminating spousal benefits means someone without an employment history would not receive a higher benefit than someone with an employment history. The downside is that under both reforms an individual who does less paid work than another may receive an equivalent benefit. Moreover, more people would be eligible for the Worker and Resident minimum benefits than were eligible for spousal benefits, so they would reduce overall employment incentives. Some may argue that altering SSI is the best way to improve benefits for the poorest elderly Americans. But less than half of those eligible for SSI receive benefits. It does not reach all of the neediest elderly Americans. Systems that combine a minimum benefit with an earnings-related benefit are most effective at reducing poverty (Korpi & Palme, 1998). Means tested benefits, like SSI, generally cannot generate the broad support a more universal social insurance program like Social Security generates. Having a minimum benefit within Social Security shifts resources toward the needy within a popular, mostly universal program. In terms of implementing targeted or more universal minimum benefits, targeting may reduce waste, but it fails to protect many needy beneficiaries. Universal approaches may benefit some who do not need them, but they catch almost everyone who does. Ultimately, more universal minimum benefit proposals best meet the goal of benefit adequacy. And while some will query whether such proposals are politically viable, before politicians, policy makers, and the American citizens decide what is politically viable, the research that explains all of the options must be there. References Advisory Council on Social Security. (1979). Report of the 1979 Advisory Council on Social Security. Washington, DC: Government Printing Office. Bachu, A., & O Connell, M. (2000). Trends in labor force and employment characteristics of married-couple families with or without children. Washington, DC: U.S. Bureau of the Census. Bloom, D., & Bennett, N. (1990). Modeling American marriage patterns. Journal of the American Statistical Association, 85, Butrica, B., & Iams, H. (1999). Projecting retirement income of future retirees with panel data: Results from the Modeling Income in the Near Term (MINT) Project. Social Security Bulletin, 62, 3 8. Congressional Research Service. (2001). Summary of major changes in Social Security. Washington, DC: Government Printing Office. Duan, N., Manning, W., Morris, C., & Newhouse, J. (1983). A comparison of alternative methods for the demand for medical care. Journal of Business and Economic Statistics, 1, Edin, K., & Lein, L. (1997). Making ends meet. New York: Russell Sage. 22 The Gerontologist

12 Favreault, M., Sammartino, F., & Steuerle, C. (Eds.). (2002). Social Security and the family. Washington, DC: Urban Institute. General Accounting Office. (1996). Social Security: Issues involving benefit equity for working women. GAO/HEHS General Accounting Office. (1979). Minimum Social Security benefit: A windfall that should be eliminated. HRD Goldstein, J., & Kenney, C. (2001). Marriage delayed or marriage forgone? American Sociological Review, 66, Gustman, A., & Steinmeier, T. (2000). How effective is redistribution under the Social Security benefit formula? Boston, MA: National Bureau of Economic Research. Haider, S., & Solon, G. (2000). Nonrandom selection in the HRS Social Security earnings sample. Rand Working Paper. Retrieved October 15, 2002, from Harrington Meyer, M. (1996). Making claims as workers or wives. American Sociological Review, 61, Herd, P. (2003). Crediting care, citizenship, or marriage?: Gender, race, class, and Social Security reform. Presented at the 2003 American Sociological Association Meetings, Atlanta, GA. Herd, P. (in press). Reforming a breadwinner welfare state: Gender, race, class, and social security reform. Social Forces. Iams, H., & Sandell, S. (1997). Projecting Social Security earnings: Past is prologue. Social Security Bulletin, 60, Kolbe-Stenholm Plan, H.R. 1793, 107 th Cong. (2001). Korpi, W., & Palme, J. (1998). The paradox of redistribution. American Sociological Review 63, Korpi, W. (2000). Faces of inequality: Gender, class and inequalities in different types of welfare states. Social Politics 7(2), McGarry, K. (2000). Guaranteed Income. NBER, WP#7574. Cambridge, MA. Porter, K., Larin, K., & Primus, W. (1999). Social Security and poverty among the elderly. Washington, DC: CBPP. President s Commission to Strengthen Social Security. (2001). Strengthening Social Security and creating personal wealth for all Americans. Government Printing Office. Proctor, B., & Dalaker, J. (2002). Poverty in the United States. Current Population Reports, Rubin, D. (1987). Multiple imputation for nonresponse in surveys. New York: John Wiley. Smeeding, T., Estes, C., & Glasse, L. (1999). Social Security reform and older women, CPR Income Security Policy Series, no. 22. Syracuse, NY: Syracuse University. Toder, E., Uccellow, C., O Hare, J., Favreault, M., Ratcliffe, C., Smith, K., et al. (1999). Final report: MINT projections of retirement income through 2020 for the birth cohorts. Washington, DC: Urban Institute. Williamson, J., & Rix, R. (1999). Social Security reform: Implications for women. Boston, MA: Center for Retirement Research at Boston College. Received September 15, 2003 Accepted April 20, 2004 Decision Editor: Linda S. Noelker, PhD Appendix Table 1. Population Reweights for Projecting Women No Degree High School Degree College Degree Variable White Black White Black White Black Not working, with children Married Divorced Widowed Never married Not working, no children Married Divorced Widowed Never married Working, with children Married Divorced Widowed Never married Working, no children Married Divorced Widowed Never married Vol. 45, No. 1,